LOGIC2015Belgrade,21‐23May,2015

Proceedingsofthe

2ndLogisticsInternationalConference

Editors:

MiloradVidovićMiloradKilibardaSlobodanZečevićMomčiloMiljuš

GordanaRadivojević

Technicaleditor:

DraženPopović

Belgrade,2015

Editors: Prof.Dr.MiloradVidović Prof.Dr.MiloradKilibarda Prof.Dr.SlobodanZečević Prof.Dr.MomčiloMiljuš Prof.Dr.GordanaRadivojević UniversityofBelgrade,FacultyofTransportandTraffic Engineering,VojvodeStepe305,Belgrade,SerbiaTechnicaleditor: DraženPopović UniversityofBelgrade,FacultyofTransportandTraffic Engineering,VojvodeStepe305,Belgrade,SerbiaPublisher: UniversityofBelgrade,FacultyofTransportandTraffic Engineering,VojvodeStepe305,Belgrade,Serbia http://www.sf.bg.ac.rsInbehalfofpublisher: Prof.Dr.BranimirStanić DeanofFacultyofTransportandTrafficEngineeringPublishingeditor: Prof.Dr.DragoslavKuzmanovićYear: 2015Conferenceorganizer: UniversityofBelgrade,FacultyofTransportandTraffic Engineering,DepartmentofLogistics,Belgrade,SerbiaSupportedby: MinistryofEducation,ScienceandTechnological DevelopmentoftheRepublicofSerbiaISBN: DRAFTAllrightsreserved.©Copyright2015by theUniversity ofBelgrade, Faculty ofTransport andTrafficEngineering,VojvodeStepe305,Belgrade,Serbia.Theindividualessaysremaintheintellectualpropertiesofthecontributors.

PartI OPTIMIZATIONAND

MODELLINGINLOGISTICSANDTRANSPORTATION

ANEXPERTFUZZYMODELFORTHEDETERMINATIONOFTHEAMOUNTOFPURCHASE

SinišaSremaca*,BojanMatića,MilošKopića,GoranTepića

aUniversityofNoviSad,FacultyofTechnicalSciences,Serbia

Abstract:Inthispaperanexpertmodelforthedeterminationoftheamountofpurchasehasbeendevelopedonthebasisofthetheoryoffuzzylogic.Fuzzylogichaslargelybecomeasubstituteforconventionaltechniquesinnumerousapplications,includingtheareaofmanagingcertainlogisticsprocesses. In the paper, it has been used formodeling a complex logistics system inwhich it isdifficult todetermine the interdependenceof thepresentedvariablesapplyingclassicalmethods.Experienceofanexpertandinformationontheoperationsofthecompanyforacertaingroupofitemshavebeenusedtoformthemodel.Analysisofthevalidityofthemodelresultswasperformedon the basis of the average relative error and it has showed that the expert fuzzymodel fordeterminingtheamountofpurchaseimitatestheworkoftheexpertintheobservedcompanywithgreataccuracy.

Keywords:fuzzylogic,determinationoftheamountofpurchase,logisticsprocess.

1.INTRODUCTION

The determination of the amount of purchase is a logistics process that has a significantinfluence on the successful operation of a company [3]. From the logical aspect, thedetermination of the amount of purchase requires an adequate attention, since inadequatepurchase can additionally burden the company’s business [1]. On the other hand, in order toachieve ahigh level of service for the client, all purchase shouldbe realized independentlyoftheirvalue.Therefore,thedeterminationoftheoptimalamountofpurchaseisimportantfortherationalrealizationoftheprocessoftransport,manipulationandstorageinthedeliverychaintothefinalcustomer[2].

Inthepaper,anexpertmodelforthedeterminationoftheamountofpurchase(DAPmodel)wasdevelopedfollowingthefuzzylogic.DAPmodelisusedtoestimatethequantityforpurchaseinabusinesspractice.ThesoftwarepackagesMatlabandFuzzyLogicToolboxwereusedtoformthemodel.

2.DESCRIPTIONOFTHEEXPERTFUZZYDAPMODEL

Themainproblemin formingtheexpert fuzzyDAPmodelwastodeterminethebaseof fuzzyrulesandmembershipfunctionparameters[7].Todefinerules,thedataobtainedbyalogistics

* sremacs@uns.ac.rs

experts’ survey were used. Expert knowledge on the process of determining the amount ofpurchasewasexpressedusingacertainnumberoflinguisticrules.

The selection of the type and parameters of themembership functionwas performed on thebasisofthepositiveexperienceof individualauthors[4]andsubjectiveevaluationbyauthors.Theoverviewofliteraturehelpedtodeterminethatthehighestprecisionofoutputvaluescouldbe obtained by applying the Gaussian membership functions [5]. For that reason, that curveshapewasgenerated inthemodel,while itsparametersweredeterminedusingthesubjectiveauthors’ evaluation. Amount intervals of the input and output variables were defined on thebasisofrealvaluesinthepractice.ThemodelwasestablishedontheMamdanifuzzyinferencesystem and the min‐max inference method, while the centroid method was applied for thedefuzzificationprocess[6].

The fuzzy expert DAP model has three input variables: value demand, stock level and price(Figure 1). The output variable is the amount of purchase. Stock level and price have threevalues,whilethevaluedemandandtheamountofpurchasehavefivevalues.Theinputvariablevaluedemandhas the following values: very small (XS), small (S),medium (M), large (L) andverylarge(XL),whilethestocklevelhasthefollowingvalues:small(S),medium(M)andlarge(L).The inputvariablepricecanbedescribedwith the followingvalues: low(L),medium(M)andhigh(H).Theoutputvariableamountofpurchasehasthefollowingvalues:verysmall(XS),small (S), medium (M), large (L) and very large (XL). Most of the linguistic values were notnecessaryduetothefactthatthesatisfactoryoutputgradationandprecisionwereachievedinchangingtheinputvalues.

Figure1.Expertfuzzyinferencesystem

3.THEDEVELOPMENTOFMODEL

Realvaluesweremappedusingfuzzificationintothemembershipfunctions.Theinputvariablevaluedemandhasthevaluesintheinterval[0,250],stocklevelisintheinterval[0,200],whilethepriceispresentedinlinguisticvaluesintheinterval[1,3].Theintervaloftheoutputvariableamount of purchase is [0, 250]. These interval values were obtained on the basis of thecompany’sbusinessfortheobservedarticle.

ThevaluesoftheinputandoutputvariableshavetheGaussianmembershipfunctions(1)thatcanbedefinedasfollows:

,forxϵ[0,c] (1)

GaussianfuzzynumberisdescribedwithtwoparametersA=(σ,c).Thefirstnumberpresentsthe left and the right distribution of the Gaussian curve along the abscissa,while the secondnumber presents the value on the abscissawhere the Gaussian curve has the value 1 on theordinate.Membership functions of the input variable valuedemand (2) are definedusing theparameters XS [30; 0], S [30; 60],M [30; 125], L [30; 190] and XL [30; 250] for x ϵ [0, 250](Figure2a)):

(2)

Membership functions of the input variable stock level (3) are defined with the parametersS[50;0],M[50;100]andL[50;200]forxϵ[0,200](Figure2b)):

(3)

Membershipfunctionsoftheinputvariableprice(4)aredefinedusingtheparametersL[0,5;1],M[0,5;2]andH[0,5;3]forxϵ[0,3](Figure2c)):

, , , (4)

Membershipfunctionsoftheoutputvariableamountofpurchase(5)aredefinedwithXS[30;0],S[30;60],M[30;125],L[30;190]andXL[30;250]forxϵ[0,250](Figure2d)):

(5)

a) b)

c) d)

Figure2.Fuzzysetmembershipfunctions:a)valuedemand,b)stocklevel,c)priceandd)amountofpurchase

InthefuzzyexpertDAPmodel,therearethreeinputlinguisticvariables,whereonevariablehasfive values,while the remaining two variables have three values each. The combination of alllinguisticvaluesprovidesthebaseoffuzzyruleswith45rules(Table1).

Table1.Fuzzyrulematrixif:a)priceisL,b)priceisMandc)priceisH

a) b) VALUEDEMAND

XS S M L HL

STOCK S M L XL XL XL

M S M L XL XL

L XS S M L XL

VALUEDEMAND

XS S M L HL

STOCK S S M L XL XL

M XS S M L XL

L XS XS S M L

c) VALUEDEMAND

XS S M L HL

STOCK S XS S M L XL

M XS XS S M L

L XS XS XS S M

The centroid method was applied for the defuzzification process. The result of the modeldefuzzificationistheselectionofonevalueoftheinputvariableamountofpurchase(Figure3).

Figure3.Modeldefuzzification

4.RESULTSOFMODEL

Figure 4 presents the graphic overview of the output variable amount of purchase independenceontheinputvariables.

a) b)

c)

Figure4.Amountofpurchasedependingon:a)valuedemandandstocklevel,b)valuedemandandpriceandc)stocklevelandprice

TheDAPmodelistestedon50examplesfordeterminingtheamountofpurchasefortheknowninputvalues.Figure5presentstheamountofpurchaseobtainedbytheexpertfuzzyDAPmodelandtherealamountofpurchaseinacompanyforthegiveninputdata.

Figure5.AmountofpurchasebasedontheexpertdecisionandtheexpertfuzzyDAPmodel

VALUEDEMANDVALUEDEMANDAMOUNTOFPURCHASE

AMOUNTOFPURCHASE

AMOUNTOFPURCHASE

PRICE

PRICESTOCKLEVEL

STOCKLEVEL

Expert ExpertfuzzyDAPmodel

The evaluation analysis of the DAPmodel resultswas conducted on the basis of the averagerelativeerroroftheobtainedresults inrelationtotherealresults.Thetestingof50examplesprovidedtheaveragerelativeerrorof5.6%.Basedontheanalysis,itcanbeconcludedthatthereisahighcompatibilitybetweenrealanddemandedresults,andthattheexpertfuzzyDAPmodelprovidesvalidresultsandcancompletelyimitatetheworkofanexpert.

5.CONCLUSION

The applied concept of artificial intelligence is utilized for presenting, manipulating andimplementing human knowledge on the efficient management of certain logistics processes.Fuzzylogichasproventobeavaluableartificialintelligenceconceptindeterminingtheamountof purchase that is designed using intuition and assessment of a logistics expert. Fuzzy logicenabledtheexplanationofthesystemdynamicsviaalinguisticpresentationofknowledgeonalogisticsprocess.Itwasusedformodelingacomplexlinguisticsysteminwhichitisdifficulttodeterminetheinterdependenceofthepresentedvariablesapplyingotherclassicalmethods.Inthepaper, aDAPmodel for solving a concreteproblem in abusinesspracticewasdeveloped,followingthetendencyincontemporaryscientificresearch.Themodelwastestedandverified,andhenceitcanbepracticallyapplied.Theproposedmodel,withsomeminormodification,canbeappliedinanycompanydealingwiththegoodsflowrealization.

ACKNOWLEDGMENT

Thiswork is a part of the research project TR 36030 funded by theMinistry of Science andTechnologyofSerbia.

REFERENCES

[1] Capkun,V.,Hameri,A.P.,Weiss,L.A, (2009).On theRelationshipBetween InventoryandFinancialPerformanceinManufacturingCompanies.InternationalJournalofOperations&ProductionManagement29(8),789‐806.

[2] Ferišak, V., (2013) Nabava – politika, strategija, organizacija, management. Vlastitoizdanje,Zagreb.

[3] Garcia,N.,Puente,J.,Feandez,I.,Priore,P.,(2012).Evaluationoftheappropriatestrategicproductsuppliersusingafuzzyapproach.InternationalConferenceonIndustrialLogistics,Zadar,145‐153.

[4] Giannoccaro,I.,Pontrandolfo,P.,Scozzi,B.,(2003).Afuzzyechelonapproachforinventorymanagementinsupplychains.EuropeanJournalofOperationalResearch149(1),185‐196.

[5] Griffis, S.E., Bell, J.E., Closs D.J., (2012). Metaheuristics in Logistics and Supply ChainManagement.JournalofBusinessLogistics,33(2),90‐106.

[6] Negnevitsky,M.,(2005).ArtificialIntelligence:AGuidetoIntelligentSystems.2ndEdition,EdinburghGate,England.

[7] Sremac, S., (2013).Goods flowmanagementmodel for transport‐storageprocesses,PhDdissertation.

COSTANALYSISOFOPENVRP

ZuzanaČičková,IvanBrezina,JurajPekára

aUniversityofEconomicsinBratislava,Slovakia

Abstract:Classicalroutingproblemsaimatdesigningtheminimumcostofroutesoriginatingandfinishingfromacentraldepotforsatisfyingcustomerdemand.Theopenversionsdonotconsiderthevehiclesneedtobereturned.Theimportanceofopenversionsoftheknownroutingproblemsarisese.g.fromcar‐hireoptions.Althoughhiringvehiclescouldbemoreexpensiveperunitdistancetraveled,theirusecouldleadtoconsiderablesavingduetoreductionofthetotalcostoftheroute.Open models of routing problems allow describing different real‐world applications, e.g.consideringoneormorevehiclesofhomogenousorheterogeneousfleetstructure(ownorleased)startinginoneofdifferentdepotsthatresulttodifferentlogisticcosts.Thepaperisfocusedoncostanalysis thatmay aid in the decision to use own and hired vehicles based on regional data inSlovakia.

Keywords:OpenVehicleRoutingProblem,MixedIntegerProgramming,MathematicalModel.

1.INTRODUCTION

Logistics costs constitute amajor share of the total costs of almost every organization.Manyvariants of routing problems can be very rewarding in the field of logistics. Thewell‐knowncapacitatedvehicleroutingproblem(CVRP)isoneofthemostdiscussedproblemsinoperationsresearchČičkováandBrezina(2008),Čičkováetal.(2013),Čičkováetal.(2014),Desrochersetal. (1992), Fábry et al. (2011). The significance of that problem is evidently depended of itscomputationalcomplexity,butalso the importance follows fromagreatpracticalapplicability,thustheproblemcanbeappliedinmoregeneralway.

This paper describes, based on regional data in Slovakia, the version of open vehicle routingproblem (OVRP) that allows combining CVRP together with the possibility to realize openroutes, e.g. Čičková et al. (2014). In general, OVRP is a popular problem in the field ofdistributionmanagementandcanbeusedtomodelmanyreal‐lifeproblems.Forexampleifthecompanydealswithoutitsownvehiclefleetandhastohiresomevehicletodeliveritsproductstocustomers,itisnotconcernedwhetherthevehiclereturnstothedepot,anddoesnotpaythetravelingcostsbetweenthelastcustomerandthedepot.ThestandardOVRPcanbedescribedasfollows:Consideradepotfromwhichsomeproductshavetobedeliveredtoasetofcustomers.Products are loaded on a vehicle (vehicles) at the depot and afterwards load needs to betransported to the customers. The capacity of vehicle (or fleet of vehicles) is (are) known (ifmore than one vehicle are used, the same capacity of all of them is supposed), so that allcustomers’demandneedstobeservedwiththeuseofsomevehicle(allthedemandsaremetinfull).Itisassumingtheknownshortestdistancebetweendepotandeachcustomer’slocation,aswellasbetweeneachpairsofcustomer’slocation.Thegoalistofindtheoptimalshortestrouteforavehicle(vehicles)sothateachcustomerdemandismet.

Practicalapplicationsoftenconsiderthedealtospecifythenumberofownvehiclesrelativetocostofrentedvehicle.Theunitcostperkmofownvehiclesconsistnotonlyoffuelcostbutalsoincludee.g.amortization,taxesanddriver’swage.Ontheotherside,therentalcostdependsoncostperkmandalsoincludesthefixedcostpercar.Presentedanalysisisbasedonexampleofdistribution in regions in Slovakia. The practical significance lies in modeling centrallycontrolled logisticprojectsthatconsistof twostages(e.g.organizingofvotedistribution).Thefirst stage consists in transferringelements (e.g. ballots) to regional capitals and their furtherdistributiontorelevantdistricttowns.

2.MATHEMATICALMODEL

Themathematical formulation based on abovementioned can be stated as follows: ConsidergraphG=(N0,A),whereN0=0,1,…nisthesetofallnodesinthegraph,sothatN0=N0,wheretheset 1, 2... N n representsthesetofservednodes(customers)andthenodeindexed

0representsorigin(depot).ThesetA=(i,j):i,jN0,ijisthearcsetofG.Ashortesdistancedij is associatedwitheveryarcof thegraph.Let theparametersgi, i N representdemandofcustomerandparametergrepresentsthecapacityofvehicles.Furtheron,thecost(perkm)ofownvehicles(co)andalsothecost(perkm)ofrentedvehicles(cr)areknown.Thefixedcostareasociated only with number of rented vehicle and they are designated as cf. Mathematicalprogrammingformulationrequirestwotypeofbinaryvariables:thevariablesxij, 0,i j N withafollowingnotation:xij=1ifcustomeriprecedescustomerjinarouteoftheownvehicleandxij=0otherwiseandthevariablesyij, 0,i j N withafollowingnotation:yij=1ifcustomeriprecedescustomer j inarouteof therentedvehicleandyij=0otherwise.Furtheron,wewillapplythevariablesui, i N thatbasedonwell‐knownMiller–Tucker‐Zemlin’sformulation,e.g.Milleretal.(1960),ofthetravelingsalesmanproblem.Thatvariableswillrepresentcumulativedemandofcustomersononeparticularroute.

Themathematicalmodelcanbestatedasfollows:

0 0 0 0 0

0min , ij ij ij ij ji N j N i N j N j N

i j i j

f co d x cr d y cf y

X u (1)

0 0

1ij iji N i N

x y

, j N , i j (2)

0 0

1ij ijj N j N

x y

, i N , i j (3)

(1 )i j ij ju q g x u , 0i N , j N , i j (4)

(1 )i j ij ju q g y u , 0i N , j N , i j (5)

0 0

0ij jii N i N

x x

, j N , i j (6)

i iq u g , i N (7)

0 0u (8)

0, 1ijx , 0, 1ijy , 0,i j N , i j (9)

Theobjective function(1)modelsthetotalcost forall theusedvehicles.Equations(2)ensurethatonlyoneofthevehicles(ownorrented)enterseachcustomerexactlyonceandequations(3)ensurethatthevehicledoesnotneedtodepartfromeverycustomer,becausetherouteofrentedvehicleendsafterservingthe lastof them.Equations(5)and(6)avoidthepresenceofsub‐tour (for own and rented vehicle) and also calculates the real cumulative demands ofcustomersofthenextnodeontheroutebasedonpreviousnode.Equations(7)ensurethatalldemandsontheroutemustnotexceedthecapacityofthevehicle.Thefixvaluesofvariablesu0aresetupbyequation(8).

3.OPENVEHICLEROUTINGPROBLEMINSLOVAKREGIONS

Theproblemdeals about thedistribution scheduling in the regions in Slovakia,whereoriginswere situated in the regional capitals. The distribution have to be assured using the own orrentedvehicleswiththesamecapacity.Thegoalwastodeterminehowmanyownandrentedvehiclesmustbeusedineachregionsothatdemandofalldistricttownsmustbemetwiththeminimalcost.Thetotalcostconsistofcostassociatedwiththeownvehicles(Fig.1a)andcostofrented(Fig.1b).

Fig.1aFig.1b

Slovakiaisdividedinto8regions:RegionBratislava(BA),RegionBanskáBystrica(BB),RegionKošice (KE),RegionNitra (NR),RegionPrešov(PO),RegionTrenčín(TN),RegionTrnava(TT)andRegionŽilina(ZA):

Region Bratislava (BA) is divided into 4 districts: Bratislava (0), Malacky(1), Pezinok(2),Senec(3).

Region Banská Bystrica (BB) is divided into 13 districts: Banská Bystrica(0), BanskáŠtiavnica(1), Brezno(2), Lučenec(3), Detva(4), Krupina(5), Poltár(6), Revúca(7), RimavskáSobota(8),VeľkýKrtíš(9),Zvolen(10),Žarnovica(11),ŽiarnadHronom(12).

Region Košice (KE) is divided into 7 districts: Košice (0), Gelnica(1), Michalovce(2),Rožňava(3),Sobrance(4),SpišskáNováVes(5),Trebišov(6).

RegionNitra(NR)isdividedinto7districts:Nitra(0),Komárno(1),Levice(2),NovéZámky(3),Šaľa(4),Topoľčany(5),ZlatéMoravce(6).

Region Prešov (PO) is divided into 13 districts: Prešov(0), Bardejov(1), Humenné(2),Kežmarok(3), Levoča(4), Medzilaborce(5), Poprad(6), Sabinov(7), Snina(8), StaráĽubovňa(9),Stropkov(10),Svidník(11),VranovnadTopľou(12).

RegionTrenčín(TN)isdividedinto9districts:Trenčín(0),BánovcenadBebravou(1),Ilava(2),Myjava(3), Nové Mesto nad Váhom(4), Partizánske(5), Považská Bystrica(6), Prievidza(7),Púchov(8).

origin

1 3 3

origin

0 0

cos ow n ij iji N j N

i j

t co d x

0 0 0

0cos rental ij ij ji N j N j N

i j

t cr d y cf y

2

5

4

1

2

5

4

Region Trnava (TT) is divided into 7 districts: Trnava(0), Dunajská Streda (1), Galanta(2),Hlohovec(3),Piešťany(4),Senica(5),Skalica(6).

RegionŽilina(ZA) isdivided into11districts:Žilina(0),Bytča(1),Čadca(2),DolnýKubín(3),Kysucké Nové Mesto(4), Liptovský Mikuláš(5), Martin(6), Námestovo(7), Ružomberok(8),TurčianskeTeplice(9),Tvrdošín(10).

Inputdata:8regions(8problemssolved),numberofdeliverydistricttownsandregionalcapitalineachregion(BA–4,BB–13,KE–6,NR–7,PO–13,TN–9,TT–7,ZA–11),distributioncentre(regionalcapital)ineachproblemisindexedasi=0,theshortestdistancesbetweenalldistrict towns and between distribution centre and each district towns is designated as cij,vehiclecapacityVwasset to12, thedemandofdistrict townswasassociatedwithnumberofinhabitant(0.00006m3perinhabitant),costperkmofownvehiclewassettoco=0.5€,costforrentedvehiclewassetasfollows:fixedcostperroutecf=15€,costperkmcr=0.6€.Thegoalwastominimizethetotalcostsforthedistributionineachregionandtodeterminethenumberofownandrentedvehicles,withrespectthefollowingrestrictions:theorigin0istheinitialnodeandalsothefinalnodeforeachownvehicleroute,thefinalpointincaseofrentedvehicleisthelastservednode.

The computational experiments were provided on the base of before mentioned data. ThemathematicalmodelwasimplementedinsoftwareGAMS(solverCplex12.2.0.0)onPCwithIntel®Core™i7‐3770CPUwithafrequencyof3.40GHzand8GBofRAMunderMSWindows8.TheresultsareshowninTable1a,1b.

Table1aRegionaldistribution

RegionBratislava

RegionBanskáBystrica

RegionKošice

RegionNitra

Route1 0‐1‐2‐3‐0 0‐10‐4‐1‐11‐12‐0 0‐1‐5‐3 0‐3‐0Route2 0‐2‐7‐8 0‐2‐4 0‐5‐0Route3 0‐3‐6‐5‐9 0‐6 0‐6‐2‐0Route4 0‐4‐1Totalcost 52 261.7 186 165Costforownvehicles 52 68.5 0 102Totalcostforrentedvehicles 0 193.2 186 63Fixcostforrentedvehicles 0 30 45 15Variablecostforrentedvehicles 0 163.2 141 48

Table1bRegionaldistribution

RegionPrešovRegionTrenčín

RegionTrnava RegionŽilina

Route1 0‐7‐0 0‐1‐3‐4‐0 0‐2‐0 0‐1‐4‐2‐0Route2 0‐4‐6‐3 0‐2‐8‐6‐0 0‐4‐3‐0 0‐6‐9‐5Route3 0‐9‐1‐11‐10‐5 0‐5‐7 0‐1 0‐8‐3‐10‐7Route4 0‐12‐2‐8 0‐5‐6Totalcost 255.2 159.1 158.1 198.7Costforownvehicles 14 104.5 58.5 38.5Totalcostforrentedvehicles 241.2 54.6 99.6 160.2Fixcostforrentedvehicles 45 15 30 30Variablecostforrentedvehicles 196.2 39.6 69.6 130.2

TheTable1providesthesolutionineachregioninSlovakia.Thefirst4rowsofthetableshowtheobtainedroutesinregion(iflastnodeisoftherouteisthenodeindexed0thantherouteisrealizedbyownvehicleandotherwisebyrentedvehicle)and5throwshows the totalcostofdistribution(objectivefunction)inselectedregion.ThedistributionsystemisdepictedonFig.2,wherethesolidlinesrepresenttheroutesofownvehiclesanddottedlinesrepresenttheroutesofrentedvehicles.

Fig.2

4.CONCLUSION

Thispaperconsiderstheversionofopenvehicleroutingproblem(OVRP)thatallowscombiningCVRPtogetherwiththepossibilitiesofopenroutes.Themathematicalformulationwasprovidedonthebaseofmixedintegerprogramming(MIP)withlinearobjectivefunctionandconstraints.So that formulation allows the use of standard software for solving MIP problems. ThecomputationalexperimentswerebasedonregionaldatainSlovakia.Thedistributionhastoheassuredfromregionalcapitaltodistricttownsineachregionusingtheownorrentedvehicles.Thegoalwastodeterminehowmanyownandrentedvehiclesmustbeusedineachregionso

thatdemandofalldistricttownsmustbemetwiththeminimalcost(thetotalcostconsistofthecostofownvehiclesandofthecostofrental).SoftwareimplementationwasrealizedinGAMS(solverCplex).

ACKNOWLEDGEMENTS

Thispaper issupportedbytheGrantAgencyofSlovakRepublic–VEGA,grantno.1/0245/15„Transportationplanningfocusedongreenhousegasesemissionreduction“.

REFERENCES

[1] Čičková,Z.,Brezina,I.:Anevolutionaryapproachforsolvingvehicleroutingproblem.In:QuantitativemethodsineconomicsmultiplecriteriadecisionmakingXIV.IURAEDITION,Bratislava,(2008),40–44.

[2] Čičková, Z.,Brezina, I.,Pekár, J.: Solving theReal‐lifeVehiceRoutingProblemwithTimeWindows Using Self OrganizingMigrating Algorithm. Ekonomický časopis 61(5/2013),497–513.

[3] Čičková,Z.,Brezina,I.,Pekár,J.:Openvehicleroutingproblem.In:MathematicalMethodsinEconomics2014.Olomouc:FacultyofScience,PalackýUniversity,2014124–127.

[4] Desrochers, M., Desrosiers, J., Solomon, M.: A New Algorithm for the Vehicle RoutingProblemwithTimeWindows.OperationsResearch.40(1992),342‐354.

[5] Fábry, J.,Kořenář,V.,Kobzareva,M.:ColumnGenerationMethod for theVehicleRoutingProblem–TheCaseStudy.In:MathematicalMethodsinEconomics2011[CD‐ROM].Praha,PROFESSIONALPUBLISHING,(2011),140–144.

[6] Miller, C.E., Tucker, A.W., Zemlin, R.A.: Integer programming Formulation of TravelingSalesmanProblems.JournaloftheACM,7(4/1960),326‐329.

FACILITYLOCATIONDECISIONUNDERDEMANDUNCERTAINTYANDTRAVELTIMEFLUCTUATION

EcemYıldıza,YelizEkincib

aInstituteofNaturalandAppliedSciences,IstanbulBilgiUniversity,Istanbul,Turkey

bDepartmentofIndustrialEngineering,IstanbulBilgiUniversity,Istanbul,Turkey

Abstract:This study solves the facility location problem of an IT service companywhosemainproblem is to arrive at the customer’s site in the shortest time when support is needed. Thisproblem shouldconsiderboth thedemandof thecustomersand thedistanceand/or travel timebetweenthecustomersandthecompany.Traveltimesanddemandarenotconstantsincetheyareaffected by many factors. Therefore, one should take into account uncertainty. This studyformulatestheproblembyconsideringthedemandas fuzzyandtraveltimeasvaryingbasedondifferenttimeintervals,whicharedefinedhourly,dailyandseasonally.Moreover,weconsiderthecases ofminimum speed, average speed andmaximum speed for travel time.We illustrate theapplication of the proposed framework using data of a company.We compare the proposedframeworkwiththetraditionaldistancebasedoptimizationapproachandshowtheadvantagesoftheproposedmethod.

Keywords:transportation‐locationproblem;fuzzydemand;traveltimefluctuation.

1.INTRODUCTION

Considering theprevious literature, thisstudycontributes to the literaturewidelybydecidingon the optimal facility location and route selection at the same time. It is based on bothuncertaintyofthetraveltimefromfacilitynodetothedemandnodeanddemanduncertainty.Moreover,theuncertaintiesaredefinedhourly,dailyandseasonally,whichcoversall thetimeintervalsthatthecompanyprovidesserviceinayear.Thus,allpossibleconditionsoftrafficanddemandvariabilityaretakenintoaccount.Whenthetransportationliteratureisexamined,itcanbeseenthateffectsofweatherconditionshavebeencommonlymodeledinthetransportationforecastingliterature[1,2,and3].

Moreover, timeofday is treatedasasignificant factorthatrepresentsthe levelofdemandforthetransportationservicesandthetrafficsituationduringdifferenthoursoftheday[4,5,6,and7].Similarly,manystudieshavedemonstratedthevariationoftrafficvolumeindifferentdaysofweekandvacationtimes[2,7].Besidesthis,thetraveltimeismodeledin3levels,i.e.worst,bestandmostlikely,inordertohandletheuncertaintybetter.

Forillustration,theproposedapproachsolvesthetransportationlocationproblemforaservicecompanywhichprovidesinformationtechnologies(IT)supportforitscustomers.Theproblemconsidersa“onefacilityandmanycustomers”case.Thecompany’smainproblemistoarriveatthe customer’s site in the shortest time when support is needed. Therefore travel time anddemand variability should be considered while selecting the optimal facility location and

optimal routes from the facility to the customers. The company and the customers are inIstanbulandIstanbulisoneofthemostcrowdedcitiesinEuropewithhightrafficdensity,whichmakestheproblemnecessarytobesolved.

The paper consists of three more sections. The second section introduces the proposedframeworkwhilesection3presentstheapplicationoftheproposedframework.Conclusionandfurtherresearchfinalizethepaper.

2.PROPOSEDFRAMEWORKFORTHETRANSPORTATION‐LOCATIONPROBLEMUNDERTRAVELTIMEFLUCTUATIONANDDEMANDUNCERTAINTY

Thisstudyproposesaframeworkforthetransportation‐locationprobleminvolvingtraveltimeand demand uncertainties. The travel time varies for each route based on time, hence it isuncertain.Moreover,thedemandofthecustomerisnotknownsincethereisnopastdataanditisasubjectofuncertainty,too.However,itisapparentthatitvariesbasedontimeintervalsforcompanies thatservetheircustomersoncall. In theproblemsettingthat isconsidered in thisstudy, thenumberof facilities tobe located isone,but thenumberofcustomers ismorethanone.Thecompanyservesthecustomerattheirlocationandwhenacallfromacustomercomestothecompany,anagentinthecompanydrivestothecustomer.Thecustomerdoesnothaveaconstraintaboutthearrivingtime;however,theagentshouldbeatthesiteofthecompanyassoonaspossible.Theassumptionintheproblemsettingisthat,thereisnocapacitylimitandthecompany always has enough employees to send to the customers for service.When all thesefactsareconsideredtogether,thefollowingframeworkisproposedtosolvethefacilitylocationproblem:

Step1.Determinealternativefacilitylocations.

Step2.Findtheroutesthatconnectthecustomerstothealternativefacilities

Step3.Determine thenumberof time intervals (r) thatwill cover all the servicehoursof thecompany.

Step4.FindtheprobabilitiesofdemandfromcustomeriattimeintervalkTheseprobabilitiesshowtheprobabilityofcustomerdemand.Inthisstudy,thecasethatthereisnopastdataaboutthecustomerdemandisconsidered.However,theexpertsinthecompanyhave some experience about the customer demand. Therefore, FRBS supported by expertopinion isused inorder tomodel thedata.AfterconstructingFRBS, thedemandprobabilitiescorrespondingtoeachtimeintervalkisderivedfromthissystem.

Step5.FindthetraveltimetocustomeriwhenroutejisselectedattimeintervalkThisinformationcanbefoundfromthetrafficdataandGeographicInformationSystems(GIS).Step6.Formulatetheproblemasamathematicalmodelandsolvetheproblemminimizingthetraveltimexdemand.Findtheoptimalroutesthatshouldbefollowedatspecifictimeintervals.

Given that there are n customers, m routes, r time intervals, and u alternative facilities, theproblemcanbeformulatedasanintegerprogrammingmodelbythefollowingequationset(1):

min∑ ∑ ∑ ∑ p ∗ t ∗ y 1

subjectto ∑ y 1∀ ∀ ∀ , ∑ z 1, y z ,y , z ∈ 0, 1 2

where

:probabilityofgettingacallfromcustomeriattimeintervalk

:traveltimetocustomeriattimeintervalkfromfacilitylwhenroutejisselected

:binaryvariablewhichdenoteswhetherroutejisselectedinordertogotocustomerifromfacilitylattimeintervalk

zl:binaryvariablewhichdenoteswhetherfacilitylisselected

The first constraint in the model ensures that only one route is selected from facility l tocustomer i at time intervalk.The secondconstraint is tomakesure thatonlyonealternativefacility is selected. The above mathematical model can be solved optimally by using integerprogrammingsolutiontechniques.

Step7.Comparetheresultswiththedistancebasedapproachwheretrafficdensityisnotconsidered.

3.APPLICATIONOFTHEPROPOSEDFRAMEWORK

The illustrationof theproposed framework isperformed foran information technologies (IT)supportcompanyinIstanbul.

Step1. The alternative facility locations are determined by the experts in the company. Thealternative facilitiesare selectedby consideringavailableoffice locations, the closeness to thecustomers, commercial areas, and closeness to the main transportation roads. There are 3alternativefacilitylocations,twoofthealternativelocationsareintheAsiansideofIstanbulandone is in theEuropeansideof the Istanbul.Twoof themarechosen fromtheAsiansidesincemostofthecustomersarethere.

Step2.Thecompanyhassixcustomersandthereare43alternativeroutesintotalthatconnectthe3alternativefacilitylocationstothe6customers.SinceIstanbulisacitythatconnectstwocontinents (namely, Europe and Asia), the customers and alternative facility locations are onbothsideofthecity,andtherearetwobridgesthatconnectthetwocontinents,thereareseveralalternativeroutesforeachcombination.

Step3.Determine thenumberof time intervals (r) thatwill cover all the servicehoursof thecompany.

Thenumberofthesetimeintervalsisdeterminedas160,whichcoversallthedaysinaweek,allhourlytimeintervalsinadayandallseasonsinayear.Therearefiveweekdays,8hourlytimeintervals that the companyserves (10.00‐11.00,11.00‐12.00,12.00‐13.00,13.00‐14.00,14.00‐15.00,15.00‐16.00,16.00‐17.00,17.00‐18.00)and4seasons.Hencethecombinationresults in160different time intervals. The aimwas to cover all possible time intervals since the trafficconditions differ substantially among weekdays, hours in a day and weathers. For instance,during rush hours, traffic is really crowded in Istanbul. Moreover, in winter the severityincreasesduetoincreasednumberofaccidentsandnecessityofdrivingslow.

Step4. FRBS isused to find thedemandprobabilities. Inorder to identify thevariables in theFRBSandtofuzzifythevariables,expertsareconsulted.ThefuzzyinferencesystemofferedbyMATLAB7.6.0fuzzylogictoolboxisusedinthisstudy.UsingtheIF‐THENrules,theprobabilityofgettingaservicedemandataspecifictimeintervalisestimatedbasedontheinputvariablesofweather condition, day ofweek and time of day. Thedetermined variables have effects onboththedemandofthecustomersandtrafficconditions.Theexpertsstatethatdemandofthecustomervariesacrossdayofweekandtimeofday.Italsovarieswithseasonssinceinsummer,mostofthepeopleareonvacationanddemanddecreasesrelatively.

In the fuzzification step, the membership functions of the input variables are defined usingexpertopinion.Timeofthedayisrepresentedbythelinguisticvariablesofearly,midandlate.Earlyisstandingforthebeginningoftheweekandlatefortheendoftheweek(weekendisnotconsideredsincethecompanydoesnotworkatweekends).Demandisalsorelatedtothetimeofthedaysothediscretehoursofthedayhavebeendefinedbyfourlevelsoflinguisticvariablesnamelyearly,mid,lateandverylate.Finally,thenumericalvalueof“probabilityofgettingacallfrom the company” has been used as the demand level, which is the output of the model.Demandlevelsarerepresentedbytheverylow,low,medium,highandveryhighterms.Usingtherule‐basedsystem,demandprobabilityderivationisachievedandusedintheoptimizationmodel. In the literature, triangular and trapezoidal fuzzy numbers are frequently utilized forfuzzy applications. In this study, both triangular and trapezoidal fuzzy numbers are used toconsider the fuzzinessof thedecisionelements.Themembership functionsaredefinedby theexpertsworkinginthecompany.FunctionsofoneoftheinputvariablesandoutputvariablearegiveninFigure1.

Figure1.Membershipfunctionsofoneinputvariable:dayofweek(ontheleft),andoutputvariable:demandlevel(ontheright).

Therulesaredefinedforeachcustomerandforeachseason(summer,fall,winter,spring).Themain difference among these customers is that, the demand level is never very high forCustomer 4. After constructing FRBS, the demand probabilities corresponding to each timeintervalkforeachcustomerarederivedfromthissystem.

Step5.We find the travel times for each route at time different time intervals from the pasttrafficdataof IstanbulMetropolitanMunicipality.Weconsider thecasesofminimum,averageandmaximumtraveltimeforeachrouteinordertomakeasensitivityanalysis.Hence,wecanmakethedecisionsundertheworstcase,bestcaseandthemostlikelycase.

Step6.Havingtheaboveexplanationsandmotivation,weusethetraveltimedata(takenfromIstanbulMetropolitanMunicipalityforlastthreeyears)anddemanddata(fromFRBS)tosolvethemathematicalmodel.Forallthescenarios,theresultsshowthatAF_3(AlternativeFacility3)hastheminimumobjectivefunctionvalue(bestscenario:87.4mostlikelyscenario:95.9,worstscenario:111.2).InfactitisexpectedtoselecteitherAF_2orAF_3,sincetheyareroughlyinthemiddleofall thecustomers.Theresultingoptimalroutes forAF_3areshown inFigure2.TheexplanationsabouttheoptimalroutesatdifferenttimeintervalsfromAF_3tothecustomersaregivenbelow.

Figure2.Alternativeroutesfromalternativefacility3(AF‐3)tocustomers

FromAF_3toC_1(Customer1),theadvantageousrouteisroute2inalltimeintervalsandinallscenarios while the optimal route from AF_3 to C_2 is route 1 in all time intervals for allscenarios.However,forthebestscenarioforAF_3andC_2,insummerroute1androute2havethesamevalueswhichyieldtoalternativesolutionsforthisseason.Thisresultisduetothefactthatmostofthepeopleareonvacationinsummerandtheroadsarelessbusy,sotraveltimesontheroutesareclosetoeachother.

Fortheworstscenario,theoptimalroutefromAF_3toC_3inspringandinwinterisroute2inthemiddlehoursoftheday,butinothertimeintervals,route1isoptimal.However,insummerinalltimeintervalsroute1isoptimal.Infall,itisadvantageoustouseroute2inlatehoursofday. However, for the best andmost likely scenarios, the optimal route is route 1 in all timeintervals.

For all scenarios, the results show that in spring fromAF_3 toC_4, the advantageous route isroute1 in themiddlehoursof thedaywhile it is route2 inother time intervals.Theoptimalrouteisusuallyroute2insummerforallscenarios.Fortheworstscenario,forfallandwinter,itis difficult to make a generalization since the optimal route changes among days and hours.However,forthebestandmostlikelyscenarios,infallandwintertheoptimalrouteisroute2.

For all time intervals and for all scenarios, the optimal route from AF_3 to C_5 is route 3.Thereforeindependentfrombeingpessimisticoroptimistic,thedrivercanselectroute3atalltimes.Again for all time intervals and for all scenarios, theoptimal route fromAF_3 toC_6 isroute2.

Step7.Whenwesolvetheproblembyusingsolelythedistancesinsteadoftraveltimes(asinthetraditionalapproach),wefindthatagainAF_3 is theoptimal location.Weseethattheoptimalroute for C_1 is route 2, for C_2 it is route 1. These results are similar to the results in theproposedapproach.However,forthebestcaseandforsummersometimesusingroute2isalsoadvantageousforC_2intheproposedapproach.ForC_3,theshortestdistancebelongstoroute

1.Intheproposedapproachtheoptimalrouteswereselectedasroute1androute2fordifferenttimeintervals.TheproposedapproachshowstheoptimalroutesforC_4aresometimesroute1andsometimesroute2.However,route2 isshorterthanroute1. In fact thesetworoutesusedifferentbridgesthatconnecttheAsiansideandEuropeansideofIstanbul,andsincethefirstbridge that isused in route2 isusuallybusier, it is sometimesadvantageous touse theotherroute(route1).Theshortestroutesarethesameastheoptimalonesintheproposedapproachforcustomer5and6.

4.CONCLUSIONANDFURTHERRESEARCH

Theresultsrevealedinthestudyshowthattheoptimalrouteandfacilitylocationdecisionsaretimedependent.Itisalsoseenthattraveltimevolatilityanduncertaintyhasasignificanteffectontherouteselections.Thisstudycontributesthefacilitylocationproblemwiththeselectionofoptimal routes for different time intervals simultaneously. The results show that, timedependent traffic conditions significantly affect the optimal route decision and uncertaindemand is also an important factor. Additionally, FRBS is seen as a useful technique for thedemand prediction. Moreover, the optimal routes change for different time intervals fordifferentscenariosandtheoptimalfacilitylocationisrobusttothesescenarios.

Best, worst and most likely scenarios are considered in this study in order to address theuncertaintyandfluctuation.Inthebestscenario,thetraveltimeonarouteatthatspecifictimeinterval is theminimumrealizedtravel time in therealdataset. In theworstscenario it is themaximumandinthemostlikelyscenarioitistheaverage.

Nowadays,technologyoffersustheavailabilityoftrafficdata,thereforeweshouldmakeuseofitmore.Thisstudyisanattempttoachievethis,byusingthepastdataonthetrafficconditionsondifferentroutesinaverybigandhighly‐populatedcity,namelyIstanbul.Furtherresearchcanbecarriedformanylocationstomanycustomerscase.Theimplementationislimitedinthisarticleanditcanbesolvedforabiggercaseinordertoseesignificantresults.Moreover,thestudycanbewidenedbyaddingthereturnroutes,too.

REFERENCES

[1] Tang,L.,andThakuriah,P.(2012).Ridershipeffectsofreal‐timebusinformationsystem:acasestudyintheCityofChicago,Transport.Res.PartC22146–161.

[2] Lai,W.‐H.andTsai,C‐T.(2009).Fuzzyrule‐basedanalysisoffirm’stechnologytransferinTaiwan’smachineryindustry,Exp.Syst.Appl.,36(10),12012–12022.

[3] Khattak,A.J.andPalma,A.(1997).Theimpactofadverseweatherconditionsontheplant,Proc.Inst.Electr.Eng.,121(12),1585–1588.

[4] Bowman, J.L. and Ben‐Akiva, M.E. (2001). Activity‐based disaggregate travel demandmodelsystemwithactivityschedules,Transport.Res.PartA:PolicyPract.35(1),1–28.

[5] Bates, J.,Polak, J., Jones,P., andCook,A. (2001).Thevaluationof reliability forpersonaltravel,Transport.Res.PartE37(2–3),191–229.

[6] Chen,M.C. andWei, Y. (2011). Exploring time variants for short‐termpassengerflow, J.Transport.Geogr.19(4),488–498.

[7] Yang, C.W. and Chang, C.C. (2011). Applying price and time differentiation tomodelingcabinchoiceinhigh‐speedrail,Transp.Res.PartE47(1),73–84.

FORECASTINGTHEINVENTORYLEVELOFMAGNETICCARDSINTOLLINGSYSTEM

BratislavLazića,NebojšaBojovićb,GordanaRadivojevićb*,GoranaŠormaza

aUniversityofBelgrade,MihajloPupinInstitute,SerbiabUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Forecasting the inventory levelofmagneticcards isan importantprerequisite for thefunctioning of a tolling system. This paper presents the approach to forecasting the requirednumberofmagneticcardsinthetollingsystemontheBelgrade–NišhighwaysectioninSerbia,asa prerequisite for purchasing and forming an optimal inventory level. The 1, 1, 21, 1, 0 model was developed and applied for forecasting the monthly inventory level ofmagneticcardsneededfortheyears2015and2016.

Keywords:timeseries,forecasting,magneticcardsinventories,tollingsystem.

1.INTRODUCTION

The tolling system is a complex system which enables the highway owner toll charging onvarioussections,andfastandhigh‐qualityusageofhighwaystotheusers.TollchargingintheRepublic of Serbia is within the competence of Public Enterprise "Roads of Serbia", which isresponsible for organization of work, supervision, management and technical provision forfunctioningofthesystem.

TollchargingiscarriedoutonthreesectionsofhighwaysintheRepublicofSerbia:Belgrade–Niš,Belgrade–ŠidandBelgrade–Subotica.TheBelgrade–ŠidandBelgrade–Nišsectionshaveaclosedsystemoftollcharginginwhichthetollamountdependsonvehiclecategoryandtraveldistance (PE “Roads of Serbia”, 2015). A highway user gets a magnetic card at entrance tollplazas and pays toll at the exit plaza, after card reading. The users get themagnetic cards atentranceplazas fromdistributors(automaticregime)orpre‐magnetizedcards fromoperators(manualregime).Atcollectingplazas,thecardsareputaside,namelyputinstoragesuntiltheyaredestroyed.Theentranceplazasmusthaveanadequate levelof card inventories toenableundisturbedfunctioningofvehicleentranceandtrafficonhighway.

PE“RoadsofSerbia”dealswiththeorganizationoftollcollectionandaccordinglyhastoplanthefollowing: magnetic card acquisition, making sufficient amount of pre‐magnetized cards,organization ofworkers at collection stations, the number of entrance/exit lanes at plazas aswell as regularmaintenance of toll collection system. It is necessary to precisely forecast thenumber of magnetic cards in the toll collection system for these plans to meet the future

* g.radivojevic@sf.bg.ac.rs; gordana.radivojevic@pupin.rs

requirementsofthesystem.Theforecastisbasedonthenumberofissuedmagneticcardsintheprevious time periods, namely in analyzing themonthly time series of the number of issuedmagneticcards.

AMultiplicativeSeasonalARIMAclassofnon‐stationarytimeserieswasappliedinthispaperfortheanalysisoftimeseries.Theusedtimeseriesincludesdataonthenumberofmagneticcardsfrom January 2000 till November 2014 on the Belgrade – Niš highway section. The obtainedmodel was applied for forecasting the required number of magnetic cards for the periodDecember2014–December2016.

Thepaperhasbeenorganized in the followingway.Thesecondpartdescribes the theoreticalbasis and Box‐Jenkins methodology for analyzing and modelling time series. The third partshowsapplicationoftheBox‐Jenkinsmethodologyformodellingtheinventorylevelofmagneticcards for toll collection system on the Belgrade – Niš highway section. The conclusion andpossibledirectionsoffutureresearcharegiveninthefourthpart.

2.TIMESERIESANALYSIS

Time series can be described as a set of temporal ordered realizations of a random variableduringaseriesofsuccessivetimeperiods.Byanalyzingchangesintherandomvariableintime,itissometimespossibletodefineaperformancemodelwhichenablesforecastingfuturestates.

Seasonal time series consist of periodical changeswhich repeat in certain time intervals. Thesmallest repetitive time period is called seasonal period ( ). When realization of a randomvariable repeats after seasonal period with certain regularities, it may be expected that thevalues in seasonal periodswill correlatemutually. High values of autocorrelation function onseasonal realizations are indicators of seasonal non‐stationary process. Seasonal differenceoperator 1 isusedeliminatingseasonalnon‐stationarity.

By combining seasonal difference operator with ARIMA model, Box and Jenkins defined theMultiplicative Seasonal ARIMA model (Box et al., 1994; Brockwell, Davis, 2002; Cryer, Chan,2008)as:

1 1 (1)

Inthisrelation islagoperator, and areautoregressivepolynomialandmovingaveragepolynomial,and i areseasonalautoregressivepolynomialandseasonalmoving average polynomial. Multiplicative Seasonal ARIMA model with seasonal period ismarkedasARIMA , , , , ,where and areordersofnon‐seasonalautoregressivepolynomial and seasonal moving average polynomial, and are orders of seasonalautoregressivepolynomialandmovingaveragepolynomial, isnon‐seasonaland isseasonaldifferencingdegree(Boxetal.,1994;Wei,2006;Cryer,Chan,2008).

Box and Jenkins proposed a methodology for analysing and modelling time series whichincludesfourstages(Boxetal.,1994):

ModelIdentification, ModelEstimation, ModelValidation,and Forecasting.

ModelIdentificationconsistsofaseriesofproceduresovertimeseriesdata,inordertochosethecorrespondingmodelfromARIMAmodelset.Theseproceduresincludethefollowing(Tsuietal.,2014):

Timeseriesplotting–enablesdetectingtrendoccurrenceandseasonalchangesintimeseries.Ifitisnotedthatmeanvalueisnotconstantthenlogarithmicdatatransformationisapplied.

AutocorrelationFunction(ACF)andPartialAutocorrelationFunction(PACF)–basedontheoreticalcharacteristicsofACFandPACFfunctionsforARIMAmodels, and ordersof ARIMAmodel are assumed. These diagrams can indicate seasonal changes in timeseries.

Unitroottests(AugmentedDiskey‐Fullertest,Philips‐Perrontest)–testsarecarriedoutbychanging lengthof lagoperatoruntil stationaryseries isobtained.Existenceofunitrootsindicatestothenecessityofseriesdifferencing.

Seasonal HEGY test – if there are seasonal unit roots then seasonal differencing isapplied.

ThedescribedproceduresgivethesetofvaluesofARIMAmodel , , , , , parameters.

ModelEstimation implies that coefficients of autoregressive polynomial andmoving averagepolynomialaredeterminedforeverypossibleparameter , , , , , namelyfittingofmodelis inquestion. In thisphase themethodsofnon‐linear least squares andmaximum likelihoodestimationare applied.Thebestmodel is theonewhichhas the lowest information criterion,whichmaybe(KirchgässnerandWolters,2007;Cryer,Chan,2008):

Akaikeinformationcriteria: ln 2 Baesianinformationcriteria: ln ln

NormalizedBaesianinformationcriteria: ln

Hanan‐Quininformationcriteria: ln 2 ln ln

ModelValidationinvolvescheckingresiduals–thedifferencesofrealizedandforecastedvalues.ACFandPACFresidualfunctionsareusedforthisorsomeoftheformalapproaches(eg.Ljung‐Box Q‐statistics). If there is no autocorrelation in residuals, the model is adequate. On thecontrary,modelisinadequateanditisnecessarytogobacktoidentificationstage.

Forecastingimpliesusingtheobtainedmodeltoforecastvaluesfor stepahead.Bycomparingthe obtained values with past values of series realization, which were not used whileestimatingmodelparameters, it canbeevaluatedwhether the chosenmodelhasgoodqualityforecast.Modelestimationcanbedoneaccordingtoanyofthefollowingcriteria:

Meanabsolutedeviation ∑

Meanabsolutepercentageerror ∑

Meansquareerror ∑

Rootmeansquareerror √

3.ARIMAMODELFORFORECASTINGNUMBEROFMAGNETICCARDS

In order to forecast the necessary inventory level ofmagnetic cards for toll collection on theBelgrade–Nišsection,thetimeseriesoftotalnumberofissuedmagneticcardswasusedfromtheperiodJanuary2000–November2014(PE“RoadsofSerbia”,2000‐2014).Thetimeseriesconsistsof179monthlyrealizations,outofwhichthefirst168wereusedforestimatingmodelparametersandthelast11formodelperformanceestimation.

3.1.Identificationoftimeseriesmodel

TimeseriesplottingofissuedmagneticcardsisshowninFigure1.Ananalysisofthetimeseriesplotshowsnon‐stationarytimeserieswithanincreasingtrendandseasonalvariation.Seasonal

variationsarejustifiedbythefactthatthelargestnumberofhighwayusersisintheperiodJuly‐August during the holidays, with heavy transit of traffic from Western European countriestowardsTurkey,BulgariaandGreece,andviceversa,andthatthelowesttrafficonthehighwayoccurs during winter weather conditions, especially in months with a great number ofexceptionallycolddaysandsnowfall.

Figure1.Timeseriesoftotalnumberofmagneticcardsfortheperiod2000–2014.

In order to stabilize variance, a logarithmic time series transformation was carried out. Thetransformedseries inthispaper is titled lntimeseries.Anautocorrelationfunction(ACF)andpartial autocorrelation function (PACF)were formed for ln timeseries, shown inFigures2(a)and2(b).

Figure2.Autocorrelationandpartialautocorrelationfunctionlntimeseries

Great correlation function values were marked in all time periods. Gradual decrease ofautocorrelationfunctionvaluesindicatetotheexistenceofalong‐termtrend,andpointsouttothe fact that ln time series differencing is necessary to be carried out when making model,namely that 1 in ARIMA model. High autocorrelation function values on annual periodsindicate to the need for seasonal differencing on model. By applying differencing operator1 1 stationarytimeserieswasobtained.Inthemodelidentificationphase,formaltests for unit roots presence were carried out. The results of Dickey‐Fulle test and Philips‐Persontestconfirmedstationarity.

3.2.ModelEstimation

ModelidentificationshowedthatARIMA , 1, , 1, modelsprovedtobemostsuitablefor the observed ln time series. In the model estimation stage, all possible models from

alternativemodel set,where , 0, 1, 2 and , 0, 1,were considered.ByusingMATLABEconometrics Toolbox®, the autoregressive polynomial and moving average polynomialcoefficientwereestimatedaswell as seasonalautoregressivepolynomial andmovingaveragepolynomial coefficient for all observedmodels. For each of thesemodels, Akaike informationcriterion(AIC)andBayesianinformationcriterion(BIC)valueswerefoundandLjung‐BoxQ‐testcarried out. ARIMA 1, 1, 2 1, 1, 0 model represents the best model for ln time series,because the lowest AIC and BIC criterion values was obtained. The parameter values of thismodelforlntimeseriesareshowninTable4,andthemodelrelationisgivenbyequation(2):

1 1 1 0.354878 1 0.715782 1 0.109707 0.416844 2

Table4.ARIMAmodelparametersvaluesforlntimeserie

Parameter Value Standarddeviation t‐statistics

AR1 ‐0.354878 0.210124 ‐1.6889

SAR12 ‐0.715782 0.0409527 ‐17.4783

MA1 ‐0.109707 0.210714 ‐0.52065

MA2 ‐0.416844 0.134475 ‐3.09978

Variance 0.00240232 0.000191872 12.5204

AIC ‐522.4981

BIC ‐500.6303

MAPE 0.2614

RMSE 0.0579

R‐squared 0.9433

3.3.ModelValidation

Diagnostic checking confirms that the chosen model is stationary and that there are noredundant parameters. Residuals analysis (Ljung‐Box Q=11.6542, p‐value 0.92747>0.05)indicate residuals are strictly random and that no significant autocorrelation exists betweenresidualsondifferenttimeperiods.

3.4.MagneticCardsinventorylevelforecasting

ThedifferencesbetweenanticipatedvaluesobtainedbymodelandrealizedvaluesoftimeseriesfortheperiodJanuary‐November2014aregiveninFigure3(a).

Figure3.Inventorylevelofthemagneticcardsforecast

Asthemeanabsolutepercentageerror(MAPE)androotmeansquareerror(RMSE)valuesaresmall (MAPE 10%)(Tsuiet al.,2014), itmaybeconcluded that theobtainedmodel isgoodenough for forecasting the inventory level of magnetic cards on the Belgrade – Niš highwaysection.

By using the obtainedmodel, the number ofmagnetic cards for the periodDecember 2014 –November2016wasforecasted.Forecastvalueand95%confidenceintervalaregiveninFigure3(b).

4.CONCLUSION

Theaimofthispaperwastoforecastinventorylevelofmagneticcardsneededtoservedemandof PE “Roads of Serbia”. High quality forecasting enables the company to properly planacquisitionofmagneticcards,themakinganddistributionofpre‐magnetizedcards,therequiredpersonnel and organize regular maintenance of toll charging system. The Box‐JenkinsmethodologywasappliedforthechoiceoftheadequateARIMAmodelbasedontimeseriesofmonthlynumberofissuedcards.TheobtainedARIMA 1,1,2 1,1,0 modelfortheBelgrade‐Nišhighwaysectionshowedhighperformanceandmaybeusedforhigh‐qualityforecastingofthemagneticcardsinventorylevel.Forecastingoftherequirednumberofmagneticcardsfortheyears2015and2016waseffectedbyapplyingmodel.Forfurtherimprovementofthemodel,itis necessary to also include other variables which affect traffic intensity on highway (grossnationalproduct,realwagesrate,gasprices,etc.).

ACKNOWLEDGMENT

Thisworkwas supportedby theMinistry of Education and Science of theGovernment of theRepublicofSerbiathroughtheprojectsTR36005andTR36022,fortheperiod2011‐2015.

REFERENCES

[1] BoxG., JenkinsG.,ReinselG.(1994).TimeSeriesAnalysisForecastingandControl.ThirdEdition.PrenticeHall.

[2] BrockwellP.,DavisR.(2002).IntroductiontoTimeSeriesandForecasting.SecondEdition.Springer.

[3] CryerJ.D.,ChanK.(2008).TimeSeriesAnalysisWithApplicationsinR.Springer.

[4] PE “Roads of Serbia” (2000‐2014). Publication counting traffic on public roads of theRepublicofSerbia.

[5] PE“RoadsofSerbia”(2015).www.putevi‐srbije.rs.(March,2015).

[6] Kirchgässner G., Wolters G. (2007). Introduction To Modern Time Series Analysis.Springer.

[7] MatlabStatisticsToolboxUser’sGuideR2014b.(2014).MathWorks.

[8] Tsui W.H.K, Baili H.O., Gilbery A., Gow H. (2014). Forecasting of Hong Kong airport’spassengerthroughput,TourismManagement42,62‐76.

[9] Wei W. (2006). Time Series Analysis: Univariate and Multivariate Methods. PearsonAddisonWesley.

MODELINGDELAYOFLOADINGGOODSONRIVERTRANSPORTVESSELS

IvanĐorđevića,GordanStojićb,IlijaTanackovc,SinišaSremacd

aUniversityofNoviSad,FacultyofTechnicalSciences,Serbia,ivan.djordjevic87@gmail.combUniversityofNoviSad,FacultyofTechnicalSciences,Serbia,gordan@uns.ac.rscUniversityofNoviSad,FacultyofTechnicalSciences,Serbia,ilijat@uns.ac.rsdUniversityofNoviSad,FacultyofTechnicalSciences,Serbia,sremacs@uns.ac.rs

Abstract: Transport and handling operations performed in cargo transport terminals cansignificantly influence speedand efficiencyof loading.Loadingofgoodshas itsown spatialandtemporaldimension.Managing temporaldimensioncontributes toamoreefficientrealizationoftheloadingoperationandaffectstheperformanceoftransportandrationalhandlingoperationsinfreight transport terminals.Theanalysisofparameterswhich influence the timeof loadingmaycontributetoefficientsolutionofpracticalproblemsthatslowdownloadingofgoods.Inthispaper,asimulationmodeloftheprocessofdelayinloadingbulkcargoisdeveloped.ThemodelwastestedontheroastedironpyriteinthemarineterminalportSabac.Thetestresultshaveshownthatdelayis a significant factorwhich influence loading of goods.Measuring of delay duration for eachloadedvesselwascarriedoutinrealconditionsofloadingandhandlingoperations.Astheoutputofthismethod,thearrangementofregulardurationofdelaysanddisorders isobserved.Furtheranalysisshowslinkbetweenregulardurationanddisordersandopportunitytopredictandcontroldisordersduringloadingofcargo.

Keywords:loading,delay,orderly,disorder,model

1.INTRODUCTION

Transportofgoodsbyinlandwaterwaysisthemostcapaciousandmostcost‐effectivemodeoftransport.Besides,environmentalpollutionisincomparablylower,whichisveryimportantfactin the planning of sustainable transport network as seen from the view of today's climatedisturbances.Pan‐EuropeantransportcorridorVIIisanatural,highlycapaciouswaterwaythatallows the formation of large river transport structures for the transport of mass andcontainerized cargo [6]. Rational managing and planning of reloading operations in portterminals are an important segment in business of port terminals. Possibility of predictingdurationof loadingoperations iscertainlyan importantorganizationalandpricingdimension.Theobjectiveofthepaperistomodelandpredictdelaysandtheirdurationonloadingofbulkcargoontovessels.

2.SIMULATIONMODEL

2.1Descriptionofthemodel

Themodelencompassesariverterminalforbulkcargowheretimesofloadinganddelaysweremeasuredoutforrivervessels.Attheriverterminalloadingplacethereisaconveyorof400t/hcapacity with three to five trucks for serving it. Loading times were measured continuouslyduringoneyearand250opentyperivervesselswereincluded.Measuringwasperformedforonetypeofgoods,whichismassivebulkcargo‐roastedironpyrites[7].

On the duration of times of loading and delays unfavourable weather conditions, such asprecipitation,moistureofgoods,highwaterlevel,availabilityoftrucksservingtheconveyor,etc.havesignificant impact.Delaysalso included technicaloperation for loadingcargo.Thedelayswererecordedfromthemomenttheloadingwasinterruptedtilltheloadingofthesamevesselwascontinued.Whereloadingwasstoppedmorethanonce,thedelaywascalculatedasthesumofallbreaks.

Intheanalyticalpartoftheresearch,thestatisticalparametersandborderofdisordersindelayswereestablished,respectively[3].Thedelaysthataremeasuredoutonloadinghaveextremelyhighvaluesasaconsequenceofdisorderduetounfavourableimpacts.

The occurrence of these values cannot be controlled because of unforeseeable circumstances.Forthisreasondisorderswereanalyzedasaparticulardelaycategory.Namely,thedelaysaredividedintwocategories:regulardurationanddisorder.

Onthebasisofhypothesisofexponentialdistributionforthetwocategoriesofdelayduetotheirownrandomness,thesimulationwascarriedoutforbothregulardurationofdelayanddisorder.The simulation part implies implementation of Monte Carlo numerical method and use ofsoftware to simulate great number of iterations [4]. Further analysis will show lineardependency between regular duration of delay and disorder, as well as the possibility ofpredictingdisorders.

2.2Delaysfromtheaspectofdisordersize

Thedelayinloadingofbulkcargoatthefreightterminalwasmeasuredoutinrealconditionsforeach loaded vessel. Measuring included all influential factors and they are included in finaldurationofdelay.Theresultsofmeasuring,wheredelayisacontinualrandomvariable(X),havecontinual distribution andmay take random value. Statistical data processingwas performedand it showed wide range of the measured values [3]. The Figure 1. shows the histogramcontainingthemeasureddelayscomparedwithnormaldistribution.

Figure1.Histogramwithmeasuredoutdelays

IntheTable1.theoutputresultsofthestatisticaldelayprocessingaregiven.

Sincethefrequencyofthedurationofdelaysmeasuredforallvesselsisconcentratedonlowervalues (statistical analysisgiven in theTable1. shows that themedian is =50min, and themaximumvalue of themeasured delays ismax = 6360min), the Figure 2. is obviously reallysmallpercentofextremehighdurationofdelays.Thesehaveledresearchontheconclusionthattherearedisordersinthedurationoutages.

Indeterminingclasseswhereborderofdisorderwillbedefined, theoptimalrelationbetweensquared mathematical expectation, and variance, was searched. As the disorderclasses, the values of delay above 350 min, 450 min, 550 min, 600 min were taken intoconsideration.

Table1.Descriptivestatisticofdelay

Statisticalparameters Obtainedvalues

Numberofsamples(n) 250

CentraltendencyMathematicalexpectation,E(x) 182,8Median, 50Standarderror,ε 32,6

DispersionofvaluesSrandarddeviation,σ 510,2Maximalvalue,max 6360Minimalvalue,min 0

ShapeofdistributionSkewness 8,44Kurtosis 92,16

Indeterminingofdisorderbordersindelaysthefollowingrelationistakenasoptimal:

0,82 (1)

and delays in timewith duration over 450min are declared as disorders. Duration of delaysbelow450minisregularduration.

Figure2.Cumulativefunctionofmeasuredoutdelays

Observing the cumulative function at theFigure2. it canbe concluded that 87%of all delayshavedurationbelowborderof450min.Onthiswayitisdefinedparticipationofalldisordersindelayswith13%.Disordersabove650minaretakingonly5%intotalnumberofmeasuring.

Inorder to assessmatchingwithmathematical distribution, goodnessof fit test isperformed.TheAnderson–Darlingtest(p>0.05)iscarriedoutusingprobabilityplotswith95%confidence

interval[1].TheTable2.showsAnderson–Darling(AD)statisticandtheassociatedpoweroftest(p‐value).

Table2.Anderson–Darling(AD)statisticformathematicaldistributions

Specifyofmathematicaldistribution ADstatistic p‐value

Normal 49,581 <0,0053‐ParameterLognormal 3,295 nodata2‐ParameterExponential 51,761 <0,0103‐ParameterWeilbull 8,144 <0,005SmallestExtremeValue 73,494 <0,010LargestExtremeValue 31,722 <0,0103–ParameterGamma 12,969 nodataLogistic 33,022 <0,0053–ParameterLoglogistic 5,524 nodata

FromtheTable2.itisobviousthatloadingdelayshavefeatureofrandomvariablesandthereisnomatchingwithanymathematicaldistribution(ADstatisticislargerthan0andp–valueisnot>0.05).

2.3ImplementationofMonteCarlomethod

As the statistical verification confirmed that delays does not match any mathematicaldistribution,itisassumedthatthefirstcategoryofdelay(regularduration)matchexponentialdistribution,whiledisordersmatchErlangdistribution.Bothcategoryofdelaywillbepresentedbyprobabilitydensityfunction(pdf)ofrelatedassumeddistributionandwillbesimulated.

Toperformsimulationparameterofprobability,thedensityfunctionisdetermined[5]:

1 0,02 (2)

as a border parameter between regular duration and disorder. Regular duration of delay insimulation contains value of the parameter λ>0,02, while disorder contains value of densityfunctionparameter,λ<0,02.

Onthebasisofthemeasuredvaluesofdelayforvessels,thecalculatedmathematicalexpectationofregulardurationisE(x)=62,94min,andofdisorders,E(x)=991,88min.Standarddeviationforregular duration is σ=75,57; and for disorders, σ=1093,91. Mathematical expectation is thecrucialparameterforfurtherevaluationofefficiencyofsimulation.

MonteCarlosimulationisperformedforn=10 iterationsofrandomlychosenvalues,x=[0,1]offunctionargument[2].

Insimulationofregulardurationofdelay,itwasusedprobabilitydensityfunctionofexponentialdistribution, while disorders were simulated by probability density function of Erlangdistributionofthesecondorder.

Aftern=10 iterationsthesimulatedmathematicalexpectationofregulardurationtakesvaluesthat are extremely close to mathematical expectation of the measured values (approximaterelativeerroris0.003%).Statisticalgoodnessoffittestshowspoweroftest(p=0,75)morethan0,05accordingtoconditionp>0,05toacceptassumedhypothesis.Probabilitydensityfunctionofregulardurationshowsexponentialdistribution(Figure3.).

Figure3.Histogramofprobabilitydensityfunctionforregularduration

ThecategoryofdisorderwassimulatedaccordingtotheprobabilitydensityfunctionofErlangdistribution of the second order. Also, it shows insignificant discrepancy of the simulatedmathematical expectation in regards to themathematical expectation of themeasured values(approximaterelativeerroris0.017%).Statisticalgoodnessoffittestshowspoweroftest(p=0,250) more than 0,05 according to condition p>0,05 to accept assumed hypothesis and theassumeddistributionisvalid(Figure4.).

3.MODELFORPREDICTINGDISORDERS

Aftersimulationofprobabilitydensitydistributionforbothcategoriesofdelay,thequestioniswhetherthedisorderscanbemanaged?Howtoassesspotentialdurationofdisorders?

Figure4.Histogramofprobabilitydensityfunctionfordisorders

The answers to these questions are given through the analysis of disorders in percentage inregardstoregulardurationofdelay.Disorderswereanalyzed inrangesof5%beginning fromthe mathematical expectation of regular duration (E(x) = 63 min, disorder is 0%) till themathematical expectation of disorders (E(X) = 992 min, disorder is 100%). The averagemathematical expectation (AME) in the Table 3. was estimated from 10 measurements ofsimulatedmathematical expectations for regular duration and disorders, and represents totaldurationofdelay.

Table3.CalculatedAMEinregardstosizeofdisordersinpercentage

Percentage(%) 0 5 10 15 20 25 30 35 40 45 50

AME(min) 63 109 155 201 248 294 340 386 432 478 525

Percentage(%) 55 60 65 70 75 80 85 90 95 100

AME(min) 571 617 663 709 755 801 848 894 940 992

OnthebasisofthemeasuredAMEgivenintheTable3.themathematicalmodelfordurationofdelayisdeveloped:

∗ (3)

Where:

–isthedurationofdelay;

–isthedurationofdisorders(AME=992min);

–istheregularduration(AME=63min);

p–isthepercentageofdisorders.

4.CONCLUSION

Everyreductionofdelayinloadingtimehasdirectinfluenceonreductionintotalloadingtimeper transport unit andmore efficient utilization of conveyor and staff.Modelling of disordersandtheirpredictionisimportantelementofrationalbusinessinfreighttransportterminals.

Insecondpartoftheresearch,themodelforpredictionofdisorderswasdeveloped.Thelinearregression model with extraordinary correlation was obtained. Control of unpredictableduration of potential disorders is an important factor in efficient organizing of reloadingoperationsinfreightterminals.

Theproposalforfurtherresearchistoanalyzethedelayimpactontotalloadingduration.Itisnecessary to take into consideration the possibility to predict loading delay, as well as toprognoze loading duration on the basis of potential delays. Also, further analysis of impactshould include the trucks available to serve at river terminal, the time of truck turnaround,numberofworkingmachinesandtechnologyofgoodspreparationfor loading.Onthebasisoftheseanalysesthemodelfordelaypredictingcouldbeupgraded.

ACKNOWLEDGMENT

TheauthorsacknowledgethesupportofresearchprojectTR36030,fundedbytheMinistryofScienceandTechnologicalDevelopmentofSerbia.

REFERENCES

[1] BassI., (2007).SixSigmaStatisticswithExcelandMinitab,TheMcGraw‐HillCompanies,Inc.,USA.

[2] Fishman,S.G.,(1999).MonteCarloConcepts,AlgorithmsandApplications,SpringerSeriesinOperationsResearch,Springer.

[3] Fraser,D.A.S.,(1958).Statistics–AnIntroduction.JohnWiley&Sons,Inc.,NewYork,USA.

[4] Mun, J., (2006).ModelingRisk:ApplyingMonteCarlo Simulation,RealOptionsAnalysis,Forecasting,andOptimizationTechniques,Wiley.

[5] Đorić,D.,etal.(2007).Atlasraspodela,Matematičkifakultet,Beograd.

[6] www.dunavskastrategija.rs

[7] www.elixirzorka.rs

MULTICRITERIADECISIONMAKINGFORDISTRIBUTIONCENTERLOCATIONSELECTION‐SERBIACASESTUDY

MilanDobrotaa,DraganaMacurab‡,MilicaŠelmićb

aLOGITd.o.o,Belgrade,SerbiabUniversityofBelgrade,FacultyofTransportandTrafficEngineering,SerbiaAbstract: The process of location selection for distribution center can be analyzed as amulti‐criteria decision problem and has a strategic importance for many retail companies. Theconventionalmethodsforfacility locationselectionareinadequatefordealingwiththeimpreciseor linguistic nature of assessments and measurements. To overcome this difficulty and allowincluding quantitative and qualitative criteria, fuzzy multi‐criteria decision‐making method isproposed.Theaimofthispaperistousefuzzyanalytichierarchyprocess(AHP)fortheselectionofdistributioncenterlocation.Thepresentedmethodhasbeenappliedtolocationselectionproblemof a retail company in Serbia. After determining the criteria that affect the facility locationdecisions,fuzzyAHPisappliedtotheproblemandresultsarepresented.

Keywords:distributioncenter,locationselection,fuzzyAHP

1.INRODUCTION

Distributioncenterplaysavitalroleinmodernsupplychainsanditlocationcouldaffecttothesuccess,orfailureoflogisticcompany'sbusiness.Itpresentsanorganizationalunitthatallowsdifferentiating between various types of stock in a site. Selecting appropriate location ofdistribution center is very important decision for retail firms because they are costly anddifficult toreverse,andtheyentaila longtermcommitments.Decisionsrelatedtodistributioncenter’slocationhaveanimpactonoperatingcostsandrevenues.Forinstance,apoorchoiceoflocation might result in excessive transportation costs, a shortage of qualified labor, loss ofcompetitive advantage, inadequate supplies of raw materials, or some similar condition thatwouldbedetrimentaltooperations(Stevenson1993).

The common procedure for making decision about the best location usually consists of thefollowingsteps(Stevenson1993):1)Decideonthecriteriathatwillbeusedtoevaluatelocationalternatives;2)Identifycriteriathatareimportant;3)Developlocationalternatives;4)Evaluatethealternativesandmakeaselection.

There are many criteria that influence the location decisions of a distribution center. In ourpaper,wetakeintoaccountsixgroupsofcriteriabasedonexpertopinionandliteraturereview,qualitative and quantitative. These are: capital (investment) costs, transportation costs,operating costs, strategic factors, supply chain and logistics factors and other factors (social,demographical,geographical,andpolitical).Eachofthosecriteriahastheirownsubcriteria,and

‡ d.macura@sf.bg.ac.rs

intotalweconsider24factors.Thesecriteriaareusedtochoosethebest locationamongfouralternatives(NoviSad,Šimanovci,NoviBeograd,andNiš).

Ascanbeseenfromrelevantliterature,theconventionalmethodsforfacilitylocationselectionare inadequate for dealing with the imprecise or linguistic nature of criteria and analyzedfactors. To overcome thisdifficulty and allow including variety of criteria, fuzzymulti‐criteriadecision‐making method is proposed. In this paper we use fuzzy AHP, where the ratings ofvariousalternativelocationsundervarioussubjectivecriteriaandtheweightsofallcriteriaarerepresentedbyfuzzynumbers.

Therestofthispaperisorganizedasfollows.BriefliteraturereviewisgiveninSection2.Thenin Section 3, fuzzyAHP approach is introduced. A numerical example is given in Section 4 toillustratetheproposedmethod.Andfinally,Section5containsconcludingremarks.

2.BRIEFLITERATUREREVIEW

Thissectioncontainsbrief literaturereviewon facilityselectionproblem insupplychain. Inadiscrete facility location, theselectionof thesiteswherenewfacilitiesare tobeestablished isrestrictedtoafinitesetofavailablecandidatelocations.Thesimplestsettingofsuchaproblemis theone inwhichp facilitiesaretobeselectedtominimizethetotal(weighted)distancesorcosts for supplying customer demands. This is the so‐called p‐median problem which hasattractedmuchattention in the literature(Daskin1995,DreznerandHamacher2004,ReVelleandEiselt2005).

Significant literature review of facility location models in the context of supply chainmanagementisgiveninpaper(Meloetal.2009).Theyidentifiedbasicfeaturesthatsuchmodelsmust capture to supportdecision‐making involved in strategic supply chainplanning. In theirchapter(Daskinetal.2005)outlinedtheimportanceoffacilitylocationdecisionsinsupplychaindesign.Ertugrul andKarakasoglu,2008presented fuzzyAHPand fuzzyTOPSISmethods forafacilitylocationselectionproblemofatextilecompanyinTurkey.Theauthorsofpaper(Kuoaetal. 2002) developed a decision support system for locating a new convenience store. Afeedforward neural network with error back‐propagation learning algorithm was applied tostudy the relationship between the factors and the store performance. Kahraman et al. 2003solved facility location problems using different solution approaches of fuzzy multi‐attributegroup decision‐making. The paper (Meng et al. 2009) addressed a novel competitive facilitylocation problem about a firm that intends to enter an existing decentralized supply chaincomprisedof three tiersofplayerswith competition:manufacturers, retailersandconsumers.Thepaper(Özcanetal.2011)consideredAHP,TOPSIS,ELECTREandGreyTheoryappliedonthewarehouseselectionproblem.

3.BASICASSUMPTIONSOFFUZZYAHP

The Analytic Hierarchy Process is well known multi‐criteria decision making approachintroducedbyThomasSaaty(Saaty1994).Somespecificsofthisapproachare:thehierarchicalstructureofasystem(goal,criteria,alternatives);theelementsofonelevelshouldbecomparedwith each other based on the elements of higher level; generating the pairwise comparisonmatrices;Saatyscaleisusedfordefiningtherelativeimportanceoftheelements;theabilityforverificationoftheexperts’consistency;etc.HerearethespecificsofappliedfuzzyAHPapproach.

Step1:Defineagoal,criteria,subcriteriaandalternatives.Then,theimportanceofonecriteriontotheotherbasedonthegoal,subcriteriatotheotherbasedoncriterionwhichtheybelongto,andfinally,theimportanceofonealternativetotheotherbasedoneachsubcriterionseparately.After the pairwise comparison matrices are developed, consistency validated (with eventual

adjustmentof importance), thealternatives rankcanbedetermined.For comparisonbetweentheseelements,theSaatyscaleisused(Table1).

Table1.Saatyscale

Crispvalue

Importance Fuzzyvalue Reciprocalcrispvalue

Fuzzyreciprocalvalue

1 Equal (1,1,1) 1/1 (1,1,1)2 Intermediatevalues (1,2,4) 1/2 (1/4,1/2,1)3 Weak (1,3,5) 1/3 (1/5,1/3,1)4 Intermediatevalues (2,4,6) 1/4 (1/6,1/4,1/2)5 Strong (3,5,7) 1/5 (1/7,1/5,1/3)6 Intermediatevalues (4,6,8) 1/6 (1/8,1/6,1/4)7 Demonstrated (5,7,9) 1/7 (1/9,1/7,1/5)8 Intermediatevalues (6,8,10) 1/8 (1/10,1/8,1/6)9 Absolute (7,9,11) 1/9 (1/11,1/9,1/7)

WiththeaimtopresentthebasicequationsforapplicationofthefuzzyAHPapproach,herewillbeconsiderthefuzzymatrixR.Theelementofthismatrixisfuzzytrianglenumberaij=(aijl,aijm,aijr)andshowsapreferenceofoneelementtotheother(i=1,2,..q,j=1,2,..q,whereqisthenumberofcriteria).LetusassumethatthismatrixRisthepairwisecomparisonofcriteria,basedonthegoalofamodel.

1,1,1 , , ⋯ , ,1

,1

,1

1,1,1 , ,

⋮ ⋱⋱ ⋮1

,1

,1

⋯⋯ 1,1,1

Step 2: Overall weight of elements, criteria, based on matrix R is calculated by followingequations(Jieetal.2006):

q

i

q

jijr

q

i

q

jijm

q

i

q

jijl acacac

1 13

1 12

1 11 ,, (1)

q

jijr

q

jijm

q

jijl adadad

13

1,2

1,1 (2)

qittc

dz t

it ,...,2,1;3,2,1,4

(3)

Where: c1, c2, c3 are the sum of all left, middle and right values of fuzzy triangle number aij,respectively;aijl,aijm,aijrareleft,middleandrightvalueoffuzzynumberaij,respectively;d1,d2,d3arethesumbycolumnsofallleft,middleandrightvaluesoffuzzymatrixR,respectively,ziistheoverallweightofcriteriai,andzi1,zi2,zi3areleft,middleandrightvalueofzi,respectively.

Step3:Basedoncertainequations(Jieetal.2006),thematrixofweightedperformanceofeachalternative,Pwithfuzzyelementspk,isdefined(k=1,2..,n,wherenisthenumberofalternatives).Thence, the matrix of total weighted performance, S with fuzzy elements sk, is obtained byfollowingrelation:

1

n

k kk

s p

(4)

Step4:Defuzzificationisaprocedureofchoosingaparticularoutputvalue.Thematrixoftotalweightedperformance,S, isobtainedbyequation(4).Avariable isdefinedby fuzzysetwhichhavetobetransformedintocrispvalue.Theprocessoftransformationoffuzzynumber~s=(sl,sm,sr)intocrispnumberispresentedbelow(LiousandWang1992).

* *( ) 1 * *m l m r r ms s s s s s s ,where , 0,1 (5)

Where and are the preferences and risk tolerance of decision makers, respectively. Thepessimismandoptimismofdecisionmakerscanbeexpressedbythesevalues.

4.FUZZYAHPFORDISTRIBUTIONCENTERLOCATIONSELECTION‐CASESTUDY

RetailerAthatoperatesinSerbiaforover10yearshasaretailstoresindozenofmajorcitiesbutnotinall.Amongothers,citieswithstoresareBeograd,Niš,Čačak,Valjevo,Šabac,Zaječar,andObrenovac. In order to support further expansion on the market, company has decided toinvestigate the possibility to setup a distribution center (DC) fromwhich retail storeswill besupplied.With doing so, company is looking to achieve several benefits such as being able tonegotiatebetterconditionswithsuppliers(e.g.volumeand logisticsrebates),reducethestockholding costsandbenefit fromriskpooling, increase responsivenessof supply chain, etc.Pre‐selected locations for setting up a DC are: Novi Sad, Šimanovci, Novi Beograd and Niš. Eachlocation has some advantages and disadvantages. Since locations have many different andspecificattributes,expertsagreethat theyshouldbeall taken intoconsiderationand that it isnotpossible to relay solely toquantitative criteria. In order to support suchdecisionmaking,fuzzyAHPapproachisselectedtosupportit.Basedonexperts’opinionallrelevantelementsofthemodelaredefined(Figure1).

Figure1.Quantitativeandqualitativefactors

4.1ResultsandDiscussion

Afterdescribedmethodologyhadbeenappliedandexpertshadevaluatedallscores,Šimanovcihasshowntobethebestalternative(Figure2).Thisresultisobtainedforλ=0.5andα=0.5.Itisvery closely followed by alternative‐Novi Beograd, while Niš and then Novi Sad are lesspreferredlocationsalthoughveryclosetoeachother.Speakingofcriteria,expertshavedecidedthatquantitativeindicatorsaccountfor2/3of importanceformakingdecision,andqualitativehave1/3ofimportance.Figure3displayssortedcriteriabytheirglobalweights.

Figure2.Finalalternativesrank

Figure3.Finalsubcriteriaweights

Withoutdoubt,totaloutboundtransportationcostsarebyfarthemostimportantcriteria.Thiscriteriaaloneaccountsfor1/3ofimportancefortotaldecision.Itisfollowedbyhandlingcostsandinboundtransportationcosts,sothethreecriteriabeartogether50%ofimportance.Thenthe fourth criteria is qualitative one, which is support of location for overall supply chainstrategic fit,whichdependsonhowcompanywants tocompeteon themarket, andsimilar tosixthrankedcriteriaofresponsivenessimprovement,thuswecouldsaythattheresponsivenessof supply chain is important criteria for decisionmakers. Fifth is price of acquiring land, andthese 6 criteria together account for 70% of importance, while the remained 30% is spreadbetween others. Regarding the alternatives, although Novi Beograd surely optimizes thetransportationcosts,whichisthemostimportantcriteria,butduetoloweroperatingcostsandlesscapitalneededforinvestment,Šimanovcigainedtheoveralladvantage.Thestrategic,supplychain/logisticsandotherfactorshaveveryclosevaluesforbothlocations.

5.CONCLUSION

The model for distribution center location selection is developed in this paper. AuthorssuggestedthefuzzyAHPforsolvingtheproposedproblem.Afterdefiningthepotentiallocations,asalternatives,andallrelevantfactorsintheprocessofselection,i.e.criteriaandsubcriteria,theproposedapproachisappliedtolocationselectionproblemofaretailcompanyinSerbia.Finally,

thebestalternativecomparedtootherfour,isobtained.ProposedmodelselectedŠimanovciasthebestsolution.Frommanagers’pointofviewthislocationcandecreasetransportationcosts,increase competitive advantages, etc. Moreover, proposed framework can help managers inmakingfutureDClocationdecisions,byusingthis intuitiveandeffectiveapproach.Inordertotakebothquantitativeandqualitativefactors intoconsideration,aswellasthepreferenceandrisktolerance(optimism/pessimism)factors,presentedfuzzyAHPmodelgivesasolidbasisforanalyticalbutpracticaldecisionmakinganditsapplication.

ACKNOWLEDGMENT

ThisworkwassupportedbytheMinistryofEducation,ScienceandTechnologicalDevelopmentoftheRepublicofSerbiathroughtheprojectsTR36002andTR36022fortheperiod2011‐2015.

REFERENCES

[1] Daskin, M. (1995). Network and Discrete Location: Models, Algorithms, and Applications. Wiley, New York.

[2] Daskin,M.,Snyder,L.,Berger,R.(2005).FacilityLocationinSupplyChainDesign.Chapterin:LogisticsSystems:DesignandOptimization.39‐65.

[3] Drezner, Z., Hamacher, H.W. (Eds.) (2004). Facility Location: Applications and Theory,Springer,NewYork.

[4] Ertuğrul, I.,Karakaşoğlu,N.(2008).Comparisonof fuzzyAHPand fuzzyTOPSISmethodsforfacilitylocationselection.IntJAdvManufTechnol,39,783–795.

[5] Jie,L.H.,Meng,M.C.,Cheong,C.W.(2006).Webbasedfuzzymulticriteriadecisionmakingtool.InternationalJournaloftheComputer,theInternetandManagement,14(2).1‐14.

[6] Kahraman,C.,Ruan,D.,Doǧan,I.(2003).Fuzzygroupdecision‐makingforfacilitylocationselection.InformationSciences,157,135–153.

[7] Kuoa,R.J.,Chib,S.C.,Kaoc,S.(2002).Adecisionsupportsystemforselectingconveniencestorelocationthroughintegrationof fuzzyAHPandartificialneuralnetwork.ComputersinIndustry,47,199–214.

[8] Lious, TS., Wang, MJJ. (1992) Ranking fuzzy numbers with integral value. Fuzzy SetsSystem,50(3),247‐255.

[9] Melo, M.T., Nickel, S., Saldanha‐da‐Gama, F.(2009). Facility location and supply chainmanagement–Areview.EuropeanJournalofOperationalResearch,196,401–412.

[10] Meng,Q.,Huang,Y.,Long,R.(2009).Competitive facility locationondecentralizedsupplychains.EuropeanJournalofOperationalResearch,196,487–499.

[11] Özcan, T., Çelebi, N., Esnaf, S.(2011). Comparative analysis of multi‐criteria decisionmaking methodologies and implementation of a warehouse location selection problem.ExpertSystemswithApplications,38.9773–9779.

[12] ReVelle, C.S., Eiselt, H.A.(2005). Location analysis: A synthesis and survey. EuropeanJournalofOperationalResearch,165,1–19.

[13] Saaty,Th.(1994)FundamentalsofDecisionMakingandPriorityTheorywiththeAnalyticHierarchyProcess.RWS

[14] Stevenson,W.J. (1993). Production /operationsmanagement,4thedn.RichardD. IrwinInc.,Homewood

SOLVINGVEHICLEROUTINGPROBLEMSVIASINGLEGENERICTRANSFORMATIONAPPROACH

KarloBalaa,DejanBrcanovb,§,NebojšaGvozdenovićb

aFacultyofPhilosophy,UniversityofNoviSadbFacultyofEconomics,UniversityofNoviSad

Abstract:We present a simple data structure capable to solve a broad class of vehicle routingproblems.Ourmodelingapproachinvolvesvehiclecapacityconstraintsandextendstothetacticallocationdecisionsoflocationroutingproblems.Theuniquedatastructureaccompaniedbyasinglegeneric transformation allows an effective search of the solution space.We provide simulatedannealingresultsforstandardbenchmarksthatconfirmthequalityoftheproposedalgorithm.

Keywords:Vehiclerouting,Locationrouting,Simulatedannealing.

1.INTRODUCTION

Heuristic approaches prevail as the solution approaches for various vehicle routing problems(seee.g.[7]).Althoughinventedtobesimple,fastandgenericitisoftenthecasethatapplyingheuristictoaspecificvehicleroutingproblemrequirednewdatastructuresanditerationstepsdesign,oratleastsomesignificantadjustments.

On the other hand, supply chain practice required optimization of the entire logistics valuechain, and thus led to anewgroupofproblems that combines facility locationproblemswithvehicleroutingproblems.Theseproblems,knownaslocationroutingproblems,accordingto[3],neededredesignoftheneighborhoodsearchtechniquesofthesolutionspaces.

In this paper, we present a simple data structure accompanied by a single generictransformation which leads to an effective search of the solution space. The power of thetransformation is that most of the classical transformations like insertion, swap, 2–opt, andsomeothers,canbeseenas itsspecialcases.Ourmodelingapproachinvolvesvehiclecapacityconstraintsandextendstothetacticallocationdecisionsoflocationroutingproblems.

Thepaperisorganizedin5sections.TheintroductionisfollowedbydatastructurepresentationandtransformationdescriptioninSection2.InSection3,weshowhowvehicleroutingproblemscan be modeled with the new data structure, and in Section 4 we present some results oncapacitatedlocationroutingproblems.Inthelastsectionweprovidesomefinalremarks.

§ dejanbrcanov@gmail.com

2.ROUTESANDTRANSFORMATIONSINCIRCULARBUFFER

Everyvehicleroutingproblemhasadepot,avehicleandacustomer.Weaddadepot,avehicleand/oracustomerasmarkerstothepositionsofacircularlist.

Figure1.Circularlist.

IfapositioninthelistismarkedasadepotD,thenallpositionsbeforethenextdepotposition,wherethecounterclockwisedirectionisassumed,areassociatedwiththedepotD.IfapositionisassociatedwithavehicleV,allpositionsbeforethenextvehicleposition,wherethecounterclockwise direction is assumed, belong to the vehicle V, i.e. all customers on those positionsshouldbeservedbyvehicleV(Figure2).

customer depot vehicle

1 2a b c d e f g l

First route at depot 1 Second route at depot 2

The list of routes of depot 1

Figure2.Positionsandmarkersinalist.

Notethatdepotpositionscontainvehiclemarkers,whichweinterpretasthefirstroutesstarts.

Wenextdefineageneric transformation.Thepowerof this transformation is thatmostof theclassicaltransformationslikeinsertion,swap,2–optandsomeothers,canbeseenasitsspecialcases.Inordertoperformtransformationoneshouldchoose2‘outer’positions,AandBinthelist,andtwo‘inner’positionsCandD(Figure3).Thefirststepinthetransformationistoinvertthesub‐listbetweenAandB.Inthesecondstep,weinvertthesub‐listbetweenCandD.

Figure3.Generictransformation.

Itcanbeeasilyobservedthatmakes traditional insertiontransformation ifAcoincideswithCandDandBareconsecutivepositions.IfAandCononesideandDandBontheothersideare

both pairs of consecutive positions, the transformation is simply the traditional swaptransformation.

3.MODELINGROUTINGPROBLEMS

In this sectionweshowhowseveralknownvehicle routingproblemscanbemodeledviaourapproach.Weusecapacitatedlocationroutingproblem(CLRP),consideredin[]asourreferenceproblem. As all routing problems, it contains customers, vehicles and depots. Each customerorderedacertainquantityofaproductthatneedstobedelivered.Themaingoal istomakearoutingplansuchthattheoverallwarehousingandtransportationcostsareminimized.

Moreformally,weconsiderthatthesetoflocationsofcustomersanddepots,accompaniedwiththecostmatrixcontainingallpair‐wisetravelingcostisgiven.Eachdepothasitsopeningcostofopeningandthecapacity,whilecustomerhasitsdemand.Thefleetofvehiclesishomogeneous,henceeachvehiclehastheusagepriceandthecapacity.Thetotalcostofacapacitatedlocationroutingplanisthesumofthecosts foropeningdepots,usingvehiclesandtotaltravelingcost.Theultimategoalistomakeaplansuchthatallcapacityconstraintsaresatisfied,whilethetotalcostsareminimized.

WecanmodelCLRPusingaframeworkpresentedintheprevioussection.IfDisthenumberofdepots,C thenumbercustomersandV thenumberofvehicles (V≥D),wecreateacircular listwithV+C positions,whereDpositions aremarkedwith both depot and vehiclemarkers,V‐Dwith only vehicle markers and C positions with customer markers. The scenarios of closeddepotsornotactivatedvehiclesaregivenintheFigure4.

1 2b c e f g l

1 3b c d e f g l

Not activated vehicles

2

Depot 1 is closed

Figure4.Scenarioswithemptyvehiclesandcloseddepot.

Finally,variousroutingproblemscanbeconsideredasspecialcasesofCLRP,andtheirmodelingisboilsdowntosimplificationofthecircularbuffer.Forexample,multi‐depotVRPproblemsareobtainedwhen depot opening costs are set to zero, and vehicle usage costs set according tomulti‐criteriapriorities.On theotherhand,onecanconsideradding timeconstraints,withoutaddingcomplexityinthegivenstructure,andthuscapturevehicleroutingproblemswithtimewindoworworkingtimeconstrains.

4.CALCULATIONRESULTSFORCLRP

WegivesometestresultsonstandardbenchmarkproblemsforCLRPinthissection.InTable1our algorithm that uses presented framework and relies on simulated annealing approach isnamedSIMPGEN(fromsimpleandgeneric).ThealgorithmsGRASP.MAPM,LGRSTandSALRP,togetherwiththeresultsinthecorrespondingcolumnaretheresultsfrom[6],[2],[5],and[9]respectively.

Thebenchmarkproblemsaredescribed in [1], [6], and [8]. InTable1, thecolumns ‘Problem’,‘Cust’, ‘Dep’, ‘BKS’ contain the names of instances, the number of customers, the number ofdepots and the best know solution result, respectively. For each of the 5 algorithms the totalcost,thegaptowardsthebestknownsolutionandtheCPUtimesaregiven.ThecomparisonoftheCPU times isnot relevant since tests fordifferentalgorithmswereperformedondifferentmachines.

Table1.ComparisonresultsforCLRP

3.CONCLUSION

Wepresentedaframeworkformodelingvariousvehicleroutingproblems.Itusespositionsinacircularlistandaccompaniedwithgenerictransformationofferspowerfultoolforsearchingthesolution spaces of considered problems. We tested it on CLRP benchmark problems andobtainedpromisingresultsthatarecomparabletootherapproaches.

REFERENCES

[1] Barreto S.S. (2004). Análise e Modelização de Problemas de localização‐distribuição[Analysis andmodelling of location‐routing problems]. PhD thesis, University of Aveiro,campusuniversitáriodeSantiago,3810‐193Aveiro,Portugal.

[2] Duhamel, C., Lacomme, P., Prins, C., Prodhon, C. (2008). A memetic approach for thecapacitated locationroutingproblem. InProceedingsof theEU/meeting2008workshoponmetaheuristics for logistics and vehicle routing. University of Technology of Troyes,France.

[3] NagyG.,SalhiS. (2007).Location‐routing: Issues,modelsandmethods.Eur. J.Oper.Res.177(2):649–672

[4] Laporte G., Norbert Y., Arpin D. (1986). An exact algorithm for solving a capacitatedlocation‐routingproblem.Ann.Oper.Res.6(9):293–310

[5] PrinsC.,ProdhonC.,RuizA.,SorianoP.,WolflerCalvoR. (2007).Solving thecapacitatedlocation‐routing problem by a cooperative Lagrangean relaxation‐granular tabu searchheuristic.TransportationSci.41(4):470–483

[6] Prins C., Prodhon C., Wolfler Calvo R. (2006). Solving the capacitated location‐routingproblembyaGRAPScomplementedbyalearningprocessandpathrelinking.4OR,4:221‐238

Problem Cust Dep BKS Cost Gap CPU Cost Gap CPU Cost Gap CPU Cost Gap CPU Cost Gap CPU

Christ50 50 5 565.60 599.1 5.92% 2.3 565.6 0.00% 3.8 586.4 3.68% 2.4 565.6 0.00% 52.8 565.60 0.00% 43.062

Christ75 75 10 844.40 861.6 2.04% 9.8 866.1 2.57% 9.4 863.5 2.26% 10.1 844.4 0.00% 126.8 848.85 0.53% 129.81

Christ100 100 10 833.40 861.6 3.38% 25.5 850.1 2.00% 44.5 842.9 1.14% 28.2 833.4 0.00% 330.8 833.43 0.00% 238.61

Das88 88 8 355.78 356.9 0.31% 17.3 355.8 0.00% 34.2 368.7 3.63% 17.5 355.8 0.00% 226.9 355.78 0.00% 157.33

Das150 150 10 43919.90 44625 1.61% 156 44011.7 0.21% 255.0 44386.3 1.06% 119.2 43919.9 0.00% 577.0 43919.90 0.00% 328.28

Gaspelle 21 5 424.90 429.6 1.11% 0.2 424.9 0.00% 0.3 424.9 0.00% 0.2 424.9 0.00% 18.3 424.90 0.00% 9.609

Gaspelle2 22 5 585.10 585.1 0.00% 0.2 611.8 4.56% 0.3 587.4 0.39% 0.2 585.1 0.00% 16.6 585.11 0.00% 5.703

Gaspelle3 29 5 512.10 515.1 0.59% 0.4 512.1 0.00% 0.8 512.1 0.00% 0.4 512.1 0.00% 23.9 512.10 0.00% 11.078

Gaspelle4 32 5 562.20 571.9 1.73% 0.6 571.9 1.73% 0.8 584.6 3.98% 0.6 562.2 0.00% 27.0 562.22 0.00% 3.579

Gaspelle5 32 5 504.30 504.3 0.00% 0.5 534.7 6.03% 1.0 504.8 0.10% 0.5 504.3 0.00% 25.1 504.33 0.01% 4.546

Gaspelle6 36 5 460.37 460.4 0.01% 0.8 485.4 5.44% 1.4 476.5 3.50% 0.7 460.4 0.01% 31.7 460.37 0.00% 61.218

Min27 27 5 3062.00 3062 0.00% 0.4 3062.0 0.00% 1.0 3065.2 0.10% 0.3 3062.0 0.00% 23.3 3062.02 0.00% 1.297

Min134 134 8 5709.00 5965.1 4.49% 49.6 5950.1 4.22% 110.5 5809.0 1.75% 48.3 5709.0 0.00% 522.4 5709.00 0.00% 346.36

Average 61.23 6.62 4487.62 4569.07 1.63% 20.28 4523.25 2.06% 35.62 4539.41 1.66% 17.58 4487.62 0.00% 154.05 4487.97 0.04% 103.11

Algorithms SIMPGENGRASP MAPM LRGTS SALRP

[7] Toth,P.,Vigo,D.,Eds.,(2002).Thevehicleroutingproblem,SIAMmonographsondiscretemathematicsandapplications,Philadelphia.

[8] Tuzun, D., Burke. L.I., (1999.)A two‐phase tabu search approach to the location routingproblem.EuropeanJournalofOperationalResearch,116:87‐99

[9] Yu, V. F., Lin, S. W., Lee, W., Ting, C. J. (2010). A simulated annealing heuristic for thecapacitatedlocationroutingproblem.Computers&IndustrialEngineering,58,288–299.

THECAPACITATEDTEAMORIENTEERINGPROBLEM:BEECOLONYOPTIMIZATION

DraganaDrenovaca,RankoR.Nedeljkovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Inthispaperthecapacitatedteamorienteeringproblemhavebeenappliedasatoolforcarrierstoavoidemptyreturnswhentrucksdonotworkwithfullload.Thesoftwarebasedonbeecolonyoptimization technique isdeveloped forsolving theproblem.Numericalexample issolvedandtheresultisshowntodepictthepossibilityoftheproposedalgorithm.

Keywords:orienteeringproblem,beecolonyoptimization,electronicauction.

1.INTRODUCTION

Orienteering is a sportwhich is amixture of cross‐country running andnavigation through aforest,usingamapandcompass.Anumberofcontrolpointseachwithanassociatedscoreareplacedintheforestandtheirlocationsaremarkedonthecompetitors'maps.Competitorsstartatintervalsof,forexample,aminuteandarerequiredtovisitasubsetofthecontrolpointsfromthestartpoint(node1)soastomaximizetheirtotalscoreandreturntotheendpoint(noden)withinaprescribedamountoftime(Goldenetal.1987,Tsiligirides1984).

There isanalogybetweensportorienteeringandtheproblemwhichconsistsofdeterminingapath,limitedbyTmax,thatvisitssomeofthevertices(eachwithascore)ofagraph,inordertomaximizethetotalcollectedscore.Duetotheanalogy,theproblemwasnamedtheorienteeringproblem (OP). In the literature (Laporte and Martello 1990, Gendreau et al. 1998) theorienteeringproblemisalsoknownastheselectivetravelingsalesmanproblem.Inaddition,thesame problem in the literature (Fillet et al. 2005) can be found as a variant of the travelingsalesmanproblemwithprofitsinwhichthecostisreferredtotheconstraintandthecollectedprofit to the objective function. Orienteering problem (Kim et al. 2013) is the special case ofteam orienteering problem (TOP) in which there are teams of several competitors, eachcollecting scores during the same time span. The TOP is also known as the multiple tourmaximum collection problem (Butt and Cavalier 1994) or the vehicle routing problem withprofits.ThecapacitatedversionofTOPwasstudiedbyArchettietal.(2009,2010).Theauthorstriedtodevelopmethodasadecisionsupportwhendemandandofferoftransportationservicedidnotmatch.

In thispaper the capacitated teamorienteeringproblem is solvedby applying theBee colonyoptimization.Theproposedalgorithmisusedforsolvingtheproblemwhencarrierstrytoavoidemptyreturnsbecausethefleetofvehiclestheyowndoesnotworkwithfullload.Theproblemwas successfully solved by the developed metaheuristic algorithm. The proposed algorithmmightbe a tool for carrierswhen considerbidsof potential customers through the electronicauctionprocess.

2.THECAPACITATEDTEAMORIENTEERINGPROBLEM

GivenacompleteundirectedgraphG=(V,E),whereV=1,…,nisthevertexsetandEistheedgeset. Vertex 1 is a depot form identical vehicles of capacity Q, while the remaining verticesrepresentpotentialcustomers.Anedge(i,j)Erepresentsthepossibilitytotravelfromcustomeri tocustomer j.Anon‐negativedemanddi andanon‐negativeprofitpi isassociatedwitheachcustomer i (di=pi=0). A symmetric travel time tijand cost cij are associatedwith each (i, j)E.Eachvehiclestartsandends its touratvertex1,andcanvisitanysubsetofcustomerswithatotaldemandthatdoesnotexceedthecapacityQ.Theprofitofeachcustomercanbecollectedby one vehicle at most. In the following we suppose that tij=cij for each edge (i, j). In thecapacitatedteamorienteeringproblem(CTOP)asubsetofthepotentialcustomersavailablehastobeselected.Theobjective is tomaximize the totalcollectedprofitwhilesatisfying, foreachvehicle,atimelimitTmaxonthetourdurationandthecapacityconstraintQ.

LetΩ=r1,...,r|Ω|bethesetofpossibleroutesforavehicle,thatis,thesetofroutesstartingandending at vertex 1, visiting at most once each potential customer, satisfying the capacityconstraintQandthetimelimitTmax.Letaik=1ifrouterkΩvisitscustomeri,aik=0otherwise.Let ck be the total profit generated by route rkΩ:

Viiikk pac . The CTOP can be stated as

follows:

kr

kk pc max (1)

Subject to: 1\Vi xakr

kik

,1 (2)

mxkr

k

(3)

kk r 0,1x , (4)

Thedecisionvariablesxkindicatewhetherrouterk isusedornot.Constraints(2)ensurethateachcustomer isvisitedatmostonce.Constraint (3) limits thenumberofvehiclesused tom.Solving the linear relaxation of model (l)‐(4) necessitates the use of a column generationtechnique,due to the sizeof . Columngeneration isbasedon two components: a restrictedmasterproblemandasubproblem(Archettietal.2009).

Goldenetal.(1987)provethattheOPisNP‐hard.TheTOPisalsoNP‐hard(Chaoetal.1996).This implies that exact solution algorithms are very time consuming and for practicalapplicationsheuristicsandmetaheuristicswouldbenecessary.

TherearesomeinterestingapplicationsoftheOPandtheTOPintheliterature:thecasewhentruckfleetdeliversfueltoconsumersondailybasis(Goldenetal.1984),applicationintourism(Souffriauetal.2008),militaryapplications(Wangetal.2008).

Asitissaidbefore,theCTOPwasstudiedbyArchettietal.(2009,2010).Severalheuristicsandexact algorithms for solving the problem are presented in both papers. Archetti et al. (2007)proposed two variants of a generalized tabu search algorithm and a variable neighborhoodsearchalgorithmforthesolutionoftheTOP.

3.BEECOLONYOPTIMIZATION

Thebeecolonyoptimizationmetaheuristic(Teodorović2008)wasdevelopedandusedbyLučićand Teodorović (2001, 2003). Artificial bees represent agents, which collaboratively solvecomplex combinatorial optimization problems. Each artificial bee is located in the hive at thebeginning of the search process, and makes a series of local moves, thus creating a partialsolution. Bees incrementally add solution components to the current partial solution and

communicate directly to generate feasible solutions. The best discovered solution of the firstiterationissavedandtheprocessofincrementalconstructionofsolutionsbythebeescontinuesthroughsubsequentiterations.

Artificialbeesperformtwotypesofmoveswhileflyingthroughthesolutionspace:forwardpassorbackwardpass.Forwardpassassumesacombinationofindividualexplorationandcollectivepastexperiencetocreatevariouspartialsolutions,whilebackwardpassrepresentsreturntothehive, where collective decision‐making process takes place. It is assumed that bees exchangeinformation and compare the quality of partial solutions created, based on which every beedecides whether to abandon the created partial solution and become again uncommittedfollower, continue to expand the samepartial solutionwithout recruitingnestmates,ordanceand thusrecruitnestmatesbeforereturning to thecreatedpartial solution.During thesecondforwardpass,beesexpandpreviouslycreatedpartial solutions,afterwhich theyreturn to thehiveinabackwardpassandengageinthedecision‐makingprocessasbefore.Seriesofforwardandbackwardpassescontinueuntilfeasiblesolutionsarecreatedandtheiterationends.

4.SOLVINGTHECAPACITATEDTEAMORIENTEERINGPROBLEMBYBEECOLONYOPTIMIZATION

Drenovacetal. (2014)developed thealgorithmbasedon theBCO forsolving theorienteeringproblem.Thealgorithmwasmodifiedandadapted forcapacitatedvehicle fleet.Descriptionofthealgorithmfollows.

Thehivewithartificialbees is located innode1.Beesbounce fromthehiveandbeginsearchprocessofareaofallowablesolutions.LetVibethebee’sbenefitofchoosingtheithnodetoserve.Itisacceptedthatthebee’sbenefitwhenselectinganodeisevengreaterifthescoreisgreaterandthetraveltimeisshorter.Namely,itcouldbepresentedinthefollowingway:

li

ii

t

SV (5)

whereSiisscoreinnodei,tliisthetimerequiredtoreachnodei,fromthelastselectednodel.

Letpibe theprobability thatabeewillchoosethe ithnode.Eachroutehas tostartandendatnode1.Beforethenodeselection,asubsetofthosenodesthatsatisfythemaximallengthoftheroute (their inclusion does not exceed the maximal length) as well as the maximal capacity(whendemandofpotentialnode isaddedto thepartialdemand, thesumdoesnotexceedthemaximal capacity of vehicle), has to be determined. Later, nodes are being chosen randomlyfromthesubset.

Logitmodelwasadoptedasamodelofchoiceandtheprobabilityofselectionis:

m

i

VVV

V

ieee

ep

...21

(6)

wheremisthecardinalnumberofthesubset.

During the first forwardpass eachbee chooses apredefinednumberofnodes for each route.Havingreturned to thehive,beesstart tocommunicate.Theycalculatevaluesof theobjectivefunction as the ratio of the total score and the total time required to visit chosen nodes andcompare their partial solutions. In the proposed algorithm, the objective function of eachbeecomprisestimebecausethegreaterrouteefficiencythegreaternumberofincludednodesandthe greater achieved score. Then, bees make decision about their loyalty to the solution(whethertokeepthesolutionortoabandonit).

Let j betheobjectivefunctionvaluegeneratedbythejthbee(j= b,1 ,wherebisthenumberof

bees). Letjnorm be the normalized value of the objective function value. It is calculated as

follows:

minmax

min

j

norm j,

jnorm [0,1],j= b,1 (7)

Where max and min aretheminimumandthemaximumvalueoftheobjectivefunction.

Theprobability that the jthbeewillbe loyal to itspartial solutionat thebeginningof thenextforwardpassiscalculatedasfollows:

uuj

jnorm

ep

max

1 ,j= b,1 (8)

whereu isordinalnumberoftheforwardpassandmaxnorm ismaximalnormalizedvalueofthe

objectivefunction.

Aftermakingdecisionsabouttheloyaltytotheirsolutions,beesgatherinthedancefloorarea.Thebeesthatdecidetokeeptheirpartialsolutionsstartdancingandthusrecruituncommittedbees.Uncommittedfollowerschoosewhichoftheloyalbeestofollowinthenextforwardpass.

Duringthesearchprocesstimeisupdatedandwhenbeescannolongerexpandthesolution,thesearchends.Havingcomingbacktothehiveafteranarbitraryforwardpass,thebeescancreatepartialaswellasfinalsolutions.Despiteofthat,allofthemparticipateininformationexchange,evaluationofsolutionsanddecision‐makingconcerningtheloyaltytosolutions.

If thenumberof loyalbees isequal torbefore thenext forwardpass, theprobability that theuncommittedbeewilljointhekthloyalbeeinthenextforwardpassisequalto:

rnormnormnorm

knorm

eee

epk

...21

,k=1,…,r (9)

Based on these probabilities, uncommitted bees are coupledwith committed bees. Bees thenstart a new forward pass flying together to the last node of partial solutions generated bycommittedbees.Afterthat,eachbeeindividuallyextendsitspartialsolution.

Each iteration gives a particular solution containing tours for vehicles. The best solutionobtainedduringthepre‐specifiednumberofiterationsisselected.

5.NUMERICALEXAMPLE

Spreadingofmarketsforcescarriersto lookforcustomerswiderthantheirtraditionalservicearea.Usually, long‐termcontracts are settled through electronic auctions (using the Internet),whereselectedsetofcarrierssubmitabidtocoveragivensetoflanes.Theaimofcarriersistoreducetheircostsandtodecreaseseasonaleffectswhenfleetofvehiclesisnon‐optimallyused(when trucks travelwithout any load or have a partial load). Carriers have tomake effort toexpand theirmarket area and to attract new customers. Through theweb carriers can find anumberofspotloadstofilluptrucksortoavoidemptyreturns.Theydecidewhethertopickupornotaccordingtothecompatibilitywiththeremainingcapacityofatruck.Theadditionalcostforservingthecustomermustbecompensatedbyprofit.Forplanningpotentialcustomerstheentirefleetofvehiclehastobeconsidered(Archettietal.,2009).

This problem can bemodelled as a routing problemwith profitswhere a fleet of capacitatedvehicles is given as well as a set of customers that have to be served. In addition, a set of

potentialcustomersisavailable.Theproblemistodecidewhichofthesepotentialcustomerstoserveandhowtoconstructroutesforvehiclesinsuchawaythatasuitableobjectivefunctionisoptimized. In thispaper theobjective is themaximizationof thecollectedprofit,given limitedtimeavailableforvehicles.

Inordertodemonstratetheproposedmetaheuristicapproach,thefollowingnumericalexamplewassolved.Thebenchmarkproblem,21‐nodenetworkwas taken fromthe following internetaddressandmodified:www.mech.kuleuven.be/cib/op.A time limitonthetourduration is22.The capacity constraint is 25.Demands equal to 5 are associatedwith thenodes exceptnodeone,whichisthedepotandhasdemandequalto0.The21stnodeisreplacedbythecoordinatesofthefirstnodeinorderforeachvehicletostartandtoenditsrouteatthedepot.Nodesfrom6to13represent long‐termcontracts.Thereare threevehicles.Eachvehiclestartsandends itstouratdepot.Solvingvehicleroutingproblemofregularcustomersgivesthefollowingresult:

Table1.Tourswithregularcustomers

tour tourlength demand score

Vehicle1 1,7,6,13,1 7.9968 15120

Vehicle2 1,12,10,8,1 11.5067 15

Vehicle3 1,11,9,1 10.0119 10

Itcouldbeseenthatallofvehiclesarepoorlyexploitedaswellasthetimeavailableforservice.Nodesfrom2to5andfrom14to20arepotentialcustomersthatmightbeconsideredtoserve.Regularcustomers thatneed tobevisitedaremodelledwithscores10 timesbigger thanrealones.Solutionobtainedbytheproposedalgorithmisgivenintable2.

Table2.Tourscontainingbothregularandnewcustomers

tour tourlength demand score

Vehicle1 1,11,10,9,17,14,1 21.2215 25290

Vehicle2 1,13,12,8,2,4,1 20.0413 25

Vehicle3 1,7,6,5,3,20,1 20.1056 25

Itcouldbeseenthatallregularcustomersareincludedintours.Remaining,unvisitednodesare15, 16, 18, 19. Capacity of vehicles are fully used and the available time is almost completelyused, while the score is higher. It could be said that the new, better transportation plan forcarrierscouldbeobtainedbytheproposedalgorithm.

6.CONCLUSION

In this paper capacitated team orienteering problem is solved by applying the Bee colonyoptimization.Heretheproblemisconsideredwhenthecarrierhasafleetofcapacitatedvehiclesfor the set of customers but it does not completely use the time available for the service orcapacity of vehicles. Finding new customersmay bemodelledwith the CTOP. The developedalgorithm, used for solving capacitated team orienteering problem for the first time, yieldsacceptableresults.ItmightbeconsideredasasupporttoolfordecisionmakingofcarrierswhenchoosingnewcustomersfromdatabasesavailableontheInternet.

REFERENCES

[1] Archetti, C., Hertz, A., Speranza, M. G. (2007). Metaheuristics for the team orienteeringproblem.JHeuristics,13,49‐76.

[2] Archetti,C.,Feillet,D.,Hertz,A.,Speranza,M.G.(2009).Thecapacitatedteamorienteeringandprofitabletourproblems.JournaloftheOperationalResearchSociety,60,831‐842.

[3] Archetti, C., Feillet, D., Hertz, A., Speranza,M. G. (2010). The undirected capacitated arcroutingproblemwithprofits.Computers&OperationsResearch,37,1860‐1869.

[4] Butt,S.,Cavalier,T.(1994).Aheuristicforthemultipletourmaximumcollectionproblem.ComputersandOperationsResearch,21,101‐111.

[5] Chao, I., Golden, B. L., Wasil, E. A. (1996). The team orienteering problem European.JournalofOperationalResearch,88,464‐474.

[6] Drenovac, D., Jovanović, P., Nedeljković, R. (2014). Orienteering problem: Bee colonyoptimization approach. International Conference on Traffic and Transport Engineering,November27‐28,Belgrade,Serbia,103‐108

[7] Fillet, D., Dejax, P., Gendreau, M. (2005). Travelling salesman problem with profits.Transportationscience,39(2),188‐205.

[8] Gendreau,M.,Laporte,G.,Semet,F.(1998).Abranch‐and‐cutalgorithmfortheundirectedselectivetravellingsalesmanproblem.PublicationCRT‐95‐80,Centrederecherchesurlestransports,Montreal,Canada.

[9] Golden,B.L.,Assad,A.,Dahl,R.(1984).Analysisofa largescalevehicleroutingproblemwithaninventorycomponent.LargeScaleSystems,7,181‐190.

[10] Golden, B. L., Levy, L., Vohra, R. (1987). The Orienteering Problem. Naval ResearchLogistics,34,307‐345.

[11] Kim, B., Li, H., Jonson, A. (2013). An augmented large neighborhood searchmethod forsolvingtheteamorienteeringproblem.ExpertSystemswithapplications,40,3065‐3072.

[12] Laporte, G., Martello, S. (1990). The selective travelling salesman problem. DiscreteAppliedMathematics,26,193–207.

[13] Lučić, P., Teodorović, D. (2001). Bee system: modeling combinatorial optimizationtransportationengineeringproblemsbyswarmintelligence.In:PreprintsoftheTRISTANIVTriennialSymposiumonTransportationAnalysis,SaoMiguel,AzoresIslands,Portugal,441‐445.

[14] Lučić,P.,Teodorović,D.(2003).ComputingwithBees:AttackingcomplexTransportationEngineering Problems. International Journal on Artificial Intelligence Tools, 12(3), 375‐394.

[15] Souffriau,W.,Vansteenwegen,P.,Vertommen, J.,VandenBerghe,G.,VanOudheusden,D.(2008). A personalised tourist trip design algorithm for mobile tourist guides. AppliedArtificialIntelligence,22(10),964‐985.

[16] Teodorović, D. (2008). Swarm intelligence systems for transportation engineering:Principlesandapplications.TransportationResearchPartC,16(6),651‐667.

[17] Tsiligirides, T. (1984). Heuristic methods applied to orienteering. Journal of theOperationalResearchSociety,35(9),797‐809.

[18] Wang,X.,Golden,B.,Wasil,E. (2008).Usingageneticalgorithmtosolve thegeneralizedorienteering problem. In: B. Golden, S. Raghavan, E. Wasil (Eds.), The Vehicle RoutingProblem:LatestAdvancesandNewChallenges,NewYork,Springer,263‐274.

THEIMPACTOFTHELOWCOSTCARRIERPRESENCEONTHEDATAENVELOPMENTANALYSISEFFICIENCYOFAIRPORTS

KlemenGrobina,TomažKrambergera

aUniversityofMaribor,FacultyofLogistics,Slovenia

Abstract:Thepresentpaperaimstofilltheapparentgapinexistingliteratureandproposepropermethodsfordetermininganimpactthepresenceoflowcostcarriershasonairportefficiency.Theimpactofthelowcostcarrierpresenceontheefficiencyofchosenairportswascalculatedbysimplelinearregression.Thenumberoflowcostcarriersoperatingfromcertainairportswasconsideredand independent variable while the calculated data envelopment analysis efficiencies wereconsideredasdependentvariables.Theresultsclearlyindicatethatthereisapositivecorrelationbetweenthenumberoflowcostcarriersandtheairportsefficiency.

Keywords:Airportefficiency,DataEnvelopmentAnalysis,LowCostCarriers

1.INTRODUCTION

Low Cost Carriers (LCC) in Europe first appeared in the late nineties with the help ofderegulationsofairspacethatwereputinplacethroughtheEuropeanUnion.Theyloweredtheservicesprovidedtothecustomersandstartedsellingtheticketsovertheinternet.Despiteallthis theLCCtrendmanagedtosucceedandstill seesrapidgrowth in the lastdecade(Francis,2005).ModernLCCsenablethecustomertotravelcheaplywithlowfarestoaneverincreasingnumber of destinations. The growth can be best seen in numbers of short haul passengerscarried in Europe. In 2003 LCCswere responsible for 10% of short haul passenger traffic inEurope (Francis, 2003). By year 2013 this share has increased to 26% according to trafficanalysismadebyEurocontrol(EPRS,2014).ThisnumberfallsshortofthepredictionsthatLCCswould control33%of themarket sharebyyear2010but theyarenone the less showing thetrendtowardsthegrowthofLCCpassengertraffic.

WhiletherewerenumerousstudiesconductedonLCCimpactstowardscompetition,farepricesandtrafficnumbersasclaimedbyGraham(2013),thesubjectofLCCeffectsonairportefficiencyandairportperformanceremainfewandfarbetween.MostofexistingworkmaterialisfurtherlimitedtomajorEuropeantouristareasandbiggerwellknownairports.

Aimof thisresearchpaper is to fill theapparentgap inexisting literatureandproposepropermethodsfordetermininganimpactthepresenceofLCCshasonairportefficiencyintheAdriaticregion.Theairportswehavechosenforthissamplestudyareconsideredsmallbuttheregionitselfisverydependentontheirexistence.Oneofthereasonsforthedependenceisthefactthatthe region set its future goals in developing tourism, therefore efficiency and development ofsaidairportsisoneofthekeyfactorsinsuccessoftheregioninthefuture.

2.LITERATUREREVIEW

Airtravelersatisfactionhasbeennotedtoconstantlydropinthepastdecade.Inordertocombatthis problem a constant effort has been made in measuring and comparison of airportperformancesofcompetingairportswhileatthesametimeanothertrendofAirportcongestiongrowing has been noted. Operational efficiency of airports is therefore becoming one of theimportantdeterminantsofthesystem’ssuccessinthefuture(Schaar&Sherry,2008).

Pyrialakouandcoauthorsproposedtoassess theoperationalefficiencyofairportswherehighlevels of low‐cost carrier traffic can be seen using a nonparametric method called DataEnvelopmentAnalysis(DEA)(Pyrialakou,Karlaftis,&Michaelides,2012).Usingthismethodtwomodelshavebeendeveloped.Thefirstmodelisusedinordertoassesstheterminalservicesandtheotherisusedinordertoassesstheairsideoperations.Datainputsforbothmethodsconsistof the number of gateways used, number of runways used and the number of aircraftmovementson theairport,whilenumberofenplanedpassengerswasusedasdata foroutputpart of the model. The results of this model show high correlation between enplanements,terminal efficiencies and hours of operation of chosen airports. Impacts of LCC services onefficiency of major U.S: airports have been addressed in the study done by Choo and Oum(2013).ThestudywasconductedusingtheIndexNumberApproachastheprimarymethodfordetermining the efficiency of chosen airports but in addition, the SFA approach was used toensure the robustness of obtained results. In contrast of the previouslymentioned study theresearchdetermined,thattheLCCpresenceonmajorU.S.airportshasshownanegativeeffectonoperatingefficiencyofthechosenairports.

SchaarandSherryhaveexposedthedifferencesinresultsusingvariousdifferentDEAmethods(Cooper‐Charnes‐Rhodes (CCR), Banker‐Charnes‐Cooper (BCC), and Slacks‐Based Measure ofefficiency (SBM)) (Schaar & Sherry, 2008). Results were obtained using the data from 45airportsinthetimeframebetween1996and2000.Furtheranalysishasshownawidevariationinresults,whichpromptedthediscussionoftheneedforguidelinesforselectionofproperDEAmodelsandproper interpretationofDEAresults inthepaper.ResearchconductedbyYoshidaandFujimoto focusedon testing the criticismof overinvestment in Japanese regional airports(Yoshida & Fujimoto, 2004). Researchers employed two distinctly different methods, DataEnvelopmentAnalysisandendogenous‐weightTotalFactorProductivitymethod(TFP).ResultsofbothmethodsindicatedthattheregionalairportefficiencyofJapanairportsisnotashighasthe others. At the same time the research exposed an interesting fact that the airportsconstructedinthe1990swererelativelyinefficientcomparedtoairportsbuiltindifferentpointsintime.

3.METHODOLOGY

In previous section we established the most commonly used methods for measuring airportefficiencyandproductivity.MethodsinquestionareIndexNumberMethod,DataEnvelopmentAnalysis (DEA) and Stochastic Frontier Analysis (SFA). Firstmethod (INM) is used for directmeasurement of productivity as output index over the input index. Theproductivity factor ofairport can be determined as the ratio of aggregate output and input indexes. The secondmethod in question is DEA. Its deterministic approach is based on the assumption that allobserveddeviationsaretheresultofinefficiencies.ThelastmethodnotedisSFA.Themethodisusingastochasticapproachtomodellingwhichenablesthemodeltoaccountforbothdeviationscaused by inefficiencies and deviations caused by measurements errors, random errors andstatisticaldiscrepancies.

Themethodchosen for theefficiencycalculations in thispaper is theDEAmethod.Oneof themainadvantagesofthechosenmethodisthefactthatitdoesnotrequiremoredatathaninputandoutputquantities.Theefficiencyiscalculatedrelativetothehighestobservedperformance

andnottheaverageobservedperformance(Kamil,Baten,&Mustafa,2012).AccordingtoSchaar&Sherry(2008)thechoiceoftheappropriateDEAmodeliskeyforthestudy.WehavedecidedontheusageoftheoutputorientedConstantReturnofScalemodelwhichisnotusuallysuitableforusage in caseswhere theDecisionMakingUnit (DMU)doesnot reachoptimumoperation.Wesupport thisdecisiondue to the fact that theusageof competingVariableReturnofScalemodel the largest and the smallest airport in the sample would become vastly overstated interms of efficiency.We calculated the impact of the LCC presence on the efficiency of chosenairports using the method of simple linear regression. The number of LCCs operating fromcertainairportswasconsideredandindependentvariablewhilethecalculatedDEAefficiencieswereconsideredasdependentvariables.

4.DATA

The data sample for this paper was chosen from the airports operating in the Adriatic Searegion. Not all airports are located in the coastal region but all of them are commerciallyconnectedthecoast.Forpurposesof thisresearchpaper fiveof theairportswerechosen.TheselectioncanbeseenontheFigure1.

Figure1:SelectedairportsinAdriaticregion

Criteria for airport selection demand that the selected airportmust provide scheduled flightswithatleastoneairlineandprocessattheveryleast15000passengersinayear.Welimitedthedataselectioninthetimeframebetweenyear2006andyear2013.Ourprimarysourcesofdataconsist of Air Traffic Reports provided by ACI (2014). Other sources of data are Airport‐data.com (2014)web page and data collected directly from the airport authorities. Using theprocureddatawebuiltthreedistinctOutputOrientedDEAmodelsbasedonCRSalgorithm.Themodel considers the extent inwhichoutputs couldbe increasedwhileusing the same inputs,relatively to the standards set by the competing units. The inputs and outputs used arepresentedintheTable1.

Table1:Inputsandoutputs

Inputs OutputsFlightEfficiencyModel(outputorientedCRS)Numberofrunways AircraftmovementsLengthofthelongestrunwayNumberofdestinationsPassengerEfficiencyModel(outputorientedCRS)Numberofoperatingcarriers PassengerthroughputCarparkcapacity

DistancetocitycenterNumberofpassengerterminalsNumberofgatesOverallEfficiencyModel(outputorientedCRS)Numberofoperatingcarriers PassengerthroughputCarparkcapacity AircraftmovementsDistancetocitycenterNumberofpassengerterminalsNumberofgatesNumberofrunwaysLengthofthelongestrunwayNumberofdestinations

ThesametypeofanalysiscouldalsobeusedtoanalyzetheefficiencyofairportswhilehandlingcargoaswecanseefromthealreadyestablishedanalysismeasuringthetechnicalefficiencyofairportsinLatinAmerica.Thestudyusedsimilardatasetcomparedtoourresearchpaperwiththeadded freightmovements asanoutput categoryaswell (PerlemanandSerebrisky,2012).DuetothefactthatLowCostCarriersfocusismovingpassengerswedecidedtoeliminatethisoutputvariablefromourstudy.AircompaniesconsideredasLCCandinvolvedinourstudyarethefollowing:AirBerlin,AirOne,Blueair,Blueexpress,BluePanoramaairlines,EasyJet,Flybe,Germanwings, HOP!, Intersky, jet2.com,Monarch airlines, Norweigian airline, Pegasus airline,Ryanair, Smart wings, Transavia airline, Volotea, Vueling, Wizz Air, Sky Europe, Eurolot andEdelweiss.ThenumberofairlinesatdistinctairportsovertheyearsispresentedintheTable2.

Table2:ThenumberofLCCatdistinctairports.

2006 2007 2008 2009 2010 2011 2012 2013Ljubljana 2 2 2 2 1 2 2 2Pula 1 2 3 3 3 3 4 4Zadar 0 2 2 2 2 2 2 2Split 2 2 3 4 5 5 9 9Dubrovnik 1 3 4 7 8 8 9 9

5.RESULTS

The DEA efficiency numbers were obtained with the help of LIMDEP 10 software using theFRONTIERfunctionwithALG=DEAandCRSlimiters.

5.1DEAEfficiency

IntheTable3.wecanseethecalculatedFlightEfficienciesforselectedairportsovereightyears,from2006to2013.

Table3:FlightEfficiency

2006 2007 2008 2009 2010 2011 2012 2013Ljubljana 1 1 1 1 1 1 1 1Pula 0,26398 0,266678 0,287911 0,294223 0,23532 0,26086 0,298703 0,32706Zadar 0,077127 0,078691 0,085384 0,094276 0,103199 0,114264 0,149574 0,163177Split 0,546255 0,482707 0,468695 0,488964 0,533392 0,57406 0,643143 0,693237Dubrovnik 0,362408 0,323483 0,309278 0,315274 0,365042 0,408753 0,463077 0,492216

WecanseethatSlovenianairportisrelativelymuchmoreFlightEfficientthanCroatianones.

ThecalculatedPassengerEfficiencyisshownintheTable4.

Table4:PassengerEfficiency

2006 2007 2008 2009 2010 2011 2012 2013Ljubljana 1 1 1 1 1 1 1 1Pula 0,68734 0,872912 0,860893 0,74044 0,682456 0,69237 0,664775 0,617474Zadar 0,168036 0,267714 0,343728 0,529231 0,657782 0,664192 0,875956 1Split 1 1 1 1 1 1 1 1Dubrovnik 1 1 1 1 1 1 1 1

Here, efficiency is not geographically dependent, but yet, some airports are still much moreeffectivethantheothers.

The Overall Efficiency is calculated with respect to two outputs (passenger throughput andaircraftmovements).TheresultsarepresentedintheTable3.

Table3:OverallEfficiency

2006 2007 2008 2009 2010 2011 2012 2013Ljubljana 1 1 1 1 1 1 1 1Pula 1 1 1 1 1 1 1 1Zadar 0,259897 0,275256 0,343728 0,529231 0,657782 0,664192 0,875956 1Split 1 1 1 1 1 1 1 1Dubrovnik 1 1 1 1 1 1 1 1

5.2TheimpactofLCCpresenceonEfficiency

The impact of LCCpresence on efficiency of airports is estimatedby simple linear regressionusingLIMDEP.ThenumberofLCCflightsfromcertainairportsovertheyearswasconsideredasan independent variablewhile the calculateddefinedDEAefficiencieswere considered as thedependent variable. The correlation coefficients, explaining the impact of LCCs, are listed inTable4.

Table4:Calculatedcorrelationcoefficients

FlightEfficiencyCoefficient

PassengerEfficiencyCoefficient

OverallEfficiencyCoefficient

Ljubljana 0 0 0Pula 0,56186 ‐0,37052 0Zadar 0,3877 0,54531 0,46351Split 0,88393 0 0Dubrovnik 0,6454 0 0

Theresultsshowninthelasttableclearly indicatethatthereisapositivecorrelationbetweenthe number of low cost carriers and the airports efficiency. The negative number in Pulasegmentwascausedbyarapiddeclineofnumberofthepassengersduetoeconomiccrisis.Ininstanceswhereoptimalefficiencyhasbeencalculatedthoroughthewholetimeofobservationthe correlation shown is 0 due to the way the correlation coefficient is calculated. Furtherresearch featuringmore airports andmoredata is required to further explain the correlationbetweenLCCsandairportefficiency.Afterfurtheranalyzing,theimpacttheLCCcouldhaveonfreightcargowediscoveredthatmostLCCsbased inEuropedonot thinkpursuingthe freightmarket is a good idea and they tend to stay on the familiar territory of passenger transport.AirlinessuchasNorwegianclaimthattheywillnotcompetewithestablishedcarriersinpricesdespitetryingtosucceedinfreighttransportaswell(Lennane,2013).Assuch,wecanconcludethat current day LCCs do not pose a significant factor in changing the efficiency with theinclusionof freight flows.Thishowevermight change in the futureasmoreaggressive tacticscouldbeusedbytheLCCs.

REFERENCES

[1] ACI (2014). Airports Council International. Retrieved December 2, 2014, from http://www.aci.aero

[2] Airport‐Data.com (2014). One stop Aviation Info. Retrieved December 2, 2014, fromhttp://www.airport‐data.com

[3] Choo,Y.,Oum,T.(2013).ImpactsoflowcostcarrierservicesonefficiencyofthemajorU.S.airports.JournalofAirTransportManagement(33),60‐67.

[4] EPRS(2014).Low‐costaircarriersinEurope.RetrievedDecember1,2014,fromEuropeanParliamentary Research Service: http://epthinktank.eu/2014/06/26/low‐cost‐carriers‐in‐europe/

[5] Francis, G. (2003). Airport‐airline interaction: the impact of low‐cost carriers on twoEuropeanairports.JournalofAirTransportManagement(9),267‐273.

[6] Francis,G.(2005).Wherenextfor lowcostairlines?Aspatialandtemporalcomparativestudy.JournalofTransportGeography(2005),83‐94.

[7] Graham,A.(2013).Understandingthelowcostcarrierandairportrelationship:Acriticalanalysisofthesalientissues.TourismManagement,36,66–76.

[8] Kamil, A., Baten, M., & Mustafa, A. (2012). Stochastic Frontier Approach and DataEnvelopment Analysis to Total Factor Productivity and Efficiency Measurement ofBangladeshiRice.PLoSONE,10(7).

[9] Lennane, A. (2013) Low‐cost air cargo entrant vows to hold the line on freight rates.Retrieved December 3, 2014 from http://theloadstar.co.uk/low‐cpst‐air‐cargo‐entrant‐vows‐to‐hold‐the‐line‐on‐freight‐rates/

[10] Perelman,S.,Serebrisky,T.(2012).MeasuringthetechnicalefficiencyofairportsinLatinAmerica.UtilitiesPolicy,1‐7.

[11] Pyrialakou,V.,Karlaftis,M.,&Michaelides,P. (2012).Assessingoperational efficiencyofairportswithhighlevelsoflow‐costcarriertraffic.JournalofAirTransportManagement,25,33‐36.

[12] Schaar,D.,&Sherry,L.(2008).ComparisonofDataEnvelopmentAnalysisMethodsUsedinAirportBenchmarking.RetrievedDecember3,2014, fromCenter forAirTransportationSystemsResearch:http://catsr.vse.gmu.edu/pubs/Schaar_Sherry_ICRAT.pdf

[13] Yoshida, Y., & Fujimoto, H. (2004). Japanese‐airport benchmarking with the DEA andendogenous‐weight TFP methods: testing the criticism of overinvestment in Japaneseregionalairports.TransportationResearchPartE(40),533–546.

THEVEHICLEREROUTINGPROBLEMWITHTIMEWINDOWSANDSPLITDELIVERY

MilošNikolića,DušanTeodorovića,*

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Manylogisticscompanyusethesamesetofvehiclerouteseveryday.Unexpectedlyhighcustomers’demandinoneormorenodescancauseadisruptioninvehicleoperations.Disruptionoccurs since, for one or more planned routes, capacity constraints can be broken. In such asituation,dispatchersmustdecidehowtoorganizeadistributioninthewaytominimizenegativeeffects of disruptions. The mathematical programming formulation for the vehicle reroutingproblem in thecaseof splitdelivery isproposed in thepaper.Performednumericalexperimentsshow that split delivery concept of distribution, in the case of disruptions in vehicle operations,outperformdistributionapproacheswithoutsplitdelivery.

Keywords:Vehicleroutingproblem,disruptions,splitdelivery

1.INTRODUCTION

Many logistics companies, on a daily basis, deliver various goods to the customers. In manycases,companiesdelivergoodstothesamesetofthecustomerseveryday.Insuchasituation,dispatchers usually decide to use the same set of vehicle routes every day. The benefits ofperforming the same vehicle routes on a daily basis are obvious. During the time, driversbecomefamiliarwiththerouteswheretheymakedeliveries. It isalsoeasier forcompaniestoplan fleetsizeand fleetusage.Simultaneously, thecostscanbeplanned fora longerperiodoftime.

Suddenchangesincustomerdemandscancausesituationswheretheplannedvehicleroutesarenot the best possible. Significantly increased customer demands canmake that some vehiclecapacityconstraintsarebroken.Thedisruptionsinvehicleoperationscanproducethefollowingnegative effects: there is a need to engage additional vehicles; the delivery cost could beincreased;somecustomerscannotbeserved,etc.

We consider in this paper the case when split delivery is allowed in the situation whendisruptionoccurs.Wepropose,forthiscase,themix‐integerlinearprogrammingmathematicalformulation. We clearly show that the split delivery can significantly decrease negativeconsequencesofdisruptions.

Thepaperisorganizedinthefollowingway.TheliteraturereviewandstatementoftheproblemaregivenintheSection2.MathematicalformulationoftheproblemisdescribedintheSection3.NumericalexperimentsarepresentedintheSection4.Finally,theconclusionsanddirectionsforfutureresearcharegivenintheSection5. * duteodor@gmail.com

2.LITERATUREREVIEWANDSTATEMENTOFTHEPROBLEM

The area of disruptionmanagement has been studied since 1984 (Teodorović and Guberinić,1984). In distribution systems different disruptions can occur, such as: vehicle breakdown,changes in customerdemand, traffic jam, changes indelivery locations, etc. There are severalpapersintheliteraturedevotedtothevehiclereroutingproblems.Lietal.(2009a,b)consideredthe vehicle routing problem which arise when one or more vehicles have breakdown. Theauthors proposed insertion heuristic based on Lagrangean relaxation. Mu et al. (2011)consideredsimilarproblemandproposedmodelbasedonTabusearchmetaheuristic.Wangetal.(2011,2012)proposedmathematicalformulationaswellas,heuristicalgorithmsforthecasewhenfewdisruptioneventsoccur.HuandSun(2012)proposedknowledge‐basedapproachtomitigatedisruptionsduringdistributionprocess,whileHuet al. (2013)proposed local searchalgorithmsandobject‐orientedmodeling.MuandEglese(2013)analyzedtheproblemwhenthedeliverytocentralwarehouseisindelay.TheauthorsproposedthemodelbasedonTabusearchmetaheuristic.

Weconsiderthecasewhencompaniesusethesamesetofvehiclerouteseveryday.Theproblemariseswhenthecustomerssignificantlyincreasedemandandconsequently,somevehicleroutesbecomeinfeasible.ThisproblemwasstudiedbySplietetal.(2014)andNikolićandTeodorović(2015). Spliet et al. (2014) based theirmodel on the capacity vehicle routing problem. Theyproposedthemathematicalformulation,aswellasheuristicalgorithm.NikolicandTeodorović(2015)studiedthevehiclereroutingprobleminthecaseoftimewindowsandproposedmulti‐objectivemathematicalformulation.

Inthispaper,wefurtherexpandtheresearchperformedbyNikolićandTeodorović(2015).Weconsiderthecasewhensplitdeliveriesareallowed.Inotherwords,weconsiderthecasewhennodescouldbeservedbymorethanonevehicle.

3.MATHEMATICALPROGRAMMINGFORMULATION

Letusdenoteby ANG , anorientedgraph,whereVisasetofnodes 1,,,1,0 nnV ,and

Aisasetofedges Aji , .Wedenoteby0andn+1thecentraldepot.

Wealsointroducethefollowingnotation:

Binarydecisionvariables:

otherwise.0,

servedbenotwillnodeif,1 ii

otherwise.0,

route original itsin not is node hisalthough t , node odelivery t make will vehicle theif,1 ikzik

otherwise.,0

nodetonodefromtravelsvehicleif,1 jikykij

Theotherdecisionvariablesare:

kiw ‐timewhenstartserviceatnodeibyvehiclek

fik–thefractionofdemandofnodeideliveredbyvehiclek

Thefollowingaretheinputvalues:

kijx ‐isequal1ifintheinitialsetofroutesvehiclek,afterservicingnodei,goestonodej

wi‐importance(weight)toservethenodei

ui–importance(weight)thatnodeiwillnotbeservicedintheroutethatisnottheoriginalrouteofthenodei

cij–thecostfortravelfromnodeitonodej

qi–demandatnodei

Q–vehiclecapacity

tij–traveltimefromnodeitonodej

si–durationofserviceatthenodei

M–largepositivenumber

Mathematicalformulationofthevehiclereroutingprobleminthecaseofsplitdelivery,basedonNikolićandTeodorović(2015)andHoandHaugland(2004),canbegiveninthefollowingway:

Ni

iiwF 1min (1)

Kk Ni

iki zuF2min (2)

Kk Aji

kijij ycF

,3min (3)

subject to:

NiyKk ij

ikij

1)(

(4)

Kkyj

kj

1)0(0

(5)

NjKkyyji

kji

ji

kij

,0)()(

(6)

Kkyni

kni

1)1(

1,

(7)

AjiKkyMtsww k

ijijiki

kj ,,1 (8)

ViKkbwa ikii , (9)

KkQfg

Niiki

(10)

Nif

Kkik

1 (11)

KkNify ikij

kij

,)(

(12)

NjKkzxy jjkji

kij

kij

,)(

(13)

Nii 1,0 (14)

NiKkzik ,1,0 (15)

AjiKkykij ,,1,0 (16)

Theobjectivefunction(1)thatshouldbeminimizedrepresentsthetotalnumberofunsatisfiedcustomers.Thetotalnumberofcustomersthatwillbeservedintherouteswhicharenottheiroriginal routes is calculated by objective function (2). This objective function should beminimized.Theobjectivefunction(3)thatshouldbeminimizedrepresentsthetotaltravelcosts.

Constraints (4)guarantee that ifnode iwillnotbe served, thedecisionvariable i must takevalue1.Allvehiclesleavethedepot,visitsomenodesandreturntothedepot.Thisisguaranteedbytheconstraints(5),(6)and(7).Timewindowsmustbesatisfiedaccordingtotheconstraints(8)and(9).Constraints(10)and(11)guaranteethatthevehiclecapacitywillnotbeexceededbythetotaldemandthatwillbedeliveredthroughtheroute.Becauseofconstraint(12)ifnodeiwillnotbeservedbythevehiclek , thedecisionvariable fikmusttakevalue0. If thevehiclekmakesthedeliverytothenode j,andnodej isnotincludedintheinitialrouteofthevehiclek,thenvariablezjkwill takethevalue1accordingtoconstraints(13).Constraints(14), (15)and(16)definedecisionvariablesasbinary.

4.NUMERICALEXAMPLES

Toevaluate theeffectsofallowedsplitdelivery,wehaveusedbenchmarkexamples,basedonC101 Solomon benchmark example, given in (Nikolić and Teodorović, 2015). Nikolić andTeodorović, 2015 used lexicographic method to solve the problem instances. Their results,relatedtothecasewhenthesplitdeliveryisnotallowed,aregiveninTable1.Theresultsaregivenforthecaseswhenthenumberofcustomersare25and50respectively.AllinstanceshavebeensolvedbytheCPLEXsoftware.

Table1.Theresultsobtainedwhensplitdeliveryisnotallowed(NikolićandTeodorović,2015)

Thenumberofnodes

Increasein

demand

C101Fleetsize:3 Fleetsize:4

F1 F2 F3 F1 F2 F3

25nodes

10% 0 1.12 249.00 0 1.12 249.00

20% 0 1.15 239.33 0 1.13 251.92

30% 1.02 1.15 239.27 0 2.16 271.29

40% 1.30 2.22 257.76 0 3.27 272.84

50% 2.34 1.12 243.47 0 3.29 273.30

60% 3.60 1.09 220.37 0 4.31 265.00

70% 4.78 0 185.81 1.04 3.37 271.48

80% 4.78 0 185.81 1.04 3.37 271.48

90% 6.04 2.21 192.02 2.30 5.55 284.41

100% 6.04 2.21 192.02 2.30 5.55 284.41

50nodes

10% 0 2.43 480.79 0 2.43 480.79

20% 0 3.68 469.92 0 3.66 488.66

30% 0 6.01 496.3 0 6.00 505.1

40% 1.04 7.23 563.66 0 7.16 556

50% 2.62 7.5 492.68 0 8.59 527.25

60% 4.92 8.99 610.38 1.04 11.16 552.46

70% 7.54 6.63 516.52 3.46 9.04 561.14

80% 7.54 6.63 516.52 3.46 9.04 561.14

90% 10.34 9.16 613.72 5.78 12.18 584.09

100% 10.34 9.16 613.72 5.78 12.18 584.09

InthispaperwehavesolvedallinstancesconsideredbyNikolićandTeodorović(2015).WealsouseLexicographicmethodandtheCPLEXsoftware.TheobtainedresultsaregiveninTable2.Inthe case when there are one central depot and 25 customers, when the fleet size is equal 3vehicles,forthesixinstancesweobtainedthebetterresultsthaninthepreviousresearch.Inthecasewhen the fleet size is equal 4 vehicles, we obtained better results in two cases. Similarresultsareobtainedforthecaseswherethereareonecentraldepotand50customers.Inthesecases,whenthefleetsizeisequal6vehicles,weobtainedbetterresultsforsixinstances,whileinthecaseswherethefleetsizeisequal7vehicles,threetimeswediscoveredthebettersolution.ThebetterresultsaredenotedbyboldanditalicletterinTable2.

Table2.Theresultsobtainedinthecasewhensplitdeliveryisallowed

Thenumberofnodes

Increasein

demand

C101Fleetsize:3 Fleetsize:4

F1 F2 F3 F1 F2 F3

25nodes

10% 0 1.12 249.00 0 1.12 249.00

20% 0 1.15 239.33 0 1.13 251.92

30% 0 2.21 257.28 0 2.16 271.29

40% 1.3 2.18 256.60 0 3.27 272.84

50% 2.3 3.37 253.52 0 3.29 273.30

60% 3.46 2.38 288.07 0 4.31 265.00

70% 3.66 2.07 229.53 0 4.39 278.66

80% 4.78 0 185.81 1.04 3.37 271.48

90% 5.16 3.2 258.49 1.3 6.55 304.73

100% 6.04 2.21 192.02 2.3 5.55 284.41

50nodes

10% 0 2.43 480.79 0 2.43 480.79

20% 0 3.68 469.92 0 3.66 488.66

30% 0 6.01 496.30 0 6.00 505.10

40% 1.02 7.21 531.07 0 7.16 556.00

50% 2.30 7.59 567.93 0 8.59 527.25

60% 3.66 8.96 600.50 0 12.14 590.62

70% 5.78 8.36 514.64 1.30 11.12 540.34

80% 7.54 6.63 516.52 3.46 9.04 561.14

90% 9.08 9.13 669.01 4.78 12.68 559.52

100% 10.34 8.01 615.03 5.78 12.18 584.09

5.CONCLUSIONS

Manylogisticscompaniesusethesamesetofvehiclerouteseverydayfordeliverygoodstothecustomers.Theseroutessometimescanbecomeunfeasibleinthecasesofveryhighcustomersdemand.

Whensolvingvehiclereroutingproblem,weproposedandanalyzedthesplitdeliveryconcept.Ithas been shown, that by allowing the split deliveries the negative consequences of thedisturbancescanbesignificantlyreduced.

Wewillexplorevariousheuristicapproachesinthefutureresearchthatwillenableustoattackthevehiclereroutingproblemsofbigdimensions.

ACKNOWLEDGMENT

This research was partially supported by Ministry of Education, Science and TechnologicalDevelopmentRepublicofSerbia,throughtheprojectTR36002fortheperiod2011–2015.

REFERENCES

[1] Ho,S.C.,Haugland,D.,2004.Atabusearchheuristicforthevehicleroutingproblemwithtimewindowsandsplitdeliveries,Computers&OperationsResearch,31,1947‐1964.

[2] Hu, X., Sun, L. 2012. Knowledge‐based modeling for disruption management in urbandistribution,ExpertSystemwithApplications,39,1,906‐916.

[3] LiJQ,Mirchandani,P.B.andBorenstein,D.,2009a.ALagrangianheuristicforthereal‐timevehicle rescheduling problem. Transportation Research Part E: Logistics andTransportationReview,45,419–433.

[4] Li,J.Q.,Mirchandani,P.B.andBorenstein,D.,2009b.Real‐timevehiclereroutingproblemswithtimewindows.EuropeanJournalofOperationalResearch,194,711–727.

[5] Mu,Q.,Fu,Z.,Lysgaard,J.,Eglese,R.,2011.Disruptionmanagementofthevehicleroutingproblemwith vehicle breakdown, Journal of the Operational Research Society, 62, 742‐749.

[6] Mu, Q., Eglese, R.W. 2013. Disrupted capacitated vehicle routing problem with orderreleasedelay,AnnalsofOperationsResearch,207,1,201‐2016.

[7] Nikolić, M., Teodorović, D., 2015. Vehicle rerouting in the case of unexpectedly highdemandindistributionsystems,TransportationResearchPartC,acceptedforpublishing

[8] Spliet, R., Gabor,A.F.,Dekker,R., 2014. The vehicle reschedulingproblem, Computers&OperationsResearch,43,129‐136.

[9] Teodorović, D., Guberinić, S., 1984. Optimal dispatching strategy on an airline networkafterascheduleperturbation.EuropeanJournalofOperationalResearch,15,178–182.

[10] Wang,X.,Ruan,J.,Shang,H.,Ma,C.2011.ACombinatorialDisruptionRecoveryModelforVehicleRoutingProblemwithTimeWindows,IntelligentDecisionTechnologies,10,3‐11.

[11] Wang,X.,Ruan,J.,Shi,Y.2012.Arecoverymodelforcombinatorialdisruptionsinlogisticsdelivery: Considering the real‐world participators, International Journal of ProductionEconomics,140,508‐520.

PartII INTERMODAL

TRANSPORTATION

CHALLENGESINTHERAILWAYYARDSLAYOUTDESIGNINGREGARDINGTHEIMPLEMENTATIONOFINTERMODAL

TECHNOLOGIES

IvanBeloševića*,MilošIvića,MilanaKosijera,NorbertPavlovića,SlavišaAćimovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: This paper provides a survey on railway freight yard classification and presents thechallengesindesigningandplanningofrailwayyardsregardingtheongoingtransformationandchanges. Transport service in railways can be achieved through the conventional concept ofwagonload transportationorwithin intermodal transportchain. Inbothconcepts,railwayyardsarerecognizedaskeyfactorsforthesmoothfunctioningofthefreighttransportation.Atthesametime,yardsarefacingproblemsrelatingtolayoutdesigningandoperationplanning.Therailwayyardunder thestudy,Vršacrailwayyard,hasastrategicposition in the transportationnetworkand for thatreason isrecognizedasapotential location forestablishing intermodalservice.Thepaperproposesthe layoutsdesigned forthereconstructionoftheexistingclassificationyardandtheconstructionofanewrail‐roadtransshipmentyard.

Keywords:railwayfreightyards,layoutdesigning,intermodaltransportationtechnology

1.INTRODUCTION

The accomplished transport service in railways can be achieved through the conventionalconcept of wagonload transportation or within intermodal transport chain. In both conceptsrailwayyardsplayadecisiveroleforimprovingtheefficiencyandqualityoftransportservice.Atthe same time, yards have been faced with a whole array of problems relating to layoutdesigningandoperationplanning.

Wagonload transportation, also called Single Wagon Load Service (SWL), consolidate loadscomposedofsinglewagonsandwagongroups.Thesewagonloadsarelessthanfulltrainloadswhich are collected at different customer sidings and assembled in yards to full trainson thesameroutesthroughouttherailwaynetwork.WagonloadtransportationisbasedonthesystemofmarshallingyardswhicharecomparablewithhubsinthelogisticHubandSpokeSystem.Inthisway,marshallingyardsobtainacertaindegreeofeconomiesofscalebenefitsandreducethewastages of the system. On the other hand, the overall operating efficiency is limited by thecapacityoftheyardsandtheirutilization.

Theintensificationofrailwayyardefficiencycanbeachievedwithinintermodal’’doortodoor’’principle application. The essence of the technology of combined rail‐road transport is in thecombination of the advantages of the fast, cheap and safe railway transport of freight in long * i.belosevic@sf.bg.ac.rs

distancesandfastandefficient finaldistributionbytrucks.This,aswell,enablesasustainablemodalsharewithintransportchains.Thesustainablemodalshareassumesthecontributiontothewhole transport system by shifting from one to another,more favorable transportmode.Fromtheaspectof sustainablemodal share, railwaysadvantagesarehighcapacities reserves,reliabilityandsafety,reductioninspecificenergyconsumptionandpollution.

Railwayyardmanagementtriestoreduceredundancycostsandtostrengthentheroleofyardsinthetransportchainbyimplementingtheintermodaltransporttechnologies.Onlyinthatway,railways can increase its market share and reach the targets set by the White paper in theEuropeantransportarea.Direct intermodaltrainshavethepotentialtobethebasisofanovelrailwayfreighttransportservice.Theydirectlyconnectintermodalterminalswhileloadingandunloading costs are allocated outside the railway service.Direct trains connect two terminalswithout intermediate stops in themost economical and fastest way. In addition to transporttechnologydevelopment,itisimportanttoestablishamoreefficientfreighthandlingandtrainoperatinginrailwayyards.

Thispaperpresents the challenges indesigningandplanningofVršac railwayyard regardingtheongoing transformation and changes. The railway yardhas a strategicposition in Serbianrailwaynetworkbutduetoyearsoflackofmaintenanceitisinaninadequatestatetoperformthe service relating consolidation of wagonloads and to serve as interface in the novelintermodaltransporttechnologies.Thepaperproposeslayoutsdesignedforthereconstructionoftheexistingclassificationyardandconstructionofthenewrail‐roadtransshipmentyard.

2.RAILWAYFREIGHTYARDS:OVERVIEWANDCLASSIFICATION

Yards as nodes in networks have the key role for the smooth functioning of the freighttransportation.Railwayyardsshouldbemoreusedastransshipmentpointsinsidetherailwaysystemorbetween the railwaysandanalternativemodeof transportation.Currently, railwayyards mainly consolidate the movements of freight trains on the network. Railway yardsconcentrate resources, such as track sidings and various track installations, shuntinglocomotives and freight handling equipments in order to receive, operate and depart freighttrainsforthesakeofprovidedservice.

The most of railway freight yards execute classification procedures: disassembling andreassembling trains. It can generally be acknowledged that the most of classification freightyards are inefficient and that their layouts and disposition in networks does not satisfycontemporary needs. The most of the unproductive time railcars spend in the yards, approximately about two-thirds of their turnover system time. Petersen(1977)identifiedthreetypesofclassificationfreightyards:flatyards,humpyardsandgravityyards.Flatyardsaretheyardswheretherailcarsarepushedandpulledbyashuntinglocomotivewhichsortsthemintotheassignedtrack.Thesetracks lead intoa flatshuntingneckatoneorbothsidesof theyard(for a detaileddescription of flat shunting yard operation and layout seeMarinov andViegas(2009)).Humpyardsare the largestandmostcomplexrailwayyardswhich featurethehumpoverwhichtherailcarsarepushedbytheshuntinglocomotive.Earlyanalyticalqueuingmodels(see e.g. Crane et al. (1955), Petersen (1977)) have been generally used in analyzing andevaluatingthecomplexyardoperations.Duetothedynamicandstochasticyardoperation,theapplicationof exact optimizationmethods isquite limited.Theapplicationof simulation toolsenables the analysis of different yard layouts and operation strategies in real‐time workingconditionsbyvaryingdifferentenvironmentalparameters.Theusageof simulationsand theirefficiencytoestimatecapacityrequirementsarepointedoutinseveralpapers(seee.g.Ivićetal.(2010)andBeloševićetal.(2012)).Gravityyards,thethirdtypeofconventionalrailwayyards,aredesignedwithacontinuousfallinggradientandadistincttypeoflayout.Gravityyardshaveavery large capacitybutwith lotsof thedifficulties related to safety issuesand for that reasononlyfewremaininoperationtoday.Typically,classificationfreightyardsconsistof:

Arrivalyardtoreceivefreighttrainsintherailwayyardsystem; Classificationyardtodisassemblereceivedfreighttrainsandtosortrailcars; Departureyardtodepartfreighttrainsfromtherailwayyardsystemand Maintenanceyardanddepottoserverailcarsandlocomotives.

On the other hand, the railway yardsmay serve as an interface between railways and othermode of transportation, mostly as rail‐road transshipment yards. In such a system, trainsperformonly the longhaulagewhilepre‐haulageandend‐haulage isperformedbytrucks.Forthispurpose, trainsarenotdecomposedandonly loadingunitsare transshippedbymeansofhandlingequipment(gentrycranes,reachstackersorforkliftset.).Therail‐roadtransshipmentyardswere in details analyzed considering layout planning and operational problems by e.g.BallisandGolias(2002).Toprocessfreighttrainsandloadingunits,therail‐roadtransshipmentyardsareconsistedof:

Tracksidingsfortrainarrival/departureandinspectionpurpose; Transshipmenttracksforthetrainloading/unloadingoperation; Loadinganddrivinglinesforthetrucks; Storageareaforloadingunitsand Handlingequipment.

Besidestheconventionalrail‐roadtransshipmentyardsfullyautomatedrail‐railtransshipmentyardsaredesignedtoincreasetheexchangeofloadingunitsbetweentrains.Inordertoensuretherapidtransshipment,theyardsaredesignedwithoutfloorstorageareaswhicharereplacedwithsomeoftheautomatedsystemsforsorting.Althoughthisnoveltransshipmenttechnologyisstillunderdevelopmentsomeoftheseyardshavealreadybeenconstructed(seee.g.Boysenetal.(2013)).

3.CASESTADY:VRŠACRAILWAYYARD

3.1Motivation

ThecurrentroleofVršacflatyardistorearrangefreighttrainspassingthroughtheyardandtoclassifyrailcars for furtherdistributionoverrailwaynetwork inSerbiaandbordercrossingtoRomania. Furthermore, it is expected that Vršac yard obtains additional role in Serbiantransportation network through the development of intermodal transportation. TheestablishmentoftheintermodalterminalandrevitalizationofrailwayinfrastructurearedefinedastopprioritiesforfurtherdevelopmentoftransportserviceintheMunicipalityofVršac.Inthissense, the existing railway yard in Vršac should be redesigned to fulfill the additionalrequirementsforimplementationofrail‐roadtransshipment.

3.2Thecurrentstate

The flat‐shuntedyardVršac is located in the industrialzoneofVršacandserves the followingrail lines:Pančevo – Vršac – State border toRomania andZrenjanin – Vršac – Bela Crkva. Vršac yard is open to entire passenger service and the service of part-load and wagon-load consignments. The yard is divided into the group of main tracks (the tracks for the arrival/departure of trains and the tracks for freight train classification) and a few groups of sidings. Theyardhas28tracks,7outof28trackscanbeusedforthearrivalanddeparture oftrains,whileother tracksarehandling, loading,depotandothersidings.Thecurrent layoutofVršacrailwayyardisoutlinedinFigure1.

Figure2.ThecurrentlayoutofVršacrailwayyard

3.3ReconstructionofVršacrailwayyard

Theinitialframeworkfordesigningtheintermodalterminalanditsconnectiontothepublicrailwaynetwork is based on the orientation to retain the existing location of the railway yard, which isproposedintheMasterPlanofVršacandintheDevelopmentProgramofSerbianRailways.Underthecurrent state, the railway yard in Vršac functions as a flat‐shunted yard. Conducted current‐stateanalysishasshownthattheyardisabletoservetrainswithintermodalloadingunits.Ithasalsonotedthedeficienciesofthesidings’conditionsinthetermsoftheexistenceoftheinsufficientdesignoftrackladders,unequaltracklengthsandpoorsuperstructure.

Layoutoftheflatyard

Within the reconstruction (see Figure 2), the group of main tracks should be extendedconsideringtherequirements forservingintermodaltrainswiththemaximumlengthsof600‐650m.ThisextensionofmaintracksrequirestherealignmentofBelgraderailwaylineleadingayard and the complete reconstruction of the corresponding yard ladder. The expectedintensificationof the freightrailway transportrequiresadditionalcapacity,so theyardshouldbe widened with twomain tracks. Also, two shunting necks would be constructed to realizesmooth andparallel disassembling and reassembling operations on the each side of the yard.Uponthereconstruction,theyardwoulddealwiththe12maintracksandtwoshuntingnecks.Thefirstthreetracksshouldbetheplatformtracksandtheremainingnineshouldbeusedforfreightservice.Thearrivalgroupshouldbecomposedof5tracks,whilethelast4shouldbeusedforclassification.Thearrivalgroupoftracksisintendedforreceivingthetrainsforcompleteorpartial disassembling or other shunting operations (such as locomotive exchange, technicalinspectionor customs clearance).Thearrival groupofmain tracks in theyard shouldbealsousedforreceivingin‐going/out‐goingtrainsfororfromtheintermodalterminalandthelogisticcenter.All currentlyusedhandlingsidingswouldbeallocated fromtheyardandalignedoverthelogisticcenterandnearbyindustrialzone.

Figure3.ThereconstructedlayoutofVršacrailwayyard

Layoutoftherail‐roadtransshipmentyard

The construction of the intermodal terminal could be carried out independently of thereconstructionofVršacyard.Inthebeginning,theintermodalterminalcouldbeconstructedandlaunched under the current state of the railway infrastructure in Vršac. Initially, thetransshipmentyardwouldbedesignedwiththreetracksintheareaoftheintermodalterminal(see Figure 3). Two tracks would be used as transshipment tracks for the trainloading/unloadingandonetrackwouldbeusedasapassingtrackfortherun‐aroundoperationofshuntinglocomotives.Theminimumusablelengthofhandlingtrackswouldamount650minorder to ensure the smooth reception of complete intermodal trains. For the purpose ofaccompanied transport, the RoLa loading rampwould be installed on track T1. Furthermore,aftertheestablishmentof logisticcentreandinitiationofintermodaltransportit ispossibletoexpand the capacity of the transshipment yard in terms of extending existing tracks orincreasingthenumberoftracks.

Figure4.Thelayoutoftherail‐roadtransshipmentyard

Layoutoftherail‐roadtransshipmentyardconnectionwiththeflatyard

Toensurethefullandeffectiveoperationoftheintermodalterminalitisnecessarytoenableitsconnection with the public railway infrastructure (see Figure 4). The connection to the railnetworkwouldbecarriedoutindirectlyusingVršacrailwayyard.ConnectingVršacyardtothetransshipment yard could be implemented in phases, which would be in line with thereconstructionofVršacyard.

Underthecurrentstate,thetransshipmentyardcouldbeconnectedfromthemainpassingtrackintherailwayyard.Forthereasonofincreasingsafetyonthemainline,theterminalconnectionwouldbedesignedwiththeshortdead‐endtrack(seeFigure4a).

a)Beforethereconstructionoftherailwayyard b)Afterthereconstructionoftherailwayyard

Figure5.Therail‐roadtransshipmentyardconnectionwiththeflatyard

The allocation of the entrance of the railroad fromBelgrade (in the final phase ofVršac yardreconstruction)revokes theallocationof thepreviousconnection to themainpassing track intheyard.(seeFigure4b),Theconnectionoftheterminaltothearrivalgroupoftheyardwouldbeperformedindependently(usingthetrackofexistingtriangle).Thisconnectionresultsinanindirect arrival and dispatch of the compositions to and from the terminal. Parallel to thisconnecting track it ispossible toalignanotherconnecting trackand tospread thenetworkofprivatesidingswithintheindustrialpark.

4.CONCLUSION

Thispaperprovidesasurveyonrailwayfreightyardclassificationandpresentsthechallengesin designing layouts of railway yards regarding the establishment of novel intermodaltechnologies. Railway yards are recognized as key factors for the smooth functioning of thefreight transportation. The most of railway freight yards execute only the classification ofrailcarswhichdonot satisfy contemporaryneeds.Thereforemodern railwayyards shouldbemoreorientedonthetransshipmentprocedureeitherinsidetherailwaysystemorbetweentherailwaysandalternativemodes.IntheshapeofcasestudyweanalyzedtherailwayyardinVršacwhichmainlyperformstheconventionalrearrangementoffreighttrainsoverrailwaynetworkin Serbia and border crossing to Romania. In addition, it is expected that Vršac yard obtainsmoresignificantroleinSerbiantransportationnetworkthroughthedevelopmentofintermodaltransportation.Thecatchmentareaof thepotentialVršac intermodal terminalwidelyexceedsbordersof theSouthBanatDistrictbecauseof its favorableposition inSerbian transportationnetwork,nearCorridorsX, IVandVII. In thispaperwepropose the layoutsdesigned for thereconstructionof theexistingrailclassificationyardandtheconstructionof thenewrail‐roadtransshipmentyard.

ACKNOWLEDGMENT

ThispaperissupportedbyMinistryofScienceandTechnologicalDevelopmentoftheRepublicofSerbia(no.project36012).

REFERENCES

[1] Ballis,A.,Golias, J. (2002).ComparativeEvaluationofExistingand InnovativeRail–RoadFreightTransportTerminals.TransportationResearchPartA,36(7),593–611.

[2] Belošević, I., Ivić, M., Kosijer, M. (2012). Conditions for Simultaneous Formation ofMultigroupFreightTrains.Građevinar,64(7),553‐564.

[3] Boysen,N.,Fliedner,M. Jaehn, F., Pesch, E. (2013). A Surveyon Container Processing inRailwayYards.TransportationScience,47(3),312‐329.

[4] Crane,R., Brown, F., Blanchard,R. (1955).AnAnalysis of a RailroadClassification Yard.JournaloftheOperationsResearchSocietyofAmerica,3(3),262‐271.

[5] Ivić,M.,Marković, A.,Milinković, S., Belošević, I., Marković,M., Vesković, S., Pavlović, N.(2010). Simulation Model for Estimating Effects of Forming Pick‐Up Trains bySimultaneousMethod.Proceedingsof7thEUROSIMCongressonModelingandSimulation.

[6] Marinov, M., Viegas, J. (2009). A Simulation Modeling Methodology for Evaluating Flat‐ShuntedYardOperations.SimulationModelingPracticeandTheory,17(6),1106‐1129.

[7] Petersen, E.R. (1977). Railyard Modeling: Part I. Prediction of Put‐Through Time.TransportationScience,11(1),37‐49,1977.

IMPACTOFCROATIAACCESSIONTOTHEEUONTHECROATIANRAILWAYINFRASTRUCTURE,RAILWAYAND

MARITIMEFLOWS

AnteKlečinaa,SlavkoŠtefičara,NikolinaBrnjacb,*

aPro‐railalliance,CroatiabFacultyofTransportandTrafficSciences,UniversityofZagreb

Abstract:Thegoalofthispaperistotrytoportraitthecurrentstateoftherailwayinfrastructurein Croatia,main freight railway flows in the country, the state of infrastructure regarding theCroatian accession to theEuropeanUnion and the future potential of the railway freight flowsespecially flows fromtheAdriaticportsand flowsonthesocalledeast‐westcorridor(the formerPan‐EuropeancorridorX).

Keywords:railwayinfrastructure,railwayflows,maritimeflows

1.INTRODUCTION

Croatiacurrentlyhas2.722kilometresoftherailwaynetwork.Theentirenetworkismanagedby Croatian railway infrastructure Ltd company which is 100% owned by the Republic ofCroatia.AuniquecompanycalledCroatianRailways(HŽ)“lived”until2008whenitwasdividedintofivecompaniesandin2012anewrestructuringleftonlythreeindependentcompanies.Oneof them is HŽ Infrastructure Ltd which manages the railway infrastructure in Croatia today.StatesSlovenia,CroatiaandSerbiaformaspacewhichlinkssouthofEuropewiththeEastandSouthEastpartsofEurope.The freight route thatpasses through these threecountries is theshortest connection between Italy, France, Germany and Austria with Romania, Bulgaria andGreeceandtheseasliketheBlackSeaandtheAegeanSea.Thesetwoaxeshavethepotentialtoformmainfreightflows.

The freight flowsgenerated fromtheCroatian industry isveryhardtoexpect in the followingyearsduetoaharddeclineof the industrialproduction inCroatia. Worthmentioningare thepotentialflowsfromtheportofPločetowardstheBosniaandHerzegovinaandpossiblytowardsHungary. In the end the port of Vukovar should bementioned. It is the only freight port onDanubeRiverinCroatia.ThereisapossibilitythatsomegoodscanbetransportedfromAdriaticportstotheportofVukovarandthenonwardsbyDanuberiverintoHungary,Austriaandintosouthern Germany. The importance of Vukovar was recognized with latest TEN‐T networkrevisionandtheextensiontotheWesternBalkans.Vukovarwasdesignatedasthecoreportandthe Danube River in the Balkan states became a TEN‐T corridor. Also, Sava River from SisakdownstreamtoBelgradebecameapartofTEN‐T.

* nikolina.brnjac@fpz.hr

2.CROATIANRAILWAYINFRASTRUCTURE–FACTSANDFIGURES

Croatiahasanoverallof2.722kilometresoftherailwaynetwork.Only254kilometresofthesetracks are double track. In overall 977 kilometres of the network is electrifiedwith 25kVACsystem.Main nod in Croatia is Zagrebwith its railway station. Central nods are ZagrebMainstationforpassengertransportandZagrebmarshallingyardforfreighttraffic.FirstrailwaylineinCroatiawaspassingtroughnorthofCroatia,itwasarailwaylinefromHungarianNagykanizsatoSlovenianPragersko. Itwasthe linewhichconnectedBudapestNagykanizsamain linewiththeVienna–Graz–Ljubljana–Triestemainline.

2.1InfrastructureandthecapacityontheZagreb–Rijekamainline

ThelineRijeka–Zagreb–Koprivnica–Hungarianborderisentirelyelectrifiedwiththe25kVAC system and it is 332 kilometres long. Except of 20 kilometres in the Zagreb nodwhich isdoubletrackalltheotherpartsofthelinearesingletrack.Theaxleloadonlineis22,5tonsperaxle. On the part from Moravice through Zagreb until Hungarian border there an automaticblocksystemofsignalling,whileonthe90kilometreslongstretchfromMoravicetoRijekathereis only clearance between vehicles (between the train stations) which significantly reducescapacity. The line due to its hauling characteristics can be divided into five sections. TwosectionsHungarian border – Koprivnica – Lepavina andKriževci – Zagreb – Karlovac are thesections built on flat terrain allowing good speeds and gradual inclines. That gives theopportunity to haul very heavy trains with only one locomotive. This sections could allowspeedsfrom140to160kilometreperhourorevenhigherbutthecurrentstateofinfrastructurelimitsthisfrom60to140kilometres.SomepartsthatpassthroughZagrebarelimitedtoonly60kilometresperhourduetobadtrackcondition.ThestretchfromLepavinatoKriževciisonly29kilometreslongbutofferinginclinesuptoelevenpermilewhichaddsaneedfortwolocoswhentrainsheavierthan1.200brutetonsneedtobehauled.Thesimilarhaulingconditionsapplyonthe86kilometreKarlovac–Moravice.Speedsonthesesectionsvaryfrom80–140km/h.ThelastsectionfromMoravicetoRijekaisthemostdifficultones.Thelineclimbstheretomorethan900metersabovethesea.Manypartsofthelineoffersteepinclinesfrom20toeven27permile.OneofthehardestpartsofthetrackisaninclinefromRijekatoPlasestation.Morethan95%ofthis29kilometrelongsectionhasa25permileinclineatsomepartsreachingeven27permile.One electric locomotive can haul only a 500 brut tons trains here. Modern 6 Mega Wattslocomotivecanhaulupto750tonshere.Onthisstretchaddingapushinglocomotivetotherearendofthetraincansignificantlyincreasethetonnageofthetrain.Unfortunately,althoughthishauling technique was used before, during the last ten years Croatian railways didn’t allowadding a pushing locomotive. Hopefully, this can be changed allowing heavier trains to leaveRijeka.Speedonthissectionisfrom60to80kilometresperhour.ThecurrentassessmentoftheRijekamain line full capacity differs depending on different section of the line. The Rijeka –Moravicesectionhasthemaximumcapacityaround5millionnettons(12millionbruttons)ofgoodsperyear.Thiscanbealsoexpressedwithafigurearound60trainsperday.Thesectionsof that line that lie in a flat terrain can allow almost up to three times this capacity reachingaround16millionnettons(40bruttons)peryearwithsome80freighttrainsperday.Ahighercapacityisnotpossibletoreachduetoonlysingletrackandmixedpassenger/goodstraffic.

2.2InfrastructureandcapacityontheZagreb–Vinkovci–Belgrademainline

ApartoftherailwaylineLjubljana–Zagreb–Vinkovci–Belgradeisalsoentirelyelectrified.Outof its 320 kilometres that passes through Croatia 216 kilometres is double track. Also fromZagrebtoNovskatosingletrackexist,one105andtheotheris117kilometreslong.

AllthelinepassingthroughCroatiawasbuiltinaflatterrainwithmostofthecurveswithlongradiusesallowspeedfor160ormorekm/honmostoftheline.Somesmallradiuscurvescanbefoundrestrictingspeedbetween80–120km/h.All the line isequippedwithautomaticblock

system signalling. Double track parts have a very high capacity for transporting freight. Alimiting fact is that thesectionbetweenDugoSeloandNovskaasingle track line.ThiscanbealleviatedbyusinganotherlineparalleltothisonewhichrunsfromZagrebtoSisakandthentoNovska.Itis117kilometreslongwhichisonly12kilometreslongerthanthelineZagreb–DugoSelo–Novska.Itisalsoelectrified,inflatterrainandallowingspeedfrom100–140kilometresperhour.

3.CROATIANMARITIMEANDRAILWAYFLOWS

3.1Maritimeflows

Port ofRijekahandles all kindof goods.General cargo,woodand cereals come and go to theport.Main inland destinations areHungary, Serbia and Croatia. TEUs (twenty feet equivalentunit) arealsobeingmostlydelivered to the samecountries andalso inAustria.Outof almost160.000TEUsperyearmostofthemaredeliveredbytrucks,aroundthreequartersofthem.Therestgoestorail.InportofSplitthemaintypeofgoodsaregeneralcargo,wood,cereals,cementand fuels.Mostof thesegoodare transportby trucks.Mostof thecerealsandcementgoesbyrail.InportofPločealumina,fuels,oreandsomeotherkindofbulkcargoarrives.MostofthesegoodaretransportedtoBosniaandHerzegovinaandabouthalfofthegoodstravelbyrail.IntheportofŠibenikmostlybulkcargoisbeingdeliveredfrominlandsofCroatiaandHungary.Largeshipments of phosphates for Hungary and Petrokemija in Kutina were also recorded in therecent years. Phosphates are practically 100% delivered by rail. In the port of Zadar smallquantitiesofgeneral cargoand fuelsarebeingrecorded.Up to two trainsperdaydrive thesegoodwhilemostofitgoesinlandbyroad.

3.2Railwayflows

GoodstrafficflowschangedsignificantlyinthelastthirtyyearsinCroatia.TodayinCroatiathereisonlyonerailwayfreightoperatorandthisisHZCargoLtd,whichwasestablishedin2008afteradivisionofasinglerailwaycompany,Croatianrailways.HZCargois100%ownedbythestate.Today it has around 80 diesel and 50 electrical locomotives and somewhere around 6.000wagons. Ifwe compare thedata from1996where the formerCroatian railways still had430locomotives and almost 12.000 freightwagons and that number dropped to 280 locomotivesand 7.000 wagons in 2005 we can see a clearly bad policy towards the modernisation andextensionofthefleet.Thatisalsooneofthereasonswhytherailnumbersdeclinedsignificantly.

Currently it is clear that the number of goods carried inland in Croatia is declining. One of the reasons for that is the decline of the industrial production and the economical activities in general.

Table1.InlandtransportationinCroatiain2010,2011,2012byroadandrail

Overalltransport Marketshare

Measure 2010 2011 2012 2010 2011 2012 Index

Overallgoodscarried inmillionsoftons 87.170 86.439 76.527 100 100 100 99 89

Railwaytransport inmillionsoftons 12.203 11.794 11.088 14 13,64 14,49 96,65 94,01

Roadtransport inmillionsoftons 74.967 74.645 65.439 86 86,36 85,51 99,57 87,67

OverallTKM millionsofTKM 11.398 11.364 10.981 100 100 100 99,7 96,63

Railwaytransport millionsofTKM 2.618 2.438 2.332 22,97 21,45 21,24 93,12 95,65

Roadtransport millionsofTKM 8.780 8.926 8.649 77,03 78,55 78,76 101,66 96,9

From the table 1 it ismore than obvious that the inland transportation of goods is decliningwhilethemarketshareroad/railremainsthesame.Itisobviousthatthemostrailwaytransportin Croatia is transit through the country. This transit includes also the goods transloaded inCroatianportsandthencarriedoverthecountrybordersandviceversa.

Thebiggestamountof therailwaytransport inCroatia isdoneontheEast‐Westcorridor thatincludestherailwaylinesSlovenia–Zagreb–Vinkovci–SerbiaandthemainlineZagreb–Sisak–Novska.ThesecondbusiestlineistheoneconnectingportofRijekawithZagrebandHungary.Twomain types of transport can bementioned here, the first is the one fromport of RijekatowardsHungaryandSerbiaandthesecondoneisthetransportfromBosniaandHerzegovinatoZagrebandthenonwardstoSlovakiaandPoland.This transportmostly includesores fromnorthernBosnian territory. So calledDalmatian lines are thirdwith the amount of transport.This does not come as a surprise since three Dalmatian ports depend on these lines, Zadar,ŠibenikandSplit.Althoughtheamountofgoodstransportedonthe formerVccorridorseemssmall,wemust realize that the section between port of Ploče and Bosnian border is only 24kilometres long.This lineserves theportofPloče takingeachyearalmost twomillion tonsofdifferentgoods.The lines inPodravinaandSlavoniaarenext in lineservingmanyenterprisesand mostly transport agricultural goods. On these lines there is “Viro” sugar refinery inViroviticaand“Nexegroup”factoryNašiceandtheybothonthelistof10biggestcompaniesthatuse rail in Croatia. Largest users of railway transport are AGIT d.o.o. and C.K.T.Z. Both thesecompanies are logistic operators. First company is owned by HZ Cargo, the other is private.Third largest user is Petrokemija from Kutina which is situated on the East‐West railwaycorridoraddinga lot to its transport. It is also important tonotice thateight largestusersofrailwaytransportincreasedtheirtransportin2012theoverallamountoftransportedgoodsindecreasing.Thismightindicatethatthesocalledsmallusershavestartedusingrailwayservicesevenless.TwentybiggestusersofthefreightrailwayservicesinCroatiatransport88,1%ofallthe goods transported by rail. Small users therefore transport only 11,9% of all the goodstransportedbyrailandthispercentageisgettingsmallereveryyear.Thismightindicatethatatransport of goods by rail by current conditions is very unattractive formany smaller users,someofthemevenwiththepotentialuserailmuchmorethantoday.ThatindicatesaproblemthatisbeenpresentnotonlyinCroatia,butintheotherEuropeancountriesaswellandthatisthattherailwaytransportcompaniesaregettingfurtherawayfromsmallusers.Butservingonlybigusersintheendgivesuslessandlessquantityofthegoodstransportedbyraileveryyear.Thisclearlyindicatesthatalthoughthemaincorridors(likeTrans‐EuropeanTransportNetwork‐TEN‐Tnetwork) isvery important, it isvital thatcorridor linesareverywell connectedwithregional and local lines, local transport centres and industrial sidings of various scale oftransport.This caneasilybe comparedwith large riverswhichdefinitelywouldbe sohuge ifthereweren’talargenumbersofsmallriversandcreeksupstream.

4.EXISTINGTRAFFICFLOWSITALY/AUSTRIA–SLOVENIA–CROATIAANDTHEPOSSIBILITIESFORMAKINGANEWTRANSPORTFLOWSAMONGTHESECOUNTRIES

TheexistingrailwayconnectionbetweenRijekaandVenice/TriestegoesoverPivkaandSežanainSloveniaanditissignificantlylongerthantheroadconnections.Inordertomakecompetitivepassengerandpassengeroperationanewrailway line fromRijeka toTrieste isplanned.Thisline should be a logical extension from the newHungary – Zagreb – Rijeka high speed high‐capacity line.The initiativecalledNAPA–NorthAdriaticPortAssociations isan initiative thatjoinstogetherfourNorthAdriaticportsinordertoboosttheircooperation,commonmarketingandbythattoboostpotentialofalltheports.TheseportsareVenice,Trieste,KoperandRijeka.

ThepossibilitiesofNAPAaregrowingevenstrongerafterthe1stofJulywhenCroatiajoinedtheEU.ThetransportofgoodsfromRijekatowardstheEuropeandestinationsisnowevensimpler.This initiative is a goodbasis for furtherdeeper cooperationamong theseports.This alliance

will in the future definitely need a better railway link between each other. Therefore a highcapacityandahighspeedrailwaylinkRijeka–Koper–Triesteareneededinthefuture.Sincetheseportstransloadtheamountofcargoveryclosetotheamountthat forexampleaportofHamburgdoes,withcommonoperations,betterinfrastructureandagoodmarketingthesefourportscanbecometheleadingintermodalareainEurope.

East‐west corridor in Croatia forms a very good potential link of Italy and Austria towardsRomania, Bulgaria, Turkey and Greece. The growing Turkish economy had bigger and biggerneedtodrivethegoodstowardsthemiddleandWesternEuropeandthiscorridorisoneofthebestsolutions.

Inthefreightoperationstheeast‐westcorridorinCroatiacanalsobeaverygoodlinkbetweentheNAPAportsandtheBlackSeaportsinRomaniaandBulgaria.FurtherintermodalityoptionsappearwhentheserailtrafficflowsarecombinedwiththeSavaandDanuberiversandtheportsofVukovar,Belgrade,SlavonskiBrod,etc.Agreatpossibilitiescanbeseenwithasuggestednewhigh‐capacityandhigh‐speedlinebetweenZagrebandMaribor.Thislinerailwayline,togetherwith the new Zagreb – Rijeka high‐speed line and with the new Semmering basis tunnel inAustriacouldformtheshortestandthefastestbothpassengerandfreightconnectionbetweenthe port of Rijeka, Graz and Vienna. This railway partially new railway line could offer verygradualinclineswhichmakesapossibilitytohaulheavyfreighttrainsallthewaywithonlyonelocomotive. This kind of operation saves energy, nature and of course offers lower costs oftransportaswell.

Unfortunately,newZagreb–Mariborrailwaycorridorwasnot recognized in the latestTEN‐TrevisioninMay2013.Also,by“Regulation(EU)No913/2010oftheEuropeanParliamentandoftheCouncilof22September2010concerningaEuropeanrailnetwork forcompetitive freightTextwithEEArelevance”onthe firstpage it isclearlystated“Inorder tobecompetitivewithother modes of transport, international and national rail freight services, which have beenopeneduptocompetitionsince1January2007,mustbeabletobenefitfromagoodqualityandsufficientlyfinancedrailwayinfrastructure,namely,onewhichallowsfreighttransportservicestobeprovidedundergoodconditionsintermsofcommercialspeedandjourneytimesandtobereliable,namely,thattheserviceitprovidesactuallycorrespondstothecontractualagreementsentered into with the railway undertakings.” This new railway line would cut the travellingtimes both for passengers and the freight to one third of today’s time and finally make himcompetitivewiththeroadtransport.

IntotheregulationEUNo913/2010regardingthefreightcorridorsthereisaneedtoaddnewcorridors and connections to the existing corridors in order to boost the freight connectionbetweenItaly,Austria,SloveniaandCroatia.Onthepage12onthelistofinitialfreightcorridorsunder corridornumber5 there is abranch listedGraz–Maribor–Ljubljana–Koper/Trieste.Another branch should be added Graz – Maribor – Zagreb – Rijeka – Koper/Trieste. Undernumber 6with the branchKoper – Ljubljana –Budapest – Zahony another branch should beadded: Ljubljana – Pragersko – Čakovec – Nagykanisza – Budapest. Also two new freightcorridorsshouldbeaddedtothislist,thefirstisLjubljana–Zagreb–Sisak–Vinkovci–BelgradeandtheotherRijeka–Zagreb–Koprivnica–Gyekenyes–Budapest.

5.CONCLUSION

Croatia has 2.722 kilometres of the railway network and out of those only 254 kilometres ofdoubletrackrailways.Inoverall977kilometresofthenetworkiselectrified.Thisclearlyshowsthat the railway network in Croatia was newer built to a satisfactory extent. Boostingpromotional activities,more research andmore legislative support is needed to boost transittransport, transloading inportsand intermodal transport ingeneral.Besides theseactivitiesanew infrastructure capacities must be planned in order to help the boost of the railwaytransportandintermodaltransport.Portsdefinitelyneedbetterconnectionstowardsinland,but

thesenewrailway infrastructurecapacitiesshouldnotavoidthepossibilities toserveas long‐distanceandlocalpassengerlinksandlocalfreightlinksalso.Incrossborderoperationahugepotential for interoperabilityalreadyexists.LinksbetweenCroatiaandSlovenia ispossible toovercome by using multiple system electric locomotives and links between Hungary, Serbia,Bosnia and Herzegovina and Croatia can be achieved by the existing fleet of mono systemlocomotives. Further legislative and organisational steps are needed to make this possible.Croatian Adriatic ports already have large possibilities in transloading goods. This especiallyworksfortheportsRijekaandPločewhicheventodaytransloadasignificantamountofgoods.PortofRijekaisalreadymakingpossibletotransloadaround600.000TEUsperyearbutbetterorganisationofrailtransportandbetterrailwayinfrastructureisneeded.Thestrongestrailwayflows can be recognized in three areas. Hungary – Zagreb – Rijeka axis, Slovenia – Zagreb –Vinkovci – Serbia axis andDalmatianports –Zagrebaxis. A lotof freight transport alsogoesfromport of Ploče toBosnia andHerzegovina. All these flows can be boosted evenwith theexisting railway infrastructure by promotional and organisational measures. The existinglegislativeinCroatiatodayallowstheexcessofalltheEuropeanoperatorstothenetwork.SinceCroatiaonly joinedtheEUandthenewlegislativemeasures just tookthefirststeps incannotstill be seen that other operators than Croatian HŽ Cargo is running the railway freightoperations.ButtheconcessionairesintheportofRijekaalreadyannouncedthepossibilitiestoengage the other operators for the operation of transporting goods from and to the port. ItwouldbealsoverygoodthattheplannednewrailwaylineZagreb–MariborentersTEN‐Tinthenext revision. This would offer the shortest, fastest andmost economical railway connectionbetweenRijeka,Zagreb,Maribor,GrazandVienna.Also,anewbranchesandcorridorsshouldbeadded into the list of initial freight corridors (EU No 913/2010). The branches are Graz –Maribor–Zagreb–Rijeka–Koper/TriesteandLjubljana–Pragersko–Čakovec–Nagykanisza–Budapest. Two new freight corridors should be added as well: Ljubljana – Zagreb – Sisak –Vinkovci–BelgradeandRijeka–Zagreb–Koprivnica–Gyekenyes–Budapest.

REFERENCES

[1] HŽCargo(2011).Annualreport–HŽCargo2011,HŽCargo

[2] HŽCargo(2012).Annualreport–HŽCargo2012,HŽCargo

[3] CroatianBureauofStatistics(2005).StatisticalYearbookoftheRepublicofCroatian2005,Zagreb

[4] CroatianBureauofStatistics(2012).StatisticalYearbookoftheRepublicofCroatian2012,Zagreb

[5] CroatianMinistry ofMaritime affairs, transport and infrastructure (2005). New railwaylineZagreb‐Rijeka,Zagreb

[6] EuropeanParliamentandCounciloftheEU(2012).Directive2012‐34‐EU.OfficialJournaloftheEuropeanUnion,L343,32‐77.

[7] HŽInfrastruktura(2014).ReportontherailwaynetworkinCroatia.

[8] HŽInfrastruktura(2012).HŽInfrastrukturaannualreport2012,Zagreb

[9] PortofPloče(2012).PortofPločeannualreport2012,Ploče

[10] PortofRijeka(2012).PortofRijekaannualreport2012,Rijeka

[11] RepublicofCroatia(2012).Thelawofrailways–draft––December2012,Croatia

INTERMODALTRANSPORTINTHEPOLICYDOCUMENTSCONTEXT

OliveraMedara,*,BranislavBoškovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: It is the policy analysis and policy‐making processes, along with the structure andcontent of the resulting documents thatmake strategic transport planning different from onecountry to another. A national transport policy, as an umbrella strategic planning document,should meet multiple goals the relative importance of which changes over time. A need forintermodal solutions emerged as far back as the 1970s, and political interest in intermodaltransporthasgrownconsiderablysince.Topromote intermodaltransporthasbecomeoneofthekeygoals innationalandEuropeanUnion transportpoliciesalike.StrategicplanningobjectivesandkeyelementshavebeenanalysedagainstthebackdropofseveralEuropeannationaltransportpolicies,andthesummarypresentedinthepaper.

Keywords:strategicplanning,transportpolicy,publicpolicyinstruments,intermodaltransport

1.INTRODUCTION

Strategic documents for transport development constitute a prerequisite for coordinatedinvestment activity, paving the way to the growth of a national transportation system. Indifferentcountries,however,theyarenameddifferently:(national)transportstrategy,transportdevelopment strategy, transport policy, strategic plan, master plan etc. They are invariablystructured differently, even if the name is the same. The differences in the structure andcontents of transport policy strategic documents arise from different approaches to strategictransport planning (STP), but also from a regulatory framework in which policy is created,economic,social,culturalandothersocialdivides.Theyare theresultof thecurrentsituation,capacitiesandneedsoftransportinfrastructure,institutionsandhumanresources;theexistingopportunities to define specific plans for a country’s economic and social development; thepotentials and interests of scientific and research institutes in the sector, but, more thananything, the priorities of the government and relevantministry. These differences, which inprinciple exist in all transportmodes, are also incorporated in strategicdocuments related tointermodaltransport(IT).

InordertogetanIToverviewintheSTPprocess,andtoidentifythepoliciesandinstrumentsthecountriesareusingtodaytoshapetheITsector,thetransportstrategicdocumentsofmorethan 20 European countries have been reviewed. Three different approaches have beenidentified to the role of intermodal transport in thosedocuments: (1) intermodality has beenincorporated in STP as a principle (in the countries with developed IT); (2) intermodality istreatedasaformalobjective,andtheSTPisactuallybasedontransportmodes;(3)theapproach * o.medar@sf.bg.ac.rs

broughtintolinewiththeEUpolicy,butonlyasamatterofconvenience,mostoftentobenefitfromtheEUfunds.

Thepaperhasdefinedanddescribedanationaltransportpolicyasadocument.Itoverviewsanddiscusses theobjectivesand instrumentsof ITnational transportpolicies inseveralEuropeancountriesandSerbia,offeringconclusions,too.

2.INTERMODALTRANSPORTINNATIONALTRANSPORTPOLICY

NationalTransportPolicy (NTP) is adocumentadoptedby thegovernmentor theparliament,defining clearobjectives andgeneral guidelines as tohow to achieve them. It shouldbequitegeneric,butwithclearcoursesofactiontoapplyindifferentcircumstancesinthefuture.Thisisalandmarkdocumentforallpartiesandtheirstrategicplans.TogetherwiththeNTP,follow‐upstrategic documents are the elements of a state transport policy, too. The policy instruments(laws, regulations,programmes,andactionsresulting fromthepolicy)used tomaterialize thepolicyshouldreflectandreinforceitspurposes.

ANTPisexpectedtomeetmultipleobjectivestheimportanceofwhichchangesovertime.Witha shifting focus though, problems in the fieldof infrastructuredevelopment topped the list ofpriorities in the past. Different transport modes have been considered separately. To boostefficiency,aneedforintermodalsolutionstomakeabetteruseofdifferentmodesoftransportemergedasfarbackasthe1970s,butasatisfactorylevelofinteroperabilitywithinthetransportmodesisyettobeachieved.Thetransferprocessatterminalandamultilaterallegalframeworkwhichisstilladaptedtoallmodesseparatelyisjustoneofthetypicalexamples.

The EU Common Transport Policy (CTP) has a strong impact on the NTPs of the Europeancountries,bothEUmembersandaccedingcountries.EachoneofthefourWhitePapers(1992,2001,2006and2011)definedobjectivesandinstrumentsofCTP.Theyconstantlyreaffirmtheintegration across all modes, interoperability and coordination as key preconditions for thedevelopment of IT. Practice is still modal though, but multimodal issues are now taken intoaccount in theprocessofplanning.Thedominationof road transportand therestructuringofrailwaysinordertoincreasetheircompetitivenesswillforalongtimeremainachallengeforallIT actors, especially in the context of a single market policy and transport as a commercialactivity,basedonafreechoicebyserviceusers.

The challenges of IT strategic planning have been addressed in differentways, depending onstrategic planning and the structure of objectives. In some cases there are explicitNTP goals,objectives and policies to be used for action, but sometimes they are incorporated in freighttransportorlogisticsdevelopmentdocuments.Thedifferencesarealsovisibleinthewordingoftheobjectives.Somearedefinedasgeneralobjectives,andothersasspecificobjectivesorevenindicators(Table1,Hungary:ratioofcombinedgoodstransport).ObjectivesastheyaredefinedinNTParepresentedinTable1.

The vision of transport system development, along with general NTP objectives, shape theapproach to strategic planning of freight transport and, therefore, IT. In some countries, theapproachisstillmodal,with,possibly,formalcommitmenttoIT.Thisisusuallyhappeninginthestates that established their NTPs years before the adoption of CTP 2011, such as Slovakia(MoTPT,2005). IntheNTPsadoptedshortlybeforeorafter theadoptionofCTPtheapproachhaschangedconsiderably;freighttransportisaseparateissueandtherearespecificobjectivesrelated to intermodality. On the other hand, they usually fail to offer the explicit policies andinstruments to carry out the objectives. The objectives are merely formal, like in Hungary(MoTTE,2007)orfocusedontheuseoftheEUfunds,likeinBulgaria(MoTITC,2010).Therearecountrieswhichhavedefinedtheirobjectivesnotonlyfordifferenttransportmodes,butalsoforIT(Russia(МoT,2008)andSerbia(GovSRB,2008)).

Table5.OverviewoftheNTPgoalsconcerningintermodaltransport

COUNTRY GOALS/PRIORITIES/OBJECTIVES

Bulgaria(MoTITC,2010)

- Developmentofthelogisticsinfrastructure- FulfillingtheEuropeaninfrastructureandservicesstandardsalongtheTEN‐Taxes- Assuringtransparentandharmonizedconditionsforcompetitionbetweenandwithinthe

differenttransportmodesCroatia

(MoMTI,2014)- Improvementoftransportconnectivityandcoordinationwithneighbouringcountries- ImprovementoffreightaccessibilityinsideCroatia

Finland(MoTC,2007)

- Maintainingcompetitivenessinlogistics:Thetrunknetworkdevelopmentprogrammeandhigh‐qualitylogisticsservices

Germany(FMoTDI,2008)

- Avoidingjourneys‐ensuringmobility- Moretransporttorailandinlandwaterway- Upgradingmoretransportarteriesandhubs- Goodworkingconditionsandgoodtraininginthetransportsector

Hungary(MoTTE,2007)

- Increasingtheratioofcombinedgoodstransport- Improveefficiencyofintermodallogisticcentre

Spain(MoPW,2013)

- BoosttheSpanishlogisticssectorasanengineofoureconomy- Developanintermodalnetworkthatallowsshuttlingbetweennodesandprovide

completeandintegratedlogisticsservices- MaximizeSpain'sroleasagateway

UnitedKingdom(DfT,2011)

- Leveragingshorttermprivatesectorinvestment- Improvingthelongertermcapacity,performanceandresilienceofourcongestedroad

andrailnetworks- Reducingunnecessaryregulation;- Attractingandretaininghighcalibrerecruits

Croatia,whichadoptedaNTPin2014,definesallobjectivesasanintermodallistofobjectives.Theseobjectives constitute themaingoal ‐ toestablisha sustainableandefficientMultimodalTransport System (MoMTI, 2014). Spain, which defines its policy both in terms of differenttransportmodesandIT,hasamorespecificapproach.Itdefinesactivitiesforallmodessoastopromote intermodality, first of all through a system of stimulations and support. It favourstransformationofoperatorsofdifferenttransportmodesintologisticandintermodaloperators,integrationofsmalleroperatorsintransportchainsandstrengtheningofintermodaloperatorsorexpansionoftheirroleontheEuropeanandinternationalmarket(MoPW,2005).

Thecountrieswhichhavebeenthefirsttoabandonthemodalapproach,suchasGreatBritain,FinlandandSweden,arebeingfocusedontheanalysisoftransportchains,nottheanalysisofthecondition of different transportmodes. GreatBritain shifted its attention to themovement ofdifferent types of goods, and now has a more active overview of the future needs, sources,objectives and influencesof the flowsof goods, including thekey routes. It also examines thepossibility of supporting the logistics. Strategic document Delivering a Sustainable TransportSystem:theLogisticsPerspective(DfT,2008),followingtheNTP,providesadetailedanalysisofmovement of goods within the national transport corridors and examines the possibility offacilitatingeffectivemovementofgoodsandreducingnegativeeffectsbyjointeffortsmadebythegovernmentandeconomicactors.ThisdocumentisfollowedbytheLogisticsGrowthReview(DfT,2011)includingadiversepackageofmeasurestotargetrealbarrierstogrowthidentifiedby businesses across the sector ‐ from freight transport operators to logistics users in themanufacturing,wholesale,retail,postalservicesandwastesectors.

Althoughtheyhaveasimilarapproach–a longdistancetransportdevelopment trend is tobeconcentrated on specific routes and transport corridors ‐ Finland and Sweden’sNTPs place astrongeremphasisoninnovativesolutions,exploitingthepotentialprovidedbycommunicationstechnology to createan intelligent transport system(ITS).The follow‐up strategicdocuments,FinnishandSwedishITSStrategies(MoTC,2013;SRA,2010),arethefirstofintermodalnatureinEurope.IntheFinnishITSStrategythelong‐termgoalistodigitizelogistics.

Over the past ten years, five documents related to IT strategy, policy or other form of ITdevelopment havebeen produced in Serbia. Theumbrella document is the Strategy (GovSRB,2008),witha separate chapteron thedevelopmentof IT. In anutshell, at thispoint in the ITdevelopment an emphasis should be placed on institution building, eliminate infrastructurebottlenecks,buildaterminalnetworkandlogisticcentres,improvetheorganisationofactorsintheITchain.Theroleandassistanceofthestatearevitalinthis,especiallyintermsoffundingandproperconditionsforfinancingtheITinfrastructure.Eventhoughitcameouttogetherwithaproperactionplan,itremainedadeadletter.

Besides the Strategy, another twoprojects have been carried outwithin the framework of ITdevelopmentand terminalnetworks.Theseare IMODX (TheProject IntermodalSolutions forCompetitive Transport in Serbia, 2004‐2006) and The General Master Plan for Transport inSerbia2008/2009.Astrategicdocumentno.4istheSerbiaIntermodalTransport(G2G),dealingwiththeeducationofITactorsinSerbia.However,it’sjustoneofthefiveprojects,thelastone,whichtreatedaspecificproblem‐theintroductionofaSerbia‐Austriaintermodaltrain.

3.POLICYINSTRUMENTSFORINTERMODALDEVELOPMENT

Regardlessoftheapproachtakentostrategicplanning,thedifferentleveragesusedtoinfluencefreighttransportaresimilartooneanother,aswellastothoseusedinthepast.Althoughmostcountriesstillactbasedontheinfrastructureinvestmentpolicy,fiscalpolicyandencouragingachange of behaviour, there is an increasing support to training programs, new regulatoryframeworksandregulatoryreforms,andatightersupervisionoftheenforcementofregulations.Inordertogiveasystematicpresentationofpolicyinstrumentsdefinedintheanalysedstrategicdocuments, they have been classified in different areas of activity and followed by specifiedobjectivesandpolicieswhoserealizationtheyhavebeendefinedfor(Table2).Suchanapproachgives the same importance to all of the elements of the control cycle. A minimum group ofelements includes objectives and instruments, as well as indicators, which have not beenaddressedbythispaper,whilefurtherdevelopmentsshouldincludelaws,competentauthoritiesandfinancing,whichareindispensableintherealisationofapolicy.

Table6.OverviewoftheNTPinstrumentsconcerningintermodaltransport

GOAL POLICY MEASURE/INSTRUMENTAreaofactivity:Atmosphericpollution

Environmentally‐friendlytransport

Encouragetheuseofrailandwatertransport

Continuewithlowdutyonreddiesel;zerodutyonbunkerfuels;tonnagetaxforshippingcompanies;andtheexclusionofelectricrailfreightfromtheclimatechangelevyFinancialincentivesthroughmodeshiftprogrammesIncreasedfundingforcombinedtransport

Areaofactivity:Publicfinance

Efficientpublicfinanceininfrastructure

BetteraccessibilityImprovinganddevelopingrailwayconnectionswithothermodesImprovinginlandaccesstomajorports

Fundingasthebasisforservicelevel

AdaptationofthemainrailcorridorstointernationalstandardsEquipinfrastructureformultimodalinterconnectionwithadvancedITStechnologiesInteroperabilityontherailbyERTMSRedesigntransportservicesfundinganddiscontinueseparatefundingformodesoftransportFacilitatingaccesstocapitalforcommercialinvestmentSupporttheequippingofterminalswithprogressivetrans‐shipmenttechnologiesSupporttheestablishmentofpublicterminalsCreateconditionsfortheprivatesectortobuildlogisticscentres

Table2.OverviewoftheNTPinstrumentsconcerningintermodaltransport(continued)

GOAL POLICY MEASURE/INSTRUMENTAreaofactivity:Transportmarket

Transparentandharmonizedcompetitivebusinessenvironment

Equalizeconditionsofcompetition

Improvementofthequality/priceratiooftheportofferCreatinganationalportsstrategyReviewofregulatoryframeworkfortheFreighttransportinintermodalcompetition

Improveconditionsofcompetition

CompletingtheRestructuringoftheRailwayTransportSystemEnsureavailableinformationsystemstoplanandmanagesupplyanddemandoftransportservicesCreateaprogrammeforthesupportofanexpansionofthefleetforcombinedtransportCreateaprogrammeforsubsidiesintheinitialstageofoperationofregularmultimodaltransportlines

Areaofactivity:Transportservices

Competitivenessinlogistics

Securequalifiedandprofessionallyskilledemployees

ProvidefundingforestablishingnewandinnovativeapproachestotrainingRegularSummit onlabourandeducationinfreightandLogisticsInvestinlogisticscompetenceandrelatedresearchanddevelopment

Improveefficiencyoflogisticchains

SimplificationandhomogenizationoftheexistingprocedureindifferentmodestoperformadministrativeandcustomscontrolsReviewregulatoryrequirementsforallfreightsectorsonaregularbasistoensurethattheyremainappropriateConductanassessmentofbusinesslogisticscompetitivenessineachgovernmenttermMakelogisticsservicesavailabletosmallandmiddle‐sizedbusinessesIntegratetheITSconceptofGreencorridorsandintermodalterminalsBorderbottleneckseliminationFacilitatebundlingofgoodsflowsbytheagro‐logisticspilotandtheconsolidarityprogrammePromotetheimplementationofelectronicoperatingproceduresandfidawaytoencouragesmalltransportoperatorstouseelectronicsolutionsIntegrationandimprovementofexistingcodesofgoodpracticeinthesectorofLogisticsandTransport

IntelligentlogisticsCreateanationalSingleWindowSystemcapableofprocessingmultimodalelectronicdataFullyautomateddatatransferwithinthetransportchain

4.CONCLUSION

The conclusion is that the approach, form and content of NTP documents differ significantlyfrom country to country, as well as objectives, policies and instruments related to thedevelopmentofITinthesedocuments.Consequently,thecountriescanbeclassifiedindifferentways. The most obvious are the divisions into (1) the countries that have incorporatedintermodality as a principle within their strategic planning; (2) the countries addressingintermodality at the level of a formal objective,while in practice transport planning remainsseparateforeachtransportmode;(3)thecountrieswhoseapproachfollowsthepolicyoftheEU,butonlyasamatterofconvenienceandawaytousetheEUfunds.

As for the countries with an insignificant IT share in the transport market and limited ITpotentials, as well as the countries in transition or those in the EU integration process, likeSerbia, it is vital to take into account the experiences regarding inconsistencies, thecharacteristics of the environment, realistic possibilities of the public administration, thegenuinecapacityofinstitutionsandtheirpotentialforgrowth.ItisunrealistictoexpecttheNTPto contain all the elements as defined in literature, or based on the practical experience ofdevelopedorlargerstates.Strengtheningtheeconomicandinstitutionalpotentialsofacountry

allows for the external definition of general objectives and the creation and adoption of anindispensable umbrella document. These elements should be included in the next cycle ofplanning anddrafting aNTP. Likewise, theorderof specificationof elements, aswell as theirapplication,shouldbeflexible,whichwillmakeitpossibleforthedifferentelementstobeaddedanddefinedinphases,inordertoachieveconsistencyandcomprehensiveness.

ACKNOWLEDGMENT

ThisresearchwassupportedbytheMinistryofScienceandTechnologicalDevelopmentoftheRepublicofSerbiathroughtheprojectsno.36010,36027and36022.

REFERENCES

[1] DfT, (2011). Logistics growth review, Department for Transport, British Government,https://www.gov.uk/government/publications/logistics‐growth‐review

[2] DfT, (2008).Delivering a sustainable transport system: the logisticsperspective, Crown,London,UK.

[3] FMoTDI, (2008). Master Plan for Freight Transport and Logistic, Federal Ministry ofTransportandDigitalInfrastructure,http://www.bmvi.de/(inGerman)

[4] GovSRB, (2008). Development Strategy for rail, road, inland waterways, air andintermodal transport in the Republic of Serbia 2008‐2015, Official Gazette of the RS4/2008.(inSerbian)

[5] MoPW,(2013).LogisticsStrategyoftheSpain,MinistryofPublicWork,GovernmentoftheSpain,http://www.fomento.gob.es/(inSpanish)

[6] MoPW, (2005). Strategic Infrastructure and Transport Plan 2005–2020, Ministry ofPublicWork,GovernmentoftheSpain,http://www.fomento.es/

[7] MoMTI,(2014).TransportDevelopmentStrategyoftheRepublicofCroatia2014–2030,MinistryofMaritimeAffairs,TransportandInfrastructure,http://www.mppi.hr/

[8] МoT, (2008). Transport Strategy of the Russian Federation until 2030, Ministry oftransportoftheRussianFederation,http://www.mintrans.ru/(inRussian)

[9] MoTC, (2013).Towardsanew transportpolicy. Intelligence in transportandwisdom inmobility. Finland's second generation intelligent strategy for transport, Ministry ofTransportandCommunicationsProgrammesandstrategies2/2013,http://urn.fi/

[10] MoTC, (2007). Transport 2030:Major challenges, newdirections,Ministry of TransportandCommunicationsPublicationsandStrategies2/2007,http://urn.fi/

[11] MoTITC,(2010).StrategyforthedevelopmentofthetransportsystemoftheRepublicofBulgariauntil2020,MinistryofTransport,InformationTechnologyandCommunications,https://www.mtitc.government.bg/

[12] MoTPT,(2005).TransportpolicyoftheSlovakrepublicuntil2015,MinistryofTransport,PostsandTelecommunicationsoftheSlovakRepublic,http://www.telecom.gov.sk/

[13] MoTTE, (2007). Unified Transport Development Strategy 2007‐2020, Ministry ofTransport, Telecommunication and Energy, Republic of Hungary, http://s3.amazonaws.com/zanran_storage/www.khem.gov.hu/ContentPages/45092222.pdf

[14] SRA, (2010).Multimodal ITS strategy and action plan for Sweden – Summary, SwedishRoadAdministration,http://www.irfnet.ch/

REGIONALLOGISTICSANDINTERMODALTRANSPORTSCENARIOS

SlobodanZečevića,SnežanaTadića,MladenKrstića***

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Withthecontinuousgrowthoftheinternationaltrade,transportationtothehinterlandisgainingevergrowing importancewith theaimofraising thecompetitivenessofports,regionsandeconomicsystems.Stablegrowthintheinternationalcontainerflowscreatesthepressureonportsandterminalnetworkinthehinterland.Seaportsarefacingproblemsofinsufficientstoragespace,longretentiontimeofcontainersandpoortransportandlogisticssysteminthehinterland.These situations generate demands for new distribution and logistics solutions in the portshinterlands, which are based on a combination of different modes of transport and thedevelopmentofmulti‐modal logistics systems.Thispaperpresents the scenariosof logisticsandintermodaltransportationinMontenegro,withtheaimsofexpandingthehinterlandofthePortofBar,achieving the logistics sustainability, improving the economyand efficiencyof supply chainservices,integratingthelogisticsactivitiesandstimulatingtheregionaleconomicgrowth.

Keywords:regionallogistics,intermodaltransport,concentration,integration,network

1.INTRODUCTION

Logisticsandintermodaltransportarethekeywordsinthedevelopmentstrategiesofmodernanddevelopedsocialandeconomicsystemsoftheleadingcountriesintheworld.Thesuccessfulconceptoflogisticsandintermodaltransportinvolvestheformationoflogistics,i.e.intermodaltransport networks with logistics centers that represent capable and modern places ofconnection of the various modes and technologies of transport. Logistics, i.e. multimodaltransport network is largely developed in the area of Europe. There are significant terminalswiththedevelopedinfrastructureofmaritime,rail,inlandwaterwayandroadtransportationinthe region. However, in Montenegro, Serbia and the closer region, there are certaindisadvantages in terms of logistics and transportation networks, which represent a reallimitationfortheapplicationoftheintermodaltransporttechnologies.

The paper dealswith the regional logistics and intermodal transport scenarioswhichwould,within the economic system of Montenegro and the region, create the preconditions for theinclusionintothemostimportantinternationaltradeandeconomictrends.Thescenarioneedsto offer solutions which would support the "extended gateway" concept (Iannone, 2012),logisticssusstainabillity,economyandefficiencyofthesupplychains,integrationofthelogisticsactivitiesandtheeconomicgrowthoftheregion.Inordertoidentifythewin‐winsolutions,thescenarios of regional logistics and intermodal transport should be based on the fundamental

*** m.krstic@sf.bg.ac.rs

principlesofthemodernlogistics:concentration,consolidation,cooperation, integration, inter‐modality,outsourcing,synergy,sustainabledevelopmentandthequalityoflogisticsservices.

2.DECENTRALIZEDLOGISTICSSYSTEM

Traditional development of regional and national economic units of Montenegro andneighboringcountrieshasacharacteristicofdecentralizationandspatialdispersion,withoutthepresenceofintegrativeinstitutionallinks.Theconsequenceofsuchsituationisthelargenumberof systems, institutions, i.e. transport and logistics enterprises (distribution systems, portterminals, transport companies, forwarding and agency companies, etc.). These systems oftenoccupy largeareasvaluable for thedevelopmentofprofitableprograms,use theoutdatedandinadequate technology anddonot offer integrated services in the logistics chain (Tadic et al.,2013). Lack of cooperation and consolidation affects the increasingnumberof road transportvehicles,poorutilizationofcargospace,largenumberofemptytrips,lowefficiencyoftransportand high costs of logistics. This has a negative impact on the environment, possibility of theroads, traffic safety, consumption of thenon‐renewable natural resources, and it impedes theeconomicandsocialdevelopmentatthesectoral,nationalandregionallevelineveryway.

Existing plans and development tendencies of logistics systems, traffic infrastructure andeconomicentities arecharacterizedbythedecentralizationfromthespatial,organizationalandinstitutionalaspects.AdecentralizedlogisticssystemisbasedontheexistingsystemofthePortofBarandplannedroad‐railintermodalterminalsinPodgoricaandBijeloPolje,withtheaimofdeveloping an intermodal transport on the corridor Bar‐Podgorica‐Bijelo Polje‐Belgrade.However, logistics centers, i.e. warehousing and distribution systems, retain the concept ofdecentralizationorspatialdispersioninalleconomicregionsofMontenegro(northern,centralandmaritimeregion)(Figure1).

TheconceptinvolvesthedevelopmentofregionaltransportinfrastructureandconnectiontotheCorridorXandtheso‐calledCorridorXI–Motorwayof theSeas.Apart fromthepromotionofintermodal transport on the route Bar‐Podgorica‐Bijelo Polje‐Belgrade, the construction andlinkageoftheroadinfrastructureoftheregionalandinternationalsignificanceisalsoplanned.However,systemsconfiguration, inthespatialandinstitutionalterms,aswellasthedegreeofcooperation and consolidation of flows, are not appropriate for the development of strongintegrative functions between the public transport companies and the other operators in thetransportandlogisticsmarketofMontenegro.

Existing and planned storage and distribution systems are in the function of the individualpublicandprivatecompanies,whileinthespatialandorganizationaltermstheyrelatetotheirparent companiesandsituatenext to theproduction facilities, salesoutlets, shoppingcenters,industrial areas and major traffic routes. Decentralization and the individual developmentprograms represent a limiting factor for the overall optimization and upgrading of thewarehouseanddistributionsystems(vandenHeuveletal.,2013;Wagner,2010).

Logisticsofthetransportation,warehousing, inventory,packagingandmanipulationprocessesischaracterizedbytheinsourcingstrategyorthein‐houselogistics,1PL(FirstPartyLogistics).In the individual cases, the relationship between users and providers of logistics services isbasedonthetraditional2PLprinciples(SecondPartyLogistics).Dueto the lackofequipmentand infrastructure and the reduction of costs or investments, companies mainly leave theexecution of the traditional logistics services (transport and storage) to the logistics serviceprovider. In the strategy of the decentralized system, the logistics partnership is expected todevelopatthelevelofthe3PL(ThirdPartyLogistics).

3.LOGISTICSINTERMODALNETWORSCENARIOS

Basedontrendsandpossiblerealandvisionarysolutions,tworegionaltransportandlogisticssystemscenariosareproposed(InstituteofFacultyof trafficandtransportengineering,2009;Tadićetal.,2013):

Thescenarioofregionalconcentration Thescenarioofcompleteconcentrationandintegration

Plans of regional logistics encourage the economic growth (Zing et al., 2008). Concentration,collaboration,cooperationandverticalintegrationinthetransportchainoftheporthinterland,expanditscatchmentareaandenable"doortodoor"services.Inlandintermodalterminalstakea more active role in the supply chain, leading to extended gates and extended distributioncenters(Notteboom&Rodrigue,2009;Rodrigue&Notteboom,2009).

3.1Thescenarioofregionalconcentration

Thescenariooftheregionalconcentrationinvolvestheintegrationandconcentrationoflogisticsandtransportationsystemsinthethreespatialeconomicregions.Itisbasedonthedevelopmentof the integrated logistic intermodal centers in all three regions, with smaller centers orterminalsforthereceptionanddeliveryofgoods,i.e.thedevelopmentofanetworkoflogisticscentersinMontenegro(Figure2).

North region

Central region

South region

Logistics center with intermodal terminal

Local logistics center

Figure1.Decentralizedlogisticssystem

North region

Central region

South region

Logistics center with intermodal terminal

Regional logistics sub-center

Local logistics center

Figure2.Scenariooftheregionalconcentrationoftransportandlogisticssystem

Podgorica,BarandBijeloPoljeare ina functionof the important intermodal terminalswithadistinctaspectofthespatialcentralizationoflogisticssystems.Thewarehouseanddistributionsystemsofthemajorcompaniesremainatthecurrentlocations,whilethelogisticssubsystemsof small and medium sized enterprises are directed towards the concentration and locationwithin the regional logistics centers near the developed economic centers of Montenegro.Systemswithadvanced technologieswhichwilloffer thehigh‐quality logisticsserviceswillbebuiltanddevelopedintheareaofthelogisticscenter.

Thescenarioisfocusedonthetransportcorridors,intermodaltransport,greaterparticipationoftherailwaysintherealizationofthecargoflowsandthedevelopmentofcooperationbetweenthe transport and logistics companies and the companies from the industry, trade and otherservice‐providingsectorsinthefieldoflogistics.Inaddition,scenariosupportsthedevelopmentand integrationof largepublicandprivate transportcompanieswithsmallandmedium‐sizedenterprisesandintensifiesthedevelopmentofthetransportandlogisticsserviceexchange.

Regional concentration and development of intermodal logistics networks enable intensivedevelopmentoflogisticsoutsourcingonthe3PLlevel.3PLproviderhasbetteroffers,agreaternumber of service functions, and besides logistics activities, it performs the exchange ofinformation, risks and benefits between 3PL providers and companies. With the aim ofpromotingand increasing themarket shareof the intermodal transport, the concept supportsthedevelopmentofintermodalintegrators,i.e.theoperators,whichpromote,buildandmanagethe logisticschains.RailwaysofMontenegrohaveto identifysolutionswhicharebasedonthecustomerrequirementsandwhichhaveahighervaluethantheservicesofitscompetitors,butalsotobecomeirreplaceableforroadtransportcompaniesandfreightforwarders.Throughtheintegrationandstrategicpartnershipswiththespecializedlogisticsserviceproviders,railwayswouldexpandtheirdistributionnetwork,thevolumeoftransportandthenumberofclients.

3.2Thescenarioofcompleteconcentrationandintegration

Thescenarioofcompleteconcentrationandintegrationoftransportandlogisticssystemfitstheconceptof "extendedgateway", and it isbasedon the logistics centersofdifferent structures,sizes and functions. The scenario involves the development of the logistics platform thatintegrates logisticsandsupportingsystemsandactivities inadefined, i.e.arrangedspace,andrepresentsaplaceofconcentrationofthetransportandlogisticsflows.Withinthisplatform,thefunctionsof thedryport terminals,cross‐dockingterminals,city‐logistics terminals,andotherterminals,warehousesanddistributionsystemsisbeingdeveloped(Figure3).

North region

Central region

South region

Logistics center with intermodal terminal

Local logistics center

Logistics platform with dry port, cross-dock, city logistics and intermodal terminals

Figure3.Scenarioofcompleteconcentrationandintegrationoftransportandlogisticssystem

Dryportisacomplexoflogisticalsystemsandactivitiesinthehinterlandoftheseaports(Roso,2008). The conceptwas developed in order to keep the high quality of the port services andanswers to the demands of the ever growing cargo flows. Dry port terminal is located in thehinterlandandconnectedwithoneormoreportsbyrailwayand/orroadtransport.Thetaskofthe dry port is to collect the goods for the maritime transport over long distances and thedistributionofgoodsonthelocal,regionalandinternationallevel.Theseterminalsprovidesomeadditional services such as customs clearance, warehousing, packing, repacking, update data,informationservicesetc.InthescenarioofcompleteconcentrationandintegrationoftransportandlogisticssystemofMontenegro,thedevelopmentofthedryportterminalhasthefunctionofthespatialrelaxationofthePortofBar.Theterminalwouldbemultimodalorientedanditwouldcontainallservices,facilitiesandequipmentrequiredbytheshippersandforwardersfromtheport.PortofBarandthedryportwouldbetheunifiedfunctionalandtechnologicalsystem.Theintegration of the port and the dry port enables the "extended gateway" concept and larger

shareoftherailwaysinthetransportoperations,thusimprovingtheoverallcosteffectivenessoftheentiresystem(Iannone,2012).

Logisticsplatformmayhaveabroaderregionalimportancewithsettlementofthesystemsthatwould expand the logistics service offer with the value added service (VAL, value addedlogistics), such as: labeling/marking of the goods, packing and repacking, sterilization, filling,mixing,painting,finalassembly,installation,MergeinTransit(MIT),etc.Inthelogisticscenter,as the point of connection, reconsolidation of the componentswithout holding inventories isbeing done. This could become very profitable service concept for all freight flows whichgravitate towards the countries of the region through the Port of Bar. Within the logisticsplatform and the Port of Bar, the cross‐docking and the city logistics terminals are beingdevelopedwiththeprimaryfunctionofcentralizedandintegratedsupplyoftheurbanareas.

All warehouse systems of Montenegro gravitate towards the logistics centers, i.e. logisticsplatform in the central part of Montenegro. In spatial terms, Luka Bar relieves of the largewarehousingsystemsandorientstowardsthetransit‐transshipmentsystems,i.e.cross‐dockingterminals,andperformsthefunctionofthecitylogistics.Intermodaltransportterminalsremainatthreelocations(Bar,Podgorica,BijeloPolje),andthemicrolocationofthelogisticsplatforminthecentralpartofMontenegroshouldbecarefullydefined.

Concept of the port and dry port is based on a strong transport connection, primarily by therailway,which requires thehighestdegreeof integrationbetween theportsand the railways.Transportbytheheavyfreightvehiclesislimitedtotheconnectionsbetweentheterminals,andtransport within the urban areas to the small and eco‐friendly vehicles. The warehousingsystemsarefullytransferredontotheoutsourcingstrategy.Inordertoreducetherequiredarea,modern warehousing technologies and transshipment systems are being applied for certaintypes of goods. In relation to the existing situation, the concept of concentration providessignificantsavings,reductionoflogisticssystemsandtherequiredsurfaces.

Given that scenario represents the highest level of cooperation and integration between thepublicandprivatesectors,thedevelopmentofthelogisticsoutsourcing,i.e.3PLand4PL(Fourthparty logistics) strategies are being stimulated. In institutional and organizational terms, theportandthedryportmayconstituteasingleunit,opentothehighestlevelofintegrationwiththerailwayofMontenegroandotherpublicandprivatetransportandlogisticscompanies.Thepartnership concept between the 3PL and 4PL logistics service providers and the companiesfrom the other economic fields providesmultiple effects, such as the efficiency improvement,reduction of the operational, transportation and inventories costs, improvement of thetechnologiesandthequalityofservices,reductionoftheinvestmentinthenon‐profitsystems,thedevelopmentofcoreactivitiesoftheclientcompany,divisionandreductionoftherisks,etc.

4.CONCLUSION

Transport and logistics systemofMontenegro and the region lags behind themodern trends.The gap is not only in technical and technological, but often in the planning, organization,marketingandqualitativeterms.The logisticsoftheregion iswithoutthenecessarydegreeofconcentration and consolidation of flows and providers of the logistics services, while thetransport and traffic systems are characterized by an extremely low level of intermodaltransportdevelopment.Thecausesaredifferent,andtheyaremanifestedthrough: insufficientflows volumes, a lack of investment, poor organization, poor marketing and lack of theawarenessoftheimportanceoflogisticsandtheapplicationofmoderntransporttechnologies.

Thedescribedscenariosof transportand logisticssystemdevelopmentbelongto thegroupofsolutionswithwhichMontenegrocanbeplacedamongwell‐organizedandplannedEuropeancountries.Applicationofthedryportsystemortheregionalintermodalterminals,withtheaimofconnectingtheportwiththeshippersandreceiverswithintheregion,enablesthegrowthof

competitivenessandsustainabilityofthemulti‐modaldistributioninthehinterlandofthePortof Bar, i.e. the development of Montenegro and the region. The proposed solutions have theelements of openness, multi‐functionality, multimodality, with special emphasis on thepossibility of cooperation between all forms of organization, ownership and size of thecorporations that participate in the realization of logistics services. In some of the proposedsolutions,thepartnershipsbetweenthesmallandmediumsizedenterprisesandthepublicandprivatecompaniesmayrepresentthemostefficientformofeconomic,organizational,functionalandtechnologicalcooperation.Inaddition,thedevelopmentofanetworkoflogisticscentersandahigherdegreeofthesystemintegrationallowstheexpansionofthehinterlandofPortofBar.

The successful development of logistics and intermodal transportation in the territory ofMontenegroandtheregioncanbeexpectedafteradoptingoneofthescenariosofcentralizationoflogisticssystems.However,itisnecessarytoadoptandimplementanintegratedpackageofrecommendations and measures related to the (Tadić & Zečević, 2012; Iannone, 2012):infrastructure policy, development of a network of logistics centers, improvement of the railtransport services, legislation,alterationof thecustomsprocessesandprocedures, removalofthe technical and legalbarrierswith theaimof fair competition in the transportmarket,newbusinessmodelsforthelogisticssystemsintegration,establishmentofnationalassociationsandbodies,provisionofthefinancialmeasuresandsupport.

REFERENCES

[1] Iannone, F. (2012). The private and social cost efficiency of port hinterland containerdistribution through a regional logistics system. Transportation Research Part A, 46,1424–1448.

[2] Institute of Faculty of traffic and transport engineering, Department of logistics (2009).Studija logističke integracije subjekata transportnog sistema regiona – UčešćeintermodalnogtransportaCrneGore,Belgrade,Serbia

[3] Notteboom, T., Rodrigue, J.P. (2009). The future of containerization: perspectives frommaritimeandinlandfreightdistribution.GeoJournal,74(1),7–22.

[4] Rodrigue, J.P., Notteboom, T. (2009). The terminalization of supply chains: reassessingport‐hinterlandlogisticalrelationships.MaritimePolicy&Management,36(2),165–183.

[5] Roso,V.(2008).Factorsinfluencingimplementationofadryport.InternationalJournalofPhysicalDistribution&LogisticsManagement,38(10),782–798.

[6] Tadić, S., Zečević, S., 2012.Development of intermodal transport and logistics in Serbia.Internationaljournalfortrafficandtransportengineering,2(4),380‐390.

[7] Tadić,S.,Zečević,S.,Krstić,M.(2013)Vrednovanjekoncepcijaregionalne logistike.SYM‐OP‐IS2013,Zlatibor,515–521.

[8] vandenHeuvel,F.P.,deLangen,P.W.,vanDonselaar,K.H.,Fransoo, J.C. (2013).Regionallogistics land allocation policies: Stimulating spatial concentration of logistics firms.TransportPolicy,30,275–282.

[9] Wagner,T.(2010).Regionaltrafficimpactsoflogistics‐relatedlanduse.TransportPolicy,17(4),224–229.

[10] Zing,Q.,Huapu,L.,Haiwei,W.(2008).PredictionMethodforRegionalLogistics.TsinghuaScience&Technology,13(5),660–668.

THESELECTIONOFTHELOGISTICSCENTERLOCATIONUSINGAHPMETHOD

ŽeljkoStevića*,SlavkoVeskovićb,MarkoVasiljevićb,GoranTepićb

a UniversityofEastSarajevo,FacultyofTrafficEngineeringDoboj,BosniaandHerzegovinabUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbiab UniversityofEastSarajevo,FacultyofTrafficEngineeringDoboj,BosniaandHerzegovinab UniversityofNoviSad,FacultyofTechnicalSciences,Serbia

Abstract:Theaimofthispaperisselectionofthemostacceptablelocationthroughoutthestateof facts: the ability to prove the purpose of this geographical area to form one complete andcomplex logistics systemwhere the logistics center (LC)would be an interconnection betweenproduction and consumption. Themulti‐criteria analysis, i.e. AHPmethod (Analytic HierarchyProcess)wasusedforthechoiceoflocation.Apartfromthehandcalculation,the"ExpertChoice"softwarewasusedforbetterpresentationofresults,aswellasfortheirvalidity.Basedonasetofcriteriaand their evaluation, then the evaluationofalternativesaccording to these criteria, theapplicationoftheAHPmethodleadstothemostacceptablesolution.

Keywords:logisticscenter,AHPmethod,location,multi‐criteriaanalysis.

1.INTRODUCTION

The theme of this work is chosen in accordance to the needs of the economic system of acountry,andtheneedsofallparticipantsinthatsystembecause,astheGermanssay:"Logisticsis likefueltoeveryeconomicsystem"andit iscommonknowledgethat logisticscentersareakey element of the entire logistics network. The functioning of the entire logistics systemdependsontheexistenceoflogisticscenters,whichhave,sotospeak,theroleofan"umbilicalcord" in logistics, since they connect and integrate all logistics systems, subsystems, activitiesandprocesses.

GermanagriculturaleconomistsJ.H.ThunandA.Weberareconsideredtobethepioneersofthedevelopmentoflocationtheory,whileintermsofmathematicalformulationsitisbelievedthatFermabegantheconsiderationoflocationproblems.VidovićandMiljuš(2004).

Frommethodologicalpointofview,amulti‐criteriaanalysisisasystematicapproach,andthusthemosteffectiveandmostfunctionalmethodologicalapproachtoproblemsolving.Sincethisisa multi‐criteria problem, systematic approach to its solution requires the use of methods ofmulti‐criteriaanalysistooptimizethesolutionoftheidentifiedproblem.Kovačic(2008).

AHP is especially suitable for complex decisions which involve the comparison of decisionelementswhicharedifficult toquantify. It isbasedon theassumption thatwhen facedwithacomplex decision the natural human reaction is to cluster the decision elements according totheir common characteristics. It involves building a hierarchy of decision elements and then

* zeljkostevic88@yahoo.com

makingcomparisonsbetweeneachpossiblepairineachcluster.Thisgivesaweightingforeachelementwithinaclusterandalsoaconsistencyratio.Saaty(1980)

AHPisatheoryofmeasurementthroughpairwisecomparisonsandreliesonthejudgementsofexpertstoderivepriorityscales.Saaty(2008)

AHP is used extensively for decision making in the areas of management, allocation anddistribution.Konstantinos(2014)

2.THEPROBLEMPOSTULATEANDTHEMETHODOLOGY

AccordingtoZečević(2006)developmentofanetworkofLCatnationalandinternationallevelisaprerequisiteforoptimizationoftransportandlogisticschains.Theidealsolutionwouldbethe formation of the network, where all three sites under consideration had logistics center,however,thepaperdiscussestheselectionofthemostacceptablesolutionoutofagivencriteriaset,becauseinthebeginningitisveryimportantthingtohaveit,ifonlyonelogisticscenter.

Applicationofamulti‐criteriaanalysismethodinsolvinglocationproblemsisverycommonandthereareanumberofworksdealingwith this issue, forexamplesolvingsiteselection,wheremoremethodswereapplied,outofwhichoneistheAHP,asinYildirimandÖnder(2014).AHPstands out as one of themainmethods ofMCDM (multiple criteria decisionmaking),when itcomestosolvingapproachtotheseproblemsasevidencedbyTomićetal.(2014).MorerecentlytheincreasinglyextendedAHPmethodisbeingused,ie.fuzzyAHPwhoseapplicationissolvingtheproblemsoflogisticsnaturesuchasinKayikci(2010)andTadicetal.(2015).

The choice of the logistics center location is based on an integrated decisions and riskmethodologyfortheselectionofthebestlocationswhichinvolvesgeneralstepsaslistedbelow:theinitialstepformsaschedulescollectionandacceptablespatialalternatives.Step1definesasetofcriteriafordecisionmaking,step2identifiestheinitialweightoftherelevantcriteria,step3usestheAHPasoneofthetechniques(MCDM),step4establishesarankinglistofalternatives.

OneofthemostrelevantpartoftheAHP,isrelatedwithtogiveastructuretotheproblemtobesolve through the hierarchy. In this phase, the decision group involved should divide theproblemonhisfundamentalcomponents.(Saaty1980)

AHPconsistsof foursteps.One,definetheproblemandstatethegoal.Two,definethecriteriathatinfluencethegoal.Three,usepairedcomparisonsofeachfactorwithrespecttoeachotherthat forms a comparisonmatrixwith calculatedweights, ranked eigenvalues, and consistencymeasures.Four,synthesizetheranksofalternativesuntilthefinalchoiceismade.Melvin(2012)

TheoreticallytheAHPisbasedonfouraxiomsgivenbySaaty;theseare:Axiom1:Thedecision‐maker can provide paired comparisons aij of two alternatives i and j corresponding to acriterion/sub‐criteriononaratioscalewhichisreciprocal,i.e.aji=1/aij.Axiom2:Thedecision‐maker never judges one alternative to be infinitely better than another corresponding to acriterion,i.e.aij≠∞.Axiom3:Thedecisionproblemcanbeformulatedasahierarchy.Axiom4:All criteria/sub‐criteria which have some impact on the given problem, and all the relevantalternatives,arerepresentedinthehierarchyinonego.

2.1.Relevantcriteria

There is a number of criteria that can be studied in relation to the choice or ranking ofalternatives. In order to define the relevant criteria, hierarchical structureswere established,defining the groupof high‐level and lower‐level criteria. Thehierarchical structure of criteriausedinthisresearchforthechoiceoflogisticscenter'slocationconsistsof3groupcriteriaand6criteria (table 1). Criteriawere chosen in accordancewith the standards for defining a set ofcriteriatobeusedinsolvingtheseproblems.

Table1.Thehierarchicalstructureoftherelevantcriteria

Groupcriteria CriteriaLevel TypeSpatial availablesurface

NumericalLandprice

Geographic geographicallocationLinguistic

macro‐microleveloflocationTraffic affiliationtotheformoftransportation Numerical approachwaysaccessibilityoftransportequipmenttothelogistics

centerLinguistic

2.2.Theinitialmassofrelevantcriteria

Oneofthemainfeaturesofmulti‐criteriadecision‐makingisthattheeachcriteriamaynothavethesame importance.Toavoidsubjectivity in theprocessofdetermining therelativeweights,the criteria standardization is for the purposes of this paper, carried out by Delphi methodLinstone (1975) used to identify the initialweight of the relevant criteria. The procedure fordetermining the initialweightwas carried out in three phases, as defined by the hierarchicalstructure of criteria: determining the relative weight of every criteria group (gk, k=1..3),determining the relative weight of every relevant criterion (ci, i=1..6), corrections of relativeweightofcriterionwithitsgroupweight(wii,i=1..6).

The Delphi method is based on structural surveys and makes use of the intuitive availableinformationoftheparticipants,whoaremainlyexperts.Therefore,itdeliversqualitativeaswellas quantitative results and has beneath its explorative, predictive even normative elements.There isnot theoneDelphimethodologybut theapplicationsarediverse.There isagreementthatDelphiisanexpertsurveyintwoormore'rounds'inwhichinthesecondandlaterroundsof the survey the results of the previous round are given as feedback. Therefore, the expertsanswer from the second roundonunder the influenceof their colleagues' opinions.Thus, theDelphi method is a 'relatively strongly structured group communication process, in whichmatters,onwhichnaturallyunsureandincompleteknowledgeisavailable,arejudgeduponbyexperts'.HäderandHäder(1995)

This research included traffic engineers, construction engineers and spatial planners, whichdetermined criteria weighting through a survey: w1=0.181; w2=0.421; w3=0.227; w4=0.080;w5=0.050;w6=0.040.

3.THESELECTIONOFTHELOGISTICSCENTERLOCATION

Thispapertookintoconsiderationthreealternativesites:Doboj,BanjaLukaandŠamacasareasthatwiththeircharacteristicsandgeographicpositionaresuitableplacesfortheformationofalogisticscenter.Given locationsarepartof totalnumberof locations,whichwerediscussed inthe studyof intermodal transport forBosnia andHerzegovina in2008.Definedcriteriaareofcoursecommontoallthreelocationsinordertomaketheircomparison.Theresearchrepresentonly illustrationofapplyingAHPmethodonanarbitrarysetof locations forgivencriteriaset.HierarchicalpostulateofAHPmethodforgivenlocationproblemisshowninFigure1.

Figure1.HierarchicalpostulateofAHPmethodforgivenlocationproblem

ComparisonofalternativesinrelationtothecriteriaonthebasisofSaati'sscaleSaaty(1980)isshownintable2whereyoucanseetheirimportance.Forexample,relationA2A1forK1=1/3meansthatalternativeonehaslowdominancecomparedtootheralternativeinrelationtothefirstcriterion.

Table2.Comparisonofalternativespercriterions

K1 K2 K3 K4 K5 K6A1 A2 A3 A1 A2 A3 A1 A2 A3 A1 A2 A3 A1 A2 A3 A1 A2 A3

A1 1 3 1/5 1 7 2 1 3 4 1 1 1/9 1 4 3 1 1/3 1/3

A2 1/3 1 1/7 1/7 1 1/5 1/3 1 3 1 1 1/9 1/4 1 1/3 3 1 2

A3 5 7 1 1/2 5 1 1/4 1/3 1 9 9 1 1/3 3 1 3 1/2 1

In the next step a table is formed that represents the vector of eigen values of comparingalternatives with respect to the first criterion, (table 3), which is prepared as follows:a11=1/(1+0.33+5)=1/6.33=0.158;a12=3/(3+1+7)=3/11=0.273;a13=0.2/1.343=0.149etc.

Table3.Vectorofownvaluecomparisonofalternativesinrelationtothefirstcriterion

A1 A2 A3 ΣA1 0.158 0.273 0.149 0.58A2 0.052 0.091 0.106 0.249A3 0.79 0.636 0.745 2.171

Table ΣK represents the sum of vectors of own values and in order to commence linesnormalization,sumisbeingdividedwithnumberofcolumnsie.inthiscasewiththree(sincewehavethreealternatives):

13

0.580.2942.171

0.1930.0830.724

;1 3 0.2

0.33 1 0.1435 7 1

0.1930.0830.724

0.5870.252.27

;0.587/0.1930.25/0.0832.27/0.724

3.0413.0123.135

Tocalculatethemaximalmeanvalueofcomparisons'matrixλmax,matrixistobemultipliedwithvectorofcomparisons,whatwasrepresentedbythelastmatrix.

1 3.041 3.012 3.1353

3.063;1

3.063 33 1

0.0315

0.03150.58

0.054

RI (randomindex)istakenfromSaatytable(1980)onthebaseofnumbernwhichrepresentsthealternative.Herewehavethreealternatives,ie.n=3andrandomindexvalueis0.58.HereitisveryimportanttomentionthatconsistencydegreeCR is0.054whatmeansthattheresult isvalid,becauseresultsarevalidifthisdegreeissmallerthan0.10.

Whenitcomestothefirstcriterion,i.e.theavailableareaforthelocationofthelogisticscenter,thegreatestsignificancehasthenumberthreealternative, ie.Šamaclocationbecauseithasanareaofabout40ha,whilesecondalternativehasabout6haandthefirstalternativehasabout10 ha, and so they are evaluated by the size of their fields. The method of evaluation ofalternativesinrelationtoothercriteriaisthesameandtheresultsaregivenbelow.

Ifcomparingofalternativesisdonebythesecondcriterionie.thelandprice,thecheapestoneislocationnumberonefollowedbythirdlocationandattheendthesecondone.Suchasequenceof results is a consequence of the position of locations themselves ie. location number two(BanjaLuka)asthecapitaloftheRepublicofSrpskahas,ofcourse,themostexpensivelandasthe result of a slightly higher standard than the rest of the entity.When it comes to locationnumber three the land price ismore expensive slightly compared to the price of the land at

locationnumberone,primarilybecauseit isalandalongtheSavaRiver.Whenitcomestothethirdcriterion,alternativeonehasthebestresultandthatisprimarilybecauseofcrossingroads‐ road and railway. Affiliation to the form of transportation is the fourth criterion of a givenlocationproblem,andanalternativenumberthreehasthehighestpriorityvectoraccordingtothis criterion because of the location of the port of Šamac, which has affiliation to the threeaspectsoftransport:road,railandwaterways,whiletheremainingtwositesonlyhavetheroadandrailtransport.Inrelationtothefifthcriterionlocationnumberonehasthehighestpriorityvectorbecauseit'sbasicallyasitethathassuchgeographicallocationthatcanservewithgreatsuccessinboth,microandmacroenvironment.

Afterpreviously completedsteps, result isobtainedbymultiplyingvectorsprioritiesobtainedbymutualcomparisonofcriteriawithalternativepriorityvectorsaccordingtothecriteriaandsumminguptheirmultiplicationproducts.

A1=WK1*WA1+WK2*WA1+WK3*WA1+WK4*WA1+WK5*WA1+WK6*WA1=0.467A2=WK1*WA2+WK2*WA2+WK3*WA2+WK4*WA2+WK5*WA2+WK6*WA2=0.142A3=WK1*WA3+WK2*WA3+WK3*WA3+WK4*WA3+WK5*WA3+WK6*WA3=0.391

Basedoncommencedcomparisonofallinputdata,locationnumberonewasselected.Checkofmethodologywascommencedwithassistanceofsoftwarepackage„ExpertChoice”,andresultisshown in Figure 2. It is obvious that the alternative number one ie. location Doboj has thebiggest priority vector,when observed generally and in total in relationwith all commencedcomparisons,aswellasinbasicpostulate.

Figure2.Finalresultoflocationselection

4.SENSIVITYANALYSIS

Thesoftwarealsoprovidescertainchartswhichareenablingustoperceivedifferencesbetweenevaluated solutions. Figure 3 makes it possible to see how the priorities of alternatives aresensitive to the weight changes of individual criteria. Increase in the weight of the secondcriteria(priceof land‐themost influentialofall)causesgrowingofpriorityofthealternativenumberoneie.locationDoboj,whilepriorityoftheremainingtwolocationsdecreases.

Figure3.Gradientdiagram

Figure 4. a) shows the criteria bywhich one alternative is preferred over the other. Selectedalternativesareoneandthreebecausetheyhavethehighestpriorityvectorsbyallcriteria.Forinstance Šamac location has the advantagewhen it comes to available land, belonging to theform of transport and accessibility approach of TS, while Doboj location supersedes othercriteriaandintheend,anoveralladvantagewasgivenandhighlightedingray.

Figure4.a)Differencesdiagramb)Performancediagram

Figure4.b)showsvaluesofalternativespereachcriterionie.perpriorityvectorofalternativesbyasinglecriterion.Thevariationofalternativesaccordingtothecriteriaisclearlyvisible.

5.CONCLUSION

Data available during the development and the implementation of AHP as the multi‐criteriaanalysismethod,leadstosolutionwhichrepresentsthemostacceptablesolutionofcriteriasetsfortheconstructionoftheLC.Itisthealternativenumberoneintheappliedmethodie.locationofDobojthatshowedasthemostacceptablefortheconstructionofthelogisticsnetwork'skeyelement, ie.LC.Dobojasapotential locationfortheconstructionofLChasbeenrecognizedatthetimeoftheformerYugoslaviawhenthetopexpertshavemadeprojectsofLCintheterritoryofformerrepublics,andamongthemDobojhadfounditsplace.Then,tenyearslaterprojectwasre‐made for thesame location,whichmeansthat it is justifiedtoconstruct theLC in thisverymuchurbanarea.

REFERENCES

[1] Häder M., Häder S., (1995) Delphi und Kognitions psychologie: Ein Zugang zurtheoretischenFundierungderDelphi‐Methode,in:ZUMA‐Nachrichten,vol.37,19.

[2] Kayikci Y. (2010). A conceptual model for intermodal freight logistics centre locationdecisions,ProcediaSocialandBehavioralSciences26297–6311

[3] Konstantinos, P. (2014). The Analytical Hierarchical Process, International HellenicUniversity

[4] Kovačić, M. (2008). Primjena AHP metode za izbor lokacije luke nautičkog turizma naprimjeruSjevernogJadrana,Našemore,Znanstveno‐stručničasopiszamoreipomorstvo, 55(5‐6),249‐261.

[5] Linstone,H.,Turoff,M.D. (1975).DelphiMethod:TechniquesandApplications,Addison‐Wesley,Reading(Massachusetts),

[6] Melvin A. (2012). Decision‐Making using the Analytic Hierarchy Process (AHP) andSAS/IMLSESUG

[7] Saaty,T.(1980).„TheAnalyticHierarchyProcess”,McGraw‐Hill,NewYork,[8] Saaty,T.(2008).Decisionmakingwiththeanalytichierarchyprocess2008. [9] TadićS.,ZečevićS.,KrstićM.(2015).Rankingoflogisticssystemscenariosusingcombined

fuzzyAHP‐Vikormodel,InternationalJournalforTrafficandTransportEngineering,5(1):54‐63

[10] Vidović, M., Miljuš, M. (2004). Lokacijski problemi‐značaj, vrste i načini rešavanja,Vojnotehničkiglasnik52(5),445‐457

[11] YildirimB.,ÖnderE.(2014).EvaluatingPotentialFreightVillagesinIstanbulUsingMultiCriteriaDecisionMakingTechniquesJournalofLogisticsManagement,3(1):1‐10

[12] Tomić V., Marinković D., Marković D. (2014). The selection of Logistic centers locationusing multi‐criteria comparison: Case Study of the Balkan Peninsula Acta PolytechnicaHungaricaVol.11,No.10,

[13] ZečevićS.(2006).Robniterminaliirobno‐transportnicentri,SaobraćajnifakultetBeograd

PartIII MATERIALHANDLINGAND

WAREHOUSING

COMPETITIVEACCESSANDSERVICEARRANGEMENTSOFCONTAINERMOVEMENTS:PORTBOTANYCASE

MihailoCvetojevića,*

aVesselOperationsGroup,DPWorldAustralia–Sydney

Abstract: The paper analyses New SouthWales (NSW) Government role in supply chain andnetwork efficiency as a newmodel that ensure competitive access and service arrangements ofcontainermovementsbetweenstevedoresandtransportcarriersatPortBotany.ThepaperfurtheranalysesOperationalPerformanceMeasures(OPMs)setbytheGovernmentbodythatensuresthattheport supply chain’s stakeholdersaremadeaccountable to eachother for theirperformance.Logistic chain participants are penalized, by paying a financial penalty to the other party, forperformancethatfailstomeettheOperationalPerformanceMeasure(OPM)standards.Thepaperfurther analyses the impact onDPWorld Sydney ‐ PortBotany container terminal operationalrequirements, changes of Key performance indicators (KPI) related to landside operations andstrictmonitoringof terminalperformance,both landsideandwaterside, invery tight timeframe,decisionmakingprocessforreallocatingresourcesfromwatersidetolandsideoperationsandviceversainordertominimizepossiblefinancialpenaltiesortogainfinancialbenefits).

Keywords:Containerterminal,OperationalPerformanceMeasures

1.INTRODUCTION

Governmentroleinthelogisticandcargomovementtaskinparticularshallbealsofocusedondelivering network capacity that enables supply chain efficiency. This includes removingobstaclesforachievingbestpractice,creatingcapacityand,wherenecessary,becominginvolvedinthemarketplacetoensurethenetworkoperatesefficiently(Ferreira,Bunker,2003).

Cargomovement is a basic element of logistics. InNewSouthWales (NSW), as inmost otherstatesandareas,thecargomovementismainlyundertakenonatransportnetworkwhereboththe movement of cargo and the movement of people compete for space. The governmentprimarilyprovides thephysicalnetwork,andaccess to it;however there isan inexorable linkbetweentheactionsofgovernmentandtheperformanceoflogistictasksacrosstheeconomy.InNSW the logistics industry accounts for approximately 13.8 per cent or $58 billion of NSW’seconomy;wheremorethan67billiontonnekilometresofcargoismovedannuallyandthevalueoftheproductscarriedexceeds$200billion(TransportforNewSouthWales,2015).

TheGovernmentofNSWhasmadeFreightandPortsStrategy thatwas released inNovember2012, with aim to enhance productivity and efficiency in cargo flows realization. The NSWFreight and Ports Strategy identifies where government intervention is justified to enhanceproductivityandeconomicefficiency.Government interventionwasdefined in the formof the

* Mihailo.Cvetojevic@dpworld.com

provision of physical infrastructure, coordination and control, market structure reforms, co‐investmentwiththeprivatesector,regulatoryreformandothereconomicincentives.

TheobjectiveofthepaperistoanalyzetheimpactonDPWorldSydney‐PortBotanycontainerterminal operational requirements, changes of KPI related to landside operations and strictmonitoringofterminalperformance,bothlandsideandwaterside,decisionmakingprocessforreallocating resources from waterside to landside operations and vice versa in order tominimizepossiblefinancialpenaltiesortogainfinancialbenefits.

Remaining of the paper is organized in following way. The section 2 presents landsideimprovementstrategyandNSWgovernmentregulatoryreform.Section3discussesoperationsprotocols,KPIandpracticeofPBLISregulationsapplicationinDPWorldPortBotanycontainerterminalandssection4givessomeconcludingremarks.

2.PORTBOTANYLANDSIDEIMPROVEMENTSTRATEGY–NSWGOVERNMENTREGULATORYREFORM

Port Botany is essential infrastructure asset that serve as the primary import and exportgatewaystoNSW.NSWisAustralia’s largesteconomyandhometoapproximatelyonethirdofthe Australia’s population. Port Botany is located 12 nautical miles south of the entrance toSydney Harbor and is well serviced by road and rail networks. The facilities at Port Botany(Figure1)currentlyconsistof3containerterminalswith12containervesselberthsand2bulkliquids berth, container support businesses, bulk liquid berth storage facilities and privateberths.ThevolumeofcontainersthroughPortBotanyhasapproximatelydoubledoverthepast11years,fromabout0nemillionTEUin1999tomorethantwomillionin2014(TransportforNewSouthWalesStrategy,2015).

Containervolumegrowthisforecasttocontinueatbetween5–8%perannumoverthenext25years (Source: Sydney Ports 30 Year Vision). Using a growth rate of 7%, the volume ofcontainers throughNSW portswill be about 11million TEU by 2036‐37 (Transport for NewSouthWalesStrategy,2015).

Figure1.LocationsofPortBotanycontainerterminals

IndependentPricingandRegulatoryTribunal’s(IPART)Report(IPARTReport,2008)inMarch2008highlightedan inefficiencyof the landsideoperationsatPortBotany. Inresponse to thisreport NSW Government has established the Port Botany Landside Improvement Strategy(PBLIS),which is led and coordinated by Sydney Ports and nowby Transport forNew SouthWales (TfNSW). Themain goal of the PBLIS is to improve the competitive access and servicearrangements of container movements between stevedores and transport carriers at PortBotany.

ThePIBLISRegulationprovidesperformancestandardsrelatingtoaccessbyroadcarrierstothePortBotanyContainerTerminals, theperformanceof roadcarriersat those terminalsand theperformanceofstevedoresinprovidingservicestoroadcarriersatthoseterminals.TheTruckMarshalling Area (TMA) has been developed as part of expansion of infrastructure in PortBotany; TMA is a component of the Operational Performance Measures framework and isfundamental tomoving vehicle congestion off public roads; and to provide a safe and securearea,forearlyarrivingtruckstobestagedbeforetheyareservicedbythestevedores.

2.1PBLIS–operationalperformancemeasures(OPMs)andmandatorystandards

OperationalPerformanceMeasures(OPMs),definedinPBLISwith industryfinancialpenalties,commencedonMonday28February2011.Logisticchainparticipantsarepenalized,bypayingafinancialpenaltytotheotherparty,forperformancethatfailstomeettheOPMstandards.TheOPM standards that aremeasured for truck carriers include: Early Arrivals, Late arrivals, NoShowsandCancellationofbookings(listings).Theoperationalperformancestandardsthatwillbe measured for stevedores include: Minimum Number of Slots Offered per Hour, TruckTurnaround Time, Truck Non Service and Time Zone Cancellations. The OPS integrates thestevedores’ processing data and truck tracking data, to provide an independent andcomprehensive data record of the operations of the landside interface. PBLIS Mandatorystandardscover(TransportforWales,2012):

Carrier Mandatory Standards, which sets mandatory standards regarding CarrierperformanceinrespectofaccessbytheirTruckstotheland‐basedfacilitiesandservicesattheTerminalsatPortBotany;

Stevedore Mandatory Standards, which sets mandatory standards which apply toStevedoresinrespectoftheoperationandprovisionofland‐basedfacilitiesandservicesattheirTerminalsatPortBotany;

Regulation of Charges, which regulates the extent to which a Stevedore may imposecertaincharges, includingby increasingcertaincharges, inrelation to theoperationorprovisionofland‐basedfacilitiesandservicesatitsTerminal;

DeterminingCertainMatters (Definitions) such as:whenTruckArrives,when aTruckjoins,orfailstojoin,aServiceLine,TruckTurnaroundTime,MinimumNumberofSlots.

RecordsandInformation,whichcontainsdirectionstoStevedoresregardingthekeepingofrecordsandtheprovisionofinformationbyStevedoresandCarriers

Invoicing of Financial Penalties, which prescribes certain matters with respect to theinvoicingofFinancialPenaltiespayableundertheRegulation.

2.2 Mandatory standards and penalties related to DP World Sydney – Port BotanyTerminal

PBLIS Performance Standards for Truck Carrier and Stevedore performance standards arepresentedintable1andtable2respectively.Operational Mandatory Standards also include regulations, rules and definition of terms andconditionssuchas(TransportforWales,2012):

All truck carriers visiting Port Botany are required to register their details with theTransportforNSWCargoManagementControlCentre(CMCC).

AlltrucksservicingPortBotanymustbefittedwithanRFIDtrucktrackingtag.Thistagisused to capture the movements of each truck when in the Port Botany precinct andrecordsarrivaltime,queuetime,TTT,thetimetakentobeservicedbythestevedore.

TheStevedorenotifiestheCarrierthataContainerisavailableforcollection. The Stevedore notifies the Carrier less than 4 hrs prior to the commencement of the

TimeZone inwhich theBooking is scheduled tooccur that theContainer isno longeravailableforcollection.TheCarriermakesaBookinginrespectofthatContainer.

Table1.PIBLISPerformanceStandardsforTruckCarrier

TruckCarriersPerformanceStandard

Description IndustryFinancialPenalty

1)EarlyArrival

Notruckstoarrivebeforebookedtimeslot.Trucksmaybeacceptedintotheterminalatthestevedore’sdiscretionwithoutpenalty.TruckTurnaroundTime(TTT)appliesfromstartoftimezone

$100pertripPayabletothestevedore

2)Latearrival

Notruckstoarriveafterbookedtimeslot.Trucksmaybeacceptedintotheterminalatthestevedore’sdiscretionwith“LateArrival”penaltiestoapply.Iftruckisnotacceptedbythestevedorethena“NoShow”penaltywillapply

$50perslotPayabletostevedore

3)Noshowandextendedlate

Truckfailstoarrivewithinbookedtimeslotandisnotacceptedbythestevedore

$100perslotPayabletostevedore

4)Cancellationofbookings

Restrictionsandpenaltiesapplytocancellationofbookingsforaslotwithin24hoursofthecommencementofthattimezone

Lessthan12hoursnotice,penaltyis$50perslot.Between12–24hoursnotice,penaltyis$50perslotunlessslotisbookedbyanothercarrierPayabletostevedore

Table2.PIBLISStevedoreperformancestandards

StevedorePerformanceStandards

Description IndustryFinancialPenalty

TruckTurnaroundTime(TTT)

GateIntoJobComplete50minutesforfirstcontainerplus10minutesperadditionalcontainer

PayabletothetruckcarrierIncrementsof$25pereach15minutesIfatruckisaffectedbypoorterminalperformanceduringthespecifiedperiod,(forsubsequenttripsthesametruckmaybeeligibleforextendedarrivaltime(uptothesamenumberofminutesthatthetruckwaspreviouslydelayed).

Minimumslotsoffered

Minimumof50slotsperhourmustbeoffered24/7(min1200slotsperday)Stevedoreshavetheflexibilitytooffer,andarecurrentlyofferingmorethantheminimumnumberofslots

Stevedoreissubjecttoinfringement(fine)andprosecution

Trucknon‐servicing

Stevedorefailstoserviceatruckthathasaslotbooking.TruckTurnaroundTime(TTT)alsoappliesStevedoremustprovideareplacementslotwithin24hours

$100perslotplusTTTpenaltiesapplyPayabletothetruckcarrier

Cancellationoftimezones

Closerthestevedoregetstothetimezonethemorerestrictionsapply

Greaterthantwohoursnotice,penaltyis$50perslotLessthantwohoursnotice,penaltyis$100perslotPayabletothetruckcarrier

ForeachTruckthatArrivesataStevedore’sTerminalpursuanttoaBookingandforthepurposeofreceivingTruckServicesaStevedoremustperformtheTruckServicesinfullwithintheapplicableTruckTurnaroundTime.

AStevedoremustnotprescribeaTimeZonewhichislessthan60minutes. AStevedoremustnotcancelanentireTimeZoneunlessitisduetoanUnforeseenEvent

orisnecessarytodosotoaddressreasonableconcernsregardingthesafetyofapersonorpersons.

A Carrier must not cancel a Booking for a Slot less than 24 hours prior to thecommencement of the Time Zone in which that Booking occurs Stevedore mustimmediatelymakeacancelledSlotavailabletoallCarriersforBooking.

IfaCarriercancelsaBookingafterthecommencementof theTimeZoneinwhichthatBookingoccurs theCarrierwill fail tocomplywith themandatorystandardrelating tothearrivalofTrucks.

Each Stevedoremustmake available no less than theMinimumNumber of Slots eachHour,24hoursaday,inrespectofwhichallCarrierscanmakeBookings.

TheMinimumNumberofSlotstobemadeavailablebyaStevedoreeachHourmustbemade available by that Stevedore for Bookings at least 2 Working Days prior to thecommencementofthatHourunlessithasreceivedthepriorapprovalofTfNSWtomakeoneormoreofthoseSlotsavailableforaperiodthatislessthan2WorkingDayspriortothecommencementofthatHour.

A Stevedore and, if applicable, its Vehicle Booking System (VBS) must not make aBooking,oracceptaBooking, foraContainer tobe loadedorunloadedontoor fromaTruck at that Stevedore’s Terminal unless that Booking has been made through thatStevedore’sVBS.VBSisaweb‐basedonlineslotbookingsystemdesignedforfacilitiestoorganisethereceiptanddeliveryofshippingcontainers.TruckCarriersutilisetheVBStocreatebookingsinanyofthe24TimeZones(perday).

EachStevedoremustcreate,collectandretaintherecordsanddataspecifiedOPMandeachStevedoremustprovidetoTfNSWtherecordsanddataspecifiedinOPM.

Financial penalties are issuedvia the stevedores’ invoicingprocess forwhich they areresponsible. Stevedores send an invoice to truck carriers that have not met OPMstandards detailing penalties they owe. Stevedores are also responsible to self‐invoiceforfinancialpenaltiestheyowetotruckcarriers.

TheseinvoicingprocessesaremonitoredandauditedundertheCMCC.

Financial penalties are issued via the stevedores’ invoicing process for which they areresponsible. Stevedores send an invoice to truck carriers that have not met OPM standardsdetailing penalties they owe. Stevedores are also responsible to self‐invoice for financialpenaltiestheyowetotruckcarriers.TheseinvoicingprocessesaremonitoredandauditedundertheCMCC.IftheCMCCfindthatapartyhasnotissuedapenalty,issuedapenaltyincorrectlyorhasnotpaidapenalty,theymaybeliabletoaninfringement.

3.DPWSYDNEY–PORTBOTANYLANDSIDEIMPROVEMENTSTRATEGYFOLLOWINGPBLISREGULATIONS

In response toPBLISregulationsDPWorldPortBotany terminalhasamended theoperationsprotocols, KPI and practices with increased focus on landside operations and balancing andimprovingefficiencyofboth landsideaswellaswatersideoperations.PriorPBLISregulationsthe terminal was not financially penalised for failing tomeet expected landside performancelevel and during period of restricted resources availability, the focus was usually shifted towaterside operations, as the penalties would be imposed by vessel operator if vessel failedbehindagreedtarget.Closemonitoringoflandsideoperationsbecomesimportantaspenaltiespaidtotruckcarriers,aspresentedinFigure2,canquicklyrisetosubstantialamounts.

ServicingtruckswithinPBLISmandatorystandardsisessentialtaskforlandsideoperationandTruckTurnaroundTimeiscalculatedforeachtruckbasedonnumberofcontainerstransactionsbookedforthetruckvisit(Figure2)andbasedKPIistokeepTTTbelowthestandard.

TruckTurnaroundTimeiscalculatedunderPBLISregulationfromthetimetruck“QueueIn”atService Line, rather than at “Gate In”, to the time truck leaves the terminal ‐ “GateOut” time.ServiceLanestartsabout800mfromtheterminalgateandthereforeDPWSydneyterminalhasadjustedrespectiveKPIthatincludestimetruckspentinservicelaneoutsidetheterminalgates.

Figure2.DPWorldSydney–NetCumulativePenaltiesandMandatoryStandardsTTTinrelationtonumberofcontainertransactionpertruckvisit

4.CONCLUSION

TheOperationalPerformanceSystem(OPS)integratesthestevedore’sprocessingdataandthecarriers’ trucktrackingdata,toprovideanindependentandcomprehensivedatarecordoftheoperations of the landside interface. This allows PBLIS to comprehensively analyze theperformanceof the landside interface, includingbooking information, carrier listings, etc.Thedatawillalsobepossibletoviewatthecompleteindustryleveloratthesingleoperatorlevel,acrossall timeperiods i.e. timezones,days,weeks,monthsoryears.This informationprovidethePBLISteamwithtechnicalinsightsastohowefficientlyandconsistentlyallthecomponentson the landside interface are performing. Then, this knowledge can be used to improve theefficiency,consistencyandtransparencyofthelandsideinterface(TransportforNSW,2015).

TheregulationsintroducedinDPWorldSydneyappeartocontinuetohavepositiveoutcomes,particularlyintermsoftruckturntimesandcarrierarrivalpatterns.Asignificantimprovementsarereachedcaneasilyrecognizediflookattrucksarrivalsscheduleshowingthatalmostthereisnovehiclesmorethan1minutelate.

REFERENCES

[1] Ferreira,L.,Bunker,J.,(2003).TheRoleofGovernmentsinImprovingFreightLogisticsinQueensland Working Paper 1: A Review of Evidence, Transport Research @ QUT,Technical report, web page: eprints.qut.edu.au/2768/1/2768.pdf (last accessed10.04.2015)

[2] Transport for New SouthWales, web page: freight.transport.nsw.gov.au/ (last accessed12.03.2015)

[3] AustralianGovernment,DepartmentofInfrastructureandRegionalDevelopment,StateofAustralianCities2012,Chapter3,Productivity,page68

[4] Transport for New South Wales Strategy, Action 1H, web page:freight.transport.nsw.gov.au/strategy/action‐programs/network‐efficiency/action1h.html(lastaccessed10.04.2015)

[5] IPART Report (2008), Reforming Port Botany’s links with inland transport, web page:www.sydneyports.com.au/data/assets/pdf_file/0019/7273/NSW_Government_response_to_IPART_Report_Full_report.pdf(lastaccessed10.04.2015)

[6] TransportforWales,portBotanyLandsideOperations;MandatorystandardsunderPart3ofthePortsandMaritimeAdministrationRegulation2012.

[7] Transport for NSW, Cargo Movement Coordination Centre, Weekly Web PerformanceReports(2015).webpage:freight.transport.nsw.gov.au/(lastaccessed10.03.2015)

IDENTIFICATIONOFCONTAINERHANDLINGPROCEDURES

DimitrisFolinas,DimitrisAidonis,IoannisMallidis,MarilenaPapadopoulou

DepartmentofLogistics,TEI‐CM,Kanellopoulou2,Katerini,Greece

Abstract:Considering thatcontainer shipsrepresent themostdominant transportationmode inglobalized supply chain networks, the identification and protypation of effective containerterminals’ management processes is critical for the reliable and efficient operation of thesenetworks.Furthermore,theidentificationandprotypationofcontainers’handlingproceduresisacritical success factor for the development of sea transportations and especially the containertransportationwhichisgrowinginglobalmaritimebusiness.Underthiscontext,thepurposeofthispaper is two‐fold: first, to identify the key container management processes, second, theappreciationof thenecessary ICT systems that shouldbeusedbyany terminal’smanagement inordertoprovidegeneralguidelinesforestablishingeffectivecontainermanagementoperations.

Keywords:Maritimelogistics,Containershandling,Process/logisticsstandardization.

1.INTRODUCTION

Improvingtheefficiencyofcontainerterminals(CTs)iscriticalforthedevelopmentofeffectivedoor‐to‐door supply chain networks, as these constitute the start point of responsive andresilient hinterland connections. On this basis, the optimization of their operationsmay havesignificant value added effects throughout thewhole supply chain. As the optimization of CToperations can be mainly achieved through the optimization of container managementprocedures and the employment of Information Technology (IT) systems, the purpose of thispaper is to map: (i) the critical container management processes, and (ii) the necessary ITsystems that should be employed by the port’s management, in order to develop generalguidelinesforthedesignofeffectivecontainermanagementoperations.

2.CONTAINERTERMINALOPERATIONS

Thischapterwillprovideananalysisofthecontainermanagementoperationsassociatedto:(i)conventional cargo containers, (ii) empty containers, and (iii) hazardous cargo containers astheseare,depictedinthefollowingFigure1:

2.1ConventionalCargoContainerManagement

The management of conventional cargo containers represents a range of 50‐80% of theterminal’s totaloperationsand thus, theireffectivemanagementmaysignificantly improve itsoverall performance. These operations are classified into the operations associated to theirexportsandtheirimports.Theoperationsassociatedtotheirexportsarethefollowing:

Pre‐gaterecordingofthetruckandcontainer’sweight.

Pre‐gateconfirmationofthecontainer’sshippingandcustomsrelateddocuments. Pre‐gaterecordingofthecontainer’sexportinformationintotheTerminal’sITsystem. Physical inspectionof thecontainer(inspectionof itsstamp,physicalconditionetc.)at

theterminal’sgate. Consistentcheckingofwhetherthetruckandcontainer’sidentitycomplywiththepre‐

checkstage. Indicationtothetruckdriver,ofthecontainers’dischargepositionwithintheterminal’s

interchangeareaaccordingtocontainer’ssize/type/height,departurevessel,weightandportofdischarge(Iftheterminalreceiveshoweverinstructionsfromthecustomsofficeto scan the container, the containerpasses through the scanner first and itsdischargepositionisthenindicatedtothetruckdriver).

Issuanceofthecontainer’sdischargeandstoragecommandanditstransmissiontothestraddlecarriers(ortootheralternativeequipment).

Determinationoftheexportcontainership’sberthinglocationattheterminal’squay. Development of a specific container sequence, based on the order that these will be

loadedontotheship(thatisbasedontheMasterBayPlan). Weightingoftheemptytruckattheexit(AqabaContainerTerminal,2015).

Figure6:ContainerTerminalOperations

Insomecontainerterminals,itsmanagementalreadyhaspreliminaryinformationonitsarrivalandhasthusdetermineditsstackingposition.Moreover,andIftheterminal’smanagementhasinstructionsfromthecustomsofficeforscanning,thecontainerpassesthroughthescannerfirstandisthendischargedatthedesignatedplace.Specificallyinthecaseofthecontainer’sphysicalinspection, and under the orders of the customs office, the terminal’s staff can assist theperformanceofphysical control (checks)of the cargo in the container.Then, the ship loadingorderisissuedandcountersignedbythecustomsoffice.TheTerminal’soperationsassociatedtotheimportcontainersarethefollowing:

Receiptoftheship’simportmanifest. Berthallocationtothevessel. Dischargeofthevessel’scontainersonthequayandfurtherinspectionofthecontainers

inordertoidentifywhether:(i)eachcontainer’snumberiscorrect,and(ii)itsstampisnot infringed (The scanning of a container after the Customs Office’s request may beperformedevenduringshipdischargeoperationsorsubsequently.

Electronicrecordingofthecontainer’sreceipt. Generation of a delivery sequence list based on the stacking of the container in the

containeryard. Deliveryofthecontainerstotheterminal’scontaineryard. Inthecaseofpartialcargoloads(groupage),theirreceiptisconductedbytheterminal’s

warehouse, based on the mailing list supplied by the cargo’s agent and after theirinspectionwithrespecttotheirbrands,andtheirquantitystatedatthebilloflading.

Afterthecontainers’andthepartialcargoarereceived,an"ExecutionReport"isformedandpassedtothecargo’sagentandtheCustomsOffice.

When the importer’s agent carries out customs formalities and receives the deliveryorderofeachcontainer,sealedbythecustomsofficeforrelease,thecargoisdeliveredtothePortAuthority.

TheimporterthenpaystheGeneralPortandWarehouseRights,andaninvoiceisissuedbythePortAuthority.

Theterminal’sdocumentationstaffreleases"hold"inEXPRESSforimportdeliveryafterrelevantpaymentiscleared.

DocumentationstaffupdatestheweightandB/LnumberinEXPRESSandscansacopyofB/LintoadesignatednetworksharefolderwiththeB/Lnumberasfilename.

Thetruckthatwillreceivetheimportedcontainergoesthroughaweighbridgeinordertoobtainaweightticketthatshowstheweightofemptytruck.

Truckdriverpresentsitsentrypermittothepre‐gatestaff. Thepre‐gatestaffthenprintsa"gatepass"tothetruckdriver. The in‐gate staff checks the consistency of the truck number with that on the entry

permitandgatepassandmakesagate‐in. Theyarddispatcherdispatchesequipmenttoloadtheimportcontainerontothetruck. Theweighstaffwillweighthetruckwiththecontainerandprinttheweightonthesame

weightticket. The pre‐out gate staff will check theweight of the container andcompare it with the

scannedB/Linthenetworkfolder. ThetruckwillbesealedbyCustomsanditsentrypermitwillberecorded. Out‐gate staff will compare the truck and container number with thaton the entry

permit. Truckhastowaitifitisoverloaded,orelseitdeparts(AqabaContainerTerminal,2015).

Forbothimportandexportcontainersthatemployrailtransportation,theterminal’soperationsarethesameasthosefortrucks,withthedifferencethatwenowhavewagonsinsteadoftrucks.In some container terminals, after the container discharge from the ship, the port issuesdocuments for the customs cffice and the agents of the imported containers, namely: (i) TheGeneralAct,(ii)TheNotificationActfordifferencesinseals,and(iii)TheContainerInterchangeReport.Moreover,theScanningofacontaineratthecustomsoffice’srequestmaybeperformedeven during ship discharge operations or subsequently. Customs‐cleared containers can beshippedontrucksorcanbestripped.Loadingoffullcontainersontrucksforfurthershippingisdoneonthebasisofpreviouslysubmittedbytheagentnumbersofcars.Finally,atthecontainerreleasestage,thenumbersandtheintegrityofsealsarechecked.

2.2EmptyContainerManagementProcedures

Therequiredproceduresassociatedtotheexportofemptycontainersarethefollowing:InGate–gatestaffwillcheckthecontainer’scondition,InGate–staffwillchecktheagent'srequestnoteand entry permit,andregister it into the system, Yard staff will inform the empty containerhandler to unload thecontainerto the projected position according to the line operator,thecontainer’ssizeandtype,OutGate–gatestaffwillconfirmthatthecontainerisdischarged,and

stampontheentrypermit.(AqabaContainerTerminal,2015),andtheemptycontainerswillbethen stored in the Terminal’s empty container slots until loading on the ship. The requiredproceduresassociatedtotheimportofemptycontainersarethefollowing:Theexportersendstruckstocollecttheemptyimportedcontainers.Theexporter’sagentsendsanEDImessagetothePort’smanagement indicatingthenumberofemptycontainershe intendstoreceive.Afterrelevantpaymentiscleared,theTerminalapprovestheemptycontainer’srelease,andtheyarddispatcherloadsitonthetransportationmodeemployedforitsrelease.

2.3HazardousCargoContainerManagementProcedures

Hazardous cargo can be classified to the following categories: Petroleum in accordancewithAnnex I of the International ConventionMarpol 73/78, Gases as defined inGC Code,Noxiousliquid substances/chemicals, includingwaste, asdefined in theBCHCode andAnnex II of theInternational Convention MARPOL 73/78, Solid bulk cargoes which exhibit chemical risk inaccordance to the BC Code, Harmful substances in packaging covered by Annex III of theInternationalConventionMARPOL73/78, andPackedhazardousgoods: substances,materialsandarticlesasdefinedintheIMDGCode.Thepackagingandtransportunitofhazardouscargoshould carry warning labels, depending on the category of the cargo’s risk, in order tocommunicatethepotentialdangersthroughoutallstagesofthesupplychainandthushandledaccordingly. The empty uncleaned packages should be also treated as hazardous cargoes,sufficientlypurgedbyresiduesoftheirdangerouscargoandevacuatedfromhazardousvapors.

2.3.1LegislationforHazardousCargoes

The management of hazardous cargo should cope to the following international legislativeframeworks: International Maritime Dangerous Goods Code (IMDG Code), International BulkChemicalCode(BCHCode):InternationalCodefortheconstructionandsupplyofshipscarryingdangerouschemicalsinbulk,InternationalBulkGasCarrierCode(GCCode):InternationalCodefor the construction and equipment of ships carrying liquefied gases in bulk, InternationalConvention for Safe Containers (CSC Code), International Ship and Port facility Security Code(ISPS Code), International Convention for the Safety of Life at Sea (SOLAS) and InternationalConventionforthePreventionofPollutionfromShips73/78(MARPOL73/78).

2.3.2ImportantIssuesonHazardousCargoesforPortAuthorities

Theportauthoritiesmustbeinformedinadvanceregardingthecharacteristicsofthehazardouscargo.Specifically,theshippersorconsigneesmustsubmitthefollowinginformationtothePortAuthority. Specifically about: Technical name (Proper shipping name), Class and subsidiarydanger,UNNumber,PackingGroup,Numberandkindofpackages,Totalquantityofdangerousgoods, Auxiliary descriptors such as : Marine pollutant, elevated temperature, etc. Specificinformation for classes: 1, 6.2 and 7, for some cargo types of classes 4.1 and 5.2 and forfumigatedunits(Minimumflashpoint,ifthetemperaturedropsbelow≤61Celsius,indicationofuncleanedpackagingthatcontaineddangerousgoods,andanyinformationthatisnecessaryforthesafehandlingofthecargointheterminal’sarea.

2.3.3GeneralPrinciplesfortheManagementofHazardousCargoes

Themanagementofhazardouscargorequirestheadoptionofmeasuresforitssafeloadinganddischarge.Suchmeasuresinvolve:

Identification of specific positions for the ship’s discharge and loading, which furthermeetssafetydistancerequirementsfromsitesofindustrialfacilities,residentialregions,shipbuilding areas, and anyotherwork site that couldbe affectedby thesedangerouscargoes.

Safemooringofshipsatpiers. Responsive information sharing on the dangerous cargoes that will be loaded /

dischargedorremainintheharborarea. Emergency response plan in the case of an explosion, fire, and water or gas or solid

dangerousgoodsleaks. Constant monitoring of the hazardous cargo storage areas along with appropriate

warningsignsonpotentialdangers.

Specifichazardouscargoclassesrequirespecial treatment.Regardingexplosivecargoes(Class1)thefollowingshallapply:1)Theexplosivesubstancesandcargotypes(otherthanclass1.4s)shouldbeacceptedonly for immediatedelivery/receipt,2)Theirresidence in terminal’sarearequiresspecial facilities,andTheirloadinganddischargingshouldbe:monitoredbyspeciallyauthorized personnel, conductedwith extreme caution,madewith the of use explosion‐proofequipment (e.g. electric forklifts) and the appropriate workers’ gear and clothing. Forradioactive cargoes (Class 7), the following processes should be followed: They should beacceptedonlyforimmediatedelivery/receipt,theirresidenceintheterminal’sfacility,evenfora few hours, requires a special storage area, which must be isolated and not accessible byunlicensed individuals and their loading and discharge should be monitored by speciallyauthorized personnel, and conducted with care, along with the use of appropriate gear andclothing. For temperature controlled cargo, such as automatic ignition substances (Class 4.1),organicperoxides(Class5.2),andsomeinfectioussubstances(Class6.2), thefollowingshouldapply:Thesecargoesshouldbeacceptedonlyforimmediatedelivery/receiptandtheirstoragein the terminal should be made in specific, specially designed areas that incorporatetemperature‐controldevicesfordealingwithexceptionalcasesofhightemperatureincreases.

3.ITSYSTEMSFOREFFECTIVECONTAINERMANAGEMENT

Effective containermanagementprocedures require the installationof Integrated InformationSystems, in order todiffuse thenecessary informationbetween the terminal’s staff. ThemostcriticalITsystems,aspresentedbySxoinakisandKoukouloudi,(2004),arethefollowing:

Centralized Information Management System: This system supports the dynamicmovementofdataand informationbetweenthe terminal’ssubsystems.Thepurpose isthecollectionanddistributionofdatabetweendifferentsources.

OfficialDocumentFillingSystem:Thissystemsupportsthedownloadingprocessofthenecessarydocumentsforthedeliveryandreceiptoperationsofcontainers.

Administrative work system: This system supports the issuance of exit, loading andtransit permits, themaintenance of records and the submission of customs clearancedocuments.

Pricing system: This system supports the estimation, publication and management ofinvoices,withrespecttotheterminal’sstorageandloading/dischargerates.

Entry/exitcontrolsystem:Thissystemsupportstheautomaticcontrolofcontainersanditsvehicleusingasmartcardandbyautomaticallyreleasingthegateafterasuccessfulcross‐checkofthevehicle’snecessarydocumentation.

Control of loading/discharging: This system supports the management ofloading/discharge operations through the employment of wireless terminals, and byelectronicallymonitoringtheloadinganddischargemovementsperformed.

Geographicalinformationsystem:Thissystemsupportsthegraphicalrepresentationofcontainer’smovementswithintheterminalinrealtime.

Inventorysystem:Thissystemsupportstheelectronicconfirmationofthepositionofthecontainerintheterminal’sstowagearea.

Stowage planning system: This system supports the distribution of container’s in thestowagearea.

Equipment Management System: This system supports, via wireless terminals, thestowageplan system,whilepassing thenecessary instructions to the straddle carriersforcontainermovements.

4.SUMMARY

Considering that container ships represent the most dominant transportation mode inglobalized supply chain networks, the development of effective container terminals’ is criticalforthereliableandefficientoperationofthesenetworks.Underthiscontext,thepurposeofthisreportistoidentifykeycontainermanagementprocesses,alongwiththenecessaryITsystemsthat shouldbeusedby any terminal’smanagement inorder toprovide general guidelines forestablishingeffectiveandsustainablecontainermanagementoperations.

ACKNOWLEDGMENT

This research has been co‐financed by the Joint Operational Programme “BLACK SEA BASIN2007‐2013” / Priority: “Supporting cross border partnerships for economic and socialdevelopmentbasedoncombinedresources”.SEcuringTRansitCONtainers(SETRACON).

REFERENCES

[1] AqabaContainerTerminal.(2015),https://www.act.com.jo/content/procedures.

[2] VishakaContainerTerminal.(2015).http://www.vctpl.com/images/oppd.pdf.

[3] Vlachos, D., Tsitsamis, D., Iakovou, E., & Georgiadis, P. (2006). A MethodologicalFramework for the Strategic Management of Port Container Terminals. LogisticsOrganization.

[4] Sxoinakis, M., & Koukouloudi, R. (2004). A Standard Management System of ContainerTerminalinTh.P.A.2ndInternationalConferenceonTransportResearch.

[5] Waterschoot, K. (2011) Antwerp port system http://www.wcoomd.org/en/events/eventhistory/2011/~/media/2858CC3E9403462AB0A6FA5459B39541.ashx

[6] International Navigation Association (2000) Dangerous cargoes in ports. Brussels,Belgium.

OPERATORSCHEDULINGUSINGMIN‐CONFLICTSANDTABUSEARCHBASEDHEURISTICINCONTAINERTERMINALS

VladoPopovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Inthispaperanewapproachtotheoperatorschedulingproblemincontainerterminalsispresented.Theproblemisrelatedtoassigningoperatorstohandlingequipmentwhilesatisfyingrequirementsforoperatorsovershiftandworkregulations.Itisconsideredinacaseofdisruption,when it is impossible to make a schedule so that all requirements for operators and workregulationsare satisfied. In that situation the leastbad solution isneeded.For thatpurpose theproblemisconsideredlikeamaximumconstraintsatisfactionproblem(MaxCSP)andsolvedbyaheuristicbasedonMin‐conflictsheuristicandTabusearch.

Keywords:operatorscheduling,containerterminals,Min‐conflictsheuristic.

1.INTRODUCTION

Resourceschedulingincontainerterminalsisaveryimportantactivityduetoagreatimpactontheefficiencyandthequalityofservice,andthusontheterminalbusiness.Itisespeciallyrelatedtoschedulingofhandlingequipment,butalsotoworkers(operators)whooperatewiththem.Infact, operators are highly qualified workforce and cause a big expense for terminals. On theotherhand,forthesakeoftheirsafetyandsafetyofgoods,theirworkisregulatedbynumerousstrictworkregulations.

Kim at al. (2004) pointed out that operators are usually scheduled after determining theirnumberineachshiftandschedulingoftheoperationtimeoftheequipment.Accordingly,theydefinedoperatorschedulingastheproblemofassigningthecertainnumberofoperatorstotimeslots of equipment usage while satisfying the requirement for the number of operators in aspecific time slot, various regulations for workforce utilization, and the requirements ofoperators forassignments.Considering thatdefinition, theproblemcorresponds to aCSPandKimatal.(2004)solveditusingILOGSolver.However,inthatcasethepossibilityofobtainingaschedule that does not meet all regulations and requirements is eliminated. Consequently, ifsomedisruptionsoccursuchasanunplannedoperatorabsenceoranincreasedworkload,therewouldbeachancethatnonschedulecouldbemadei.e.nonsolutioncouldbefound.

For that reason in this paper operator scheduling is considered as the problem of assigningoperatorstotheequipmentwiththeaimofsatisfyingasmanyrequirementsforoperatorsandwork regulations as possible. That means that the sum of intervals in which the requirednumberofoperators issatisfiedandsatisfiedworkregulations ismaximized,wheresatisfyingeachworkregulationareobservedforeachoperatorseparately.Accordingtothat,theproblemcorrespondstoaMaxCSP,becauseitsaimisactuallymeetingasmanyconstraintsaspossible.In

ordertosolvethatMaxCSP,aheuristicbasedonMin‐conflictsheuristic(MCH)andTabusearch(TS)isdeveloped.

Therestofthepaperisorganizedasfollows.Inthenextchaptertheproblemisdescribedandrelated literature is given. The third chapter is dedicated toMCH, whereas the fourth to thedeveloped heuristic. The fifth chapter presents the result of the application of developedheuristic to anoperator schedulingproblem in theport container terminal, and the last givesconcludingobservations.

2.PROBLEMSTATEMENTANDLITERATUREREVIEW

Environment in which the problem is considered involves the following: 1.) The handlingequipmentisrelatedtomachinessuchascontainercranes,yardcranesandyardtrucks;2.)Theshiftiscomposedofaknownnumberofonehourtimeslots.3.)Foreachpieceofequipmentineachtimeslot isdeterminedwhether it isusedornot,basedonthescheduleof theoperationtimeofequipment.4)WorkregulationsarethesameasthoseinthecontainerterminalinPusan(RepublicofKorea)andtheyarepresentedinKimatal.(2004).Theyspecifytheminimumandmaximum operating time allowed per shift (S, L), the minimum and maximum consecutiveoperating time (session) allowed for every operator ( , ), types of equipment that can beassignedtoeachoperator,theminimumlengthoftherestperiod(F),necessityfortherestaftera session. They also involve disallowance of assigning one operator to multiple pieces ofequipment in one time slot. 5.) Given the great complexity of the problem, in this paper anassumption with the aim of the problem relaxation is introduced. It is assumed that eachoperatorcanhandleanytypeofequipment.Accordingly,typesofthepiecesofequipment,whichworkinaspecifictimeslotarenolongerimportant,butthetotalnumberofthem.Thatnumberrepresentstherequirementforoperatorsinaspecifictimeslot.

Considering these characteristics of theproblem, the input data for theproblem are the totalnumberoftimeslotsintheshiftN,thetotalnumberofpiecesofequipmentM,thetotalnumberofpiecesofequipmentwhichworkbytimeslot ( 1,2, … , ),thetotalnumberofoperatorsintheshiftRandparametersintheworkregulations.Theoutputoftheproblemsistheschedulethatdefinesforeachoperatorwhenandonwhichpieceofequipmentwork.

It is adopted that it is equally important to satisfy every requirement foroperators andworkregulation,exceptof the lastregulationthattakesprecedenceoverallothers. Inmathematicalterms the problem consists of constraints that represent the work regulations and therequirements for operators. Here, all constraints are expressed by variable formulated as ( 1,2, … , ; 1,2, … , ) whose value r ( 0,1,2, … , ) represents the operator numberassignedtotimeslot ionequipment j(thedecisionvariable);and 0whennooperatorisassignedtotimeslotionequipmentj.

Inadditiontothepaper(Kimetal.,2004),LegatoandMonaco(2004)dealtwiththisproblemtoo.Theysolveditforalongerperiodinwhichtheydeterminedaysofffortheoperators,andforashorterperiodinwhichtheyassignoperatorstotheshifts,tasksandresources.However,theoperatingrulesapplyonlyindeterminingthedaysoff,freetimebetweenshifts,etc.Murtyetal.(2005) dealt with determining the number of workers by shifts in the horizontal transportprocesses,whilerespectingtheruleofbreakduringtheshift.Zampelliatal.(2013)carriedouttheassignmentofgroupofworkerstothecranesduringaspecifictimeperiod.Operatorsgroup(gang)wasviewedasaworkforce"consumed"bythecrane,exceptduringthebreakperiod.

3.MAXCSPANDMIN‐CONFLICTSHEURISTIC

Generally,CSPisaproblemthatinvolvesasetofconstraintsandasetofvariables,whereeachvariablehaseachowndomainofvalues. It issolvedbyusingvarious typesofsearch,and the

solutionisfoundifeachvariablehasavalueandallconstraintsaresatisfied.ACSPbecomesaMax CSP if a solution which does not satisfy all constraints is acceptable. In that case theobjectiveoftheproblemsolving,andhenceofthesearch,isfindingasolutionwhichsatisfiesasmanyconstraintsaspossible.According to that,methodsof localsearcharenaturallysuitableforMaxCSP,becauseMaxCSPcanbeconsidered likeoptimizationproblemwithoptimizationobjective of minimizing the number of violated constraints. Therefore, most local searchalgorithms forCSPcanbedirectlyapplied toMaxCSP, since theirevaluation functiondirectlycorrespondstotheoptimizationobjectiveofMaxCSP(Rossiatal.,2006).

Probably thebestknownandmostwidelyusedmethodof localsearch forsolvingCSP(henceMaxCSP)isMCH,duetogoodresultswhichareobtainedbyitsapplicationtothecomplexCSPs.Particularlygoodresultswereobtainedforthewell‐knownproblems:"TelescopeProblem"and"N‐Queen Problem" (Russel and Norvig, 2003). It represents an iterative improvementalgorithmwhichchangesthevalueofarandomlyselectedvariablelocatedintheconflictsetineach iteration. The conflict set consists of variables present in the constraints violatedby thesolution of the previous iteration. Each variable in the conflict set has a certain number ofconflicts,whichisequaltothenumberofviolatedconstraintswhichcontainaspecificvariable.Atotalnumberofconflictsisobtainedbyaddingupconflictsofeveryvariablefromtheconflictset and being updated alongwith the conflicts set after each iteration. The new value that isassigned to the chosen variable should provide a minimum total number of conflicts incomparisontoothervaluesfromthevariabledomain.Ifmorethanonevaluefromthedomainhas this feature, the choice between them is randomagain. The searching process is stoppedwhenthevaluesforallvariablesarefoundsothatthetotalnumberofconflictsisequaltozero,i.e.allconstraintsaresatisfied,orafterreachingapredefinednumberofiterations.

Like most iterative improvement methods, MCH can get stuck in a local optimum of theunderlyingevaluationfunction;andthereforeitisessentiallyincomplete(Rossiatal.,2006).Inorder to overcome that problem the rules such as “restart” or “random walk” or somemetaheuristics are usually used.Ametaheuristic that is oftenused in a combinationwith theMCHisTS.Ithelpthesearchprocessescapefromalocaloptimumbyforbiddingmoveswhichtakethesolution,inthenextiteration,topointsinthesolutionspacepreviouslyvisited(hence"tabu").That is implemented by putting tabu on each new pair of variable‐value, for a fixednumberofiterations.

4.HEURISTIC

Forproblemsolving,aspecialheuristicbasedonMCHandsupportedbyTSisdeveloped.MCHischosenbecauseitisgenerallyoneofthebestmethodsoflocalsearchforCSPs,butalsoduetothefeaturethattendstoprovideasolutionthat is the leastdifferent fromthe initialone.Thiscanbeusefulfortheproblemdiscussedinthispaperifthescheduleismadeadaybefore.Min‐conflictsheuristicisthereforedescribedasa"repair"or"online"technique(RusselandNorvig,2003).

Variable formulated as (i=1,2,…,N; r=1,2,…,K) is introduced for the implementation of theheuristic. Its value represents the number of constraints in time slot i which are unsatisfiedbecause of an operator r, i.e. the number of conflicts that the operator r has in time slot i.Variable increasesitsvaluebyoneiftheoperatorrisassignedtoapieceofequipmentinthetime slot i, i.e. value r is assigned to one of the variables (j=1,2,…,M), and it causes someconstraintstobecomeunsatisfied.Also,variable increasesitsvaluebyoneiftheoperatorrisnot assigned to any piece of equipment in the time slot i, and it causes some constraints tobecomeunsatisfied.Ifoneoftheseconstraintsrelatestothelastregulationitisdefinedthatthecorresponding variable enormously increases its value.Hence, it is impossible for it to beviolated in a solution. Therefore, the number of conflicts of each value r in each time slot isconsideredinthiscase,andnotthenumberofconflictsofeach variable.Thisisdonedueto

thefactthatallconstraintsregulateoperatorsworkintime,anditiseasiertomakealgorithmoftheheuristicwithvariabledefinedas .Though,inthiswayaslightlydeviationofMCHismade.

Because of using TS, variable formulated as (i=1,2,…,N; r=1,2,…,K) is also involved in theheuristic.Itdemonstratesthevalueoftabuforthecorresponding variable.Thevalueoftabuisassignedto variablewhenoneofthevariables (j=1,2,…,M)obtainsanewrvalue.Duetothat,thenumberofconflictsthattheoperatorrhasinthetimeslotiisnotconsideredduringthenextiterationswhosenumberisequaltothevalueoftabu.

The end of heuristic consists of several steps which attempt to improve the best solutionobtainedbythepartofheuristicbasedonMCH(ifthereareconflicts).Itisperformedthroughexchangingpositions(piecesofequipment)betweenanoperatorrthatcausesconflictsinatimeslotwithotheroperatorswhoalsowork in that timeslot.Thispart is introducedbecause thepreviousimpliesonlychangingachosenvaluer.However,itmayonlybeneededtoreplacethatvaluewiththecorrespondingvaluefromthesametimeslot,inordertoavoidconflicts.

Thealgorithmofheuristicispresentedinthefollowingsubsection.It involves16steps,wherethe part for improving begins from the step 11. In order to facilitate its explanation, threematricesX,YandAareintroduced.ThematrixX(NxM)containsvaluesrofthevariables andrepresents a solution of the problem in an iteration. Matrix Y (NxK) contains values of variables. The sum of the values of all variables represents the number of conflicts that asolutionXhas,andatthesametimethevalueoftheobjectivefunctionintheheuristic.MatrixA(NxK) contains tabu values that are related to variables, i.e. values of variables. Inaddition to the matrices, the sets I = 1,2 ..., N, J = 1,2 ..., M and R = 0,1, ..., K are alsointroduced.

4.1Algorithmofheuristic

1.) Set the globalminimum on a sufficiently big value (bigger than the maximum number ofconflicts),equatetheiterationnumbertozeroanddefinethetabuvalueandmaximumnumberof iterations. Assign value from the initial solution to each variable, or zero if the initialsolutiondoesnotexist, remember that setofvalues (matrixX)as thebest solution.Go to thenextstep.

2.)Increasetheiterationnumberbyone.Iftheiterationnumberissmallerthanorequaltothedefinedmaximumnumberofiterationsthengotothenextstep,otherwisegotothestep10.

3.)Calculatevaluesofall variables.Setthenumberofconflicts in iteration(nci)=maximum(maximum>themaximumpossiblenumberofconflicts)andgotothenextstep.

4.)Among variableschoosetheonewhichhas thebiggestvalue(let itbevariable ), forwhich the corresponding variable =0. If there aremore thanone such variable, choose thefirstandgotothenextstep.

5.)Findvariables ( 1,2, … , )whichhavethervalue.Ifnonehasthatvaluegotothe6astep,otherwisegotothe6bstep.

6a.)Foreachvariable ( 1,2, … , )checkhowmanyconflictswouldhaveasolutionifthervaluewouldbeassignedtothatvariable.Assignthervaluetothevariableforwhichthesolutionhasthesmallestnumberofconflicts(letitbe variable).Ifthatnumberofconflictsissmallerthanthecurrentvalueofnci:remember variableandthervalueasitsnewvalue,forgetthepreviouslyrememberedanditsnewvalue.Gotothestep7.

6b.) Choose the first variablewhichhas thervalue (let it be variable). For eachvalue ∈ \ , check how many conflict have a solution if that one would be assigned to variable.Assignthe valueforwhichthesolutionhasthesmallestnumberofconflictstothe variable (let it be the value). If the solution with a smaller number of conflicts than the

current value of nci is reached: remember that number of conflicts as a new value of nci,remember variableandthe valueasitsnewvalue,forgetthepreviouslyrememberedanditsnewvalue.Gotothestep7.

7.)Iftherearestillsome variableswhichshouldbechecked(whichhavethesamevalueasthe previously checked , and for which the corresponding tabu values are equal to zero),choosethenextandgotothestep5,otherwisegotothenextstep.

8.)Assignthenewvaluetotheremembered variable.UpdatematrixX.Markthe variableastabu,i.e.assignthetabuvaluetothe variable,wheren=newvalue.UpdatematrixAandgotothenextstep.

9.)InthematrixAdecreaseeachvaluebiggerthanzero,byone.Ifthevalueofnciissmallerthanvalueoftheglobalminimumthen:setthenewvalueofglobalminimumtobeequaltothevalueofnci:forgetthepreviouslyrememberedbestsolutionsandrememberthematrixXasthebestsolution.Ifthevalueofnciisequaltothevalueoftheglobalminimum,rememberthematrixXasthesecondfoundbestsolution.Returntothestep2.

10.)Ifglobalminimum=0,markthefirstbestsolutionasfinalandgotothestep16.Otherwisechoosethefirstbestsolutionandgotothenextstep.

11.)Choosethefirst variablewhichhasthevaluebiggerthanzero(letitbe variable)andgotothenextstep.

12.)Among variables ( 1,2, … , ) find theoneswhichhave thervalue,choose the first(letitbe variable)andgotothenextstep.Iftherearenosuchvariables,gotothestep14.

13.)Checkthetotalnumberofconflictswhichisobtainedbyreplacingvaluesbetweenvariable=randeverysecondvariable , ∈ \ .Ifoneofthereplacementsprovidessolution

whichhas thetotalnumberofconflictssmaller thanthevalueof theglobalminimum: set thatnumber of conflicts as value of the globalminimum; carry that replacement out; Update thematrix X; forget the previous improved solution and remember matrix X as the improvedsolution.Ifthevalueoftheglobalminimumisequaltozero,marktheimprovedsolutionasthefinalandgotothestep16,otherwisegotothenextstep.

14.)Iftherearestillsome variableswithvaluesbiggerthanzero,choosethenextandgotothestep12.Otherwisegotothenextstep.

15.)Iftherearestillbestsolutions,choosethenextandgotothestep11.Otherwise,marktheimprovedsolutionasthefinalone, ifexists,andgotothestep16.Ifdoesnotexist,markeachbestsolutionasthefinalandgotothenextstep.

16.)Finishthealgorithm.

5.EXAMPLE

ProblemselectedfortheexampleissimilartothesolvedoneinKimatal.(2004).Itdiffersonlyinthenumberofpiecesofequipmentwhichworkinthelasttimeslot.Itissetthatthreeofthemworkinsteadoftwo.Beforeitssolvingwiththedevelopedheuristic,itwascheckedwhetheritcanbesolvedasaCSPbyusingMicrosoftSolverFoundation(MSF).AConstraintProgrammingSolverwithinMSF did notmanage to solve it,whichmeans that there isn't any solution thatmeetsall constraints.Solving theproblemasaMaxCSP,using thedevelopedheuristics in theVisualBasicenvironment,thesolutionwithoneconflict,i.e.oneviolatedconstraintisobtained.It refers to thework regulation that regulates theminimumnumberof consequent time slotswhich an operator has to operate when starts. The solution, together with the problemparameters, isgiven inTable1. In thesolutionthenumbersarereferredto theoperators, thecolumnstothepiecesofequipment,whereastherowstothetimeslots.Theunderlinednumber

indicatestheoperatorwhoviolatestheregulation, i.e.thervaluewhichgeneratestheconflict.Therefore, the operator three violates the regulation by working on the second piece ofequipmentinthesixthtimeslot.

Table1.Exampleoftheproblemwithsolution

Problemparameters Searchcharacteristics ThesolutionN=6,M=4,K=6,tabu=25, =3, =4,=4, =3, =4,=3,L=4,S=2, =3,=2,G=1,F=2;

Thesmallestnumberofconflictsreachedduringthesearch:2;Totalnumberofsolutionswiththesamenumberofconflicts:8;Theyoccuriniterations:120,165,175,477,584,974,1040,1128;Thefinalsolutionisobtainedthroughimprovingthethirdbestsolutionandhasoneconflict.

143014365236520652140314

6.CONCLUSION

Usingthenewapproach,theoperatorschedulingproblemincontainerterminalsisconsideredasMax CSP. In order to be able to solve it, the heuristic based onMCH andTS is developed,whichisthereforethemajorcontributionofthiswork.Inthiswayitisensuredthattheproblemhas a solution in a situation when it is not possible to meet all work regulations andrequirementsforoperators.Thatsolutionhasthesmallestnumberofunsatisfiedrequirementsandregulations,whichcanbefoundbytheheuristic.Also,inthiswaytheapproachusedinKimetal.(2004)isextended.TheproblemcouldbealsosolvedasaWeightedConstraintSatisfactionProblem (WCSP) by using the developed heuristic. It would only require defining a largernumber of conflicts for the violation of constraints related to themore important regulationsandrequirements.Althoughitisnotappliedhere,thisoptionwouldverylikelybeusefulinthereallifeconditions.

REFERENCES

[1] Kim,K.H.,Kim,K.W.,Hwang,H.,Ko, C.S., (2004).Operator‐schedulingusing a constraintsatisfaction technique inport container terminals.Computersand IndustrialEngineering46(2),pp.373–381.

[2] Legato, L., Monaco, M.F., (2004). Human resources management at a marine containerterminal.EuropeanJournalofOperationalResearch156(3),pp.769–781.

[3] Murty,K.G.,Liu,J.,Wan,Y.W.,Linn,R.,(2005).Adecisionsupportsystemforoperationsinacontainerterminal.DecisionSupportSystems39(3),pp.309–332.

[4] Rossi,F.,VanBeek,P.,Walsh,T.,(2006).HandbookofConstraintProgramming,Elsevier.

[5] Russell, S., Norvig, P., (2004). Artificial Intelligence: A Modern Approach, 2nd Edition,PrenticeHall.

[6] Zampelli, S., Vergados, Y., Van Schaeren, R., Dullaert, W., Raa, B., (2004). The berthallocationandquaycraneassignmentproblemusingaCPapproach.Proceedingsof19thInternationalConference:‘‘CP2013’’,Uppsala,Sweden,pp880–896.

POSSIBILITYOFAPPLICATIONOFREMOTECONTROLLEDORDER‐PICKERFORKLIFTS

MomčiloMiljuša*,DraganĐurđevića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Warehousesandwarehousingprocesses,asasignificantcomponentofthesupplychains,areplaceswherethepossibilitiesofrationalizationarebeingcontinuouslyexplored.Order‐picking,that generates main warehousing costs, is an especially interesting process in terms ofrationalization. As such, order‐picking is a subject of permanent research in the field ofdevelopmentanduseofequipment,design,management,optimization,etc.Oneapproach to thisgoal is increasing the productivity of theworkforce by using remote controlled forklifts in themanuallyrealizedorder‐pickingprocess inorder toreduceorder‐picker’sengagement time.Thispapergivesanoverviewofthistechnologywiththeevaluationofpotentialpointsandthelevelofsavingsinorder‐picker’slabortimeforasetoftypicalorder‐pickingtasks.

Keywords:warehouse,order‐picking,remotecontrolledforklift

1.INTRODUCTION

Withtheaimoftimerationalizationoforder‐pickerengagement,differentkindsofmoreorlessefficient solutions in the technology domain are present, applicable for the purpose ofrationalizationofelementaryorder‐pickers’activities.Someofthemarerelatedto:technologiesof receiving/sending of information about ordered SKUs (Stock Keeping Units) to be picked,selectionoforder‐pickers'routes,determinationofpickinglocations,technology/placeofpickedSKUsputoff,etc.[1].

Although the spectrum of applied order‐picking technologies is verywide, [2, 3, 4], in praxissolutions based on themethod „man to goods“ aremostly dominant. Also, in thismethod socalled „low‐level“ principle is the most applied – so that the zone of picking goods is inergonomicallysuitableareaavailable toorder‐pickerwhilemoving throughof the floorof thewarehousearea[1].Thisisveryimportantduringlabor‐intensiveactivities,wherehugenumberoforderedSKUshavetobepickedinshorttimewindow.Thistypeoftasksrequiresignificantactivities of order‐pickers, where order‐pickers’ traveling time has the greatest share (5).Therefore, themoving phase is one of the focuses of research of rationalization possibility oforder‐pickersactivities,and thispaperpresentsoneof thepossibleapproacheswith thisaim,basedonnewtechnology–byuseofremotecontrolledorder‐pickerforklift(OPF).

* mmiljus@sf.bg.ac.rs

Having that inmind, this paper consists of several sections. After introductionwith problemdescription, the second section shows short reviewof themost appliedmanual order‐pickingtechnologieswith the analyze of the time structureof order‐pickers’working cycle.The thirdsection presents the characteristics of the new technology based on the appliance of OPF: itincludesworkprinciple,timeworkingstructureanddifferencebetweenthisandthetechnologyof AGV system. The fourth section gives review of potential effects through overview ofimportantmeasurers–possibilitiesoftimesavings,advantagesanddisadvantagesofapplianceofthistechnology.Inthissection,someeffectsoftheuseofthenewtechnologyarepresented.Theconclusiongivesconsiderationofup‐to‐day,relativelysmallexperiencesinthisarea,aswellas discussions about potential researches related to the use of OPF in the order‐pickingprocesses.

2.BRIEFOVERVIEWOFTYPICALMANUALORDER‐PICKINGTECHNOLOGIES

Regardlessofhighlevelofthedevelopmentoftechniquesandtechnologyinlogisticsprocesses,about 80%of the activities in thedomain of order‐picking technologies are basedonmanualwork.Byruleitistheconsequenceofthefrequentadjustmentstotherequestchangesoforder‐pickingbydifferentkindsofparameters(typeofgoods,logisticsunittype,orderpattern,etc.),when sophisticated technologies require significant time (and costs) to adapt to the newcharacteristicsofrequests.

We can see several various categorizations when we speak about manual order‐pickingtechnologies[1,6].Themostcommontechnologiesarebasedonman(order‐picker)movementwith the transport means (forklift or carts with or without motor drive) along aisles withinorder‐pickingarea(Figure2.1). In thispaper, technologyOPFpresentedon theFigure2.1.c isanalyzed.

a) b) c)

Figure2.1.Someexamplesofthemanualpickingtechnologywithdifferenttransportmeans

Order‐pickermovestothepredefinedpickinglocation,wherehe(byrule)stepsoffromtheOPFand walks to the picking location, picks requested number of ordered SKUs (Stock KeepingUnits),moveswithgoodstoOPFandputthemonpalletthatisonOPF.Thementionedprocessisbeing repeated until the completion of the order or the fulfillment of the OPF’s capacity;afterwardspickedSKUsaredispatchedtothelocationoffurtheractivitiesrealization(eventualsorting,packing,etc.).Byanalyzingorder‐picker'sworkinthistechnology,wecanseethatupondefining (and receipt) of picking order‐list, the order‐picking task on one location consists offollowingsetoftypicalelementaryactivities:

1.ReceiptofinformationofpickinglocationoforderedSKUs,2.Order‐picker’ssteppingon(ifrequired)ontheOPF,3.ActivationoftheOPF’smoving,

4. Moving the OPF with order‐picker to the required location and the OPF’s stopping andsecuring,5.Order‐picker’ssteppingofffromtheOPF,6.Order‐picker’smovingfromtheOPFtotherequiredSKUs’location,7.Searching/identifyingandpickingoftheorderedSKUs,8.Order‐picker’scarryingtheSKUstotheOPFandputthemontheOPF,9.ConfirmationoftaskrealizationforrequiredSKUs.

Duringworkingtime,theseactivitiesarebeingrepeatedforeachpickinglocation.Consideringspecificitiesandrepetitionoftheseactivities,thedistributionoforder‐picker’sengagementtimeinthistechnologywasthesubjectofthewholerangeofanalyzes([3,5,7]).Accordingtothesesources, the most dominant participation in the order‐picker’s time work structure iswalking/movement time (app. 50%‐60%), while the searching and picking time of the SKUsparticipateswith20‐25%.Therefore,possibilitiesofrationalizationofthistimemovementwerethe primary area of researches. The following section gives the overview of some of theappliance of the remote controlled OPF technologies, since they provide achievement ofmentionedgoal.

3.SPECIFICITIESOFREMOTECONTROLLEDOPFTECHNOLOGIES

Development of remote controlled OPF technologies had the goal of the elimination or therationalizationofsomeofthepartialtimesoftypicalelementaryactivities.Theprimaryideaisbasedontheeliminationoforder‐picker’stimeforsteppingon/offtheOPFaswellasgivingtheappropriate associated commands for vehicle guidance (activities 2, 3, and 5, and partiallyactivity4describedinSection2).Thedevelopedsolutionsbasedonthetwoprimaryprinciplesof giving orders for the OPF movement (i) by the order‐picker, and (ii) by WMS, will bepresentedindetailsinfollowingchapters.3.1OrdersforOPFmovementwithactiveorder‐picker’sparticipation

The principle of this technology is based on the activation of the appropriate buttons by theorder‐picker.Givenorder(start,turn,stop)istransferredtotheOPFviaradio‐connection.Thistechnologyhasseveralsolutionsandtwoofthemarepresentedasfollows.

a)OPFremoteunitbondedaroundorder‐picker'shandbythestrip

Theorder‐pickerreceivesinformationrelatedtothenextpickinglocation(OPFstop)basedonthe order‐picking list. By pressing the appropriate button on the remote unit (Figure 3.1.a,www.still.de;named“iGoRemote”)desiredorderoftheOPFisactivated.

Figure3.1.a–Overviewoftheremoteunitbondedaroundorder‐picker’shand

It isnecessary tonote that theorder‐picker’shandactivities canbepartly reducedusing thisalternative,sincetheremoteunitrequiressignificantattentionandhigherlevelofhand’suse.

b)OPFremoteunitontheorder‐picker'sglove

Theorder‐pickerreceivesinformationrelatedtothenextpickinglocation(OPFstop)basedonthe order‐picking list. The optionwith this remote unit is in the form of glove (Figure 3.1.b,http://video.tdcols.com/video/g_‐V31UL4Ww; named “Crown'sQuickPickTM”). By pressing theappropriatebutton,desiredorderofOPFisactivated,withthepossibilityofsoundsignalaswell.

Figure3.1.b–Overviewofremoteunitintheshapeoforder‐picker’sglove

In this technology, the freedom of the order‐picker’s hand activities is on the higher levelcompared to the alternative a), since the orders for the OPF require less attention and lessorder‐picker'susageofhand.Also,greater freedomofhandsusageduringpickingandputtingoffactivitiesisachieved.

3.2OPFremotecontrolbyWMS,withoutactiveorder‐picker'sengagement

Inthisalternative,theappliedWMShasdoublerole:controlsOPFandtransfersinstructionstothe order‐picker usingpick‐by‐voice technology (Figure 3.2, http://video.tdcols.com/video/g_‐V31UL4Ww;named“Pick‐n‐Go”byToyota(BT)).BasedontheWMSinstructions,OPFmovestothe location where the SKUs have to be picked and stops there. Order‐picker receivesinformationrelatedtothepickinglocationandthenumberoftheorderedSKUsviaheadset,pickthem and walk with items, putting them on the OPF. By voice – over microphone andcommunication system –WMS receives information related to the realization of the task fororderedSKUs.Basedonitandaccordingtothefulfillmentoftheorder,WMSsendsinstructiontotheOPFaboutthenextpickinglocation,order‐picker'sdecidesaboutthemodeofmovement(onOPF,orbyfootbehindit).

Figure3.2–OverviewofOPFremotecontrolusingWMS

This technology influences a range of significant factors which increase the order‐picker’sproductivity: complete freedom of hands; with the option of OPF with lifting forks, WMSadditionallyprovidespalletunit liftingonergonomicallyoptimalheight level forSKUsputtingon;usingAGVSelementsincreasesworksafetywiththeOPF,byusingalotofsensors/detectorsfornoticingobstacleson thepath, soundand light signals,etc.– this isvery importanton thetransport paths/aisles/corridors in the order‐picking area where the presence of people,vehicles,goodsandequipmentisverysignificant;WMSusingAGVsystemprovideseliminationoforder‐picker'sworkrelatedtomovementtotheOPFanddrivingtothedisposalplaceofpalletunitwhentheorderiscompleted.

4.POTENCIALEFFECTSBASEDONAPPLIANCEOFREMOTECONTROLLEDOPF

Theinfluenceofthepresentedtechnologyontimereductionoftheelementaryactivitiesoftheorder‐picker (see Section 2) is shown in Table 4.1. Thereby, the estimations are given byauthors,comparedtotheconventionaltechnology.

Table4.1Leveloftheappliedtechnologies'influenceonthetimereductionoftherealizationoftheorder‐picker'selementaryactivities

Technology t1 t2 t3 t4 t5 t6 t7 t8 t9 1 3.1.a +++ + ++ +++ + ‐ ++ 2 3.1.b +++ ++ ++ +++ + ++ 3 3.2 +++ +++ +++ +++ + + +++

Legend:‐unfavorableinfluence;noinfluence;+smallinfluence;++mediuminfluence;+++greatinfluence

The available sources fromcompanies‐producers areused since, according toourknowledge,therearenoexpertpaperswhichhaveexplicitlyanalyzedthistechnologyandpotentialeffectsof its appliance. By analyzing the available sources of several producers (Still, Crown, Toyota(BT))potentialeffectswhichmightresultcanbeclassifiedinseveralgroups:

Timesavingsarementioned as primary achieved effect – according to thedata published incompanies' materials, time of order‐picker's movements is reduced up to 70 percent by theeliminationoftherangeoftheunnecessarymovements,ordersfortheOPFoperationsetc.;

The reduction of the traveled path of order‐pickerwith load – which results with lessworkload of order‐picker, and thereby with the increase of the quality level of the basicprocessesrealization;

Thedecreaseofpotentialinjuriesandprofessionaldiseases–bythereductionofsteppingon/off OPF at picking locations, effects can be expected in the decrease of workers’injuries/diseases, having in mind the extent of these activities during work (i.e. for severalyears);

Preconditionsforworkingenvironmentwithhigherorder‐pickingdensityareachieved–inpaths/aisles/corridorsoforder‐pickingarea,itispossibleformoreorder‐pickerstoworkatthesame timewithOPF; theriskof injuriesofworkersand/orgoodsandequipmentdamage(duetoactivitydensity)arebeingreduced/eliminatedbyapplyingOPFequipmentmentionedinpreviousSection;

Flexibility – when required, it is simple to move from semi‐automatic/automatic to manualcontrolofOPF;

Cost‐effectiveness – The increase of the productivity and the decrease of the energyconsumptionhaveinfluenceoncost‐effectivenessandROItothistechnologycanbeexpectedinrelativelyshorttimeperiod.

In less extent, some of described technology appliances can be met in praxis. Most of themoutline general positive impressions and conclusions which imply to the favorableresults/effectsoftheappliancethatcannotbepresentedinthisworkduetolimitations.

5.CONCLUSION

Byanalyzingpresentedtechnologies, itcanbeconcludedthatthesearenewsolutionswithoutsignificantexperiences inpraxis.Mentionedmanufacturersstate theirobservations,especiallythe ones referring to the range of favorable effects. It is clear that these states should passtheoreticalandpracticalaudits,havinginmindseriesofspecificitiesrelatedtorealistictasksoforder‐picking, staff educations, appropriate information system support andothers.However,potentialeffectsofthistechnologyareveryinterestingforwarehousesystemdesignersanditisrequiredtohavethesetechnologiesinmindwhendevelopingalternativetechnologicalconceptsof order‐picking subsystem(s). Therefore, it is necessary to perform series of research in thisarea,which canbe related to the theoretical aswell as to thepractical aspects. Someof themmight refer to thework and time studies in different environment/conditions (type of goods,orders’ patterns, path conditions, etc.), the analysis of the specific strategies for goodsassignmentappliance,theuseofadditionaldevicesandequipmentonOPFandetc.Thatway,themorequalitativebases fordecision‐making in logistics’ systemdesignwouldbeacquired, andthroughthem,supportformorequalitativelevelofsupplychainsolution.

REFERENCES

[1] DeKoster,R.,Le‐Duc,T.,Roodbergen,K.J.,2007.Designandcontrolofwarehouseorderpicking: a literature review. European Journal ofOperational Research. 182 (2), p. 481‐501.

[2] Gudehus T., Kotzab ·H, Comprehensive Logistics, Second Revised and Enlarged Edition,Springer‐VerlagBerlinHeidelberg,2012,

[3] Frazelle,E.,H.,,World‐ClassWarehousingandMaterialHandling,McGraw‐HillCompaniesInc.,NewYork,2002.

[4] Ten Hompel, M., Sadowsky, V., Beck, M., Kommissionierung: Materialflusssysteme 2 –Planung und Berechnung der Kommissionierung in der Logistik, Springer‐Verlag BerlinHeidelberg2011

[5] Tompkins,J.A.,White,J.A.,Bozer,Y.A.,FrazelleE.H.,Tanchoco,J.T.,FacilitiesPlanning,JohnWilley&Sons,NewYork,1996.

[6] Đurđević B.D., Development of Models for Order‐Picking Area Selection and Design,PhDThesis(inSerbianlanguage),UniversityofBelgrade,2013

[7] Martin, H., Transport‐ und Lagerlogistik, 4 Auflage, Vieweg & Sohn VerlagsgesellschaftGmbH,Braunschweig/Wiesbaden,2002.

STMODELIMPLEMENTATIONONSHORTTERMOPERATIONS’PLANNINGATBULKTERMINALS

SvjetlanaHessa*,MiranoHessb

aUniversityofRijeka,FacultyofMaritimeStudies,Croatia,shess@pfri.hrbUniversityofRijeka,FacultyofMaritimeStudies,Croatia,hess@pfri.hr

Abstract:Thispaperfocusesonabulkterminalshorttermoperationsplanningproblemfacedbytheportmanagement in tacticaldecisionmaking.Loadinganddischargingof shipcargo,cargostocking,maintenanceandserviceoffacilityequipmentareregularoperationsofaseaport'sbulkterminal. Their organization is subject to difficult‐to‐predict or unforeseen influences. Thechallengethatportmanagementfacesindaytodayoperationsisinmakingthebestpossibleplanwithscheduleddurationofoperations/statesandtransitioninstants,consideringvariousinternalandexternal factors influencing terminalperformance.A statesand transitionsmodel isused toderiveeffectivesolutions forobtainingstatesorderandstatestransitiontimeofabulkterminal,with the objective of minimizing operational costs. Behavior of the terminal is tested withstochasticanddeterministicmethods.

Keywords:operationsplanning,seaterminal,generalsystemstheory,stochasticmodeling

1.INTRODUCTION

Our understanding of the traffic phenomenon is based on empirical researches and verbaldescription of traffic systems. The core concept of a systemic traffic theory is not presentlyavailableinunifiedandformalizedform.Thefieldoftrafficscienceandtechnologyisextremelybroad one, encompassingmany different disciplines and activities, thus the unification seemsimpossiblewithoutapplicationofgeneralsystemstheoryandmethodology.Thepartialuseofsystem's theories in major part of traffic literature has been only a superficial descriptionwithout precise formulations derived from concept of general or generalized system. On theother side, classical analytic approachwith bounded discipline‐oriented researches, use theirowntheoreticalconceptsandmethodologies.

Radmilović (1989) presents functioning of port facilitieswith discreteMarkov processes andproposes the application of themodelwith system of differential equations, which describestechnological processes of direct and indirect trans‐shipment of the cargo.Radić andBošnjak(1997) give the concept of the generalized traffic model using general system theorymethodology,andderiveequationsfromST‐diagramforstationarybehaviorofthesubsystem.Kiaetal.(2002)exploreportcapacityunderanewapproachbycomputersimulation.Asperenetal.(2003)proposeapossiblewayofmodelingshiparrivalsinports.

* Corresponding author

Themajor objective of port operationsplanning is to diminishport vacancy, thusminimizingoperational costs, while assuring that the service rendered to ships is in line with widelyacceptedstandards.

Even though the wide range of planning problems within the shipping industry receivedsignificant attention from researchers so far, as in Cullinane et al. (2009), there are stillproblems that have to be addressed, i.e. port operations planning under uncertainty. Theproblemof optimizationunder uncertainty exists in bulk terminals too, but is of a somewhatdifferentcharacter, seeHessetal. (2007),HessandHess(2010).The limitedstoragecapacityandfacilityoutputnecessitateplanningofterminaloperationstopreventstorageoverflowandunoccupiedterminalcapacities.

This paper focuses on terminal behavior understanding and addresses the question whetherterminaloperations showdeterministicor stochasticbehavior.Themajor contributionof thiswork is in determination of the states and transitions model (ST model) for bulk terminalbehaviorobservationwhichisbasedoncomparisonofdeterministicstatesandtransitions(DSTmethod) and stochastic states and transitions (SSTmethod).Worksheets of bulk terminal ofBakar port in two years period (2013‐2014) are analyzed. Contribution is in conclusions onwhichmethodbetterfitsrealexample.Finally,practicalextensionsareoutlined.

2.GENERALSYSTEMSTHEORYAPPLIEDTOTHEPORT

2.1Fundamentaltraits

Highest‐level generalization is axiomatic,mathematical theory of traffic system.On that level,fundamental traits and relations must be derived from a concise formal definition of trafficsystem.CollectionofconceptsanddefinitionsforfundamentaltraitsofsystemaregiveninKlir(1972). The general systems theory is applied in the state system analysis, forecasting andplanning development of dynamic systems, in selection of optimal or at least satisfyingmanaging actions and decisions. The state and the transitions between the states will beidentified,alongwiththescheme(ST‐diagram)onthebasisofwhichthemathematicalmodelisderived.Throughtheproposedmodelonecanobservetimevaryingportsystemoperation.

Fivedefinitions,eachbasedonaseparatetraitaredefinedbyKlir(1972).Eachverbaldefinitionis followed by a mathematical definition, the two indicated as (a) and (b). Definition 4 anddefinition5ofatrafficsystem(RadićandBošnjak,1997)areasfollows:

Definition4.

a) A traffic system is a given set of elements, their permanent behaviors, and a set ofcouplingsbetweentheelementsandbetweentheelementsandtheenvironment.

b) A system is 2‐member set (B, C), where: 1 2, , , rB b b b is the set of all permanent

behaviorsof elementsof theuniverseofdiscourseand ;ij ij i jC c c A A i j is the

setofcouplingsbetweenthesetofinputquantitiesAiandthesetofoutputquantitiesAj.

Definition5.

a) Atrafficsystemcanbedefinedbyitshypothetical(known)ST‐structureasasetofstatesandasetoftransitionsbetweenthestates.

b) Asystemisa2‐member(S,R(S,S)),where:Sisthesetofstates;Rarelationdefinedon(SxS)orasystemis3‐member(S,R(S,S),P(R)),where:P(R)isaprobabilitymeasuredefinedonRsuchthatif(si,sj)R thenP(sj si)isconditionalprobabilityoftransitionfromstatesitostatesj.

Aminimaldefinitionofsystemwouldhavetobeoneofthebasicdefinitions.

2.2Theuniverseofdiscourseandcouplingsoftheportsystem

AccordingtotheDefinition4,setofalltheelementsandtheirlinksinthesystem"servingshipatquay"isshownbytheUC‐structure(Fig.1).

Theelementsofthesystem"servingshiponquay"are:

theship(S)–theobjecttowhichtheactivityisdirected, thequay(Q)–theelementquaydoestheloading/unloadingoperationsoftheship, the centre for organization the technological process (COTP) – organizes, coordinates

andcontrolsthetransshipmentprocess,doesthepaper‐workregardingthecargo,givespossibilitiesforobtainingthedifferentstatisticaldata,transactsinvoice.

Thelinksbetweentheelementsofthesystemareasfollows:

L0–initiation–putsthesystemintheactivestate,andstartswiththeshiparrivalatthequay,

L1–two‐headedarrowbetweentheelementsSandCOTP,servesforinformingtheCOTPon ship arrival and for acknowledgment transmission, and for additionalcommunicationsbetweentheSandtheCOTP,

L2–two‐headedarrowbetweentheCOTPandQ, isrepresentedbythecommunicationchannelswithpurposetocoordinatetheloading/unloadingoperations,

L3 – two‐headed arrow between the elements S and Q, is represented by thecommunicationchannelsintendedforthecommunicationsbetweentheSandtheQ,

Figure1.UC‐structureofthesystemservingshipatquay

L4–two‐headedarrowbetweentheCOTPandtheenvironment,servesfortheCOTPtocommunicatewith themeteorological service, the agents, forwarders, land carrier, airandrivercarriers,

L5 – two‐headed arrow between the ship and the environment, and serves for thecommunicationbetweentheshipandtheagents,forwarders,meteorologicalservice,andsoon.

SHIP COTP

QUAY

L1

L5

L0

L2L3

L4

ENVIRONMENT

2.3Thestatesandtransitionsbetweenstatesoftheportsystem

Thesetofstatesandtransitionsbetweenthesestates(Definition5)forthesystem"servingshipatquay"ispresentedbytheST‐structure(Fig.2andFig.3).

3.THEMODEL

Theobjectiveofthepaperistoidentifytheparticularstateinwhichthebulkterminalwillbeinagivenmomentinthefuture,startingfromassumptionthatinthebeginningitwasinidlestateandthatstates'switchingoccurredwithdesignatedtransitionprobabilities.Themaingoalistoanswerthequestionwhethertheobservedbulkterminalbehavesasadeterministicsystem,i.e.according to the logical terminal'soperation flow,or as a stochastic system,meaning that theinfluencescausingstatetransitiondisorderarenotnegligible.

Twoapproachesinquantifyingthestatetransitionsareexamined.Thefirstconsistsofsettingupasystemofdifferentialequationsforterminaloperationswithassumptionthattheterminalhasdiscretestatesexpressedwithprobabilities.Thesecondapproach,bydefiningthebulkterminaloperations as Markov processes and setting up matrix of transition probabilities, yield stateprobabilities and lead quickly to acceptable solution of ST model that is essential for anypracticalapplication,HessandHess(2010).

ThestatesoftheSTmodelofterminaloperationsare:

S1– idlestate(nooperationsontheterminalexceptdataprocessing, i.e.collectionandanalysisofweatherreports,cargo/shipsrelatedinformation)

S2 – preparatory state (operations carried on the terminal just before ship arrival, i.e.preparationoffacility/cargo/longshoremenforcargooperations)

S3 – transshipment state (cargo loading and/or discharging; from economicalperspectivethemostdesirablestateoftheterminal)

S4 – closing state (operations immediately after finish of ship loading/discharging, i.e.paper‐work,shipdepartureoperation)

S5–repairandmaintenancestate(regularmaintenanceofequipment,repairincaseofmachinerybreakdown).

Figure2.ST‐structureofdeterministicterminalbehavior

Figure3.ST‐structureofstochasticterminalbehavior

Aterminalhasdeterministicbehavioriforderofstatesandtheirdurationsareexactlyknowninadvance,Hessetal.(2008).InDSTmethodthesetofstatesandtransitionsbetweenthesestates

for a bulk terminal is formed around ST‐structure shown in Fig. 2. A terminal has stochasticbehavioriforderofstatesandtransitionsdoesn'tfollowlogicalworkflowduetovariousinternalandexternalunforeseeninfluencesonregularoperations.ForresearchofsuchasystemtheSSTmethodwillbedeveloped.InthiscasetheST‐structuremaybedefinedfromFig3.

Data thatwerederived froma terminalworkwereused to assemble a problemof stochasticterminaloperations.Todefineterminal'svariousoperationsforSSTmethod,worksheetsofbulkterminalofBakarportinoneyearperiod(2013)havebeenanalyzed.Thoseoperationsincludetransportation of bulk cargo from/to terminal, loading/discharging cargo to/from ships,inspection of ship and cargo, distribution of cargo to shore stock,maintenance and repair offacility equipment, and customsprocedures. The data on frequency ofmachinery failure, badweatherandstrikecausedstoppagesofoperations,andcongestionsontheterminalwerealsotaken into consideration. These data served as a basis for population of stochastic matrix oftransitionprobabilitiesforbulkterminalbehavior.

Thesolutionwasobtainedwithcomputer‐assistedevaluationprogramWinQSB.WinQSBhasanintegrated Markov modeling and simulation tool, based on discrete space, continuous‐timeMarkovmodel. After data entry in the transition table, definednumber of periods (n=1,…, 12steps)and initial statevectorof the terminalat time t=0, 0 1,0,0,0,0P , probabilitiesof the

five terminal states are obtained and presented in Fig. 4. Starting from idle state, simulationshows in which state the terminal will appear most probably after each transition (step).Probability of the most notable state decreases with number of simulation steps and theterminalapproachessteadystateprobabilities.

Figure4.Probabilitydistributionforterminalstatesfor12steps

After obtaining state probabilities for 12 steps in SST method and deducing the stateprobabilities inDSTmethodfollowscomparisonofresultswithreal‐worldoperationsflowforyear 2014 and selection of best‐fit method for further short term planning. For example,durationofstateS1dependsonactualtimeofships'arrivalanddeparturetimeofpreviousship,while duration of state S3 is influencedwith size of ships and quantity of cargomanipulated.Consequently,durationofeachstateisdeterminedwithrealsituationinterminaloperations.

TheorderofstatetransitionsonweeklybasiswereevaluatedandthefinalresultsshowedthatdataobtainedbySSTmethodmatched thepractice in34cases(weeks),byDSTmethod in11casesandin7casesthestatetransitionsfollowedsomeotherorder.Itwasconcludedthattheobservedterminalhadstochasticbehaviorin65%casesduring2014.

Comparisonof theDSTsolutionto thecorrespondingSSTsolution indicates that the lateronebetteremulatesthelogicofbulkterminaloperationsflow.Therefore,iftheportmanagement,inexisting situation and without overtaking specific measures for improvement of operationaleffectiveness, follows the SSTmethod in terminal operations short term planning, the plan isexpectedtobemorefeasible.However,bearinginmindthattheSSTterminaloperationsdraw

0 1 2 3 4 5 6 7 8 9 10 11 12

p1 1.00 0.00 0.03 0.03 0.36 0.40 0.03 0.09 0.22 0.32 0.19 0.10 0.18

p2 0.00 0.98 0.00 0.03 0.03 0.42 0.39 0.04 0.10 0.25 0.34 0.19 0.11

p3 0.00 0.00 0.95 0.00 0.04 0.14 0.41 0.39 0.05 0.15 0.28 0.34 0.21

p4 0.00 0.00 0.00 0.91 0.00 0.04 0.13 0.39 0.37 0.05 0.14 0.27 0.32

p5 0.00 0.02 0.01 0.03 0.57 0.01 0.04 0.10 0.26 0.24 0.05 0.10 0.18

0.000.100.200.300.400.500.600.700.800.901.00

p1 p2 p3 p4 p5

longerworkingproceduresandthereforetimelostonovercomingeffectsofunforeseeneventsandconsumemoreresources,toperformevenbetterthemanagementshouldstrivetoadheretotheplanbasedonDSTmethod.

3.CONCLUSION

Inthispaperbasicsofthegeneralsystemstheoryarepresented.Thistheoryisappliedforthesystemstatesanalysis,forecastingandplanningthedevelopmentofthedynamicsystems,thenchoiceoftheoptimaloratleastadequatelycontrolsactionsanddecisions.Lackonuniformityinthecaseofcargoarrivalattheportandimpossibilitytopredictexactlytimeandthequantityofthe cargo arriving to the port, are the main reasons of the stochastic property in the portoperating. The port can be presented as physical systemswith random changes during timewhichdrawnecessityofusingprobabilitiesinitsmodeling

Themethod,presentedhere,maybeusedforshorttermtacticaldecisionmakingbyidentifyingthe particular state in which the terminal will be in a given instant. One of the majorshortcomingsoftheSSTcomparedtoDSTterminaloperationsisrepresentedbytimelostandconsumption of more resources on overcoming effects of unforeseen events, resulting inoperational costs inefficiency. The ST model presented can serve as theoretical base formodelingtechnologicaloperationsofotherportterminalsortrafficsystems.

REFERENCES

[1] Asperen,vanE.,Dekker,R.,Polman,M.,Arons,H.deS.(2003).ModelingShipArrivalsinPorts. Proceedings of the 2003Winter Simulation Conference, S. Chick, P. J. Sánchez, D.Ferrin,andD.J.Morrice(Eds.),NewOrleans,1737‐1744.

[2] Chang, Y‐L. (1998). WinQSB, Decision Support Software for MS/OM, Version 1.0. JohnWiley&Sons,Inc.,NewYork.

[3] Cullinane,K.,Song,D‐W.,Wang,T.(2005).TheApplicationofMathematicalProgrammingApproaches to Estimating Container Port Production Efficiency. Journal of ProductivityAnalysis,24(1),73‐92.

[4] Hess, S., Hess, M. (2010). Predictable uncertainty about terminal operations in the sea.Transport,25(1),148–154

[5] Hess,M.;Kos,S.;Hess,S.(2007).QueueingSysteminOptimizationFunctionofPort'sBulkUnloadingTerminal.Promet–Traffic&Transportation,19(2),61‐70.

[6] Hess,M.; Hess, S.; Kos, S. (2008). On Transportation SystemWithDeterministic ServiceTime.Promet–Traffic&Transportation,20(5),283‐290.

[7] Kia,M.;Shayan,E.;Ghotb,F.(2012).Investigationofportcapacityunderanewapproachbycomputersimulation.Computer&IndustrialEngineering42,533‐540.

[8] Klir,G.J.(1972).TrendsinGeneralSystemsTheory.JohnWiley&Sons,NewYork.

[9] Radić,Z.,Bošnjak,I.(1997).GeneralizedTrafficModelandTrafficEquationsDerivedfromST‐Diagrams.ModernTraffic,17(3/4).

[10] Radmilović, Z. (1989). Analytic Modelling of the Port System by Means of the DiscreteMarkovChains.Traffic,36(10),823‐841.

PartIV LOGISTICSCONCEPTSAND

STRATEGIES

AIRCARGOFLOWANALYSISINTHEEUROPEANUNION

DanicaBabića*,JovanaKuljanina,MilicaKalića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Although the demand for air cargo services in theworld has been showing signs ofacceleratedgrowthsince2010,thelevelofcargotrafficintheEuropeanUnionisstillbelowitspre‐crisis2007level(measuredinfreighttones‐kilometres).Notwithstanding,aircargovolumeintheEuropeanUnionaccountsforupto20%ofthetotalaircargoworldwide.InthispapertheauthorsexamineaircargonetworkandaircargoflowsonscheduleroutesintheEuropeanUnionin2013.AircargoflowsincludeflowswithintheEuropeanUnionaswellasflowsfromtheEuropeanUniontootherregionsintheworld.Additionally,mainEuropeanUnioncargoairportsareidentifiedandcargomovementsbetweenthemareanalyzed.

Keywords:Aircargo,flows,cargoairports.

1.INTRODUCTION

Theimportanceofaircargototheeconomyandconsumerscouldbeperceivedthroughthefactthataircargorepresentslessthan1%oftheweightofallinternationalcargo,whileatthesametimethissegmentrepresentsaround35%ofthetotalworldwideshipmentvalue(Boeing,2014).Theair transportmode found its roleon themarket to shipgoodsbetweencountriesquicklyand efficiently compared to othermodes of transport, especially on the longer routes wheresavings in time is most noticeable. This advantage makes air transport more suitable fortransportofgoodswhichareeithertime‐sensitiveorhigh‐value.Butunfavourableeconomicaldevelopments in recent years have had a direct impact on people’swillingness to spend and,thus,onworldtradegrowth.Asignificant,negativeimpactoftheeconomiccrisis isnoticeableontheEuropeanmarketbecause,asamaturemarket, ithasagreatnumberofconsumers forhighvalueproducts(whichtendtouseairfreight).Thus,ithashadimportantconsequencesonairfreight.

Developmentoftheairlineindustrywasthesubjectofmanyresearchpapers.Thefocusoftheseresearch papers was mainly on the passenger side of the airline industry, putting the cargosectordevelopmentaside.Onlyafewoftheresearchpapersdealswithaircargo,Zhang(2003),Huietal.(2004)andHwang(2011),whichfocusingontheEasternandSouthernAsiamarkets.

In 2013, cargo airlines in the European Union (EU) took some cyclical upturn and improvedprofitability,afteryearsofstagnation(IATA,2014).Butthenegativeeffectsofglobaleconomicrecessionthatbeganin2007/2008weresostrongonaircargodemandthat,intermsoffreighttonne‐kilometres(FTKs),thetrafficlevelswerestillbelowthepre‐crisis2007level(AEA,2013).Additionally,highfuelpricesandincreasedcompetitionfromothermodesoftransportandnon‐EU airlines still make the European market a challenging environment. However, the EU air * d.babic@sf.bg.ac.rs

cargoindustryisoneofthemostcompetitiveintheworld.AirfreightaccountsforasignificantshareofEUcargointermsofFTKs.Almost20%ofworldaircargoFTKsaretransportedthroughthe EU (WorldBank, 2015).Measured in tonnes of goods, the share of EUmember countriesaccountsforupto30%inthetotalworldaircargotransport(in2013,airlinestransported49.8milliontonnesofgoods)(IATA,2014).

Inspiteoftheslowrecoveryandlowgrowthrates(IATA,2014),aircargoisstillveryimportantinfacilitatingtradeandeconomicactivityofEUmembercountriesanddevelopmentofefficientand sustainable transport network should not be neglected. Analyzing transport flowswithinthenetwork is very important for understanding air cargomovements anddeveloping futurestrategies for airlines, airports and policymakers. This paper presents statistical data on aircargohandledatairportsintheEuropeanUnion(EU)in2013.ThemainobjectiveofthispaperistocontributetoabetterunderstandingofaircargonetworkflowsintheEU.It,also,examinesaircargomovementsfocusingoncountrieswiththehighestaircargovolumeswithintheEUandindividual results for theirmainairports, aswell as the strongestair cargo flowsbetweenEUmembersandregions inotherpartsof theworld.Theremainderof thepaper isorganizedasfollows. Section 2 consists of three parts that describe air cargo flows in EU on domestic,international and intercontinental level. This section investigates distribution of air cargovolumefromEUbyregionsandcountriesandprovidesareviewandoutlookfortheexpecteddevelopmentsofmainaircargoflows.Finally,Section3containsconcludingremarks.

2.AIRCARGOFLOWS

ThispaperrepresentsthereviewandanalysisinaircargotransportintheEUbycountriesandairportswithemphasisonmaincargoflowsinsideandoutsidethisregion.TheresearchisbasedondatacollectedfromtheEurostatdatabase(statisticaldatarelatedtocarriageofgoodsbyair)intheperiodfromJanuarytoMarch2015.Eurostatdatabasecontainsallaircargomovementsinthe28EUmembercountries.Thetotalvolume,measuredinweightofaircargo,transportedthroughEUmembercountriesamountedto14.6milliontonnesin2013.Accordingtothisresult,aircargovolumeremainedmoreorlessstablefrom2012to2013,witha0.04%decreaseinthetotal gross weight of goods. The total air cargo volume based on geography allocation isdistributedasfollows:

72.64%of the totalair cargovolume is tradedbetweenEUmembercountriesand thecountriesoutsidetheEU,

23.64%ofthetotalaircargovolumeistradedbetweenEUmembercountries,

3.72%ofthetotalaircargovolumerepresentsnationalcargotrade.

Germanyalonehadnearly4.5milliontonnesin2013,thatis30%oftheEUaircargotransport.GermanywasfollowedbytheUnitedKingdom(UK),FranceandtheNetherland,withsharesof16%,12%and11%oftheEUtotal,respectively.HighconcentrationofaircargotransportintheEUisalsonoticeablethroughthefactthatfivemembercountriesaccountformorethan75%ofthe total air cargo volume (Table 1). FrankfurtMain Airport is the largest airport due to thecargotonnesmovedbybothbellyanddedicatedfreighters,followedbyAmsterdam’sSchiphol,London’sHeathrow andParis’ Charles deGaulle that recordedmore than1million tonnes in2013 (Table 1). Besides Frankfurt Main Airport, there are two more airports in Germany(LeipzigandKoln)thatbelongtotoptenEUairports intermsof tonnagethroughput in2013.AltogetherthosethreeGermanairportsaccountfor25%oftotalaircargotonnesintheEU.The10largestairportsaccountedforabout72%ofthetotaltonnageofaircargohandledintheEUcountriesin2013.ThisresultindicatesthataircargointheEUisconcentratedintoseveralkeyairportsthatareclosetoregionalcentresofproductionandconsumption.

2.1DomesticandInternationalAirCargoFlows

The total air cargo that is transported within the EU comprises approximately 3.0% of theworld’s air cargo tonnage, but only 0.8%of theworld’s tonne‐kilometres (Boeing, 2014).Theroutes insidetheEUareprimaryshorthaul,mostlybetween900and1,200kilometers,whichfavorsothermodesoftransportduetolowercosts.

Table1.EUmembercountriesandairportsbyaircargovolume,2013,(Source:Eurostatdatabase,2015)

Country 2013(t) % Airport 2013(t) %

Germany 4,421,995 30.11 Frankfurt/Main 2,160,824 14.71

UnitedKingdom 2,362,259 16.08 Amsterdam/Schiphol 1,565,755 10.66

France 1,775,342 12.09 London/Heathrow 1,512,246 10.30

Netherlands 1,619,819 11.03 Paris/CharlesdeGaulle 1,491,287 10.15

Belgium 1,004,808 6.84 Leipzig/Halle 895,270 6.10

Italy 814,024 5.54 Köln/Bonn 766,131 5.22

Luxembourg 673,445 4.59 Luxembourg/Luxembourg 673,445 4.59

Spain 580,492 3.95 Liege/Liege 560,470 3.82

Austria 221,518 1.51 Milano/Malpensa 430,342 2.93

Finland 192,421 1.31 Brussels 400,282 2.73

Other 1,020,882 6.95 Other 4,230,983 28.81

Totalfreight 14,687,005 100 Total 14,687,005 100

Air cargo flows in theEUcanbeobserved in respect todomestic and international transport,separately.Freightondomesticflightsaccountforonly4%ofallEUairfreight(545,988tonnesin2013)sincenarrow‐bodiedaircraftwithminimalbellycapacityareusedfortheseflightsandroadandrailtransportwithintheEUaremorecosteffective.AsmallproportionofcargothatistransportedbyairindomesticmarketsofEUcountriescanalsobeobservedthroughitssmallshareinthetotalaircargotransportedwithintheEUborders,thatis16%.ThecountrieswithhighestdomesticaircargovolumesareFrance(156,238t),Germany(119,485t)andtheUnitedKingdom (116,933 t), whilemain airportswith domestic cargo are EastMidlands (73,146 t),CharlesdeGaulle(67,069t)andLeipzig’sHalle(60,985t).

AircargotransportwithintheEUachievedaveryslowgrowthsince2011(2%annually),afterhavingrecoveredfromtheeconomiccrisisin2008.PrimaryroutesforairfreightinsidetheEUare the routes fromGermany to theUnitedKingdom, France, Italy and Spainwhere the totaltonnagebyroutewasmorethan100,000tin2013(Table2).ThehighesttotalcargotonesflownonrouteswithintheEUwerebetweenGermanyandtheUnitedKingdomin2013(224,815t).TherouteswithinEUwithlowertonnagei.e.between50,000tand100,000twere:Germany‐Belgium(62,554t),theUnitedKingdom‐France(59,590t),Germany‐Sweden(59,567t)andtheUnitedKingdom‐Belgium(57,692t).

Table2.AircargoflowswithinEUmembercountries,2013,(Source:Eurostatdatabase,2015)

Aircargoflow Totalaircargo(t)

Aircargoflow Totalaircargo(t)

Aircargoflow Totalaircargo(t)

UK‐Germany 224,815 Austria‐Germany 38,755 Germany‐Hungary 22,528

France‐Germany 177,192 UK‐Ireland 38,429 Spain‐Netherl. 21,829

Italy‐Germany 134,524 Germany‐Netherl. 36,210 Sweden‐Belgium 20,143

Spain‐Germany 100,244 Spain‐France 32,033 Portugal‐Germany 17,479

Belgium‐Germany 62,554 UK‐Italy 31,432 UK‐Netherland 16,841

UK‐France 59,590 France‐Belgium 30,050 Sweden‐Finland 15,743

Sweden‐Germany 59,567 Denmark‐Germany 29,074 Hungary‐Luxemb. 15,335

UK‐Belgium 57,692 UK‐Spain 24,339 Ireland‐Germany 13,526

Poland‐Germany 47,276 Spain‐Belgium 24,206 Spain‐Portugal 12,860

Italy‐Belgium 44,752 Greece‐Germany 23,760 Germany–Slovakia 12,468

Italy‐Luxemb. 43,393 UK‐Luxemb. 23,347 Denmark‐Netherl. 12,266

Italy‐France 41,647 Germany‐Finland 22,856 Denmark‐France 10,503

2.2IntercontinentalAirCargoFlows

Intermsofweight,EUmembercountriesexported(52%)moreairfreightthantheyimported(48%). In absolutenumbers, they exported5,559,173 tonnesand imported5,109,115 tonnes.PrimaryroutesforairfreightinandoutoftheEUarethetransatlanticroutestoandfromtheUnitedStatesofAmerica(USA)forbothimportsandexports,andalsoroutestoandfrommajorAsianregions(EasternAsiaandNearandMiddleEastAsia)(Figure1a).

Figure1.Totalfreightin2013byinternationalroutesbetweenEUmembercountriesand:a)partnerworldregions;b)partnerworldcountries,(Source:Eurostatdatabase,2015)

Afteraperiodofdecliningdemand(2008‐2010),theEU–NorthAmericaflowstartedtorecoverin 2011, but in 2012 air trade fell by 5.47% and 3.34% in 2013. The total air cargo tradebetweentheEUandNorthAmericaaccountedfor2.76milliontonnesin2013andthisvolumewasstillbelowitspeakof3.3milliontonnes in2007.However, theEU–NorthAmerica flowisstill the strongest air cargo flow within the regions and accounts for approximately 6.6% of

24%

11%

8%

5%5%3%

3%3%

3%

2%

2%

2%

2%2%

2%

23%

United States

United Arab Emirates

China (except Hong Kong)

Russia

India

Hong Kong

Qatar

South Korea

Japan

Turkey

Canada

Saudi Arabia

Brazil

Thailand

Azerbaijan

Other

26%

22%

19%

6%

5%

4%

4%

3%2%

2%2% 2% 2%

1%Northern America

Eastern Asia

Near and Middle East Asia

European Republics of the former Soviet UnionSouthern Asia

Other european countries

South America

Asian Republics of the former Soviet UnionEastern Africa

Western Africa

Southern Africa

Central America and Caribbean

Northern Africa

Central Africaa) b)

world air cargo tonnage and 8.4% of theworld’s tonne‐kilometres (Boeing, 2014). Air cargoflowstoAsianregionsweremorestableintherecentyears.Afteraslightdecreasein2009,aircargotradebetweentheEUandtheEasternAsia,aswellasbetweentheEUandtheNearandMiddle East Asia, increased continuouslywith an average growth rates of 5.16% and 5.69%,respectively.CountrieswithhighestaircargotradewiththeEUaretheUnitedStates,theUnitedArab Emirates, China and Russia that together account for almost 50% of the totalworld aircargotrade(Figure1b).TheUnitedStateshad90%shareofNorthAmerica’saircargoexportsfromtheEUand91%shareofaircargoimportsfromthisregionintotheEUduring2013.ThetotalaircargotradeontheEU‐to‐USdirectionfellby6%in2012and3.25%in2013.Analysingonacountrylevel,themajoraircargotraderoutesarebetweentheUSAandGermany,theUSAandtheUnitedKingdom,GermanyandtheUnitedArabEmirates,GermanyandRussia(Table3).AllthesemainroutesbetweenEUmembercountriesandcountriesoutsidetheEUaremoreorlesstwo‐waybalanced(cargovolumethatisexportedisclosetocargovolumethatisimported)except the routes Germany‐India and the United Kingdom‐India where the import wasmorethan50%higherthanexportandtherouteGermany‐SouthKoreawhereexportwasmorethan70%higherthanimport.AircargovolumeonallmainroutesfromGermanytoothercountriesoutsidetheEUdecreasedin2012,exceptontheroutebetweenGermanyandHongKong,whereaircargovolumeincreasedby12%.

Table3.MainaircargoflowsbetweenEUmembercountriesandothercountriesintheworld,2013,(Source:Eurostatdatabase,2015)

Countries Total(t) Import(t) Export(t)

Germany‐USA 726,274 309,275 416,999

UnitedKingdom‐USA 675,571 330,884 344,687

France‐USA 280,756 118,544 162,212

Netherland‐USA 275,553 149,152 126,401

Luxemburg‐USA 162,657 76,126 86,531

Belgium‐USA 144,867 84,258 60,609

Germany‐UnitedArabEmirates 357,526 189,479 168,047

UnitedKingdom‐UnitedArabEmirates 230,235 118,961 111,274

Netherland‐UnitedArabEmirates 134,056 68,241 65,815

Germany‐Russia 324,729 177,119 147,610

Germany‐China(exceptHongKong) 290,017 123,569 166,448

Netherland‐China(exceptHongKong) 259,922 136,113 123,809

France‐China(exceptHongKong) 108,580 47,388 61,192

Germany‐India 232,573 140,076 92,497

UnitedKingdom‐India 126,510 77,137 49,373

Germany‐HongKong 112,118 59,037 53,081

Germany‐SouthKorea 110,802 40,699 70,103

Germany‐Turkey 106,678 54,199 52,479

In2013, flows fromGermany toChinaand India recoveredandrecordedagrowthof12.74%and4.75%,respectively.AircargotradebetweenGermanyandHongKongstayedstablein2013with a growth of 13.53%. The strongest air cargo flow on the intercontinental level is stillbetweenGermanyandtheUSA,despitetherepeateddeclineinvolumetonnesrecordedin2012and2013(5%and6%,respectively).

3.CONCLUSION

The recovery of air cargo transport in EU has been very slow after a strong impact of theeconomiccrisisin2008.Unlikepassengerairlines,airlinesinaircargotransportarefacedwithadditionalproblemssuchas lowloadfactorthat isbelow50%andlowyields,bothcausedbyovercapacityintheairfreightbusiness.However,recentforecastgivenbyreputableinstitutionsand companies (IATA, Boeing, AEA, Airbus, etc.) are suggesting stronger performance of aircargointheyearstocome(worldwideairfreightvolumeswillexpandatanannualrateof5.2%through 2030). Carriers in all other regions already improved their transport and financialresults,whichisalsoexpectedforcarriersintheEUdespitethechallengestheyarefacedwith(the ongoing sanctions against Russia, the Eurozone economy recession etc.). The possiblemeans,butnotleast,bywhichEUcargoairlinescanimprovetheirperformancesareincreasedconnectivityandefficiency through theirnetworks(e.i.wherepossible introducingconnectingflights to cover airports where demand is insufficient to justify non‐stop flight and bettermatchingcapacitywithdemandusingnewfreighters).Moreover,resultspresentedinthispapershowthattheyshouldfocusontheroutesthathavemoreeventwo‐waybalanceandtobecomemorecompetitiveonthedomesticandintra‐EUroutes.SignificantgrowthofaircargodemandisexpectedontheroutesbetweentheEUandAsia(5.3%annually,Boeing2014) indicatingthatthisregionwillstillbethekeydriverforshapingthepatternofEUairlinenetworksinthefuture.

ACKNOWLEDGMENT

This researchhasbeen supportedby theMinistryof Science andTechnologicalDevelopment,RepublicofSerbia,aspartoftheprojectTR36033(2011‐2015).

REFERENCES

[1] AEA, (2013). Traffic Growth ‐ Will the Infrastructure Cope?, Association Of EuropeanAirlines.http://www.aea.be/news‐media‐room‐media‐centre.html,(December,2013).

[2] Boeing,(2014).WorldAirCargoForecast2014–2015.www.boeing.com/,(January,2015).

[3] Eurostat.http://ec.europa.eu/eurostat/data/database,(January‐March,2015).

[4] Hui,G.W.L.,Hui,Y.V.,Zhang,A.,(2004).AnalyzingChina'saircargoflowsanddata.JournalofAirTransportManagement,10(2),125‐135.

[5] Hwang, C. C., Shiao, G. C., (2011). Analyzing air cargo flows of international routes: anempiricalstudyofTaiwanTaoyuanInternationalAirport.JournalofTransportGeography,19,738–744.

[6] IATA, (2014). Annual Review 2014, International Air Transport Association,www.iata.org/.../iata‐annual‐review‐2014‐en.pdf,(January,2015).

[7] WorldBank,(2015).http://data.worldbank.org/,(March,2015).

[8] Zhang, A., (2003). Analysis of an international air‐cargo hub: the case of Hong Kong.JournalofAirTransportManagement,9,123–138.

ANOVERVIEWOFEUROPEANAIRCARGOTRANSPORT:THEKEYDRIVERSANDLIMITATIONS

JovanaKuljanina,*,MilicaKalića,SlavicaDožića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Theaircargoindustryhasbeenrecordingarapidgrowthinlastthreedecadeswithanaveragerateofalmost4.9%peryear,thusremainingthevitalcomponentofoverallglobecargoactivity.Economicexpansionofcountries in theFarEastand their intensive international tradewith other regions (primarily North America, Europe and the Middle East) have boosted thedemandforefficientandreliablecargotransportrevealingtherisingimportanceofairtransport.Inthispaper,regressionanalysisisemployedtoexplorethemostsignificantfactors(suchasGDP,internationaltrade,export‐importsofgoods,etc)thatcouldhaveacrucialimpactondemandforaircargotransportinthecountriesoftheEuropeanUnion.Althoughfeaturedbyanevidentdropinthe volume of air cargo traffic caused by severe economic crisis occured in 2008, Europe stillconstitutesanimportantlinkinthechainbetweentheEastandtheWest.

Keywords:Aircargo,aircargodemand,regressionanalysis.

1.INTRODUCTION

Air cargo sector plays a prominent role in the era of risingmobility and trade of goods andservicesworldwide.Theneweconomyproducts,consistingofsmall,light,compact,highvalue‐to‐weight parts and assembled components have been shipped internationally by air in areliableandfastmanner(KasardaandSullivan,2006),whichplacesaircargosectorinthefirstplace by value of today’s world trade. In its report, Boeing (2014) estimates that air cargoconstitutes only a 1% of world trade calculated by tonnage, but about 35% of world tradecalculatedbythevalueofgoodsshipped.

However, several factors have been crucial in catalyzing air cargo’s† consistent developmentthroughthesecondhalfofthe20thcentury.First,accordingtoAirTransportWorld(2014)theintegration of global markets followed by transatlantic and transpacific air shipments ofelectronicshadanenormous impactonrebuildingEuropeanddeveloping Japan,SouthKorea,Taiwan,HongKongandSingaporeintoeconomicpowersinthepost‐WorldWarIIera.Second,aneconomicgiantsuchasChinathathasevolvedfromacentrallyplannedeconomytoonewitha greater share of free enterprises has become a leading country around the globe by totalexportvolumeinthe1990sandearly2000s.Inalignmentwiththeprofoundtransformationofeconomy,whichcouldbecharacterizedasmarket‐oriented,Chinagraduallydeveloped fromalow‐income country to ahigher income levelone.These twochangeswill shape theair cargoindustryintheyearstocome(Huietal.,2004).ThehighlypopulatedcountriessuchasIndiaand * j.kuljanin@sf.bg.ac.rs † Inthisresearchtheterms“aircargo”and“airfreight”willbeusedinterchangeably.

Brazilhavealsoadoptedmoreopentradingpolicieswhichinducedsubstantialdemandforaircargotransport.

Certainly, the challenging circumstances (i.e. severe economic crisis) prevailing in the pastseveralyearshadweakenedtheentireworldeconomyandtradegrowthwhichcausedaircargotostagnate,particularlyduringtheperiodfrommid‐2011toearly2013.Notsurprisingly,alongwith the recovery of world economic activity (primarily in the United States and China)measuredbyGrossDomesticProduct(GDP),IATA(2014)claimsthattheoutlookforaircargomarketshasstartedimprovingagain.

Theaimof thispaper is to revealmajoreconomicdriversatmacro level that couldaffect thedemandforaircargosectorinEurope,havinginmindtheirhistoricallyfierceinterdependency.Abriefoverviewofkeydrivers isgiven inSection2.Dataandmethodologyusedaregiven inSection 3, followed by subsection which explains the obtained results. Finally, concludingremarksarepresentedinSection4.

2.ECONOMICFACTORSANDTHEIRIMPACTONAIRCARGODEMAND

Performanceforaircargotransportishighlysusceptibletothestateofglobaleconomyanditsvolatilenature.Forexample,fuelpriceshavehistoricallybeenapersistentproblemforaircargowhichdivertedalargeportionofgeneralcargotolessexpensivecompetitormodeoftransportduringtheperiodsofoilcrisis.Furthermore,theglobaleconomicdownturndrovefreightyieldsdownto22.4%in2009(Boeing,2014).Ontheotherhand,thepositiveeffectofeconomyonaircargodemandare evident in the countries of theFarEast, particularlyChinawhichdomesticmarket will be the fastest growing with a 6.7% average annual growth according to Boeing(2014) forecast for the period from 2013 to 2033. The rising economy of China and othercountriessituatedintheFarEastwiththeirintenseinternationaltradingactivitieswillalsohavea positive stimulus to world air cargo growth in the future. It is worth mentioning that theleading east‐west air cargomarkets are those betweenAsia–NorthAmerica, Europe–Asia andEurope–North America, each of which accounting for 21%, 20% and 8% of total 208 billionrevenuetonne‐kilometers(RTK)in2013.Aircargotrafficinaspecificcountryhighlyreliesonthe volume of export and import with other countries, as well as on transportation cost,exchangeratesandrelativeprices(Fig.1). Inthatsense,aircargotransportconstitutesoneofthemajorfacilitatorstotradeactivitiesbetweencountries,mainlythedistantones.

Figure1.Tradingactivitybetweencountriespullstheaircargotraffic(Boeing,2014)

Conversely, the development of air cargo could boost the global economy, as it enables theconnectivity of a country to other markets in an efficient and reliable manner. The relationbetween air cargo demand and economic activity has been very well documented in theliterature(KasardaandGreen,2005;ChangandChang,2009)indicatingthataircargovolumeandGDPpercapitaaremutuallyinterdependentandcausal.Althoughhighlyinterdependent,air

cargogrowthmostlyoutpacedtheGDPgrowthandtradegrowth.Theevidenceforthiscanbefound in the analysis by Kasarda and Green (2005) who used World Bank database for 68countriesandshowedthattheGDPgrewby72%,tradeby132%andaircargoby302%intheperiodbetween1980and2000.Alongwithliberalizationofmarketsstipulatedthroughalargenumber of bilateral agreements between countries, air cargo sector could significantlycontribute to inflows of foreign direct investment (FDI) and consequently foster county’seconomic activity. In the near past, this was the case particularly with Dubai Airport in theUnitedArab Emirates that use its advantageous geographic position, placed halfway betweenEurope and Asia, and airport free trade zone to attract companies looking to invest in theEmirates.Nowadays,DubaiAirportisthelargestaircargocenterintheMiddleEastandoneofthelargestreexporthubsintheworld.

3.OVERVIEWOFEUROPEANAIRCARGOSECTOR

About14.4milliontonnesofairfreight(bothnationalandinternational)werecarriedthroughairportswithintheEU‐28in2013,markingaslightincreaseof0.4%whencomparedwith2012(Eurostat, 2014). However, the share of air freight in the total freight transported by eachcountry within the EU remained sufficiently low ranging between 0.01% for the countrieslocatedintheCentralandEasternEurope(CzechRepublic,Slovakia,Slovenia,Bulgaria,CroatiaandRomania)toalmost1%forLuxembourg.Cargolux,oneofthemajoraircargocarriersintheworld, has its hub at Luxembourg’s Findel Airportwhich significantly facilitates internationaltradebyair.Ontheotherhand,Germanyisaleadingcountryintermsoftotalvolumeoffreighttransportedbyairaccounting for4.4milliontonnesof freight in2013, followedbytheUnitedKingdom(2.3million tonnes) andFrance (1.8million tonnes).These three countries togetherwith Italy and the Netherlands have consistently accounted for approximately 70% of allEuropean air trade with North America. Nevertheless, the intra –Europe air cargo marketcomprises approximately 3% of world’s air cargo tonnage, but because the region isgeographicallycompactitcomprisesonly0.8%oftheworld’stonne‐kilometers(Boeing,2014).

Europeisstillanimportantpartnerforthemajorityofregionsintheworld.TheEuropeanUnion(EU) remains an important tradepartner for LatinAmerica, secondonly to theUnited States,and is also the region’s leading source of FDI (Boeing, 2014). Furthermore, overall air cargotraffic between the Middle East and Europe has been growing over the last 10 years by anaveragerateof5%,accountingforavarietyofcommoditiesshippedonbothdirectional flows(such as telecommunication equipment, machinery, and finished goods, perishables andgarments).EuropeistheprimarydestinationforAfricanaircargoaccountingforabout2/3ofthe total. African exports are typically counter‐seasonal cut flowers and other perishables toEurope, but there is relatively little return cargo (WorldBankGroup, 2009). Finally, Europe–Asiamarketcomprisesapproximately19.6%oftheworld’saircargotrafficintonne‐kilometersand10.0%intonnage(Boeing,2014)andisexpectedtohavesteadygrowthinthefuture.

Fromalltheabovefacts,itisevidentthateconomicactivity,followedbyindustrialactivity,playakeyroleingeneratingdemandforaircargoinanycountry.EuropeanUnioncountriesareverydiverse in terms of the level of economic development and broader commercial policyenvironment which strongly affect their respective international trade, and subsequently thevolume of freight carried by air. For example, nominal GDP per capita that can be used as arough indicator for the relative standard of living among member states, with Luxembourghavingthehighest(EUR83,400)andBulgariahavingthelowest(EUR5,500)in2013.Betweenthese two extremes, there are a number of variations across EU counties and some of themseverelysufferedfromtheEuropeandebtcrisisthateruptedinthewakeoftheGlobalfinancialcrisis(Greece,Spain, Ireland,CyprusandPortugal).Ontheotherhand,Germany,asoneofthestrongesteconomyintheworld,continuestoachieveverypositivefinancialresultsevenintheperiodofcrisis.HavinginmindthatGermanyisthethirdlargerexporterintheworld,itisnot

surprisingthatairfreightsectorrecordedanimpressivegrowthrateby6.5%intheperiodfrom2002to2013.

Regressionanalysiswasemployedinordertocapturetheimpactofunderlyingeconomicfactorson thedemand forair cargoacross the countriesof theEuropeanUnion.Themodeldesignedassumes a uni‐directional relation between the air freight volume and several economicindicatorsdeemedascrucialdrivers.

3.1Dataandmethodology

In order to create a robust model for air cargo demand on the European Union level, theimportance of several explanatory economic indicators was examined. As it can be found inprevious researchon this issue,GDPper capita, ForeignDirect Investments and Internationaltrade (measured as the sum of import and export) have been selected as key drivers for aircargodemand.Thegrowthratesofthosefourvariablesfortheperiodbetween2004and2013areshowninFigure2.

Figure2.Growthrateforaircargofreight,GDPpercapita,FDIandInternationaltrade

As it can be observed from Figure 2, growth in International trade has substantiallyoutperformedGDPgrowth,whichisalsothecasewithaircargogrowth(exceptforthelastthreeyearswhenaircargoinEuropehasexperiencedasteadydecline).

Thecross‐paneldatawasusedtoexploreanydifferencesbetweenvariablesselectedintermsoftime period. Three temporal intersection including 2002, 2007 and 2013 were selectedprimarily to investigate potential changes in correlation between air cargo demand and itsexplanatory variables across EU member countries. The temporal intersections are inaccordance with EU enlargements that have occurred in the last decades. In 2002, the EUconsistedofonly15memberstatesfollowedbytheentranceofformercommunistcountriesoftheCentralandEasternEuropein2004,2007and2013.Sincenosignificantchangeshavebeenrecorded during these three years in terms of correlations between variables, for the sake ofsimplicitythemodelwillbeestimatedfor2013only.

AirfreightdataofthecountriesintheEuropeanUnionoriginatesfromEUROSTATdatabase.Airfreightstatisticsarecollectedforbothfreightandmailloadedandunloadedonallcommercialflights in scheduleandnon‐schedule service.Air freightdatacanbedivided intonational andinternationallevel,dependingonthescopeofresearch.AnnualdataareavailableforthemostofEUmemberstatesfortheperiodfrom2003onwards,whilesomecountrieshaveprovideddataever since 1993. The statistics that are collected are also available formonthly andquarterlyperiods,butforthepurposeofthisresearchonlyannualdataareused.

Inadditiontoairfreightdata,EUROSTATprovidesanabundanceofhistoricaldataoneconomicindicatorsfromwhichfourabovementionedhavebeenselected‐GDPpercapita,ForeignDirect

Investments(netinflows),ExportsofgoodsandservicesandImportsofgoodsandservices.Thevariable called International trade is derived as a sum of Exports and Imports of goods andservices.AllvariableshavebeenexpressedinthevalueofcurrentEUR.

3.2Results

In order to avoid multicollinearity in the regression model, it is necessary to explore thecorrelationbetweeneachpairofvariablesselected.Multicollinearityisaphenomenoninwhichtwoormoreexplanatoryvariablesinamultipleregressionmodelarehighlycorrelatedcausingspuriousresults.Table1presentsthebasiccorrelationbetweenthevolumeoffreightandmailand threeexplanatoryvariables: (1)GDPper capita, (2) International tradeand (3)FDI–netinflows.Basedontheseresults, itcanbeconcludedthataircargovolumeacrossEUcountrieshighly correlateswith the international trade (r=0.96) confirming the previous findings fromrelevantliteraturethatcountrieswithhigherexportandimportactivityarelikelytoshowmoreneeds for air cargo transport. As expected, FDI is seen as a good indicator with correlationcoefficient accounting for 0.75. However, the correlation coefficient between FDI andinternationaltradeissufficientlyhigh(r=0.72),whichisaseriousindicationtoomitoneofthesetwo variables from the further regression model. Although GDP per capita is in positivecorrelation with air cargo volume, this coefficient (r=0.34) is still low which disenables thisvariabletobe includedintofurtherconsideration.Thiscanbeexplainedbythe fact thatsomecountrieswithslightlylessGDPpercapita(suchasGermany,France,theUnitedKingdometc.)havehighervolumeofair freight compared to thosecountriesplacedon the topof the list intermsofGDPpercapita(suchasScandinaviancountriesandLuxembourg),butwithlessvolumeofaircargo.

Table1.Correlationbetweenselectedvariables

Freightandmailbyair

GDPpercapita Internationaltrade

FDI,netinflows

Freightandmailbyair

1

GDPpercapita

0.34 1

Internationaltrade

0.97 0.28 1

FDI,netinflows

0.76 0.38 0.72 1

Therefore, the regressionmodel proposed consists of only one variable– International trade.SkewnesscoefficientforAirCargovolumeandInternationaltrade(Int.Trade)account2.7and2.5 respectively. From the statistical perspective, highly skewed variables have to betransformed into one that ismore approximately normal. The convenientmethod to do so islogarithmictransformationwhichwascarriedoutinthiscase.Therobustmodelispresentedbythefollowingequation:

0.92 1.14 . (1)

The accompanying statistics are the following: R2=0.75, Fstatistics=82.36 (with significance of2.1887E‐09).Theinterpretationof“log‐log”regressionmodelisgivenasanexpectedpercentageof change independentvariablewhen independentvariable increasesbysomepercentage. Inotherwords,multiplyingInternationaltradeby10willmultiplytheexpectedvalueofAirCargo

volumeby101.14.Finally,byknowingtheinternationaltradeofagivencountry,onecaneasilypredictthevolumeoffreightcarriedbyairinthefuture.

4.CONCLUSION

Thepaperprovidesarobustregressionmodeltoestimateaircargodemandacrosscountriesinthe European Union in 2013. The results reveal that level of international trade remains themostimportantdriverthathighlycontributestogenerationofadditionaldemandforaircargotransport.Inotherwords,aslongastheexportandimportincreases,thedemandforaircargowillrise.Althoughminorintermsoffreightvolumecomparedtoothermodesoftransport,aircargo sectorwill present the vital component in the region’s overall economic activity in thefuture.However,theeconomicgrowthintheEuropeanUnionwillseeseveralmajorlimitationssuchasagingpopulations,uncompetitivelabormarketsandeconomiccrisiswhichwillcertainlyhave substantial influenceonair cargoperformance in the long‐termperiods.Nevertheless, itwouldbechallengingforfurtherresearchestoinvestigatereverserelation‐theimportanceofaircargoexpansiononeconomicgrowthacrosscountriesoftheEuropeanUnion.

ACKNOWLEDGMENT

This researchhasbeen supportedby theMinistryof Science andTechnologicalDevelopment,RepublicofSerbia,aspartoftheprojectTR36033(2011‐2015).

REFERENCES

[1] AirTransportWorld,(2014).Glimmersofhope.

[2] http://atwonline.com/airlines/glimmers‐hope.(February,2015).

[3] Boeing,(2014).WorldAirCargoForecast2014–2015.www.boeing.com/(January,2015).

[4] Chang,Y.‐H.,Chang,Y.‐W.,(2009).Aircargoexpansionandeconomicgrowth:Findingtheempiricallink.JournalofAirTransportManagement15(5),264‐265.

[5] Eurostat,(2014).Freighttransportstatistics.

[6] http://ec.europa.eu/eurostat/statistics‐explained/index.php/Freight_transport_statistics(January,2015).

[7] Hui,G.W.L.,Hui,Y.W.,Zhang,A.,(2004).AnalyzingChina’saircargoflowsanddata.JournalofAirTransportManagement10(2),125‐135.

[8] IATA (2014). Annual Review 2014, International Air Transport Association,www.iata.org/.../iata‐annual‐review‐2014‐en.pdf.(January,2015).

[9] Kasarda, J.D.,Sullivan,D.L.(2006).AirCargo,Liberalization,andEconomicDevelopment.AnnalsofAirandSpaceLaw,31,1‐26.

[10] WorldBankGroup,(2009).AirFreight:AMarketStudywithImplicationsforLandlockedCountries.

[11] http://siteresources.worldbank.org/EXTAIRTRANSPORT/Resources/515180‐1262792532589/6683177‐1268747346047/air_cargo_study.pdf(January,2015).

DELIVERYPLANNINGINCITIESFORCEPITEMS

DraganaŠaraca*,JovanovićBojana,KopićMiloša,DumnićSlavišaa

aUniversityofNoviSad,FacultyofTechnicalSciences

Abstract:Planning thedeliveryanddistributionofcourier,expressandparcel (CEP) itemshastendmore‐than‐currenti.e.implicitconnectionwiththedevelopmentofinformationsocietyinsidewhich telecommunications progress lead to restructuring delivery time and space in away toensurethesustainabilityofclassicpostalservicesinalongerperiodoftime.Implementationofnewtechnologies in thedeliveryprocessofCEP itemsdeservesmoreattention inurbanplanninganddevelopment of the information society and e‐commerce. Standard delivery of CEP items is nolongersatisfactory.ThispaperdealswithplanningthedeliveryofCEPitemsincitiesbyautomatedpostalbooths.Thepaperwillconsiderexamplesoftheapplicationofthistechnologyintheworld.Itwillpresenttheresultsofcustomersurveyanditwillproposelocationparametersforautomatedpostal booths. The aim is to propose concrete solutions and consider the justification for usingthem.

Keywords:planning,delivery,courier,express,parcel

1.INTRODUCTION

Global electronicmarket isdeveloping rapidly.As thenumberof e‐users grows, sodoes theirexpectations.Itisforecastedthatinthenextthreeyearstheinternationalelectroniccommercewillincreaseby150%,whilethesalesthroughe‐commercewillreachseveraltrillionUSdollarsby2016.Growthofe‐commerceimpactssurgeinpopularityofalternativewaysofdeliveryanddistributionofCEPitems.

AlternativewaysofdeliveryanddistributionofCEPitemsaregrowingandtheyprovideeasiercollectionofitems.Accordingtoresearchinthepostoffice,userswanta24/7solutionsthatarefastandaffordable.StandarddeliveryofCEPitemsisnolongersatisfying.

Themarketofe‐commerceisturningtowardself‐servicedevicesforthedeliveryofitems.

These devices eliminate most common problems faced by users, postal operators and other.Thereisnoneedtowaitforacourier,deliveryissuccessfulinthefirstattempt,shippingcostsaredecreasedandtheavailabilityofpostalservicesisincreasedbecausecustomerscancollecttheiritemonthewayhomeorataconvenientlocation,within24hours,7daysaweek.Usingmodern technology to deliver CEP items resolves problems of returning items, fees andproblemsofreverselogistics.

This paper deals with planning the delivery of CEP items in cities by the implementation ofautomatedpostalbooths.Thepaperwillconsiderexamplesoftheapplicationofthistechnologyintheworldandpresenttheresultsofacustomersurvey.Wehaveanalyzedthesavingsin:the * dsarac@uns.ac.rs

costofoperatorswhoprovidethisservice,inthecostsofoperatorswhowouldusethesedevicesonthebasisofacontractwhichallowsnetworkaccess,thecostofuserscollectingtheir itemsduetofaileddelivery.Theaimistoproposeconcretesolutionsandconsiderthejustificationforusingthem.

2.24/7SOLUTIONAPPLICATIONINTHEDELIVERYOFPACKAGES

Postalnetworkof automatedpostalbooths is the largest internationalnetworkof self‐servicemachineswhichenabledispatchandcollectionofpackagesataconvenientlocationfor24hoursaday,sevendaysaweek.

These solutions can already be found in 21 countries: Australia, Chile, the United Kingdom,Ireland, Iceland, Italy, France, Latvia, Lithuania, Ukraine, Estonia, Poland, Russia, Cyprus,Slovakia,CzechRepublic,Colombia,SaudiArabia,CostaRica,SalvadorandGuatemala.Therearecurrentlymorethan3,500unitsofpostalboothsthatarebasedonpostaltechnology.Inthenearfuture,additional6,000automatedpostalmachineswillbeputinEurope,Asia,America,NorthAfricaandtheMiddleEast(www.postaltechnologyinternational.com).

In the last 12 years, the companyKebaAGhas released around3000KePol packetmachineswith over 250.000 boxes. KePol machines are considered the best in the market due to thelargest number of installedmachines in theworld, top quality and a high level of safety andvalueretention.Thecreationofautomatedsolutionsforthefirstandlastmileisamilestoneinthelogisticsof itemtransportation.Kepolsystemispresent inGermanywithmorethan2,500devices(DHLPackstation),Denmark(Doegnposten)withover300devicesaswellasinNorway(Postautomat),inVienna,Austria,Russia,Turkey,Spain,Lithuania,theCzechRepublicandothercountries.

By automating item delivery, costs of delivery services are being reduced while providing amorecomfortabledeliverytousers.Theconceptof thisdeliverysystemusingpostalbooths issimple and acceptable from the standpoint of both users and the postal operators. User canordergoodsovertheInternetandpackagewithhisgoodswillbedeliveredthroughasystemofpostal booth. Via e‐mail or SMS message, the user sets requirements on how they want thepackagetobedelivered.Collectingyourpackageisindependentoftheworkinghoursofthepostofficeanddeliveryservice.Systemscontaincompartmentsofvarioussizes,whosenumbercanbeadaptedtouserneeds.

Automatingproductdeliveryhasnumerousadvantagesbothforusersandpostoffices:

it complements the classic way of delivery and physically separates the courier fromusers;

systemisindependentofthepostofficeworkinghoursandisavailable24hours,sevendaysaweek;

onesystemmaybeusedbyvariousdeliveryservicesand systemiseasily integratedintoexistingdeliverysystemincludingtrack&tracesystem

tomonitortheshipments.

Almost a year ago,AustrianPost has decided to provideuserswith additional possibilities oftheir self‐service with a so‐called "Pick ‐ up stations". "Pick‐up station" actually representsparcelautomatsthatarespeciallydesignedinaccordancewithKebastandardsfor indooruse.Theyareoptimizedintermsofsizeanddesignfor installationin locationssuchasself‐serviceareas,shoppingcentersorpostalshops.

Denmark Post has over 300 parcel automats which mean that automated postal machinesalmostcompletelycoverthecountrythroughout.JointcooperationbetweenDenmarkpostandleading retail chain "Coop" has opened a brand new office space for Keba machines. Parcelautomats in storeswith 35,000 employees and sales revenue of nearly 6.7 billion euros have

made that "Coop" Denmark becomes the largest retail chain in the country. "Pakkeboksen"(Figure1.)havebeeninstalledin300"Coop"storesandaremanagedbyDenmarkpost,i.e.theyallowuserstocollectandsendtheirpackages.InDecember2014,theyhavereachedanall‐timehigh in volume of packages in transfer ‐more than 10,000 parcels a day at parcel automats.ClientsofDanishpostcancollectandsendtheirpackagesduringshoppinginCoop,while"Coop"customerscancollectproductsthattheyorderedonlineattheirlocalstore."Coop"Denmarkhasalsobeenclearlydefinedinthemarketthroughthisservice.Theplanistoexpandthissystemtoatotalof490locations(www.keba.com).

FollowingtheexampleofGermany,Austria,Denmark,Luxembourg,Lithuania,Switzerland,theCzechRepublicandFrance, thenationalpostal serviceofSpainCorreoshasdecided toexpandits serviceswith Kebaparcel automats. Theywill operate under the name "CityPaq"(Figure 2. (www.postaltechnologyinternational.com))andatotalofsixty‐systemswillbeinstalledin2015.ParcelautomatswillbeinstalledingasstationsandinothersimilarplacesinMadridandtownsthroughoutSpain(www.keba.com).

.

Figure1.“Pakkeboksen” Figure2.“CityPaq”(www.keba.com)

"Kouzelné Almara" (Magic Box) is the solution of the Czech Republic,which has been on themarketsinceApril2013andisdesignedforfutureexpansion(Figure3).Itisusedtodelivertheproducts of the selected e‐shoppartners. Themarket of e‐commerce in theCzechRepublic isgrowingrapidly,whichcorresponds to thecurrent increases in thevolumeof items.KouzelnéAlmaraprovides its customers the possibility to take their products at a time that is mostconvenient,24hoursaday,sevendaysaweek.

Figure3.“KouzelnaAlmara”(www.keba.com)Figure4.“AlzaBox”(www.logistika.ihned.cz)

ThissolutionusesKePolFS/08systemwhereeachcabinetisfittedwith54boxes.Thebiggestinternet shop in the Czech Republic (Alza) opts for Keba system. Alza is the largest store ofelectronic goods but also a pioneer in e‐commerce in the Czech Republic. The success ofAlza’sinternetshopsisbasedonthegrowingmarketofe‐commerceintheCzechRepublicand

Slovakia,whichstronglyinfluencedthecompanytobecomeoneofthelargestinEurope.Alzahasactually launched their own businessmodelwith online store for electronics, but now offersgoods of all descriptions. Until recently, Alza customers had to pay for the delivery or topersonally collect their packages from one of the company stores, which also have limitedworkinghours.However,thecompanyrecentlydecidedtooffercustomerstheoptiontocollecttheirpackageatpackagestationswhicharecalled"AlzaBoxes"(Figure4.)whenevertheywant(24/7).Sincethebeginningof2011,theLithuanianPosthasestablishedthelargestnetworkof71self‐serviceterminalscalled"ExpressLP24"in41citiesinLithuania(www.keba.com).

3.THEDEVELOPMENTOFELECTRONICCOMMERCEINSERBIAANDUSERATTITUDES

Assumptions about the development of electronic commerce and information technology, asspecified in the paper of Tomić and Petrović (2011) can be divided into five categories. Firstcategory is based on technical assumptions, electricpowersupply and development oftelecommunications network (speed, the Internet quality and coverage in urban and ruralareas),representationofICTdevices(computers,mobilephones,etc.),availability(price)ofICTinfrastructureandequipment,developmentofroadsandlogisticsinthecountry,inthepaperofVukićević and Drašković (2014).Secondcategoryconsistsoflegalandpoliticalassumptionsinthe form of laws and regulations in the field of e‐commerce: adequate legal protection ofparticipants in e‐commerce, existence of the law on e‐document and the legality of digitalsignatures, the judiciary and legislature follow the development of e‐commerce, promoting e‐commercebygovernmentinstitutions.Thethirdcategoryconsistsofinstrumentalassumptions,which are realized through the prism of services. The fourth group consists of utilitarianassumptions, the usefulness of e‐commerce for customers. The fifth category includes socio‐cultural assumptions: personal preference of electronic communication or communicating inperson;willingness toacceptnewtrends; theexistenceof initiatives,willingness to takerisks,trust in abstract systems, government institutions, the judiciary, people, lawenforcement andconfidenceinthedataconfidentiality,ageandfinancialcapabilitiesofcustomers,needtoseetheliveproductwhichisbeingpurchased.

Asurveyofusers for thispaperwasconducted inorder tocomprehend thesignificanceofallthesecategoriesfromtheviewpointofuserhimself.Theaimofthesurveywastoanalyzetheattitudes of users towards the influence of certain factors on the use of e‐commerce. Thepurposeofthissurveywasalsotoshowhowmanyofexistingusersofpostalparcelandexpressservices is ready to replace the traditional method of delivery with a newmethod of postaldelivery. Based on the data obtained from a survey sample, we have obtained the estimatednumberofuserswhowouldusethemethodofdeliveryviaautomatedpostalbooths.

Usersweresupposedtoprovideanswersto10questions(statements).Answerstooktheformof Likert’s 5‐point scale of attitudes ("1‐strongly disagree", "2‐disagree", "3‐neither agree nordisagree,""4‐agree","5‐stronglyagree").Questionsaregroupedintofivecategories,accordingto factors influencing the development of e‐commerce. The results of the survey will not bedisplayed individually. It is important tonote thatstatements in thecategoryofsocio‐culturalassumptionshavethelowestlevelofagreement.Thestatement"IamwillingtotakesomeriskofbuyingovertheInternet,"hadthesmallestdegreeofagreement,only2.6.Thestatement"It isnotnecessarytoseetheliveproductbeforepurchasingit,"only3.4.Usershaveagreedwiththestatementfromthecategoryofusefulnessofservicestousers,sothatthestatement"Electroniccommerceisacomfortablewayofbuyinggoodsandservices"gotthescoreof4.3.Theanswerstoquestionsfromthefirstcategoryoftechnicalassumptionsarecertainlyworrying.Mostusersespecially in rural areas are not satisfied with the development of telecommunicationsinfrastructure, speed, reliabilityof linksandservices.Therefore,ourresearchwill continue itsfocus on urban areas of cities in Serbia.When it comes to the development of transport andlogisticsinfrastructure,attitudesweredivided,andbasedontheadditionalquestionsduringthe

directcontactwithcustomers,itwasconcludedthatattitudesdependonpreviousexperiencesofusersandoperatorswithwhomtheycollaborated.Mostobjectionsrelatetothedeadlinetodelivertheitem(longerthanthreedays),thepossibilityofcollectingshipmentsinthepremisesofoperators (someoperatorsdonothave theirbranches in smaller townsand thedelivery isdonesolelyat theuser’saddressat the timewhen it isnotconvenient to theuser),payextrapostageforthenextdelivery,etc.Mostoftheusers,namely98%ofthosewhousetheservicesofe‐commerce,haveagreedthattheuseofpostalboothwouldpositivelyinfluencetheirdecisiontopurchase in the future.Whileaccording to the surveyofRAPUS,90%ofusershaveagreedthat their postal item with personal delivery can be delivered to a member of the samehousehold,howeveraccording toour research100%ofusershaveagreed to receivemail viaautomatedpostalbooth(RAPUS,2014).

Havingcompletedthesurvey,wehaveconcludedthefollowing:

itisnecessarytoreducetheriskofbuyingovertheInternet; isnecessarytoprovidetheoptionofreturningtheproduct, isnecessarytoprovidegreateraccesstoICT, itisnecessarytoconductregularqualitycontrolofpostalandlogisticsservices, itisnecessarytoincreasetheavailabilityofnetworksandservicesofpostalandlogistics

operators, it isnecessarytoreducethecostofproductdeliveryandintroducealternativewaysof

delivery.

4.DELIVERYOFCEPSHIPMENTSINCITIESTHROUGHAUTOMATEDPOSTALBOOTHS

Currently in theRepublicofSerbia,onlyPEPostofSerbia is licensed toprovide theuniversalpostal service and therefore is obliged to provide these services throughout the country.Development of alternative modes of delivery of shipments to customers has a specialsignificance to Post of Serbia. However, the Postal Services Act does not allow delivery ofshipments to be done via self‐service machines, such as packet automat (Republic of Serbia,2005).Therefore, furtherdevelopmentofautomateddeliverydependsonchanges in the legalframework.However, recently adopted Regulations on access to the postal network (RATEL,2014)isinfavoroftheapplicationofthesedevices,whichmakesitpossibleforthePosttooffertheresourcesofpublicnetwork tootherpostaloperators.Postalautomatedboothsshouldbepartofthepublicpostalnetwork,designedtodeliverthepackage(whichfallswithinthescopeof universal service) and other items, such as express shipments or logistics shipments. ThepossibilitythatthePostofSerbiaoffersunitsofpublicnetworktootheroperatorsisinfavorofthe sustainabilityofuniversalpostal service.ThePostof Serbiahas about5.2millionexpressand package items per year. Other postal operators have approximately 10,800,000 expressitems per year. Participation of private postal operators is around 68% which is highercomparedtoThePostofSerbia.ThecostofthefirstdeliveryofCEPshipmentsincitiesislowerthanthecostsofCEPdeliveryofshipmentsinruralareas.However,successfulfirstdeliveryinruralareasisalmost100%.Numberofitemstobedeliveredinrepeateddeliveryinurbanareasisover7%.Deliverycostscomprisingthepriceoftheconsignmentare45%,about50dinarspershipment. Of the total number of items 1,200,000 parcels are delivered in repeated delivery,which annually cost operators in Serbia an extra 60,000,000 dinars. For users, regardless ofwhether their item was delivered in the first or repeated (second) delivery, waiting for thecouriermeanswasteoftimeandmoney.Thepotentialmarketfortheseservicesandthenumberof end users is much larger than customers who receive shipments in the second repeateddelivery.SizeofthemarketisproportionaltothenumberofusersthatreceiveCEPshipments.Most CEP shipments come from electronic commerce (about 70% of shipments), and in oursurveythesepeoplehavedeclaredthatthisservicewascrucialforthem.Theconclusionisthat

more than11millionshipmentsperyearmaybedeliveredviapacketautomats,andshippingcostswillbereducedtoaminimum.

5.CONCLUSION

The increase inpostalserviceshascaused theconstant improvementof thequalityofservingcustomers, which is related to the development and deployment of cutting‐edge technicalequipment in the units of the postal network. The growth of e‐commerce contributes to agrowing number of parcelswhich requires implementation of new services, such as deliverysystem of parcels twenty‐four hours seven days a week, which is the basis of this paper.Automationofthedeliveryprocessofpostalitemscontributestoreducingcostsandsavingalotoftimewhichinthepostalactivityisthemostimportantqualityparameter.Withthisserviceitisnolongernecessarytowaitforcourieratthehomeaddress.Thepackagecanbecollectedatanytimeattheconvenientlocationfortheuser.Whenintroducingnewtechnologies,ThePostofSerbia can take advantage of their existing network infrastructure andmarket coverage. ThePostmustadapttothemodernwayofdoingbusiness,primarilyby investing innewtechnicalsolutions.

ACKNOWLEDGMENT

ThisresearchissupportedbytheMinistryofScienceofSerbia,GrantNumber36040.

REFERENCES

[1] Republic Agency for Electronic Communications and Postal Services – RATEL, (2014).Regulationsonaccesstothepostalnetwork.

[2] Republic Agency for Postal Services – RAPUS, (2014). Research on customer needssatisfactiondegreeoftheuniversalpostalservice‐Individuals.

[3] RepublicofSerbia,(2005).Lawonpostalservices.

[4] TomićP.N.,Petrović,D.(2011).,Pravneikulturološkepretpostavkezarazvojelektronsketrgovine u Republici Srbiji, XXIX Simpozijum o novimtehnologijama u poštanskom itelekomunikacionomsaobraćaju–PosTel2011,Beograd,06.i07,65‐72.

[5] Vukićević,S.,Drašković,D.(2014).Analysisofe‐commerceinSerbiabasedonfactorsthatinfluenceitsadoption,YUINFO,1‐6

[6] www.postaltechnologyinternational.com

[7] www.keba.com

[8] www.logistika.ihned.cz

ECO‐DRIVINGAWARENESSANDBEHAVIOUROFCOMMERCIALDRIVERS

MarkoVeličkovića,*,ĐurđicaStojanovića,GoranAleksićb

aUniversityofNoviSad,FacultyofTechnicalSciences,DepartmentforTrafficandTransport,SerbiabSrbijatransporta.d.,Belgrade,Serbia

Abstract:Unwanteddrivingroutines(harshacceleration,harshbreakingandsuddensteering)areamong the key issues for increased fuel consumption, maintenance cost, air pollution andgreenhousegasemissionsreduction.Thispaperfocusesoncommercialvehicledrivers’awarenessabout eco‐driving impacts on environment and costs, their current driving behaviour andapplication of eco‐driving measures by companies in transportation and logistics sector. Eco‐drivingindevelopingcountriesishighlydesirableandreasonableduetoitslowcostsandalmostpermanent effects in comparisonwithother costlyand time‐consumingmeasures (fleet renewalwithhybridandelectric vehicles,environmental impact strategiesandgreen logistics concepts).Fortheresearchpurpose,aquestionnairesurveyisconductedtocollectdataoneco‐drivingskillsandstrategiesoninlogisticsandtransportationsectorinSerbia.

Keywords:Eco‐driving,Fuelconsumption,Greenlogistics,Serbia

1.INTRODUCTION

Eco‐drivingprogramsattempttochangeadriver’sbehaviourthroughgeneraladvice,suchas:donotdrivetoofast;donotacceleratetooquickly;shiftgearssoonertokeepenginespeedlower;maintain steadyspeeds; andkeep thevehicle in goodmaintenance (e.g. check forproper tyrepressure frequently) (Barth & Boriboonsomsin, 2009). Some of these actions drivers doapplicatewithoutknowingwhateco‐drivingmeans.However,notalldriversareawareofallofthe techniques of eco‐driving. Awareness of eco‐driving is the first step towards this conceptimplementation and further development. The aimof this paper is to estimate a level of eco‐drivingawareness intransportationandlogisticssectorand levelof(un)consciousapplicationofeco‐drivingtechniquesonroadsinSerbia.

This paper is organised as follows. After a brief introduction, an insight into the eco‐drivingprinciples and benefits, as well as a brief overview of the recent eco‐driving initiatives andprojects are given in Section 2. The research methodology, main results and discussion arepresented inSections3,4and5,respectively. Finalremarksandconclusionsaregiven in thelastsection.

* marvel@uns.ac.rs

2.RESEARCHBACKGROUND

Eco‐drivingisatermusedtodescribetheenergyefficientuseofvehicles,andreferstothesetofrules, techniques and behaviour that drivers can employ in order to reduce their fuel usage,emissions of CO2 and other pollutants. Vehicle efficiency is not constant and changing driverbehaviourcansignificantlyreducefuelconsumption.Suchbehaviourisoftendescribedaseco‐driving(Mensingetal.,2013).Wehaveidentifiedafewtermsinuse‐rationaldriving,moderndriving, smarter driving, etc. ‐ all referring to similar actions. Additionally, “energy efficientdriving” or “environmentally friendly driving” are the terms often associated with “slowingdown” which is counterproductive, rather than supportive for changing driving behaviour(Schulte,2012a).Therefore,wedecidedtouseterm“eco‐driving”todescribesuchbehaviour.

Numerouspossiblebenefitsofeco‐drivingaredividedintofourgroups:environmental(reducedgreenhouse gas emissions, local air pollutants and noise), financial (reduced fuel, vehiclemaintenance,andcostsofaccidents),social(moreresponsibledriving,lessstresswhiledriving,higher comfort for drivers), and safety (improved road safety, enhanced driving skills)(ECOWILLproject, 2010‐2013).The essenceof eco‐driving is to reduce fuel consumption andstudiesshowedthatsavingsare5‐10%onaverageinlong‐termperiod(Barkenbus,2010).

In transportation and logistics sector, eco‐driving also contributes to avoid situations whendriver of the vehicle with company logo shows too aggressive driving style, which harm thereputation of the company. In order to achieve listed benefits, driving behaviour and drivingtechniquesshouldbeadaptedtotrafficconditionsandtothetypeofvehicle.Ingeneral,thereisarangeofrules,tipsandtricksabouteco‐drivingthatdriversshouldknow.However,fivebasic("golden")rulesofeco‐drivingfordriverstofollowareidentified(Schulte,2012b):

I Anticipatetrafficflow;II Maintainasteadyspeedatlowrpm;III Shiftupearly;IV Checktyrepressuresfrequently;V Consideranyextraenergyrequiredcostsfuelandmoney.

Eco‐driving initiatives in transportation and logistics sector can be differed by levels ofapplication:Policylevel–creatingaunifiedplatforminordertostimulateeco‐driving;Companystrategiclevel–enhancedrouteplanninganddecisionmaking;andCompanyoperationallevel–drivereducation,trainingsandmotivation,aswellasvehicleequipmentandmaintenance.Onlythe last one is considered in this paper. Also, different types of eco‐driving activities can bedistinguished, but the optimal scenario is a blend of these activities (SenterNovem, 2005): 1.Awarenessraising;2.Disseminationanddistributionofinformation;3.Trainingprograms.Weconsidered the awareness of commercial vehicle drivers about eco‐driving as the first andnecessarysteptowardsothermeasuresandactivities.

Therehavebeensomeactionsinthepasttostimulateeco‐drivinginEurope,fromthelegislative(Directive 2003/59/EC, 2003), to the implementation and education area. Target groups of alargenumberofinitiativesweremainlythedriversofprivate(passenger)cars.Greaterattentionto eco‐driving in freight transportation is paid in several projects (ECOEFFECTproject, 2011‐2013; ECOSTARSEurope project, 2011‐2014;RECODRIVEproject, 2007‐2010). The results ofthoseprojects showedenvironmental, financial, socialandsafetybenefits.Thepositiveeffectsareparticularlyrelatedwithemissionreductions,fuelconsumptionandcostssavings.

The first steps of eco‐driving in Serbia were made under the UNDP project "Support toSustainable Transportation System in the City of Belgrade", 2011‐2015 (Plevnik, 2013).However, only five truck drivers were involved into program. Training and testing of truckdriversresultedwithabout7%savingsoftheaveragefuelconsumption(VukovićandĐorđević,2014).Toincludeotherpositiveeffectsandobtainmorereliableconclusions,itisnecessarytocontinuetheresearch.

3.METHODOLOGY

Ashortstructured interview(2‐3minutes) isused fordatacollection.ThesurveywascarriedoutinthefirstweekofMarch2015.Thedriversofallcommercialvehicletypeswereincludedinto the survey. The interviewwas conducted at two level crossingswith the barriers in thesurroundingof thecityofNoviSad, inperiodswhenthebarrierswereclosed.Oncethetrafficflowisstopped,randomdriverswereaskedbytrainedinterviewerstoparticipatethesurvey.Ifthey agreed, the interviewers asked the questions from the questionnaire, explained them, ifnecessary,andfilledintheobtainedanswersfast"insitu".Theinterviewsamplesizeaccountsthe total number of 113 drivers. This method allowed the highest level of response. Only 3questionnaireswerenot finishedduetotheopeningofthebarriers inthemiddleof interviewandtheyareexcludedfromfurtherresearch,sothetotalnumberofanalysedanswersare110.

The questionnaire consisted of three sets of questions. At the beginning, drivers were askedabout theirage,drivingexperience, vehicleand routebasicdata.Furtherquestionsaddresseddrivingbehaviour,habitsandskills,implementationofeco‐drivingmeasuresbycompaniesandwhatmeasuresdriversprefer.Finally,driverswereaskedtoassessdrivingskillsofcommercialdriversinSerbiaandwhethertheyarefamiliarwiththeterm“eco‐driving”.Thepossibilitythatdrivers could give desirable/expected answers from the viewpoint of their company isminimized, because the survey was not conducted in firms. The interviewers have alsoconductedabriefvisualinspectofthevehicleequipmentrelatedwithaerodynamicdesign.Thiswas considered as an indirect supporting indicator of an overall approach of respondents'companiestoenvironmentalissues.

4.MAINRESULTS

Thesurveyintendedtocollectthedataoneco‐drivingawarenessandbehaviourofcommercialvehicledrivers.Thebasic informationaboutdriversandvehicles isshowninFigure1.Awiderangeofdriversiscoveredbythesurveywithdifferentlevelsofexperience–young,old,moreor less experienced. About 62%of drivers hasmore than 10 years of driving experience andmorethan68%ofdriversareolderthan35.Asignificantlyhighpercentageofoldvehiclescanalsobenoticed.Aboutaquarterofrespondentsusuallydriveonthelong‐haulroutes.

Figure1.Basicdataondriversandvehiclesincludedinsurvey

We have investigated at what amount of revolutions per minute (rpm) commercial vehicledrivers usually perform gear shifting, depending on the type of vehicle they drive (Figure 2).Driversofvehiclesequippedwithtachometer(around74%)answeredabouttheenginespeedsthey usually shift gears. Twenty three percent of total number of respondents drive vehicleswith automatic transmission system and a few vehicles with manual transmission are notequippedwith tachometer. The two thirds of the light commercial vehicle (LCV) drivers shiftgear between 2000 and 2500 rpm. Drivers of vehicleswith load capacitymore than 3.5 tons(middleandheavycommercialvehicles–MCVandHCV)mostlyshiftsbetween1250and1500rpm.Whenmaintainingasteadyspeeddriversactdifferent:40%tendtomaintainspeedatlow

rpm,other40%maintainacertainrpm(themostat1250,buttherewereanswersupto3000rpm)and20%maintaindesiredspeedregardlessamountofrpm.

Figure2.Drivers’opinionaboutpreferredenginespeedforgearshifting

Electrical energy is converted from extra fuel burned in a combustion engine, so electricalequipmentcostsextraenergy.Airconditioningisbasicelectricaldevicethatsignificantlyimpactfuel consumption.Driverswereaskedhowoften theyuse theair conditioner–60%neverorrarelyuseit,28%sometimesand12%oftenorconstantly.

DriverswereaskedtoratedrivingbehaviourofotherprofessionaldriversinSerbiaonascalefrom 1‐poor to 5‐excellent. The rough assessment of driving behaviour resultedwith 72% ofanswerswithaverageandaboveaveragerates.Also,alittlemorethan60%ofdriverssaidtheyhad heard of the term eco‐driving, but only 20% of those had a good explanation of what itinvolves.

Attheend,wehaveinvestigatedlevelofeco‐drivingimplementationbycompanies.Even75%ofdriversansweredthattheircompanydonotimplementanyofthemeasures.However,50%saidthat implementation of eco‐driver recognition and rewarding system is highly desirable. Theinvestment into the airdrag reductionaids additionally indicate that the company careaboutenvironmentalimpact.Around30%ofvehicleswerenotequippedwithairdragreductionaidsatall.Roofdeflectorswereon97%andsideextenderson39%ofvehiclesequippedwithanyaid. When it comes to tyre pressure checking, research results shows that 35% of driversperform frequent checks and around 40%do it occasionally. In few cases, another person incompanywasresponsibleforvehicleoperationalmaintenance,sodriverswerenotobligatedtochecktyrepressure.

Airdragreductionaids(roofdeflectors,sideextenders,sidetankfairingsandendfairings)andrecommended tyrepressure reduce fuel consumption, especiallyona long‐haul routesdue tohigherdrivingspeeds.Roofdeflectorsandsideextendersare inmostcasesaddedinavehicleproduction process and are not necessary a result of companies paying attention on air dragreduction.Thesidetankfairingshadonly1%ofvehiclesandthattherewasnovehiclewithendfairingsattached.

5.DISCUSSION

The variety of respondents in terms of age, driving experience, company they work for androutes theyusuallydrivegivesstrengthto thisresearch.Differentshifting techniques for lightandheavycommercialvehicles(seeFigure2)areconsequenceofdifferentpowersourcesanddifferentdesignsoftransmissionsystem–HCVaremostlypoweredwithbigdieselengineswithnarrowpowerbandandalotofLCVarepoweredbygasolineengineswithwiderpowerband.Althoughtheresultsaboutpreferredenginespeedforgearshiftingarenotoutputoftestingandmeasurement,butthepersonalopinionsofdrivers,wecanconcludethatdriversofcommercial

vehiclesperform(ortendtoperform)gearshiftingineco‐drivingmanner.Thesameconclusioncanbemadeaboutmaintainingasteadyspeed.Themostofthedriverswhichaimtomaintaincertain engine speed cited low number of rpm. Based on those results,we can conclude thatalmost80%ofdrivers (ofbothvehicle loadcapacities ‐ lessandmore than3.5 tons)actwithregardtoeco‐drivingrule“lowrpmatsteadyspeeds”.

Ahighpercentofanswersaboutneverorrarelyuseofairconditioner(60%)indicateitsuseinthe eco‐driving manner. This is partly influenced by a high number of vehicles without airconditionerinstalled.Itshouldalsobenoticedthatsurveywasconductedinwinterdays,whichmight have subconscious influence on answers or question misunderstanding. Further, apurpose of commercial vehicle sometimes requires additional (special) electrical equipment.Thisresearchquestionnairewasnotdesignedtocollectspecificdataaboutequipmentusageofsuchdedicatedvehicles(concretemixers,refrigeration,wastecollection,etc.).

Average and above average assessment of other drivers’ behaviour leads to conclusion thatprofessionaldriversgenerallythinktheirdrivingbehaviourisgood,butthatthereisstillaroomforimprovements.Littlefamiliaritywitheco‐drivingindicatesthateducationanddisseminationof eco‐driving concept should be carried out. This can be achieved in different ways:presentations, press andmedia, current and new driver trainings in driving schools, etc. Thedrivers'interestingforthismatterencouragestheconsiderationofthesemeasures.

The average age of vehicles indicate a high level of emissions and therefore, a greaterimportance of the question how they are exploited. While the fleet renewal is cruciallysignificantforenvironmentalimpact,itisinthesametimethemostexpensivemeasure,hardlyextensivelyapplicablebymajorityofSerbiancarriers.Hence,theimportanceofeco‐drivinggetsanadditionalsignificanceinsuchchallengingeconomicenvironment.

The adoption of eco‐driving techniques by drivers may be stimulated by their belief thatcompany policy is eco‐oriented. A systematic implementation of eco‐driving initiatives andprogramsbycompanies in transportationand logisticssector inSerbia isona low level, fromdrivers’ point of view. However, respondentsmostly agreed that company should implementsomemeasures.Eco‐driverrecognitionandrewardingschemeisthemostwantedmeasurebydrivers,soitislikelythatthismeasurewillbewellaccepted.Alowlevelofapplicationofaddedair drag reduction aids may indicate low willingness of companies to invest into theenvironmentallyfriendlymeasures,eveniftheysupportreductionoffuelconsumption.

This research also has a set of limitations. The survey method always indicate a subjectiveviewpointsof respondents.Further, traffic flowanticipationminimizesharshaccelerationandharshbraking,but thisaspectcouldnotbeanalysedby thequalitative research.Therefore, infurtherresearchECUoraccelerometerreadingsshouldbeincludedintotheanalysistoobtainamorecomprehensivepictureofeco‐drivingbehaviourandobtainmoreexactresults.Additionalresearch could be conducted about eco‐driving activities on company strategic level andnational level. Dedicated vehicles should be analysed in order to assess additional electricalequipmentusageandimpactonfuelconsumption.

6.CONCLUSION

Eco‐driving reduces fuel andmaintenancecosts, aswell as the traffic accidents costsdue toasaferwayofdriving.Consequently,risingtheeco‐drivingawarenesscontributetothebusinessperformance improvementof companies.Due to its social andenvironmental significance,wehave conducted a research on eco‐driving awareness of commercial drivers in Serbia. Thegeneralconclusionisthatdriversarenotfamiliarwitheco‐drivingconcept,buttheyoftenusetoapplyoneormoreeco‐drivingtechniques.

While majority of environmental measures in transportation and logistics require additionalcostsandefforts,eco‐driving isamongtherareoneswhichhavemostlycoupledcost‐effective

andenvironmentalpositiveeffects.Thispointmaybeofcrucialimportanceforwiderinvolvingcompaniesintotheeco‐drivingactions.Anincreasingawarenessaboutenvironmentaleffectsofeco‐drivingmightadditionallyincreasethemotivationforwiderimplementationoftechniquesin full range from driver, to company and wider level. This is particularly important fordeveloping countries, with weak economies, where transport and logistics industry facenumerous challenges. Therefore, eco‐driving concept should be disseminated in all possiblewaysanditshouldalsobeapartoftrainingindrivingschools.

ACKNOWLEDGMENT

This paper was realized as a part of the project TR36030, supported by the Ministry ofEducation,ScienceandTechnologicalDevelopmentoftheRepublicofSerbia.

REFERENCES

[1] Barkenbus, J.N., (2010). Eco‐driving: An overlooked climate change initiative, EnergyPolicy,38(2),762‐769.

[2] Barth,M.,Boriboonsomsin,K., (2009).Energyandemissions impactsofa freeway‐baseddynamic eco‐driving system, Transportation Research Part D: Transport andEnvironment,14(6),400‐410.

[3] DIRECTIVE 2003/59/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL,(2003).Theinitialqualificationandperiodictrainingofdriversofcertainroadvehiclesforthecarriageofgoodsorpassengers,OfficialJournaloftheEuropeanUnion,L226/4.

[4] ECOEFFECTproject,(2011‐2013).Link:http://ecoeffect.org/

[5] ECOSTARS Europe project, (2011‐2014). Link: http://ecostars‐europe.eu/en/Home/Home/

[6] ECOWILLproject,(2010‐2013).Link:http://www.ecodrive.org/en/home/

[7] Mensing, F., Bideaux, E., Trigui, R., & Tattegrain, H., (2013). Trajectory optimization foreco‐driving taking into account traffic constraints, Transportation Research Part D:TransportandEnvironment,18,55‐61.

[8] Plevnik, A., (2013). Mid‐term evaluation of the GEF/UNDP project "Support to thesustainabletransportinthecityofBelgrade",FinalReport,UrbanPlanningInstituteoftheRepublicofSlovenia,Ljubljana.

[9] RECODRIVE project, (2007‐2010). Rewarding and Recognition Schemes for EnergyConserving Driving, Vehicle procurement and maintenance. Link:http://www.recodrive.eu/index.phtml?id=1013&ID1=&sprache=en

[10] Schulte, K., (2012a). Ecodriving in Learner Driver Education ‐ Handbook for drivinginstructors,GermanRoadSafetyCouncil,ECOWILLLevel1,DeliverableD3.1,pp.28.

[11] Schulte, K., (2012b). Short Duration Training (SD‐Training) ‐ Handbook for Trainers,GermanRoadSafetyCouncil,ECOWILLLevel2,DeliverableD3.3,pp.24.

[12] SenterNovem, (2005). Ecodriving ‐ The smart driving style, TREATISE project, pp. 31,Utrecht.

[13] Vuković,V.,Đorđević,Ž.,(2014).Uticajobukeekonomičnevožnjenasmanjenjepotrošnjegoriva, XII International Symposium „Road Accidents Prevention 2014”, 09‐10 October,BorskoJezero.

FACILITATINGCROSSBORDERMOVEMENTOFGOODSINTHEREGIONOFWESTERNBALKAN–THEEVIDENCEFROMSERBIA

JelicaPetrovićVujačića*,OliveraMedarb

a,bUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Theauthorselaborateproblemsofbordercrossingpractices in the regionofWesternBalkan countrieswith special focus on Serbia. Among themany barriers to the efficient cross‐bordermovementofgoodsarethecomplexityofprocedures,expensesinbothmoneyandtime,andinsufficienciesininfrastructuresandoperations.Customsandotherborder‐relatedcontrolsimposecostson tradersand induce considerable frictions in cross‐border supply chains.ThepaperalsopresentsandelaboratessomeofresultsoftheprojectConductingsurveysatbordercrossingpointsdone for theBelgradeChamberofCommerceasapartner to theACROSSEEproject.Thesurveysincludedstructuredinterviewsoftruckdriversandtrafficcountingatbordercrossingpoints.Theresearch was conducted in 2013 at 7 border crossings with Croatia, Bosnia, Montenegro,Macedonia,HungaryandRomania.

Keywords:tradeflows,transportflows,bordercrossing,supplychains,Serbia.

1.INTRODUCTION

Intensifying and facilitating trade and transport flows between Western Balkan countriesrepresentsapreconditionfortheeconomicdevelopmentofeachofthem,aswellas,meetingtheconditionsforentryintotheEuropeanUnion(EU).Oneofthemostimportantissuesconcerningtradefacilitation is therelationshipbetweentradecostsandprivatesectorgrowthandexportcompetitiveness, especially for countries in transition. Regional economic cooperation,partnerships in forming industrial clusters, mutual support in meeting the requirements ofcandidacyandthenmembershipintheEUareonlyapartofthehugeagendaofmutualsupportandcooperationintheregionofWesternBalkan(WB).

The mutual interdependence of trade and transport flows points to the need for integrated,coordinatedandharmonizedtransportpolicies.GreatertradeintegrationpriortobecomingpartoftheEUhasmultiplebenefits:(i)countrieswillneedtoaligntotheEUacquisintrade‐relatedareas(freemovementofgoods,servicesandpeople,customs); (ii) firmswillbebetterable tocope with the competitive pressures within the EU; and (iii) national administrations arebuildingupcapacityinregionalcooperation(Handjiskietal.,2010).Higherinterregionaltradewould lead to, other factors being equal, to a higher level of foreign direct investment as theregionwould be perceived as onemarketwith a broader potential. Export led growth is thesolutionforthecountriesofWB(VujačićandPetrovićVujačić,2012).

* j.petrovic@sf.bg.ac.rs

Tradeand transport facilitationaddressesawideagenda in economicdevelopmentand tradethatmayincludeimprovingtransportinfrastructureandservices,reducingcustomstariffs,andremoving non‐tariff trade barriers including administrative and regulatory barriers. Businesscosts may be a direct function of collecting information and submitting declarations or anindirectconsequenceofborderchecksintheformofdelaysandassociatedtimepenalties.Thesecostscanresultinforgonebusinessopportunitiesandreducedcompetitiveness.

2.TRADEFLOWSANDTRADEFACILITATION

Serbia,as theothersWBcountries (Albania,BosniaandHerzegovina,Croatia,FYRMacedonia,Montenegro, UNMIK‐Kosovo) has implemented substantial trade liberalization reforms,including reduction of tariff and non tariff barriers, the elimination of import quotas, thereduction of import licensing requirements and prohibitions, and the restructuring andsimplificationofcustomsprocedures. Inaddition,Serbiahasalso introducedanupdated tariffschedule to reduce trade restrictions faced by nearly all agricultural, fish, and food productsimportedfromtheEU.

Themost importanteconomicpartnerofSerbia is theEUand thenCEFTAwhichabsorbover90%ofSerbianexportsandaccount for3/4ofSerbianimports(Figure1and2). InDecember2006, the countriesof theWesternBalkans signed theCentralEuropeanFreeTrade (CEFTA),whichsubstitutedanetworkofbilateralfreetradeagreements.AfterbecomingmembersoftheEU,RomaniaandBulgarialeftCEFTAin2007,andCroatiainJuly2013.Thecountriesremainingin CEFTA are: Albania, Bosnia andHerzegovina, FYRMacedonia,Montenegro, Serbia, UNMIK‐KosovoandMoldovawhichmakeupforacommonmarketofcloseto23millionconsumers.

Figure1:Importsbycountriesoforigin,2013,USDmillion

Figure2:Exportsbycountriesofdestination,2013,USDmillion

Source:Statisticalofficedatabases

ThestructureofthetradeingoodsofSerbiawiththesecountrieshasbeenfairlyconstantoverthe last years. The exports of Serbia to the WB countries are primarily processed foods,agriculturalrawmaterials,electricpower,coatedmetals,chemicalsandtextiles.Ontheimportsidethelargestshareistakenbyoilandderivatives,naturalgas,paper,cardboardandcelluloseproducts,vegetablesand fruit, ironandsteel.Serbia isrepresentedontheWBmarketswithabroadarrayofproductswithwhichitwouldnotbecompetitiveontheEUmarketsduetohigherstandardsrequired there. In thisrespectSerbianexportsaredominatedbyasmallnumberofproductssuchasiron,steel,raspberries,corn,tires.

415.9

478.5

490.4

576.2

785.7

851.5

1062.7

1201.5

1735.1

2379.3

Croatia

Slovenia

USA

FRYMacedonia

Romania

Montenegro

RussianFederation

BosniaandHerzegovina

Germany

Italy

582.1

603.8

630.7

888.2

972.7

1014.5

1509.6

1903.8

2255.8

2358.7

Romania

France

Austria

Kazakhstan

Poland

Hungary

China

RussianFederation

Germany

Italy

IntradewiththeCEFTAcountriesSerbiahasatradesurplusthankstoagoodmarketpositioninBosniaandHerzegovina,Montenegro,theFRYMacedoniaandUMNIKKosovo,whiletheratioofexportstoimportsbycountrydecreaseswiththegeographicaldistancefromSerbia.Tradeflowsare hindered by too many different types of non‐tariff barriers (complicated procedures onborder crossings, cumbersome administrative procedures and uncoordinated custom andinspectionactivities,aninadequatenumberofinternationallyaccreditedcertificationbodies,aswellas,accreditedlaboratoriesandinstitutions,unrecognizedcertificatesofquality,corruptionandcounterband.ItisnecessarytoupgradetheinfrastructuretothelevelwhentheSerbianandtheproductcertificatesofotherWBcountrieswillberecognizedinallcountriesoftheEUandCEFTA.ThereisalackofinstitutionalizedaccreditationbodiesduetowhichitisnotpossibletoconsistentlyapplytheCEFTAagreement(Nikolićetal.,2011).

The Enabling Trade Index (ETI) was developed within the context of the World EconomicForum’s Industry Partnership Programme for the Logistics and Transport sector. The ETImeasures the extent towhich individual economies have developed institutions, policies, andservicesfacilitatingthefreeflowofgoodsoverbordersandtodestination.TheETIcategorizesthe obstacles into four categories: market access, border administration, infrastructure andoperating business environment (Drzeniek Hanouz et al., 2014). These subindexes arecomposedofthesevenpillars:domesticandforeignmarketaccess,efficiencyandtransparencyof customs administration, transparency of border administration, availability and quality oftransport infrastructure, availability and quality of transport services, availability and use ofICTs,andoperatingenvironment.

IfwecomparetheETIindexfor2014(Table1)andpreviousyearsforSerbiawecannoticethatthesituationhasbecameworse,especiallyfromthepointofviewofmarketaccess,availabilityand quality of transport infrastructure and business/operating environment (Petrović VujačićandMedar,2012).

Table7.EnablingTradeIndex(ETI)2014forSerbia

EnablingTradeIndex2014Rank

(outof138)Score(1–7)

EnablingTradeIndex2014 89 3.7SubindexA:Marketaccess(25%) 112 3.2Pillar1:Domesticmarketaccess 107 4.0Pillar2:Foreignmarketaccess 79 2.3

SubindexB:Borderadministration(25%) 78 4.2Pillar3:Efficiency&transparencyofborderadministration 78 4.2

SubindexC:Infrastructure(25%) 69 3.8Pillar4:Availability&qualityoftransportinfrastructure 103 2.6

Pillar5:Availability&qualityoftransportservices 55 4.3Pillar6:Availability&useofICTs 54 4.4

SubindexD:Operatingenvironment(25%) 104 3.7Pillar7:Operatingenvironment 104 3.7

Source:WorldEconomicForum,2014,p.276.

Thedata fromthe2014ReportshowthatSerbia,according to thesurvey, is ranked89,whileBosnia and Herzegovina is ranked 78, Albania 69, FYR Macedonia 63, Croatia 56, andMontenegro49.

3.TRANSPORTFLOWSANDBORDERCROSSINGSINEFFICIENCES

RoadtransportisthevehicleforthelargestpartofSerbianforeigntrade.In2013itaccountedfor 46.6%of all transport in the import of goods and 60.4% of all transport in the export of

goods.Thetotalvolumeofinternationaltransportofgoodswasslightlyhigherin2013thanin2012andamountedto19.5milliontonsofgoods:5.65milliontonsinimports,6.75milliontonsinexports(Figure3)and7.08milliontonsintransit.

Figure3:Volumeoftotaltradeingoodsandroadtransport,milliontons,2013

Source:StatisticalofficedatabasesandSORS(2014),p.2

The international haulage is facingmany challengeswith a direct impact on productivity andquality of the service. One of the most severe is considerable time losses during thetransportation process. Studies done in 2007, 2008 and 2011, presented in Medar andManojlović (2011) and Petrović Vujačić and Medar (2012), highlighted inner customs andborder crossings as amajor source of considerable time losses. The records of time spent ininner customs and at border crossings (hold‐up times) in 2008were: for border crossings –leavingSerbiawas1hourandenteringSerbiawas6.5hours.Eventhoughthiswasapilotstudy,theresultsclearlyindicatethathold‐uptimesinSerbiaarelongerthaninothercountries.Theresults from a similar sample in 2011 show a slight improvement, i.e. acceleration of theprocedure.

More improvement,specially inenteringSerbia,canbedetectedfromthereportsof thestudyConducting surveys at border crossing points (Ipsos Strategic Marketing, 2013) done for theBelgradeChamberofCommerceasapartnertotheACROSSEEproject inwhichtheauthorsofthis paper participated. The research included road and railway transport. The surveys doneaccordingtothedefinedmethodologywithintheACROSSEEproject(ACROSSEE,2013)includedstructured interviews (questionnaire) of truck drivers and traffic counting at border crossingpoints.TheresearchwasconductedinJune2013at7bordercrossingswithCroatia(Batrovci),Bosnia andHerzegovina (Mali Zvornik),Montenegro (Gostun),Macedonia (Presevo), Bulgaria(Gradina),Hungary(Horgos)andRomania(Vatin).

Outofmorethan7000countedtrucksenteringandexitingSerbiaduringthestudyperiod,1823truckdriverswereinterviewed.Mostofthem,25.5%wereatBatrovciborderstation(Table2),54%exitingSerbiaand76%drivingheavygoodsvehicles.Transitfrequencyofthesametripis:weekly39%(average1.8times),monthly46%(2.8times),andyearly8%(7.7times).

Table8.Truckdriverinterviews‐characteristicsofthesample[%]

BorderstationTotal

Batrovci Gostun Horgos GradinaMali

ZvornikPresevo Vatin

DirectionEntering 20 17 14 14 12 13 10 46Exiting 30 15 15 12 14 8 5 54

Vehicle'sclassification

LGV 14 14 22 11 13 20 6 4HGV 25 20 15 7 16 10 7 76TIR 31 12 38 1 9 8 20

Total 25.5 16.1 14.6 13.2 12.9 10.2 7.4

Source:IpsosStrategicMarketing(2013)

0 5 10 15

Export

ImportRoadTransport

Total

Yearlyaverageis54.2trips–almostoneeveryweek.Forthistripthetotalaveragetimefromoriginpointtodestinationpointis53hoursandtotalapproximatedistancefromoriginpointtodestinationpointis1048km(IpsosStrategicMarketing,2013).TheshortesttripsareassociatedwithcrossingtheborderwithBosniaandHerzegovina,andthelongestarecrossingborderwithHungary(IpsosStrategicMarketing,2013).Onaveragetheycarry16.3tonesofgoods.

Accordingtothemethodologydefinedwithintheproject(ACROSSEE,2013)waitingtimeswererecordedseparatelyfortheperiodbeforethebeginningoftheprocedureandfortheprocedureitself and entire hold‐up. The mean of the waiting time in queue before the start of theproceduresis59minutesandthemostperceivedisupto10minutes–33.1%(Figure4).Presevoborderstationhasthehighestvalue–55%ofrespondentsstatedmorethan40minutes.Vatinisthebestwith53%answersupto10minutes.Themeanforthetimeneededforallthecontrolsisbelowthequeuingtimeandamountsto47minutes.Theanswerupto10minuteswasgivenby37.8%respondents(Figure5).Theresultsfortotalwaitingtimeattheborderstationduringthistrip revealsproblemsatHorgoswhere70%statedmore then60minutes.Themean for totalwaiting time is 69 minutes. Results for hold‐ups at border crossing show a better situationcomparedwithpreviousstudiesbutsomeofthecrossingshaveresultsthatneedtheattentionofthegovernment.

Figure4:Waitingtimeinqueuebeforethestartoftheprocedures,%

Figure5:Timeneededforallthecontrols,%

Source:IpsosStrategicMarketing(2013)

Thetimelossescertainlyhaveanimpactontheperformanceofindividualtrips,haulersandtheinternationalhaulagesector,aswellason theirusersandsocietyasawhole.Thereasons formajor hold‐ups can be categorized in several groups: (i) flaws and deficiencies in the legalframeworkanditsimplementation;(ii)poororganizationofthecustomsofficesandassociatedinspection services; (iii) unprofessional and unmotivated staff; (iv) corruption; (v) lack oftechnical equipment and technology and (vi) the low capacity and poor condition ofinfrastructure.Theinterviewwithtruckdriversshowsthereasonsaslonglines/trafficdensity– 28%, procedures /paperwork/ long processing of documents –18%, slow work – 6%,inactivity/indifference/ unprofessional officers, change of officers/long change of shifts, notenough lanes, organization of work (breaks, number of officers...), inspections (procedures,slowness,workinghours...),systemerror,smallcrossingcapacity,congestionontheothersideoftheborder,corruption(thisquestionhasmultipleopenendedanswers).

All this timeandmoney losses inborder crossingcause supply chain fragmentation,weakneslinks and reduce the efficiency of supply chains. Improving connectivity and reducingfragmentationof supply chains impliesa renewedpush fornational improvementsandcross‐border integration in such areas as infrastructure standards, trade facilitation, and serviceregulation.

33.1%

30.8%

25.2%

10.8%

Upto10min

11‐40min

Morethan40min

Noanswer

37.8%

29.0%

29.2%

4.0%

Upto10min

11‐40min

Morethan40min

Noanswer

4.CONCLUSION

In Serbia the legislation and corresponding regulations that ease trade have been passed.Nevertheless,muchremainstobedoneindifferentareas.IntheperiodaheaduntiltheybecomeEU member states, the countries of the WB would benefit from some measures that wouldcontribute to easier and faster trade and transport flows. Thesemeasures should lead to thecreationofconditionsfortradeandtransportsimilartothoseintheEU.Substantialassistanceinthis regard is provided by the EU,World Bank, international donors and other internationalinstitutions. Although positive improvements have been made in facilitating cross‐borderpractices andbetter cross border regional cooperation, joint projectswith the goal of furtherimprovementof tradeand transport flows in theWesternBalkan region areneeded.There isstill a lot to do in Serbia and other WB countries in the area of economic and institutionalreformswhichshouldleadtomoreefficientsupplychains,abetterbusinessenvironment,tradeandeconomicgrowth.

ACKNOWLEDGMENT

ThisresearchwassupportedbytheMinistryofScienceandTechnologicalDevelopmentoftheRepublicofSerbiathroughtheprojectsno.36006,no.36010,andno.36022.

REFERENCES

[1] ACROSSEE, (2013). Accessibility improved at border CROSsings for the integration ofSouth East Europe, SEE/D/0093/3.3/X, project co‐financed by the EU TransnationalCooperationProgramme"South‐EastEurope".

[2] Drzeniek Hanouz, M., Geiger. T., Doherty. S. (eds.), (2014). The Global Enabling TradeReport2014,WorldEconomicForum,Geneva,Switzerland,www.weforum.org

[3] Handjiski, B., Lucas, R.,Martin, P., Sarisoy Guerin, S., (2010). Enhancing Regional TradeIntegration in Southeast Europe, World Bank Working Paper 185, Washington, D.C.,InternationalBankforReconstructionandDevelopment.

[4] IpsosStrategicMarketing, (2013).Conducting surveysatborder crossingpoints, partoftheprojectACROSSEE:WP.5.CrossBorderanalysis–Methodologyforcommonstandardsanalysisoncrossborderpoints,Belgrade,Serbia.

[5] Medar,O.,Manojlović,A., (2011). InternationalRoadHaulage Industry inSerbia:CriticalIssuesAnalysis.JournalofAppliedEngineeringScience,9(1),243‐252.

[6] Nikolić,G.,Jovanović,N.,TodorićV.,(2011).CEFTA2007‐2010,Experience,Potential,andPerspective,Belgrade,Serbia:NewPolicyCenter.(InSerbian)

[7] SORS, (2014). Entry, exit and transit of freight road vehicles, by countries of vehiclesregistration,2012–2013,Belgrade,Serbia:StatisticalOfficeoftheRepublicofSerbia.(inSerbian)

[8] Petrović Vujačić, J., Medar, O., (2012). Towards trade and transport facilitation in theWestern Balkans – the case of Serbia. Conference Proceedings: 15th InternationalConference on Transport Science, UL Fakulteta za pomorstvo in promet, Portorož,Slovenia.

[9] Vujačić,I.,PetrovićVujačić,J.,(2012).TheEUCrisisandInstitutionalreform.FromGlobalCrisistoEconomicGrowth–WhichWaytoTake?Vol.1,Economics,(eds.)B.Cerović,M.Jakšić,Z.Mladenović,A.Praščević,FacultyofEconomics,Belgrade.pp.75‐101.

MULTICRITERIAANALYSISAPPROACHFORHUMANITARIANPREPAREDNESSOPERATIONS

D.Aidonisa,Ch.Achillasa,b,D.Folinasa,D.Triantafylloua

aDepartmentofLogistics,TechnologicalEducationalInstituteofCentralMacedonia,BranchofKaterini,60100Katerini,GreecebSchoolofEconomics&BusinessAdministration,InternationalHellenicUniversity,14thkmThessaloniki‐Moudania,57001Thermi,Greece

Abstract: Strategic, tactical and operational decisions in humanitarian logistics require thesimultaneousconsiderationofanumberofusuallymutuallyconflictingcriteriainordertocomeupwithoptimaldecisions.Ourworkaims topresenta conceptualapproachbasedonmulticriteriaanalysis fortheselectionoftheoptimal location forthedevelopmentofahumanitarian logisticsdepot.Inthislight,decision‐makersareprovidedwithadecisionsupportsystemtowardsoptimalpreparednessforfacilitatingsecurerelieftothevictimsofnaturaldisastersinatimelyandreliableway.Theproposed framework is implemented for the caseofa specific typeofnaturaldisaster(earthquake).However, theprocedurecouldbeeasilyadopted inorder tocomeupwithoptimaldecisions in respect to types of natural disasters other than the one examined in the presentanalysis

Keywords: Location of logistics depot; relief operationsmanagement;multicriteria analysis;decisionsupportsystem

1.INTRODUCTION

Themanagementofhumanitarianlogisticsisacriticalfactorforfacilitatingsecurerelieftothevictims of natural or man‐made (e.g. terrorist attack) disasters in a timely and reliable way.Humanitarian logistics management is a complicated issue dealing with different situations,times and responses (Kovacs and Spens, 2007). In contrast to commercial logistics wherenumerous analytical optimization models have been constructed, humanitarian logisticsconstitute a rather unexplored field. Depending on the type of disaster, different ways oftreatment need to be planned and implemented, making disaster relief management a case‐specificapproachwithhigh levelsofuncertaintyandrisk.Thispaperpresentsamulticriteria‐based strategic approach towards optimal location for the development of a humanitarianlogistics depot. The aim is to provide an easy‐to‐use decision support system thatsimultaneously accounts social, technical and economic concerns. The rest of the paper isorganizedasfollows.Section2accommodatestheproposedmethodology,whileinSection3,acasestudyisprovided.Finally,inSection4weprovidethemainconclusionsofthepaper.

2.METHODOLOGICALFRAMEWORK

Strategic,tacticalandoperationaldecisionsinhumanitarianlogisticsoften,ifnotalways,requiresimultaneous consideration of a number of usually mutually conflicting criteria in order todecide theoptimalalternative. Inpractice,decision‐makersaim towards thebest compromisebetweenavailablealternatives.Towardsthisdirection,anumberofMulti‐CriteriaDecisionAid(MCDA) techniques have been proposed to assist decision‐makers not taking decisions basedonly on personal thoughts, views or experiences, but on a robust and rational approach. Theproposedmethodologicalframeworkisdividedintotwophases;before(PhaseI)andafterthedisasterstrikes(PhaseII).AsaprimarystepofPhaseI,potentiallocationsforthedevelopmentof a humanitarian logistics depot need to be identified. Suchdepots can alternatively be builtwithinarmycampsorotherpublicowned infrastructure.Themethodologycontinueswith itsPhaseII,whichcommencesexactlyafterthedisasteroccurs.Operationalcriteriaaredeterminedaccordingthenatureandthespecialcharacteristicsoftheincident.Itshouldbeunderlinedthattheexactnumberofbothstrategic(PhaseI)andoperationalcriteria(PhaseII)dependsonthedecision‐maker.Performancesofallpotentialsitesonthosecriterianeedtobequalitativelyorquantitativelyquantified.Quantifiedperformancesofalternativesitesovertheselectedcriteriaareenteredintoaperformancematrix.Thenthequantifiedparametersarepreferablyscaledinordertofacilitatemonitoringanddirectcomparisonbetweenindividualcriteria.Asa laststepbefore the formationof themathematicalmodel,weighting factorsneed tobedetermined fortheselectedcriteria.Thisstepallows the incorporationofspecificstrategicgoalsaccording tothe decision‐makers philosophy to the decision‐making process. Weighting factors of theselectedcriteriaarecasespecific,basedonthetypeofincidentandtheseverityofitsimpacts.However, due to the humanitarian nature of the problem under study, weightings of criteriausually“favor”socialparametersfollowedbytechnicaloreconomicconcerns.Themethodologyconcludes with the model runs and the determination of the humanitarian depot’s optimallocation.Thisisfurtherreinforcedbyasensitivityanalysiswhichenablesthedecision‐makertoidentify the robustness of the calculated ranking, as well as which are the parameters thatmostlyinfluencetheoptimalselection.Acriticalturningpointinthedecision‐makingprocessisthe choice of themulticriteria analysis technique to be employed. The literature lists a largenumber ofmulticriteria analysis techniques. There are no universally preferablemulticriteriaanalysistechniques,butsometechniquesaremoreorlessappropriateaccordingtotheproblemunder study and its specific characteristics. In the proposed method, theEliminationEtChoixTraduisant la REalité III technique (Roy, 1978), commonly referred to asELECTRE III, is used. ELECTRE III requires the determination of three thresholds; preferencethreshold(p),indifferencethreshold(q)andvetothreshold(v).Withtheinclusionofthethreeaforementionedthresholds,ELECTREIIIisconsideredtobewell‐adaptedforuncertainties.Thetechnique uses three pseudo‐criteria to represent all aspects of a problem, and it begins bycomparingthealternatives'criteriascores.Theresultsareaggregated,andamodelofthefuzzyoutranking relation, according to the notion of concordance and discordance, is built. Themethod, in the second phase of the fuzzy relation exploitation, constructs two classifications(complete pre‐orders) through descending and ascending distillation procedures. A finalclassificationconsistsof the intersectionof thetwocompletepre‐orders.Asensitivityanalysistests theresultbyvaryingthevaluesof themainparametersandobservingtheeffectsonthefinaloutcome.

3.CASESTUDY

3.1Theincident

The applicability of the proposedmethodology is demonstratedwith its implementation in avirtualcasestudyforthedevelopmentofahumanitarianlogisticsdepotafteran8.4magnitudeearthquake in Pella, Greece. The epicenter of the earthquake is located at Kromni in Mount

Paiko, 3.2 kilometers below ground, 60kilometers north east from the city of Thessaloniki.ΑlternativesitesthatareavailableforthedevelopmentofthehumanitarianlogisticsdepotareoperationalarmycampsinGoumenisa,Giannitsa,EdessaandSkidra,andschoolyardsinAridea,ExaplatanosandAravisos,respectively.

3.2Determinationofcriteria

After identifyingallalternativepotential sites, thecriteria tobe taken intoaccountneed tobedecidedupon.Inanystudyconcerningsitelocations,themostimportantquestionisdefiningthecriteriatobeconsidered.Inthislight,acriticalnumberofrelevantstakeholders,allexpertsonthethematicareaunderstudy,werepersonallyinterviewedinordertodecidewhichcriteriatouseinthevirtualcasestudy.Asalreadystatedinthemethodologicalframework,thosecriteriacanbeeitherstrategicoroperational.Onastrategicbasis,theselectedcriteriaincluded;(CS1)local population, (CS2) distance from centralized warehouse, (CS3) quality of infrastructure,(CS4)distancefromclosestport,(CS5)distancefromclosestairport,(CS6)distancefromclosestheliport and (CS7) local hospital capacity. As already mentioned in the previous section, allabove criteria can be quantified before the incident occurs in order to come up with timelydecisionsinthecaseaneventstrikesinthearea.Performancesofthesevenalternativelogisticsdepots’ sitesaresummarized inTable1.References for thedata isalsodepicted. It shouldbementioned that quality of infrastructure is measured in a 1 – 10 scale, based on experts’opinions. Value “1” refers to totally inappropriate infrastructure, while “10” refers toinfrastructureequippedwithstate‐of‐the‐artfacilities.

Table1:Alternativesites’performanceonstrategiccriteria

Alternativesite

Population(CS1)

Distancefromcentralizedwarehouse

(CS2)

Qualityofinfrastructure

(CS3)

Distancefromclosest

port(CS4)

Distancefromclosestairport(CS5)

Distancefromclosestheliport

(CS6)

Hospitalcapacity(CS7)

(#) (km) (1‐10) (km) (km) (km) (#ofbeds)Reference HSA,2012 Google,2012 HGSCP,2012 Google,2012 Google,2012 Google,2012 TCG,2010

Edessa 25,729 40 2 91 56 7 150Aridea 19,970 60 4 98 67 13 18Exaplatanos 8,852 63 4 91 60 24 ‐Aravisos 7,509 47 8 64 33 6 ‐Skidra 15,633 32 6 75 38 12 32Goumenisa 6,677 82 9 70 42 7 ‐Giannitsa 31,782 39 5 52 21 9 248

Similarly,morecriterianeed tobedefinedonanoperationalbasis.To thatend, the followingoperational criteria were selected; (CO1) distance from the earthquake’s epicenter, (CO2)condition of local infrastructure after the earthquake,(CO3) number ofmissing people, (CO4)numberofhomelesspeopleand(CO5)numberofwoundedpeople(Table2).Thosecriterianeedto be quantified after the event, thus those cannot be quantified in advance similarly to thestrategiccriteria. It shouldbementioned that theperformancesof theoperationalcriteriaarebasedonvirtualcasestudyandarenotrepresentingreal lifedata.Asregardstheconditionoftheinfrastructureaftertheearthquake(CO2),inthepresentworkthisischaracterizedona1–4scale with “1” referring to totally destroyed local infrastructure and “4” referring to localinfrastructureingoodconditionand/orfullyoperational.Thevalues“2”and“3”refertomajorand minor destructions in the local infrastructure, respectively. Moreover, as regards thecriterionreferringtothenumberofwoundedpeople(CO5),thisisfurtherdividedintoseverityof wounds with Cat I to represent the number of slightly wounded citizens, while Cat III torepresent thenumberofheavilywoundedones.Lastbutnot least,quantificationofvalues forcriterionCS7isbasedonhospitalbedscapacitiesofGeneralHospitalofEdessa,GeneralHospitalofGiannitsa,MedicalCentreofArideaandMedicalCentreofSkidra,respectively(TCG,2010).

Table2:Alternativesites’performanceonoperationalcriteria

Alternativesite

Distancefromepicenter(C01)

Conditionofinfrastructure

(C02)

Missingpeople(C03)

Homeless(C04a)(C04b)

Wounded

CatI(C05a)

CatII(C05b)

CatIII(C05c)

(km) (1‐4) (#) (#) (%) (#)Edessa 46 3 87 16,724 65% 5,000 1,000 50Aridea 47 3 367 15,976 80% 6,000 1,000 40

Exaplatanos 39 1 458 8,409 95% 1,500 200 15Aravisos 16 2 309 7,134 95% 1,000 90 10Skidra 34 3 ‐ 13,288 85% 1,000 100 15

Goumenisa 30 2 207 5,675 85% 1,200 90 10Giannitsa 22 4 4 6,356 20% 2,000 800 30

Towardsoptimalsitelocation,thevaluesofalternativesoncriteriaCS1,CS3,CS6,CO2,CO3,CO4andCO5needtobemaximized,whereasthevaluesofalternativesoncriteriaCS2,CS4,CS5,CO1needtobeminimized.Asdiscussedinthemethodologicalsection,thevaluesofTables1and2are scaled in a range from 1 to 10 using the aforementioned Nk(A) index (eq. 1). ScaledperformancesaredepictedinTable3.

Table3:Alternativesites’scaledperformances

Alternativesite

CS1 CS2 CS3 CS4 CS5 CS6 CS7 CO1 CO2 CO3 CO4a CO4b CO5a CO5b CO5c

Edessa 7,8 2,4 1,0 8,6 7,8 1,5 6,4 9,7 7,0 2,7 10 6,4 8,2 10 10Aridea 5,8 6,0 3,6 10 10 4,5 1,7 10 7,0 8,2 9,4 8,2 10 10 7,8

Exaplatanos 1,8 6,6 3,6 8,6 8,6 10 1,0 7,7 1,0 10 3,2 10 1,9 2,1 2,1Aravisos 1,3 3,7 8,7 3,3 3,3 1,0 1,0 1,0 4,0 7,1 2,2 10 1,0 1,0 1,0Skidra 4,2 1,0 6,1 5,5 4,3 4,0 2,2 6,2 7,0 1,0 7,2 8,8 1,0 1,1 2,1

Goumenisa 1,0 10 10 4,5 5,1 1,5 1,0 5,1 4,0 5,1 1,0 8,8 1,4 1,0 1,0Giannitsa 10 2,3 4,9 1,0 1,0 2,5 10 2,7 10 1,1 1,6 1,0 2,8 8,0 5,5

Forthecaseunderconsideration,weightingfactors,indifference,preferenceandvetothresholdsarepresentedinTable4.Weightingfactorsandvetothresholdsarecalculatedasaveragesofthecorresponding views of the 13 experts that were interviewed. As regards preference andindifferencethresholds,thosewerecalculatedasfollows:

(1)

(HaralambopoulosandPolatidis,2003)and

(2)

(Kourmpanisetal.,2008),where:

: MaximumperformanceofalternativeAforcriterionk,

: Minimumperformanceofalternativewforcriterionkand

n=numberofavailablealternatives.

Inthatsense,indifferenceandpreferencethresholdsaredirectlylinkedtothequantifiedvaluesofthealternatives’performancesintheselectedcriteria.Sincevaluesoftheperformancetable(Table3)arescaled(inarange1‐10),indifferenceandpreferencethresholdsaresimilarforallcriteria.However,thosescaledvaluesrefertototallydifferentactualvalues.

),...,,(),,...,,(),(1

521721 cOSSAkAkk CCCkAAAAVVn

pMINMAX

),...,,(,3.0 521 cOSSkk CCCkpq

MAXAkV

MINAkV

Table4:WeightingfactorsandthresholdsforthedevelopmentofahumanitarianlogisticsdepotinPella,Greece

CS1 CS2 CS3 CS4 CS5 CS6 CS7 CO1 CO2 CO3 CO4a CO4b CO5a CO5b CO5c

Weightingfactor5.3% 4.1% 6.0% 1.6% 1.8% 4.9% 9.8% 7.7% 7.2% 14.9% 9.7% 1.4% 5.1% 7.8% 12.7%Thresholdofindifference

1.43

Thresholdofpreference 0.43

Vetothreshold ‐ ‐ 6.1 ‐ ‐ ‐ ‐ ‐ 5.7 2.8 4.8 ‐ ‐ ‐ 5.4

3.3Resultsanddiscussion

As a next step, following the determination of alternative sites and criteria, themathematicalThe optimal location (Aravisos) can be interpreted as a result of the specific site’s excellentperformanceinanumberofdifferentcriteria.Morespecifically,Aravisosisthesiteclosesttotheearthquake’sepicenter(CO1)amongalternatives,whilealsothelocationisoneoftheclosesttoport(CS4),airport (CS5)andheliport (CS6)so that resourcesandhumancapitalcanbequickerandmoreefficientlytransported.Thisisfurtherreinforcedbythespecificsite’sexcellentqualityofinfrastructure(CS3),sincethespecificareacorrespondstoanewlyrenovatedarmycamp.Inordertoovercomesubjectivityissues,thesensitivityanalysisthatfollows,aswellastheeasetore‐calculate optimal solutionwithmodified parameters, provides the decisionmakerwith aneasy‐to‐use tool. Sensitivity analysis is considered as a significant merit of the presentedmethodological approach on the grounds that in real‐world applications, input data originatefrom estimations which, although assumed constant, are sometimes more or sometimes lessreliable.Especiallyintimesofsevereincidentsastheonehereindescribed,thequalityofinputdata is furtherquestionable.Generalsourcesof individualuncertaintiescouldcomefromdataseries uncertainties, synergies and idiosyncrasies in the interpretation of ambiguous orincompleteinformation.Inanycaseitshouldbeunderlinedthatthesimultaneousconsequencesofpotentialvariationsofparametervalues,decisionvariablesandconstraintscouldbestudiedbynewmodel runs, since the lowcomputational timegives theopportunity for fast reformedoptimalsolutions.Onthisbasis,ELECTREIIIispreferable,sinceitisconsideredtobetteradaptto uncertainties (Roy and Bouyssou, 1993).In this light, for sensitivity analysis purposes, theproblem under study is resettled with modified parameters (weighting factors andthresholds).Together with the “basic” scenario, twelve more parameter‐based scenarios withdifferentiated preference and indifference thresholds were examined. Those scenarioscorrespondtoincreaseofpreferenceandindifferencethresholdsby5%,10%,20%,30%,40%and50%, respectively.The rest six investigated scenarios similarly correspond todecreaseofpreference and indifference thresholds by the same percentages. For all above scenarios,Aravisos appears as the optimal location for the development of the logistics depot, whichprovides thedecision‐makerwithadditional confidenceon the robustnessof the result.Apartfromtheaforementionedtwelvethreshold‐basedscenariosexamined,sensitivityanalysisisalsoconducted with the modification of criteria weighting factors also by 50% (increasing anddecreasing).Forallweightingfactors‐basedscenarios,Aravisosappearstheoptimallocation.

4.CONCLUSIONS

Incasesofearthquakes,floods,volcanoes’explosions,tsunamis,aswellasothernaturalorman‐madedisasters,timelyandefficientorganization,planningandexecutionofallnecessaryactionstowards humanitarian relief is most critical. Notwithstanding the fact that all such differentincidentsheavilydifferentiateamongeachotherintermsoftherequiredreliefoperations,inall

cases those operations need to be carefully organizedwell ahead the incident occurs. In thislight,responsiblebodiesneedtobeadequatelypreparedinordertoefficientlycopewithsuchdisastroussituation.Towards thisdirection,oneof themost strategicdecisions is theoptimallocationof ahumanitarian logisticsdepot tobedeveloped.Suchadecisionmayrepresent thecornerstonetowardshumanitarianreliefandmayfurtherinfluenceallotherimportantstrategic,tacticalandoperationalactions thatare required.Optimal locationof ahumanitarian logisticsdepot involves simultaneous consideration of usually mutually conflicting parameters whichmaybeeitherrelatedornottotheepisode’sepicenter.Thelattercanbeassessedwellaheadtheincidentstrikesinordertheresponsiblebodytobebetterpreparedforrapidresponse.Ontheother hand, the decision‐maker needs to wait until the disaster actually happens in order tocome up with the assessment of the parameters that are directly influenced by the type ofdisastrousincidentanditsepicenter.Thispaperpresentsaneasy‐to‐usetoolthatcouldfacilitatesecure relief to the victims of disasters in a timely and reliable way. The proposedmethodological framework is implemented for the case of a virtual earthquake incident innorthernGreece.However,theprocedurecouldbeeasilyadopted,withslightmodificationsandadjustmentstothespecialrequirementsoftheproblemunderconsideration,inordertosurfaceoptimal decisions in respect to types of natural or man‐made disasters other than the oneexaminedinthepresentanalysis.

ACKNOWLEDGEMENT

The presented research was partially conducted in the context of the project entitled“InternationalHellenicUniversity(Operation–Development)”,whichispartoftheOperationalProgramme “Education andLifelongLearning”of theMinistryofEducation,LifelongLearningandReligiousaffairsandisfundedbytheEuropeanCommission(EuropeanSocialFund–ESF)andfromnationalresources.

REFERENCES

[1] Haralambopoulos,D.,Polatidis,H.(2003),“Renewableenergyprojects:structuringamulti‐criteriagroupdecision‐makingframework”,RenewableEnergy,Vol.28,No.6,pp961‐973.

[2] Kourmpanis,B.,Papadopoulos,A.,Moustakas,K.,Kourmoussis,F.,Stylianou,M.,Loizidou,M.(2008),“An integrated approach for the management of demolition waste in Cyprus”,WasteManagementandResearch,Vol.26,No.6,pp573‐581.

[3] Kovacs, G., Spens, K.M. (2007),“Humanitarian logistics in disaster reliefoperations”,InternationalJournalofPhysicalDistribution&LogisticsManagement,Vol.37,No.2,pp99‐114.

[4] Overstreet, R., Hall, D., Hanna, J., Rainer, R.K. Jr. (2011), “Research in humanitarianlogistics”, JournalofHumanitarianLogisticsandSupplyChainManagement,Vol.1,No.2,pp114‐131.

[5] Roy,B.(1978),“ElectreIII;Algorithmedeclassementbasesurunerepresentationflouedespreferencesenpresencedecriteresmultiples”,CahiersdeCERO,Vol.20,No.1,pp3‐24

NEWTRENDSANDSTRATEGIESINPOSTALLOGISTICS

ĐorđijeDupljanina*,DraganaŠaraca,MomčiloKujačića

aUniversityofNoviSad,FacultyofTechnicalSciences

Abstract:Thispaperwillpresentthe latesttrendsandstrategies inpostal logisticsandwillalsogive the experiences ofmost developed postal operators in the field of logistics.Postal businessnetwork of Serbia is the largest infrastructuraland logisticalnetwork. Serbianposthasawiderangeoflogisticsservicesandpostisaleaderintheprovisionoflogisticsservices.Wewillprovidean adequate prediction of the future development of postal logistics and assess the currentsituationinthefield.

Keywords:Logisticstrends,logisticsstrategies,experiences,Serbianpost,prediction

1.INTRODUCTION

Therapidtechnologicaldevelopmenthasgivenanumberoftrendsthathaveaffectedallareasoflogistics activities. Besides the technological trends there are and social‐business trends. Incombinationwiththetechnologicaltrendstheycreateuniquevalueforbusinessimprovement.Social‐business and technology trends are the result of megatrends such as digitization,globalizationandsoon.UniversalPostalUnionmonitorsallmodern trends, implements themandadjustsforpostallogistics.

The aimof this paper is to summarize themost important trends aswell as strategieswhichpassingthroughfromthesetrendsinthefieldofpostallogistics,givetheexperienceofleadingoperatorsinpostallogistics,andthenanalyzesthestateofpostallogisticsprovider"PostSerbia"to accept new trends as leading postal logistic provider. Through SWOT (StrengthWeaknessOpportunities Treats) analysis it will be determine the readiness to accept new trends andperformtheappropriateconclusions.

2.TRENDSINPOSTALLOGISTICS

According to researchconductedbyKückelhausandWegner (2013) from theDepartment forInnovation of DHL in cooperationwithDetecton consulting, itwas concluded that logistics isinfluenced by a numerous of trends. Themost important of them are labeled asmegatrends.Someofthesemegatrendsare:

Digitalization E‐substitution Sustainability Continuingglobalization Outsorcing

* ddjordji@uns.ac.rs

Demographicchangeandurbanization.

Inadditiontotheidentifiedmegatrends,itisimportanttonotetheimpactofnewtechnologiesand their development trends that appear in logistics. The greatest impact have these newtechnologies:

Mobildevicesofnextgeneration:smartdevicesandtablets Cloudbasedmarketplaces Internetofeverything Internetofthings Mobilapplication In‐memorycomputingetc.

Bycombiningthemegatrendsandnewtechnologies,DHLcompanywasabletoidentifytrendsthatarepresentinthelogisticsanddividedthemintotwocategories:

Socialandbusinesstrends Technologytrends

HeutgerandKückelhaus(2014)fromDHL'ssector for inovationhavealsoconducted identicalstudyandinthefollowingtablearegiventhemostimportanttrendsfromtheaforementionedtwocategories,ie.trendswhichhavethegreatestimpactonlogisticsproviders(Table1):

Table1.Logisticstrends

Logisticstrends

Socialandbusinesstrends

Technologytrends

SupergridLogistics BigData/OpenData

Real‐TimeServices CloudLogistics

AnticipatoryLogisticsAutonomousLogistics

UrbanLogistics 3Dprinting

Omni‐ChannelLogistics InternetofThings

2.1Socialandbusinesstrends

Supergridlogisticsisaveryimportantlogistictrend.Thereisatendencyinmajorurbanareastoconnectwitheachotherandcreatesupergridtransportationnetworks,andthatnetworkswillbe oneday connected to large internationalmega‐cities,whichwill act as logistics hubs. Thistrendhasalreadybegun.UnitedStatesPostalService(2014)imposedoneproblem.Problemistheexistenceof"lastmilelinks,toconnectpeoplefromruralareastotherestoftheworld.Thatsegmentwillalsobeveryinterestingforpostallogisticproviders.

Heutger and Kückelhaus (2014) said that in the trend of real time services, there are twodominantinfluence:realtimetrackingservices(forexample,servicesthatprovidesinformationaboutpositionsofitems)andrealtimeordermanagement(orderingalongtheway).Thereareavarietyofmobileapplicationsforthislogistictrendinpostallogistic.

Anticipatory logistics is improved method for predicting request for items. It is especiallyapplicable formilitary operations. Predictive logistics planning is particularly important in e‐commerce. Companies analyze the user's search, shopping history, desires and prepare theorderbeforeordering.

By resarch of Federal Ministry of Education and Research (2014), 85 percent of Europeanpopulation will live in urban areas in 2050. The existing infrastructure does not represent aguaranteeforasustainablesupplychainofgoodsandservices.Postal,parcel,expressandotherlogisticscompanieswillconsidernewopportunitiesthatarespecialdesignedforurabanareas.Urbanlogisticswilloccupyaveryimportantplaceinthelogisticsindustry.

Omnichannellogisticsismadeofbetterlinkingshopande‐commerceopportunities.Thereareseveraloptions,oneofthemisthattheuser,forexample,cangotothestoreandseesanobjectthathelikesandtoorderforhomedeliveryortoorderonlineandpickupinstore.Thistrendisbeginningtoappearinthepostallogistics.

2.2Technologytrends

UniversalPostalUnion(UPU)over140yearsofcollectingdataonglobalpostalexchange.Inthepostalterminology,thistermhasbecomeknownasbigpostaldata.InReportofUniversalPostalUnionforDecember(2013)wasconcluded:“whetheryoumakeaphonecall,sendaletter,youleave some footprints digital character”. The private sector can use this data to improvebusiness.PostalbigdataisthekeytoglobaldevelopmentandUPUrealizedthatdata ispublicgood.

IBM researchers have demonstrated how big data can be used for optimization of publictransportinAfricancitiesmatchingofbusrouteswithtrafficfrommobilephones.Thatallowedthemtocreatethenewroutesthatpassengerswillsave10%ofthetimeintransit.

Postaland logisticscompaniescanbenefitbyshifting ITcomponents to thecloudacross theirentire value chain.Most postal and logistics companies already started their cloudmigrationjourney. TNT uses cloud to support both front‐ office and back office operations. Postal andlogisticcompaniescangeneraterevenuefromofferingtheirservicefromthecloud.Servicesareaccessibleanytimeandanywhere.(Deloitte2012)

Kückelhaus andWegner (2013) indicated that autonomous logistics is the implementation ofunmanned aerial vehicles (UAVs), cellular transport systems and other similar systems thatprovide a variety of opportunities in the sector of storage and transport. Such systems in thecoming years will become a reality and the postal and logistics sector will be impossible toimaginewithouttheiruse.Allcompaniesthatare involved inexpressdeliveryareconsideringautonomouslogisticsintheair.Thepotentialapplicationisinhardtoreachareasorformedicalpurposes.(figure1‐DHLtestedparcelcopter)

Figure1.DHLparcelcopter

3D Printing was presented as an automatic method that will be used for the production ofprototypes.Thereareseveraltechnologiesandmostofthemworkontheprincipleofbuilding

http://www.aircargonews.net

uplayersofmaterial.Thematerialsusedinthe3Dprintingaremostlyofplasticsandceramic.3Dprintingisatechnologythatwillchangethelivesofpeople.Bearingthisinmind,3Dprintingwillhaveabiginfluenceonthelogisticscompany.Manners‐BellandKenLyon(2012)indicatedthat this trendwill create new opportunities andwill develop new sector thatwill dealwithstorageandmovementofrawmaterialusedby3Dprinters.Withtheavailable3Dprinters,foroperations in remote locations is only required electronic library with items and computer.Combinationof3Dscanners,computers,3Dprintersallowyoutoprintthedesiredobject.

Universal Postal Union saw the significance of the trend which is known as the "internet ofthings". Term “internet of things” is adapted for postal vocabulary and become “internet ofpostal things”(Figure2) .When itsays"internetof things"usuallyrefers to theuseofsensorsthat enable physical objects, for example letters and packages that collects a variety ofinformation in real time through Internet. In thismannerareobtained"bigdata" foranalysis.Thiswillhelpanycompanytopromotetheirbusinessortodevelopanewsetofservices*.

Figure2‐Internetofpostalthings**

3.STRATEGIESINPOSTALLOGISTICS

Inthissection itwillbegivensomebasiccharacteristicsofstrategiesof leadingpostal logisticoperator:DHLandSwissPost.

ThestrategyDHLisbasedonthreepillars:Focus,Connect,Grow.Theobjectiveistoremaininpositionlogisticsproviderfortheworld.Eachofthethreepillarshasverycleargoals.Focusonthethingsthatmakethemsuccessful.Thecompanycontinuestofocusonlogisticsasaprimaryactivity,andplansto85%ofrevenuecomesfromlogisticsactivities.Theyaretryingtoexpandservices for e‐commerce market. Connect pillar is based on linking the organization. This isachieved by making the exchange of internal know‐how expertise. Connect pillar allows toincreaseservicequality.

Toensurethedevelopmentofthecompanytheycreatedagrowpillar.Withit,thecompanyisexpandingintonewmarketsegments,especiallyinthesphereofe‐commerce***.

SwissPost isamodern logisticscompany.Everydaydeliveredhalfamillionpackagesand15million letters. Itstrivestomeetuserneeds,andat thesametimetoremaineffectivenationalcompany.SwissPostisgoingtoincreasetheefficiencyfromcollectiontodelivery.SwissPost’slogisticstrategyprovidescommonbasisfordevelopmentforthecompany’sdifferentareas.Thisappliestoallfieldsoftheprocesschain.(Swisspostandpolitics2014).

* http://news.upu.int/no_cache/nd/internet‐of‐postal‐things‐offers‐food‐for‐thought ** http://nfpe.blogspot.com/2014/10/upu‐news‐forum‐on‐internet‐of‐postal.html *** http://www.dpdhl.com/en/about_us/strategy.html

4.POSTALLOGISTICSINSERBIA

PostSerbiaoffersnumerous logistical services. Services includesetofpostal, logistical and ITservices.Postusemoderntechnologytoimprovebusiness.InthispaperwedealwithabilityofPostSerbia toacceptnewtrends, technologicalandsocial‐business trends *. Byanalyzing theexisting capacities of post and history of accepting new trends in postal sector we havesummarizedthemostimportantfeaturesintheSWOT(Table3)matrix.

Table2.SWOTAnalysisfornewlogisticaltrendsinSerbianPost

SWOTanalysis

Strength

Goodinfrastructureandcapacitytoreceivetechnologicaltrendsandsocial‐businesstrends

Strategicapproachtobusiness

Educatedstaffwhocanquicklyimplementsocial‐businessandtechnologicaltrends

Weakness

Timeofimplementationofnewtrendsduetoregulationsandlaws

Thehighcostofnewservices

Culturaldifferencesandtimetocustomeradaptationonnewtrends

Opportunities

Theappearanceandavailabilityofnewusers(byusingautonomouslogistic‐reachhardlyaccessibleareas,deliveryofdrugsetc)

Implementationofnewtechnicalandtechnologicalsystems(increasecompanybrandineyesofcustomer)

Newserviceswithadditionalvalues(realtimeservices)

Treats•Theappearanceofcompetitorswithlowercosts•Theappearanceofcompetitorswithahigherlevelofquality•Theintroductionofnewlawsthatincreasethecostofbusiness•Changesdesiresandhabitsofusers

5.CONSLUSION

Attheendofthepapercanbesummarizedthatthenewtrendsinpostallogisticswillshapethedifferentwaysofdoingbusiness,whichcanleadtothecreationofnewservicesthathaveaddedvaluetotheuser,andthusincreaseprofitsforthelogisticcompany.InSerbia,inthenextperiodisexpecteddevelopmentofe‐commerce.Userswanttohavetheirorderedproductsdeliveryin * http://www.posta.rs/default‐eng.asp

the shortest possible time. This can be achieved through a combination of new trends,particularly the application of cloud logistics in the postal sector and providing variouselectronicservices,whichareattractive forusers.PostSerbiahasall thecapacitytobecomearegional leader intheprovisionof logisticsservices,aswellasveryeasyto implementcertaintrends such as cloud logistics, urban logistics, autonomous logistic especially in the sphere ofprocessing shipments. Future research in this sector we will focus on the analysis of otherprovidersofpostallogisticsandthedevelopmentofnewservicesinthisdomain.

ACKNOWLEDGMENT

This paper was supported by the Ministry of Science and Technological development of theRepublicofSerbia,throughtheprojectTR36040.

REFERENCES

[1] Cristiane,B.(2014).UrbanSupply,FederalMinistryofEducationandResearch.

[2] Deloitte,(2012).Cloudcomputinginpostalandlogisticsector,PostExpo.

[3] DHL,(2014).Strategy2020,Focus.Connect.Grow,DHL.

[4] Heutger,M., Dr. Kückelhaus,M., Detecon consulting, (2013). Logistic Trend Radar, DHLCustomerSolutions&Innovation.

[5] http://www.aircargonews.net

[6] http://news.upu.int/no_cache/nd/internet‐of‐postal‐things‐offers‐food‐for‐thought/)

[7] http://www.posta.rs/default‐eng.asp

[8] Manners,J.,Lyon,K.(2012).Theimplicationof3Dprintingforthegloballogisticindustry,TransportIntelligence.

[9] Swiss Post, (2008). Post and Politics, Logistic Strategy, Efficiency from collection todelivery,SwissPostCorporateCommunication.

[10] UPU, (2013). EDI messages help human development new stats bring good news,UniversalPostalUnion.

[11] U.S. Postal ServiceOffice of InspectorGeneral, TheGlobal LogisticRevolution:APivotalMomentForLogisticSector,(2013).USPostalService.

[12] Wegner, M., Dr. Kückelhaus, M., Detecon consulting, (2013). Logistic Trend Radar, DHLCustomerSolutions&Innovation.

SOCIALNETWORKSINLOGISTICSSYSTEMDECISION‐MAKING

BrankoArsića,PetarSpalevićb,LjubišaBojićc,AdelaCrnišanind

aUniversityofKragujevac, FacultyofScience,SerbiabUniversityofPrištinainKosovskaMitrovica,FacultyofTechnicalSciences,SerbiacUniversityofLyon,InstituteforPoliticalStudiesLyon,FrancedUniversityofNoviPazar,DepartmentofTechnicalSciences,Serbia

Abstract: Social networks, such as Facebook, Twitter and LinkedIn have been becoming verypopularduringthelastfewyears.Facebookiscurrentlytheworld’smostpopulous“country”withmore than 1.3 billion “inhabitants”. According to the statistical data, the users share theirimpressionsdailyintheformofstatusesaboutupcomingeventsandpresentstateofaffairs,theirproblems, plans, novel experiences about the products, political stances, and alike. Having thepossibilitytoextractthe informationof interest fromahugeamountofhand‐createddataabouttheusers’personalaffinitiesand theirusagewithin logistics system, it is facilitated tomeet thecustomers’ needs. In this paper we present a procedure for finding and analyzing valuableinformation related to the specific products, and its effect on logistics system decision‐making.Filteringisbeingdonebyalreadydevelopedsoftwareforneurolinguisticssocialnetworkanalysis‐“Symbols”. This software offers graphical representation of statistical data for selected brandsbasedon the socialnetwork statuses, its implications,aswellas targetgroupdemographicandterritorial structure. The results obtained point out possible increasing/decreasing demandsamongseparateusergroups,thereforegivingafactualbasisforlogisticschanges.

Keywords:socialnetworks,informationretrieval,logisticssystemdecisions.

1.INTRODUCTION

Enterprises createanddeliverproductsandservices through increasinglyglobalandcomplexsupply chains (Binder and Clegg, 2007). Nowadays, business requests are demanding andenterpriseshavetocontinuouslyseekwaysforbusinessimprovement.Improvementindicatorsare numerous and vary from decreasing operational costs, providing satisfactory customerservice, tominimizing existing disruption risks. These pointers can be achieved bymeans ofefficientsupplychaindesignandmanagement.

Inthetraditionalanalysesperformedinlogisticsandsupplychainmanagement(SCM)researchrelationships among buyers and suppliers are observed as linear (Cox et al., 2006). Thetraditional approachputsanactor (individualororganization)participating ina supply chaininto the focus, thus isolating a unit of the analysis. Contrarily, social network analysis (SNA)focusesonactor‐actorrelationshippatterning.Relationshipsamongindividualswithinasocialnetworkcanrevealhighly indicativeresults thataconventionalsurvey in the fieldof logisticsandsupply chainmanagement couldneveryield.These relationsmaybeestablished through,for instance, friendship, liking, ‘‘talked to over the last month’’, ‘‘sent e‐mail to’’, workflow,

moneyflowortheexchangeofgoodsamongactors(Scott,2000).Basically,flowsandtransfersarethemostimportantkindsofrelations.Yet,theyarerarelysurveyedandarepresumedfrominteractionsorsocialrelations.Incaseofrealdatawant,sincetheyarenotcollected,scrutinizedoravailableduetocompanysecurityandprivacypolicy,wehavefocusedoncompanyandfanprofilesonInternetsocialnetworks,suchasFacebook.

Unlikeclassicalofflinetechniques, Internetoffersuniqueandimmensepossibilitiesformarketresearch, giving real‐timedata access and insight into thepeople’s changingpreferences thusproviding room for innovation in the field. In this paper, we demonstrate how to use socialnetwork analysis and online communication for logistic problem investigation, advisingamendmentforcertainsupplynetworks,theiractivitiesandplans.Filteringofcollectedstatusesthroughout news feed was done by software for neurolinguistics social network analysis“Symbols”which is presented inmoredetails in the fourth section. Finally,we comparedourresultswiththeresultsofexistingsurveysforbeveragesbrandsavailableonInternet.

The remainder of the paper is structured as follows. Section 2 gives an overview of theliterature. Section 3 presents the details of our software “Symbols”. Section 5 describes ourresearch methodology. In Section 6, we provide interdisciplinary research opportunities forusing our software in supply chain system (SCS) and enterprise system research. Section 7concludesthestudy.

2.LITERATUREREVIEW

Inter‐connected companies which are integrated into supply networks participate inprocurement,use,andtransformationofrawmaterialstoprovidegoodsandservices(Harlandet al., 2001).Traditional approaches in analyzing supply chains are characterizedbydifferentissueswhichcanbeovercomewithviewingsupplychainsasanetworkforallactors.Ifwetakealookatonecompany, its “supplynetwork”will consistsof relations to itsdirectsuppliersandcustomers,andrelationsbetweenthemandtheirdirectsuppliersandcustomers,andsoon,thusforminganevermorecomplexnetwork.

Relativepositionofindividualfirmswithinnetworkinfluencesstrategyandbehavior(BorgattiandLi,2009),thereforeitisincreasinglyimportanttoanalyzethenetworkstructureofsupplyrelationships.A relationbetween two firmscanbeestablishedbasedon their collaboration, aproduct development, sharing the trade organization or money exchange for services etc.Different metrics, node‐level and network‐level, provide researchers with a descriptive andstatistical method for positioning and connecting the SCS actors (Wasserman, 1994). Degreecentrality, closeness centrality, and betweenness centrality are different types of centralitymetricsandtheyidentifynodesthatareimportant.

MostoftheexistinginterorganizationalSNAresearchhasoccurredinthestrategicmanagementarena(Pettigrew,1992).SNArepresentsavaluabletoolforanalyzingstructuresandrelationsindifferent areaswhich could be transformed in network, includingbusiness studies, sociology,computerscience,physics,andpsychology,knowledgetransfer,andinnovation.Kimetal.usedsocialnetworkanalysisto investigatethestructuralcharacteristicsofthreeautomotivesupplynetworks reported in (Choi and Hong, 2002). SNA use could be seen in the analysis ofcommunication patterns in organizations involved in humanitarian and tourist logisticsoperations(Holguín‐Verasetal.,2012;Pesonen,2011).Generally,SNAhasnotbeenappliedinan empirical study of real supply networks. A general paucity of SNA applications in supplymanagementarisefromthelackofconceptualclarificationastohowthekeySNAmetricscanbetheoreticallyinterpretedinthecontextofsupplynetworks(Carteretal.,2007).

Here we tried to overcome mentioned issues by offering additional information about thecompany'sbusiness.UnlikeSNA,wheretheobligatoryunitofanalysisisanorganization,firm,orbusinessunit,ourapproach isbasedonFacebookprofilesnetwork.Manysocialnetworksare

extremelyrich incontent,andtheytypicallycontainatremendousamountofdatathatcanbeused for analysis (Aggarwal, 2011). There are no studies employing demographic data fromFacebook for product popularity improvement. Following different parameters enables us topredict demands at a location, for a target group ‐ the need for building newwarehouses orrearrangingtransportroutesandallaccordingtothemarketchangesdetectedonline.

3.“SYMBOLS”DATACOLLECTION

WhenanalyzingFacebookdata,datacollectionwasamajorchallenge.Inourapproach,weaskeda group of people for the permission to access their data. This method results in a strongsamplingbiasandmakesitdifficulttoacquirelargesamples.

For the purposes ofweb application named “Symbols”,we developed a Facebook applicationSSNA(SoftwareforSocialNetworkAnalyze).Appusershavetoexplicitlyagreethattheappispermitted to access thepart of theirdata classified into two groups, static anddynamicdata,explainedbelow.Theywere informed that theirdatawill beusedonly for scientificpurposesand that it will not be given away to any third party. The app has had 46 activated, basicFacebookprofilesduringthedatacollectionperiod,12.12.2009–tothepresenttime.Onepartoftheaskedpermissionsreferstothefriends’data;therefore,thedataof106,434userscouldberetrieved.Inthebackground,twotime‐basedjobschedulers(CRON)areprocessingeverycoreprofile.Onejobistodealwiththestaticdata(education,birthday,city,job,fanpage,etc.),andanother one with dynamic data (statuses, likes and comments, all friends’ data). The dataincludes the friendship network (friendship relations) and the communication network (Like,Comment,PostandTagrelations).

4.METHODOLOGY

SamplesizeformineralwaterbrandsinSerbiatakenintoconsiderationare422forKnjaz,249for Aqua Viva, 859 for Rosa and 127 for Jana. Profiles who like brands’ pages on Facebookconstitutesamples.Forinstance,pagestakenintoaccountforAquaVivamineralwaterareAquaViva,AquaViva‐vodasaukusombreskve,AQUAVIVABreskvaHydroactive,AquaVivaSport.

Market Network (2014) is a source of information aboutmineralwatermarket in Serbia for2014.Thiswebsiteprovides research findings related tomineralwater consumption.MarketNetworkenlistssalechannels.ThisisbasedonstandardizedresearchinSerbiathatlastssinceNovember2002anditencompasses1500families inthecountry.Amongothers,thiswebsitelists themostpopularwaterbrandsbased.Basedonthis, listofmineralwaterbrands for thisresearch has been created. Other reports about mineral waters include text written by Smit(2009).Asforsodas,someoftherelatedresearchfindingscouldbeseenonwebsite“Tvojstav”(Online istrazivanja, 2010). Although similar research inquiries to our social networkstratificationhavenotbeendoneinSerbia,goalofthisparagraphistolistsomeofthefindingsofclassicalmarketresearch.

5.RESULTS&DISCUSSION

Goal of this paper is to see how real time data from social networks can be used in logistics.Figure1depictsage,educationlevelandrelationshipstatusofthosethatlikedifferentmineralwaterbrandsonFacebook.Thesedataaredifferent foreachbrandso itwouldbepossible toconcludethatelderlikeKnjaz(30yearsoldinaverage)whiletheyounglikeAquaViva(Figure1‐a).Thisawarenesscanoptimizemarketingbybeingleveledtochosentargetgroupsinordertogetcustomersfromdifferentagegroups.

Figure1.Statisticsfora)Ageb)Educationallevelc)Relationshipstatus

Figure 1‐b on the other hand, depicts that themost educated consumers ofmineralwater inSerbiaare thosethat like JanamineralwaterbrandonFacebook,while the leasteducatedarethose that like Knjazmineralwater. Consequently, advertisements addressed to those peopleshouldbefittinglycreated.Knjazmineralwaterbrandmanagershouldwanttoattractyoungerconsumers by creating advertising campaigns that talk to youth more than to the aged, forexample. The similar conclusion could bemadewhen looking at Figure1‐cwhichdepicts thatthosewholikeAquaVivaarethemost"single"intermsofrelationshipstatuswhencomparedtoother mineral water consumers. On the other hand, most married people like Rosa mineralwater.Additionally,weobtainedthatwomenpreferJanamineralwaterwhilemenpreferKnjazmineralwater.Similarly like incasesof first threegraphs,dataobtained fromFigure4canbeusedinmarketing.Whetherthismineralwaterbrandwouldgoforoneortheotheradvertisingstrategyitdependsfromdecisionofthecompanyinquestionbutthemainpointwouldbethatsocialnetworkswithuseof adequate software canprovidevaluable logistics formarketing interms of advertisements. Obviously, it is important that products in questions arementionedonline,andthis isexpectedbecausetheyarewidelyconsumedandtheyarepartsofeverydaylife. In thisway,online identificationandquantificationof logisticsdemandscanbeuseful forbrands.

Mentions of product category indicate the present needs for it. For example, water may bementionedmoretimesatsomelocationbecauseofhotweatherconditions.Ontheotherhand,watermay bementionedmore in one place because of the problems inwater supply. Thesecircumstances can be registered on social networks by increasedmentions of water productcategory, but also there might not be any apparent reason for the increase in demands.Measuringthisonlineincreaseinneedswouldbevaluableforlogistics.Thisreal‐timestatisticsmay be an input for topmanagement in deciding for highermarketing efforts, increasing the

numberofproductsonthemarketandinimprovingpositionofbrandswithintheexistingones.These actions would result in better sales. This approach has its use in lower branches ofmanagement, as well. They would use the input for selecting where and how to implementdecisions of top management (install water machines near appropriate institutions, how toorganizedistributionroutes,designproductfortargetpopulation).

Figure4.StatisticsforlocationandschoolsinKragujevac

Nonetheless, this method of research may have its implication in delivery logistics. Wheninquiredintohowmanytimespeoplementioncertainmineralwaterbrandduringaweek,thefindings show thatwater gets themostmentions from Kragujevac, which is a city in centralSerbia(Figure4).Hence,productcategoryistalkedaboutmoreinKragujevacthaninthecapitalcity, Belgrade. This may imply that water brand should increase supply in this part of thecountry.

When looked more closely, Figure 4 shows which population from different schools inKragujevaclikemineralwaters.Basedonthisgraphitispossibletoconcludethatsupplyshouldbeincreasedinthevicinityoftheschoolswherepupilsmostlikemineralwaters,whileforthoseschoolsinwhichbrandinquestionisnotsopopular,advertisingeffortsshouldbeincreased.Bygainingthesepiecesofinformationcompanycan,forexample,decidewheretosendpromotersofmineralwateraswellaswheretoincreaseotherkindofmarketingefforts.

6.CONCLUSION

Internetsocialnetworksmaybehidinganabundantsourceofopportunitiesgivingspacetothe“parallelworld”whichcanandinmanywaysdoessurpasstherealty.UnlikeSNA,inthispaperwechanged the focus fromorganizationasaunitofanalysis tocompanyand fansprofilesonFacebooksocialnetwork.Yet, thisresearch isnot independent frompreviousSNAresearch; itonly provides additional information about company’s business. The result is a demographicrepresentation offering newmarketing advices for specific brands, in our particular case, forwaterbrands. In the same time,we comparedour resultswith someonline surveysmadebydifferentmarketing companies.The resultspresented in this studygive affirmative answer tothe research question whether it is possible to use Facebook demographic data for productpopularityimprovement,andpointoutthenoveldemandsappearinginlogistics.

Theadvantagesofthisapproacharereal‐timeinformationwithoutconductingasurvey.Peopleimpressionscanbedividedintoinformationsetsfordifferentpopulationstructuressuchasage,

location,gender,educationallevelandmanymore.Thesesetsinfluenceonproductdesignandmarketingtargeting.However,everynewdecisionmustbeconfirmedwith“field”data.

Our future goal is to expand core Facebook profiles and to adapt application to the newestFacebookAPI2.0.Finally,anapplicationupgradeforTwitternewsfeedisrealnecessity.

ACKNOWLEDGMENT

ThispaperwassupportedbytheMinistryofEducation,ScienceandTechnologicalDevelopmentoftheRepublicofSerbia(scientificprojectsTR35026,ON174033andIII44006).

REFERENCES

[1] Aggarwal, C. C. (2011). An introduction to social network data analytics (pp. 1‐15).Springer

[2] Binder, M., Clegg, B. (2007). Enterprise management: a new frontier for organisations.InternationalJournalofProductionEconomics,106(2),409‐430.

[3] Borgatti,S.P.,Li,X.(2009).Onsocialnetworkanalysisinasupplychaincontext.JournalofSupplyChainManagement,45(2),5‐22.

[4] Cox, A., Sanderson, J.,Watson, G. (2006). Supply chains and power regimes: toward ananalyticframeworkformanagingextendednetworksofbuyerandsupplierrelationships.JournalofSupplyChainManagement,37(2),28–35.

[5] Choi, T. Y., Hong, Y. (2002). Unveiling the structure of supply networks: case studies inHonda,Acura,andDaimlerChrysler.JournalofOperationsManagement,20(5),469‐493.

[6] Carter,C.R.,Ellram,L.M.,Tate,W.(2007).Theuseofsocialnetworkanalysisinlogisticsresearch.JournalofBusinessLogistics,28(1),137‐168.

[7] Holguín‐Veras, J., Jaller,M.,VanWassenhove,L.N.,Pérez,N.,Wachtendorf,T. (2012).Onthe unique features of post‐disaster humanitarian logistics. Journal of OperationsManagement,30(7),494‐506.

[8] Kim,Y.,Choi,T.Y.,Yan,T.,Dooley,K.(2011).Structuralinvestigationofsupplynetworks:Asocialnetworkanalysisapproach.JournalofOperationsManagement,29(3),194‐211.

[9] Market Network (2014). Vode. Marketnetwork.rs. Accessed on 15th March 2015. URL:http://marketnetwork.rs/roba‐siroke‐potrosnje/analiza/63‐rsp‐analiza‐presek‐trzista/3167‐voda

[10] Online istrazivanja (2010). Coca Cola ili Pepsi. Tvojstav.com. Accessed on 15th March2015. URL: http://www.tvojstav.com/results/frADOXqyG83Sdn85uZbM/coca‐cola‐ili‐pepsi

[11] Pettigrew, A. M. (1992). On studying managerial elites. Strategic management journal,13(S2),163‐182.

[12] Pesonen, J. (2011).Tourismmarketing in facebook:Comparing rural tourismSME’s andlargertourismcompaniesinFinland(pp.537‐546).SpringerVienna.

[13] Scott,J.(2012).Socialnetworkanalysis.Sage.London

[14] Smit,F.M.(2009).Negaziranevode.Progressivemagazin.rs.Accessedon15thMarch2015.URL:http://progressivemagazin.rs/online/negazirane‐vode‐2/

[15] Wasserman, S. (1994). Social network analysis: Methods and applications (Vol. 8).CambridgeUniversityPress,Cambridge.

TRENDSINAIRLINECARGOFLEET

SlavicaDožić*,DanicaBabić

UniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Althoughtheglobalaircargo industryhasrecordedsignificant increase, it isexpectedthat itwillbemore thandoubled in thenext twodecade.The fleet involved inworldair cargotransport consists of freighters and passenger’s aircraft carrying cargo in its belly.The fleet ofpassengeraircrafthasalargernumberofaircraftandlowerbellycapacitycomparedtofreighters.Thereforeitisnotunexpectedthatmorethanhalfofworld’scargoiscarriedbyfreighters.Inthispaper, the cargo fleet structure and changes related to fleet size and fleet types are analysed.Bearing inmind thatmore than half of freighters are converted passenger aircraft it isworthnotingaircraftconversion/productionandretirementinthecargofleet.Theauthorsalsogiveanoverview of air cargo fleet both worlds and European, pointing the factors that influence onfreighterfleetgrowth.

Keywords:AirlineCargoFleet,AirCargoTransport,AircraftCapacity.

1.INTRODUCTION

Thecargotransportplaysasignificantroleintheworldtransportsystemduetothefactthatitmakes more money for a transport company in comparison with transport of passengers(Wensween, 2007). The air cargo transport has not achieved this success yet, but still isunavoidablemeansoftransportforspecificgoods.Thespeedofairtransportandthelowerriskof losingordamagingof the freightmakeair transport favourable forcarryingtheperishable,time‐sensitiveandvaluablecargodespitethehighcharges.Thevalueoffreightcarriedbyairisoverathirdofthetotalvalue,whilethetonnagerepresentsonlyfewpercentofthetotalfreightcarriedintheworldbyallothermodesoftransport(Boeing,2014).

Aimingtoachievethegroundandwatermeansoftransport,aircargoisthefastest‐growthareain the world cargo transport. Its growth depends on international trade and it is easilyinfluencedbythechangesinworld’seconomy.Afterfewyearsofstagnation,duringtheworldeconomic crisis and economy recovering, the air cargo traffic started to grow. Although theglobalaircargoindustryhasrecordedsignificantincrease,itisexpectedthatitwillbemorethandoubledinthenexttwodecade(Airbus,2014).

The fleet involved inworld air cargo transport consists of freighters and passenger’s aircraftcarrying cargo in its belly. The fleet of passenger aircraft has a larger number of aircraft butlowerbellycapacitycomparedtofreighters.Thereforeitisnotunexpectedthatmorethanhalfof world’s cargo is carried by freighters. In this paper, the cargo fleet structure and changesrelated to fleet size and fleet types are analyzed. Bearing in mind that more than half offreightersareconvertedpassenger,aircraftconversion/productionandaircraftretirementalso * s.dozic@sf.bg.ac.rs

shouldbenoted.AnaircargofleetoverviewbothworldsandEuropeanisgiven,pointingoutthefactorsthatinfluenceonfreighterfleetgrowth.

2.THEHISTORYOFAIRCARGOFLEET

The aircraft is constructedmainly for carrying passengers (Wensween, 2007). Therefore, thefact thatpassengers transportbyair ismoreprofitable thanaircargotransport isreasonable.Dependingon aircraft used in the fleet for cargo transport, developmentof air cargo servicescouldbedivided into threeparts. Taking a lookback in thepast it canbe seen that air cargotransportstartedafterWorldWarIIinordertousesparecapacityoftransportmilitaryaircraftwiththepistonpowerplant.Theseaircraftwerecharacterizedby lowspeed,smallrange, lowpayloadandlowavailablevolume.Hence,theywereemployedtocarryvaluableshipmentsuchasgems,preciousmetals,smallpackagesetc.

The secondpartof air cargohistory is related toappearanceofnarrowbodyaircraftwith jetpowerplant.Theseaircrafthavehigherspeed, largerrangeand largerpayload.Theyareusedfortransportofpassengersinthecabinwhilecargoiscarriedintheirbellyassupplementtofullpayload,enablingtomakeadditionalearningsforanairline.Thenarrowbodyaircraftofferedapossibilitytotransportdifferenttypesofcargo,whilethetransportcostperunitofcargowerereduced.Thetypeofcargothatcouldbetransportedislimitedduetothecargodoordimensionsand belly volume, as well as due to the volume and mass of cargo consignment. Thedisadvantages of narrow bodies are related to the fact that there aremany different units ofcargowhichmeansthatthegreatdiversityofhandlingequipmentisneeded.

Thethirdpartofaircargohistoryhasstartedwiththeappearanceofwidebodyaircraft(inthesixties) that gaveopportunity to transport a varietyof goods.At thebeginning the cargowasshippedbypassengeraircraftinthebelliesofwidebodiesassupplementtofullpayload.Later,the aircraft that carried cargo in bothbelly andmaindeck and allowedeven the transport ofunusualshipmentswerecame.Theseaircraft(knownasfreighters)offerpossibilitytotransporta largeamountofgoods(up to250 tonneson thesame flight).Theyareequippedwithcargoloadingsystems(mechanicalandpowered),whichenablesbetterstackingandsecuringofcargoandreductionofcargo loadingandunloading timeaswell.With these improvements theunitcostsoftransportarefurtherreduced.Thewidebodyfreightersprovidetheabilitytotransportalargenumberofcontainersandpalletsofdifferentsizes,aswellasthetransportofintermodalcontainers.Theintermodalcontainersareverysuitablehavinginmindthataircargotransportincludes inter‐modality because the cargomust bemoved to and from the airport by surfacemodeoftransport.Also,highcapacityoffreightersisconvenientfortransportofoversizedandheavycargo(e.g.helicopters,aircraftfuselages,trainsetc.).

Thewidebodyaircraftareproduced incombiandquickchangeversion.Bothpassengersandcargoarecarriedonthemaindeckofcombiaircraft(passengersmayoccupythefrontsectionofthe aircraft, with cargo occupying the back section), while quick change aircraft are used aspassengersaircraftbydayandasfreighterbynight.Thequickchangeaircraftareequippedwithquickchangesystemsthatincludepassengerseatsonflooringpanelswhichareputonandofftheaircraftlikecargocontainers.

ThefreighteraircraftaredividedintothreecategoriesaccordingtoBoeing(2014)(largewide‐body,mediumwide‐bodyandstandard‐body)andAirbus(large freighters,mid‐size freightersand small freighters). Each freighter category corresponds to particularmarket. As stated byBoeing(2014),theshareofwide‐bodyfreightersinthecurrentairlinecargofleetisabout65%andtheyofferabout95%ofthetotalcargofleet’scapacity.Duetotherangelimitationofsmalland mid‐size freighter, large freighters are usually employed on long‐haul routes with highmarket demand. Mid‐size freighters operate the regional routes where the usage of largefreighterspresentsacapacityrisk,anddemand isnotsmall.Small freightersareversatileand

flexible, thus they are used when terrain is geographically difficult (mountains, forests orislands).

Customers that sent their shipments by air consider the speed, reliability and quality of airtransportation asmain advantages and crucial factorswhile choosing themode of transport.AccordingtoHsuetal.(2009)itisveryimportanttoselectnotonlyoptimaltransportmodebutalso the optimal carrier that will deliver the goods to customer. The authors noted that thelogisticscostoftheshipperisinfluencedbythecharge,flightfrequencyandtransittimeofthecarrieraswellasbyproductcharacteristics(highpricingand/orperishablecharacteristicsleadtoahighinventoryloss).Theyshowedthattheshippersforwardingtheproductofhighvalueonshortdistanceconsidertheshippingchargeasimportantfactorandprefertheaircargocarrierthatoffershigherfrequencies.Ontheotherhand,thefreighteroperatorsaresensitivetotonne‐kilometrecostsduetothecompetitionwithothermodesoftransport.Theyalsoshouldadjusttheirsupplyattributestotheproductcharacteristicsinordertoincreasethemarketshare(Hsuetal.,2009).Whilechoosingtheirfleet,airlinesmusttakeintoconsiderationboththeinterestofcustomersandtheirowninterests,whichareusuallyconflicting.Theartoffleetplanningliesintheabilitytoharmonizetheseinterestsinthemostconvenientway.

3.FACTORSINFLUENCINGTHEFREIGHTERFLEETGROWTH

According to Airbus (2014) the key drivers of air freight growth are: 1) economic and tradegrowth that strengthen or induce cargo flows, 2) globalisation of exchanges and free tradeagreementsthatenablessmoothandeasyflowsofgoods,3)expandingmiddleclassdemandinghigheraddedvalueproducts(itisforecastedthatthemiddleclasswillgrowfrom33%in2013to over 60% in 2032), 4) increasing urbanisation and industrialisation that require specificgoods to be delivered in right place and right time from central warehouse, without localwarehousing(thereisnolargeinvestmentsinwarehousingandinventories).Also,theincreaseofpassengerdemandanditsimpactonaircargodemandshouldnotbeomitted,too.

Themarketconditionshavestronginfluenceonworldfreighterfleetgrowth.Theconsequenceoftheglobaldownturnreflectsindeclineofcargoflowsaswellasdeclineoffreighterdemand.Freighteroperatorssolvetheproblemrelatedtosurplusofcapacitybyparkingcertainnumberof theiraircraft.AccordingtoBoeing(2014)different factorsoftencausecontraryeffects.Thehighfuelpricesincreaseaircargotransportcosts,whichfurtherdecreasedemandforservices.Thecontraryeffectisthathighfuelpricesandfuelcostsarereasontoreplaceoldlessefficientfreighters (usually converted passenger aircraft that constitute more than half of freighters)with the newer ones, which further induce the demand for new freighters. This demand forfreighters decreaseswhen yields, load factor andutilization are reduced. The newwide‐bodypassenger aircraft with large cargo compartment also affects the demand for freighters, byreducingitslightly.However,thefreightersofferspecific,uniqueadvantageonthemarketthatcould not be offered by passenger aircraft. Competitivemode of transport and belly capacitycould influence on demand for mid‐size freighter by reducing demand for air cargo serviceofferinglowerprices.Theintercontinentalflightsareoperatedbywide‐bodyfreighters.

4.WORLD’SAIRCARGOFLEET

Airbus(2014)predictsincreaseinaircargofleetfrom2014to2033(Figure6).Theyexpectthat328 of 341 small freighters will be replaced, 13 will stay in service, while the fleet will beenlargedwith284aircraft(in2033,625smallfreighterswillbeinservice).Thefleetofmid‐sizefreighterswillbeincreasedfrom750to1242.Only97aircraftwillstayinservice,653willbereplaced,while growthwill be492 aircraft. Finally, 337 large freighterswill be replaced, 177willstayinservice,whilethefleetwillbeenlargedwith264largefreighters.Thetotalfreighter

fleetwillincreasefrom1605to2645aircraft.Only287aircraftwillstayinservice,while1318willbereplaced,thereforeenlargementoffleetis1040freighters(Figure6).

Considering the conversion – new‐built share, it could be noted that the conversion of largefreighters isnotusual, becausenewbuilt aircraft aremore reliable andhave loweroperatingcosts.Smallfreightersaresubjecttotheconversionbecauseoftheloweracquisitioncosts.BothmanufacturersAirbusandBoeingsimilarlyforecastedtheconversion–new‐buildshareinthetotalfleetinthenext20yearsofapproximatelyonethirdoffleettobenewbuilt,whiletherestwillbeconversion.Alsobothmanufacturersnotedthataircargofleetwillbedoubledinthenext20years.

Figure6.World’sfreighterfleetforecast(DatasourceAirbusGMT2014)

Boeing (2014) expects that theworld cargo capacitymeasured in available tonne‐kilometres(ATKs1) will be more than doubled in the next two decade (more than 1000 billion ATKs).Capacity in combi aircraft will not be offered. Although the increase of the cargo capacity inpassengeraircraftmeasuredinATKsispredicted,thededicatedfreighterswilltransportmorethanhalfrevenuetonne‐kilometres(RTKs).Thededicatedfreightersarepreferableduetotheiroperatingadvantages,especiallyonlonghaulroutesorforunusualshipmenttransport.Italsoshouldbenotedthatthebellycapacityhasincreasedwiththesizeofaircraft,whiletheavailablebellycapacityperpassengerhasremainedstableduringthelast25years.

5.EUROPEANAIRCARGOFLEET

Theaircargotransportisinfluencedbyeconomicconditionsonthemarket.TheeffectsofworldcrisiscanbeobservedthroughtheEurostatdatarelatedto32Europeancountriesfortheperiodfrom2001to2012.Itcanbeseenthatthetotalnumberofcargoaircraftinconsideredcountries

1 Available tonne kilometres (ATKs) is a measure of an airline's total capacity. It is calculated bymultiplyingcapacityintonnesandkilometresflown.

isreducedfrom427in2008to416in2012.Thenumberofcargoaircraftbycountriesin2012isgiven on Figure 1, while Figure 2 presents the share of cargo aircraft in the total fleet bycountries in 2012. Germany had the largest fleet consisting of 68 aircraft,while Portugal andEstoniahadonlyonecargoaircraft(Figure1).Theeightcountries(accordingtoEurostatdata)didnothavecargoaircraft in their fleet in2012.Serbiaalsodidnothavecargoaircraft in itsfleet,thusaircargowascarriedinthebellyofpassengeraircraft.ThelargestfleetintotalcanbeobservedinUnitedKingdomandGermany,consistingof1229and1146aircraftrespectively.

Consideringtheshareofcargoaircraft inthetotal fleet(Figure2)Luxembourghasthelargestshare of about 24%, that could be explained by the fact that large cargo airline Cargolux isoriginated from this countries. The share of cargo aircraft in Portugal fleetwas less than onepercent,whileinFranceandItaly,thepercentagewas2.3%.AlthoughtheshareofcargofleetinFrance in 2012was small, the number of cargo aircraft in its fleet is not negligible (14 cargoaircraft).

Figure1.Numberofcargoaircraftin2012(Datasource:Eurostat)

Figure2.Theshareofcargoaircraftinthetotalfleetin2012(Datasource:Eurostat)

AccordingtoEurostat(datarelatedto32Europeancountries),thenumberofcombiaircraftisinsignificant.Unlikethesedata,ZhangandZhang(2002)pointedthatmostpassengercarriersinAsiahavesubstantialcargobusinessandoperatecombi fleets.Thequickchangeaircraftwereused inEuropean countries. Its number in2012 is presentedonFigure 3.Thenine countriesownedquickchangeaircraftintheirfleet,andtheleaderswereFranceandUKwith15and14aircraftrespectively.

Figure3.Numberofquickchangeaircraftin2012(Datasource:Eurostat)

Figure4.ThegrowthofEuropeanfleet Figure5.FreightersdeliveryinEurope

(Datasource:AirbusGMT2014)

TheforecastedgrowthofEuropeanfreightersfleet(Airbus,2014)fornext20yearsispresentedonFigure4.Thefleetwillexpandfor95aircraft,222freighterswillbereplaced,while47willstayinservice.ConsideringdeliveriesinEurope(Figure5)itcanbeseenthatallsmallaircraftwillbeconvertedpassengersaircraft.Theshareofnewmid‐sizefreightersis23%,whilelargefreightersarein65%newones.

6.CONCLUSION

Thefleetplanningistheproblemofstrategicimportanceforanairline.Aircraftacquisitionandfleet renewal enables an airline to reduce the costs, becausenewer aircraft aremore flexible,fuelefficientandbettermatchthedemandonthemarket,especiallyonshort‐haulandmedium‐haul routes. New generation of freighters are designed to be more adoptable to the marketconditions,thereforeallowingtheairlinetobemorecompetitivewithothermodesoftransport.Forecasted trends, both in Europe and world, indicate the growth of air cargo fleets, whichmeansthataircargotransporthasbeenrecoveringfromrecession.

ACKNOWLEDGMENT

This research has been supported by the Ministry of Education, Science and TechnologicalDevelopment,RepublicofSerbia,aspartoftheprojectTR36033(2011‐2015).

REFERENCES

[1] Airbus, (2014). GMF 2014 Freight: Flying on demand, available athttp://www.airbus.com/company/market/forecast/,(March2,2015).

[2] Boeing, (2014). World Air Cargo Forecast 2014‐2015, available athttp://www.boeing.com/assets/pdf/commercial/cargo/wacf.pdf,(March2,2015).

[3] Eurostat,http://ec.europa.eu/eurostat/data/database,(March2,2015).

[4] Hsu, C.I., Li, H.C., Liao, P., Hansen, M.M., (2009). Responses of air cargo carriers toindustrialchanges.JournalofAirTransportManagement,15(6),330‐336.

[5] Zhang,A.,Zhang,Y., (2002). Issueson liberalizationofaircargoservices in internationalaviation.JournalofAirTransportManagement,8(5),275‐287.

[6] Wensveen, J.G., (2007). Air Transportation: A Management Perspective, AshgatePublishingLimited,England,AshgatePublishingCompany,USA,6thed.

PartV INFORMATIONAND

COMMUNICATIONTECHNOLOGIESINLOGISTICS

APPLICATIONOFREVENUESHARINGCONTRACTINTELECOMMUNICATIONSINDUSTRYSUPPLYCHAINS

BrankaMikavicaa*,AleksandraKostić‐Ljubisavljevića,VesnaRadonjićĐogatovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Revenue sharing contracts have been used in a number of industries and play animportantrole inmanagementof supplychains.Thispaperanalysespossibilityofapplicationofrevenue sharing contracts in telecommunications industry. The telecommunications industry ischaracterized by complex supply chain, with frequent services and tariffs introduction. In thisfiercelycompetitive industry,newtechnologiesandnewservicesrequirehighcapital investmentsin order to provide emerging services to the customers. Other important issues intelecommunications industryareuncertaintyofdemandandservice lifecycleshortening.Insuchdynamicenvironment,supplychainmanagementisessential.TelecommunicationssupplychainindeliveringcontentfromContentProvider,throughServiceProviderstothecustomers,isobserved.The goal of the analysis is formulation of revenue sharingmechanism that reduces provider’sincentivestoincreaseretailpriceandstimulatesimprovementofcustomerbaseandmarketshare.

Keywords:Revenuesharing,contract,telecommunicationsindustry,supplychains

1.INTRODUCTION

The telecommunications industry is characterized by rapid development of new technologiesand new services. Large capital investments are needed in order to make content available.Market players often introduce new services and tariffs to increase their market share.Considering fierce competition, these systems need to be effective, flexible and scalable toefficiently manage an increasingly complex service portfolio. Telecommunications industrymarketincludesresidentialcustomers,smallbusinessesandbigcorporatecustomers,accordingtoGupta(2008).Intheresidentialcustomersmarket,competitorsrelyheavilyonretailpricestoincreasetheircustomerbase.Successdependsonreputationandinvestmentinbillingsolutions.Thecorporatemarkethasdifferent featuresincomparisonwithresidentialcustomers.Mainly,big corporate customers are less price sensitive when value added services are beingconsidered.Telecommunicationsindustrysupplychainsareverycomplex,dynamicsystemsandingeneral involvemanydifferent typesofplayersanddeliverdifferentoutputsbyAgrelletal(2004). The roles and responsibilities in the supply chain are often changed due to theuncertaintyofdemand.

*b.mikavica@sf.bg.ac.rs

Coordination of these supply chains can be achieved applying revenue sharing contracts. Themainobjectiveofthispaperisobtainingtheappropriaterevenuesharingcontractthatreducesproviders’incentivestoincreaseretailpricesandimprovesproviders’marketposition.

Thepaperisorganizedasfollows.Afterintroductoryremarks,Section2givesbriefoverviewonrevenue sharing coordination of supply chains. Two types of revenue sharing contracts arepresented in Section 3. In this Section, we introduce two relevant parameters, provider’sreputationfactorandcustomer’swillingnesstopay.NumericalexampleispresentedinSection4.Finally,concludingremarksaregiveninSection5.

2.COORDINATIONOFSUPPLYCHAINSBYREVENUESHARINGCONTRACTS

Coordinationofsupplychainscanbepursuedbycentralizedordecentralizeddecision‐makingapproachaccordingtoGiannoccaroandPontrandolfo(2004).Ifthereisauniquedecisionmakerwithin the supply chain, decision‐making approach is centralized. Decentralized decision‐making approachoccurswhen there are several independent actors at thedifferent stages ofsupplychain.Supplychaincontractsareausefultooltomakeseveralactorsofadecentralizedapproachtobehavecoherentlyamongthemselves.Inacentralizedcontrolmechanism,auniquedecisionmaker possesses all relevant information on thewhole supply chain and contractualpower to implement such decisions. The centralized control assures the system efficiency.However, centralized control is often considered as not realistic. In such conditions,decentralizedcontrolismoreconvenient.Itincludesseveraldecisionmakers,pursuingdifferentobjectives,possiblyconflictingamongeachother.Locallyappropriatebehaviourisoftengloballyinefficient.Hence,coordinationmechanismsareneededinordertohavelocaldecision‐makerspursuechannelcoordination.Suchcoordinationmechanismsincludethesupplychaincontracts,which control transactions between actors in supply chain. In particular, the risk and therevenue,arisingfromdifferentsourcesofuncertaintyandchannelcoordinationrespectively,areshared by all actors of supply chain. Different models of supply chain contracts have beendeveloped, including quantity flexibility contracts, the backup agreements, the buy back orreturn policies, the incentivemechanisms, the revenue sharing contracts, the allocation rulesandthequantitydiscountsbyGiannoccaroandPontrandolfo(2004).RevenuesharingcontractshaveanimportantroleinthesupplychainsmanagementbyKrishnanandWinter(2011).Thesecontracts can be applied to industries where their demand is uncertain, forecasting demanddoesnotfollowawell‐definedtrendanddistributorscaneasilycontrolretailer’sprofit.Revenuesharing contracts inwhich retailers pay specified amount of revenue to suppliers arewidelyadopted,especiallyinthevideodistributionandmovieindustry.Accordingtotheseagreements,a retailer purchases a product from a supplier for a fixed price and shares a portion of therevenuewitha supplier.Thesupplieroffers thepurchaseprice for theproduct to theretailerwho defines the quantity of the product to be ordered by Parsule‐Desai (2013). Consideringcontent delivery from Content Provider (CP) to Service Providers (SPs), due to thecharacteristicsof telecommunicationssupplychain,shortservice lifecycle,andtheabsenceofinventory, there is no difference between purchase and demand. Schematic presentation oftelecommunication supply chain is shown in Figure 1. A major Electronics ManufacturingServicesProvider(EMS)maysupportmanyof theOriginalEquipmentManufacturers (OEMs),such as Nokia, Ericsson, Alcatel, etc. Global operators often choose different OEMs to supplydifferent countries and use more than one supplier within a country. Second and third tiersuppliers are supplying not only the next step in the supply chain, but also companiesdownstream.

Figure1.Schematicpresentationoftelecommunicationsupplychain

Theuncertaintyonthetelecommunicationsmarketandtheshortservicelifecyclemakereliableforecasts of supply chain characteristics more difficult. The reason lies in the uncertainty ofdemand,newoperatoremergenceandintroductionofnewtechnologies.

3.PROBLEMSTATEMENT

Let us consider telecommunications supply chain as shown on Figure 1. Emphasis of ourresearch is theanalysisof the lastsegmentofsupplychain,consistingofCP,SPandcustomerandtheirmutualrelations.Assumethattotalcustomerpopulation,denotedas ,isfixedduringobserved time interval. SPs compete on the common retailmarket for provisioning the samecontentfromCP.TotalnumberofSPsisdenotedbyN.Assumethatcontractsarebeingagreedamongprovidersforasinglecontent.QualityofService(QoS)isassumedtobesatisfactoryandall SPs offer the same service class. In order to obtain appropriate revenue sharing contractamong CP and SPs, such that provide market stability and ensure customers protection byreducingincentivesofSPstoincreasetheirretailprices,weanalysetwotypesofcontracts,staticanddynamic.Theaimofdynamicrevenuesharingcontractistostimulateproviderstoimprovetheirmarketshare,tostrengthenmarketpositionandthusincreasetheirrevenue,ratherthanbypriceenhancement.Relevantparametersinthesecontractsarecustomers’willingnesstopayspecificcontentbythegivenSPs’retailprice,andSPs’reputation.Customers’willingnesstopayreferstotheportionoftotalcustomers’populationthatareabletoaffordthemselvesobservedcontentbytheretailprice, .Regardingtothepricesensitivity,twotypesofcustomerscanbedistinguished,moreandlesspricesensitive.Thus,customers’willingnesstopaymathematicallycanbeexpressedasfollows:

,

, , ∈ 0,1 (1)

Parametersinthisequation(1), , , and dependonsocioeconomicstructureofcustomers,targetgroups,butalsoonsubstitutabilityandpopularityofcontentdeliveringfromCP,throughSPs,tothecustomers.WillingnesstopayisinverselyproportionaltothetotalnumberofSPsatthemarket.Figure2presentscustomers’willingnesstopayforarbitrarychosenvaluesforparametersandprice.Lesspricesensitivecustomersarelessflexibleintermsofpricevariation,andtheslopeoftheirwillingnesstopayhasslowdecayincomparisonwithmorepricesensitivecustomers.

Figure2.Pricedependenceofcustomers’willingnesstopay

Pricereductionleadstoenhancementofcustomers’willingnesstopayspecificcontentforbothtypes of customers.As a result, number of SPs’ end customerswill increase. Thus, one of themost important providers’ business goals, improved market share, will be satisfied. Anotherimportant parameter in proposed models is SPs’ reputation SPi’s reputation is denoted by, ∈ 0,1 .ThisvalueisestablishedonthebasisoflongtermbusinessexistenceofSPionthe

market,and∑ 1.ThismeansthatvalueforSP’sreputationisnormalizedandsumofthevaluesforallSPsequals1.WeassumethatreputationofallSPsonthemarketisknown.Higher

reputationisreasonwhyanumberofcustomersarewillingtopaycontentathigherpricerate,althoughQoSrequirementsaresatisfiedbyallSPs.Staticrevenue‐sharingcontractdefinesfixed,predetermined, portion of generated SP’s revenue from provisioning CP’s content on retailmarket that must be paid to CP. Thus, CP always obtains the same portion of revenue fromcontracts with SPs for the given content. Revenue of from provisioning content to thecustomers and CP’s revenue in accordance with the static revenue sharing contract can be,respectively,writtenasfollows: ,1 XrwpR iii

si

si (2)

N

iiii

si

sCP XrwpR

1

(3)

whereΦ , Φ ∈ 0, 1 presents fixed portion of revenue that, by the contract, iSP pays to

under static revenue sharing contract. Dynamic revenue‐sharing contract defines flexibleportionofrevenuethatSPpaystotheCP,dependingonSP’sretailprice.Portionof ’srevenuepaidtoCPunderdynamicrevenuesharingcontractcanbeexpressedasfollows: s

iidi p 1 (4)

In this equation (4), Δ presents variation of ’s retail price . Thus, revenue of fromprovisioning content to the customers and CP’s revenue under dynamic revenue sharingcontractcanbe,respectively,writtenasfollows:

,1 XrwpR iiidi

di (5)

N

iiii

di

dCP XrwpR

1

(6)

4.NUMERICALEXAMPLE

LetusconsiderthesituationinwhichrevenuesharingcontractshavebeenappliedtoasingleCPandtwoSPsontheretaillevel.CPoffersasinglecontentandnegotiateswithSPsontheportionof revenue thathas tobepaid on the revenue sharingbasis.We assumed that one SP is newentrant in the observed market, thus having lower reputation factor than other. Regardingcustomers’pricesensitivity,weassumedtwoscenarios.ThefirstscenarioreferstothesituationwhenSP1increasesitspriceuptothelevelofSP2’sretailprice,whileSP2remainsretailprice.The second scenario describes price reduction by the SP’s offering higher retail price. Thissituationiscommonwhenpromotionsanddiscountsarebeingapplied.Valuesforparametersinthe expression for customers’ willingness to pay are specified according to assumed marketsituation. We assumed that values of relevant factors for calculation of revenues are thefollowing: 200000, Φ Φ 0.5, 100, 120, 0.3, 0.7. All obtainedrevenuesareexpressedinmonetaryunits[MU].

Figure3presentsrevenuesofSP1understaticanddynamicrevenuesharingcontracts,formoresensitivecustomers,whenSP1increasesitsretailprice,whileSP2remainsitspriceatthesamelevel. Maintaining SP2’s retails price keeps its revenue the same. It can be noted that staticrevenue sharing increases SP1 revenue, while dynamic reduces for bothmore and less pricesensitive customers.Hence, under dynamic revenue sharing contract, SP1has no incentive toincreaseitsretailprice.

(a) (b)Figure3.RevenuesofSP1whenretailpricevariesfor(a)moresensitivecustomers,and(b)less

sensitivecustomers

Figure4presentsrevenuesofSP2whenitsretailpricedecreasesuntilitreacheslevelofSP1’sretail price. For both more and less sensitive customers, dynamic revenue sharing contractenhanceSP2’srevenue.SP1maintainsitsrevenue,sincetherewasnopricevariation.

(a) (b)Figure4.RevenuesofSP2whenretailpricevariesfor(a)moresensitivecustomers,and(b)less

sensitivecustomers

Retail prices enhancement ensures greater revenue for CP. However, it leads to reduction ofcustomers’willingnesstopayandreductionofmarketshareinlongterm.Also,pricevariationmustbeinaccordancewithpriceregulationinforce.Dynamicrevenuesharingcontractensureshigherrevenueincomparisonwithstatic,asshowninFigure5.

(a)(b)

Figure5.RevenuesofCPwhenSP1retailpriceincreasesfor(a)moresensitivecustomers,and(b)lesssensitivecustomers

SP2’spricereductionleadstothereductionofCP’srevenueforbothstaticanddynamicrevenuesharing contract, as well. This situation is shown in Figure 6 for both observed types ofcustomers.Staticcontractensuresgreaterrevenueincomparisonwithdynamic.

(a)(b)

Figure6.RevenuesofCPwhenSP2retailpricedecreasesfor(a)moresensitivecustomers,and(b)lesssensitivecustomers

5.CONCLUSION

This paper analyses characteristics of telecommunications supply chains and possibility ofapplicationofrevenuesharingcontractforsupplychaincoordinationbetweenContentProviderand Service Providers on the given market. Two types of contracts are observed, static anddynamic.TheformerestablishesfixedportionofrevenuethatServiceProviderpaystoContentProvider.ThelaterdependsonretailpriceandinvolvesfixedportionofrevenuethatSPpaystoCP,butinvolvesvariablepartwhichreflectsretailpricevariation.Theaimofdynamicrevenuesharing contract is to enlarge customer base by price reduction, and thus improve marketposition.Regardingpricesensitivity,twotypesofcustomersareobserved,moreandlesspricesensitive.Forallanalysedscenarios,dynamicrevenuesharingcontractsatisfiesgivengoals.OurfurtherresearchinthisfieldwillbedirectedtowardsintroductionofcompetitionattheContentProvider level, and introduction of content differentiation through content popularitydistinction,aswell.

ACKNOWLEDGMENT

This work is partially supported by Ministry of Education, Science and TechnologicalDevelopmentoftheRepublicofSerbiaunderNo.32025.

REFERENCES

[1] Agrell,P.J.,Lindroth,R.,Norrman,A.,(2004).Risk,InformationandIncentivesinTelecomSupplyChains.Int.J.ProductionEconomics,90(1),1‐16.

[2] Giannoccaro, I.,Pontrandolfo,P., (2004).SupplyChainCoordinationbyRevenueSharingContracts.Int.J.ProductionEconomics,89(2),131‐139.

[3] Gupta,A.,(2008).PursuitofthePerfectOrder:TelecommunicationsIndustryPerspectives.BPTrends.

[4] Krishnan, H.,Winter, R.A., (2011). On the Role of Revenue‐Sharing Contracts in SupplyChains.OperationalResearchLetters,39(1),28‐31.

[5] Parsule‐Desai, O.D., (2013). Supply Chain Coordination Using Revenue‐DependentRevenueSharingContracts.Omega,41(4),780‐796

BIGDATA:CHALLENGESANDOPPORTUNITIESINLOGISTICSSYSTEMS

BrankaMikavicaa*,AleksandraKostić‐Ljubisavljevića*,VesnaRadonjićĐogatovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Advancements in telecommunications and computer technologies have led toexponentialgrowthandavailabilityofdata,bothinstructuredandunstructuredforms.ThetermBigDatamainlyrefers toenormousdatasetscontaining largeamountofunstructureddata thatrequiremore real‐time analysis.Great potential and very useful values are hidden in this hugevolumeofdata.The influenceofBigData is recognized in logistics services, turning large‐scaledatavolumesintoauniqueassetcapableofboostingefficiencyinareasofthebusiness.Thispaperanalyses benefits and opportunities of Big Data in logistics systems. Challenges and risks thatlogistics systems, affected with this phenomenon deal with, are highlighted. The paper alsoproposessomeefficientwaysofexploitingthevalueofBigDatainlogisticssystems.

Keywords:BigData,logisticssystems,challenges,opportunities

1.INTRODUCTION

Withthepermanentincreaseofdata,scalinguptonoteworthyamounts,generatedbyInternet‐basedsystems,BigDatahasemergedasanewresearchfield.ThecoreofBigDataparadigmisthe extraction of knowledge fromdata as a basis for intelligent services and decisionmakingsystems. It encompasses many research topics, disciplines and it investigates a variety oftechniques and theories from different fields, including data mining, machine learning,informational retrieval, analysis etc. Big Data has indubitably become important trend inlogisticssystems.Optimizationofserviceproperties,suchasdeliverytime,resourceutilizationandgeographicalcoverage,isapermanentchallengeinlogisticssystems.Large‐scaleoperationsin logistics systems require data in order towork efficiently. Optimization results depend oninformationaccuracyandavailability.Integrationoflogisticsprovidersandcustomeroperationsprovides comprehensive knowledge related to supply chain risks. The transport and deliverynetworkisimportantdatasource.Besidenetworkoptimizationitself,networkdatamayprovideinsightonglobalflowofgoods.Hence,BigDataanalyticsemphasizeamicroeconomicviewpoint.Big Data concept provides multilateral analytics in order to enable the insight on customerexperienceandproductquality.

Thepaper isorganizedas follows.After introductoryremarks,asurveyofsomeproposedBigDatadefinitions,toolsandtechniquesisgiven.

* b.mikavica@sf.bg.ac.rs

The third section analyses potential benefits and opportunities of Big Data application inlogisticssystems.StepsforefficientexploitationofthevaluesinBigDataarealsopresentedinthissection.Theforthsectionobserveschallengesandrisksthatlogisticssystemsneedtocopewith in order to ensure efficientbusinessplatforms. Concluding remarks are given in the lastSection.

2.BIGDATADEFINITIONS,TOOLSANDTECHNIQUES

Considering vast increase of global data and advancement of information technology whichenabledatageneration,thetermBigDataismainlyusedtodescribeenormousdatasetswhichtypically includemassesofunstructureddata requiringmore real‐timeanalysisbyChenetal.(2014). Itprovidesnewopportunities fordiscoveringnewvalues, betterunderstanding thosevalues,butalsoraisesnewchallengesincludingorganizationandmanagementofthesedatasets.Besidemassesofdata, therearemanyother featuresdetermining thedifferencebetweenBigDataand“massivedata”or“verybigdata”.Thesedifferencesanddifferentconcernsinscientificand technological aspect cause different definitions of BigData. BigData can be described asdatasetswhichcouldnotbecaptured,managed,andprocessedbygeneralcomputerswithinanacceptablescopebyChenetal.(2014).Inaccordancewiththisdefinition,BigDatapresentsthenext frontier for innovation, competition, and productivity by MGI (2011). It refers to suchdatasetswhichcouldnotbeacquired, stored,andmanagedbyclassicdatabasesoftware.ThismeansthatvolumesofdatasetsconformingtothestandardofBigDataarechanging,increasingovertimeorwithtechnologicalimprovement.Also,thesevolumesindifferentapplicationsdifferfromeachother.MostcommonexplanationofBigDatais3Vs(Volume,Velocity,Variety)modelbyLaney(2001).Volumereferstovastdatascaleincreasingwiththegenerationandcollectionofmasses of data. Timeliness of Big Data is described by Velocity. Hence, data collection andanalysismustbe rapidlyand timely conducted inorder toutilize the commercialvalueofBigData. Variety indicates the various data types, including unstructured, semi‐structured andtraditionalstructureddataaswell.AnotherdefinitiondescribesBigDataasanewgenerationoftechnologiesandarchitectures,designedtoeconomicallyextractvaluefromverylargevolumesofawidevarietyofdata,byenablingthehigh‐velocitycapture,discovery,and/oranalysisbyIDC(2012).Thus,characteristicsofBigDatacanbesummarizedas4Vs(Volume,Velocity,VarietyandValue).ThisdefinitioniswidespreadsinceitemphasizesthemeaningandnecessityofBigData including exploration of huge hidden values from datawith an enormous scale, varioustypes, and rapid generation.AddingVeracity as a key characteristic ofBigData, 5Vsmodel iscreatedbyHassanienetal(2015).MoredetaileddefinitionbyBeyerandLaney(2012)describesBigDataashigh‐volume,high‐velocity,and/orhigh‐varietyinformationassetthatrequirenewforms of processing to enable enhanced decision making, insight discovery and processoptimization.Hence,datasetcanbereferredasBigDataifitisformidabletoperformcapture,storage,distribution,management,analysisandvisualizationonitatthecurrenttechnologies.

TherearemanyproposedtechniquesandtechnologiestocapturevaluefromBigData,analyzeand visualize these data by Philip Chen and Zhang (2014). However, they are not entirelycapable to meet the requirements. These techniques and technologies combine a number ofdisciplines,includingcomputerscience,economics,mathematics,statisticsandotherexpertise.MultidisciplinarymethodsareneededfordeterminationofvaluableinformationfromBigData.These techniques involve many different disciplines, such as statistics, data mining, machinelearning, neural networks, social network analysis, signal processing, pattern recognition,optimizationmethods, visualizationapproachesand theyoverlapwitheachother. Statistics isusedtocollect,organize,interpretdataandtoprovidenumericaldescriptions.Thesetechniquesare used to determine correlations and causal relationships between different objectives.However, standard statistical techniques are not entirely convenient for management of BigData.Manyextensionsofclassicalstatisticaltechniquesorcompletelynewmethodshavebeenproposed by Di Ciaccio et al (2012). Data mining is a set of techniques to extract valuable

information from data, including clustering analysis, classification and regression. Big Datamining is more complex in comparison with traditional data mining algorithms. Machinelearning is aimed to design algorithms that allow computers to evolve behaviors based onempirical data. Important characteristics of machine learning are discovery knowledge andmaking intelligent decisions automatically. Considering Big Data, both supervised andunsupervisedlearningalgorithmsneedtobescaledup.

3.BENEFITSANDOPPORTUNITIESOFBIGDATAAPPLICATIONINLOGISTICSSYSTEMS

Thebarriers toextractingvalue fromBigDatacanovercomethroughsystematicplanbyWEF(2014).The firststep is todefineresponsibilitiesandroles forcollectionandanalysisofdata.ThenextstepisdeterminehowBigDatamightbevaluable,sincetheaimofBigDataisnotdatabyitself,butdiscoveringinsightsthatcanleadtovaluableoutcomes.Valuablebusinessinsightcan be obtained from various sources, including social media, activity streams, machineinstrumentation,operationaltechnologyfeedsanddatacurrentlyunusedbuthavealreadybeencaptured.Thesesourcesneedtobeexploredinordertofindnewwaysofcapturinginformation,suchascomplexeventprocessing,videosearchandtextanalytics.BigDatainitiativesshouldbelaunched in business functions for which the potential payback is high. Functions such asmarketing,customerservice,supplychainmanagementandfinancearepoisedforthegreatestgrowth.AnotherstepistomatchBigDatainitiativeswithcompatiblebusinessfunctions.IssueofgreatimportanceisdeterminationweatherBigDatawillyieldvaluableinformationunavailablethroughtraditionalbusinessanalytics.Also,complexities,prioritiesandtechnologyarchitectureneed to be assessed accordingly. Finally, multidisciplinary team of business and technologyexpertsisthekeyforsuccess.

TheadvancementsintechnologicalandmethodologicalaspectofBigDataprovidegreatbenefitstothelogisticssector.Logisticsprovidersmanageenormousflowofproductsthuscreatingvastdatasets.Origin,destination,size,weight,contentandlocationofshipmentsoneverydaybasisare tracked across global deliverynetworks.There is greatunutilizedpotential for improvingoperational efficiency, customer experience and creating new business models. Big Dataanalytics provides competitive advantage through propertieswhich highlightwhere Big Datacanbeeffectivelyapplied in the logistics industryaccording toDHL(2014).Someof themostimportantfieldsofBigDataapplicationinlogisticssystemsareshowninFigure1.

Figure1.BigDataApplicationinLogisticsSystems

Big Data analytics can accelerate business processes and increase the level of operationalefficiency enabling last mile optimization. This goal can be achieved through real‐timeoptimizationofdeliveryroutesorutilizingdataprocessinginordertocontrolanentirelynewlast‐mile deliverymodel. For suchpurposes sensor‐baseddetection of shipment items canbeused or automatically changing of delivery routes according to current traffic conditions. TheautomatedcontrolofalargenumberofrandomlymovingdeliveryresourcesrequiresextensivedataprocessingcapabilitiesofBigDatatechniques.Thiscandecreaselast‐miledeliveringcosts,

especially in rural and seldom populated areas. Optimal utilization of resources is one of themostimportantcompetitiveadvantagesforlogisticsproviders.Hence, logisticsprovidersmustapplyresourceplanningonstrategicandoperationallevel.Planningonstrategiclevelinvolvesthelong‐termconfigurationofthedistributionnetwork,whileoperationallevelplanningscalescapacitiesonadailyormonthlybasis.Inbothcases,BigDatatechniquesimprovethereliabilityof the planning and enable logistics providers to optimally match demand and availableresources.Inordertosignificantlyincreasepredictivevalue,amuchhighervolumeandvarietyofdataisexploitedbyadvancedregressionandscenariomodelingtechniques.Thisresultsinanewqualityofplanningwithgreaterforecastperiods.Hence,theriskoflong‐terminfrastructureinvestmentsandcontractedexternalcapacitiesisreduced.Inthecaseofplanningonoperationallevel, transit points and transportation routes must be efficiently coordinated on day‐to‐daybasis. Operational planning tasks are often based on historical averages or even on personalexperience.Asaresult,resourceutilizationisinefficient.Real‐timeinformationaboutshipmentsis aggregated to predict the allocation of resources. This data is automatically sourced fromwarehousemanagementsystemsandsensordataalongthetransportationchain.Also,detectionofchangesindemandisderivedfromexternallyavailablecustomerinformation.Thepredictionof resource requirements enables scaling capacity in each location. In addition, it revealsupcomingcongestionsonroutesorattransitpointswhichcannotbeaddressedbylocalscaling.Thedistributionnetworkcanbecomeself‐organizinginfrastructureusingBigDataanalytics.

Acquisitionofcustomer insight isofgreat importance intheaspectofBigDataanalytics.Datafrom the distribution network carries meaningful value for the analysis and management ofcustomer relations. Application of Big Data techniques enables understanding of customerdemand. Big Data analytics allow a comprehensive assessment of customer satisfaction bymerging multiple extensive data sources. In order to achieve the insight across the entirecustomerbase, the logisticsprovidermustmergemultipledatasources.BigDataanalyticsareessential increatinganintegratedviewofcustomerinteractionsandoperationalperformance,ensuring satisfaction of both sender and recipient. In order to get accurate results fromcustomer feedback evaluation, it is necessary to aggregate information from as many touchpoints as possible. The open research problem in logistics systems and supply chainmanagementcanbeanalyzedfromtheperspectiveofmanagerialbusinesscomponentsandthedifferent category of stakeholder, where main business functions are forecasting, inventorymanagement, transportation management, transport and human resources by Robak et al(2014).Issuessuchaspredictionoftimedelivery,timelyresponsetocustomerexperience,real‐time planning of capacity availability, inventory, customer and supplier relationshipmanagementcanbeaddressedbyBigData.Consideringdatanotonlyasinformationasset,butalsoasastrategicasset,organizationsinsupplychainmanagementcanrealizeeconomicvaluein the data using Big Data analytics through revenue generating activities by Rozados andTjahjono(2014).

4.CHALLENGDESANDRISKSOFBIGDATAINLOGISTICSSYSTEMS

The effective use of Big Data techniques introduces great advantages in economytransformation, but also raises many challenges, including, among others, difficulties in datacapture, storage, searching, shearing, analysis and visualization. These challenges need toovercome in order to exploit capabilities of Big Data. Computer architecture is one of thegreatest challenges. Central Processing Unit (CPU) performance is doubling each 18 months,accordingtotheMoore’slawbyPhilipChenandZhang(2014).Performanceofthediskdrivesisalso doubling at the same rate, but rotational speed of the disk has slightly improved. Inaddition,theamountofinformationincreasesexponentially.Thishasabigimpactonlimitationofreal‐timevaluesdiscoveryfromBigData.AnotherimportantchallengerelatedtotheBigDataanalysis includes data inconsistence and incompleteness scalability, timeliness and datasecurity. Hence, data must be appropriately constructed and a number of preprocessing

techniques,suchasdatacleaning,dataintegration,datatransformationanddatereductionneedtobeapplied inorder toalleviatenoiseandcorrect inconsistencies.BigDatahassignificantlychanged data capturing and storing, including data storage device, data storage architecture,data access mechanism. The knowledge discovery process puts the highest priority on theaccessibility ofBigData. In that sense,BigData shouldbe accessed efficiently and enabled tofullyorpartiallybreaktheconstraintofcomputerarchitecture.Direct‐attachedstorage(DAS),network‐attached storage (NAS), and storage area network (SAN) are often used storagearchitecture. However, they have severe drawbacks and limitations in large‐scale distributedsystems. Optimizing data access is common way of improving the performances of data‐intensive computing. This includes data replication, migration, distribution and accessparallelism.Whendata volume is enormous, networkbandwidth capacity is thebottleneck incloud and distributed systems. Another issue related to cloud storage is data security. Datacuration is aimed at periodically data discovery and retrieval, data quality assurance, valueaddition, reuse and preservation. This includes authentication, archiving, management,preservation,retrievalandrepresentation.

Inlogisticssystems,acomprehensiveanalyticsframeworkrequiresintegrationofsupplychainmanagement, customer management, after‐sales support and advertising by Kambatla et al(2014). Enormous amounts of multi‐modal data including customer transactions, inventorymanagement,store‐basedvideofeeds,advertisingandcustomerrelations,customerpreferencesand sentiments, sales management infrastructure, and financial data among others.Comprehensive deployment of RFIDs to track inventory, links to supplier’s databases,integration with customer preferences and integrated financial systems provide improvedefficiency. Big Data approach facilitates exploitation of RFID‐enabled manufacturing data forsupporting production logistics decision‐makings by Zhong et al (2015). These applicationsmostly have relatively well structured and integrated datasets. Since infrastructure and dataanalysisisbeingperformedinthesamesecuritydomain,privacyandsecurityissuesareeasiertohandle.Themajorbottleneckinthisdomainisthedevelopmentofanalyticscapabletoscalevastamountsofmultimodaldata.

With increasing data volume, the probability that the data contain valuable and confidentialinformationraises.Thus,informationstoredforthepurposeofBigDataanalyticsisvulnerableforcybercriminalbyKshetri(2014).Besides,availabilityofpersonaldatacanbeusedtocreatevalue.Anotherimportantissueisdeterminationofrelevancewithinenormousdatavolumeandusage of Big Data analytics for creating value from relevant data. Using such data, differentproducts can be offered to different groups through quality discrimination and differentialpricing. BigData analytics enables determination of variableswith amuch higher correlationthaninnon‐BigDatatechniques.Italsodesignsofferingsandsetpricesbasedonsuchvariables.Combination of structured and unstructured data for various sources, hidden connectionsbetweenoutwardlyunrelateddatacanberevealed.Securityproblemsalso include intellectualpropertyprotection,personalprivacyprotection,commercialsecretsandfinancial informationprotection by Philip Chen and Zhang (2014). Data protection laws are already established inmost developing and developed countries. For Big Data related applications, data securityproblemsarehardertodealwithbecauseofextremelylargeamountofBigDataandmuchmoredifficultworkloadofthesecurity.

5.CONCLUSION

Thispaperpresentsoverviewofbenefits,businessopportunitiesofBigDatainlogisticssystems,butalsoemphasizeschallengesandpotentialrisksinthisdomain.BigDataanalyticsisstillintheinitial phase of development, since existing Big Data tools and techniques are not capable toentirelymeettherequirements.Moreeffortsfromvariousfieldsofexpertiseneedtobeinvolvedin order to more efficiently exploit all hidden values in vast datasets. There are numerous

obstaclesandchallengestodealwith,suchasdataquality,privacy,technicalfeasibility,amongothers,beforeBigDatacanachievewidespreadinfluenceinthelogisticssector.Inthelongrun,thesechallengesarelikelytobesolved,consideringventuresomecharacterofBigData.BigDatahave a potential to improve operational efficiency, customer experience and to create newbusinessmodels.Therefore, it isreasonable topredict thatBigDatawillbe importantsuccessdriverinlogisticssector.

ACKNOWLEDGMENT

This work is partially supported by Ministry of Education, Science and TechnologicalDevelopmentoftheRepublicofSerbiaunderNo.32025.

REFERENCES

[1] Beyer,M.,Laney,D.(2012).TheImportanceofBigData:ADefinition.Gartner.

[2] Chen,M.,Mao,S.,Liu,Y.(2014).BigData:ASurvey.MobileNetworksandApplications,19(2),171‐209.

[3] DHL.(2014).BigDatainLogistics

[4] DiCiaccio,A.,Coli,M.,Ibanez,A.,Miguel,J.(2012).AdvancedMethodsfortheAnalysisofLargeData‐Sets,Springer.

[5] Hassanien,A.E., Snasel,V.,Kacprzyk, J. (2015).BigData inComplexSystems,ChallengesandOpportunities,Springer.

[6] IDC.(2012).MarketAnalysis,WorldwideBigDataTechnology2012‐2015Forecast.

[7] Kambatla,K.,Kollias,G.,Kumar,V.,Grama,A.(2014).TrendsinBigDataAnalytics.JournalofParallelandDistributedComputing,74(7),2561‐2573.

[8] Kshetri, N. (2014). Big Data’s Impact on Privacy, Security and Consumer Welfare.TelecommunicationPolicy,38(2014),1134‐1145.

[9] Laney.D.,(2001).MetaGroup.ApplicationDeliveryStrategies.

[10] McKinseyGlobalInstitute.(2011).BigData:TheNextFrontierforInovation,Competition,andProductivity.

[11] Philip Chen, C.L., Zhang, C.Y. (2014). Data‐intensive applications, challenges, techniquesandtechnologies:AsurveyonBigData.InformationSciences,275(2014),314‐347.

[12] Robak, S., Franczyk, B., Robak, M. (2014). Research Problems Associatedwith Big DataUtilization in Logistics and Supply Chain Design and Management. Annals of ComputerScienceandInformationSystems,3(1),245‐249.

[13] Rozados,I.V.,Tjahjono,B.(2014).BigDataAnalyticsinSupplyChainManagement:Trendsand Related Research. 6 International Conference on Operations and Supply ChainManagement,Bali.

[14] WorldEconomicForum.(2014).TheGlobalInformationTechnologyReport:RewardsandRisksofBigData.

[15] Zhong, R.Y., Huang, G.Q., Lan, S., Dai, Q.Y., Zhang, T. (2015). A Big Data Approach forLogisticsTrajectoryDiscovery fromRFID‐enabledProductionData. International JournalofProductionEconomics,acceptedmanuscript.

MACHINE‐TO‐MACHINECOMMUNICATIONSTOWARDSMARTLOGISTICSSYSTEMS

NatašaGospić,BojanBakmaz

UniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Transport and logistics companies face continuous competitive pressure tomaximizecapacity and improve efficiencies of their infrastructure as well as to meet the increasinggovernmental regulation and compliance demands. Deployment of new communicationstechnologies, e.g. cloud computing, Internet of Things (IoT) and Machine‐to‐Machine (M2M),representsaprospectivesolutionforachievingtheseobjectives.Owingtolowpower,costefficiencyandlowhumanintervention,M2Mcommunicationhasbecomeamaindrivingforceforanumberofwidevarietyofreal‐timeapplications.Theexpectedhugenumberofinterconnecteddevicesandthesignificantamountofavailabledataopennewopportunitiestocreateservicesthatwillbringtangiblebenefits to the logistics companies,aswellas to theendusers.Thispaper representsacomprehensive survey on M2M communications considering some open research issues andchallengestowardtherealizationofsmartlogisticssystems.

Keywords:Logisticssystems,M2Mcommunications,mobilenetworks.

1.INTRODUCTION

Machine‐to‐machine (M2M) communications have emerged as a paradigm which providesubiquitousconnectivityamongdevices(objects)withouthumanintervention.ThemainideaofM2Mcommunicationsistoenablemechanicalandelectricalcomponentstobeinterconnected,networked,andcontrollableremotely,withlow‐cost,scalable,andreliabletechnologies.Currentmarket penetration and recent predictions confirm that M2M system deployments areincreasing exponentially (Cisco, 2014). This is driven by the needs of industries to automatetheirreal‐timemonitoringandcontrolprocessesaswellas the increasingpopularityofsmartapplicationstoimprovethelivingstyle.

Achievingbettercostefficiency,M2Mcommunicationshasbecomeamarket‐changingforceforawidevarietyofreal‐timeapplications,suchassmartenvironments,industrialcontrol,securityandemergencies,e‐healthcare,transportandlogistics,etc. InVodafone(2014)report,transportand logisticsareconsideredasprospective leaders in theuseofM2Mcommunications, takingintoaccountsignificantbenefitsofimplementation.Accordingcomprehensivemarketresearch,12% of all companies in this field already deployed some solutions ofM2M communications.Moreover, the fact that great majority of them (67%) achieved significant return of theirinvestments for a very short time, is of great importance. Major benefits of M2Mcommunications deployment in transport and logistics sectors can be identified as costreduction, better customer service, business agility, improved productivity, consistency of

delivery across markets, etc. Generally, M2M communications can be observed as promisingtechnologicaltoolforoptimizationofcomplexindustrialprocesses.

This paper is organized as follows. First, general end‐to‐end architecture for M2Mcommunications is presented. Then we focus on some specific properties of M2McommunicationswhichareofimportanceforapplicationsofM2Msystemsinlogisticsservices.Finally, prospective solution for realization ofM2M communications overmobile networks ispresented.

2.GENERALNETWORKARCHITECTUREFORM2MCOMMUNICATIONS

The characteristics of M2M communications are quite different from those of conventionalnetworks(Kimetal.,2014).M2Mnetworksarecomposedoflargenumbersofnodes,sincethemain subject participating in communication is amachine (device). Becausemost devices arebatteryoperated,energyefficiencyisthemostimportantissue.Asforthemachinesenses,itselforitssurroundingphysicalenvironment,thetrafficgeneratedbydeviceisverysmall.However,data are generated from a large number of objects, and because the data generation period,amount, and format are all different, a large quantity of data is generated. While M2Mcommunicationcanoccurwithouthuman intervention,operational stabilityandsustainabilityarealsorequired.

In2009,theEuropeanTelecommunicationsStandardsInstitute(ETSI)hasestablishedtheM2MTechnical Committee with the purpose to develop an end‐to‐end architecture for M2Mcommunications.AccordingtoETSI,anM2Msystemiscomposedofthefivekeyelementswithfollowingfunctions:

M2Mcomponent,embeddedinasmartdevice,transmitsdataorrepliestorequests. M2M gateway enables connectivity between the M2M components and the

communicationnetworks. M2M server works as a middleware layer to pass data through various application

services. M2M area network provides connectivity between M2M components and M2M

gateways. M2M communication network provides connection betweenM2M gateways andM2M

servers.

These key elements constitute the general M2M communication architecture in the threeinterlinkeddomains, i.e., theM2Mdevicedomain,networkdomainandapplicationdomainasshowninFig.1.

Figure1.GeneralM2Mcommunicationarchitecture

IntheM2Mdomain,apotentiallylargenumberofnodesandM2Mgateway(GW)areintegratedto enable automated and diverse services. Each embedded node as flexible and smart device

should be equippedwith various functions, such as data acquisition, data preprocessing, datastorage, distinctive address, communication interface, power supply, etc. They can makeintelligentdecisionandtransmitthesensorydatatotheGWinsingle‐hopormultihopmanner.TheM2MGW is an integrateddevice.After collecting thepackets fromembeddednodes, it isabletointelligentlymanagethepacketsandprovideefficientpathsforforwardingthesepacketstotheremoteback‐endserverviawired/wirelessnetworks.

In the network domain, a large number of heterogeneous points of attachment potentiallycoexist.Here,convergenceofheterogeneousnetworksinanoptimalwayprovidescost‐effectiveandreliablechannelsforsensingdatapackettransmissionfromM2Mtotheapplicationdomain.

Finally, in the application domain, various real‐time services for remote managementmonitoringareprovidedandcanbeclassified intoseveral categories, suchas traffic, logistics,business,home,etc.Back‐endserveristhekeycomponentforthewholeM2Mcommunicationsystem.ItformstheintegrationpointforallcollecteddatafromM2Mdevicedomain.

M2Mdevicescanbeeitherstationary(e.g., smartmetersathomes,vendingmachines,etc.)ormobile(e.g., fleetmanagementdevicesinvehicles,e‐healthsensors,etc.).TheaccessnetworksconnectM2M devices to the core network using eitherwired orwireless links. Although thewiredsolutioncanprovidehigherdatarates,reliability,securityandlowlatency,itmaynotbeadequate for the all M2M applications due to it cost ineffectiveness, lack of scalability andmobility support.Wireless access can be either capillary/short range (e.g., ZigBee, Bluetooth,Wireless Fidelity ‐ WiFi, etc.) or cellular (e.g., Long Term Evolution ‐ LTE, WorldwideinteroperabilityforMicrowaveAccess‐WiMAX,etc.).Wirelesscapillarysolutions,mainlyusedforsharedshortrangelinks,arerathercheaptorollout,andgenerallyscalable.However,smallcoverage,lowdatarates,weaksecurity,severeinterference,andlackofuniversalinfrastructurepose restriction on their applications to M2M communications. On the other hand, mobilesystems offer wide coverage, mobility support, satisfactory security, and ready‐to‐useinfrastructure, making cellular networking a promising solution for M2M communications.Therefore,currentandnextgenerationmobilesystems(i.e.,LTEandLTE‐Advanced)areinthemainfocusofrecentresearchesasstatedbyGhavimiandChen(2015).

3.PROPERTIESOFM2MCOMMUNICATIONSANDAPPLICATIONSINLOGISTICSSERVICES

The properties of M2M communications are quite different to traditional human‐basedcommunications.Whilehuman‐to‐human(H2H)andhuman‐to‐machine(H2M)communicationsobey a certain session length, data volume and interaction frequency, M2M communicationsfollow some very specific traffic properties. M2M devices generate heterogeneous trafficpatterns, including periodic (uniform), event‐driven (stochastic), and multimedia streaming(heavytailed),dependingontheirapplications.Thesevarioustrafficpatternsaregeneratedbythemassivenumberofnodes,whichhavedifferentpatterns,generationperiods,andgeneratedintensities.AsmostofM2Mdevicesare reportingsensordata, suchas temperature,pressure,humidity, etc., the transmitted packets consists of themeasured data plus the correspondingprotocol overhead. In general, this overhead is kept as small as possible, whereas the actualpayloaddiffersaccordingtoitsapplication.Althoughtheexpectedintensityofgeneratedtrafficper device is relatively low (in order of few kbit/minute) the share ofM2M traffic will soonexceed5%ofglobalInternettraffic,havinginmindhugenumberofdevices.

The combination of M2M communications and intelligent objects can largely improve thesupervisionof logisticprocesses(Palafox‐Albarranetal.,2012).Withan increasingnumberofvehiclesontheroad,thetransportationandlogisticsserviceswillbecomebigmarketforM2Mcommunication technology.VehiclesequippedwithM2Msensorsandactuators,becomeM2Mcommunicationentities.Furthermore,roadsandtransportedgoodsuseM2Msensorsandtags(e.g.,RadioFrequencyIdentification‐RFIDandNearFieldCommunication‐NFC)thatcanalsosend valuable information to the M2M control centers and logistics companies to route the

vehicles,monitorthestatusofthetransportedgoods,seamlesslytrackthephysicallocationsoffleetvehicles,anddeliverupdatedscheduleinformationtocustomers.

Today, a large number of container cargo ships are traveling through international waters.Thesecontainercargodeliveryservicesmayrisktheft,physicaldamage,deliverydelays,piracy,and even ship sinking. M2M technology provides solutions that are being used in fleetmanagementtoacquireabettercontrolwherecargoscanberapidlydeliveredacrossdifferentcontinents.TheM2Mapplicationsenablethetrackingofvehiclesandcargocontainerstocollectthedataon locations, fuel consumption, temperature, andhumidity, in order to increase fleetsafety, reduce the accident rates, and increase the productivity of a logistics company. Withmore precise information, greater control, better resource management, and higher costeffectiveness,afleetbusinesscanbeabletomaintainitscompetitivenesswiththehelpofM2Mtechnology.

Goods supply chain can work in a more efficiency way as M2M communications providepossibility to track the status of goods in real‐time. The M2M logistics enables ubiquitoussurveillanceonthestatusofproducts,rawmaterials, transportation,storage,saleofproducts,and aftersales services by keeping an eye on temperature, humidity, light, weight, etc. If thestatushassomeproblem,theM2Mdevicescanautomaticallysendanalert totheM2Mservervia the LTE/LTE‐A core network. Furthermore, it is also possible to track the inventory in awarehouse so that stockholders and enterprises can respond to themarket dynamics and todecidewhentorefillandwhentogoonsale.Therefore,thiscansignificantlyreducethesizeofwarehouse,thewaitingtimeofcustomers,andthenumberoftheemployeesinordertosavetheoperationalcostsforbusinessentities.

4.HIERARCHICALCELLULAR‐CENTRICM2MARCHITECTURE

As previously stated, cellular networks are considered as a prospective solution for M2Mapplications, especially for high mobility services such as smart transportation systems andlogistics. CurrentM2M solutions already use second and third generation (2G, 3G) ofmobilenetworks,butwithexpectedexponential increasingofM2Mdevicesdensityandconsequentlytraffic intensity, their capacity will soon be overcome. Thus, implementation of M2Mcommunicationsinnextgenerationcellularsystems(i.e.,LTE/LTE‐A)isunavoidable.Moreover,M2M communications are declared by Boccardi et al. (2014) as one of the five disruptivetechnology directions for the fifth generation (5G) mobile networks. Recently, flagshipstandardization bodies in this field, Third Generation Partnership Project (3GPP, 2012) andInstitute of Electrical and Electronics Engineers (IEEE, 2012) have introduced someenhancementsofcellularsystemsarchitectureinordertosupportM2Mcommunications.

Based on ETSI general architecture, complemented with 3GPP LTE‐A and IEEE 802.16penhancements, it is possible to define a cellular‐centric architecture proposed by Lo at al.(2013). It can provide useful reference architecture for designing a complete M2Mcommunications system. As shown in Fig. 2, the resulting M2M network architecture ishierarchical,consistingoffourtiers.FirsttierconsistsoftheM2Mapplication(M2M‐A)servicesandserver(M2M‐S). Insecondtier,anewfunctionalentityM2Mrelay(M2M‐R)isintroduced.M2M‐RisanextensionoftheconventionalLTE‐ArelayfunctionalityandcanbeusedasanM2Mdataaggregator. In thiscase,M2M‐Raggregatesdataunits frommultipleM2Mdevices (M2M‐Ds)intoasinglelargepacketfortransmissiontothesameordifferentM2M‐Ss.M2M‐Rincreasesystemcapacityand,moreimportantly,reducestransmissionpowerofM2M‐Dsthathavepowerconstraints. At third tier, the M2M gateway (M2M‐G) ensure M2M‐Ds interworking andinterconnection to theM2Mnetworkdomain.UnlikeM2M‐R, theM2M‐Gsupportsmulti‐radioaccesstechnologies(e.g.,RFID,ZigBee,lowpowerBluetooth,WiFI,etc.)inadditiontoLTE‐A.Inprinciple, the M2M‐G can assume the role of M2M‐R. However, such a solution might notnecessarily lead tooptimalperformancebecause thedeploymentofM2M‐G isdictatedby the

coverage of the supported wireless technologies. On the other hand, an M2M‐R can bestrategicallydeployedtoachievegoodcoverageandsignalquality,resultingwithbetterqualityofservice.Typically,anM2M‐Rcanbepositionedtoattainline‐of‐sightwiththeLTE‐AevolvednodeB(eNB).Thus,legacy(non‐3GPPcompliant)M2M‐DsareconnectedtotheM2M‐Gatfourthtier,whilelowmobility,powerandlocation‐sensitive3GPPM2M‐DsareconnectedtoM2M‐Ratupper tier. Furthermore, M2M‐Ds and M2M‐Gs with high mobility requirements are directlyconnectedtotheLTE‐AeNBatsecondtier.

Figure2.Hierarchicalcellular‐centricM2Marchitecture

WhenanM2M‐AdataunitissentfromtheM2M‐DtotheM2M‐S,everyprotocollayeroneachnetwork entity adds its own header. In addition to the entity headers, IP‐in‐IP tunnelingestablishedbetweenM2M‐RandeNB,andbetweeneNBandcorenetworkaddsthreeprotocolheaders (i.e., network, transport, and tunneling). This means a significant portion of thebandwidthiswastedontransmittingoverheads,thusreducingspectralefficiency.ItisshownbyLoetal. (2013)that intheworstcase, thetotaloverheadconsumesmorebandwidththantheactualM2M‐Aprotocoldataunit.AlthougheNB incurred thehighest overheads, theoverheadproblem is more pronounced on the M2M‐R to eNB transmission than on the eNB to corenetwork link due to limited frequency spectrum. Moreover, M2M‐Ds share the frequencyspectrumwiththemobileterminals.However,theoverheadproblemcannotbeneglectedwhentheeNBservesseveralM2M‐Rs,andeachofwhichservesalargenumberofM2M‐Ds.

Rationalsolutionsforsolvingconsideredproblemaremainlyrelatedtothetrafficaggregation.In research provided by Lo at al. (2013) a tunnel‐based data unit aggregation scheme isproposed.M2M‐Adata‐unitaggregationentitycanbeintroducedatM2M‐R,eNB,corenetwork,aswellasatM2M‐G.Dataunitsareclassifiedintotwodistinctclasses,highandlowpriority.Anoutgoing first‐in first‐out (FIFO) queue is defined for each of them. The high priority queuebuffers delay‐sensitive data units (e.g. accident events), while delay‐tolerant data units

(ordinary reports) are placed in the low priority queue. Obtained results show a significantreductioninprotocoloverheadsusingaggregationschemesinceLTE‐AisoptimallydesignedforH2H communication. Although aggregation causes data‐unit delay, it rapidly decreases as thenumber of M2M devices increases. On the other hand, Ahmad et al. (2014) proposed amethodology for facilitation of logistic processes by exploiting the M2M‐R functionality foraggregationandmultiplexingofM2Mdata traffic.By integrating trafficdemands fromseveralM2M devices in a single physical resource block, system performance can be considerablyimproved in terms of lower delay and higher throughput, in comparison to the case ofunaggregatedM2Mtraffic. In thisapproach, trafficclassification isnotconsidered,so it seemsthat increased delay can be expected for stochastic (i.e., event‐driven) demands such asemergencyandaccidentevents.

5.CONCLUSIONANDFUTUREWORK

ItisobviousthatconvergenceofM2Mcommunicationsandlogisticsservicesisoneofthepillarsof the IoT environment. The expected benefits are significant for both, telecommunicationoperators and logistics companies. In this paper some specific properties of M2McommunicationswhichareofimportanceforapplicationsofM2Msystemsinlogisticsservicesareemphasized.Having inmind these characteristics and fact thatmobile systemsofferwidecoverage,mobility support, satisfactory security, and ready‐to‐use infrastructure, hierarchicalcellular‐centric architecture is analyzed as a prospective solution for M2M‐driven logisticsservices.Moreover,challenging issueregardingoverheadproblem isnoted,andsomerationalsolutions based on traffic aggregation are presented. As main direction of future researches,M2Mtrafficcharacteristicsandcomprehensivesimulationanalysisoftheirinfluencesonmobilenetworkperformancesareenvisaged.

REFERENCES

[1] 3GPP (2012). System Improvements forMachineTypeCommunications, Release11, TR23.888.

[2] Ahmad, F., et al. (2014). Machine‐to‐Machine Sensor Data Multiplexing using LTE‐AdvancedRelayNodeforLogistics.Proc.LDIC,Bremen,Germany.

[3] Boccardi, F., et al. (2014). Five disruptive technology directions for 5G. IEEECommunicationsMagazine,52(2),74‐80.

[4] Cisco(2014).VisualNetworkingIndex,www.cisco.com.

[5] Ghavimi, F., Chen, H‐H. (2015). M2M Communications in 3GPP LTE/LTE‐A Networks:Architectures,ServiceRequirements,ChallengesandApplications. IEEECommunicationsSurveys&Tutorials,acceptedforpublication.

[6] IEEE (2012), Air Interface for Broadband Wireless Access Systems ‐ Amendment 1:EnhancementstoSupportMachine‐to‐MachineApplications,IEEE802.16p.

[7] Kim,J.,etal.(2014).M2MServicesPlatforms:Survey,Issues,andEnablingTechnologies.IEEECommunicationsSurveys&Tutorials,16(1),61‐76.

[8] Lo,A.,etal.(2013).ACellular‐CentricServiceArchitectureforMachine‐to‐Machine(M2M)Communications.IEEEWirelessCommunications,20(5),143‐151.

[9] Palafox‐Albarran, J., et al. (2012). CombiningMachine‐to‐Machine CommunicationswithIntelligentObjectsinLogistics,Proc.ImViReLL,Bremen,Germany,102‐112.

[10] Vodafone(2014).TheM2MAdoptionBarometer,m2m.vodafone.com.

RADIOFREQUENCYIDENTIFICATION:TECHNOLOGIESANDAPPLICATIONSINLOGISTICS

SvetlanaNikoličića*,JasminaPašagićŠkrinjarb,DejanMirčetića,PredragAtanaskovića

aUniversityofNoviSad,FacultyofTechnicalSciencesbUniversityofZagreb,FacultyofTransportandTrafficSciences

Abstract: Monitoring and management of complex logistics processes could not be realizedwithoutappropriate informationandcommunicationssystems.Awell‐structuredand technicallysafe information flow of valid data has come to form an essential element for designing andextending materials‐flow systems. Radio frequency identification technology (RFID) plays animportant role in supporting logisticsand supplychainprocessesdue to theirability to identify,trace and track information throughout the supply chain. The paper focuses on the technicalcharacteristics of RFID systems, their opportunities, challenges, barriers and standards.Furthermore, the paper discusses the results of RFID application in the realization of logisticsprocesses.Themaingoalof thepaper is tostimulate further interest in thisareabyprovidingacomprehensivereviewofRFIDapplication,whichwillbeagoodresourceforresearchesinterestedintheRFIDsystems.

Keywords:logistics,RFID,supplychainmanagement

1.INTRODUCTION

The realization and management of logistics processes includes a number of technologies,primarily transport, reloading, storage and information technologies.All of themareessentialfor logistics, and their development in the previous period has caused reoccurring waves ofchangesintheorganizationandmanagementoflogisticsprocesses.Technicalpossibilitiesofthemeans in transport, storage and reloading in the area of the goods flow have been greatlyresearchedandemployed;hence,thedirectionsforfurtherrationalizationsinthisareaarebeinginvestigated in the new approaches for the realization of logistics processes and in newstrategiesoflogisticsmanagement.Theimportantfactorinthecreationandrealizationofnewlogisticsstrategiesare thecontemporary informationandcommunications technologies.Theirsignificanceisevengreaterconsideringthatthecompaniesinthelastdecadehavebeguntobemoreandmoreorientedtowardstheirkeycompetencies,causinginitsturnthetrendtowardstheoutsourcingof businessprocesses, and thus enlarging thenetwork for addingnewvaluesand increasingthe logisticscross‐sectionbetweenpartnersexchanginggoodsand information(Nikoličić,2011).

Manifold transport, reloading and storage processes during the logistics process realization

* cecan@uns.ac.rs

demand for a fast goods identification and stock updates. The unambiguous identification ofgoods and the related data exchange have presented the basis of the efficient processes inlogistics processmanagement and control (Guenthner, 2004). Identification systems are onlythe first step, i.e. the link between goods and computers, directly linked to the supervisingcomputersystemforprocessmanagement.Inthispaper,theemphasisisontheauto‐IDsystemsthatenablewirelessoption foracceptingand transferringdatausingradiowaves, i.e. systemsfor radio frequency identification (RFID).Thepaperpresents the technicalpropertiesofRFIDsystems, as well as the advantages, drawbacks and examples of the RFID application in thelogisticsprocess realization.Themaingoalof thepaper is tostimulate further interest in thisareabyprovidingacomprehensivesurveyoftheRFIDapplication.

2.RFIDTECHNOLOGY

RFIDtechnologyenablesautomaticandnon‐contactidentificationofobjects,peopleandanimalsusingradiosignalsbeingemittedatacertainfrequency(VDInachrichten,2004a).AtypicalRFIDsystem consists of tags and readers, application software, a computing hardware, and amiddleware. Tag (transponder) is the basic component of the RFID system, presenting in theessence themicroprocessor chipand containing theelectronicallymemorizeddata.Generally,whenatagisinthefieldofaRFIDreader,itisactivatedandittransmitsthedatastoredonitsmemory chip to the reader. The tag has an identity that can be broadcast to a reader that isoperatingonthesamefrequencyandunderthesametagprotocol.Thereaderthenconvertstheradiowaves returned from the tag intodigitaldata and forwards them toa computer systemthatcollectsandprocessesthem(Figure1).

Microprocessor/memory

Analyser

Data

Cycle

EnergyInterface

Antenna

RFID read/writeunit(analyser)

AnalyserPC/database

PackageTransponder

Figure1.FunctioningprincipleofanRFIDsystem(VDI4472,2006)

RFIDsystemshavemoredeterminantsfordifferentiationandforestablishingtheareaoftheirapplication.ThefollowingcanbestatedasthebasicdeterminantsofanRFIDsystem:workingfrequency (LF,HF,UHF,MF), tagpower source (active,passive, semi‐passive tags),data inputpossibilities(readonly,writeoncereadmany,read‐write),memorycapacity(from32Byteto1kByte), reading range (from 0.15m to 30m), etc. The frequency on which the RFID systemoperatesdesignatestheintensityoftheradiowavesusedtotransmit informationandisakeyfactorindeterminingtheperformancelevelsandapplicationsforthesystem(Tajima2007).AllthisdemonstratesthewiderangeoftheRFIDandthepossibilitiesfortheirimplementationintodiverse areas (animal detection, logistics and supply chainmanagement, wastemanagement,museums, retailing, etc.). The general RFID system characteristicsmay be the following: highreadingvelocityandrelativelylargedistancebetweenthereaderandtheobjecttobeidentified;possiblereadingoutsidethevisualrange;simultaneousmultipledatareading;automateddataregistration; identification and monitoring of individual products, containers and means oftransportation,eveninextremeconditions;possibilitytomemorizegreatamountsofdataandthecommunicationdirectlyontheproductlevel;decentralizeddatastorageintheproductitself;abilityofwritinginformationonthetag(dependingonthetype),etc.(VDInachrichten,2004a).

InadditiontotheaboveadvantagesofRFIDsystem,therearealsosomebarriers:theprice,theshareofmistakes, radio interference, removalproblems, underdeveloped conscience of users,security and privacy issues. The following are four frequently cited barriers for widespreadadoption of RFID: a lack of return on investment (ROI), technical risks, the popularity of barcodes,andprivacyconcerns(Tajima,2007).

3.RFIDSYSTEMSINLOGISTICS

Although commercial applications of RFID date back to the 1960s, the use of RFID in supplychain management is relatively new (Tajima, 2007). At the beginning of this century, asignificant stimulus for developing and introducing RFID in supply chain management camefrom theworld’s leading retailers suchasMetroGroup,Wal‐Mart,Tesco,Carrefour, etc. Since2002,thereoccurringpapertopicinEuropeanjournalshasbeenrelatedtothediverseaspectsoftheRFIDapplicationinlogistics(Behrenbeck,2004).Thissectioncomprisesfromtwoparts:thefirstpartprovidesashortoverviewofacademicpublicationswherediverseaspectsof theRFID system application in logistics were researched, while the second part presents thedescriptionoftheexamplesoftheRFIDsystemapplicationsinpractice.

3.1RFIDresearches

Theresearch in theapplicationofRFID in logisticsandsupplychainmanagementarediverseand oriented towards different segments: technical problems related to the implementation,methodologicalprocedureofimplementation,aswellastheevaluationanddeterminationofthetechnology usefulness for different participants in the supply chains. The overview and thesystematizationofacademicliteratureareprovidedbyseveralauthors(Tajima,2007;Saracet.al., 2010; Zhu et al., 2012). Very interesting and useful is the general classification of papersaccording to themost used approaches and themain topics of the publications on the RFIDapplicationsinasupplychain(Saracet.al.,2010),presentedinTable1.

Table1.Typesofpublications

Publications Mostusedapproaches MaintopicsPracticalpapers Pilotprojects

CasestudiesROIanalyzes

InventorymanagementLogisticsandtransportationAssemblyandmanufacturingAssettrackingandobjectlocationEnvironmentsensors

Academicpapers AnalyticalapproachSimulationapproachCasestudiesROIanalyzesLiteraturereview

InventoryinaccuracyBullwhipeffectReplenishmentpolicies

BottaniandRizzi(2008)quantitativelyevaluatedtheeffectsoftheRFIDtechnologyandtheEPCsystem(ElectronicProductCode)onthemajorprocessesoffastmovingconsumergoodssupplychains.Afeasibilitystudywasconductedonthebasisofqualitativeandquantitativedatarelatedtothelogisticsprocessesofallparticipants.Asaresult,itwasconcludedthattheuseoftheRFIDtechnology on the level of pallets produced positive effects from the income aspect for allparticipantsinthechain;ontheotherhand,attheproductlevel,negativeeconomicresultswererecorded.

Ustundag and Tanyas (2009) used a simulation model for the calculation of the expectedbenefitsoftheRFIDsystemsintegrationintoathree‐stagechain.Itwasconcludedthatthevalueof products and demand had a significant impact on the expected benefits of the RFID

technology. The increase of the product value increased the total chain cost savings, and theincreaseddemanduncertaintyreducedsavingsinthechain.

Inordertoimprovetheinventorymanagement,anexperimentwasperformedfordeterminingtheimprovementintheinventoryrecordaccuracybeforeandafterimplementingRFID‐enabledadjustments to the inventorymanagementsystem(Hardgraveetal.,2011).Asa result, itwasconcluded that the effectiveness of the RFID tagging was not homogenous for all products.Reductionsinthepercentageofstock‐outsrangedfrom21%to36%,dependingonthecategory.

TheSupplyChainRFIDInvestmentEvaluationModelthatisbasedontheclassiceconomicorderquantity (EOQ) model considers the possibilities of the RFID investments to improve orderefficiency,JITefficiencyandoperationalefficiency(LeeandLee,2010).Fortheaforementionedefficiency factors, analytical procedures were developed to determine the optimal level ofinvestmentintheRFID.

Forthepracticalusage,theVDIguideline(VDI4472,2006)canbeveryuseful,containing,apartfromthedetailedcostslist,alistofindividualprocesseffectsinthesupplychain.TheessentialprerequisitefortheimplementationoftheRFIDtechnologyarethepositiveeconomiceffects;inthatsense,therecommendationistohaveaconsistentandclearsystemforevaluation.

Kokatal.(2008)conductbreak‐evenanalysisoftheRFIDtechnologyimplementationfocusingonthecostsresultingfrominaccurateinventorydata.Thebreak‐evenanalysisalsoincludesthemainfactorslike:technologycosts,lengthofreviewperiod,etc.Theresultsarepresentedwithexact analytical expressions for the cost‐effectivepriceofRFID, and show that there is ahighcorrelationbetweenthecostofRFIDtechnologyandthevalueofproductslost.

Between the participants in the supply chains there are disagreements on the necessaryinvestments and achieved benefits from the RFID technologies (Karkkainen and Holstrom,2002).Mostsavingsareachievedintheretail,whilethesuppliersareburdenedwiththecostsofintroducingRFID(Smith,2005).

3.2RFIDinpractice

Thepotentialsof theRFIDapplications in logisticsprocessesaremanifold;however, therearestillmanybarriersforthewiderapplicationinpractice(e.g.necessaryinvestments,popularityof barcodes, and question of privacy). This section of the paper describes the experiences ofcompaniesthatimplementedtheRFIDforlogisticsprocessesmonitoringandmanagement.

In 2003,Metro AG began the project namedMetro Extra Future Store that implemented andtested theRFID technologies in real surroundings (Behrenbecketal.,2004). In2004, the firstphaseoftheimplementationofthissystemwascompleted,andthetestdemonstratedthatthecostsofout‐of‐stock,rangingfrom9%to14%,couldbereducedfor17%;theftswerereducedfrom11%to18%,andtheworkingcostsfrom8%to11%.In2006,another300supplierswereincluded.In2008,theRFIDprojectwasexpandedto200Metrocentres.Thatprocessincluded1.3millionpalettesannually(LOG.m@ilNewsletter,2008).

InWal‐Martprojects,asoneofthepotentialoptionsfortheimprovement,theyinvestigatedtheinfluence of RFID on the out‐of‐stock reduction (OOS situation) that provided a number ofbenefitsforretail,suppliersandconsumers.Asaresult,itwasconcludedthattheapplicationofRFID reduced the out‐of‐stock for 26% (Hardgrave et al. 2005). Furthermore, the surplus ofstockinthesupplychainwasminimized(LOG.m@ilNewsletter,2007).

Best Buy tested the RFID usage in the storage level in two distribution centres. During thetesting, 70 suppliers used RFID for palettes and packaging, and the data were processedautomatically in the software system for business management. Very positive results wereobtained,i.e.stockinstoragewasreducedfor20%,whiletheparticipationofthecross‐dockingprocesswasincreased.

Apartfromfollowingthegoodsflowinordertorationalizeit,RFIDcanbeappliedforsecurityreasons as well. Hutchinson corporation, with the annual turnover of 44 million containers,introduced sensors inside and outside for every container, which could detect, among other,light,humidity,airpressure,certainchemicalparameters,etc.,beingreadyinanygivenmomenttosendane‐mailalarmtothecentralunit.Insomeports(e.g.LosAngelesandLongBeach),theactiveRFID system isused for security reasonsand for solvingproblemsof the congestion inports.AlltrucksenteringtheportsneedtohaveRFIDtagsbeingreadinentranceandexit(VDInachrichten,2004b).

SSAMarine,oneoftheleadingworldwidecargocompanies,introducedtheRFIDtechnologyforcontainermonitoring in severalportson theWesterncoast.On the containerhoist, there is areader,while the active tag is on the container, enabling the identification of the appropriatecontainerandloadingontotheappropriatetruck(Voyles,2005).

ThecompanyIvecoincooperationwithKuehne+NagelusesRFIDtomanagesparepartsintheautomotive sector.The system in theplant inTorino simplifiesorders forparts replacementsfrom suppliers. Kuehne + Nagel receive the parts and add EPC Gen 2 marks coded with theidentificationnumber andprinted togetherwith the barcode, hencedealers canuse them fortracingandmonitoring(Wessel,2010).

DHLintegratedRFIDintotheirtrucksfordeliveryinordertoachievefasterparceldelivery.Thisinitiative was named SmartTruck and it achieved a number of goals. They accomplished theexpected savings in fuel consumption andCO2 emission, and increased the accuracy in parcelcollectionanddelivery(Neely,2009).

4.CONCLUSION

RFID systems show a great potential for the improvement of processes and the reduction ofcostsassociatedwiththesupply‐chainmanagement.However,someofthebasicproblemsoftheRFIDsystemapplicationincludetheintroductionanddevelopmentcosts.Opinionsofexpertsonthe implementationof theRFIDsystems insupplychainsarestilldivided, though, theopinionthatthetimeoftheRFIDtechnologyisstilltocomeseemstodominate.

The paper demonstrated the basic principles in utilizing the RFID system, together with thepotentialapplication in logistics.AshortsurveyofpublishedpapershasdemonstratedagreatinterestofscientistsforresearchingdiverseeffectsofRFIDinlogistics,whiletheexamplesfromthepractice emphasise that theRFIDsystemapplication ispossible indiverse companiesandthatitisdirectedtowardstheincreaseinthelogisticsprocessefficiency.ItisonlyhopefulthatthispaperwillrepresentavaluablebasisfortheresearchersinterestedinresearchingtheRFIDsystems.

ACKNOWLEDGMENT

ThestudywasfinanciallysupportedbytheSerbianMinistryofEducationandScience(ProjectNo.TR36030).

REFERENCES

[1] Behrenbeck,K.,Küpper, J.,Magnus,K.H.,Thonemann,V., (2004).ZwischenEuphorieundSkepsis.LogistikHeute,12/2004.

[2] Bottani,E.,Rizzi,A.,(2008).EconomicalassessmentoftheimpactofRFIDtechnologyandEPCsystemonthefast‐movingconsumergoodssupplychain.Int.J.ProductionEconomics112,548–569.

[3] deKok,A.G.,vanDonselaar,K.H.,vanWoensel,T.,(2008).Abreak‐evanAnalysisofRFIDtechnology for inventory sensitive to shrinkage. Int. J. ProductionEconomics, 112, 521–531

[4] Guenthner,W., (2004). RFID‐Technologie als wegbereiter neuer Logistikstrukturen. TIL2004,Niš.

[5] Hardgrave, B.C., Goyal, S., Aloysiu, J.A., (2011). Improving inventorymanagement in theretailstore:TheeffectivenessofRFIDtaggingacrossproductcategories.OperManagRes4,6–13.

[6] Hardgrave, B.C., Waller, M., Miller, R., Walton, S. M., (2005). Does RFID reduce out ofstocks?WorkingPaper.UniversityofArkansas,Fayetteville,Arkansas.

[7] Karkkainen,M.,Holstrom,J.,(2002).Wirelessproductidentification:enablerforhandlingefficiency, customisation and information sharing. Supply Chain Management: AnInternationalJournal,7(4),242‐252.

[8] Lee,I.,Lee,B.‐Ch.,(2010).AninvestmentevaluationofsupplychainRFIDtechnologies:Anormativemodelingapproach,Int.J.ProductionEconomics,125,313–323.

[9] LOG.m@ilNewsletter(2008).Metro,9/08.

[10] LOG.m@ilNewsletter(2007).Wall‐MartwillEinsatzvonRFIDausweiten,19/07.

[11] Neely,B.,(2009).DHLSaysItsSmartTrucksSaveMoney,TimeandCO2.RFIDJournal,Sep30.

[12] Nikoličić,S., (2011).Acontribution to the improvement inexistingmethodsandmodelswith a view to establishing the performance of the retail logistics system. (in Serbian).Dissertation.UniversityofNoviSad,Serbia.

[13] Sarac, A., Absi, N., Dauzere‐Peres, S., (2010). A literature review on the impact of RFIDtechnologiesonsupplychainmanagement.Int.J.ProductionEconomics,128,77–95.

[14] Smith,A.D.,(2005).Exploringradiofrequencyidentificationtechnologyanditsimpactonbusinesssystems.InformationManagement&ComputerSecurity,13(1),16‐28.

[15] Tajima, M. (2007). Strategic value of RFID in supply chain management. Journal ofPurchasing&SupplyManagement,13(4),261–273.

[16] Ustundag, A., Tanyas,M., (2009). The impacts of Radio Frequency Identification (RFID)technologyonsupplychaincosts.TransportationResearchPartE45,29–38.

[17] VDInachrichten,49(2004a).RFIDTechnikfehlenStandards,Berlin,BeuthVerlag

[18] VDI nachrichten 51/53, (2004b). Contanierschutz verteuert welweite Logistikketten,Berlin,BeuthVerlag.

[19] VDI 4472 (2006). Part 1, Part 2 Handbuch – Requirements to be met by transpondersystemsforuseinthesupplychainGeneral,Berlin,BeuthVerlag.

[20] Voyles, B., (2005). SSA Marine Adds RFID to Quicken. RFID Journal, (available at:http://www.rfidjournal.com/articles/view?1488).

[21] Wessel,R., (2010). Iveco toExpandRFIDSystem forManagingReplacementParts.RFIDJournal,(availableat:http://www.rfidjournal.com/articles/view?7352).

[22] Zhu, X., Mukhopadhyay, S.K., Kurata, H., (2012). A review of RFID technology and itsmanagerialapplicationsindifferentindustries.J.Eng.Technol.Manage,29,152–167.

PartVI SIMULATIONOFLOGISTICS

PROCESSES

BULLWHIPEFFECTANALYSISBYSIMULATIONEXPERIMENTSINECHELONUNDER(R,s,S)INVENTORYPOLICY

SamirŽica,TončiMikaca,IrenaKosb,JasminaŽicb

aUniversityofRijeka,FacultyofEngineering,CroatiabVoraxd.o.o.,Rijeka,Croatia

Abstract:Inordertoimprovetheirperformanceonmarket,moderncompaniesandsupplychainsarefocusedtoreduceallexcesscosts,especiallythoserelatedtoinventoriesandlogistics.However,reducingsuchcostsmustbetaken intoconsiderationseriously,sincetherearemanydifficult‐to ‐predictinterdependenciesbetweenvariablesofperiodic(R,s,S)reviewinventorysystems.Inorderto set up the inventory system in most optimal manner or predict its behavior, simulationexperimentsshouldbeexecuted.

Keywords:Bullwhipeffect,Supplychain,Simulationexperiment,Inventorymanagement

1.INTRODUCTIONANDPREVIOUSRESEARCHES

Modern companies and supply chains compete globally in a large number of independentvariableswithnon‐trivialdependencies.Togetherwithconstantfluctuationsininventorylevels,structural and logistical characteristics of a supply chain also change. Therefore, influentialvariablesof supply chain inventorymanagement systemsneed tobemonitored continuously;forecastedandoptimalbusinessscenariosinresponsetofutureeventsshouldbecalculated.Inorder to optimize the properties of the supply chain inventory systems it is necessary tothoroughlyinvestigatetheinterdependenceofvariablesininventorymanagementsystem.

ResearchesfromHoulihan[1987],Taylor[1999],andFransooandWouters[2000]describethebullwhipeffectinMake‐to‐stocksupplychains.WorksfromBaganhaandCohen[1998],Graves[1996], and Cachon [1999] are dealing with the increased deviation of demand informationwhentransferredawayfromthemarket,andChenetal.[2000]andMetters[1997]weretryingto analytically quantify the consequences of the bullwhip effect. Forrester in hiswork [1958]describestendenciesofbusinesssystemstoamplify,causedelayanddistortinformationaboutactualmarketdemandas this informationmoves furtheraway from it.Thisphenomenonwasfirstrecognizedbythemanufacturerswithowninventoriesofrawmaterials,semi‐finishedandfinished products intended for sale. Lee et al. [1997a], [1997b] call this phenomenon theBullwhip effect since small changes or variances in customer’s demand often cause a largeincrease inordersreceived incompanieswithin thesupplychain towards thedirectionof themanufacturer.

Asmallvariance inactualconsumerdemandcanresult in thevariouscompaniesoperatingatdifferentstagesofasupplychainbeingsubjecttothebullwhipeffect(Fig.1).Forrester’sworkconcluded that thebullwhipeffectoccursdue tounstablebusiness conditionsandproposesamethodology of simulation of dynamic models of business systems. The study by Lee et al.

showedthatthebullwhipeffectiscausedbythefollowingeffects:(1)Forrestereffect,orlead‐timesanddemandsignalprocessing,(2)Burbidgeeffect,ororderbatching,(3)Houlihaneffect,orrationingandgaming,and(4)promotioneffect,orpricefluctuations.

Sterman’swork[1989a]determinedanadditionalcauseofthebullwhipeffectwhichoccursduetothelimitedhumanpossibilitiesofperceptionandmanagementofcomplexdynamicsystems.The above, aswell as numerous other studies have shown the negative impact of behavioralcausesonthesupplychain.

Increaseof customerdemandgives rise to anumberofproblems: theaccumulationof excessinventoriesatcertaintimesfollowedbyseriousinventoryshortagesandlowcustomerfillrate,excess or insufficient capacity, unstable and inefficient production leading to higher costsresultingfromthecorrectiveactionsthathavetobetaken.

Figure7.Oscillationdemandinthesupplychainsinthedirectionofthemanufacturer

This type of deviation from the actualmarket demand causes significant costs and ultimatelycalls intoquestiontheexistenceofthesupplychainforcontinuouslyincreasedcosts,overloadproduction, increased logistics requirements, insufficient or excessive inventories and thusinsufficientstorageandlogisticscapacities,resultingwitha lowpercentageofmeetingmarketdemands.Duetothesignificantandnegativeimpactofthebullwhipeffectinsupply,andthusinmanyotherbusinessindicators,thebullwhipeffectbecomestheresearchsubjectofnumerousscientistsandbusinessexperts.

Thispaperisorganizedasfollows.Section2presentsthesetupofsimulationexperimentsandexperimental work with a 300 simulation experiments of periodic (R, s, S) review inventorysystem.Insection3,theconclusionsareprovided.

2.EXPERIMENTALWORK

Demandwithinthesupplychainisoftenrandomandunpredictable.Whereitisnotpossibletoaccuratelypredictmarketdemand,simulationexperimentsareusedtoexaminetheoccurrenceandsizeofthebullwhipeffect.ThepositiveresultsoftheapplicationofsimulationexperimentswereconfirmedbyworksofOuyangandDaganzo[2008]andBaccadoroetal.[2008].

Chen et al. [2000] suggested that the bullwhip effect could bemeasured by the ratio of / between the input and output flows in each echelon in a supply chain; or between the final

demandandthemanufacturerwhenthewholesupplychainistobeevaluatedascanbeseeninequation(1).

(1)

where:

Orepresentsorders;Drepresentsdemand ‐varianceofdemandobservedinechelon μ–meandemandobservedinechelon

2.1.Setupofsimulationexperiments

Experiments will simulate performance of periodic (R, s, S) review inventory system of oneechelonintraditionalsupplychain.Monitoringperiod is5000days(several times longerthansimilarresearcheshaveworkedwith),meandemandμ=99,998,varianceofdemandobservedinechelon =1857,532anddistributionofdemandisgeneratedasnormal,independentandidenticallydistributed. Simulationexperimentswere created andanalyzedbyOptimInventorysoftwareincaseswherefillrateof99,9%±0,01%isachieved.Thistighttolerancesareneededinorder to reduce the number of results and to get better understanding of (R, s, S) inventorypolicy.Echelondoesnotbacklogunfulfilleddemand.

Variablesonthesuppliersideofechelonareas follows: lead time is3days,observedechelonanditssupplierworkeverydayinweekandminimalorderquantityis100products.

OptimInventory is a computer application that simulates real world inventory conditions ineither individual echelons orwhole supply chain. It offers in depth and intelligent analysis ofmultipletypesofinventorysystemsallowingsignificantreductionsincostsandsimultaneouslyincreasingcustomersatisfactionandfillrateonthemarket.

2.2.Experimentalwork

Bestsimulationexperiment(SE)isconsideredtobetheonethathas lowest levelof inventorybut manages to satisfy desired fill rate of 99,9%±0,01%. This simulation experiment will beassigned with ID value 1, after which the software will increase level of inventories andrecalculatefillrate.

Itisimportanttoknowthatwithinventorylevelincrease(especiallyorderuptolevel–S);manydifferentfactorsaffecttheinventorysystem.Onlythoseexperimentsthatsatisfyfillratewillbeanalyzed and the softwarewill calculate following variables in each of them: lower inventorylevel (s), upper inventory level (S), exact fill rate, minimal and maximal stock level, averageinventorylevel,numberofshipments,totalnumberofproductsshipped,bullwhipfactoretc.

Forthispaper,300simulationexperimentswereexaminedandresultsareshownonfigures2to6.

Figure8.ResultsofSEssortedbyascendingsimulationexperiment’sID

Figure3.ResultsofSEssortedbyascendingvaluesofaveragedailyinventorylevel(averageI.L.)

Figure4.ResultsofSEssortedbydescendingvalueofrequirednumberofshipments(No.S.)

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

300

400

500

600

700

800

900

1000

1100

1200

1300SEID8 16 24 32 40 48 56 64 72 80 88 96 104

112

120

128

136

144

152

160

168

176

184

192

200

208

216

224

232

240

248

256

264

272

280

288

296

LowerI.L. UpperI.L. AverageI.L. No.S. Averageshippings. BWcoefficient

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

300

400

500

600

700

800

900

1000

1100

1200

1300

SEID 38 56 42 21 117

257

243 47 176 61 66 266

183

193

217 24 25 33 78 86 94 102

110

127

144

135

284

177

292

150

165

188

204

189

239

230

241

LowerI.L. UpperI.L. AverageI.L. No.S. Averageshippings. BWcoefficient

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

300

400

500

600

700

800

900

1000

1100

1200

1300

SEID 76 84 92 100

108

138

167

199

207

224

232

132 16 26 34 130

144

161

240

158

184 46 176

182

140 4 5 39 238

244

243

255

263

284

274

286

288

LowerI.L. UpperI.L. AverageI.L. No.S. Averageshippings. BWcoefficient

Figure5.ResultsofSEssortedbyascendingvaluesofminimalrequiredaverageshippingsize

Figure6.ResultsofSEssortedbyascendingvalueofBullwhipcoefficient(secondaryYaxis)

3.CONCLUSION

From these 300 simulation experiments it is noticeable that important variables of inventorymanagement systemshowsignificantoscillations resulting inmoreor lessnegativeeffectsonechelon.ThisconcludeswithpreviousworkonSEsusedin(R,s,S)inventorysystems.Itisalsoimportant to notice that these simulations analyzed only one product in echelon and if weassumethatbiggerretailerstorescanholdupto80.000productsit isclearthatfinetuningofinventorysystem is impossiblewithoutadvancedsimulationsoftwaresuchas theoneused inthispaper.

Another important thingnoticeable from figures 2‐6 is the fact thatwith reducingnumber ofrequiredshipments,bullwhipeffectsignificantlyreducesinanabsolutevalueanditsvarianceissignificantlyloweredtoo.Thereforethisresearchshowsthatbysettingcharacteristicinventorylevels higher than needed and respectively reducing number of deliveries to specific echelon,onecanactuallyreducethebullwhipeffect.Ifthissetupismonitoredregularly,negativeeffectssuchasincreaseinaverageinventorylevelcanbeavoided.

Thislevelofanalysisiscomputationallyveryintensivebutoffersbenefitsthat,inthelongrun,offersignificantsavingsandmakeechelonsandsupplychainsmoreeffectiveandcompetitiveon

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

300

400

500

600

700

800

900

1000

1100

1200

1300SEID 80 88 96 104

112

138

167

199

207

224

232

179 16 26 34 130

144

160

187

159

184 46 176

182

140 4 5 39 238

244

243

255

263

284

274

286

288

LowerI.L. UpperI.L. AverageI.L. No.S. Averageshippings. BWcoefficient

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

300

400

500

600

700

800

900

1000

1100

1200

1300

SEID 289

238

213

253

254

259

250

115

258 1

174 59 192

183 8

118

158

186

161 68 52 129 36 34 148 53 221

165

228

236

204

188 78 86 94 102

110

LowerI.L. UpperI.L. AverageI.L. No.S. Averageshippings. BWcoefficient

globalmarket.Inthefutureweplantoexplorethebehaviorofperiodic(R,s,S)reviewinventorysystemwithdifferentdeliveryterms,batchingsizeandbacklogperformance.

REFERENCES

[1] Baganha M., Cohen M., The stabilizing effect of inventory in supply chains, OperationsResearch,46(3),72–83,1998.

[2] Boccadoro M., Martinelli F., Valigi P., H‐infinity control of a Supply Chain model,Proceedingsofthe45thIEEEConferenceonDecisionandControl,4387–4392,2008

[3] Cachon G., Managing supply chain demand variabilitywith scheduled ordering policies,ManagementScience,45(6),843–856,1999.

[4] Chen F., Drezner Z., Ryan J., Simchi‐Levi D., Quantifying the bullwhip effect in a simplesupplychain:Theimpactofforecasting,leadtimes,andinformation,ManageScience,46,436‐443,2000.

[5] Forrester J., Industrial dynamics, a major breakthrough for decision makers, HarvardBusinessReview,36,37–66,1958.

[6] Fransoo J.,WoutersM.,Measuring thebullwhip effect in the supply chain, SupplyChainManagement:AnInternationalJournal,5(2),78–89,2000

[7] Graves S., A multiechelon inventory model with fixed replenishment intervals,ManagementScience,42(1),1–18,1996.

[8] Houlihan J., International supply chain management, International Journal of PhysicalDistributionandMaterialsManagement,17(2),51–66,1987.

[9] LeeH.,PadhamanabhanV.,WhangS.,Informationdistortioninsupplychain:thebullwhipeffect,ManagementScience,43(4),546–558,1997a.

[10] Lee H., Padhamanabhan V., Whang S., The bullwhip effect in supply chains, SloanManagementReview,38,93–102,1997b.

[11] Metters R., Quantifying the bullwhip effect in supply chains, Journal of OperationsManagement,15,89–100,1997.

[12] OuyangY.,DaganzoC.,Robusttestsforthebullwhipeffectinsupplychainswithstochasticdynamics,EuropeanJournalofOperationalResearch185(1),340–353,2008.

[13] Sterman J., Modeling managerial behavior: misperceptions of feedback in a dynamicdecisionmakingexperiment,ManagementScience,35,321–339,1989a

[14] TaylorD.,Measurementandanalysisofdemandamplificationacrossthesupplychain,TheInternationalJournalofLogisticsManagement,10(2),55–70,1999.

SIMULATIONASATOOLFOREXTRACTINGKEYPERFORMANCEPARAMETERSINPOSTALLOGISTICCENTERS

DejanMircetica,SvetlanaNikolicica,MarinkoMaslarica,MilosStanojevica

aUniversityofNoviSad,FacultyofTechnicalSciences

Abstract:Today,postaloperatorsarebecomingmoreandmoremarket‐orientedandcustomer‐driven.Asaconsequenceofthat,manyofthetraditionalpostalcentersexpandedtheirservice inarea of logistics and transform itself in postal logistics centers (PLC‐s). Prior orienting to theproviding logistic services management of PLC needs to determine current key performanceparameterswhile performing basic postal activities, driven from defined business objectives ofpostalservice.InordertodeterminePLCcapabilitiesforadditionallogisticservices,firststepistoknowhowmuch resourcescanbe redirected to the thosenew logistic services.Thispaperdealsexactlywiththiskindofproblem,andweproposethesimulationofpostalprocessesinPLC,usingdiscreteevent simulation (DES),asamethodologyapproach forextractingandrepresenting theinfluence of differentworker layouts on overall processing time inPLC.Accordingly, simulationmodelofcurrentoperationconditionsinPLCwascreated(AS‐ISmodel).

Keywords:postal logisticscenters,discreteeventsimulations,mails,postalprocesses, logisticsperformances.

1.INTRUDUCTION

Postalserviceplaysasignificantroleinthedevelopmentofaleadingcommercialandfinancialregions, providing communications between individuals, business and government. In recentyears, postal operators are facing with the challenges of rapid technological development,marketliberalization,segmentationandincreasingcompetition.Insuchconditions,theindustryhasevolvedtoincludetraditionalpost(likepackageandmaildelivery),courierservices,freightservicesande‐services(Chanetal.,2006)inlogisticsfrightflows,andleadingEuropeanpostaloperators(suchas:DeutschePostDHL,LaPosteandRoyalMail)haveexpandedtheiroperationsin the logistics sector. This development of the postal sector is conditioned by the rapiddevelopmentofe‐commerce,which,amongotherthings,includesastronglogistics.

PLCareimportantlinkinthedeliveryofshipmentsbetweenthesenderandtherecipient.Theirmain function is collecting and distribution of large quantities of shipments in the postalnetwork. Postal processes that are carriedout in these centers, arising fromdefinedbusinessobjectives of postal service and generally include: sorting of postal shipments, organizing thetransportandprovisionofpostalandlogisticsservices.HeterogeneityofpostalshipmentsinthePLC causes various postal activities in order to process different mails. Depending on theshipments type, depends which resources will execute the processing and in what manner.Differentmailshavedifferentleadtimes,andacquiredifferentresourcesforprocessing,which

implies different costs of processing. Bearing in mind the global determinants of postaloperatorsandthestructureandcharacteristicsofthebusinessprocessesinpostalservice,itisclearthatthelogisticsasaninstrumentofdifferentiationandrationalization,shouldbeintegralcomponentofthepostaloperatorsbusinessstrategy(Kujačićetal.,2013).Inordertodeterminethe levelof efficiencyof executedprocesses inPLC‐s, it isnecessary toestablishprocedure toextract themeasures of logistic performances of basic postal operations in the PLC. For thatpurposeweusedDESofbasicpostalprocessesinPLC,andwefocusedondeterminingoverallprocessingtimeofmails,takingintoaccountengagednumberofworkers.

2.STRUCTUREOFTHEPOSTALNETWORKANDBUSINESSPROCESSESINTHEPOSTALSERVICE

Thekeycharacteristicsofpostalservicesisreflectedinthemassiveuserdemandfortransferofpostalitems,whichisrealizedthroughpostalprocessesthatgeneratephysicalmovingofmailswith aim of their transfer to the recipient. From an organizational point of view, for theimplementationofpostalservicesthatdemandsthemarket(nationaloperatorsareconditionedtoalsotoprovideauniversalservice)itisnecessarytoestablishauniformpostalnetwork(onnational and international level), and use of unique technologies and standardization ofequipment. The structure of the postal network and its equipment is conditioned by theeconomic capabilities of operators, traffic volume, and in the case of a national operator,governmentinvestmentpolicies.

Fromthelogisticalpointofview,aPLCperformsthefunctionofintralogistics(LisecandRihter,2007),andasaunitsofpostalnetwork,PLCarelocatedinthetraffichubsinordertoachievetheconcentration and diffusion of shipments on the geographical area that they cover, andconsequently,theyhaveakeyroleintheconcentration,processing,deliveryandtransshipmentof receivedmails.Recently,SerbianPostsconstructed threenewregionalPLChubs(Belgrade,NoviSadandNish), andperformed automatizationofmailsprocessing. In thePLCsNoviSadandNishautomatizationiscariedoutonlyfortheletterswithstandardsizes(dimensionsfrom90x140 mm to 120x235 mm), while in PLC Belgrade automatization also included theautomatization of parcel deliveries. For more details about organization of postal‐logisticsprocesses,especiallyaboutorganizationofpostalserviceinSerbia,referto(Kujačićetal.,2013).

3.CASESTUDYOFPLC

The case study was carried out on the example of PLC Novi Sad (Republic of Serbia). In asimplifiedanalysisPLCNoviSadcanbeconsideredasposthubwithbasicfunctionofreceiving,sorting, delivery and transshipment of incoming mails. In the observed PLC different mailclassesareprocessed(prioritymailexpress, firstclassmails,standardmailsand internationalmails). Prioritymail express represent servicewhich allows fastest transfer all kind ofmails(letters, parcels, advertisements, bills, etc.). This service also provides tracking of shipmentsduringtransfer,andserviceisavailableonlyindomestictraffic(i.e.withintheboundariesoftheRepublicofSerbia).Firstclassmailsincludesvaluableparcelsandvaluableletters.Bybusinesspractice of Post of Serbia, for valuable parcels medium and large packages are considered.Valuablelettersarespecialkindofpostalshipments,whichareprocessedinseparatesectionofPLC, since they are confidential contracts between companies, they can be internationalshipments and government shipments. By standard mails Post of Serbia consider:advertisements, circulars, newsletters, small parcels, merchandise, bills, etc. Standard postalitems are classified on the basis of their dimensions, and in practice of Serbian Postsclassification is performed on the ordinary and nonordinary mails. Ordinary mails includeletterswith standarddimensionswhichare consisted from recommended letters andprioritymails.Recommendedlettersaresimilarshipmentsaspostmailexpress.Theyalsoallowtrackingoflettersandbuttheyareslowerthanthepostmailexpressshipments.Prioritymailsaremails

which have priority in delivering to the final consumer, and usually there are consisted frombillsforelectricenergy,TV,andotherkindofgovernmenttaxes.Accordingly,duringprocessingin thePLCdifferent shipments (i.e.mail classes)havedifferent sequenceof activities throughPLC,andbyaverage7000000mailsareprocessedinPLCNoviSad,foronemonth.TypicalflowofmailsthroughPLCNoviSadisshowninFigure1.

Figure1.FlowofmailsthroughPLC(adaptedfrom(Kujačićetal.,2013))

3.1As‐Is

Simulation of current operation conditions in the PLC (AS‐ISmodel), is carried out using theseveraldatabasesandexperiencegainedfromvocationaltraininginthePLC.Followingthelogicof Figure 2, and using data about time duration for each activity from document “Postalstatistics”(Đakovački,2006)andalsodatarelatedtoactuallayoutofactivities(determinedfromvocational training inPLC),PLCNoviSad is simulated.Asa resultAS‐ISmodel is formed,andpresented in the Figure 2. For building the simulation scenarios iGrafx software was used.Simulation scenarios include simulation of PLC for a onemonth (March 2014). As a input insimulation, real shipment quantities of themail classes and different number ofworkers areused. Each simulation lasted for approximately half an hour and twelve different simulationwere performed. During each simulation 7,048,126 mail shipments were simulated. In AS‐ISmodel,everyactivityisassociatedwith:particularresourcewhichisexecutingthatactivity,andwith timeneeded forexecutionofobservedactivity.Forcreating theAS‐ISprocessmapbasicelements and reasoning of business process modelling is used, and while performingsimulationsAS‐ISprocessmapwasfixed(i.e.theinterconnectionsbetweentheelementsandthequantity of postal shipments that flow between them). Only parameter that was changing ineachsimulationwasnumberofworkersandtheirposition inAS‐ISprocessmap,accordingtotheTable1.FormoredetailsonbusinessprocessmodellingusingiGrafx,referto(Grozniketal.,2008).

Process map for AS‐IS model presents four sectors in which incoming mails are processedaccording to mails type (i.e. mail class). Each mail class has different specificities andrequirementsforprocessing.Greencolorobjectsrepresentthegeneratorsofmailclasses.Mailsare thendirectedtoparticularsectors,accordingto theirrequirements for furtherprocessing.GraycolorobjectsrepresentstheactivitieswheretimedurationoftheactivitiesaredeterminedbymeasuringatthePLCduringthevocationaltrainingcarriedoutinPLCNoviSad.Reasonformeasuring timeofexecutionofactivities isbecausePLCNoviSad isnewlyconstructedcenterand Posts of Serbia didn't yet perform measuring and update the official document "PostalStatistics"(Đakovački,2006),withtimedurationsfortheabovementionedactivities.Bluecolorobjectsrepresents theactivitiesrelatedwithmanualsortingof incomingmails,wereacceptedstandard is thatoneworkersorts the1000mailsperhour.Browncolorobjectsrepresent theactivitieswiththeirtimeofexecutiontakenfromthe"Postalstatistics"(Đakovački,2006).

Figure2.SimulationofcurrentoperationconditionsinthePLC(AS‐ISmodel)

3.2Overallmailsprocessingtimevs.engagedworkers

First step in analyzing the operation capabilities of observed PLCwas performing of severalsimulations with different number of workers in particular sectors in order to determinerelationship between engaged workers and processing time of center. Simulations wereperformed based on several different layouts of workers, and there were twelve differentscenarios.Resultsofsimulation,accompaniedbydifferentlayoutsofworkers,arepresentedinTable1.FromTable1it isobviousthatengagingthedifferentnumberofworkersleadstothedifferentprocessingtimeofshipments,bearinginmindthedefinedworkerslayout.

Figure3presentsmentionedrelationship,fromwhichwecanconcludethatprocessingtimeismonotonically decreasingwith engaging large number ofworkers. But, fromFigure 3we canconclude that there is a bottom limit, and that drop of processing timewith engaging largernumberofworkersstopsatsomepoint.Reasonforthatisbecauselaboroperationcapacityhasphysicallimitations,andengagingmoreandmoreworkersisn'taccompaniedbysignificantdropin processing time. Which leads to conclusion that engaging more workers isn't alwayseconomically justified,sincethedecreaseinprocessingtimeisnotbigenough,comparedwithcostsrelatedwithengagementof additionalworkers.

Table1.Simulationresultswithtwelvedifferentlayoutofworkers

Differentsimulationscenarios I II III IV V VI VII VIII IX X XI XII

Supportworkers 3 3 3 3 3 3 3 3 3 3 3 6

Workersinsectorfirstclassandprioritymails 8 8 8 8 8 9 9 9 9 12 9 12

Controller 1 1 1 1 1 1 1 1 1 1 1 1

WorkersinCustomssector 9 9 9 9 9 10 10 10 10 12 10 15

Workersinsectorfordelivery 10 10 10 10 11 12 12 13 15 15 18 21

Workersinstandardsector 18 20 22 25 26 27 28 30 32 30 36 40

Totalnumberofworkers 49 51 53 56 58 62 63 66 70 73 77 95

Averageprocessingtime(minutes) 3137 2042 881 232 172 148 70 64 50 57 31 20

Figure3.Therelationshipbetweentheprocessingtimeandthenumberofengagedworkers

The dependence of the processing time and the number of workers engaged is numericallyestimatedbyequation(1):

2,1529105,1)( xxf

(1)

Where x is number of engaged workers, and 95% confidence intervals for coefficients are (3030 106,1;103,1 ) and (‐17,61; ‐12,8). Equation (1) can be used by managers in order to

determine on how big drop in processing time can they expect when they engage differentnumberofworkers.This is important,especially indaysofpeakload,whicharedayswithbigincrease of incoming mails. PLC have problems in processing this huge amount of incomingmails in small period of time (usually peak load is at the beginning of themonth, due to theincreaseofstandardmailsshipments).InthosedaysPLCmanagersneedtorearrangeworkersactivities(transferofworkersfromothersectorstothesector forstandardmails)andengageadditionallaborforcetodealwithpeakload.

4.CONCLUSION

Efficiencyof theexecutedprocessesdirectlyaffectsonoverallefficiencyof theentire logisticssubject. This paper directly dealswith this kind of problem.Themain researchhypothesis inpresentedpaperwas to examine thepossibility of applicationofDESonbasicpostal‐logisticsprocesses in PLC. Papers aim was to create framework of PLC in which different simulationscenarioscouldbeperformed,inordertotestvariousdesignsolutionsonlogisticperformancesofparticularPLC.Basically,modelcanbeusedforanykindof“whatif”analysis,whichisfrominterestforPLCmanagers.Besideframework,inthispaperwealsofocusedontheproblemofadditionalworkersengagementandtheir impactonprocessingtime.Asaresultwepresentedtheequation(1),derivedfromsimulatingtheseveraldifferentworkerslayouts.Equation(1)canservetothemanagersasaguidingtoolfordecidinghowmuchworkersshouldbeincluded,inwhichsectors,inordertoobtaindesiredprocessingtime,andinthatwayfulfilldefinedlevelofservice.

ACKNOWLEDGMENT

ThisresearchwassupportedbyMinistryofeducation,ScienceandTechnologicalDevelopment,throughprojectTR36030.

REFERENCES

[1] Chan, F. T., Chan, H., Lau, h. C. & Ip, R.W. (2006). An AHP approach in benchmarkinglogisticsperformanceofthepostal industry.Benchmarking:AnInternationalJournal,13,636‐661.

[2] Đakovački, V. (2006). Pravilnik o statistici i normama, Javno preduzeće PTT‐saobraćajaSrbija.Beograd:SlužbeniPTTglasnik.

[3] Kujačić,M.,Nikoličić,S.,Jovanović,B.,Mirčetić,D.(2013).Logisticsperformanceinpostallogisticscenters.1stLogisticsInternationalConference,Belgrade,Serbia,209‐214.

[4] Lisec, A., Rihter, A. 2007. Logistical operations in postal logistics centres. Logistics &SustainableTransport,1(2),1‐12.

[5] Groznik, A., Kovačič, A., Trkman, P. (2008). The role of business renovation andinformationine‐government.JournalofComputerInformationSystems,49(1),80‐88.

SIMULATIONBASEDLIFE‐CYCLEANALYSISOFAVEHICLEFLEET

IvanDjokica,LjubomirLazicb,AleksandraPavlovica,*,AldinaAvdica

aStateUniversityofNoviPazarbMetropolitanUniversityofBelgrade,

Abstract:Throughouttheyearsthelogisticexpertshavedevelopedavarietyofsimulationmodelsforspecificapplications.Thispaperaddressesadiscrete‐eventsimulationmodel,whichestimatestheoperationalavailabilityandmaintenancecostofavehiclefleetthroughoutcompletelifecycle,underacertainmaintenancescenario.Themodelgivesallnecessaryparameterstocomputethetotalvehicle lifecyclecost, linking reliability,operational tempoandmaintenance scenariowithvehicleacquisitionandmaintenancecosts.Basedonsimulationresultsonecanmakecost‐effectivedecisionrelativetobuyingadequatevehiclesandorganizingproperfleetmaintenance,whichgivesrequiredoperationalavailabilityatlowestcosts.Inthisworkwehaveanalyzedapplicationofourmodelonanexample–alighttacticalvehiclefleet.

Keywords:vehicle,availability,maintenance,cost,simulation.

1.INTRODUCTION

Thetechnicalperformanceofvehicles(suchasspeed,range,stability,fuelconsumption,powergeneration) has improved significantly over the last several decades. On the other hand,suitability parameters (such as reliability, availability, and maintainability) have not beenanalyzed and improved. Suitability determinants are generally not addressed early enoughduringprogramdevelopmentandarenotprioritizedwiththesameseriousnessanddisciplineas performance parameters. The cost of operating and maintaining a vehicle fleet is a largeexpense for theowner,andsuitabilityperformance isamajor factoraffectingthesecosts.Theexisting off‐the‐shelf and best‐practice methods to select maintenance strategies are mainlybasedonexperienceandmanufacturersproposals.Improvementsareusuallydoneinatrialanderrormannerwithouttakingintoaccountcosteffectiveness.Withinthelastfewyears,pressureon costs and delivery on time have dramatically gained importance to optimizemaintenanceprocess. Very often logistics and maintenance objectives are separately optimized andoptimizationresultsaremoderate.

Inthisworkanintegrationofrequiredavailabilityandmaintenanceoptimizationisdone,basedon vehicle life cycle simulation. The simulation model is a very flexible one, and gives theopportunitytochangeavarietyofparameters:fleetsize,vehiclereliability,operationaltempo,work allocation between maintenance levels, time‐to‐repair distribution, maintenance cost –vehicle age correlation, preventive maintenance strategy, etc. The simulation results give adetailed insight into fleet life cycle: obtained availability of the fleet and every vehicle, every

* apavlovic@np.ac.rs

vehicle and fleet daily and total path, maintenance facilities utilization, queues, logisticsadministrative time,maintenance labor,maintenance cost (minimal,maximal,mean, standarddeviation).

2.RELATEDWORK

Thedevelopmentoflife‐cyclemodelsisnecessarytoidentifykeyfactorsthataffectoperationalreadinessandcostofrequiredreadiness.Modelingneedscomplexandtimeconsumingresearchto examinemany input parameters and possible scenarios, andmodels usually cover specificsystemor only apart of a life‐cycle.One approachusesoptimizationsmaximizing availabilityandprofitabilityoftheproductionsystembyvaryingbothmaintenancestrategiesandlogisticsfactors. The obtained results indicate that a joint optimization of logistics and maintenancestrategiesprovidesbetterresultsthanoptimizingthoseelementsindependentlyandhighlightstheneedforacomprehensivesophisticatedmodel(Achermann,2008).Someworksarefocusedon a particular influencing parameter. The authors stress the problem of influence of thereliability parameters for final system functional measures (required time of delivery). Thepresentedproblemispracticallyessentialfordefiningorganizationofvehiclemaintenanceandtransport system logistics (Walkowiak and Mazurkiewicz, 2007). To establish a relationshipbetween achieved reliability improvement and reduction in support cost, onemodel uses theCost Analysis Strategy Assessment (CASA), combining the two relationships—investment inreliability to reliability improvement and reliability improvement to support cost reduction(Long at al., 2007).Manyworks are dedicated to transportation systems. One paper suggestsanalysisbasedonthemodelingandsimulatingofthesystembehavior.MonteCarlosimulationisusedtoencouragereliabilityandstochasticfunctionalparameters.ThesimulatorisbuiltusingScalable Simulation Framework (SSF) (Walkowiak and Mazurkiewicz, 2008). In some otherworks the primary goal is to conduct comparison of different support strategies of a system.Smith(2011)analyzesthecostperunitusageandoperationalavailabilityofthemilitarytacticalvehicle.Transportationsystems,withdifferentlevelsofimportance,areanalyzedviasimulation.Naidu at al. (2010) used a discrete event simulation to highlight the difficulties involved in atypicalfour‐wheelerservicecenter.

Thisworkisaimedatdevelopingsimulationtoolforrevealingthemutualimpactofacquisitionand proving evidence for benefit of a joint optimization of logistics and maintenance,incorporatingavailability,logistics,andfinancialaspects.Inthispaperwesuggestasimulationmodel,basedonGeneralPurposeSimulationSystem(GPSS),whichallowsintegratedanalysisofcompletevehiclefleetlifecycle,fromacquisitiontoretirement.

3.SIMULATIONMODEL

ThesimulationmodelwasbuiltusingGPSSinsuchawaythatsoftwarenotonlyprovidestoolsformodelingandsimulationofawidevarietytomaintenanceservices,manufacturing,butalsohaspossibility toshape inputdataandcarryoutoutputstatistics(Schriber,1974).Simulationmodeldescribesamoderntwo‐levelmaintenanceconcept,showninFigure1.

The first maintenance level includes preventivemaintenance and correctivemaintenance forminor failures, and all maintenance activities are done via one of M available mobilemaintenance stations. The second level is dedicated to serious maintenance actions (repairscaused by serious failures), and maintenance depot with N working places is the availableinfrastructureofthislevel.Inputandoutputparametersofthesimulationmodelareasfollows:Inputparameters:numberofoperationalvehicles,fleetreserve(numberofvehiclesintendedtoreplace faulty vehicles),vehicle serial numbers, operational tempo, vehicle reliability,failuredistribution between minor and serious failures, maximal path between preventive

maintenance actions, firstmaintenance level capacity (M), secondmaintenance level capacity(N), time to execute preventivemaintenance, time distributions to repair vehicles, labor andsparescostformobilestations,laborandsparescostfordepotmaintenance,maintenancecost–vehicleagefactor.

Figure1.Two‐levelMaintenanceConcept

Output parameters: fleet availability, vehicle availability, vehicle usage histogram, number ofpreventive maintenance actions, vehicle preventive maintenance histogram, number of firstlevelcorrectiveactions,numberofsecondlevelcorrectiveactions,mobilemaintenancestationutilization,mobilestationsqueue,depotutilization,depotqueue,totalpreventivemaintenanceworking hours, total first level corrective maintenance working hours, total second levelcorrectivemaintenanceworkinghours,vehiclefailurehistogram,vehicledailypath,vehicletotalpath,vehiclemaintenancecost.

Operational availability. Operational Availability is a measure of the percentage of the totalinventory of a system operationally capable (ready for tasking) of performing an assignedmissionatagiventime,basedonmaterielcondition.

(1)

Operational Availability also indicates the percentage of time that a system is operationallycapableofperforminganassignedmissionandcanbeexpressedasuptimedividedbyuptimeplusdowntime,wheretuptimeisthetimewhenasystemisreadyforoperation,andtdowntimeisthemaintenancedowntime,whichincludesrepairtime,administrativeandlogisticsdelaytimes.In

this model Operational Availability is calculated for every vehicle and complete fleet usingequation (1). Fleet tuptime and tdowntime are calculated as the sums of each vehicle’s tuptime andtdowntime.Operationaltempo.Operationaltempoisameasureofthedynamicsofanoperationintermsofequipmentusage.Operationaltempocanbechangedbyincreasing/decreasingdailynumberofdriving hours. The driving speed depends on operational environment. Higher than expectedutilizationratesand fatiguecausedbyoperatingenvironmentareresulting inreducedservicelife.Maintenance.Maintenancedepicts theentityof all technical, technological, organizational, andeconomicactionstodelaywearoutand/orrecoveryoffunctionalcapability,includingtechnicalsafety,of a technical system.Two‐levelmaintenancestrategy, shown inFigure1,withregularscheduledmaintenanceactionsarerestoringthelostcapabilityofsubsystemimpairments,andthesemaintenanceactionsallowthevehiclestomeetoperationalstandardsandrequirements.Thatmeans‐vehicle failure intensitycanbeconsideredconstantthroughoutcompleteservicelife. In the case of extreme operational tempo (wartime operations of military vehicles, forexample),maintenancedone to counter the effects of it, to somedegree, but regardless of themaintenanceor “reset” completed, it doesnotbring the vehicle to a true “zero‐km” condition(USADoD,2008).Vehiclemaintenance cost.Themaintenance cost in ourmodel is defined by equation (2), andconsistsoftwocomponents,costofmaintenanceactivitiesdonebymobilemaintenancestations(Imaintenancelevel)andcostofdepotlevelcorrectivemaintenance(IImaintenancelevel). MCTotal=MCMobSt+MCDepot (2)

MCMobSt=WHMobStxCLabor(1+SMobSt)xKAge (3)

MCDepot=WHDepotxCLabor(1+SDepot)xKAge (4)Calculations of the mobile stations and the depot maintenance level costs are defined byequations (3) and (4). The maintenance cost depends on working hours spent on eachmaintenanceaction(WHMobStandWHDepot),averagecostofonehourlabor(CLabor),averagecostof spares replaced during one working hour (CLaborxSMobSt and CLaborxSDepot), and agingmaintenance cost correction factor (KAge). Having in mind that older vehicles show greaterprobabilities of having repair costs, our cost model introduces the aging maintenance costcorrection factorKAge,whichconnectsmaintenancecostandvehicleage.Valuesandshape forKAgemaintenancecostfactorisadoptedfromresearchreport(Pintatal.,2008).

3.EXAMPLE

Importanceofsimulationanditsuseinoptimizationliesinthefactthatmanyproblemsaretoocomplex to be described in mathematical formulations. Nonlinearities, combinatorialrelationships or uncertainties often give rise to simulation as the only possible approach tosolution.Oursimulationmodelistestedthroughrelativelycomplexexample:findthecheapestlife cycle solution for a fleet of 220 light tactical vehicleswith required availability 0.89.Lighttacticalvehiclesareplatformscapableforsmall‐unitcombatandtacticaloperationsincomplexurbanandruralenvironments,convoyescort,trooptransport,explosiveordinancedisposal,andambulancemissions.Fleetconsistsof200operationaland20reservevehicles.Thekeyreliabilityparameter is Mean Km Between Failure. This parameter depends on vehicle reliability andoperational conditions. For HMMWV (High Mobility Multipurpose Wheeled Vehicle)examplevehicle,undersupposedoperationalconditions,simulatedMeanKmBetweenFailurewas1300km.Maintenancealternativesaregeneratedbyvaryingnumberofmobilemaintenancestationsand number of depot working places. Nine alternatives were tested. Level I is the field

maintenance, consisting of preventive maintenance and 70% of corrective maintenanceactivities.LevelIIisthedepotmaintenance,covering30%ofvehiclefailures(seriousfailures).Scheduled preventive maintenance labor was fixed ‐ 2 hours per vehicle. Correctivemaintenanceworkinghoursfollowlognormaldistribution,withmeanequal4hoursforminorfailures,and10hoursforseriousfailures.

Figure2.Fleetavailabilityfordifferentvehicle‐maintenancealternatives

Figure3.Averageannualvehiclemaintenancecost

Eachmaintenancealternativewastestedin20operationalconditions(atotalof180scenarios).Thepresentedsimulationresultsare fromonepeacetimescenario.Theresults show,Figure2,that6of9maintenancealternativessatisfyrequiredfleetavailability,onealternativeismarginal(3d1m = 3 depot working places + 1 mobile station) and 2 alternatives do not satisfyrequirements.Foradoptedmaintenancealternative(1d2m=1depotworkingplace+2mobilestations) under supposed operational tempo, during 22 year service life average vehicleavailabilitywas0.8493with0.0509standarddeviation.Figure3showsatimedependentvehiclemaintenance cost (average 6065 $ with 927 $ standard deviation).Results suggest that theeffects of fleet aging on annual maintenance cost are significant. It is easy to observe thataverage annual maintenance cost for a vehicle 5 years old is 30% of the average annualmaintenancecostattheendofvehiclelifecycle.

4.CONCLUSION

Results of the simulation runs confirmed the assumption that system availability alone is aninsufficientobjectivefunctionforoptimizingamaintenancestrategy.Availabilityconsiderationshavetobemergedwithfinancialaspectstoachieveoptimalmaintenancestrategythatsatisfiesboth,therequiredavailabilityandlowestpossibletotallifecyclecosts.Oursimulationmodelisflexible enough to cover variety of scenarios: different requirements, different fleet size,different vehicles, different operational tempos, different maintenance strategies andinfrastructure, etc. In addition to the classic optimization criteria, as minimizing costs andmaximizing fleet availability, some other characteristics can be taken as supplementaryobjectivefunctions.Themodelfunctionalitywasdemonstratedthroughanillustrativeexample,fleetof220lighttacticalvehicles.

ACKNOWLEDGMENT

This research was partially supported by Ministry of Education, Science and TechnologicalDevelopmentofSerbia,underthegrantsTR32023andTR35026.

REFERENCES

[1] D.Achermann,(2008).Modelling,SimulationandOptimizationofMaintenanceStrategiesunderConsiderationofLogisticProcesses,AdissertationsubmittedtotheSWISSFederalInstituteofTechnologyZurichforthedegreeofDoctorofTechnicalSciences.

[2] T.Walkowiak, J. Mazurkiewicz, (2007). Discrete transport system simulated by SSF forreliability and functional analysis, Proceedings of the 2nd International Conference onDependability of Computer Systems(DepCoS‐RELCOMEX'07), Szklarska Poreba, Poland,352‐359.

[3] E.A.Long,J.Forbes,J.Hees,V.Stouffer,(2007).EmpiricalRelationshipBetweenReliabilityInvestments andLife‐Cycle Support Cost, Report SA701T1, LMIGovernmentConsulting,USA.

[4] T. Walkowiak, J. Mazurkiewicz, (2008). Availability of Discrete Transport SystemSimulatedbySSFTool,ProceedingsofThirdInternationalConferenceonDependabilityofComputerSystemsDepCoS‐RELCOMEX2008,SzklarskaPoreba,Poland,430‐437.

[5] D.A.Smith,(2011).AssessingtheSuccessoftheArmyHMMWVIntegratedLogisticsPlan,SubmittedinpartialfulfillmentoftherequirementsforthedegreeofMasterofscienceinOperationalresearch,NAVALPOSTGRADUATESCHOOL,USA.

[6] Ch.V.Naidu,P.M.Valli,A.V.S.R.Raju,(2010).ApplicationofSimulationfortheImprovementof Four Wheeler Service Sector, International Journal of Engineering and TechnologyVol.2(1),16‐23.

[7] T.J.Schriber,(1974).SimulationUsingGPSS,JohnWiley&Sons.

[8] USADoD(2008).MilitaryEquipmentUsefulLifeStudy‐PhaseII,FinalReport,DoD,OfficeoftheUnderSecretaryofDefense(Acquisition,TechnologyandLogistics)andOfficeoftheUnderSecretaryofDefense(Comptroller).

[9] E.M.Pint,L.PelledColabella,J.L.AdamsandS.Sleep,(2008).ImprovingRecapitalizationPlanning ‐ Toward a Fleet Management Model for the High‐Mobility MultipurposeWheeledVehicle,RandCorporation,USA,ISBN978‐0‐8330‐4174‐6.

SIMULATIONMODELFORIRPINPETROLSTATIONREPLENISHEMENT

DraženPopovića,*,MiloradVidovića,NenadBjelića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,DepartmentofLogistics,Serbia

Abstract:Qualityofatwo‐echelondistributionsystem,comprisedofvendorandasetofclients,isgreatlydependent on vehicle routingand inventorymanagement. Simultaneous optimization ofthese two segments in distribution systems is commonly known as Inventory Routing Problem(IRP),wheredecisionmakermustdeterminedeliveryquantitiesandvehicleroutes inaplanninghorizonof severaldays.For successful implementationof IRP,aVendorManagement Inventory(VMI)conceptmustexistinwhichvendorisresponsibleforinventorymanagementatclients’side.Many models for solving the IRP in different systems are based on deterministic input data,although in real life these systemshave some levelofuncertainty. In thispaperweobserve IRPapproachinpetrolstationreplenishment,wherewepresentsimulationapproachforapplicabilityanalysisofdeterministicsolutiontosystemswithstochasticnature.

Keywords:simulation,inventoryroutingproblem,heuristics,petrolstationreplenishment.

1.INTRODUCTION

InventoryRoutingProblem(IRP)isarelativelynewfieldthatisgaininganincreasingattentioninrecentyearsbytheinternationalresearchcommunity.TherearemanydifferentdefinitionsofIRPinavailableliteratureandperhapsthemostcomprehensivedefinitionisgivenbyCoelhoetal.(2014):“IRPcanbedescribedasthecombinationofvehicleroutingandinventorymanagementproblems, in which supplier has to deliver products to a number of geographically dispersedcustomers, subject to side constraints”. IRP comprises of two subsystems with a stronginterdependence, the vehicle routing and the inventory management, which are beingsimultaneously optimized. Andersson et al. (2010), based on the literature survey in variousindustries, defines three basic benefits that can be achievedwith the IPR: economic benefits,flexibilityofservice,andimprovedrobustnessofthesystemduetobettercoordination.IRPcanbe observed as an extension to the vehicle routing problem in a sense of taking intoconsideration the inventorymanagement segment, usually in the planning horizon of severaldays. The basic prerequisite for the implementation of IRP is the VendorManaged Inventory(VMI) concept, which implies the application of modern information and communicationtechnologies. To solve IRP, many authors develop models that use deterministic data input(deterministic consumption) to simplify the problem and enable the solving of the probleminstances.Inrealsystems,consumptionhasstochasticcharacterandthissimplificationmayleadtoadverseeffects,primarilyasshortagesorexcessesof inventoriesandunplannedchanges in

* d.popovic@sf.bg.ac.rs

the realizationof thedeliveryplan.Applicability analysis of thedeterministic solutions to therealsystemofstochasticnaturecanbemadewiththeuseofsimulation.

Thispaperisonepartofthedoctoraldissertation(Popović,2015)wherewepresentheuristic’ssolutions(whichpresumesdeterministicconsumption)applicabilityanalysistothesystemwithstochasticconsumptionusingsimulation.Weobservepetrolstationreplenishmentprobleminwhichfuelmustbedistributedfromonedepottoasetofpetrolstations,whereeachstationhasstochastic consumption in real life. In the proposed simulation, stochastic consumption canresultinthepossibilityoftwonegativeeventswhichrepresenttwoadditionalperformancesforevaluating the quality of solutions (in addition to inventory and routing costs): inventoryshortagesduetounplannedhighconsumption,andtheneedforurgentdeliveriestominimizetheinventoryshortages.

The rest of this paper is organized as follows. Section 2 contains brief literature reviewwithmain focus on stochastic IRP. Section3presents IRP inpetrol station replenishmentproblemformulationwith deterministic consumption. In Section 4we present proposed heuristic andsimulation models used for applicability analysis of deterministic solution to systems withstochasticfuelconsumption.TestinstancesandcomputationalresultsarepresentedinSection5.ConclusionsandfutureresearchdirectionsaregiveninSection6.

2.LITERATUREREVIEW

TwomostrecentIRPreviewpapers(Anderssonetal.2010,Coelhoetal.2014)madesimilarIRPclassificationbasedon themost importantproblemcharacteristics,presented inTable1. IRPrelatedpapersusuallyobservedeterministicsystemstosimplifytheproblemandtoobtainthesolution,whereconsumption isrepresentedwithsomeexpectedvalue.Oneof the firstpapersthatobserved IRPwithdemanduncertaintywasbyFedergruenandZipkin (1984)where thisuncertainty can lead to inventory stock‐out. Golden et al. (1984) also observed IRP withconsumption uncertainty where they developed heuristic model based on clients’ serviceemergency.BermaniLarson(2001)analysedsomewhatdifferentcaseofstochasticIRPwheredriversdon’tknowexactclient’sdemandquantitybeforevisitingthatclient.

Table1.IRPclassification(derivedfromAnderssonetal.2010,Coelhoetal.2014)

CHARACTERISTIC POSSIBLEOPTION

Timehorizon Finite Infinite Structure One‐to‐many Many‐to‐many Many‐to‐one Routing Direct Multiple Continuous Inventorypolicy Fixed Stock‐out Lostsale BackorderFleetcomposition Homogeneous Heterogeneous Fleetsize Single Multiple Unconstrained Numberofproducts Single Many

InsomepapersauthorsusesimulationtoanalyzeapplicabilityofdeterministicIRPsolutiontoreal‐lifesystemswithstochasticnature.OneofthosepapersisJailletetal.(2002)whereheatingoildistributionfromonedepottoasetofcustomerswasobserved.RollinghorizonframeworkwasusedtoobtainsolutionforlongertimeperiodandMonteCarlosimulationwasusedtotestthe solutions. Customers’ consumption was randomly generated from a truncated normaldistribution. Hemmelmayr et al. (2010) used simulation approach to analyze differentdistributionconceptsinblooddeliverytohealthcareinstitutionswherebloodconsumptionwasrandomly generated from a truncated normal and uniform distribution. From stochastic IRPliterature, some general conclusions can be made that are applicable to petrol stationreplenishment and therefore incorporated in the approach presented in this paper: solutionfrom deterministic IRP is obtained by an approximation of stochastic consumption/demand;

simulation can be used to analyze applicability of deterministic IRP to stochastic systems;preventive and correctivemeasures arenecessary todecreasenegative effects ofuncertainty;rollinghorizonframeworkisusedtoobtainsolutioninlongertimeperiod.

3.PROBLEMFORMULATION

TwomaindecisionstobemadeinIRPpetrolstationreplenishmentare:(1)fuelquantitiestobedeliveredpereach fuel typetosetofpetrolstations inpredefinedplanninghorizonofseveraldays;(2)accordingtodeliveryquantities,routeconstructionforfueldeliveryfromonedepottoset of petrol stations. Objective function, in the process of decisionmaking, tries tominimizetotalcostswhicharecomprisedofroutingandinventorycosts.Routingcostisdefinedbytotaltravel distance of all vehicle routes, while inventory costs are defined by average inventorylevelsofdifferentfueltypesatpetrolstations.ObservedIRPcanbedescribedasdistributionofseveral fuel types (j 1, 2, ..., J) in one‐to‐many systemwith homogenous fleet of vehicles inplanninghorizonofseveraldays(t 1,2,...,T).Ineachdayoftheplanninghorizononepetrolstation i 1,2, ...,Icanbeservedwithonlyonevehicle.EachpetrolstationhasconstantdailyconsumptionpereachfueltypeqijandundergroundtankofknowncapacityQij(oneforeachfueltype). It is not allowed that inventory levels in petrol stations for any fuel type fall belowpredeterminedsafetystock level.Fuel is transportedbya fleetofvehiclesFwithvehiclesthathave compartments. Only full compartments can be delivered to petrol stations. Number ofcompartmentsinavehicleforfueldistributionusuallyvariesfrom4to6andwetestproposedmodelwiththreevehicletypes.Giventhetotalnumberofcompartmentsinvehiclesandmorethanonefueltype,inpracticeonevehiclecanserveuptothreepetrolstationsinasingleroute(Cornillieretal.2009),andthislimitationisappliedinourmodel.

4.HEURISTICANDSIMULATIONMODELS FORIRPINPETROLSTATIONREPLENISHMENT

ObservedIRPinpetrolstationreplenishmentisacomplexcombinatorialproblem,asisthecasewithmostIRPproblems(Andersssonetal.,2010),whereoptimalsolutioncannotbeobtainedinacceptable computational time. Therefore, a heuristic approach is a necessity and we havedeveloped a Variable Neighbourhood Search (VNS) model (Figure 1.a). This VNS modelconsidersfuelconsumptionasadeterministicvalueinordertoobtaindeliveryplanforagivenplanning horizon. Since fuel consumption is uncertain in real life, we developed simulationmodel to analyze applicability of a VNS solution to the stochastic case (Figure 1.b). Both,heuristic and simulation models were implemented using C++ programming language on PCIntel(R)Core(TM)i3CPUM380@2.53GHzwith6GBRAMmemory.

4.1VNSheuristic

TheVNSmetaheuristicwasintroducedbyMladenovićandHansen(1997)withthebasicideaofsystematic change of neighbourhoods in the process of finding improvement of current bestsolution.Therefore,VNS isbasedonmultipleneighbourhoods inwhichalgorithmattempts tofindlocalbestsolution.Searchoflocalminimumofmultipleneighbourhoodsimproveschancesto find a global minimum. We have developed general VNS heuristic that has the followingprocedures:initialsolution,shakingprocedure,localsearchprocedure.Initialsolutioniscreatedbydeliveringminimumfuelquantity(onefullcompartment)atthelatestpossiblemoment(dayof planning horizon) regarding level of safety stock. Then, vehicle routes (for each day ofplanning horizon) are created using Sweep algorithm which are improved by Swap andReallocate/Insertion (VND‐route) procedures. Last step of obtaining the initial solution isCompartmenttransferprocedurewhichtriesto improvethesolutionbychangingthedayofafuelcompartmentdelivery.Thisinitialsolutioncanbeadditionallyimprovedbytwoinventorylocal search (VND‐IR)procedures: reallocationof a single fuel compartmentdeliverybetween

daysofplanninghorizon;reallocationofallfuelcompartmentsinapetrolstationbetweendaysofplanninghorizon.Shakingprocedureisusedtopartiallyandingraduallyintensifiedmanner“destroy” (parameters of destruction are increased) current best solution and then toreconstructittoobtainsomenewsolutionwhichislocatedinanothersegmentofsolutionspace(comparedtocurrentbestsolution).Fromthisnewsolution,VNDprocedureswithrandomizedorderofneighbourhoods(RVND‐routeandRVND‐IR)areusedtoobtain improvement.Finally,newsolutioniscomparedwiththecurrentbestsolution.If thenewsolutionisbetterthanthecurrentbestsolution,thenitbecomescurrentbestsolution(andshakingprocedureparametersarereseted).VNSheuristicalgorithmispresentedinFigure1.a.

4.2Simulationapproach

Havinga sufficient amountof fuel inpetrol stations that canmeet the consumption is abasicrequirement for solution excellency in petrol station replenishment problem. In the case ofstochastic consumption there is a risk of stock‐out due to the possibility of excessiveconsumption.Toavoidstock‐outs,safetystocksareused(asapreventivemeasure)whichhavethe role ofmeeting the extreme fuel consumption deviations. Larger consumption deviationsrequireahigherlevelofsafetystockthatcauseshigherinventorycosts.Asolutionwhichwillbebasedonlyonraisingsafetystock isnotcost‐effective (highcosts related tocapital stock).Toavoid a high level of safety stocks, with simultaneous satisfaction of stochastic consumption,conceptofemergencyfueldelivery(asacorrectivemeasure)forthosepetrolstationswhichareindangerofstock‐outisanecessity.Simulationisbeingusedfortheanalysisoftheapplicabilityofdeterministicsolutionsintermsofroutingandinventorycosts,aswellasadditionalnegativeeffectsofinventorystock‐outsandthenumberofemergencydeliveries.

Figure1.VNSheuristicandsimulationalgorithms

The basic idea of simulation approach is to obtain solution for a planning horizon using VNSheuristic, and then to simulatedaily fuel consumptionusingappropriate randomdistribution.TosimulateconsumptionweusedNormaldistributionqsijt∼N(µ,2)(truncatedleftfromzerotoforbid negative consumption), as it was used in some other papers (see Jaillet et al. 2002,Hemmelmayretal.2010).Aftertherealizationofdeliveriesandsimulatedconsumptions,modelcheckstheinventorylevelofalfueltypesinallstations.Ifinventorylevelfallsbelowsafetystock

level, an emergency delivery is activated. Any excessive fuel consumption that cannot besatisfied is registered as stock‐out. Length of simulation planning horizon TSIM is limited bycomputational time,and thiswas thereasonwhywe introduceda rollingplanninghorizon inwhichweusedVNSheuristictosolvesuccessiveshorterperiodsT<TSIM.Inthesimulation,fromeach solution of a shorter period T only first day delivery was executed (expect the shorterperiod T at the end of TSIM) where this first day takes into account expected costs of entireshorterperiodT.SimulationalgorithmispresentedinFigure1.b.

5.COMPUTATIONALRESULTS

Intheproposedsimulationapproachweobservedfollowinginstances’characteristics:15petrolstations; 3 fuel types; fleet size isF=4; transportation cost per travelled kilometre is 2€/km;dailyinventoryholdingcostis1€/1000l;dailyfuelconsumptionpertypeqijcanrandomlytakevalues from intervals [1000‐3000 l], [800‐2000 l], [500‐1500 l]; petrol stations’ undergroundtanksper fuel typecanhave randomvalues from intervals [30000‐50000 l], [20000‐40000 l],[20000‐30000 l]; petrol station inventory level per fuel type at the start of simulation canrandomlytakevaluesfrom[2*qij,8*qij];petrolstationsarerandomlylocatedinasquare[‐100,100]km,whiledepotislocatedat[0,0];simulationplanninghorizonisTSIM=16days,whileeachshorterperiodisT=4days.Wehaverandomlygenerated100testinstanceswith10simulationrunspereachinstance.Additionally,wetestedtheimpactoffollowingparametersonsolutions’quality:

differentconsumptiondeviation:=0.2*µ,=0.3*µ,whereµ=qij; differentvehicletypes[K–numberofcompartments,Qo–compartmentcapacity] [4,

8800l],[5,7000l],[6,5800l]; differentlevelsofsafetystocks:V=0.5*qij,V=1.0*qij,V=1.5*qij,V=2.0*qij.

Simulation results are presented in Table 2. Results show that increase in safety stock levelconsequentlyincreasesbothinventoryandroutingcosts(duetoincreaseindeliveryquantities).Depending on the level of consumption deviation different safety stock level is required toreduce stock‐outs to practically zero (between 1.0*qij and 1.5*qij). Additionally, number ofemergencydeliveriesishigherwithhigherlevelofconsumptiondeviation.Vehicleswithmorecompartmentshavelowertotalcostsbutalsomoreemergencydeliveriesandstock‐outs.

Table2.Simulationsresultsfor100testinstances(averagewith10repetitions)

Vehicletype[K,Qo]

V[qij]

Routingcosts[€]

Inventorycosts[€]

Totalcosts[€]

CPUtime[sec]

Emergencydeliveries

Stock‐outs[l]

0.3* [4,8800l] 0.5 10968.9 5348.5 16317.5 8.6 6.76 292.31.0 11323.6 5875.0 17198.6 8.8 7.33 22.41.5 11720.2 6447.3 18167.5 8.5 7.06 0.82.0 12122.1 6987.7 19109.8 8.7 6.96 0.0

[5,7000l] 0.5 11149.9 4883.2 16033.1 11.3 7.29 323.01.0 11485.3 5393.5 16878.9 11.8 7.94 24.21.5 11947.7 5954.8 17902.5 11.4 7.59 1.12.0 12328.7 6481.9 18810.6 11.7 7.37 0.0

[6,5800l] 0.5 11397.5 4598.3 15995.8 14.1 7.82 379.81.0 11731.2 5090.1 16821.3 14.7 8.42 32.11.5 12178.7 5644.8 17823.5 14.4 8.02 0.12.0 12540.8 6177.9 18718.7 14.8 7.87 0.0

0.2* [4,8800l] 0.5 11059.7 5317.7 16377.4 8.7 4.59 47.21.0 11388.9 5848.0 17236.9 8.9 4.92 0.71.5 11813.1 6416.7 18229.9 8.6 4.77 0.02.0 12189.2 6956.4 19145.6 8.7 4.53 0.0

[5,7000l] 0.5 11231.1 4858.7 16089.9 11.5 4.81 52.11.0 11534.6 5364.6 16899.2 12.0 5.31 0.61.5 11987.0 5914.5 17901.5 11.6 5.22 0.02.0 12356.0 6445.2 18801.2 11.9 5.12 0.0

[6,5800l] 0.5 11424.9 4560.0 15984.9 14.3 5.23 56.91.0 11773.6 5059.8 16833.4 15.0 5.65 2.11.5 12224.2 5607.5 17831.7 14.6 5.33 0.02.0 12547.9 6141.9 18689.8 15.0 5.36 0.0

6.CONCLUSIONS

Simulation can be used to analyze petrol station replenishment problem in differentenvironmentsandtochoosethemostsuitabledistributionvariant(e.g.vehicletype,safetystocklevel).Basedonsimulationresultspresentedinthispaper,generalconclusioncanbemadethatthe total inventory and routing costs decrease with the increase of the number of vehicles’compartments (smaller capacity of a single compartment). On the other hand, looking at theaspectofservice(numberofemergencydeliveriesandstock‐outs)vehicleswithsmallernumberofvehicles’compartments(largercapacityofasinglecompartment)aremoresuitable.Decisionmakersmustdetermine thedegreeof importancebetweenqualityof service (possible lossofrevenue) and the total inventory and routing costs, and based on that to design distributionstructure and policy. Potential research directions are adjustments of the proposedmodel todifferent petrol stations replenishment distribution structures as well as modifying theproposedmodelsforuseindifferentbusinesssystemswithsimilarcharacteristics,suchasthecollection of waste and recyclables, transport of live animals, collection of milk or olive oil,distributionoffrozenproductsthatarekeptatdifferenttemperatures,etc.

ACKNOWLEDGMENT

Thisworkwassupportedby theMinistryofEducation,ScienceandTechnicaldevelopmentoftheGovernment of theRepublic of Serbia through theprojectTR36006, for theperiod2011‐2015.

REFERENCES

[1] Andersson,H.,Hoff,A.,Christiansen,M.,Hasle,G.,Lokketangen,A., (2010). Industrialaspectsandliteraturesurvey:Combinedinventorymanagementandrouting.Computers&OperationsResearch,37,1515‐1536.

[2] Berman,O.,Larson,R.C.,(2001).DeliveriesinanInventory/RoutingProblemUsingStochasticDynamicProgramming.TransportationScience,35(2),192–213.

[3] Coelho, L.C., Cordeau, J.‐F., Laporte, G., (2014). Thirty Years of Inventory‐Routing.TransportationScience,48(1),1‐19.

[4] Cornillier, F., Laporte, G., Boctor, F.F., Renaud, J., (2009). The petrol station replenishmentproblemwithtimewindows.ComputersandOperationsResearch,36,919‐935.

[5] Federgruen, A., Zipkin, P., (1984). A Combined Vehicle Routing and Inventory AllocationProblem.OperationsResearch,32(5),1019‐1037.

[6] Golden,B.,Assad,A.,Dahl,R.,(1984).Analysisofalargescalevehicleroutingproblemwithaninventorycomponent.Largescalesystems7(2‐3),181–190.

[7] Hemmelmayr,V.,Doerner,K.F.,Hartl,R.F.,Savelsbergh,M.,(2010).Vendormanagedinventoryfor environmentswith stochastic product usage, European Journal of Operational Research,202,686–695.

[8] Jaillet, P., Bard, J.F., Huang, L., Dror,M., (2002).Delivery Cost Approximations for InventoryRoutingProblemsinaRollingHorizonFramework.TransportationScience,36(3),292–300.

[9] Mladenović,N.,Hansen,P.,(1997).VariableNeighborhoodSearch.ComputersandOperationsResearch,24(11),1097‐1100.

[10] Popović,D.,(2015).Inventoryroutingproblem:modelingandperformanceanalysis.Doctoraldissertation,UniversityofBelgrade,FacultyofTransportandTrafficEngineering,DepartmentofLogistics,Serbia.

SIMULATIONMODELOFAQUEUINGSYSTEM:THECASESTUDYOFAFAIRTRADEMANIFESTATIONINNOVISAD

VladimirIlina*,DraganSimića,NenadSaulića

aUniversityofNoviSad,FacultyofTechnicalSciences,Serbia

Abstract: The main idea of this paper is to evaluate supply processes at a Fair Trade (FT)manifestationinNoviSad.Tosuccessfullyorganizesuchamanifestationdetailedpreparationplanand transparent exhibitors’ arrival schedule are required. In our case, exhibitors’ arrivals arerandom, which may result with long queues and bottlenecks. Simulation model is created inMATLAB to evaluate whether the FTmanifestation was overload with queuing during supplyprocesses. According to obtained results all vehicles that entered manifestation are servicedwithouttheneedforlongqueuing.

Keywords:FTexhibition,simulationmodel,supplyprocesses,performanceevaluation.

1.INTRODUCTION

ThispaperanalyzesperformanceevaluationofaFairTrade(FT)manifestation.AccordingtotheWorldFairTradeOrganization,FT isatradingpartnership,basedondialog, transparencyandrespect, which seeks greater equity in international trade (WFTO, 2015). Observed FTmanifestationislocatedinthesecondlargestcityinSerbia–NoviSad.Itspreadsontheareaof226.000m2andtheindoorexhibitionareacovers60.000m2.Thereare37halls,amongwhichthemostup‐to‐dateistheMasterHall,whichspreadson5.970m2ofexhibitionarea.TourismFTmanifestationwasinthemainfocusofthisanalysis.ItwasheldintheMasterHall.

Themainideaofthispaperistoprovideananswershouldthestochasticscheduleofexhibitors’arrivals be reorganized and replaced with deterministic timetable? The main dilemma iswhether there is a short time window in which exhibitors’ set up their exhibits? If so, theexecution of supply processeswill be dysfunctional. Simulationmodel provides calculation ofvariousperformanceparameterswhichshouldgivetheanswertokeyquestioninthisanalysis.Performance parameters (1) utilization (U), (2) rejected exhibitors (R), (3) mean number ofvehiclesinthequeue(Lq),and(4)meantimevehicleshavespentinthequeue(Wq)areselectedasthemainindicators.

The rest of the paper is organized in the followingway. Section 2 describes a discrete eventsimulation and queue modeling in logistics. Section 3 considers the methodology. Section 4explainssimulationmodelandprovidessimulationresults.ConcludingremarksandfutureworkfollowinSection5.

* v.ilin@uns.ac.rs

2.DISCRETEEVENTSIMULATIONANDQUEUEMODELINGINLOGISTICS

Discrete‐eventsimulation(DES) ismodelingapproachwidelyusedasdecisionsupport tool inlogisticsandsupplychainmanagement.Specificexamplesofthe issuesthattheseDSSaddressaresupplychaindesignandreconfiguration, inventoryplanningandmanagement,productionscheduling and supplier selection (Tako and Robinson, 2012). Simulationmodels are usuallybuilt to understand how systems behave over time and to compare their performance underdifferent conditions. DES can be applied at operational and tactical level (Figure 1). Severalpapershave suggested thatDES is not suitable for strategicmodeling as it doesnotnormallyrepresent systems at an aggregate level (Baines andHarrison, 1999). For strategicmodeling,systemdynamics (SD) ismore adequatemodeling approach. In some cases the use of hybridsimulationapproachescombiningDESandSDgivesthebestresults(Rabeloetal.,2005).

Figure1.Orderingoflogisticsissuesintostrategicandoperational/tactical(TakoandRobinson,2012)

DESmodelssystemsasanetworkofqueuesandactivitieswherestatechangesoccuratdiscretepoints of time. In DES entities (e.g. objects, people) are represented individually. Specificattributesareassigned toeachentity,whichdeterminewhathappens to themthroughout thesimulation.Statisticalcalculationsareusuallyperformedtodescribeentities’attributes.

Queuingmodelsarewidelyusedinservicefacilities,production,material‐handlingsystems,andin situations where congestion or competition of scarce resources can occur. Generally, aqueuing system isdescribedbydefining its population, thenatureof arrival, the service timeandmechanism,queuingbehavior,andthequeuingdiscipline(Jahnetal.,2010).Thepurposeofqueuing models is to provide information about important performance parameters such asqueue lengths, response times,waiting times,utilization,probability thatanydelaywilloccur,probabilitythatsomecustomerswillberejected,probabilitythatallservicefacilitieswillbeidle,etc.(BhaskarandLallement,2010).

Inourapproach,U,R,Lq,andWqaretargetedperformanceparametersinthesimulationmodel.Performanceparameter‐U‐shouldrevealhowhighthepercentageofthesystemusageis?Thehighervalueofutilization is shown, thehighervalueof systemusage is implied.Performanceparameter‐U ‐measureswhatpercentageofdailysupplyactivitieshasbeenoperational.Thekeyparameterwastimeonthebasisofwhichsimulationresultsarecalculatedandproposed.

Performance parameter ‐ R ‐ should show whether there were rejected exhibitors fromservicing.Exhibitorsarerejectedincaseofexcessivequeuing.Thosesituationsareunacceptableinpractice.Consideringthatexhibitorsarriverandomly,organizersoftheFTmanifestationneedto make available sufficient number of places for vehicles that are waiting in a queue.Performance parameter ‐Lq ‐ should indicatewhat is the average number of vehicles duringqueuing.Correspondingly,performanceparameter‐Wq‐shouldshowwhatistheaveragetimethatvehicleshavespentinthequeue.

3.METHODOLOGY

Themost efficientmethod for data collectionwas quantitative research. The total number ofmonitoredvehiclesattheTourismFTwas56attwoservicingchannels.Servicingchannelsareplaces in front of entrances into exhibitionhalls. They canbe comparedwith regular parkingplaces,buteachservicingchannelcanserveonlyonevehicleatonce.Thesizeofentrancesintoexhibitionhallsandvehicles’dimensionsrestrictthatonlyoneexhibitorcansetupitsexhibitssimultaneously.ObservedFTmanifestationismediumsizedexhibitions.Itlastedforthreedays.A sampling frame was identified through Novi Sad exhibition web site and organizationalschedulingprogram.Table1showstheframeworkofthisresearch.

Table1.Basicinformationabouttheresearch

ObservedFTmanifestation

TourismFairTrade

Typeofresearch Quantitativeresearch

Theaimofresearch Toidentifymaincharacteristicsofarrivalrates(λ)andservicerates(µ)duringsupplyprocesses

Targetgroups vans,carsObtainedresults 56vehiclesrecordedResearchperiod October2013.

With the FT management support, the lead author and three research helpers were able toapproach exhibition on‐site and conduct the research at servicing channels. After arriving atexhibition area, the helpers were assigned to different entrances to Master Hall to ensuresystematiccoverageofsupplyprocesses.TheillustrationoftheFTmanifestationispresentedtovisualizesimilaritywiththequeuingsystem(Figure2).

Figure2.MasterHallattheFTexhibitioninNoviSad

It can be noticed that there are two entrances into Master Hall in front of which exhibitorsunload their exhibits. Both entrances are observed in the context of one queuing system, notseparately.Ifmorethanonevehiclearrivesatthesametimequeuingwilloccur.OrganizersofFTmanifestationneedtoprovidesufficientspaceforvehicleswhenqueuingoccurs.Exhibitors’

arrivals are random. All supply processes are organized one day before the FTmanifestationbecomesopenedforvisitors.

4.SIMULATIONMODELANDRESULTS

Collecteddatafromthequantitativeresearcharemergedintoonedataset.Thefirststepwastoperform statistical analysis and define the input parameters for the simulation model. TodetermineprobabilitydistributionsofcollecteddatasoftwareMINITABhasbeenused.Themaincriterion during statistical analysiswas that the p‐value from obtained statistic results be, atleast,at the5%of thesignificance level.Thep‐value indicates theprobability that testeddataareincorrelationwithselecteddistribution.ObtainedresultsfromstatisticalanalysisimplythatarrivalratesareincorrelationwithNormaldistribution(p‐value=0.743)andserviceratesareincorrelationwithWeibulldistribution(p‐value=0.187).

SoftwareMATLABand its graphical programming tool Simulinkareused to create simulationmodel. MATLAB provides a relatively easy‐to‐use, versatile, and powerful simulationenvironment for investigating the basic, as well as advanced, aspects of dynamics systems(Hung,1998).Simulinkissimulationsoftwaredevelopedforsimulatingandanalyzingdynamicanddiscrete systems,which iswidelyusedwithin industry for representingprocessbehaviorandcontrolsystems(Asbjörnssonetal.,2013).ThesimulationmodeliscreatedusingSimulink’sSimEventssectionwithdifferentblocks.Thegraphicaldrag‐and‐dropinterfaceinSimulinkwasenvironmentinwhichadiscrete‐eventsimulationmodelwasbuild(Figure3).Thekeyfeaturesinclude (1) Libraries of predefined blocks, such as queues (first‐in‐first‐out (FIFO) queue),servers (N‐server), generators (Time‐based entity generators) and sinks (Entity sink) formodeling system architecture, (2)Built‐in statistics such as number of entities timed‐out andutilization, and (3) M‐function for calculating the main results (Inyiama, 2012). Event‐BasedRandomNumberblockgeneratesrandomnumbersfromspecifieddistribution,parameterandinitialseed.Dependingonthetypeofdetermineddistributioninputparametersforarrivalsandservicing are statistically determined. Time‐Based Entity Generator block generates entitiesusing integration times that satisfy specified criteria.The integration time is the time intervalbetweentwosuccessivegenerationevents.StartTimerblockassociatesanamedtimertoeacharrivingentity independentlyandstart thetimer.ReadTimerblockreadsthevalueofa timerthat theStartTimerblockwaspreviouslyassociatedwith.FIFOQueueblockstoresentities insequence first‐in, first‐out for undetermined length of time. N‐Server block stores up to Nentities,servingeachoneindependentlyforaperiodoftimeandthenattemptingtooutputtheentitythroughtheOUTport.EntitySinkblockacceptsallblocksentitiesandprovidesawaytoterminateanentitypath.

Figure3.Basicblocksinthesimulationmodel

Arrivalratesandserviceratesaremeasuredduring12hoursperiod,aslongassupplyprocesseslasted.Insomecases,supplyprocessesmaylastlongerdependingoftheexhibitors’dynamicsofarrivalsandexhibitionsize,ingeneral.Themaincriterionfordeterminationoftheupperlimitofpredictionperiodwasaveragedurationofsupplyprocessesofmedium‐sizedFTmanifestations.According toFTmanagers’ expertise and experience theupper limits is five additional hours.

Therefore, thesimulationandpredictionresultsofparametersU,R,Lq,andWqarecalculatedcorrespondingly(Table2).

Table2.Simulationandpredictionresultsfromthesimulationmodel

TimeintervalsKeyperformanceparameters 12h 13h 14h 15h 16h 17h

R 0 0 0 0 0 0

U 67,7% 68,2% 68,8% 69,1% 69,5% 69,8%

Lq 3,41 3,62 3,84 3,91 4,02 4,13

Wq 10,43min 10,64min 10,73min 10,85min 11,06min 11,22min

Table2showsthatperformanceparameter‐R‐hasvalue0forsimulationperiodof12hours,aswellas forallpredictionperiods.Performanceparameter ‐U ‐ shows thatsystemhasn’tbeenused up to its limits, considering that obtained values from the simulationmodel are smallerthan 100%. Performance parameter ‐Lq ‐ shows average value of 3,41 vehicles in the queueduringsimulationperiod.Predictionresultsfor‐Lq‐indicatethatthereisnolargeincreaseinobtained values considering that the greatest value is 4,13. Correspondingly, performanceparameter ‐Wq ‐ shows average value of 10,43 minutes for each vehicle during queuing.Similarly, prediction results for ‐ Wq ‐ do not show a significant difference compared tosimulationresult.

5.CONCLUDINGREMARKSANDFUTUREWORK

Theaimofthisanalysiswastoprovideananswershouldthestochasticscheduleofexhibitors’arrivals at the FT manifestation be reorganized and replaced with deterministic timetable?Accordingtoobtainedresultsfromthesimulationmodelallvehiclesthatenteredmanifestationareservicedwithouttheneedforlongqueuing.Performanceparameter‐R‐hasvalue0forallsimulation periods, which implies that there are no rejected exhibitors from servicing.Performanceparameter ‐U ‐suggests thatmanifestationhasnotbeenoverloadwithqueuingduringsupplyprocesses.PredictionresultsimplythatatthecurrentoperatingregimeTourismFTmanifestationcanefficientlyfunctionfor17hoursconsideringthat30%oftimeduringdailysupply activities system hasn’t been used. Performance parameter ‐ Lq ‐ also supportseffectivenessofFTmanifestationbecauseFT’smanagers easily canorganizeup to10waitingplaces in front of each entrance intoMasterHall. Performance parameter ‐Wq ‐ implies thatexhibitors need to wait to unload their exhibits around 10minutes. Unloading of exhibits atTourism FT manifestation is relatively fast, because the exhibits are mostly composed ofpromotional tourismmaterial. Simulation results of performance parameters imply that at apresentoperatingregimestochasticscheduleofexhibitors’arrivalsatTourismFTmanifestationisquiteefficientsupplystrategy.

Future research can be focused towardsmore comprehensive analysiswith larger number ofperformance parameters, such as mean number of vehicles on servicing, mean number ofvehiclesinthesystem,meantimevehicleshavespentonservicingandmeantimevehicleshavespent in the system. Also, different sized manifestations may be analyzed and compared.Additional data are required in that case. Proposed simulation model is universal for eachmanifestationheldattheFTinNoviSad.Additionalblockscanbeaddedincaseswhenlarger‐sizedmanifestationsarebeinganalyzed.

ACKNOWLEDGMENT

TheauthorsacknowledgethesupportforresearchprojectTR36030,fundedbytheMinistryofScienceandTechnologicalDevelopmentofSerbia.

REFERENCES

[1] Asbjörnsson,G.,Hulthén, E., Evertsson,M. (2013).Modelling and simulation of dynamiccrushingplantbehaviorwithMATLAB/Simulink.MineralsEngineering,43‐44,112‐120.

[2] Baines,T.S.,Harrison,D.K.(1999).Anopportunityforsystemdynamicsinmanufacturingsystemmodeling.ProductionPlanningandControl,10(6),542–552.

[3] Bhaskar, V., Lallement, P. (2010). Modeling a supply chain using a network of queues.AppliedMathematicalModelling,34(8),2074‐2088.

[4] Hung, G.K. (1998). Dynamic model of the vergence eye movement system: simulationsusingMATLAB:SIMULINK. ComputerMethods andPrograms inBiomedicine, 55(1), 59–68.

[5] Inyiama, H.C., Chidiebele, U., Okezie, C.C., Kennedy, O. (2012). MATLAB Simevent: AProcess Model Approach for Event‐Based Communication Network Design. Journal ofBasicandAppliedScientificResearch,2(5),5070–5080.

[6] Jahn,B.,Theurl,E.,Siebert,U.,Pfeiffer,K.P.(2010)TutorialinMedicalDecisionModelingIncorporatingWaitingLinesandQueuesUsingDiscreteEventSimulation.ValueinHealth,13(4),501‐506.

[7] Rabelo, L., Helal, M., Jones, A., Min, H. (2005). Enterprise simulation: a hybrid systemapproach.InternationalJournalofComputerIntegratedManufacturing,18(6),498–508.

[8] Tako,A.A.,Robinson,S. (2012).Theapplicationofdiscrete event simulationandsystemdynamicsinthelogisticsandsupplychaincontext.DecisionSupportSystems,52(4),802‐815.

[9] WFTO (World Fair Trade Organization), 2015. About WFTO. http://www.wfto.com(accessed13.03.2015.).

PartVII QUALITYMANAGEMENTAND

PERFORMANCEMEASURESINLOGISTICS

CITYLOGISTICSPERFORMANCE

SnežanaTadića*,SlobodanZečevića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Problem identification, modeling, planning and management of the city logisticssolutionsrequiresknowledgeabout theparametersof logistic flowsandsystems,generatorsandlogisticsdemandsoftheurbanenvironment,i.e.theenvironmentinwhichrequirementsarebeingrealized.By combining the parameters, city logistics performances are being obtained.A set ofperformances and the way of their identification are not standardized, while the permanentmonitoring of some part of the parameters is in fact an exception present in certain cities.Performancesareanalyzedanddefinedatthedifferentlevelsofresearch,whilebeinginlinewiththe basic objectives and scope of the research. In this paper, the most commonly quantifiedparameters and performances of city logistics are presented. The necessary parameters arecollectedandprocessedindifferentwayssotheirreliabilityandcomparabilityarequestionable.

Keywords:citylogistics,performance,research,reliability.

1.INTRODUCTION

Problemidentification,modeling,planningandmanagementofthecitylogistics(CL)requiresalargenumberofparameters thatdescribe the logistical flows and systemswhichenable theirimplementation, aswell as theparametersof the logisticsdemandsgenerators and theurbanenvironment, i.e. theenvironmentinwhichtherequirementsarebeingrealized.Bycombiningtheparameters,CLperformancesarebeingobtained (Tadić,2014).Considering that logisticalflowsdependontheparametersoftheurbanenvironment(infrastructureconditions,economicstructure,etc.),andthattheirimplementationaffectstheparametersoftheurbanareas(trafficcongestion, accessibility, air pollution, etc.), in addition to their establishment, continuousmonitoringandupdatingofparametersofthesurroundingandperformancesofthecitylogisticsisnecessary.However,asetoftheCLparametersandperformancesthatneedtobemonitored,aswellasthewayoftheirdetermination,arenotstandardized.Performancesarebeingdefinedaccording to the specific research objectives, afterwhich the necessary parameters are beingcollected,processed,analyzedandquantified.Acomprehensivestudyofthelogistics,intermsofspace,activities,typesofgoodsflows,activitiesandprocessesandparticipants,isaconceptthathasnotbeenimplementedinanystudyofacertaincityintheworld.Themainreasonsaretheextremelyhighcomplexityandthelackofunderstandingoftheproblem,butalsothedifferentgoalsandinterestsofdirectandindirectparticipantsoftheCL.Thepaperpresentsasetofthemostcommonlyusedcitylogisticsperformances,aswellasthemethodsfortheirdetermination.

2.IMPORTANCEOFTHECITYLOGISTICSPERFORMANCE

* s.tadic@sf.bg.ac.rs

IdentifyingCLproblemsrequiresknowledgeonthecharacteristicsoflogistics,goodsandtrafficflows, as well as the systems which enable their realization. The volume of flows is directlyrelatedtothesizeofthecity.However,inadditiontotheconsumergoods,itisalsonecessarytoexaminetheflowsoftheintermediaterawmaterials,constructionmaterialsanddifferenttypesofcargo.Ontheotherhand,thevolumeofflowslargelydependonthegeographicandphysicalcharacteristics,spatialorganizationoftheurbanfunctions, logisticsinfrastructureandtheroleoftheobservedzoneintherealizationofcargoflowsonahigherlevel(Tadićetal.,2015).

DuetothelackofresearchandconstantmonitoringoftheCLperformances,urbanauthoritiesandplannersdonothaveaclearandcompletepictureofthelogisticsactivitiesandprocesses.On the other hand, without the knowledge of the current situation, and without theidentification and quantification of the CL performances and problems, it is not possible tosearchfortheeffectivesolutions(Figure1).

Statisticalbulletins

CITYLOGISTICSPERFORMANCES

DATASOURCES

BusinessRegisters

Urbanplans

Consultationsandforums

Businessentities

documents

Logisticsactivitiesrecording

Flowgeneratorsanalysis

Logisticsprovidersanalysis

Urbanenvironmentparameters

Logisticsflowsparameters

Logisticssystemsparameters

Flowgeneratorsparameters

Modellingofthecitylogisticssystem

SelectionoftheCLconception

CITYLOGISTUCSINITIATIVES

CITYLOGISTICSPERFORMANCES

Figure1.Determinationofthecitylogisticsperformancesandtheirimportance(Tadić,2014)

Theparametersarebeingobtainedusingthedifferentdatasources,suchas localandnationalstatistical bulletins, transport statistics, business registries, urban plans, vehicle diaries andwaybills, discussion forums, consultations with local governments and lawmakers. However,most of the information is being obtained by a dedicated research of the logistics activities(recordingoftheloading/unloadingzones,GPSvehicletracking,etc.)andparticipantsoftheCL(generators and logistics providers). By combining the parameters, the performances thatdescribe thestateof theCLarebeingobtained.Performancesmaybeused for identifying theproblems and modeling of the flows and CL system. On the other hand, the success of theinitiatives, measures and concepts of the CL is determined by comparing the performanceefficiencyandthesustainabilityofthesystembeforeandafteritsimplementation.Applicationof

the certain CL initiatives and concepts requires the change of the urban plans, infrastructuresystems, legal regulations, price policy, traffic organization andother elements that representdata sources; it changes the parameters of the urban environment, attributes of the flowgeneratorsandparametersof the logistics flowsandsystems, i.e.CLperformances(Figure1).The lack of monitoring of the CL performances is more or less present in all the cities. Theexceptions are the statistical data on freight traffic in the cities, but they are also relativelyuseless.Statisticsonlykeeptherecordsonthetripsoftheregisteredcarriers,andonlyforthevehicleswiththegrosscapacityofover3.5tones,whilelightcommercialvehicles(lessof3.5t)havethemostsignificantgrowthinthenumberofvehicle‐kminthecity(Browneetal.,2007)andmakeupthelargestpartofthecargovehicles,usuallyover80%(DfT,2012).Inaddition,theresearchdoesnotincludeshoppingtrips,whichinthelargeEuropeancitiesmakeupbetween45and55%oftheurbangoodsmovement(Gonzalez‐Feliuetal.,2012)andcanparticipatewith15‐20%ofthetotalvehicle‐km(Dablanc,2009).

2.1Theparametersoftheurbanenvironment

The parameters of the urban environment include spatial, demographic and economiccharacteristics of the city and they are generally publicly available. Themain sources are thepublicationofthepublicbodiesatthelocalandnationallevels(statisticalbulletins,urbanplans,commercial registers, etc.). For the analysis of the CL performances the most importantparameters are: size, population, family size, age structure, economic structure, number ofemployees,serviceprice,taxesandthenumberofjobs.Theseparametersaredeterminedforthewholecityandtheurbanzonesandhaveagreatimportancefortheassessmentandmodelingofthe CL flows and systems (Tadić, 2014). Logistics activities in the city largely depend on thecharacteristics of the traffic network: categories of roads, dedicated lanes, organization andregulationofthemovement,taxistands,location,organizationandparkingcapacity,etc.Theseparameters affect the structure of the vehicle in the realization of the commodity flows, theintensity of commercial vehicles and traffic burdening, delivery vehicles' route planning,durationoftheloading‐unloadingoperations,operatingcostsofthecarriers,serviceprovidersandothers.

2.2Theparametersofthelogisticssystem

Mostoftheflowsthatstartorendinthecityarerealizedthroughthevariouslogisticssystems.InordertodeterminetheperformancesandconceptualmodelingsolutionsoftheCL,oneneedstoknowtheirnumberandstructure,purpose(typeofgoods),ownership,size,location,capacity,transshipment volume, the available modes of transport, but also the technologies of thelogistics subsystems for storage, transshipment and transport of the incoming and outgoinggoodsflows(Tadić,2014).Parametersofthelogisticssystems,especiallythosethatarethepartof the public infrastructure, are generally available and could be found in the planningdocuments, reports and publications of the city administration. However, determining theparametersoftheprivatesystemsrequiresresearchandconsultationwiththeowner.

2.3Theparametersofthelogisticsdemandsgenerators

Thegenerator,i.e.thefacilitythatrequiresthedeliveryand/orthecollectionofgoods,maybeseen as an entity with the attributes (parameters) which describe it. The parameters differaccordingtotheeconomicsector,butalsowithinthesameindustry.Themostcommonlyusedparametersofthegeneratorsare(Zečević&Tadić,2006):businessactivity,size,ownershipandlocation,thestructureofthegoodsintheincomingandoutgoingflows,usedsystemsoforderingandsupplying,thesizeofthestorageareainthefacility, thetimeofreceivinganddispatchingthegoods,sizeandfrequencyofdeliveriesandshipments, thestoppingplaces forthevehiclesperformingtheoperationsofloading/unloadingofgoods,typesofvehiclesperformingdelivery

etc.Inmostcities,onlysomeoftheseparametersarebeingrecorded(e.g. location,ownership,business category, total surface area), while the largest number of the attributes are beingobtained in theone‐off researches,mostly through the interviewswith the employees andbyrecordingtheloading/unloadingoperations(Tadić,2014).

2.4Theparametersofthelogisticsflows

At the national level, the basic parameters of the commodity and transport flows are beingmonitored(volumeandstructureofthegoodsandtransportflows,transportwork).Incertaincountries, someotherparameters, suchas (Browne&Allen,2006): transport intensity, trafficintensity, energy consumption; average transport distance; loading factor; number of emptyruns etc. are being determined. However, none of these parameters is being determinedexclusively for the urban freight transport, and the cases in which some of them are beingstudied apart from the national researches are rare. This is likely the result of the weakunderstandingoftheCLimportanceatalladministrationlevels.Themostimportantparametersof the logistic flows are: type of vehicle, number of trips, time of realization, tour duration,numberof requestsper tour, tripduration,dwell time in frontof theobject, the lengthof theroute, the distance traveledby the vehicle, the vehicles' operation system, type of goods, fuelconsumptionetc.(Tadić,2014).Parametersarebeingdeterminedforallflowcategories(supply,recycling materials and service activities) and require extensive research (recording of thelogisticsoperations,driving logs andwaybills analysis, surveyof theparticipantsof the cargoflows‐logisticsproviders,carriersanddrivers,etc.).Duetothelackofcontinuousmonitoringofgoodsandtrafficflowsinthecity,themodelsthatenabletheobtainingofthepartofthelogisticsflowsparametersonthebasisoftheone‐offresearchesaredeveloped.

3.BASICCITYLOGISTICSPERFORMANCE

CLperformancesdescribethelogisticsactivitiesandprocessesandenabletheidentificationoftheproblemsandthequantificationofthespecificmeasuresandinitiativesapplicationeffects,and they arebeingdefined in accordancewith theobjectivesof the specific researches.Theiranalysis,whiledefiningthespatial,urbananddevelopmentplans,canimprovethesustainabilityoflogisticsandtheurbanenvironment.Generatorsstructureandthenumberofvehiclelaunchrequirements are theparameters that directly affect thenumber, capacity and location of theinfrastructure systems for loading/unloading of the commercial vehicles. It is often wronglyassumed that the large retail chains are the largest generators of the freight traffic.However,they generally use a centralized system of supplying, bigger freight vehicles and have apredefineddynamicsanddeliverytimes.Ontheotherhand,smallshopsandspecializedstoresmayberesponsibleforasignificantpartoftheactivitiesoffreightvehicles,mainlyvans.Inordertoplan the logisticsactivitiesand theCLconcepts,butalso todefine theurbanplans (spatial,traffic),itisnecessarytoknowvariousCLperformances,someofwhichare:

Thedeliveryfrequencydiffersfromcitytocity,betweentheeconomicsectors,butalsowithinthe same industry. At the city level, it depends on the economic structure, economicdevelopment,market conditions, shareof the informal sector, shareof the large retail chains,presenceofthelogisticsproviders,users'behavior,timeoftheyear,policiesandregulationsofthelocalgovernmentetc.

The delivery time is an important parameter for the planning activities of the shipper andreceiverofgoods,aswellasforthelogisticsproviders,i.e.thecarriers.Itdependsonregulationsthat define the access for the delivery vehicles (intervals of the restricted movement of thecertaincategoriesofvehicles,restrictionsrelatedtothestoppingandloading/unloadingofthevehicles, etc.); deliveries of the time‐sensitive goodswhich loses its value after a certain time(dailynewspapers,freshdairyandbakeryproducts,etc.),aswellasthetimeintervalsinwhichthefacilitycanreceivegoods(duringtheworkinghoursandoff‐hours).Suppliers(shippers)and

carriers(logisticsproviders)havethegreatestinfluenceonthedeliverytime,whileasubstantialpartoftherecipientshasnoeffectonthetimeofreceivingthegoods(Cherrettetal.,2012).

Thestructureofthedeliveryvehiclesdependsonthebusinesscategory,thesupplyingsystem,requirements forgoods,delivery frequency, infrastructuralconstraints,regulationsthatdefineconditionsofaccessetc.Inthecentralizedsupplysystem,thedeliveriesaremainlyrealizedbythe different vehicle categories, while in the decentralized they are mainly realized by vans.Passengercarsdominateinsupplyingthefacilitiesoftheinformalsector.

The return loading of the delivery vehicles shows the frequency of the recollection of thegoodsfromthefacility(reusablepackaging,defectedgoodsorthedateexpiringgoods,goodsfortheresupplyinsomeotherfacilities)tothedistributioncenter,suppliers'depotorsomeotherfacility in the route.Using thedelivery vehicles for realization of the reverse flows, transportefficiency is increasing, the number of vehicles and vehicle‐km is reducing,, as well as thenegative impactson theenvironment.However,asmallpartof thedeliveryvehiclesregularlyrealizethereverseflows,whilealmosthalfofthegeneratorsdonotusetheservicesofreverselogisticsbytheproviderthatrealizesthedeliveries(Cherrettetal.,2012).

The dwell time of the delivery vehicles affects the deliveries coordination, vehicle routing,planningandbetteruseof loading/unloadingzonesorstreetparkingplaces.Researchesshowthattheaveragetimespentattheplaceofunloadingthevehicledependsonthevehiclecategory(longer for larger vehicles) and service provider (longer in insourcing strategy). Researchesshow that there is not a strong correlation between the average dwell time of the deliveryvehicleandthesizeofthefacility,i.e.thesupplysystem(Cherrettetal.,2012).

The place of unloading activities depends on the urban area. In areas with large shoppingcenters (mainly peripheral zones) delivery vehicles mainly stop on the reserved off‐streetparkingspaces,whileinthecentral,shoppingstreets,theystoponthestreet.Placeandtimeofthevehiclestopping,fortheloadingandunloadingoperationsatthefacility,areaffectedbythetypeandformofthegoods.Somegoodsrequiretheuseofspecialvehiclesorspecialreloadingequipment,sothevehiclesarestoppingascloseaspossibletothefacility.

Thenumberofdeliveriespertourdependsonthelogisticsprovider,carrier,deliverysizes,therequired delivery time, supply system etc. By applying the system of handling the multiplefacilitiesinasingletour,thevehiclesaretravelinglongerdistancesbecausetheydonotusetheshortestroutestoeachfacility,butontheotherhandthenumberoftripsissignificantlysmallerthan in the system of the direct delivery. In practice, the number of vehicle‐km is usuallyminimizedbytheuseofthedirectdeliveryandsmallervehicles.However,thedirectdeliveryisoftennotacceptableforseveralreasons:asmallamountofgoods,longdistances,etc.

Thefrequencyofservicevisitstothefacilitydependsonthebusinesscategoryandtypeoftheservice.Serviceflowscanhaveasignificantimpactonthetotalnumberofcommercialtrips.Onaverage,thefrequencyoftheservicevisitstothefacilitiesis7.6onaweeklybasis(Cherrettetal.,2012).Thehighestfrequencieshavemaildeliveriesandwastecollections,whileothersarenot that often (maintenance of computer equipment, copier devices, elevators and escalators,etc.).

Vehicle types used for service visits and dwell times indicate that these activities may beresponsibleforasignificantconsumptionofparkingspacesandpavementsintheurbanareas.About70%oftheservicevisitsarerealizedbythemotorvehicles,halfofwhicharevans.Thevehicles'dwelltimedependsonthetypeofservice,andtheUKresearchesshowthatthemeandwelltimeforallservicesis35min(Cherrettetal.,2012).

AwidesetofperformancescanbeobtainedbycombiningthebasicCLperformanceswiththeparameters of the urban environment and logistics systems. The number and type of theobtained CL performances depends on the research objectives, but each of them helpsunderstandingthesituationoflogisticsandidentificationofcriticalelementswithrespecttothe

business, time interval, logistics organization (insourcing and outsourcing), performers of thedelivery,vehicle type,etc.For theassessmentofefficiencyandsustainabilityof the initiatives,measureand conceptual solutionsof theCL, themost commonlyquantifiedperformancesare(Zečević&Tadić,2006;Tadić,2014):thenumberoftrips, i.e.thenumberofvehicleslaunchedfor the distribution of goods; the number of vehicle‐km; the required number of deliveryvehicles; driving time; delivery size; loading factor; delivery reliability; time and frequency ofloading/unloadingoperations; therequiredcapacityof thestoragesystems; fuelconsumption;emissions;distancetraveledperunitofdelivery;operatingcosts;etc.

4.CONCLUSION

ResearchesoftheCLsignificantlydifferbythenumberandqualityofthequantifiedparameters.The informationconcerning the collectionandprocessingofdataaremainly scarce, thereforetheirreliabilityandcomparabilityarequestionable.EveninthecountrieswithasignificantbaseoftheCLparameters,mostoftheinputsareobtainedbythedisaggregationofthedatacollectedfor a wider geographical area. On the other hand, the parameters are being collected by thedifferentpublic sectororganizations,aspartsof theshort‐term(one‐off) studiesandprojects,but alsoby theprivate sectororganizations, including industrial, retail, utilities and transportcompanies, tradeassociationsandcompanies involved in themarket research.However, theirstudiesarenotcoordinatedandthereisalargenumberofthedatasourcesthatvaryinqualityand methodology of the assessment. This makes the comparison or combination of theparametersverydifficultorimpossible.Eveninthecitieswherethelargenumberofparametersiscollected,itisstillimpossibletogetaclearpictureoftheCLsystemwhenthedataunites.

REFERENCES

[1] Browne,M.,Piotrowska,M.,Woodburn,A.,Allen,J.(2007).LiteratureReviewWM9:PartI‐UrbanFreightTransport,UniversityofWestminster.

[2] Browne,M.,Allen, J. (2006).Urban freightdatacollection‐synthesisreport.BESTUFSII,BESTUFSConsortium.

[3] Cherrett, T., Allen, J., McLeod, F., Maynard, S., Hickford, A., Browne, M. (2012).Understandingurbanfreightactivity–keyissuesforfreightplanning,JournalofTransportGeography,24,22–32.

[4] Dablanc,L.(2009).FreightTransportforDevelopmentToolkit:FreightTransport.WorldBank.

[5] DfT (2012). Road traffic (vehicle‐kms) by vehicle type and road class in Great Britain.DepartmentforTransport,London,GreatBritain.

[6] Gonzalez‐Feliu,J.,Ambrosini,C.,Pluvinet,P.,Toilier,F.,Routhier,J.L.(2012).Asimulationframework for evaluating the impacts of urban goods transport in terms of roadoccupancy.JournalofComputationalScience,3(4),206‐215.

[7] Tadić,S. (2014).Modeliranjeperformansi integrisanihcity logističkihsistema.FacultyofTransportandTrafficEngineering,UniversityofBelgrade,Serbia,PhD.

[8] Tadić, S., Zečević, S., Krstić, M. (2015). City logistics – status and trends. Internationaljournalfortrafficandtransportengineering,inpress.

[9] Zečević, S., Tadić, S. (2006). City logistika. Faculty of Transport andTraffic Engineering,UniversityofBelgrade.

FORECASTINGDEMANDINTHELOGISTICSMARKET:ACASESTUDYOFLOGISTICSCENTERVRŠAC

MiloradKilibardaa,MilanAndrejica,*,MladenKrstića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: This paper proposes a newmethodological approach for estimation of demands andflows in the logisticsmarket.Proposedapproach isdeveloped ina comprehensiveway tobetterunderstandandevaluateperspectivetrendsoflogisticsflowsanddemandsforlogisticsservicesinconditions of great uncertainty, variability and unpredictability of geopolitical, economic,commercial,transportandtrafficfactorsandopportunitiesinthelogisticsmarket.Theprocedureis applied in the concrete example for the forecasting of demands for logistics services andsubsystemsoffuturelogisticscenterVršac. Inthecasewhenthereisalargevariabilityindemandandwhentherearenoobvioustrendsthatcanbeappliedinthefuture,theauthorshavedecidedtocarryoutthepredictionofthreepossiblescenariosforeconomicandsocialdevelopment. Thegoalwas to develop approach thatwill use the larger field of possible future events on the logisticsmarket.

Keywords:logisticsmarket,logisticsflows,demands,forecasting,logisticscenter

1.INTRODUCTION

The logisticsmarket is definedas theplacewhere flowsofmaterial and serviceproducts arecreated, realized and finished. The subjects of trade in thesemarkets arematerial goods andlogisticsservicesrelatedtotheefficientflowofgoods.Theelementsofcommodityandtransportflowsare:generatorsofcommodityflows‐users(industry,trade,cateringandcraftenterprises,residents,economicandnon‐economicinstitutionsandorganizations,etc.);demandandmarketrequirements for a particular structure and volume of material and service products (rawmaterials, intermediate goods, consumer goods, finished products, transport and logisticsservices,etc.);offerofthecertainstructureandvolumeofmaterialgoods,transportandlogisticsservices for product delivery to the final consumers; carriers of the transport and logisticsservices,whichareable toofferandprovidespecificservices in therealizationofcargo flows(transport companies, freight terminals and freight villages, shipping companies, logisticsoperators,serviceprovidersandothers);places inwhichtheycarryoutcommodity flowsandprovide transport and logistics services (area, location, relations); timeof theoccurrence andrealizationofcommodity flows, transportationand logisticsdemandsassociatedwiththe flowanddeliveryofgoods.(Kilibarda,2011).

Logistics flows and requirements are extremely variable categories and sizes in terms of allparameters that characterize them. Markets, customers, place of origin and the end of thelogistics flow are variable. Structure, volume and appearance of logistics demands are also * m.andrejic@sf.bg.ac.rs

variable,aswellas technologyandcarriersof logisticsproviders.Variabilityand instabilityofcommodityflowsistheresultofanumberoffactorsthatdirectlyaffecttheappearanceandtheflow dynamics and intensity. In situation of variability and unpredictability of geopolitical,economic,commercial,transportandtrafficfactorsandopportunitiesinthelogisticsmarketitisverydifficulttogiveareliableassessmentoftheperspectivetrendsofthelogisticsflows.

In contrast to all of these difficulties the successful development of logistics systems and thecompaniesisnotpossiblewithoutforecastingandthinkingaboutwhatwillhappeninthefuture(Leeuw et al. 1998). Logistics systemsmust analyze the futuredemands for services inmoredetail,andpredicttherelevantmarket,economicandlogisticstrends.However,becauseoftheextremespecificityandcomplexityofrelationshipsinthemarket,inthelogisticsliteratureandpractice,therearenouniversalforecastingmodelsandmethods(Mcgivern,2003;Mcneil,2005).Also, standard methods and techniques often can not be directly applied. It is necessary todevelop specific research and forecastingmethodologies of logistics flows anddemandwhichwill includeall relevant factorsandpossiblescenariosof thedevelopment inaccordancewiththe characteristics of specific markets and logistics systems. In this paper mentionedmethodologyisproposed,whichwasdevelopedinordertoassesstheperspectivetrendofthelogisticsflowsanddemandsforservicesandsubsystemsoflogisticscenter.

2.METHODOLOGICALAPPROACHFORLOGISTICSDEMANDFORECASTING

Bearinginmindall theproblems, limitations,aggravatingfactors,aswellastheimportanceofthis problem, the authors have decided to develop and implement a complexmethodologicalapproach for forecasting. The process involves a series of interrelated and coordinated steps(figure1):commodityflowsmarketdefinition,theanalysisofcommodityflowscharacteristics,definition of development scenarios and factors, estimation of commodity flow perspectivetrend,commodity flowredistribution, logisticscenterand intermodal terminalVršacdemandsdetermination.

In the catchment area of logistics center and intermodal terminal in Vršac, it is possible toextract the four typical types ofmarkets: M1 ‐Vršacmarketwhich includes users of logisticsservices and commodity flow generators which are located in the municipality of Vršac.Economicentitiesareprimarily located in the technologyparkand in the industrial zone,butalso in thewide area of Vršac;M2 – Narrow catchment area include userswho are spatiallydistributed in the South Banat andmiddle Banat District, and gravitate towards the logisticscenterandintermodalterminalVršac;M3–WidercatchmentareaincludesallusersoflogisticsservicesintheareaofVojvodina,BelgradeandSerbia,forwhichitisreasonablyexpectedtoberealized through theborder crossingVatin, andmunicipalityofVršac.Mainly it is import andexportflowsbetweencertainareasofSerbia,ontheoneside,andRomania,Russiaandeasternmarkets,ontheotherside;M4–Internationalmarket,whichreferstocustomerswhoareontheterritoryofRomaniaandneighboringcountries,withflowscanreasonablybeexpectedtoservelogisticscenterand intermodal terminalVršacsubsystems.Fordefinedmarketseconomicandmarket potentials are perceived, characteristics of commodity flows, particularly in terms ofrealizationoverthelogisticscenter.Adetailedanalysisofstrategiesandpotentialofeconomicdevelopmentwasmade.Plansandbusinesstrendsofeconomicdevelopment inarediscussed.Relationbetweenthedevelopmentpotentialontheperspectivetrendoftradeflowsfordifferentdevelopmentscenariosareinvestigated.

For the described market, a detailed identification and quantification of import, export andtransit commodity and transport flows is done. Trade flows were investigated and analyzedaccordingtoallrelevantproperties,suchas:structureandintensityofflow,typeandquantityofgoods, direction and relations, etc. Commodity flows are investigated and analyzed for theperiodof2005‐2012.Commodityflows,currenttrendsandfuturedevelopmentplans,aswellastheavailablelogisticalresourcesoftherelevantbusinesssubjectsinthecatchmentareaare

identified and quantified through the survey and interviews. Specific requirements of thecompaniesforlogisticscentresubsystemsaredetermined.

Several factors which reflect market, business, economic, logistic and transport environmenthave strong influence on the appearance and realization of commodity flows in the region(Mentzer and Schroeter, 1997). In volatile and uncertain conditions that a longer period ischaracterizedbycommercial,economic,geopoliticalandsocialtrendsintheregion,itisdifficulttoreliablypredict theeffectofcertain factorson thecommodity flowtrend.Due to theglobaleconomiccrisis, therestructuringprocess, it isnotpossible identifyclear trendsthatcouldbecontinuedinthefuture.Forthesereasons,theresearchteamdecidedtoestimatecertaintrendsinrelationtothevariousscenariosofpossiblesocialandeconomicdevelopment.

Figure1.Procedureofestimationcommodityandtransportflowperspectivetrends

In order to cover asmuchof thepossible future events, the threedevelopment scenarios aredefined:basic(realisticallyexpected),pessimisticandoptimisticscenario.Fivegroupsoffactorsare identified: F1 ‐ structure and character of themarket (from limited to expansive growthmarket),F2–corporateandeconomicdevelopment(fromslowtointense),F3‐political,marketand economic integration (lack ofmarket linkages to accelerated European integration), F4 ‐development of transport infrastructure (from underdeveloped and unrelated to the fullydevelopedandassociatedtransportand logisticsnetwork)andF5‐developmentofa logisticscenter with intermodal terminal (from slow to rapid construction and development). Thesefactorshavedirectimpactonthecertainscenarios(table1).

Table1.Scenarioandeconomicfactordevelopment

FACTORS

DEVELOPMENTSCENARIOSPessimistic Basic(Real) Optimistic

F1‐structureandcharacterofthemarket

Limitedmarket Mediumexpansionofthemarket

Expansivemarketgrowth

F2–Economicdevelopment Sloweconomicdevelopment

Rapideconomicdevelopment Intensiveeconomicdevelopment

F3–Marketintegration Lackofmarketandeconomicintegration

Regionalmarketintegrationandconnectionwithcountriesinregion

AcceleratedEuropeanintegration

F4–Transportinfrastructuredevelopment

Lowlevelofdevelopmentofroadtransportinfrastructure

Revitalizationofroadandrailinfrastructure

RapidconstructionandconnectionwithEuropeanandPanEuropeantransportnetwork

F5–Developmentoflogisticscenterandintermodalterminal

Slowconstructionoflogisticscenterandintermodalterminalandweaklyattractingofcommodityflows

ModeratedevelopmentofthelogisticscenterandintermodalterminalVršac

Fastconstructionanddevelopmentofthelogisticscenterandintermodalterminalandstrongattractionofcommodityflows

Estimationofcommodity flowperspective trend is realized inaccordancewithdifferent flowscategories,typesofgoodsanddifferentmarkets.InthelogisticscentreandintermodalterminalVršaccatchmentareatherearetwotypesofcommodityflows:flowswithstablecharacteristics,which in the longruncontinuouslyappearwithgreaterorsmalleroscillationsand flowswithextremelyvariablecharacteristics,whichinoneperiodoftimeappearinasignificantshapeandintensitywhileinthenextperiodalmostdidnotappear.

Inaccordancewiththeabove,expertestimationsofcommodityflowforperiodofnexttenyearsare determined. Each estimation has a confidence interval in which can be found totalcommodity flow. For each assessment number of different options and correspondingprobability of possible states are defined. The final assessment of the size of transportQ is arandomvariable,anditisdefinedforthreedefinedscenarios.Inthisway,forcommodityflowisobtainedpessimistic, realisticandoptimisticexpectedvalue.Perspective trendsareestimatedforallgoodscategoriesandall regions.Forecastedvalues forchemicalproductsareshown inFigure2,whileforecastingvaluesforSouthBanatregionarepresentedinFigure3.

Figure2.Perspectivetrendsofcommodityflowsinnarrowcatchmentarea(a‐import;b‐export)–exampleofchemicalproducts

a) b)

It is realistic to expect that import flows of chemical products in narrow catchment in thebeginningof the timehorizonSerbia import25000 t,while at the endof timehorizonSerbiaimport35000tthroughnarrowcatchmentarea.Inobservedtimehorizonexportflowswillfrom15000 t to 20000 t. Optimistic perspective trends for import and export flows in widercatchmentareaare50000tto65000tforimportandfrom20000tto27000tforexportflows.Mentionflowsareveryimportantforlogisticscenterterminaloperating.

Figure3.Perspectivetrendsofcommodityflowsinwidercatchmentarea(a‐import;b‐export)–exampleofSouthBanatregion

After estimation commodity flow prospective trend redistribution of commodity flows ontransportroutes,modesoftransportandintermodaltransporttechnologies isdone(figure4).Redistribution is based on next factors: type and quantity of goods, transport distance, theadvantages and shortcomings of certain types of transport, transportation resources andfacilities, development of transport infrastructure, etc. Several criteria like transport time,transportcosts,resourcesutilization,servicequalityandenvironmentalcriteriaareused.Inthispaper two main redistribution categories are presented: according transport distances andaccordingmarkets.

Figure4:Commodityflowredistribution(a‐accordingdifferenttransportdistances;b‐accordingmarkets)

Foreachflowfromthepreviousstepassessmentofpossibleaffiliationtologisticscenterisdone.Affiliationwasmadeonthebasisof:categoriesandvolumeofcommodityflow,goodstypes,theoriginandtheendofcommodityflow,directionanddistance,thelocationofthegeneratorandconnectionsoftransportinfrastructurewithlogisticscenterandintermodalterminal.Asshownin table 2 fourmarketswith different zones are analyzed: a zone of very high affiliation (themarket of municipality of Vršac), zone of high affiliation (narrow catchment area), zone ofmedium affiliation (wider catchment area) and the zone of lower affiliation (internationalmarket).

The estimation took into account the position of the logistics center and intermodal terminalVršactocompetingcentersandterminalsincatchmentarea,suchasPančevoandTimisoara.Forbasic subsystems logistics center of Vršac estimation of demands volume with differentprobabilitiesthatcanbeexpectedfromthefourtypesofmarket.Theassessmentwasperformedaccordingtothepossibledevelopmentscenarios(pessimistic,expected,andoptimistic).Inthis

a) b)

a) b)

paperwarehouseandcustomsterminalareanalyzedinmoredetails.Accordingtable2itiseasytoseethatthemostofthedemandsforcustomsandwarehouseterminalarecomingfromVršacmarket(M1)andnarrowgravityarea(M2).

Table2.Estimationofdemandsforsubsystemsoflogisticscenter(exampleofcustomsandwarehouseterminal)

Markets(Pallets/year)

Customsterminal Warehouseterminal

Pessimisticscenario

Basicscenario

Optimisticscenario

Pessimisticscenario

BasicscenarioOptimisticscenario

M1‐Vršacmarket 11960‐17680 14040‐9760 20800‐26400 11900‐15500 15300‐18600 19100‐23300

M2‐Narrowcatchmentarea

11960‐15600 14040‐9760 16700‐23200 12400‐15500 13500‐16700 15000‐18100

M3‐Widercatchmentarea

6240‐8320 10400‐4040 12200‐16500 3800‐5600 5700‐9200 14300‐18000

M4‐Internationalmarket

2600‐3640 3640‐4680 4300‐5500 2500‐4600 4100‐6400 9200‐11900

TOTAL 32760‐45240 42120‐8240 54000‐71600 30600‐41200 38600‐50900 57600‐71300

3.CONCLUSION

Anewmethodologicalapproachforforecastingofdemandsandflowsonthetransportmarketisproposed in this paper. The main advantage of the proposed approach lies in the fact thatapproach includes all relevant factors of themarketing environment for the logisticsmarket.Obtainedresultsshowgreatpracticalapplicabilityof themodel.Resultsalsoshowthatmodelhasuniversalapplicabilityforforecastingdemandsforservicesandsubsystemsofplannedandprojectedlogisticscenters.Theabove‐mentionedresultsrepresentabasisforfurtherresearchtowardstheidentificationandquantificationofrequirementsforthesubsystemsandstructuralelementsofthelogisticscenter,thedeterminationofthecorrespondingvaluesfordimensioningsubsystem and determining expected revenues and evaluation of financial and economicfeasibilityoftheproject.

REFERENCES

[1] Feasibility study for Logistic Centre and Intermodal Terminal Vršac, University ofBelgrade, Institute of the Faculty of Traffic and Transport Engineering, University ofBelgrade,2014.

[2] Kilibarda,M. (2011).Marketing in logistics,UniversityofBelgrade, FacultyofTransportandTrafficEngineering,Serbia.

[3] Leeuw,S.Donselaar,K.,andKok,T.(1998).ForecastingTechniquesinLogistics,AdvancesinDistributionLogistics,Volume460,pp481‐499.

[4] Mcgivern, Y. (2003). The Practice of Market and Social Research: An Introduction, 2ndEdition,PearsonEducationLimited,London

[5] Mcneil,R. (2005).Business toBusinessMarketResearch:Understanding andMeasuringBusinessMarkets,KoganPage,UnitedKingdom.

[6] Mentzer,J.T.andSchroeter,J.(1997).Integratinglogisticsforecastingtechniques,systems,andadministration:themultipleforecastingsystem,JournalofBusinessLogistics,Vol.15,pp.205‐225.

LOGISTICSFAILURESINDISTRIBUTIONPROCESS

MilanAndrejica*,MiloradKilibardaa,VladoPopovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: The importance of failures identification, monitoring and elimination in logistics isrecognized in literatureandpractice.Failureseliminationdirectlyaffectsthe increaseof logisticsservices quality. This further has impact on costs reduction, increasing customer satisfaction,loyaltyand finally revenue increasing. In thispaper logistics failures inproductdistributionareidentified.Methodological approach for identification and correction of failures in distributionprocessisproposedinthispaper.Failuresinwarehouse,transport,inventoryandotherprocessesaredescribed inmoredetails.Also themain causesare identified. In thispaper the failuresareanalyzedin13distributioncentersofonetradingcompanythatoperatesinSerbia.Themodelfordistribution centers rankingbasedonPrincipalComponentAnalysis (PCA) is tested.The resultsshowgreatapplicabilityoftheproposedapproach.

Keywords:failures,distribution,logistics,PrincipalComponentAnalysis

1.INTRODUCTION

In order to survive on the market and achieve profitability, the companies need to meetcustomer requirements andperform their activities in an efficientway.Quality and efficiencyindicatorsareveryimportantforcompanies’operationsanalysis.AccordingChristopher(2002)logistics process becomesmore andmore complex andwithmuch higher levels of demands,especiallywhenrelatedtoachievingacompetitiveadvantage.Inthatmanner,companiesneedtofollowuponservicesofferedtocustomers.Veryimportantaspectisfailuremanagement.Thefailures are present in each system. Even more the most efficient distribution systems havefailures.Thesefailuresdirectlyaffectcomplaints.Servicerecoverymaybeunderstoodasasetofactivities that a company performs to resolve complaints and to change the attitude ofunsatisfied customers trying to keep them as loyal customers. Fleury et al. (2000), studyingwholesale suppliers and retailers in the Brazilian Industry, found that failure recovery isconsidered by customers, the fifth most important customer service feature. Thus, it is veryimportant toplan, followupandevaluate the failurerecoveryprocess (keycustomer servicesattribute) throughout the supply chain and not only in the final chain stage when there is acontactwith the end consumer (Flores andPrimo, 2008).Only if systematic effort ismade tomonitor,mitigateandeliminatefailuresthateffectcomplaintsindistributionprocesscompaniescan keep customers. There are numerous quality and efficiency indicators in logistics. Theultimate goal of each company is customer satisfaction. Satisfied and loyal customersmean asecure income for the company. On the other side unsatisfied customers and customer’scomplaints create additional costs. Quality indicators are rarely used for ranking distributionsystems (Andrejić et al. 2013). In this paper, more attention is paid on logistics failures in * m.andrejic@sf.bg.ac.rs

distribution process. The next section describes the methodological approach for failuresidentificationandcorrection.Thethirdsectionisdevotedtothefailureanalysisofdistributioncenters of trading company. In the same section model for ranking distribution centers isproposed.Concludingremarksandfuturedirectionsaredescribedinthelastsection.

2.METHODOLOGICALAPPROACHFORFAILURESIDENTIFICATIONANDCORRECTION

Failures in the transport,warehouseandothersubsystemsrepresentquality indicatorswhichmay be the cause of dissatisfaction and complaints of the customer. In distribution processesandlogisticsingeneraltherearedifferenttypesoffailures.Theyarecharacterizedbydifferentparameters: time and place of origin, caused consequences, size, etc. It is very important toidentify failures as soon as possible, before they pass on and increase the consequences. It isnecessary to correct observed error in the shortest period. This paper proposes newmethodologicalapproachforfailuresidentificationandcorrection.Thebasicstepsare:

Step1:Distributionprocessdecomposition Step2:Identificationofpotentialplaces(systems)offailuresappearances Step3:Failuresidentification(accordingdefinedsystems) Step4:Identificationofresponsibility Step5:Failurescausesdetection Step6:Assessmentoftheconsequences Step7:Definitionofpreventiveandcorrectivemeasures

Distributionprocessdecompositionisshownonfigure1.Thefirstprocessisproductorderingwithtwobasicaspects.Thefirstaspectofproductorderingisorderingfromsuppliers,whilethesecond is customerordering.All activities in thisprocess relate to information flow.Thenextprocess iswarehousing.Activities in thisprocessmaybedivided in twosegments. In the firstsegment are activities of goods receiving, putting away and storage while in the second areactivitiesoforderprocessingandpreparingfordelivery.Warehousinglargelydependsonspeedof information exchange. Order picking process is the crucial process in warehouses. Thefollowing is the process of packaging. This process is realized through merging goods fromdifferent segments, forming transport units, goods inspection, aswell as the loading goods invehicles.Packagingis indirectrelationwiththeorderprocessinganddistribution(transport).Thetransportiskeyprocessesintheproductdistribution.Thisprocesslargelyaffectscustomersatisfaction. The process that is related to allmentioned processes is inventorymanagement.Thelastprocessisunloadinggoodsintheretailstores.Accordingtothepointoforiginfailurespotential places of failures appearance can be divided into: failures in logistics systems(failuresinDC,factoryfailure,supplierfailure,etc.);failuresinlogisticssubsystems(failuresinwarehouse, failure in transport, failure in inventories, freight forwarding, reverse logistics, IT,etc);failuresinlogisticsprocess(orderprocessingfailure,orderpickingfailure,shippingfailure,etc.); failures in logistics activity (labeling mistake, picking failure, transport failure, forkliftdriverfailure,truckdriverfailure,etc).Thenextstepisfailuresidentification.Warehousesanddistributioncentersareintegralcomponentsofsupplychains.Processesinwarehousesareverylabor and cost intensive. In that manner there are different and numerous failures inwarehouses.Wang (2002) found five causes of complaints inwarehouse:wrong label (18%),later or wrong delivery (45%), damaged cargo (18%), input error (12%) and system errorrespectively (7%).Transport failuresgreatly affectdeliveryprocessandcustomercomplaints.Therearedifferentaspectsoftransportfailuresinliterature.AccordingWang(2002)therearefivemain transport failures. Research shows thatwrong calculation (10%),wrong ormissedprocess (7%), input error (7%), delayed shipping (20%) and document error (56%) areresponsible forthecomplaints in freighttransportsector. Inventorymanagementhasastronginfluenceonthefailuresinwarehousesanddistributioningeneral.

Figure1.Distributionprocessdecomposition

In that sense there are different causes of complaints inwarehouses that relate to inventorymaintenance and control: product identification, stock‐outs, contaminated products, obsoletemerchandise, wrong customer’s goods, back orders, percent of demand filled, omission rate,number of incomplete orders, minimum order size, minimum frequency allowed, inventory

YES

NO

Makingcorrections

Goodsstorageonwarehouselocation

Warehousing

Inventorymanagement

PackagingReceivingtransport

tasks

Informationaboutrealiseddelivery

Transport

Ordercreating

Sendingordertosupplier

Creatingreceivingdocument

Entringorderininformationsystem

Creatingreceivingstoragereport

Receivingorder

CreatingorderinIS

Itemseparation

Creatingacount

OrderentryinWMS

Supplierregistration

Unloadingofgoods

Goodscontrol(quantity,quality,expirationdate,

damage)

Goodslabeling

Receivingandpraparationorderpicking

tasks

Placingofgoodsinshippingzone

Assigningthetaskorderpicking

Routing

Extractionandputtinggoodsonpallets

Orderfinished

YES

NO

Merginggoodsfromdifferentsegments

Packaging–creatingshipmentunits

Sendinginformationaboutshippinggoods

Goodsextraction

Fillingorderpickinglocations

Inventoryalignment

Ordering

YES

NO

Creatingstaskforfillingthestorage

locations

Informationaboutgoodsrelocation

Confirmationmovingatorderpickinglocation

Taskforinventoryalignment

Confirmationofalignment

Ordernecessary

Tasksdefinition

Tasksassignment

Routing

Goodsloading

Vehiclesending

Distribution

Unloading

Sendinginformationaboutgoodsinspection

Informationsaboutloading

Goodsanddocumentinspection

backup during promotions, new product, introductions and competitive tests. There aredifferent criteria for identification of failures in order – picking process. According place ofidentification thereare internal (inhouse) andexternal (outside) failures.External failures inthemost cases cause customer complaints. In the literature there are fourbasic categories offailures: typing failures (addition, confusion, etc.), failures in amount (shortage, excess, etc.),omission failures and condition failures (damage, lackofpackaging, labeling).The failure ratedepends of type of order picking system: pick‐by‐voice (0.08%), voucher (0.35%), labels(0.37%), pick‐by‐light (0.40%), mobile terminals (0.46%), mob. terminals + labels (0.94%)(Lolling 2002). Identification of responsibility and cause is the next step in the proposedmethodology.Warehousesareresponsiblefor55%ofthetotalfailuresindistributionprocess,while35%ofthefailuresrelatestotransport.Togethertheystandforaround90%ofthetotalfailures in distribution logistics. According causes there are four main distribution failures:methodcaused failures,management caused failures,processcaused failuresandmancausedfailures. It is very important to identify the responsibilities of employees for each failure.Assessment of the consequences is very important step. The major causes of customercomplaints are failures in logistics subsystemsandprocesses. If failures aredetectedon time,theywillnotcausecustomercomplaints.Eachfailuremakesitscostsbutthefailuresthatcausecomplaints are more expensive. In the distribution process, not all failures have equalimportance.Forexample, themostdamages inwarehousecannot causecustomercomplaints,because they are detected before delivery. On the other side bad inventorymanagement cancauseshortexpirationdatewhichmayresultsincustomercomplaints.Therearealsofailuresinorder picking process like shortage/excess or articles mix‐up that cannot be detected inwarehouse. They can cause customer complaints. Definition of preventive and correctiveactions is the last step in theproposedmethodology.Thereareproblems in thewarehousingprocesswhensuppliersupplies thegoodsof lowqualityandshortexpirationdate.Oneof thebasicstepsistodefinethelevelofqualityanddimensions(specificchecklists)ofeachunitofgoodsforeachsupplier.Arelativesmallnumberofemployersinthisprocesslimitthelevelofcontrol.Puttingaway isvery importantactivity inwarehouse.A largenumberofmistakesaregenerated in this process. In real systems, order pickers realized this activity. Frequentlyrelocationoforderpickersfrompickingtoputtingawayprocessgreatlyaffectstheoccurrenceoffailuresandreducingthelevelofcustomerservice.Theyrealizethisprocesswithinsufficientattention.Assignmentofsmallernumberofworkersthatwillrealizeonlyputtingawayprocessshould reduce failures tominimum. Inappropriate organization of spacemay affects failures.Managersinwarehouseoftenhavethegoalofminimizingthespacefororderpicking. Oneofthemainaimistoreducetheeffortintheorderpickingprocess.However,alargenumberofsimilaritems at very short distances can cause failures. Order picking process is work and laborintensive process. Failures may be reduced if the order pickers strictly follow informationsystem,anddonotmakedecisionsalone.Likeintheprocessoforderpickingprocessthesamesituation is in the order processing, packaging and loading. It is very important to assignworkersforparticularprocesses.Reducingfailuresintransportreferstothereductionoflossesin money, time, and users that are caused by theft, damage to goods, delays in delivery.Warehouse failures often are transferred in transport process. Failures in transport can bedecreasedwith good organization and process planning. Delivery delays can be overcome bybettermotivationofdriversandcontrolofthemovementofvehicles,aswellasgoodplanningroutes andpredictionsof traffic congestion.Theft canbepreventedwithmodern systems forprotectionofthecargospace.Systemsofdriverrewardandpunishmentcanadditionallyreducethenumberoffailures.Failuresininventorymanagementdirectlyaffectthewrite‐offofexpiredgoods,whichcreatessignificantlossesintheobservedcompanies.Afailureintheinventoryisalso the lack of goods required by the user. Failures can be overcome on different ways:inventory monitoring with advances information systems, definition of delivery priorityaccordingexpirationdate,morepreciseestimationof theexpecteddeliveriesof suppliersandexpected demands of customers. Video monitoring is one of the basic systems of protectionagainsttheft,etc.

3. BENCHMARKING LOGISTICS FAILURES IN DISTRIBUTION PROCESS OF TRADINGCOMPANY

Asmentionedbeforeinthispartofthepaperfailuresin13distributioncenters(DCs)oftheonetradingcompanyareanalyzed.AsshowninTable1observedcompanymonitorsnineindicatorsthatrelatetoquality:totallosses,failuresinwarehouse,failuresintransport,writeoffexpiredgoods,totalvalueofcomplaints,numberofcomplaints,complaintarticles,approvedcomplaintsand not approved complaints. Managers also monitor other indicators like number ofemployees, number of vehicles, pallet places, demands and turnover. Complaints caused bylogistics failures represent multiple losses logistics services providers. The first are costs offailurecorrection.Afterthataretimelosses,andsometimesthemateriallossesifthegoodsarepermanently damaged or lost. Additionally, losses caused by customer’s dissatisfaction canappear.Thevalueofcomplaints in total turnover inaveragecountsabout0.2%,whichat firstglancedoesnot represent a significantpercentage.However, complaints are significant lossesfor companies. Received complaints are analyzed, checked the causes, time of occurrence,executors,value,etc.Basedonallthis,adecisiononwhetherthecomplaintwillbeapprovedornot. According values in Table 1 in analyzed distribution centers 60% of complaints areapproved.Theanalysisoftheobservedsampleshowsthatthenumberofcomplaintsdependsonthetypeofcustomers.Thedistributioncentersthatsupplyexternalcustomersapprovedabout77%of total complaints.On theotherhand, thedistribution centers that supply retail outletsthat are owned by the same retail chain only 42% of complaints are approved. This can beexplained by the fact that the company focus is on the external clients (easier approvecomplaints), but also the fact that retailers, as well as members of the same retail chain,frequently (easier) complain. As can be seen monitored DCs have different combinations ofinput and output variables. The aim of this paper is the ranking of DCs according fourteenvariableswithspecialemphasisonfailures.Therearedifferentapproachesforrankingdecisionmaking units in the literature. Considering relatively large number of indicators and therelatively small number of decision units in this paper PCA approach for ranking DCs isproposed.PCAapproach forrankingDCshasseveralsteps(PetronandBarglia,2000):Step1:Data normalization and reorientation (failures ‐ negative outputs); Step 2: Definition ofindependent output/input ratio (di);Step3: ExtractingPrincipal Components (PCs) for alldi ;Step4:Foreachdideterminationofcorrespondingweightcoefficientwi;;Step5:MultiplyingeachcoefficientbythevalueofthecorrespondingvariablediforeachDCtogetthefinalscore;Step6:Finalrankingbasedontheobtainedscores.RankingresultsofobservedsetofDCsareshowninTable2.

Table1.Descriptivestatisticsof13DCs

Indicator Average St.Dev.Inputs

Employees(number) 53,54 31,04Numberofvehicles(number) 8,77 7,06Palletplaces(number) 2974,49 1552,43Demands(number) 3486,46 2142,25

OutputsTurnover(106m.u.) 182,78 107,49Totallosses(103m.u.) 589,63 436,92Failuresinwarehouse(103 m.u.) 129,16 107,06Failuresintransport(103m.u.) 222,97 197,63Writeoffexpiredgoods(103m.u.) 169,57 142,55Totalvalueofcomplaints(103m.u.) 360,21 343,04Numberofcomplaints(number) 169,00 137,30Comlaintarticles(number) 350,38 441,70Approvedcomplaints(103 m.u.) 214,97 205,28Notapprovedcomplaints(103m.u.) 145,26 258,97

Table2.Rankingscoresof13DCs

DC PCASCORE Rank

DC11 302,01 1

DC10 180,80 2

DC5 114,09 3

DC12 83,31 4

DC1 16,56 5

DC3 15,63 6

DC9 11,51 7

DC13 10,47 8

DC6 8,77 9

DC4 7,04 10

DC8 4,91 11

DC2 2,88 12

DC7 0,95 13

Basedontheresults itcanbeconcludedthattheDC11hasthebestcombinationof inputandoutput variables. This means that this center with a relatively small number of resourcesimplementedrelativelylargeturnoverwithaminimumleveloffailures.IncontrasttoDC11,DC7hasthesmallestrankingscore,whichcanbeexplainwithrelativelysmallturnoverandlargeleveloffailures.

4.CONCLUSION

This paper identified the basic steps for identification and correction of failures in thedistribution process. The importance of failure elimination and increasing logistics servicequalityisequallyimportantforcustomersandproviders.PCAapproachforDCsrankingisusedinthispaper.Theanalysisoftheresultsconfirmedtheconvenienceofitsapplication.Therearepapers in literature that deals with efficiency in logistics. On the other side, there are alsoapproaches that investigate failures in logistics and supply chains. However, to the best ofauthorknowledge, there isa lackofpapers in literature that investigate relationshipbetweenefficiencyandfailuresinlogistics.Inthefutureresearchitisnecessarytodevelopnewmodelsformeasuring and improving quality efficiency. The futuremodelsmust include new qualityindicators,withthespecialemphasizeoffailures.

ACKNOWLEDGMENT

This paper was supported by the Ministry of Science and Technological development of theRepublicofSerbia,throughtheprojectTR36006,fortheperiod2011–2014.

REFERENCES

[1] Andrejić,M., Bojović, N., Kilibarda,M., (2013). Benchmarking distribution centres usingPrincipal Component Analysis and Data Envelopment Analysis: a case study of Serbia,ExpertSystemswithapplications,40(10),pp.3926‐3933.

[2] Christopher, M. (2002). Logistics and Management of the Supply chain: strategies forreductionofcostsandimprovementofservices,SãoPaulo:Pioneira.

[3] Fleury, P.F., Wanke, P., and Figueiredo, K. (2000) Logistics Enterprise: the BrazilianPerspective,1ª.Edition,SãoPaulo,Atlas.

[4] Flores, L.A.F.S., Primo,M.A.M., (2008). Failure RecoveryManagement in Performance oflogistics Services in a B2B Context: A Case Study using the 3Pl Perspective, Journal ofOperationsandSupplyChainManagement1(1),pp.29‐40.

[5] Lolling, A. (2002) Fehlerraten verschiedener Kommissionersysteme in der Praxis. In:Zeitschriftf¨urArbeitswissenschaft56,pp.112–115

[6] Petroni, A., Braglia, M., (2000). Vendor selection using principal component analysis,JournalofSupplyChainManagement,36(2),pp.63‐69.

[7] Wang, J. (2002).How to implementCustomerRelationshipManagement system in thirdpartylogisticscompanies,Masterthesis,NationalInstituteofSingapore.

PERFORMANCEINDICATORSFORPROFESSIONALDRIVERS’EVALUATIONINSUPPLYCHAIN

MilanCvetkovića*,VladimirMomčilovića,BrankaDimitrijevića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: Contemporary road vehicles are equippedwith state‐of‐the‐art devices and on‐boardcomputersthatrecordparametersmainlyfromvehicleconditiondiagnosticsandtheiroperation.Along with the growing use of business information systems and with the implementation oftelematicssystemsonvehicles,largeamountofdriveractivitydatabecomesavailabletothefleetandtransportmanager,whichenablesandeven facilitatesthe insight intodrivers’activitiesandtheir evaluation. Naturally, besides the above mentioned electronic data, the company stillmaintainssomepaper‐orcomputer‐basedregistersanddocumentsallowing further insight intodriverbehaviour.Suchcompleteoverviewofdriveractivity fromtheaspectofdrivingbehaviour,vehicletreatmentandotherbusinessandtransportrelatedperformances,aswellasthepossibilityof these parameters’ evaluation, ultimately leads to a better human resource management,lowering transportactivityandvehiclemaintenancecosts, increasing trafficsafetyand loweringvehicleanddrivernegativeenvironmentalimpact.Keywords:drivers’performance,on‐boarddevices,drivingskills

1.INTRODUCTION

Professional drivers of road motor vehicles represent one of the most important factorsregardingvehicleoperationcosts,transportationanddeliveryservicequalityandinthelongrunthecompanyimage.

Automotive industry is involved in increasing vehicle fuel economy by introducing moderntechnologytolowerharmfulgasemissionsanddecreasefuelcosts,aswellason‐boarddevicesallowing vehicle and driver operations monitoring and analysis. In general, the latestimplemented technology invehicles isnotsupportedbyadequatedrivers training,on theonehand compulsory driving licensing and on the other continuous and periodic knowledgeupgrade.Thisisespeciallytrueforcommercialvehiclesinvolvedinthesupplychain,wheretherelated driver influence is more important having inmind their higher vehicle value, annualmileageandfuelconsumption.

The objective of this paper is to present differences betweenprofessional drivers through anoverviewandsystematizationofoveralldriverperformanceindicators(PIs),basedonliterature,today’s practices and specify those indicators that could be identified andmeasured byusingmodern vehicle and telematics technologies applied in the supply chain. This certainlyrepresentsthefirstandthemostimportantstepinmodels’buildingwhichwouldbelaterused

* m.cvetkovic@sf.bg.ac.rs

fordriverselectionwhileemploying, theirrankingconcerningrewarding/penalizingorevennoticingindividualweaknessesthatcouldbecorrectedbyacustom‐madepersonalizedtrainingprogram.

The rest of the paper beginswith a review of driver performances recognised as relevant inliterature,aswellastheirmeasurementoptions,potentialeffectsand limitations.Nextsectiondealswith PIs obtained from dataset collected as a result of driversmonitoring over certainperiod inSerbian logistics companyusinganon‐boarddevice.Finally concluding remarksarepresentedalongwithdirectionsforfutureresearch.

2.LITERATUREREVIEWONPROFESSIONALDRIVERSPERFORMANCESEVALUATION

Supply chainare complex systemswheredriversare required topossessdifferentknowledgeand skills except from driving. Those requirements depend on their interaction with othersupplychainactors.Thereoforiginatesthefact thatdriver’sperformanceusedtomeasurethequality of work is by nature multidisciplinary and roughly could be divided into technical(driving skills and other professional service skills) and nontechnical (social and other softskills).

Technicaldriverperformance,whichwillbeconsideredinthispaper,regarding:

Driving skills, involve relations between: Driver and vehicle (vehicle technologycomprehension, vehicle care and maintenance, ability to solve some unexpectedproblems, etc.), and Driver and vehicle in traffic (driving style, knowledge of trafficregulations,avoidingincidentsandcriticalsituations,etc.);

Otherprofessionalserviceskills, involveinteractionsbetween:Driverandclient(clientrequest resolution, timely appearance, client feedback, etc.), Driver and vehicle onterminal / in warehouse (loading, unloading, goods handover, paperwork,administration, etc.), and Driver and company (task realization, obeying procedures,etc.).

Many papers are dedicated to other drivers’ professional service skills, whose evaluation ismostlybasedon individualperceptionof their quality and importance.Meanwhile, thispaperfocusesondrivingskillsbasedpurelyonmeasureddatawithanattempttoobtainasexactandobjectiveperformancesaspossible.

Certain authors in their papers and studies highlight the importance of driving performance,presenting its measuring methodology and/or determine the acceptance thresholds, whichcouldbeaswellperceivedasdrivers’PIs.Those indicatorsaremostly focusedonquantifyingroadsafetyandfuelconsumptionand/orharmfulorgreenhousegasemissions.

KimandChoi(2013)statedthefactthatthespeedisamainfactortoestimatemacroamountoffuel consumption and emission of vehicles. However they have focused on acceleration anddetermination of critical values of aggressive acceleration influencing fuel consumption andemission significantly by testing vehicle speeds range from 10 km/h to 80 km/h consideringdifferentdrivingpatternsinurbanareas.Harietal.(2012)developedandtestedretro‐fittabledriver behaviour improvement device in realworld conditions (15 vans on over 39,000km).Thatdeviceprovidedreal‐timeaudioandvisualfeedbacktothedriverstoimprovetheirdrivingstyle. Analysis showed that fuel savings of an average 7.6% were obtained as a result ofreduction in harsh accelerations and early gear shifting into higher gears. El‐Shawarby et al.(2005) evaluated the impact of vehicle cruise speed and acceleration levels on vehicle fuelconsumptionandemissionratesusingfielddatagatheredunderreal‐worlddrivingconditions.An on‐board emission‐measurement devicewas used to collect emissions of nitrogen oxides,hydrocarbons,carbonmonoxide,andcarbondioxideusingalight‐dutytestvehicle.Theanalysisdemonstrated that fuel consumptionandemission ratesper‐unitdistanceareoptimum in the

range of 60–90km/h, with considerable increase outside this optimum range. The studydemonstrated thatovera sufficiently long fixeddistance, fuel consumptionandmobile‐sourceemission rates per‐unit distance increase as the level of acceleration increases. Rolim et al.(2014)assessedtheimpactsofecodrivingtrainingandon‐boarddevicesthatprovidereal‐timefeedbackon600busdrivers'behaviourregardinghardstopsandstarts,extremeaccelerationsandbrakes,timespentidlingandspeeding.Thisanalysisconsidereddriverscharacteristics(ageandtimeworkingatcompany)andvehiclecharacteristics(busageandtype).Zarkadoulaetal.(2007)showedtheeffectsofeco‐drivingtrainingforthreedriversleadingtoadecreaseoffuelconsumptionbyupto17.8%whiletheaveragedecreaseinfuelconsumptionforallbusdriverswas10.2%.

Barret al. (2011) in their studyassessed the impactofdriver fatigueondrivingperformanceusing the naturalistic data of 42 local/short‐haul truck drivers from two companies thatparticipated in the field study that lasted forapproximately twoweeksperdriver.The truckswereinstrumentedwithcamerasanddatacollectionequipmentconsistingofsensorstomonitorvehicle parameters: speed, lateral and longitudinal acceleration, steering position, and brakepedalactivation.Predictivemodelsweredevelopedtodeterminethedrivercharacteristics(e.g.age, years of commercial driving experience, sleep quality/quantity) and external orenvironmental factors(e.g. timeofday,weather, traffic) that influencethe likelihoodofdriverfatigue occurring on the job. Twomeasures of driver performance ‐ lane‐keeping and speedmanagement‐wereevaluatedinanefforttocorrelatedriverfatigueandperformance.

However, in tasks performed by drivers within supply chain, the sole fuel consumptionindicator,thatisoftenrecurredtobyfleetmanagers,isnotsufficientlygoodindicatorsinceitisinfluenced by other factors that could not be easily determined (such as landscape, trafficcongestion,idling)andcouldevenleadtoinaccurateperceptionofdifferentdriverquality,evenin similar activities and environments (since not all parameters match). Often the driver’simpactonvehiclemaintenancecostsand loweringvehicleand itscomponents lifecycleperiodandcostsisdisregardedorneglected.In(TIAX,LLC,2009)aresetfundamentaldrivingsegmentsthatU.S.medium‐andheavy‐dutyvehicledriversshouldpayattentionto,inordertolowerfuelconsumption:speedfluctuation,enginebraking,shiftoptimizationandgearselection,idling,tirecondition and inflation, speed, cruise control, clutch control, trip planning, blockshifting/skippinggears,aerodynamics,overfillingthefueltank,andmaintenance.

Morteet al. (2013)dealswithmulti‐criteria rankingofdriversbasedon twoquantitativeandninequalitativecriteriaobtainedasdecision‐makers’impressionsonthequalityofeachdriver(outof thetotalof31drivers)usingnarroworwidescales.Theyselectedthe followingsetofcriteria:technicalknowledge,labourlegislationknowledge,accidents,fuelconsumption,abilitytosolveunexpectedproblems,timelydelivery,internalrulescompliance,customer'sstandardscompliance,information,conflictresolution,expectationfit,responsibility,andavailability.Thispaper’smainshortcomingisagreatnumberofexpert‐basedqualitativecriteriavalues,withoutacleardeterminationapproach,leadingtosubjectivityindriverevaluationandfinalranking.

Fromthepresentedreviewitisobviousthatthereisagapofacomprehensivemodelfordriverperformance evaluation and ranking. Since in previous evaluationmodels there is an issue ofindicators’determinationobjectivity,theauthorsoptedfordatacollecteddirectlyfromvehiclesensors and processed in order to determine and evaluate driver PIs, which is significantlyfacilitated by todays’ vehicles in conjunction with state‐of‐the‐art on‐board devices andtelematicssystems.

3.ON‐BOARDDEVICESINEVALUATINGDRIVERPERFORMANCES

Inorder tousevehicledata fordriverperformanceevaluation, authorshavechosena fleetofdiesel‐poweredvansofoneexpresscourierdeliverycompanyfromSerbia.Threeidenticalvans(bymake,model, and age)were equippedwith an on‐board diagnostic device, OBDMATRIX

(TEXA),which inthissetting(compression‐ignitionengine)was limitedtothefollowingsetofdata:enginespeed,vehiclespeed,engineloadandenginetemperature.Datawererecordedforaperiodof5daysateveryquarterof a second(at the largestpossible frequencyof4Hz)whilevehicleignitionwason,whichresultedinlargeamountofdatatobeprocessed.

FrommeasureddatatheauthorshavetriedtodetermineasetofrelevantandimpartialPIstoevaluatedrivers inviewofbasic fleetmanagers’objectives: fuelconsumption,roadsafetyandvehiclemaintenance.Forthatpurposeenginespeedandvehiclespeed,relevantbythemselves,wereusedasrawdata.Accelerationanddecelerationwerederivedfrominstantaneousvehiclespeed.Eachofthesevaluesareratedbasedonadevelopedscale,showninTable 9.ForeachPI,driver quality decreases with rating increase. Rating scales for acceleration and decelerationwere set according to Beusen et al. (2009) and Kim and Choi (2013) researches, along withrecommendations of fleet managers from the considered company and field experts. Enginespeed range for the rating scale was set consistent with van manufacturer’s specificationsregardingmaximumpowerandtorquetoenginespeeddiagram.Thevehiclespeedlimitvaluesaresetinaccordancewithtrafficspeedlimitsinvehicleoperatingarea.Asconsideredcompanyvansoperateoutsidemotorways,theauthorshavesettheinitialspeedlimitat80km/h(with10km/hincrement).

Table9–Driverratingscalefordifferentparametervalues

Acceleration(a) Deceleration(b) Enginespeed(n) Vehiclespeed(v)Range(m/s2) Rating Range(m/s2) Rating Range(rpm) Rating Range(km/h) Rating0<a≤1.7 0 ‐2.3≤b<0 0 0≤n ≤2600 0 0≤v≤80 01.7<a≤2.3 0.2 ‐3.4≤b<‐2.3 0.5 2600<n ≤2800 0.2 80<v≤90 0.22.3<a≤3.4 0.5 ‐4.5≤b<‐3.4 2 2800<n ≤2850 0.4 90<v≤100 0.53.4<a≤4.5 2 ‐5.6≤b<‐4.5 4 2850<n ≤3100 0.6 100<v≤110 24.5<a≤5.6 4 ‐6.7≤b<‐5.6 10 3100<n ≤3350 0.8 110<v≤120 4a>5.6 10 b<‐6.7 20 n >3600 1 v>120 10

Basedonpreviousparametersandtheirratingscales,eightPIsweredefinedandcalculatedforeachdriver:

I. Start‐up acceleration – Average acceleration rating per start‐up (start‐up consists ofinitialtensecondsofdrivingfromstandstill).

II. Start‐upenginespeed–Averageenginespeedratingperstart‐up.III. Haltdeceleration–Averagedecelerationratingperhalt(haltconsistsoflasttensecond

ofdrivinguntilvehiclestop).IV. Acceleration in continuous driving conditions – Average acceleration rating per

acceleration time of all continuous driving periods (continuous driving consists of theperiodbetweeninitialandfinal10secondsofdriving, i.e.afterstart‐upperioduntilhaltperiod).

V. Deceleration in continuous driving conditions – Average deceleration rating perdecelerationtimeofallcontinuousdrivingperiods.

VI. Engine speed in continuous driving conditions – Average engine speed rating perdurationofallcontinuousdrivingperiods.

VII. Idling–ratiobetweenidlingtime,determinedasenginerunningtime(enginespeed≠0)withsimultaneousvehiclestandstill(vehiclespeed=0),andtotalenginerunningtime.

VIII. Speeding–Averagespeedratingperdrivingtime.

Authorshaverecognized threecharacteristicdrivingperiods:start‐up,continuousdrivingandhalt.Start‐upandhaltevents,asseparatetimesequences,allowforthedriverstobeobjectivelycompared on the basis of their skills to operate vehicle during start‐up and halt. Besides,observingdrivers’behaviourduringcontinuousdrivingconditionallowstoidentifythosewhoaccelerateordecelerate aggressivelyoroperate thevehicle athigh engine speedsbeyond thestart‐stopregimewheresuchinadequatedriveractionsoccurmostcommonly.Averageratings

wereusedfordifferentcharacteristicdrivingperiodsinordertominimizetheimpactofrouteconfiguration and traffic conditions (mileage, travel time, unloading points, etc.) on driverperformancerating.

Start‐up acceleration and start‐up engine speed are related to fuel consumption and vehiclemaintenance by influencing engine, drivetrain and tire wear. Halt deceleration directlyinfluencesvehiclemaintenancebybrakewearand indirectly fuel consumptionas indicatorofinsufficient driver anticipation leading to unnecessary stop‐and‐go situations. Extremeaccelerations and decelerations in continuous driving conditions denote aggressive driverbehaviour leading to road safety issues, increased fuel consumption and eventually highermaintenance costs. High engine speeds in continuous driving conditions influence fuelconsumption,showinglackofadequateknowledgeandskillsregardinggearshiftingtechniques,which could be resolved by appropriate driver training. It could also imply that the driver isaggressive or negligent, tending tomaintain vehicle acceleration capacity. Idling impacts onlyfuel consumption suggestingdriver’s negligent behaviour. It includes idling times longer than120 seconds, in order to avoid accounting for congestion delays and waiting times at trafficlights.Speeding (drivingover thespeed limit)primarily influencesroadsafety, aswell as fuelconsumption and reflects driver awareness and complying with rules and traffic regulations.ThespeedingindicatorwouldbemuchmoreaccurateifavailablespeeddatawouldbematchedtospeedlimitsonthenetworkbyusingGPSdata.

In order to examine vehicle data utility to obtain reliable PIs for driver comparison andevaluation,theauthorsdecidedtomonitor,recordandprocessvehicleoperationdataforthreedrivers.Basedonfleetmanagers’opiniontwomoderateandoneaggressivedriverwereselectedtobe furtherexamined.Driver1wascharacterisedasmoderateandattentivedriver.Driver2hadareputationofanaggressivedriver,oftenexceedingspeedlimits.Driver3hadareputationofamoderatedriver,confident inhisowndrivingskills.AfterdeterminationofdriverPIs, foreasiercomparison,theirnormalizationwasmade(Table 10).

Table10–Normalizedpriorityofdrivers’performanceindicators

DriversStart‐up

acceleration

Accelerationincontinuousdriving

conditions

Haltdeceleration

Decelerationincontinuousdriving

conditions

IdlingStart‐upenginespeed

Enginespeedincontinuousdriving

conditions

Speeding

1 0.3101 0.3519 0.3705 0.5208 0.1865 0.1365 0.4054 0.00582 0.2294 0.2375 0.1834 0.1258 0.1375 0.2308 0.0354 0.00043 0.4605 0.4106 0.4461 0.3534 0.6760 0.6327 0.5592 0.9938

From Table 10, it is obvious that there is a clear distinction of examined drivers in all PIs.Obtainedresultsgivedetailedandmeasurableillustrationofdifferencesbetweentheirdrivingskillsandbehaviour.Veryhighdifferenceinspeedingindicatorbetweenthethirddriverandtheothers isresultofdifferencesbetweentheiroperationareas.Thethirddriveronlyoperates incentralurbanareawithspeedlimitat50km/h,opposedtoothersoperatinginsuburbanareaswherethespeedlimitof80km/hisattainable,butstillrestricted.

4.CONCLUSION

Today, in supply chain fleets too much attention is given to driver evaluation upon easilyavailableparameters,asfuelconsumption,whichinauthors’opinionisnotaparameter,butanobjective. Fuel consumption does not represent a reliable comparison parameter, as besidesdriving skills, it also depends on other conditions resulting from transport task,loading/unloadingtechnology,operatingconditions,payloadutilisationandanumberofother,oftenunidentifiableindicators.

Inthispaperthedriverevaluationisattemptedbymoreobjectiveindicatorsofferedbymodernvehicle technologies. Based on a simple on‐board device and a small number of recordedparameters a larger set of PIs was derived that are not correlated among each other andtherefore justify their application allowing high‐quality and comprehensive perception ofdriving skills. For data processing and results analyses MS Excel was used, which causedimportanttimedelayswithavailabledatasetimplicatingtheneedforamoreadapteddatabasesoftwareforlargerdatasets.AsidefromdescribedPIsitispossibletogenerateotherindicatorsobtained from used or more complex on‐board devices. Likewise, for driver evaluation it isessential to involve other quantitative indicators accessible to fleetmanagers, such as: driversafety and accident costs records, different test results, vehicle maintenance costs asconsequenceof driver’s behaviour, but also somequalitativePIs related tootherprofessionalservice skills. It was earlier explained that PIs for driver evaluation are related to multiplelogistics company objectives, hence next research step is creation of drivers’ evaluation andrankingmodelincludingallmentionedPIsandobjectives,theirrelationsandimportance.

ACKNOWLEDGMENT

ThispaperisbasedontheresearchwithintheprojectsTR36006andTR36010supportedbytheMinistryofEducation,ScienceandTechnologicalDevelopmentoftheRepublicofSerbia.

REFERENCES

[1] Barr,L.C.,Yang,C.Y.,Hanowski,R. J.,Olson,R. (2011).Anassessmentofdriverdrowsiness,distraction,andperformanceinanaturalisticsetting,ReportNo.FMCSA‐RRR‐11‐010.

[2] Beusen,B.,Broekx,S.,Denys,T.,Beckx,C.,Degraeuwe,B.,Gijsbers,M.,Scheepers,K.,Govaerts,L.,Torfs,R.,Panis,L.I.(2009).Usingon‐boardloggingdevicestostudythelonger‐termimpactofaneco‐drivingcourse,TransportationResearchPartD:TransportandEnvironment,14(7),514‐520.

[3] El‐Shawarby,I.,Ahn,K.,Rakha,H.(2005).Comparativefieldevaluationofvehiclecruisespeedand acceleration level impacts on hot stabilized emissions, TransportationResearch Part D:TransportandEnvironment,10(1),13‐30.

[4] Hari, D., Brace, C.J., Vagg, C., Poxon, J., Ash, L. (2012). Analysis of a driver behaviourimprovement tool to reduce fuel consumption. In: 2012 1st International Conference onConnectedVehiclesandExpo,ICCVE2012,2012‐12‐12‐2012‐12‐16,Beijing.

[5] Kim,E.,Choi,E.(2013).Estimatesofcriticalvaluesofaggressiveaccelerationfromaviewpointoffuelconsumptionandemissions.In:2013TransportationResearchBoardAnnualMeeting,2013,Washington,DC.

[6] Morte, R., Pereira, T., Fontes, D.B. (2013). MCDA applied to performance analysis andevaluation of Road drivers: A Case Study in theRoadTransport Company. InBS#8217; 13ThirdInternationalConferenceonBusinessSustainability.

[7] Rolim,C.,Baptista,P.,Duarte,G.,Farias,T.,Shiftan,Y.(2014).QuantificationoftheImpactsofEco‐driving Training and Real‐time Feedback on Urban Buses Driver's Behaviour,TransportationResearchProcedia,3,70‐79.

[8] TIAX, LLC. (2009). Assessment of Fuel EconomyTechnologies forMedium‐ andHeavy‐DutyVehicles.Cambridge,Massachussets:ReporttotheNationalAcademyofSciences.

[9] Zarkadoula,M.,Zoidis,G.,Tritopoulou,E.(2007).Trainingurbanbusdriverstopromotesmartdriving: A note on a Greek eco‐driving pilot program, Transportation Research Part D:TransportandEnvironment,12(6),449‐451.

THEROLEOFALLOCATIONANDORGANIZATIONOFCUSTOMERSERVICEDIVISIONSASANELEMENTOFMARKETCOMPETITIVENESSOFTELECOMUNICATIONOPERATORS

MarijanaVukašinovića,ZoricaLeposavića

aTelekomSrbijaAD

Abstract: This paper deals with the impact of customer service divisions allocation andorganization to theservicesand telecommunicationoperatorspositioningatcompetitivemarketconditions. The estimation in this paper is based on an analysis of usersmovement level incorrelationwithconcreteallocationandcustomerservicedivisionorganizationasamarketplayerstrategy to respond to different customer demands and expectations so as tomake themarketpositionofoperatorsstrongerandmorecompetitive.

Keywords:Allocation,organization,services,customer,market.

1.INTRODUCTION

Formany organizations, contact centermanagement system and CRM (customer relationshipmanagement)representsthefirstlineofcontactwithcustomers.Assuch,theyprovideauniqueopportunity to increase customer satisfaction level and develop long‐term relationshipsbetween customers and the operator in order to meet the rising user expectations. For thispurpose,organizationsmustdevelopaunifiedviewoftheuserthatisoftenharderthanitlookslike.

Contactcenterisaunitoforganizationfocusedonraisingthequalityofservicesandcustomersatisfaction level. The employees in contact center as a first line of service providing arerepresentativesof thecompany.Twoveryimportantparameters–retentionandchurnareindirectcorrelationwiththequalityoftheirwork.

Philip Kotler points out that each organization needs to control and improve the level ofcustomersatisfaction.Thereareseveralreasonswhyitismoredifficulttoacquirenewcustomerthantokeeptheexistingone:

The price of acquiring a new customer is five to ten times higher than the price ofkeepingtheexistingone,

Theaverageannualchurnincompaniesrangesfrom10%to30%. Churnreducinginlevelof5%reflectstoprofitincreasinginlevelbetween25%to85%

accordingtothespecifictypeofindustry. Therateofprofitpercustomerisinproportiontothetimeofhisretention.

Contact center is a central place of implementation of external communication and businessstrategy of the company. The fact is thatmajority of customers rather call the contact center

thengettotheshop.Thecorrelationbetweenashopandacontactcenterissuchthatallthatcanbe done in the shop can be realized by calling the contact center. Allocated shops cover asignificantpartoftheterritory,butcontactcenterprovidesservicesthroughthewholeterritoryirrespectiveofauserlocation.Previously,contactcenterswereorganisedasallocatedunitsonageographical basis serving customers on a particular location. Development of technologybrought new possibilities for completely different work organization of contact centers. Themainadvantageofanewmodelorganizationisreflectedintheimplementatonofaprincipleinwhich free agent is available for a user regardless of the allocation of both subjects – contactcenter and a customer. The further estimation will be done through a case study includingTelekomSerbiacontactcenter.

2.RESEARCHMETHODS

Thesamplesizeinthisstudycaseare240callcenteragentsin8allocatedcallcenters.Inordertoreviewtheentireprocessofcommunicationbetweencontactcenteragentandtheusers,thefollowingmethodshavebeenused:

2.1Themethodofcontentanalysis

The method of content analysis involves analysis of documents, internal documents andrelevant reports related to the communication with users. Present subjects are related tostatisticalreportsofquantitativeandqualitativeparametersofoperator’scontactcenterworkonamonthlyandquarterlybasis,andtheseare:

Quantitativecriterionforcallduration–ifthecalldurationexceeds11%oftheadoptedvalue indicator then theeffect isbelowexpetations. If the calldurationdoesn’t exceed10% of the adopted value indicator the effect is within the expected limits. Thepecentageof avarage call durationwhich isbelow99%of theadoptedvalue indicatorshowstheeffectwhichisaboveexpetations.

Quantitativecriterionfortheaveragenumberoftheansweredcalls–ifthepecentageofthe answered calls is less than 89% of the adopted value indicator then the effect isbelow the expetations. The percentage between 90% and 110% related to the valueindicatorshowstheeffectatthelevelofexpetationsandallthatexceeds111%isabovetheexpectations.

Qualitative criterion for evaluating the flow of conversation – refers to the art ofinterviewingandprofessionalismincommunication,andtheyaremeasuredbythecallresponsespeedaswellasbythesuccessfulimplementationofuserrequirements.

2.2Thecomparativemethod

Thecomparativemethod involvescomparing thespecificparametersof contact center in thesituation before and after the introduction of a new technology i.e. new contact centerorganizationofworkinanobservedperiodofthreemonthsbeforeandthreemonthsaftertheintroductionofchanges.

2.3Themethodofcasestudy

Themethodof case studydealswith the researchof communicationbetweenTelekomSerbiacontactcenterandcustomers.

2.4Statisticalmethod

Statisticalmethodinvolvesdataprocessingandtheirpresentationintablesandgraphs.

3.HIPOTHESIS

Inaccordancewiththedefinedobjectexplorationwasappointedhypotheticalframeworkwhichconsistsofthemainhypothesesandindividualthatwillbepartiallyorfullyconfirmed.

3.1Mainhypothesis:

Astheorganizationofcontactcenterisonahigherleveltheneedforallocatedshopsislesser.

3.2Specifichypotheses:

As the possibility to perform calls overflow to the first available agent, regardless ofphysicallocation,isbigger,theusersatisfactionlevelishigher.

Astheusersatisfactionwithservicesprovidedishigher,theuserswillbemorewillingtoacceptnewservices.

Asthetimeofresolvingthecustomercomplaintisshorter,thelevelofusersatisfactionishigher.

Astheusersatisfactionisonahigherlevel,themarketpositionofoperatorisstronger.

4.THEOBSERVEDPARAMETERS

In order to perceive the level and type of the impact of new implementation of workorganizationonthequalityofservice,inanalysishavebeenusedmeasurableparameterswhichidentifythequalityoftheservicesprovided.Theusedparametersareasfollows:

4.1Percentageofansweredcalls

Percentage of answered calls represents the number of answered calls in relation to totalnumberofcallstothecontactcenter.TheresultsshowninFigure1.indicatethatthepercentageof answered calls is 36% higher in observed period after implementation of unified contactcenterthaninobservedperiodwhencontactcenterwasallocated.

Figure1.Parametersbeforeandafterunifiedcallcenters

4.2Servicelevel

Servicelevelisthepercentageofansweredcallswithin20seconds.Theresultsofthisanalyzedparameter are shown in Figure 1. It can be seen that the service level is 53% higher in the

observedperiodaftertheimplementationofaunifiedcontactcenterthanintheobservedperiodwhencontactcenterwasallocated.

4.3Waitingtime

Waiting time represents the time that a user spends waiting for the first available agent toanswer the call.Theresults inFigure2. indicate that thewaiting time is 50% shorter in theobserved period after the implementation of unified contact center than in observed periodwhencontactcenterwasallocated.

Figure2.Waitingtime

4.4Percentageofcomplaintsonthequalityofservices

Itrepresentsthenumberofcomplaintsonthequalityofservicesinrelationtothetotalnumberoftheansweredcalls.TheresultsofanalysedparametersareshowninFigure1.Thepercentageofusers’complaintsis0,7%smallerintheobservedperiodaftertheimplementationofunifiedcontactcenterthanintheobservedperiodwhencontactcenterwasallocated.

4.5Complaintsresolvingtime

Complaints resolving time is the time needed for resolving of complaints. The results of thisparameter are shown in Table 1 where it can be seen that the needed time for resolving ofcomplaints is22%shorter in theobservedperiodafter the implementationofunifiedcontactcenterthanintheobservedperiodwhencontactcenterwasallocated.

4.6Talktime

Talktimeisdurationof conversation betweenanagentand auser. Theresultsofanalysisofthisparameter indicate that this parameter remained the same in both models of contact centerorganization.

4.7Percentageofretention

Percentage of retention in the observed period represents the percentage of users whocontinued to use the service after the application for suspension of it in relation to the totalnumberofuserswhohaveapplied.

36

18

Waiting time to answer in seconds, allocated contact centers

Waiting time to answer in seconds, unified contact center

4.8Percentageofchurn

Percentageofchurniscorrelationbetweenthenumberofuserswhocanceledserviceandthetotalnumberofuserswhohaveappliedforsuspensionofservice.

5.RESEARCHRESULTS

ResearchresultsareshownintheTable1.

Table1.Parametersandresults

Parameters allocated unifiedThepercentageofansweredcalls 39% 61%

Servicelevel 42% 89%

Waitingtimetoanswerinseconds 36 18

Percentageofcomplaints 2,3% 1,6%

Timeresolutionofcomplaintsindays 3,2 2,5

Durationoftheconversation,inseconds 245 236

Percentageofretentionduring3months 35% 56%

Thepercentageofcustomerswhocanceledservice 65% 44%

Havinginmindthatfirst individualhypothesisrelatestothecorrelationbetweenthespeedofthe response call and customer satisfaction, the values shown in the tablewhich refer to theparametersoftheservicesprovidedsupportsthesethypothesis.

Second individual hypothesis set in this research relates to the level of customer satisfactionwiththeservicesprovidedandhiswillingnesstoacceptnewservices.Havinginmindtheresultsof the parameters from the table which relate to customer retention and customers whocancelledtheservices,itisevidentthattheaforementionedhypothesisislargelyconfirmed.Itisexpectedthattheretainedcustomerswillbepotentialusersofnewservices.

Thirdsethypothesisthatrelatestothecorrelationbetweenthetimeofcomplaintsresolvingandalevelofcustomersatisfactionistotalyconfirmedbytheresultsgivenwhichcanbeseeninthevalueofparametersofcomplaintsresolvingandinthevaluerelatedtoretentionandchurn.

Forth set hypothesis that represents correlation between a level of customer satisfaction andoperator’smarket position is confirmed through all parameterswhich relate to the quality ofservicesprovidedaswellasthroughthevaluesofparametersrelatedtochurnandretention.Asthe parameters of the quality of services provided are better, the churn is lesser, and theretentionofthecustomersisbiggerwhichdirectlyreflectstotheoperator’smarketposition.

Having in mind that all individual hypothesis are confirmed, thus the main hypothesis isconfirmedanditisevidentfromtheinformationintheTable1.

All of the parameters above speak in favour of unified contact center. The results of analysisrepresentthatthepercentageofretentionishigherandthatthepercentageofchurnislesserinconcept of unified contact center. Those are very important indicators of operator’s marketpositionandcompetitiveness.

6.CONCLUSION

There isnodoubt that there is a strongcorrelationbetween thequalityof staff comunicationand service users and a level of user satisfaction with services provided. For this reason its

permanent improvement is necessary primarily through the selection of appropriateorganizationalmodels.

The aim of this paper was to indicate the importance of selecting an appropriate model ofcontact center organization for implementation of external communication and businessstrategyofthecompany.

Comparingtheresultsofanalyzedparametersduringthetimeofallocatedcontactcenterwithobservedwork period of unified contact center the conclusion is that the implementation ofnew model of organization gives much better results. The direct effect of new modelorganizationreflectsinthefactthattheuserswhoarefarfromshopshavetheopportunitytorealizetheirneedsinthesametimeperiodandatthesamelevelofqualityastheuserslocatedneartheshop.

Thesecondimportanteffectisreflectedintheefficiencyoftroubleticketingandconsequentlyintheir resolving. The observed case study leads to the conclusion that in case of efficientorganization of unified contact center there is noneed for allocation of units on geographicalbasis. Appropriate organization of contact center including ongoing staff training andmarkettrendsmonitoringareofgreatimportanceforretentionofexistingandacquisitionofnewusersaswellasforpreservingandincreasingofuserbasisastherealmarketpowerofoperatorwhichdetermineshismarketposition.

Inadditiontoconsideringofconcretecase,theobjectiveofthisstudywastoidentifyandpublishadditional scientific understanding about interdependence of business communication andsatisfaction of a user with the services provided through description of all phenomena andprocessesrelatedtotheprocessofcommunicationbetweenanorganizationandserviceusers,classificationofconcepts,phenomenaandprocessesrelatedtotheprocessofcommunication,aswellasthroughdescriptionofmutualinfluences,connectionsandrelationshipsintheprocessofcommunicationbetweenanorganizationandserviceuser.

Havinginmindthatinthispapertheresearchisorientedtotheoverviewofparameterswhichmaintain the impact of communication on the achievement of business policy, the impact ofcommunication on customer satisfaction and finally the impact of a user satisfaction on thecompany image, present research may have even wider social contribution because it couldserveasabasisforfurtherresearchandapplicationinotherorganizationsthatwanttoimprovebusiness communication and operations of their organizations, and thereby strengthen itsmarketposition.

REFERENCES

[1] Ruth N. Bolton, (2008). Business & Economics, The Maryland School, University ofMaryland.

[2] KotlerP.,Armstrong,G.(1994).PrinciplesofMarketing,PrenticeHall,NewYork.

[3] Milijana M. Zlatković, (2013). Survey of customer satisfaction: The example of freightdistributioncenter,FacultyofTransportandTrafficEngineering,UniversityofBelgrade,Belgrade,Serbia.

[4] Zorica Leposavić, (2013). Master thesis: Influence of business communication onsatisfactiontelecommunicationsusers,UniversityMegatrend,Belgrade.

PartVIII SUPPLYCHAIN

MANAGEMENTANDREVERSELOGISTICS

AMODELINGAPPROACHOFDESIGNINGANDMANAGINGSUPPLYNETWORKS

PetrFialaa,*

aUniversityofEconomics,Prague,CzechRepublicAbstract:Supplychainmanagementismoreandmoreaffectedbynetworkanddynamicbusinessenvironment.The fundamentaldecisions tobemadeduring thedesignphaseare the locationoffacilitiesandthecapacityallocatedtothesefacilities.Anapproachtodesigningsustainablesupplynetworksistodevelopandsolveamathematicalprogrammingmodel.Thesustainabilityofsupplynetworks can bemeasured bymultiple objectives, such as economic, environmental, social, andothers.Traditionalconceptsofoptimalityfocusonvaluationofalreadygivensystems.Newconceptofdesigningoptimal systems isapplied. Searching fora betterportfolio of resources leads to acontinuousreconfigurationandreshapingofsystemsboundaries.Torespondtorapidlychangingmarketconditions,firmsmustbeabletodynamicallyformanddissolvebusinessinteractions.Usingdouble combinatorial auctions can solve the problem. The proposed combined approach ispromisingfordesigningandmanagingsupplynetworks.

Keywords:Supplynetwork,DeNovooptimization,multiplecriteria,combinatorialauctions

1.INTRODUCTION

Supplychainmanagementhasgeneratedasubstantialamountofinterestbothbymanagersandresearchersandismoreandmoreaffectedbynetworkanddynamicbusinessenvironment.TheSupply network is defined as a systemof suppliers,manufacturers, distributors, retailers andcustomers where material, financial and information flows connect participants in bothdirections (see for example Fiala, 2005). There aremany concepts and strategies applied indesigning and managing supply networks. The sustainability of supply networks can bemeasured by multiple objectives, such as economic, environmental, social, technological, andothers.

Traditionalconceptsofoptimality focusonvaluationofalreadygivensystems.Multi‐objectivelinearprogramming(MOLP)isamodelofoptimizingagivensystembymultipleobjectives.Newconceptofdesigningoptimalsystemswasproposed(Zeleny,1990).Asamethodologyofoptimalsystem design can be employed Multi‐objective De Novo linear programming (MODNLP)problemforreshapingfeasiblesetsinlinearsystems.TheapproachisbasedonreformulationofMOLP problem by given prices of resources and the given budget and searching for a betterportfolio of resources. The paper presents approaches for solving the MODNLP problem fordesignofsustainablesupplynetworks.

How to coordinate the decentralized supply network to be efficient as the centralized one?There are many concepts and strategies applied in managing supply networks. Auctions are * pfiala@vse.cz

importantmarketmechanisms for theallocationofgoodsandservices.Anauctionprovidesamechanism for negotiation between buyers and sellers. Combinatorial auctions are thoseauctionsinwhichbidderscanplacebidsoncombinationsofitems(seeCramtonetal.,2006).Inforwardauctionsasinglesellersellsresourcestomultiplebuyers.Inareverseauctions,asinglebuyer attempts to source resources from multiple suppliers, as is common in procurement.Auctionswithmultiple buyers and sellers are called double auctions. Auctionswithmultiplebuyersandsellersarebecomingincreasingpopularinelectroniccommerce.Weproposetouseamodelofdoublecombinatorialauctionsastradingmodelbetweenlayersofsupplynetwork.

2.SUSTAINABLESUPPLYNETWORKS

In thenextpartasustainablesupplynetworkdesignproblem is formulated.The fundamentaldecisions to be made during the design phase are the location of facilities and the capacityallocatedtothesefacilities.Anapproachtodesigningasustainablesupplynetworkistodevelopandsolveamathematicalprogrammingmodel.Themathematicalprogramdeterminestheideallocations for each facility and allocates the activity at each facility such that the multipleobjectivesareconsideredandtheconstraintsofmeetingthecustomerdemandandthefacilitycapacityaresatisfied.AgeneralformofthemodelforthesustainablesupplynetworkdesignisMulti‐objectivelinearprogramming(MOLP)model.Aspecificmodelispresentedbelow.

Themodel of a supply network consists of 4 layerswithm suppliers,S1,S2,…Sm,n potentialproducers,P1,P2,…Pn,p potentialdistributors,D1,D2,…Dp, andr customers,C1,C2,…Cr.Thefollowingnotationisused:

ai=annualsupplycapacityofsupplieri,bj=annualpotentialcapacityofproducerj, wk=annualpotentialcapacityofdistributork,dl=annualdemand‐customerl,

Pjf =fixedcostofpotentialproducerj,

Dkf =fixedcostofpotentialdistributork,

Sijc =unittransportationcostfromSitoPj,

Pjkc =unittransportationcostfromPjtoDk,

Dklc =unittransportationcostfromDktoCl,

Sije =unitenvironmentalpollutionfromSitoPj,

Pjke =unitenvironmentalpollutionfromPjtoDk,

Dkle =unitenvironmentalpollutionfromDk

toCl,

Sijx =numberofunitstransportedfromSitoPj,

Pjkx =numberofunitstransportedfromPjto

Dk,

Dklx =numberofunitstransportedfromDktoCl,

Pjy =binaryvariableforbuild‐upoffixedcapacityofproducerj,

Dky =binaryvariableforbuild‐upoffixedcapacityofdistributork.

Usingtheabovenotationstheproblemcanbeformulatedasfollows:Themodelhastwoobjectives.Thefirstoneexpressesminimizingoftotalcosts.Thesecondoneexpressesminimizingoftotalenvironmentalpollution.

Min 11 1 1 1 1 1 1 1

p p pn m n n rP P D D S S P P D Dj j k k ij ij jk jk kl kl

j k i j j k k l

z f y f y c x c x c x

Min 21 1 1 1 1 1

p pm n n rS S P P D Dij ij jk jk kl kl

i j j k k l

z e x e x e x

Subjecttothefollowingconstraints:

theamountsentfromthesuppliertoproducerscannotexceedthesuppliercapacity

,

1

1, 2, ..., ,n

ij ij

x a i m

theamountproducedbytheproducercannotexceedtheproducercapacity

1

, 1, 2, ..., ,p

jk j jk

x b y j n

theamountshippedfromthedistributorshouldnotexceedthedistributorcapacity

1

, 1, 2, ..., ,r

kl k kl

x w y k p

theamountshippedtothecustomermustequalthecustomerdemand

1

, 1, 2, ..., ,p

kl lk

x d l r

theamountshippedoutofproducerscannotexceedunitsreceivedfromsuppliers

1 1

0, 1, 2, ..., ,pm

ij jki k

x x j n

theamountshippedoutofdistributorscannotexceedquantityreceivedfromproducers

1 1

0, 1, 2, ..., ,n r

jk klj l

x x k p

binaryandnon‐negativityconstraints , 0,1 , , , 0, 1, 2, ..., , 1, 2, ..., , 1, 2, ..., , 1, 2, ..., .j k ij jk kly y x x x i m j n k p l r

Theformulatedmodelisamulti‐objectivelinearprogrammingproblem(MOLP).

3.FROMOPTIMIZINGGIVENSYSTEMSTODESIGNINGOPTIMALSYSTEMS

Multi‐objectivelinearprogramming(MOLP)isamodelofoptimizingagivensystembymultipleobjectives. InMOLP problems it is usually impossible to optimize all objectives together in agivensystem.Trade‐offmeansthatonecannotincreasethelevelofsatisfactionforanobjectivewithout decreasing this for another objective. Trade‐offs are properties of an inadequatelydesignedsystemandthuscanbeeliminatedthroughdesigningbetterone.Thepurposeisnottomeasure and evaluate tradeoffs, but tominimize or even eliminate them. An optimal systemshouldbetradeoff‐free.Multi‐objectiveDeNovolinearprogramming(MODNLP)isaproblemfordesigninganoptimalsystembyreshapingthefeasibleset(Zeleny,2010).

3.1Optimizinggivensystems

Multi‐objectivelinearprogramming(MOLP)problemcanbedescribedasfollows.

“Max”z=Cxs.t.Ax≤b,x≥0.(1)

where C is a (k, n) – matrix of objective coefficients, A is a (m, n) – matrix of structuralcoefficients, b is an m‐vector of known resource restrictions, x is an n‐vector of decisionvariables.InMOLPproblemsitisusuallyimpossibletooptimizeallobjectivesinagivensystem.Formulti‐objectiveprogrammingproblemstheconceptofnon‐dominatedsolutionsisused(seefor example Steuer, 1986). A compromise solution is selected from the set of non‐dominatedsolutions.Thereareproposedmanymethods.Mostofthemethodsarebasedontrade‐offs.

3.2Designingoptimalsystems

BygivenpricesofresourcesandthegivenbudgettheMOLPproblem(1)isreformulatedintheMODNLPproblem(2)

“Max”z=Cxs.t.Ax‐b≤0,pb≤B,x≥0.(2)

whereb is anm‐vectorofunknownresource restrictions,p is anm‐vectorof resourceprices,andBisthegiventotalavailablebudget.From(2)followspAx≤pb≤B.

Definingann‐vectorofunitcostsv=pAwecanrewritetheproblem(2)as

“Max”z=Cxs.t.vx≤B,x≥0.(3)

Solvingsingleobjectiveproblems

Maxzi=cixi=1,2,…,ks.t.vx≤B,x≥0(4)

z*isak–vectorofobjectivevaluesfortheidealsystemwithrespecttoB.Theproblems(4)arecontinuous“knapsack”problems,thesolutionsare

ij

iij jjvB

jjx

i,

,0,where

)/(max),...,1( j

ij

ji vcnjj .

Themeta‐optimumproblemcanbeformulatedasfollows Minf=vx

s.t.Cx≥z*,x≥0.(5)

Solving theproblem (5)provides solution:x*, B*=vx* ,b*=Ax*. The valueB* identifies theminimumbudgettoachievez*throughsolutionsx*andb*. ThegivenbudgetlevelB≤B*.Theoptimum–path ratio for achieving the best performance for a given budget B is defined as

*1B

Br .

Theoptimum‐pathratioprovidesaneffectiveandfasttoolfortheefficientoptimalredesignoflarge‐scalelinearsystems.OptimalsystemdesignforthebudgetB:x=r1x*,b=r1b*,z=r1z*.

3.3DeNovoapproachforsustainablesupplynetworks

TheDeNovo approach can be useful in the design of the sustainable supply network.Only apartialrelaxationofconstraintsisadopted.Produceranddistributorcapacitiesarerelaxed.Unitcostsforcapacitybuild‐uparecomputed:

PjP

jj

fp

b =costofunitcapacityofpotentialproducerj,

D

D kk

k

fp

w =costofunitcapacityofpotentialdistributork.

Variablesforbuild‐upcapacitiesareintroduced: Pju =variableforflexiblecapacityofproducerj,

Dku =variableforflexiblecapacityofdistributork.Theconstraintsfornon‐exceedingproduceranddistributorfixedcapacitiesarereplacedbytheflexiblecapacityconstraintsandthebudgetconstraint:

1

0, 1, 2, ..., ,p

Pjk j

k

x u j n

1

0, 1, 2, ..., ,r

Dkl k

l

x u k p

1 1

.pn

P P D Dj j k k

j k

p u p u B

4.MANAGINGSUPPLYNETWORKSBYDOUBLECOMBINATORIALAUCTIONS

Weproposetouseamodelofdoublecombinatorialauctionsastradingmodelbetweenlayersofsupplynetwork.

4.1Doubleauctions

Auctions with multiple buyers and multiple sellers are becoming increasing popular inelectroniccommerce.Doubleauctionsarenotsooftenstudied inthe literatureassingle‐sidedauctions (see Xia et al., 2005). For double auctions, the auctioneer is faced with the task ofmatchingupasubsetofthebuyerswithasubsetofthesellers.Theprofitoftheauctioneeristhedifferencebetweenthepricespaidbythebuyersandthepricespaidtothesellers.Theobjectiveistomaximizetheprofitoftheauctioneergiventhebidsmadebysellersandbuyers.

We present a double auction problem of indivisible items withmultiple sellers andmultiplebuyers.LetussupposethatmpotentialsellersS1,S2,...,SmofferasetRofritems,j=1,2,…,r,tonpotentialbuyersB1,B2,...,Bn.AbidmadebysellerSh,h=1,2,…,m,isdefinedasbh=C,ch(C),abidmadebybuyerBi,i=1,2,…,n,isdefinedasbi=C,pi(C),where

C⊆R,isacombinationofitems,ch(C),istheofferedpricebysellerShforthecombinationofitemsC,pi(C),istheofferedpricebybuyerBiforthecombinationofitemsC.

Binaryvariablesare introduced formodel formulation:xi(C) isabinaryvariablespecifying ifthecombinationCisassignedtobuyerBi(xi(C)=1),yh(C)isabinaryvariablespecifyingifthecombinationCisboughtfromsellerSh(yh(C)=1).

1

n

i

C R pi(C)xi(C)‐

1

m

h

C R ch(C)yh(C)max

s.t.1

n

i

C R xi(C)≦

1

m

h

C R yh(C), j∊R, (6)

xi(C)∊0,1, C⊆R, i,i=1,2,…,n,yh(C)∊0,1, C⊆R, h,h=1,2,…,m.

Theobjectivefunctionexpressestheprofitoftheauctioneer.Theconstraintsensuresforbuyerstopurchasearequireditemandthattheitemmustbeofferedbysellers.

4.2Solvingofdoubleauctions

The formulated combinatorial double auction can be transformed to a combinatorial single‐sided auction. Substituting yh(C) , h = 1, 2, …,m, with 1 ‐ xi(C), i = n+1,n+2, …, n+m, andsubstituting ch(C), h = 1, 2, …, m, with pi(C), i = n+1,n+2, …, n+m , we get a model of acombinatorialsingle‐sidedauction.

1

n m

i

C R pi(C)xi(C)‐

1

n m

i n

C R pi(C)max

s.t.1

n m

i

C R xi(C)≦m, j∊R, (7)

xi(C)∊0,1, C⊆R, i,i=1,2,…,n+m.

Themodel(7)canbesolvedbymethodsforsingle‐sidedcombinatorialauctions.Complexityisafundamental question in combinatorial auction design. The algorithms proposed for solvingcombinatorialauctionsareexactalgorithmsandapproximateones.Manyresearchersconsideriterativeauctionsasanalternative.

Onewayof reducing someof the computational burden in solving theproblem is to set up afictitious market that will determine an allocation and prices in a decentralized way. In theiterativeapproach,therearemultipleroundsofbiddingandallocationandtheproblemissolvedin an iterativeand incrementalway. Iterative combinatorial auctionsareattractive tobiddersbecausetheylearnabouttheirrivals'valuationsthroughthebiddingprocess,whichcouldhelpthem to adjust their own bids (see Parkes, 2001). The key challenge in the iterative

combinatorialauctionsdesign is toprovide informationfeedbacktothebidders(seePikovskyand Bichler, 2005).We propose to use an iterative approach for combinatorial auctions. Theprimal‐dual approach is used for solving. For the problem (7) we will formulate the LPrelaxationanditsdual.Theschemecanbeoutlinedasfollows:

1.Chooseinitialpricesforsellersandbuyers.2.Announcecurrentpricesandcollectbids.3. Computethecurrentdualsolutionbyinterpretingthepricesasdualvariables.Trytofinda

feasibleallocation,anintegerprimalsolutionthatsatisfiesthestoppingrule.Ifsuchsolutionisfound,stopanduseitasthefinalallocation.Otherwiseupdatepricesandgobackto2.

5.CONCLUSION

DeNovo approachwas applied for sustainable supply network design problem and providesbetter solution than traditional approaches applied on fixed constraints. The design problemwasformulatedasMOLPproblem.Theeconomicandenvironmentalobjectiveswereusedinthemodelbutmultipleobjectivescanbeusedingeneral.DeNovoprogramming(DNP)approachisopenforfurtherextensionsasfuzzyDNP,intervalDNP,complextypesofobjectivefunctionsandcontinuousinnovations.Thenumerousapplicationsinelectroniccommercehaveledtoagreatdealofinterestindoubleauctions.Thepaperpresentsamodelofcombinatorialdoubleauctionsformanaging supply networks. The formulatedmodel can be transformed to a combinatorialsingle‐sidedauctionandsolvedbymethodsforsingle‐sidedcombinatorialauctions.Weproposeto use an iterative approach to solving combinatorial double auctions. The primal‐dualalgorithmcanbetakenasadecentralizedanddynamicmethodtodetermineequilibrium.

ACKNOWLEDGMENT

The research project was supported by the grant No. 13‐07350S of the Grant Agency of theCzech Republic and by Grant No. IGA F4/54/2015, Faculty of Informatics and Statistics,UniversityofEconomics,Prague.

REFERENCES

[1] Cramton, P., Shoham, Y. and Steinberg, R. (eds.), (2006). Combinatorial Auctions. MITPress,Cambridge.

[2] Fiala,P.,(2005).Informationsharinginsupplychains.Omega,33,419‐423.

[3] Parkes, D. C., (2001). Iterative Combinatorial Auctions: Achieving Economic andComputationalEfficiency.PhDthesis,UniversityofPennsylvania.

[4] Pikovsky,A,Bichler,M.,(2005).InformationFeedbackinIterativeCombinatorialAuctions.InConferenceproceedingsWirtschaftsinformatik,Springer,Bamberg.

[5] Steuer,R.E.(1986),MultipleCriteriaOptimization:Theory,ComputationandApplication.Wiley,NewYork.

[6] Xia, M., Stallaert,J., Whinston, A. B., (2005). Solving the combinatorial double auctionproblem.EuropeanJournalofOperationalResearch,164,239–251.

[7] Zeleny, M., (2010). Multiobjective Optimization, Systems Design and De NovoProgramming.In:Zopounidis,C.,Pardalos,P.M.(ed.):HandbookofMulticriteriaAnalysis.Springer,Berlin.

[8] Zeleny, M., (1990). Optimal Given System vs. Designing Optimal System: The De NovoProgrammingApproach.InternationalJournalofGeneralSystem,17,295‐307.

LOCATIONROUTINGMODELFORDESIGNINGPLASTICSRECYCLINGNETWORK

MiloradVidovića,BranislavaRatkovića,*,GordanaRadivojevića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Thispaperproposeslocation‐routingmodelfordesigningrecyclingnetworkwithprofit.Proposedmodel simultaneously determine collection points' locations with distance‐dependentreturns, location of intermediate consolidation points (transfer centers) and the route of thecollectionvehiclesoastomaximizeitsprofitfromthecollectionofrecyclables.Routingpartofthemodelwasformulatedasamultiplematchingproblem.

Keywords:location‐routingproblem,recyclingnetwork,MILP

1.INTRODUCTION

The location‐routingproblems (LRP)merges facility location andvehicle routing into a singleproblem where strategic location and tactical/operational routing decisions are takensimultaneously.Thisintegratedapproachhasfoundtobeusefulinseveralreal‐lifeapplications(Nagy and Salhi, 2007). In this paper we consider a LRP in plastics recycling. Namely, it isestimatedthat66.5milliontonsofplasticwillbeplacedontheEuropeanUnion(EU)marketin2020 and global plastic production could triple by 2050 (Green Paper, 2013). Once in theenvironment,plasticwastecanpersist forhundredsofyears (GreenPaper,2013). Inorder todealwith theproblemof plasticwaste, EU introduced legislation like thePackagingDirective94/62/EC and Framework Directive on waste 2008/98/EC. The Packaging Directive has aspecific recycling target for plastic packaging,while the FrameworkDirective onwaste sets ageneral recycling target forhouseholdwastewhich coversplasticwaste (GreenPaper, 2013).For achieving imposed recycling targets, it is necessary to establish appropriate logisticsnetworkstructures.This logisticsnetworksmustbeconvenient forendusers(González‐TorreandAdenso‐Díaz,2005),sincetheparticipationofendusersiscrucialforasuccessfulachievingany recovery target set by legislation, because they are responsible for separation of theseproductsat theirresidenceandcarrying themtodesignatedcollectionpoints (CPs).Although,efficient source segregation collection can contribute significantly to maximizing materialrecycling, but can represent up to 70% of the entire cost of waste management (Dogan andDuleyman,2003).

From here, the main intention of this paper is to propose a model for designing recyclinglogisticsnetwork (RN)withprofit.TheLRmodel fordesigningRNwithprofitproposedhere,has the followingspecificities.Locationpartof themodel includesdecisionsof thepositioningbothCPsasa lower levelof thenetworkand transferstations (TSs)at thehigher levelof the

* b.ratkovic@sf.bg.ac.rs

network.Therevenueobtained fromquantityofrecyclables forspecificCP isrelatedwiththeproximity of the CP to the end users. So, we introduced distance dependent quantity ofrecyclablesdroppedofftoCPsintroducingcollectionrateasafunctionofdistance.Particularly,we consider aproblemwhere endusers are located in cityblocks.Mostof the todaymoderncitieshavesocalledblockstructures,characterizedwithbuildingsinwhichresidents live,andinternalstreetsnetworkwithin.Inordertocollectrecyclablesinsuchenvironment,CPsmustbelocatedalongtheseinternalstreets.Therefore,routingpartofthemodelgivesopportunityforconsideringwholeblocksasanetworknodestobevisitedbycollectionvehicles.Internalstreetnetwork in city blocks is usually simple, so it is possible to determine optimal route throughblocks in advance.Another specificity of theproposed approach is thatwhileperformingonecollection route,due to its limitedcapacity, vehicle canvisitonly smallnumberof cityblocks.This facts gives opportunity to formulate and solve routing part of the problem as multiplematchingproblem.

Theremainderofthepaperisorganizedasfollows.Section2presentstheproblemdescription,and mathematical formulation of the described problem. Section 3 gives numerical example.Finally,section4concludesthestudy.

2.MATHEMATICALFORMULATION

Mostofthetodaymoderncitieshavesocalledblockstructures.Theseblockscanbeinavarietyofsizesandshapes,andtheyarecharacterizedwithbuildingsinwhichresidentsliveandroadnetworkwithin.Inthispaper,termenduserreferstoabuildinginsidethespecificcityblock(weaggregatedresidentstoitsresidentialbuildingconsideringthemasasingleenduser).Eachenduserischaracterizedbythevolumeofwasteproducedperdaywhichcorrespondstothetotalquantity generated in all households residing in a building. Potential locations of CP arecharacterizedbyitscapacityanddistancestoallendusersineachcityblock.InordertomodeltheinfluenceofdistancebetweenendusersandCPsonthecollectingofrecyclables,weassumethatrecyclablescollectionrate 1,0df isaknownfunctionofdistance.Thisfunctionmodelstheinfluence of distance between end users and CPs, in way that collection rate is inverselyproportionaltodistance(Bermanetal.,2003).Wedefinetwocharacteristicdistanceslandu(l<u),between theenduserandCPwhere l represents the lowerandu upperboundofwalkingdistance to CP for each end user.When the distanced from the end user to the closest CP is

ld 0 thenf(d)=1,whileincasewhen ud ,f(d)=0.IfthedistancebetweentheenduserandCP

is udl ,weassumedthatthecollectionratecorrespondstolu

dudf

)( .Intheroutingpart

of the problem, there is a capacitated vehicle that has to visit CPs located in city blocks,originatingfromTS.Transferstationrepresentsafacilityinwhichrecyclablesareinspectedandconsolidatedfor furtherprocessing.Wehaveasetofpotential locationsforTSs,characterizedbycostsofopeningTSs.Thelengthofinnerstreetsincityblocksmaydifferdependingofthecityblockshapeandsize,butoncethevehicleentersthecityblockitcouldonlybetraversethroughthese inner streets, alongwhichCPsare located.The lengthof the route throughcityblock isalwaysthesame,regardlessofthenumberofstopsperCP.ThisfactenablesustoroutedistanceonlyfromTStocityblocks,whileroutepartwhenvehicletraversecityblockispredeterminedand included in routing costs. More importantly this fact, gives us opportunity to formulatevehicleroutingpartoftheLRproblemasamultiplematchingprobleminsteadofclassicalVRPformulations. Also,we include idling time at each CP in costs calculation. Assumptions of theproposedLRmodelare:

TSisassumedtobeuncapacitated TherouteincityblockcanbepredefinedastheCPsareplacedalongtheinternalstreets

withinthecityblock,whosenetworkisshortandsimple,andtheoptimalinternalroutewithintheblockcaneasilybedeterminedbysolvingarcroutingproblem.Theinternal

route has constant lengthwhich is passed alwayswhen block is visited and does notdependsonthenumberofCPsopened

Quantitiesofrecyclablesthataregeneratedduringtheobservationperiodincityblockdonotallowthevehicletoservemorethanfourblocksinoneroute

InordertoproposeaMILPmodelfortheproblemthefollowingnotationisintroduced.

Sets:

IiI ...,...1 setofendusers

BbB ...,...1 setofcityblocks

KkK ...,...1 setofpotentialsitesforCPs

JjJ ...,...1 setofpotentialsitesfortransferpoints

Parameters:

R revenuefromsellingcollectedquantityofrecyclables

kbid walkingdistancebetweenenduseritok‐thCPinacityblockb

Zikb collectionratefоrthedistancedikb,

udwhen

udlwhendf

ldwhen

Z

ikb

ikbikb

ikb

ikb

,0

),(

0,1

Fk costsofopeningCPs Kk

Fj costsofopeningtransferpoints Jj

αkb idlingtimecostsatCP Kk incityblock Bb

Qr capacityofvehicle(orroute)

Qj capacityoftransferstation Jj

Qkb capacityofCP Kk incityblock Bb

Qib availablequantityofrecyclablesatenduser Ii incityblock Bb

Cjpqwe,Cjpqw,Cjpq,Cjp costs of visiting CPs in blocks Bewqp ,,, in a single route (includingcostsfrom/totransferstation Jj andcostsinsidecityblocks),respectivelyforroutesvisitingfour,three,two,andonecityblock Bb .Costofroutesthroughcityblocksareaddedtothesecost.

BigMnumber(sufficientlylargenumber)

Decisionvariables:

e0,otherwis

BbblockcityinopenedisKkpointcollectionif1,Ykb

e0,otherwis

openedisJjstationtransferif1,Y j

e0,otherwis

JjpointtransferfromroutesametheinmergedareBp,q,w,enodesif1,Y jpqwe

e0,otherwis

JjpointtransferfromroutesametheinmergedareBp,q,wnodesif1,Y jpqw

e0,otherwis

JjpointtransferfromroutesametheinmergedareBp,q,nodesif1,Y jpq

e0,otherwis

JjpointtransferfromroutedirecttheinservedareBp,nodeif1,Y jp

Xikb≤1definesfractionofrecyclablesbroughtfromenduserto Ii collectionsite Kk incityblock Bb .

ThemathematicalformulationoftheproposedMILPmodelisgivenbelow:

jp

j Bpjpjpq

j Bp pBqjpqjpqw

j Bp pBq qpBwjpqw

jpqwej Bp pBq qpBw wqpBe

jpqwej

jjb k

kbkbkbb k i

ikbib

YCYCYC

YCYFYFXRQMax

// /

/ / /)(

(1)

s.t.bkiZXQ ikbikbib ,,, (2)

biXk

ikb ,,1 (3)

bkQYXQ kbkbi

ikbib , (4)

bkYX kbikb , (5)

pYYY

YYYY

YYY

jpj pBq

jqpj pBq

jpq

j pBq qpBwjqwp

j pBq qpBwjqpw

j pBq qpBwjpqw

j pBq qpBw wqpBejqwep

j pBq qpBw wqpBejqwpe

j pBq qpBw wqpBejqpwe

j pBq qpBw wqpBejpqwe

,1//

/ // // // / /

/ / // / // / /

(6) wqpBeqpBwpBqBpJjYY jjpqwe /,/,/,,, (7)

qpBwpBqBpJjYY jjpqw /,/,,, (8)

pBqBpJjYY jjpq /,,, (9)BpJjYY jjp ,, (10)

wqpBeqpBwpBqBpJj

YQYMXQXQXQXQ jpqwerjpqweikei k

ieikwi k

iwikqi k

iqikpi k

ip

/;/;/;;

,)1(

(11)

qpBwpBqBpJj

YQYMXQXQXQ jpqwrjpqwikwi k

iwikqi k

iqikpi k

ip

/,/,,

,)1(

(12) pBqBpJjYQYMXQXQ jpqrjpqikq

i kiqikp

i kip /,,,)1(

(13)BpJjYQYMXQ jprjpikp

i kip ,,)1(

(14) wqpBeqpBwpBqBpJjYY jpqwe

kkp /,/,/,,,

(15) wqpBeqpBwpBqBpJjYY jpqwe

kkq /,/,/,,,

(16) wqpBeqpBwpBqBpJjYY jpqwe

kkw /,/,/,,,

(17) wqpBeqpBwpBqBpJjYY jpqwe

kke /,/,/,,,

(18) qpBwpBqBpJjYY jpqw

kkp /,/,,,

(19) qpBwpBqBpJjYY jpqw

kkq //,

(20) qpBwpBqBpJjYY jpqw

kkw /,/,,,

(21) pBqBpJjYY jpq

kkp /,,,

(22)

pBqBpJjYY jpqk

kq /,,, (23)

BpJjYY jpk

kp ,, (24)

0,1,1,0,1,0,1,0

,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0

ikbikbjpjpqjpq

jqwpjqpwjpqwjqwepjqwpejqpwejpqwejkb

XXYYY

YYYYYYYYY

(26)

Theobjective functionmaximizes theprofitwhich is composedof obtained revenue from thecollected recyclables (term 1 in objective function), minus the fixed cost of opening CPs andidlingtimecostsineachcityblock(term2),costsoflocatingtransferpoints(term3),anroutingcosts(terms4,5,6and7).Firstconstraintset(2)representsthequantityofrecyclablestobecollectedunder ''coveragedecay function''.Constraints(3)ensures that fractionofrecyclablesassigned to CP k is less or equal to quantity generated at end user i. Quantity of recyclablesassigned to CP cannot exceed the capacity of CP (4), while constraints (5) ensures thatrecyclablesareassignedtoonlyopenedCPs.Constraints(6)prohibitmultiplevisitsofthesamenode(cityblockb),andprovidethateachcityblockmustbevisitedexactlyonce.Constraintsets(7)‐(10)ensuresthatvehiclesstartstouronly fromtransferpointwhich isopened.Constraintsets (11)‐(14) ensures that vehicle capacity isn't exceeded. Constraint sets (15)‐(24) ensuresthat only cityblocks inwhichCPs areopened are visited in the route, Finally, constraint sets(26)definenatureofthevariables.

3.NUMERICALEXAMPLE

To test proposed MILP model for designing RN with profit, we generated three differentproblem instance sets: small (S), medium (M) and large (L) instances. The main differencebetweeninstancesisinnumberofendusersandspatialdistributionincityblocks.Forallsetofproblemswegenerated240instances(Table1).Theexecutionoftheinstances(CPUtime)hasbeen limited to 7200 seconds. Models have been solved with the usage of Cplex 12.6, andinstanceswererunonanIntel(R)Core(TM)i5‐4200U(2.30GHz,RAM8Gb).

Table1.Inputparametersofthemodel

Inputparameters S M L

Fk(€/day) 0.5Fj(€/day) 0Costsperstop(€/perstop) 0.002revenue(€/kg) 0.12264Qk(kg) 20Qr(kg) 200 300 800l(m) 50u(m) 400Timehorizon 2.5dayNo.ofcityblocks 4,5,6,8,10,12,15,20No.ofpotentiallocationfortransferpoints 2Numberofresidentsineachcityblock Beta(2,5)distributionintherange[48,200]Generatedquantityofrecyclablesin(Qib) Uniformdistribution(0.8;0.1)*P(P‐%ofplasticsinmunicipalwaste)NumberofpotentialCPs (Qib/Qk)+1Distancesbetweencityblocks(m) Randomlygeneratedintherange[100,700]Distancesbetweentransferpointsandcityblocks(m)

Randomlygeneratedintherange[1000,10000]

Distancesinsidecityblocks(m) Randomlygeneratedinrange[400,1200]DistancesbetweenendusersandpotentiallocationsforCPs

Beta(2,5)distributionintherange[15,400]

Costperkmtraveled(€/km) 0.0006(insidetheblockx3)Numberofendusersineachcityblock(Randomlygeneratedintheshownrange:)

[4,7] [5,25] [10,70]

DuetothecomplexityofproposedLRmodel,theinstancesthatcanbesolvedtooptimalityaretypically of small size. When solving large and medium instances in preliminary tests,computationaldifficultieswerefacedduetothehardcombinatorialproblemcharacteristics.Formediuminstances,instancesupto8cityblock(notallofit)couldbesolvedtooptimality.Incaseoflargeinstances,eventhesmallestnumberofcityblockscauseddifficulties,whilethebiggestnumber of city blocks could not be loaded into computer memory (Table 2). Although thisresultspresentthebeginningoftheresearchinthisarea,it isclearthatproposedMILPmodelneedsdevelopmentofappropriateheuristicsorapplicationofmetaheuristicsforsolvingmodelsoflargersize.

Table2.ResultsoftheproposedLRmodel

No.ofcityblocks

Sinstances Minstances Linstances

#opt** Fopt* CPU‐opt(s)* #opt** Fopt*CPU‐opt(s)*

#opt** Fopt*CPU‐opt(s)*

4 10/10 2.04 12.59 9/10 24.32 313.81 1/10 134.37 11.80

5 10/10 ‐0.26 12.16 9/10 40.37 104.20 0/10 / /

6 10/10 ‐0.05 0.83 3/10 47.85 20.67 0/10 / /

8 9/10 5.95 55.98 1/10 53.3 98.83 0/10 / /

10 10/10 3.40 25.33 0/10 / / 0/10 / /

12 10/10 7.244803 67.42 0/10 / / 0/10 / /

15 10/10 9.268272 632.88 0/10 / / 0/10 / /

20 10/10 15.56321 537.26 0/10 / / a* a* a**averagevalue;**timelimit7200s;a*couldn'tbeloadedintothememory

3.CONCLUSION

This paper presents a possible approach to define the optimal network for recycling plasticwaste, by presenting LR formulation for this problem. Preliminary testing of the proposedapproaches is promising, but numerous aspects of the problem and application of approachproposed need future research. Future research directionmay include examining the systembehaviorwithdifferentsystemparameters(differentvehiclecapacity,differentcapacityof thecontainers,etc).Butmostimportantly,developmentofappropriateheuristicsforsolvingmodelsoflargersize.

ACKNOWLEDGMENT

This paper is partially supported by the Ministry of Education, Science and TechnologicalDevelopment,no.TR36006.

REFERENCES

[1] Nagy,G.,Salhi,S.(2007).Location‐routing:Issues,modelsandmethods.EuropeanJournalofOperationalResearch,177,649–672

[2] GREEN PAPER On a European Strategy on Plastic Waste in the Environmenthttp://ec.europa.eu/environment/waste/plastic_waste.htm

[3] González‐Torre,P.L.,Adenso‐Díaz,B.(2005).Influenceofdistanceonthemotivationandfrequencyofhouseholdrecycling.WasteManagement,25,15–23.

[4] Dogan, K., Duleyman, S., (2003). Cost and financing of municipal solid waste collectionservicesinIstanbul.WasteManagementandResearch,21(5),480–485.

[5] Berman,O.,Krass,D.,Drezner,Z.,(2003).Thegradualcoveringdecaylocationproblemonanetwork.EuropeanJournalofOperationalResearch151,474–480

DYNAMICANALYSISFOREND‐OF‐LIFEVEHICLESMANAGEMENTSYSTEMS:ANINEXACTMULTI‐STAGE

PROGRAMMINGAPPROACH

VladimirSimića,*,BrankaDimitrijevićb

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,vsima@sf.bg.ac.rs

Abstract:Nowadays,exportofusedvehicles represents themost significantbarrier to themoreefficient vehicle recycling in the EU, becausemillions of vehicles which are expected to go todomesticvehiclerecyclingfactoriesareexported.Asaresult,howtoallocatelimitedandfrequentlyinsufficient quantities of collected end‐of‐life vehicles (ELVs) to satisfy vast demands of vehiclerecyclingfactoriesbecomesasignificantconcernofmanywastemanagementsystemsthatcontroltheELV collectionand treatmentnetworksacross theEU. In thispaper,amulti‐stage interval‐stochastic programming (MSISP) model is developed for supporting the management of ELVallocation under uncertainty. The MSISP is a hybrid of inexact optimization and multi‐stagestochasticprogramming.TheformulatedMSISPmodelcandirectlyhandleuncertaintiesexpressedaseitherprobabilitydensityfunctionsordiscreteintervals.Inaddition,itcanhandleuncertaintiesthroughconstructingasetofscenariosandreflectdynamicfeaturesofthesystemconditions.

Keywords:End‐of‐lifevehicle;Decisionmaking;Multi‐stage stochasticprogramming; Intervalprogramming.

1.INTRODUCTION

Theamountofmaterialpassingthroughtheend‐of‐life(ELV)recyclingnetworksallovertheEUhasbeenreducedduetoanincreasedexportofusedvehiclesforreuseassecond‐handvehiclesor as sources of used parts andmaterials. Nowadays, export of used vehicles represents themost significant barrier to themore efficient vehicle recycling in the EU, becausemillions ofvehicles,whicharesupposedtogotodomesticvehiclerecyclingfactories,areexported.Illegaltreatment facilities represent another problem for the EU vehicle recycling industry sector,especially in some new EUmember states (Tavoularis et al., 2009). Illegal operators can paymoretothelastownersforanELV,becausetheyhavenointentionofincurringthecostofeitherdepollutingthevehicleorattemptingtoreachtheeco‐efficiencytargetspromulgatedbytheEUELV Directive (EU, 2000). Abandoned vehicles represent another major problem and theintroduction of the EU ELV Directive initially increased their number. Consequently, how toallocatelimitedandofteninsufficientquantitiesofcollected,decontaminatedandflattenedELVsto satisfy vast demands of vehicle recycling factories becomes a significant concern of manywastemanagementsystemsthatcontroltheELVcollectionandtreatmentnetworksallovertheEU.

* vsima@sf.bg.ac.rs

Owingtothecomplexityofthevehiclerecyclingsubject,averysmallnumberofresearcharticleshave been published. Reuter at al. (2006) formulated a nonlinear model to optimise theperformancesoftheELVrecyclingsystem.QiandHongcheng(2008)proposedamixedintegerlinearprogrammingmodel fordesigningELVrecoverynetwork.Cruz‐RiveraandErtel (2009)constructedanuncapacitatedfacilitylocationmodelinordertodesignacollectionnetworkforELVsinMexico.Vidovicetal.(2011)presentedmodellingapproachthatcouldbeusedtolocatecollection facilities for ELVs. Stoyanov (2012) formulated a multi‐source capacitated facilitylocation model in order to design a network of dismantling centers for ELVs in Bulgaria.Gołębiewski et al. (2013) proposed a simulation approach that could be used to determinelocations forELVdismantlers. SimicandDimitrijevic (2013)developeda riskexplicit intervallinearprogrammingmodelforoptimallong‐termplanningintheEUvehiclerecyclingfactories.Mora et al. (2014) proposed amixed integer linear programmingmodel for ELV closed‐loopnetworkdesign.

Fromthereviewofpreviousliterature,itisevidentthatanumberofsystemsanalysismethodsweredeveloped for solving variousELVmanagementproblems.However, the abovemethodscan hardly solve problems where multi‐stage decisions need to be made, particularly whenrandomvariablesexist inthevehiclerecyclingsystemwhiledecisionshavetobemadebeforetherandomeventoccurs.Infact,nopreviousstudywasreportedontheapplicationofthemulti‐stage stochastic programming technique to ELVmanagement problems. Therefore, in viewofthe limitations in previous works, the multi‐stage interval‐stochastic programming (MSISP)modelisformulatedandpresentedinthispaper.

The remaining part of the paper is organized as follows: Section 2 describes the consideredproblemandpresentsthemulti‐stageinterval‐stochasticprogrammingmodelforsupportingthemanagementofELVallocationunderuncertainty.Section3presentsconclusionsofthework.

2.METHODOLOGY

Considerawastemanagementsystemwherearecyclingmanager isresponsible forallocatingELVsfrommultipleregionstomultiplevehiclerecyclingfactorieswithinamulti‐periodplanninghorizon. The available quantities of ELVs in each region are random variables with knownprobabilities. Decisions about ELV allocation targets must be made at an earlier stage whenvaryingprobabilitylevelsexist,becausemanagersofallvehiclerecyclingfactoriesneedtoknowinadvancewhichquantityofELVstheycanexpectinordertocreateadequateproductionplans.Therefore,guaranteedquantitiesofELVsfromeachregiontoeachvehiclerecyclingfactoryarepromulgated in advance. On the other hand, if the guaranteed quantities of ELVs cannot bedistributed to vehicle recycling factories due to insufficient number of collectedELVs, vehiclerecyclingfactorieswillhavetoimportELVsatahigherpriceand/orworkatreducedcapacity.To all vehicle recycling factories which have not fully been provided with preliminary ELVallocation targets penaltiesmust be reimbursed, as specific compensation for acquiring ELVsfrommoreexpensivesourcesorbecauseofsmallerprofitduetoreducedworkingcapacities.

The quantities of collected, depolluted and flattened ELVs during every planning period arerandomvariables, and the appropriateELVallocationplanwouldbeofdynamic feature.As aresult, ELV allocation decisions need to be made periodically. Thus, the problem of ELVallocationneedstobesolvedusingamulti‐stagestochasticprogrammingapproach,becauseitprovidesthepossibilitytorepresentvariousuncertaintiesinaformofmultilayerscenariotree.

Uncertainties exist also in economic parameters, ELV allocation targets and demand fromvehiclerecyclingfactories.Observingaforementionedmodellingparametersasintervalvaluesispurelynatural.Inordertoreflectsuchuncertainties,interval‐parameterprogrammingneedstobeintroducedintothemodellingformulation.

Thus, the described problem can be formulated as a multi‐stage interval‐stochasticprogrammingmodelforplanningend‐of‐lifevehiclesallocationasfollows:

1 R V S( ( ) )

t

Tvrt vrt vrt vrt rst vt vrst

t r v sMax f D Z Z p K M (1a)

subjectto:

11

V( ) , R; S ; ; 1

vrt vrt vrt vrst rst rs' t t sv

Z Z M Q H r s s' t ,...,T (1b)

10= ,r rH П

Rr

(1c)

1V

1

( ) ,

R; S ; ; 1

rst rst vrt vrt vrt vrst rs' tv

t s

H Q Z Z M H

r s s' t ,...,T

(1d)

R( ) ,

vrt vrt vrt vrst vt min

rZ Z M I V; S ; 1tv s t ,...,T (1e)

0,vrt vrt vrt vrstZ Z M V; R; S ; 1tv r s t ,...,T

(1f)

0,rstH R; S ; 0 1tr s t , ,...,T

(1g)

0 1,vrt V; R; 1v r t ,...,T

(1h)

where: t is index of time period, 1 ; t , ,T R is set of considered regions; V is set of

considered vehicle recycling factories; St is set of scenarios in period t; 1S S

T

tt

is set of

scenariosinplanninghorizon; A S S s',s s' ,sissetofconse‐cutivescenarios,whichis

determinedinlinewiththestructureofthescenariotree;

1 A S s s' s',s , s' issetof

scenariosthatprecedescenarios;Tisnumberofanalyzedtimeperiods;f isintervalvalueof

the expected profit to ELV management system over the planning horizon; , R rП r is

interval value of initial inventoryweight of ELVs collected, depolluted, flattened and piled in

regionr;

, V vt minI v isintervalvalueofsafetyinventorylevelofvehiclerecyclingfactoryvin

periodt; , V, R vrtD v r isintervalvalueofrevenuetoELVmanagementsystemperweight

unit of ELVs allocated from region r to vehicle recycling factory v in period t; , V vtK v (

, vt vrtK D v,r ,t )isintervalvalueofloss(i.e.penalty)toELVmanagementsystemperweight

unit of ELVs not delivered to vehicle recycling factory v in period t; , R, S rst tp r s is the

probabilityofoccurrence for scenario s in regionr andperiod t; , R, S rst tQ r s is interval

valueofquantityofELVscollected,depollutedand flattenedbyauthorizedtreatment facilities

from region r in scenario s and period t; , V, R vrtZ v r is interval value of fixed ELV

allocation target from region r to vehicle recycling factory v in period t; , V, R vrt v r is

decision variable that is used for identifying an optimized set of ELV allocation targets vrt vrt vrt vrtZ Z Z ,where vrt vrt vrtZ Z Z and [0 1] vrt , ; , V, R, S vrst tM v r s

is interval value of quantity bywhich ELV allocation target prescribed between region r and

vehicle recycling factoryv isnotmet in scenario s andperiod t; , R, S rst tH r s is interval

valueofweightofELVspiledinregionrandscenariosattheendofperiodt.

In Model (1), the objective function (1a) seeks to maximize the expected profit of the ELVmanagementsystemthroughallocatingELVsfrommultipleregionstomultiplevehiclerecyclingfactoriesoveramulti‐stagecontext.Intheobjectivefunction,thefirsttermcalculatesrevenuetoELVmanagement system fromallocatingELVs tovehicle recycling factories.The second termrepresentstheloss(i.e.penalty)toELVmanagementsystemforviolatingthepromulgatedELVsallocationtargets.Constraints(1b)enforcethatunderallpossiblescenariosthequantityofELVsallocated from some region to vehicle recycling factories handled by the ELV managementsystem cannot be larger than the sum of quantity of piled ELVs across that region and thequantityofELVscollected,depollutedandflattenedbyauthorizedtreatmentfacilitieslocatedinthat region.Constraints (1c) initialize inventories in regionswhichcovers the consideredELVmanagementsystem.Constraints (1d)enforce the inventorybalances.Constraints (1e)ensurethe safety inventory levels in vehicle recycling factories handled by the considered ELVmanagement system in order to protect their shredders from starvation. Finally, constraints(2f)(2h)definethevaluedomainsofdecisionvariablesusedintheproposedmodel.

Model (1) can be decomposed into twodeterministic sub‐models corresponding to the lowerandupperboundsofthedesiredobjectivevalue,andsolvedusinganinteractivealgorithm.The

sub‐modelcorrespondingto f canbefirstlyformulatedasfollows:

1 R V S( ( ) )

t

Tvrt vrt vrt vrt rst vt vrst

t r v sMax f D Z Z p K M (2a)

subjectto:

11

V, R; S ; ; 1

vrt vrt vrt vrst rst rs' t t sv

Z Z M Q H r s s' t ,...,T (2b)

10= , R r rH П r

(2c)

1V

1

( ) ,

R; S ; ; 1

rst rst vrt vrt vrt vrst rs' tv

t s

H Q Z Z M H

r s s' t ,...,T

(2d)

R,vrt vrt vrt vrst vt min

rZ Z M I

V; S ; 1tv s t ,...,T

(2e)

0,vrt vrt vrt vrstZ Z M V; R; S ; 1tv r s t ,...,T

(2f)

0,rstH R; S ; 0 1tr s t , ,...,T

(2g)

0 1,vrt V; R; 1v r t ,...,T

(2h)

where vrstM ,

rstH and vrt aredecisionvariables.Let

optf , vrst optM ,

rst optH and vrt opt

be the solutions of sub‐model (2).Then, the second sub‐model corresponding to f can beformulatedasfollows:

1 R V S( ( ) )

t

Tvrt vrt vrt vrt opt rst vt vrst

t r v sMax f D Z Z p K M (3a)

subjectto:

11

V( ) , R; S ; ; 1

vrt vrt vrt opt vrst rst rs' t t sv

Z Z M Q H r s s' t ,...,T (3b)

10= , R r rH П r

(3c)

1V

1

( ) ,

R; S ; ; 1

rst rst vrt vrt vrt opt vrst rs' tv

t s

H Q Z Z M H

r s s' t ,...,T

(3d)

R,vrt vrt vrt opt vrst vt min

rZ Z M I

V; S ; 1tv s t ,...,T

(3e)

,vrt vrt vrtopt vrst vrstoptZ Z M M

V; R; S ; 1tv r s t ,...,T

(3f)

0 ,rst rstH H R; S ; 0 1tr s t , ,...,T

(3g)

where vrstM and

rstH aredecisionvariables.Let optf ,

vrst optM and rst optH besolutionsof

sub‐model(3).Thus,theprimalsolutionsforModel(1)are:

, ,opt opt optf f f

(4a)

, , vrst opt vrst opt vrst optM M M V; R; S ; 1tv r s t ,...,T (4b)

, , rst opt rst opt rst optH H H R; S ; 1 tr s t ,...,T (4c)

TheoptimalELVallocationschemeovertheplanninghorizonis:

, vrst opt vrt vrst optA Z M V; R; S ; 1tv r s t ,...,T (4d)

where vrst optA iscalculatedquantityofELVsallocatedbetweenregionrandvehiclerecycling

factoryvinscenariosandperiodt.

3.CONCLUSIONS

Thispaperintroducesthemulti‐stageinterval‐stochasticprogrammingmodelforplanningend‐of‐life vehicles allocationwhich is applicable across vehicle recycling industry that processesdozensofmillionsofELVseveryyear.Theformulatedmodelisbasedonmulti‐stagestochastic

programming and interval linear programming approaches. Thus, it can directly handleparameteruncertaintiesexpressedasbothprobabilitydensityfunctionsanddiscreteintervals.

The presented model is capable of incorporating multiple policies within the optimizationframework. Itpermitscomprehensiveanalysesofvariouspolicysituations thatareassociatedwithdifferentlevelsofeconomicpenaltiesandsystemfailureriskswhenthepromulgatedELVallocation targets are disregarded. Compared with the conventional multi‐stage stochasticprogramming approach, thepresentedmulti‐stage interval‐stochastic programmingmodel forplanning end‐of‐life vehicles allocation can incorporate much more uncertain information.Finally, it can be outlined that the proposed model is more than effective in tackling hard,uncertainty existing waste management problems. Future research will focus on extensivetestingoftheformulatedmodel.

ACKNOWLEDGMENT

ThisworkwaspartiallysupportedbyMinistryofScienceandTechnologicalDevelopmentoftheRepublicofSerbiathroughtheprojectTR36006fortheperiod2011–2015.

REFERENCES

[1] Cruz‐Rivera,C.,Ertel,J.,(2009).Reverselogisticsnetworkdesignforthecollectionofend‐of‐lifevehiclesinMexico.Eur.J.Oper.Res.196(3),930–939.

[2] EU, (2000). Directive 2000/53/EC of the European parliament and of the council of 18September2000onend‐of‐lifevehicles.Off.J.Eur.UnionL269,34–42.

[3] Gołębiewski, B., Trajer, J., Jaros,M.,Winiczenko, R., (2013).Modelling of the location ofvehiclerecyclingfacilities:acasestudyinPoland.Resour.Conserv.Recy.80,10–20.

[4] Mora,C.,Cascini,A.,Gamberi,M.,Regattieri,A.,Bortolini,M.,(2014).Aplanningmodelfortheoptimisationoftheend‐of‐lifevehiclesrecoverynetwork.Int.J.Logist.Syst.Manag.18(4),449–472.

[5] Qi, Z., Hongcheng, W., (2008). Research on construction mode of recycling network ofreverse logistics of automobile enterprises. In: Proc. of Int. Conf. on InformationManagement,InnovationManagementandIndustrialEngineering,Taiwan,p.36–40.

[6] Reuter,M.A.,VanSchaik,A.,Ignatenko,O.,deHaan,G.,(2006).Fundamentallimitsfortherecyclingofend‐of‐lifevehicles.Miner.Eng.19(5),433–449.

[7] Simic,V.,Dimitrijevic,B.,(2013).Riskexplicitintervallinearprogrammingmodelforlong‐termplanningofvehiclerecyclingintheEUlegislativecontextunderuncertainty.Resour.Conserv.Recy.73,197–210.

[8] Stoyanov, S., (2012). A theoretical model of reverse logistics network for end‐of‐lifevehicles treatment in Bulgaria. MSc Thesis, Aarhus School of Business, University ofAarhus, Denmark. <http://pure.au.dk/portal‐asb‐student/files/45645575/Master_Thesis_Svilen_Stoyanov.pdf>(accessed28.03.15).

[9] Tavoularis,G.,Lolos,Th.,Loizidou,M.,Konstantinopoulos,G.,Iordan,C.,Mihai,C.,(2009).Management of the end‐of‐life vehicles stream in Romania. In: Cossu, R., Diaz, L., &Stegmann, R. (eds.). Proc. of the Twelfth Int. Waste Management and Landfill Symp.,Cagliary,Italy,5‐9October.CISA,Italy.

[10] Vidovic,M.,Dimitrijevic,B.,Ratkovic,B., Simic,V., (2011).Anovel covering approach topositioningELVcollectionpoints.Resour.Conserv.Recy.57,1–9.

OPTIMIZATIONOFINTERNALLOGISTICSINAWASTEMANAGEMENTCOMPANYWITHQUEUINGTHEORY

DavorZaveršnika,TinaCvahtea,*,KlemenGrobina,BojanRosia

aUniversityofMaribor,FacultyofLogistics,Slovenia

Abstract:Thenumberofcompaniesforcollection,processinganddisposalofwasteisincreasing,as is theamountofwasteweproduce.Biggercompanieshave theirownvehicle fleet, theirowncollection centersand theyoften cover largegeographicalareas.For suchnetworkof supplyofservices,aqualitylogisticorganizationisneeded.ThepaperfocusesonacasestudyofaSlovenianwastemanagementcompany,whichhasawell‐organizedexternal logistics systemononehand,butalackingorganizationofinternallogisticsontheother.Thepaperwillfocusonpotentialforoptimizingtheprocessofvehicleweighingandfuelfilling,sincethisprocessrepresentsabottleneckofthe internal logistics.Congestionsappearoftenduetoexcessivevicinityoftheweighingdeviceandthegaspump.Duringtherefuelingofabiggertrucktheweighingdevicebecomesuselessasthe truck trailer is still standing on the balance. The paperwill present different improvementscenarios for internal logistics,whichwill be calculated using stochastic processes and queuingtheory.

Keywords:wastemanagement,internallogistics,queuingtheory,casestudy

1.INTRODUCTION

Logistics functions canbedivided into external and internal, and this paperwill focus on thelatter. Internal or in‐house logistics is comprised of transport in a company, all internalprocessesandallinternalflowsofmaterialsandpeople.Manycostscanbeavoidedwithoptimalplanning of these processes. In the past it has been shown that in the short term, manyoptimizationchangesarenotsensible,but inthe longrunmaymeanasignificantmoveinthedirectionofoptimallogistics.Thisisthecasealsointheresearchedcompanyforthecollection,treatmentanddisposalofwaste.Becauseofthebottleneckintheprocessofweighingtrucksonreturning to thecollectioncenterandrefuelingwithgas, lossesarecausedbywaiting.Hourlyrateofthedriver,vehicledepreciation,opportunitycosts,longeroperatingcostsandmuchmoreare costs, which the company is not aware of and so far have not beenmeasured. The focalcompanyhasbeenincreasinglyfacedwithspatialproblemsinrecentyearsduetotheincreasedvolumeofwork,which in turnaffectspoororganizationof internal transportat thecollectioncenter.Theaimoftheresearchistoexaminetheprocessesofinternaltransportatthecollectioncenter of the company and thus show the current situation. Processeswill be examined by athoroughanalysisof thedata,mostly focusingonthebottlenecksituations.Thefinalgoal is todemonstrate how using queuing theory can affect a company’s performance and improveprocesseswhich in turnmeans reducedoptimized internal logistics to saveon the costof theorganization,laborcosts,costofresourcesandtime.

2.METHODOLOGY

The case company is located in Slovenia and covers a large portion of its vicinity in terms ofgeneral wastemanagement services. Additionally, they perform services of specializedwastemanagementforknowncustomers,suchasmedicalormetalwastedisposalandrecycling.

Research for this paper is performed in three segments. First, it is necessary to analyze thecurrent situation with the use of longitudinal data, topological layout analysis and internaltransportprocesses.Withthis,theactuallossesoftimeindifferentsituationswillbepossibletocalculate.Basedonthis,aqueuingmodelwillbeassembledinordertofurtherassessprocessesofinternaltransport.Finally,possiblescenariosforimprovementswillbepresented.

Theprocessesofinternallogisticsinthecasecompanycanbedescribedasqueuesandthereforehavethepotentialtobeanalyzedandoptimizedusingqueuingtheory.Queuingsystemsincludeoneormoreservers,queuingnodes ‐customersrequiringservice,andtheprocessofserving.Where it is not possible to serve all customers at the same time, customersmustwait to beserved inaqueue (Hudoklin‐Božič, 1999,p. 136). In thepresentedcase, the customersof thequeue are the drivers of cargo vehicles, which carry the waste and other material into thecollectioncenter.Serversarethescalewherevehiclesareweighedandthefuellingtank.Ifthenumber of vehicles weighing in and refueling is more than one, this results in queues. Thearrivalsofvehiclestothecollectioncenterarenotatregularintervals,thereforethesuccessionofarrivalsofvehiclestothecenterformsaPoissonprocess.

AccordingtoHudoklin‐Božič(1999,p.139),theM/M/1queuingsystemissubjecttoaninputstreamofcustomersdescribedasaPoissonprocess,servingtimesareexponentiallydistributed,thenumberofserversis1.Suchqueuingsystemscaappearinpracticealsoinaslightlymodifiedormorecomplicatedversion–whenthequeueisinfactanetwork,consistingoftwostationsinatandemwithafiniteintermediatebuffer(seeTsiotras,Badr&Beltrami,1987).

Applied to the situation in the case company, such a system can be used to determine theoperatingcapacityoftheweighingscaleandfuelpumpastwoindependentunits,whicharealsoco‐dependentbecauseoftheirplacementinthetopologyofthecollectioncenter.Thesearethebasic arguments and assumptions that are based on an assessment of internal logisticsprocessesatthecollectioncenterofthecompany.Duetothistopography,thequeuingsysteminthe company is a two‐stage tandem queuing system with two stations and a finite bufferbetweenbothstations,where thescale is the firststationand theweighingunit is thesecondstation,interdependentonthescales.

3.ANALYSISOFINTERNALLOGISTICSINTHECASECOMPANY

Basicandalsothemainactivityofthecompanyiscollecting,processinganddisposalofwaste.Locationoftheircollectioncenterislessthan5kmoffthemainhighway,whichdoesnotcauselogisticalproblems.Thebiggestlogisticalproblemisrepresentedbyinternallayoutofbuildingsandlackofmaneuverabilityareasforlargevehicles.

In Figure 1, internal flow of transport is shown. Vehicles that enter the collection center areobliged toweigh in, except if they are empty. In this case, they turn left, past the scale.Uponcompletionoftheweighingprocess,certainvehiclesrefuel,othercontinuetothedestinationinthe center. The vehicles which are owned by the company must always refuel only at thecollectioncenterandnotonanyotherexternalfuelstations.

Figure1.Layoutofthecollectioncenterandinternaltransportflows(Source:Internaldatafromthecasecompany,2013)

2.1Refuelingandweighing

During thecompany'sdailyoperations,onaverage83vehiclesareweighed,while13vehiclesaveragely refuel. Towards the end of the month, that number rises for as much as 50%.Bottlenecksandlossesaregeneratedmostlyintheprocessofweighingandrefuelingthevehicle.Every vehicle in the center brings waste, must necessarily be weighed first. Because of this,driversmustwaitupto30minutestobeabletoweighthevehicleandmustnotcompleteanyotheroperationprior to that.Weighingvehicles takesonaveragenotmore than twominutes,meaning that the average waiting period should not be more than 6 minutes. The flow ofvehicles isstopped ifanyvehicle isrefueling.Dueto lackofspace in theareaof thescaleandrefuelingunit,vehiclescannotleavethescaleifavehicleisrefuelingandcannotevenpassby.

Thefirststepwastoanalyzelongitudinaldataconcerningweighingandrefueling.Table1showstherecordeddataonthenumberofrefuelsduringthemonthsofJanuary2013throughOctober2013withatotalsampleof272workingdays.Theaverageflowofthefueltankis70l/min,withanaverageofvehiclespumping155,4litersoffuelatatime.Theanalysisofrefuelingtimesin relation to the entire sample shows thatmost refueling occurs in two peaks.Most refuelsoccurinthemorningbetween6and7am,whenthevehiclesareleavingthecenter,andanotherpeakisinthehoursbetween14and17pmafterthevehiclesreturn.Agraphicalrepresentationis shown inFigure2.Additionally, in‐depth observations of internal logistics on the companylocation was performed in the period between 30. September 2013 and 6. October 2013.Analysisofobserveddatashowsthatthehighestaverageofarrivalsforrefuelingperhouris3vehicles,whichisintheperiodbetween14and15pm.Averageservingtimeforrefuelingwas5,25minutes.

Figure2.Distributionofrefuelingoccurrencesbyhoursoftheday

The same analysis was prepared for the arrival of vehicles to the weighing scale. ThisdistributionbyhoursofthedayisshowninFigure3.Thedatashowstheentireperiodof2013,i.e. 365 days. On the graph of the number ofweighings by hours it can be observed that thedensitydistributionisalittledifferentthanthedistributionofarrivalforrefueling.Mostoftheweighingsoccurbetween13to14hours,alittlelessinthehoursbeforeandafterthistime.Theaveragetimeforweighingis2minutes.

Figure3.Distributionofweighingoccurrencesbyhoursoftheday

Ifthetwodistributionsarecompared,itisevidentthatthegreatestpotentialforcongestioninfrontofthescalesisbetween13and15pm.

2.1Queuemodel

Foreach server,namely theweighing scaleand fuel tank, stochasticprocesses canbeused toanalyzeandoptimizetheservingprocesses.M/M/1modelofqueuingisthebasicmodelwithstochasticprocesses,basedonan infinitepopulationofcustomersandservingdiscipline"firstcome,firstserved".Thisisthecaseintheweighingandrefuelingprocessesinthecasecompany.Thevehiclewhich firstmovesonto the scalewillbeweighed first, and thevehiclewhich firstmoves to the fuelpumpwillbe the first to refuel. Sinceeachserver is treatedseparately, it issufficienttocalculatethebasicM/M/1model.

Forthecalculationofthequeuingmodel,thefollowingcalculationsareneeded(Cloud&Rainey,1998,p.66):

utilizationofservers;∙

(1)

probabilitythattheserversarebusy; , ∙!∙

∑ ∙!

∙!∙

(2)

averagenumberofwaitingvehicles∙ ,

(3)

averagewaitingtimeinthequeue∙

(4)

Strengthofarrivalsisobtainedfromasimplecalculationoftheaveragevalueofthenumberofarrivals,whichwasmeasuredonly duringpeak frequencies from6pm to 17pm, in thedaysfromMondaytoFriday.Intheobservedtimeperiod,279vehiclesperweekwereweighed.Inthesametimeperiod,88vehiclesperweekrefueled.Therefore,theintensityoftheinputcurrent–arrivalstotheweighingscalesis0,1,theintensityofarrivalstorefuelingis0,05.Usingthisdata,theinputsandoutputsofthequeuemodelwerecalculatedusingequationsaboveandareshowninTable1.

Table1.ResultsoftheM/M/1modelfortheweighingscaleandrefuelingstation

weighingscale refuelingstation

c=numberofservers 1 1

X=arrivals(vehicles/min) 0,1 0,05

Ws=averageservingtime(min/vehicle) 2 5,25

Ro=utilizationofservers(portion) 0,2 0,2625

C=probability,thattheserverisoccupied(portion) 0,2 0,2625

Lq=averagenumberofvehicleswaitingtobeserved 0,5 0,0934

Eq=averagetimeofvehicleswaitingtobeserved(minutes) 2,5 7,1186

Theweightinglocation,whichistreatedasserver1,hasa20%occupancyrate.Thecomputationshows that the scale could be even more burdened is the layout of the center would allowpassagepastthescales.Sincetheoccupancyoftheserverislessthan1(or100%),thelikelihoodthattheserverisbusyisthesameastheoccupancyrate.Inthecaseofanoccupiedserver,theaveragewaitingtimeis2,5minutes,theaveragenumberofwaitingvehiclesis0,5.

Fortheserver2,thefueltank,thesituationisverysimilar.Adistinctionismadeinrelationtothe arrivals of vehicles and occupancy of the server. The occupancy of the refueling server is26%.Alsoheretheprobabilitythattheserverisoccupiedis26%astheoccupancyislessthan100%.Averagenumberofvehicleswaitinginthequeueislessthan0,1.Fromthedataitisseenthat despite the smaller number of vehicles arriving to server 1, server utilization is slightlyhigherduetothelongerservingtime.Allofthesetimesareofferingareserveforoptimization:ifthevehiclemustwaitinlineforrefueling,theaveragewaitingtimeis7,1minutes.

Withthesecalculationsitwasprovedthatbothserversallowsforthesmoothfunctioningoftheinternal logisticsprocesses inrelation to thecurrent flow.Thedatagainedbyobservations inthe companyshowanopposite story.Driversare losing toomuch timewaiting foroneof theservers, theweighing scaleor the fuel tank, tobe free.From this it canbe concluded that themainreasonfordelaysinthecollectioncenteristheincorrectplacementofbothservers,i.e.onedirectlyaftertheother.

Thecollectioncenteradditionallyhasmorecircumstanceswhichpreventthesmoothunwindingofinternaltransport.Failuretocomplywithtrafficregulationsinthecenterisamajorproblemdue to lack of space for manipulation of vehicles. Since vehicles are driving in the oppositedirectionondesignatedone‐wayareas,thereissignificantcongestionandcertainvehiclesmustrecede backwards, which is not easy due to space restrictions and interruptions to other

businessoperations.Anotherproblemistheparkingofvehiclesandothermeansoftransportinplaceswherethisisexpresslyforbidden.

4.CONCLUSION

Moreandmorehouseholdsandproductionfacilitiesaregeneratingandmorewaste,whichhasto be removed and managed by specialized companies. An example of such a company wasdiscussed in this paper. The case company has a problem with internal logistics at theircollectioncenter.Atpresentthereisalotofovercrowding,sincetheamountofgatheredwastematerial is relatively large given the amount of available space. The biggest bottleneck of thecollectioncenterofthecompanyconstitutesthesystemofweighingandrefuelingthevehicles.

Currentlytheprocessofinternaltransporttakesplacesothatthevehiclearrivingtothecenterhastobeweighed.Theweighingscaleisnotnecessarilyfree,asitcanbeoccupiedbyavehiclewaiting foranothervehicle torefueldue to the topologyof thecenterwhere the fuelpump islocated immediately behind the scale. During this time there is a possibility that a queue ofvehiclesformsthatshouldbeweighedorrefueled.

Themainproblemofthepaperwastoidentifywhetherthereisapossibilityofreducingthesequeues with the given parameters. The company is aware that the forming queues bringunnecessarycosts,butdidnotknowhowtoevaluatethem.Theresearchshowedthatinthecaseofachangeoftopology,thequeueswouldmoreorlessceasetobeproblematic,sincetheservercapacitiesaresufficienttosupportthecompany’sneeds.

Moreover,someotheroptimizationsolutionsexist.Thefirstproposalistoconsideroutsourcingtherefuelingservicetoanexternalprovideronremotefuelingstationsoutsideofthecompany.Taking intoaccountall investment factors, fixedcostsand thepotential saleof the fuelpump,thissolutioncouldbringareductionofexpenses(excludingthefuel itself)ofasmuchas75%withinfiveyears.Thefollowingproposedsolutionistoupdatethesystemforweighinganddataentry. In this case, the use of automaticweighing systemwith RFID terminal is proposed. Bycalculating the savings, it can be assessed that it is possible to achieve 50% cost reductioncomparedwith the existing systemwithin five years. Both these solutionswould additionallycontributetothereductionofqueuingproblemsatbothserversandeventheirdiminishment.

REFERENCES

[1] Cloud, D.J. & Rainey, L.B. (1998). Applied Modeling and Simulation: An integratedapproachtodevelopmentandoperation.NewYork:McGraw‐Hill.

[2] Hudoklin‐Božič,A.(1999).Stohastičniprocesi.Maribor:Modernaorganizacija.

[3] Internaldatafromthecasecompany,2013.

[4] Tsiotras, G., Badr, H. & Beltrami, E. (1991). Blocking Probabilities for a Class of Two‐StationTandemQueuingModels.IIETransactions,23(4),pp.331–345.

EXPLOITATIONINDICATORSOFRETREADEDTIRESDEPENDINGONTHENUMBEROFRETREADINGS

SvetlanaDabić‐Ostojića,*,MomčiloMiljuša

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: The usage of retreaded tires is becomingmore common due to the effects that areachievedbythismethod.Inpractice,thedistancetravelledbyaretreadedtireisakeyparameterthat shows the effects of retreading itself. This parameter is not independent of the number ofprevious retreading,and,asa rule, it isa random variable.Thispaperpresentsa setofoutputresultsobtainedbythestatisticalanalysisofthedatabaseoftravelleddistancedependingonthenumberofpreviousretreadingbutalsoonthepreviousexploitationhistoryofthetire.Inpractice,theseanalysesmaybeofimportanceinmakingdecisionsaboutfurtherexploitation–retreadingoftires.

Keywords: retreaded tires, number of retreading, number of retreading of tires, statisticalanalysis.

1.INTRODUCTION

Tiresbelongtothecategoryofwearingproducts/partswhichappearduringtheexploitationofa vehicle.Once theworking lifeof a tirehas expired, it isusually classified intoa categoryofwaste,whichmeansthatitgetsitsindexnumber(191204)determinedaccordingtotheWasteCatalogue(Belgrade,2010).Whenitisdisposedonthegroundaswaste,thatis,atalandfill,itsdecomposition lasts for more than 150 years and as such it represents a threat to theenvironment(Hammondatal.(2009)).Oneofthetreatmentmodesofthesepartsofthevehicle,allowing to use them again – restoring their function after their tread has beenworn out, iscalledretreading–aprocedureconsistingofapplicationofanewtreadonaprepared/treatedtire(Hammondatal.(2009)andDabić‐Ostojić(2014)).

Theuseofretreadedtiresforcommercialroadvehicleshasbeengrowing,bothworldwideandin our country. The reason lies in a number of positive effects reached by this procedure.Retreaded tiresare cheaper than thenewoneswithequivalent standardandalmost identicalquality;retreadingoftiresleadstoimportantsavingsinmoneyrangingtoabout45%(costofaretreadedtirecomparedtothecostofanewone).Thisprocedurehasalsoapositiveeffectontheenvironment–about5loffuelisusedfortheprocessoftireretreadinginsteadof35lforthemanufactureofanewone(Dabić‐Ostojić(2014)andDabićatal.(2013)).

Aspecialgroupofproblemsconcernstheexploitationofretreadedtires.Regardingthisaspect,oneof themost importantparameters are thenumberof retreadings (NRT) and thedistancetravelledbyatire(TD)(ofanewtireandaftereachretreading).Itcanbededucedthatthereare

* cecad@sf.bg.ac.rs

few papers dealing exclusivelywith the decisionmaking during the exploitation of retreadedtires. This is the reasonwhy the authors of this paper have decided to devote themselves totheseissues.Inpractice,thequestionisoftenifretreadingisusefulandhowmanytimesitcanbedone andwhat decision shall bemade? It is especially important to bear inmind that thecostsoftiresareamongthehighestexploitationcostsrelatedtocommercialvehicles(Gavrićatal.(2009)andBoustaniatal.(2010)).

In order to decidewhether a tire shall be retreaded or not and if yes howmany times, it isnecessarytocarryoutappropriatestatisticalanalysesandobtainappropriateresultsrelatedtoitsexploitation(Dabić‐Ostojićatal. (2013)).Theaimof thispaper is to identifycertaincausalconnectionsconcerningtherelationshipbetweenTDandNRTonthebasisofasampleoftiresbelonging to a subsetof thevehicleparkof the largest company forpublic city and suburbanpassengertransportinBelgrade.

Havingthisinmind,thepaperincludesdifferentparts.Aftertheintroduction,thesecondpartofthe paper presents statistically processed data, mathematical expectations (µ) and standarddeviations (σ) for the subset of the above mentioned sample of tires with the highestparticipationoftiresaccordingtothenumberofretreadings(upto3)andTDclasses(from20to40.000km,from40to60.000kmand60to80.000km).TherehasbeenobservedthemutualdependencebetweenNRTandTDandtherehasbeendiscoveredthattherearesomedeviationsregarding this relationship which require more detailed statistical analyses. They have beencarriedoutwithin thispartof thepaperdivided into twounits.Within the firstunit,wehaveobserved the change of parameters µ and σ after the I, II and III retreading for the abovementionedTDclasses.Thesecondunitofthispartofthepaperincludesonlytheanalysisoftheparameter µ according to the different TD classes after the I and then after the II and IIIretreadingindividually.Thefourthpartofthepaperincludesaconclusionbasedontherealizedanalysiswhichhasalsoindicatedpotentialdirectionsofthefutureresearch.

2.STATISTICALPROCESSINGRELATEDTOASUBSETOFTIRES

When the treadofa tire isobsoleteordamaged, thedecisionrelated to itsnext retreadingordismissshouldbemade.BeukeringandJanssen(2001)assumethatTDsofretreadedtiresarethe sameas thenewones. Ferrer (1996)presumes that theTDof a retreated tire is reducedafter every retreading process and that total NRT also depends considerably on themanufacturer – it is not unlimited. In that case, there are two important issues, the first onerelated to the NRT by that moment and the second one concerning the TD during theexploitation.Itcanbeassumedthatthesetwoparametersarerandomvalues,sincetheydependon a number of factors (manufacturer, driving conditions, road quality, type/load of vehicle,modeofdriving…).Itisclearthattheseparametershavebeenobservedforhomogenousgroupsof tiresandconditionsof theirexploitation.As illustration, theTable1presents theresultsofstatisticalprocessingofTDofatiredependingontheNRTfortheanalysedsample.

Table1.AverageTDofatireasanewoneanddependingontheNRT

NRT averageTD(km)dependingonNRT

0(asanewone) 702561 352372 328303 369714 30738

TheTable1generallyshowsthattheincreaseofNRTcausesthedecreaseofTDofatire.Thiscorrelation can be described by the exponential trend with the correlation coefficient,R2=0.7586.

Figure1.Dependenceof(TD)bytireonNRT–exponentialtrend

The diagram of the Fig. 1. shows that TD decreases by the second retreading, then increasesslightly after the thirdone and thendecreases again. These oscillations have initiated amoredetailed analysis of the abovementioned dependence in order to obtain a basis for a betterdecisionmakingregardingretreadinganditwillbecarriedoutfurtherdowninthispaper.

3.DATAANALYSISFORTHESUBSETOFARETREADEDTIRESAMPLE

InordertosearchforthespecificitiesoftherelationshipbetweenTDandNRT,inthispartofthepaper we have carried out some additional analyses regarding the change of TD after acharacteristic NRT. The first analysis included the application of statistical parameters of thenumberoftires(µandσ)especiallyaftertheI,IIandIIIretreading,accordingtoTDclasses.Thesecondanalysisincludedachangeoftheparameterrelatedtothenumberoftires(µ)butonlyforcharacteristicTDclasses.

3.1Analysisofstatisticalindicatoraftereachretreading

Within this unit of the paper, the analysis has shown the frequencies of the number of tiresaccording to the TD classes depending on NRT. As input values, there have been used thefrequenciesofthenumberoftires(obtainedfromthedatabase),giveninTable2(Dabić‐Ostojić(2014)).

Table2.FrequencyofthenumberoftiresdependingontheTDaftertheI,IIandIIIretreading

TDafterIretreading(000km)TDofnewtire(000km) 20‐40 40‐60 60‐8020‐40 25 36 1040‐60 43 55 1360‐80 14 10 1 TDafterIIretreading(000km)TDafterIretreading(000km) 20‐40 40‐60 60‐8020‐40 22 17 240‐60 22 20 360‐80 7 4 0 TDafterIIIretreading(000km)TDafterIIretreading(000km) 20‐40 40‐60 60‐8020‐40 9 6 240‐60 4 6 060‐80 0 2 0

StatisticalprocessingofdatafromTable2hasprovidedtheparameters(µandσofthenumberoftires)givenintables3,4,5forthedifferentTDclassesaftertheI,IIandIIIretreading.ThesevaluesarepresentedbyadiagramontheFig.2.

y=60724x‐0.4575

R2=0.7586(exp.trend)

0

20000

40000

60000

80000

0 1 2 3 4

NRT

Td(km)

Table3.ParametersrelatedtoIretreading

TDafterIretreading(000km) µ(tires) σ(tires)from20to40 34.69 37.32from40to60 36.96 39.74from60to80 38.46 41.77

Table4.ParametersrelatedtoIIretreading

TDafterIIretreading(000km) µ(tires) σ(tires)from20to40 31.77 35.39from40to60 32.03 35.45from60to80 27.86 29.62

Table5.ParametersrelatedtoIIIretreading

TDafterIIIretreading(000km) µ(tires) σ(tires)from20to40 20.60 25.47from40to60 20.98 26.60from60to80 18.53 21.44

Figure2.µofNRTdependingonTD

In order to obtain additional basis for decision making regarding retreading, it has beennecessarytoanalysethestatisticalindicatorµ,thatis,itsvalues,butfortheabovementionedTDclasses.Thishasbeendoneinthefollowingunitofthispartofthepaper.

3.2AnalysisoftheparameterµaftereachretreadingforthecharacteristicTDclasses

Basedontheanalysiscarriedoutaftereachretreading,butseparatelyforeachoftheTDclasses,thetable6presentsthechangeoftheparameterµ.Itcanbeconcludedthatµofthenumberofretreaded tiresdecreaseswith the increaseof theNRT,namelywithineachof theTDclasses,whichisshownbythediagramsofthefigures3,4and5.

Table6.µoftires(forTDclasses)forI,IIandIIIretreading

TD(000km)µ(tires) afterIretreading

µ (tires) afterIIretreading

µ (tires) afterIIIretreading

from20to40 34.69 31.77 20.60from40to60 36.96 32.03 20.98from60to80 38.46 27.86 18.53

In order to test the change of the parameter µ there has been done a dependence analysisdescribingthisrelationship.TherehasbeenobservedthatforTDintheclassfrom20to40.000km (Fig.3.) this correlation (withR2=1) is bestdescribedby thepolynomial trendof the2nd

38.46

27.86

20.6 20.9818.53

34.69 36.96

32.0331.77

0

5

10

15

20

25

30

35

40

45

1 2 3

NRT

µ

µafterIretreaded(000km)µafterIIretreaded(000km)µafterIIIretreaded(000km)

degree.ThesametrenddescribesthecorrelationofµandNRTwhenTDisintheclassrangingfrom40to60.000km(Fig4.),aswellaswhenTDisintheclassfrom60to80.000km(Fig.5.).

Figure3.µofretreatedtires(TDfrom20‐40.000km)

Figure4.µofretreadedtires(TDfrom40‐60.000km)

Figure5.µofretreadedtires(TDfrom60‐80.000km)

4.CONCLUSION

Decisionmakingofatransportcompanydecidingwhethertobynewtiresorretreadtheusedonesisespeciallyimportantfortheuseofretreadedtiresthathavereachedthewearlimitofthetread.Therefore, it isnecessarytoknowtheexploitationparametersof tires,while theaccentshall be on the specificity of the change and relationship between NRT and TD. The resultsobtained in thispaperhave indicated important stochastic character of theseparameters, but

31.77

20.6

34.69

y=‐4.125x2+9.455x+29.36

R2=1

0

5

10

15

20

25

30

35

40

1 2 3

NRT

µµofretreadedtiresfrom20to40(000km)

Poly.(µofretreadedtiresfrom20to40(000km))

32.03

20.98

36.96

y=‐3.06x2+4.25x+35.77

R2=1

0

5

10

15

20

25

30

35

40

1 2 3

NRT

µ

µofretreadedtiresfrom40to60(000km)

Poly.(µofretreadedtiresfrom40to60(000km))

32.03

20.98

36.96

y=‐3.06x2+4.25x+35.77

R2=1

0

5

10

15

20

25

30

35

40

1 2 3

NRT

µ

µofretreadedtiresfrom40to60(000km)

Poly.(µofretreadedtiresfrom40to60(000km))

alsoahighlevelofcorrelationbetweenthesevaluesalthoughtheconcernedsampleoftireshashomogenousexploitationconditions.

Accordingtotherealizedanalyses,therehasbeenobservedthattherearecertaincorrelationsrelated to the decrease of the TD of tires depending on the NRT,which puts in question thehypothesisofBeukeringandJanssen(2001)sayingthatthesevaluesareindependentfromeachother(DabićOstojić(2014)),aswellasthehypothesisofFerer(1996)thattheNRTisunlimited.

Based on the results obtained by sample processing, it is concluded that in real workingconditionsofatransportcompanyitisnecessarytocarryoutcontinuousanddetailedstatisticalanalyses in this field. Suchadatabaseand itsprocessingwould indicate thebehaviourof therelevantparameters.

For other combinations of the above mentioned exploitation conditions (non‐homogenousvehiclepark,characteristicsofaroad…)thenecessaryanalysescouldbedoneinanappropriatemanner.Thiswouldensureanappropriatebasefordecisionmakingonretreadingforeachtireseparatelywhichisofextremeimportanceforrationalbusinessofatransportcompany.

ACKNOWLEDGMENT

ThisworkwassupportedbytheMinistryofEducation,ScienceandTechnicaldevelopmentoftheGovernmentoftheRepublicofSerbiathroughtheprojectTR36006,fortheperiod2011‐2015.

REFERENCES

[1] Beukering,P.J.H.andJanssen,M.,(2001),TradeandrecyclingofusedtyresinWesternandEasternEurope.ResourceConservationandRecycling,33(3),235–265.

[2] Dabić‐Ostojić, S. (2014), Logistički aspekti menadžmenta protektiranja pneumatika,Doktorskadisertacija,SaobraćajnifakultetUniverzitetauBeogradu,Beograd

[3] Dabić,S.,Miljuš,М.,Bojović,N.Vidanović,N, (2013),DecisionSupport forTheChoiceofTireManufacturer,FMETransactions,41(1),FacultyofMechanicalEngineering,Belgrade,Serbia,72‐76.

[4] Dabić‐Ostojić, S., Miljuš, М., Bojović, N. Glišović, N., Milenković, M, (2014) Applying amathematical approach to improve the tire retreadingprocess,Resources, ConservationandRecycling,86(1),107‐117.

[5] Ferrer G., (1997), The economics of tire remanufacturing. Resource Conservation andRecycling;19(4),221–55.

[6] Gavrić, P., Danon, G., Momčilović, V. i Bunčić, S. (2009), Eksploatacija i održavanjepneumatikakomercijalnihvozila,Istraživanjaiprojektovanjazaprivredu,Beograd,25,1‐10

[7] Katalogotpada‐Uputstvozaodređivanjeindeksnogbroja,RepublikaSrbija,Ministarstvoživotnesredineiprostornogplaniranja,Agencijazazaštituživotnesredine,Beograd,2010

[8] Hammond,P.,Lindquist,K.,Wendt,M., (2009),RetreadedTireUseandSafety,Synthesis,WashingtonStateDepartmentofTransport,USA

PartIX STUDENTS'PAPERS

CITYLOGISTICSOFBELGRADEWATERFRONT

MarijaJakovljevića,*,SlobodanDragojeraca,AleksandraVukovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract: The importance of city logistics is increasing with urban areas' sprawling anddevelopment. Considering that logistics activities have a great influence on the systemcompetitivenessand sustainability, it isnecessary topay specialattention to theirplanning.Thebeginning of „BelgradeWaterfront” project realization requires defining the plan of logisticsprocessesandactivities.Thispapershowstheproblemsandgivessomepotentiallogisticssolutionsinbothconstructionandplannedfacilitiesfunctioningphases.

Keywords:citylogistics,constructionlogistics,supplylogistics

1.INTRODUCTION

Beingacitysituatedontworivers,thepotentialofBelgradeisunderused.Thebottomlandvalueispriceless,butitisnotproperlyexploited.Inthepast,thevariousprojectswereplannedonthisarea and they were aimed at the development of more profitable business activities,modernizationandattractivenessgrowthofthecity.The„BelgradeWaterfront“ iscurrentandadopted reconstruction project of a part of Sava's bottomland. On the location of outdatedindustry,wheresomeof thenationalminorities foundhisharbourage, intheveryheartof thecity,Master plan envisages the construction of attractive residential and commercial facilities(http://www.rapp.gov.rs).Considering thatBelgrade isacitywitharichhistoryandbeautifularchitectural heritage, the certain number of buildings will be renewed within „BelgradeWaterfront“projectandwillretaintheiroriginalappearance.

The construction of the planned complex is a great challenge for the city logistics. The largeconstruction site, with numerous contractors and subcontractors, will generate the intensivegoodsandmaterialflowsforalongerperiod.Thelackofplanning,managementandcontrolofsupply flows and returnable andwastematerial extraction from this area could have seriousconsequences for the functioning of the whole city, especially of central and urban area ofBelgrade. On the other hand, the construction by phases of planned residential‐commercialfacilities allows their settlement by phases. The activating of various urban functions(residential, trading, catering etc.) will increase the logistics demand heterogeneity and theproblemsofpeopleandgoodsflowrealization.Bymovingthepeopleintobuildings,thedemandfor their supplying, i.e. the various goods, material and freight delivery, will emerged. Thedominant role of road transport in these flows’ realization and their adverse environmentaleffects,especiallyintermsofnoise,airpollutionandtrafficsafety,willcausethedeclineofurbansystemefficiencyandqualityoflife.Inordertoavoidthisscenario,itisnecessarytodefinethecity logistics conceptions which will allow the undisturbed construction work and efficient * marijlobb@gmail.com

functioningofnewlyconstructedsystems,providetheattractivenessofthesepartofthecityandimprove the quality of life in it. The conceptions should be defined in accordance with theinterestsofallparticipants,citylogisticsstakeholders(Zečević&Tadić,2006;Tadić&Zečević,2009;Tadićetal.,2014).

Thispaperaimstoperceivetheproblemsandgoals,i.e.theimportanceoflogisticsinurbanareaconstruction and functioning. The problems related to „Belgrade Waterfront” constructionlogistics and the supplying of planned facilities, aswell as thepossible solutions of them, arepresented.

2.BELGRADEWATERFRONTPROJECT

According to the project, “BelgradeWaterfront“ (Figure 1) covers an area of 1.85millions ofsquared meters and includes the space on right Sava’s riverside, from Railroad Bridge toBranko’s bridge. The project construction is planned to be implemented in four phases(http://www.rapp.gov.rs).

The firstphase includes the constructionof the space situated amongOld SavaBridge, riverSavaand futureSavaBoulevard.Two170mtallbuildingsanda shoppingmall,aswellas theotherresidentialandbusinessfacilities,areplannedtobeconstructedonthisarea.

ThesecondphaseincludesthespacebetweenBranko’sbridgeandOldSavaBridge,aswellasthespacebetweenGazelaBridgeandtheFair.Atthissiteitisbeingconstructedthefacilitiesofcultural and historic importance, aswell as the cycling bridgewhich alsomay have a role insupplyingthecitywitheasiershipments.

The third and fourth phases envisage the construction of space situated among Old SavaBridge,GazelaBridge,SavaStreetandfutureSavaBoulevard.Inthisphase,thevariouspurposebuildingsareconstructed,butthenumberoffloorsissignificantlysmallercomparedtothatinpreviousphases.

Figure1.“BelgradeWaterfront”project(http://www.beobuild.rs)

This project will engage numerous Serbian construction companies, improve the economicdevelopment and the city and region attractiveness, and its realization requires a seriousconceptofcitylogistics.

3.CONSTRUCTIONANDSUPPLYLOGISTICSOFBELGRADEWATERFRONT

Theallkindsofgoods,freightandmaterialflowsarepresentinthecity.Dependingonthetypeof generator running the flow, thequantities and formsof appearancevaryand the flowsarepresentpermanently,onceorseveraltimeduringadayoroccasionally.Duringtherealization,theflowsarepassingthroughthevarioussystemsandrequirevariousservices(Zečević&Tadić,2006).

Due to the high concentration of urban functions, the central urban areas generate thesignificantpartoflogisticsactivities,primarilythefreighttransport.Inthecurrentsituation,thegenerators fromthecentralareaofBelgrade initiate therunningofseveral thousandsof roadfreightvehiclesofvarioustypes,onadailybasis(Tadićetal.,2014).Theirpresenceinducesaseriesofeffectsonbothbusinessperformanceofparticipantsandenvironmentandqualityoflife.IntheattractivepartofSava’sbottomland,the“BelgradeWaterfront”projectenvisagesthedevelopmentofvariousresidentialandbusiness‐commercial facilitieswitharchitecturallyandvisually modern buildings. The project realization and the new plan require new logisticssolutions.On theonehand, thedemand forproviding the siteswith constructionmaterials isemerging,andalongwiththis,thesupplyingthefacilitiesthatwillbeactivatedbyphases.

3.1Constructionlogisticsof“BelgradeWaterfront”

Theconstructionlogisticsincludesplanning,organization,coordinationandcontrolofflowsandlogisticsactivitiesthatarerelatedtoconstructionprojectrealization(Duiyongetal.,2014).Thescopeofenvisagedworksandthesiteof“BelgradeWaterfront”project indicatetheadditionalcomplexityandproblemsoflogisticsinbothnarrowandbroaderareasoftheconstructionsite.

In thecurrentsituation, thehigh trafficdensity in thisurbanareacauseshugeproblemsfromthe aspects of both congestion and road safety and ecology etc. The street network is notadjustedfortheefficientfreightvehiclemanagement.Thisproblemisparticularlypresentinthemorning hours when the transport flows overlaps the people flows, thus resulting in theirmutualdisturbance.Therequestsforconstructionmaterialdeliverywillsignificantlyworsenthesituationbyrunningtheadditionalheavydutyvehicles.AspecialproblemisJIT(JustinTime)deliveries, which are dominant in the construction industry. In addition, it will emerge theproblemsofthematerialdisposal,i.e.ofitsstorageatthesite.Duetothehightrafficcongestionandwithaimtodeliverthegoodstimely,thevehiclecomesearliertothedeliverypointandwaitfor unloading, thus worsening the congestion and disturbing other road users. The efficientdistribution of the construction material requires the supply chain management and thecooperation between construction companies and suppliers and their adaptability to variouschanges that may occur (Ahmetasevic & Samuelsson, 2014). The lack of the planning,cooperation and coordination of the logistics flows and activities generated by futureconstruction sitemayhave serious consequences for the functioningofBelgrade’surban areaandfortheprojectrealizationefficiency.

In order to reduce the number of vehicle runs and traffic jams generated by freight vehiclesduring a construction material delivery to the construction site and to improve living andworkingconditions,itisproposedtointroducethecitylogisticsterminalfortheconsolidationofsmallermaterial deliveries. The participation of numerous contractors and subcontractors inproject realization involves the numerous daily deliveries of various materials at differentlocationswithinaconstructionsite.Thecityterminalhastobeabletoacceptallsmalldeliveries,toconsolidatethemanddeliverytoaspecificlocationaccordingtotheworkschedule,aswellastoprovidethetemporarystorageoftheconstructionmaterial.Inordertoavoidtrafficjams,thesupplierorconstructionmanagerhastoannouncethevehiclearrivalandthedeliverywhichisnottakingplacethroughthecityterminal.Basedonthedeliveryplan,eachsupplierisprovidedwithatimeintervalwithinwhichthedeliveryshouldberealized(Zečević&Tadić,2006).

This system allows the coordinated delivery flows management and timely supply ofconstructionsites,withouttheadditionalcongestionwithinandneartheconstructionsite.

There are numerous examples of city terminals supplying a construction site, e.g. in London(Transport for London, 2008) and Stockholm (Ottosson, 2005). The stated advantages of theconsolidationcentreinvolve(TransportforLondon,2008):

Reductioningoodsandvehiclemovementsattheconstructionsite Reduction in deliveries realized by smaller vehicles; the higher‐capacity vehicles are

used AbsorptionofCO2,noiseandvibrations Increaseofthesupplynetworkcapacity Deliveryreliabilityimprovement Savingsfordrivers;theydeliverthegoodstotheconsolidationcentreratherthantothe

constructionsite Savingsforcontractorswaitingforthedelivery Reductioninthewastematerial Wastematerialandpackagingarecollectedinthecentre Centreprovidestheopportunityforrepairsandrecyclingservicesetc.

However, the successful implementation of the consolidation centre concept requires(TransportforLondon,2008):

Trainedandinformedlabourforce;tendencytotheerrorelimination Controlofdeliveredgoodsfromtheconsolidationcentretotheconstructionsite Insuranceofgoodsifitisprofitable Temporarywarehouseconstruction

By studying the disadvantages, the clearer picture of the opportunity to implement aconsolidation centre concept for the construction logistics realization is obtained. Thesedisadvantagesaremanifestedbythelackofknowledgeinpeoplethatwillbetheworkleadersattheconstructionsite.Consideringthataspect,itshouldemphasisehowitisnecessarytoinvolvethelogisticsthinking(TransportforLondon,2008).

3.2Supplylogisticsof“BelgradeWaterfront”

Asacapitalcity,Belgradeisanadministrativeandeconomiccentrewhichmakesitthelargestgenerator of goods flow. Although, the problems of logistics activities and goods flows aresolvingpartiallyandindividually,withoutconsideringtheentirecitylogisticssystemandplan.Theincreaseinthenumberofgeneratorsintheurbanareaofthecityandtheconstructionof“BelgradeWaterfront”willresultinsignificantworseningofthelogisticsproblemsandadverseeffectsontheenvironmentandqualityoflife.

Supplyandextractionlogisticsisessentialfortheurbanareaprosperity,butsimultaneouslyitisa source of problems from the aspect of the environmental protection, traffic safety andavailability.Increasingrequirementsintermsofthespeed,flexibility,reliabilityanddiversityoflogisticsservicesandthe lackofplanningactivitiesand long‐termlogisticsplans influencetheincreaseofcommercialvehiclevolumeandthe lossofurbanvitality.Withoutappropriatecitylogisticssolutions,thesupplyingofnew“BelgradeWaterfront”facilitieswillworsenthecurrentsituation.Ontheotherhand,themodernurbanconceptoftheurbanriversideareadevelopmentrequires the modern logistics concept. Herein defined conceptions are compatible with thedevelopment plan of this attractive urban area and involve the implementation of new, oldsolutions for the logistics chains realization in the city. Inorder tomeet thenewdemand, i.e.plannedbusiness‐commercialfacilitiesintheattractivepartofcentralurbanareas,butalsothesolutionsofthecurrentproblems,twologisticsconceptionsaredefined:

Conception1:Constructionofacitylogisticsterminalforconsolidatedsupplyofgeneratorsinthegravitationarea.Thegoodsdeliveryfromlogisticscentres,situatedatotherlocations,tothecitylogisticsterminalwouldberealizedbyimplementationofcargotramway,whiledistributionto the generators in the gravitation area would be realized by electric vehicles. The mainadvantages of this conception involve (Tadić & Zečević, 2009): delivery consolidation andminimumofvehiclemovements;cargotramwayimplementationforreducingtheurbanstreetnetwork congestion, thus having a positive impact on traffic safety, air pollution and roadmaintenance costs; the use of electro vehicles significantly reduce the adverse environmentaleffects. Besides these significant advantages, this conception has some disadvantages. First, apartofgoods(perishablegroceries,dailynewspaperetc.)hastobedeliveredintheconventionalmanner,byconventionalroaddeliveryvehicleswithouttranshipmentandconsolidation(Tadić&Zečević,2009).Inaddition,thecargotramwayimplementationimpairstheflexibilityofa“JustInTime”deliveryduetotheuseofexistingfixedinfrastructure(Robinson&Mortimer,2004).

Conception 2: Construction of underground logistics system for supplying “BelgradeWaterfront”.Thegoodswouldbedeliveredtothelogisticscentreattheedgeofthisurbanareaaccordingtothecitylogisticsconceptionofthewholecity.Thelogisticscentreisalinkwiththeenvironment and offers the storage, sorting and commissioning services as well as thepreparation of units for the underground system and delivery to the generators. The systemwould have several stations for accepting/shipping the goods. The flows between the stationandgeneratorwouldberealizedbynon‐motorizedtransportmodes(bywalking,cyclingorbyhandcarts).Theenvironmentaladvantagesofundergroundingfreighttransportarenumerous:streetnetwork clearing, reduction in traffic congestion, energy consumption, hazardous gasesemissions, noise, traffic safety improvement and more rational use of the existing space. Inaddition, there are significant advantages for the logistics: faster delivery, less damage to thegoods, weather conditions not undermining flow realization. Besides all the advantages, thissystem implementation requires a long construction time and large investments, so it isnecessarytodetermineitsjustifiability.

Considering that both conceptions have some advantages and disadvantages, the final choicerequires more detailed analysis and estimation from the aspect of numerous criteria. Thecriteriashouldbedefinedinaccordancewitheconomic,environmentalandsocialsustainabilityindicators.

4.CONCLUSIONS

Acityisaconcentrationpointofvariousfunctionsandsocial‐economicsystems.Thesuccesfulfunctionning requires planning of logistics systems and activities and defining city logisticsplans.Thecitylogisticsshouldprovideanefficientandenvironment‐friendlyrealizationofthegoods flows. “BelgradeWaterfront” generate new logistics demand in both construction andplanned facilities’ activation phases and requires solving numerous logistics problems.Inadequatelogisticsorganizationwillleadtotheseriousdisturbancesinthefunctioningofthisareaandthewholecity.

ACKNOWLEDGMENT

Thispaperwaswrittenon thebasisof theseminarwork for thesubjectCity logistics.Wearegrateful to Prof. Slobodan Zečević i Asst. Prof. Snežana Tadić for their trust, consulations andsuggestionsduringtheelaborationofthispaper.

REFERENCES

[1] Ahmetasevic, M., Samuelsson, S. (2014). Management of construction logistics inStockholm, Identifyingways of improvement for construction logisticswithin the innercity of Stockholm, The Faculty for Technology, Construction engineering, UppsalaUniversity.

[2] Duiyong,C.,Shidong,J.,Mingshan,S.(2014).Engineeringconstructionprojectsitelogisticsmanagement,JournalofChemicalandPharmaceuticalResearch,6(7),353‐360.

[3] Ottosson,M. (2005).Evaluation report‐NewConcepts for theDistributionofGoods (WP9).TrendsetterReport,7,EnvironmentandHealthAdministration,Stockholm.

[4] RepublicAgencyforSpatialPlanning,http://www.rapp.gov.rs,(Accessed:15.04.2015.).

[5] Robinson,M.,Mortimer,P.(2004).UrbanFreightandRail.TheStateoftheArt.Logistics&TransportFocus.JournaloftheInstituteofLogisticsandTransport,6(1),pp.46‐51.

[6] Tadić, S., Zečević, S. (2009). Izbor optimalnog scenarija razvoja logističkog sistema. In:ProceedingsofSYM‐OP‐IS2009,333‐336.

[7] Tadić,S.,Zečević,S.,Krstić,M. (2014).RankingofLogisticsSystemScenarios forCentralBusinessDistrict.Promet–Traffic&Transportation,26(2),159‐167.

[8] Transport for London. (2008). London Construction Consolidation Centre Final Report.TransportforLondon.

[9] Zečević, S., Tadić, S. (2006). City logistika, Faculty of Transport andTraffic Engineering,UniversityofBelgrade,Belgrade.

CONSTRUCTIONLOGISTICSOFBELGRADEWATERFRONT

MarkoStokića,*,BoškoRadovanovića

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:“BelgradeWaterfront”istheoneofthemostimportantprojectsofthecityofBelgradeand currently the largest investment project in this part of Europe. The undisturbed projectrealizationrequiresseriousplanningandmanagementofthecomplexandintensiveflowsofgoods,freight,material,peopleandinformation,aswellastheflowsofvehiclesandmechanization,soitisthe major challenge for the city logistics and construction logistics. This paper describes theconstruction logisticsproblems of “BelgradeWaterfront”and possible solutions.The location oftemporary logistics centres, theproblemsof supplying the construction siteswithmaterials, thesupply chainmanagementand the roleof the reverse logisticduring theproject realizationareconsidered.

Keywords:constructionlogistics,deliveryflows,reverselogistics,logisticsconsolidationcentre

1.INTRODUCTION

The construction logistics involves all technologies, rules, methods, knowledge and solutionsrelatedtothemovementofmaterialandothertypesofresourcesinconstructionflows(Duiyongetal.,2014).Fromtheaspectoftheprojectrealizationscheduling,itcouldbedividedintothreesegments(Duiyongetal.,2014).Thefirstsegmentisrelatedtothelogisticsbeforestartingtheconstructionandinvolvessupplyingaconstructionsitewithnecessarygoods,labourforceandmechanization to start the work. This construction logistics’ segment is very important andallowsundisturbedconstructionstarting.Thebasicproject is followingby the logisticsduringthe construction, where the good communication among all construction site sectors is ofparticularimportanceforefficientlogisticsflowsrealizationandundisturbedconstruction.Thethirdconstructionlogistics’segmentisrelatedtotheperiodafterthecompletionofconstructionworks and involves the activities related to the removal of various types of waste for theirproperdisposalatlandfills,recyclingandotherrenewaloptions.Thispaperpresentsthegeneralprojectoverview,constructionlogisticsproblems,possiblesolutionsandthelogisticsconceptof“BelgradeWaterfront”project.

2.“BELGRADEWATERFRONT”PROJECT

According to the project, “Belgrade Waterfront” covers an area of 1.85 millions of squaredmetersandenvisagestheconstructionofabout200facilitiesforvariouspurposesandfunctionsontherightriversideofSava,betweenRailroadBridgeandBranko’sBridge,uptoKaradjordjevaand Savska Street. The project implementation was conceived to be realized in four phases

* markostokic12@yahoo.com

(Figure1).Thebiggestpartofthearea,about60%,isplannedfortheconstructionofresidentialunits, 17% for commercial properties, 8% for hotels and a shoppingmall, separately, 5% forshopsandabout1%ofplannedareasforculturalandleisureactivities.Thebuildingsstandingout are “Belgrade Tower”, 170 m tall, and the shopping centre which will cover more than148,000 of squaremeters.When it comes to the street network, it was planned to keep theexisting street matrix which will be linked by newly constructed boulevards. The projectappreciatestheculturalandhistoricheritageofthecityofBelgrade,thussomebuildingswillnotberazedduringtheworks(http://www.rapp.gov.rs).

Figure2.Phasesof“BelgradeWaterfront”projectrealization

3.CONSTRUCTIONLOGISTICS

Theconstruction logistics involvesplanning,organization,coordinationandcostsof flowsandlogisticsactivitiesthatarerelatedtotheconstructionprojectrealization(Duiyongetal.,2014).The large construction project realization requires precise and clear organization of alloperationsattheconstructionsite,implyingdevelopedsupplychain.Thesupplychainincludesall logistics activities, production operations and process coordination management andactivitiesrelatedtomarketing,sale,productdevelopment,financeandinformationtechnologies,buttheirrelationshipstoo,aswellasthecoordinationandcollaborationamongpartnersbeingin the same chain (Miljuš M., 2015). The supply chain may have a different impact on theconstructionsitedependingonthepointofview(VrijhoefR.&KoskelaL.,2000).Itispossibletoconsider the inherent relationshipbetween the supply chainandconstructionsite, the supplychainasanindependentunit,activityshiftingfromtheconstructionsitetothesupplychainandtheconstructionsites(VrijhoefR.&KoskelaL.,2000).Theexpertsinthisfieldhaveattemptedtodefine the main problems with construction work realization (Rogers P., 2005.): underusedloading space of vehicles, waiting at the transshipment front, qualified staff engaged forunqualified tasks such asunloading,material areoften transferred fromone to anotherplacewithintheconstructionsite,poorlyplannedworktimescheduleforprofessionaltasks,thelargeamount of material and waste and very poor training of logisticians for the work in theconstructionindustry.Figure2showstheproblemsintheconstructionprocess.

Poor logistics implementationmaycausea seriesof consequencessuchas:unnecessarycosts,poor image of the construction industry, poor constructionquality, prolongeddesigning time,additionalhealthandsecurityrisksetc.(RogersP.,2005).Ontheotherhand,byimplementationof good logistics themanypositive effects couldbeachieved, suchas: reduction inmovementvolume,reductionincredit‐linkedcapital,reductioninwasteamount,rapidconstructiontime,improved quality, lower health and security risk of theworkers,more efficient use of labourforce results in construction cost decreasing and, generally, in the improvement of industry’simage (Rogers P., 2005). The good logistics can reduce the unnecessarymaterialmovements,shorten the production process and increase the productivity up to. It is possible tomonitorsystem performances (both quantitative and qualitative) in two ways: 1) observation andmonitoring of systems and its functioning; (2) comparison of implemented solutions, i.e.

monitoringhowmuchtheappliedsolutionsareappropriateandtowhatdegreethey improvethesystem(Wegelius‐LehtonenT.,2001).

Figure3.Problemsintheconstructionprocess(VrijhoefR.&KoskelaL.,2000)

When designing, it is possible to use “Last Planner” method. The reasoning is to meet allprerequisites that are necessary for performing various construction tasks before these tasksare assigned to aworking group.By thismethod, all the tasks related to theproject couldbeclassifiedintofourcategories(Ala‐RiskuT.&KarkkainenM.,2006):

SHOULD– tasks thatneed tobeperformed in thenear futureaccording to theoverallprojectplan.

CAN – tasks that have all their prerequisites ready: e.g. previous project steps arecompleted,necessarymaterialsareathand,andworkforceisavailable.

WILL–thetasksthatarecommencedbeforethenextplanninground. DID–thetasksthatarecompleted.

4.CONSTRUCTIONLOGISTICSOF“BELGRADEWATERFRONT”

Theconstructionof“BelgradeWaterfront”involvestheprocessesoftransport,storage,loading,unloading,packagingandstockmanagementrelatedtonumerousofvarioustypesofmaterialsandequipment. It isnecessarytoprovidethatallactivitiesare takingplace inasimple,quick,continualandcontrolledmanner.Inordertoprovidematerialavailabilityinthemomentoftherequestoccurrenceattheconstructionsite,itisnecessarytodevelopbothlong‐andshort‐termplans for the material delivery which could be adjusted and upgraded during the projectrealization.During thisperiod, eachconstructionsector isobliged to submit a listof requiredmaterialsforthenextperiod(day,weekandmonth)tothelogisticcentre.Inordertomakethissystemfunction,itisnecessarytohaveawell‐developedsupplychain.

4.1Supplychain

Theimplementationofthesupplychainintotheprojecthasmanyadvantages:shortertimeoftherequestrealization,preciseamountofmaterial fortheproduction,decreasinginstockandtrappedcapitallevel,quickrealizationofthelogisticsprocesses,electronicdataprocessingandexchange,qualityofserviceimprovement,reductionintotalcostsetc.Consideringthatthisisa

long‐term project, it is necessary to make contracts among all participants in “BelgradeWaterfront”constructionprocessanddefinetheirrightsandobligationsinordertosecuretheundisturbedsupplychainfunctioning.Tomakesupplychainfunction, it isprimarilynecessarytomakeaspecialITplatforminordertoprovideahigh‐levelcommunicationanddataexchangeandprocessing.Thematerialandinformationflowshavedifferentmovementdirectionswithinasystem.Toconsolidatesmallerdeliveriesandprovidethetimelymaterialdeliverytodifferentlocations within a construction site, it is proposed to implement the construction materialconsolidationcentre.Itsimplementationistemporaryanditisdeterminedbyprojectduration.

4.2Thelocationoftemporarylogisticscentre

After detailed analysis of the project and schedule of “Belgrade Waterfront” realization, theauthors identified some possible location of the logistics consolidation centre. Based on theestimation by several criteria (the available surface area, time distance from all constructionphases, transport infrastructure development, connection tomain roads etc.), the location offuture “BelgradePark” is proposed.However, the location of the centremayvary indifferentphases of the project construction if it is economically justified. The chosen location covers40,000m2 of space, occupies the central position on theprojectmap, being situatednear theriversideand connected toother constructionphasesbymain roads.Within the centre, therewill be the entry/exit check point, administration building and warehouse for constructionmaterial,parking facility for freightandconstructionvehicles, fuel tanks,maintenancesystem,the space for waste and unused material disposal. The planned area for the logistics centreinstallationisalsoconvenientfromtheaspectofthearrangementafterthecompletionofwork.Nevertheless, the choice of a logistics centre location requires quantitative and qualitativeanalysisofpotentiallocationsfromtheaspectofalargernumberofcriteriaandtheapplicationofoperationalresearchmethods.

4.3Typesandamountsofmaterial

Withtheaimof flowvolumeestimationand for thepurposeof thispaper,onlybasicmaterialcategories are considered in the facility construction process. Besides the steel poles, whichpresents the frame and gives shape to the facility, this paper involves steel floor coverings,throughwhichtheconcreteispouredforthepurposeofthefireprotection,armature,bricksforroom separation, sheetmaterial (marble, ceramic tiles, parquet) andwindows that provide afuturisticlooktothebuildings.Thematerialamountsareestimatedbyphases(Table1),basedonthescopeofwork,i.e.theplannedareainm2.

Table11.Estimatedmaterialamountsfor“BelgradeWaterfront”construction

TypeofmaterialEstimatedscope

Phase1555.000m2

Phase2832.000m2

Phase3278.000m2

Phase4185.000m2

Total1.85mil.m2

With5%ofreserve

Steelpoles(t) 120.000 39.000 14.500 21.000 203.000 215.000

Floorcoverings(t) 30.000 3.500 1.300 2.300 37.960 40.000

Concrete(m3) 2.175.000 330.000 66.000 135.000 2.724.000 2.860.000

Armature(t) 110.000 16.500 3.500 6.700 137.400 145.000

Brick(t) 640.000 80.000 30.000 53.000 822.000 863.000

Sheetmaterial(m2) 2.100.000 245.000 91.000 164.000 2.660.000 2.795.000

Windows(m2) 95.000 24.000 8.900 16.000 143.900 151.000

Among estimated amounts, the dominantmaterial is the concrete,whichwill be used for thepurposeofbuildingstability,panelingofsteelpoles,formingoffloorconcreteslabsandpublicroads too. The large amounts of concrete and armature will be used in the construction ofcentralsupportpillarofBelgradeTower,onwhichthesteelframewillberelied.

Besides timelymaterialdelivery, thesuccessfulprojectrealizationrequires thedeliveryof theconstruction mechanization such as: earthmoving machines (bulldozers, excavators, loaders,levelers, vibratory rammers, rollers and pavers), lifting (cranes and elevators) and transportmachines(trucks,tippers,dumpers)andmixers(towingmixerandtruckmixer).

4.4Problems,possiblesolutions,directionsandmethodsofmaterialdelivery

Itiswell‐knownthatthecityofBelgradehasdifficultieswithtrafficjamsonthebridgesandthestreet matrices in the old part of the city are mainly narrow and unidirectional, with theprohibitionoflargetransportmeans’movements.Nightdeliveryisnotfeasibleforallflows,sothe construction site supply will be a large traffic problem. Since “Belgrade Waterfront” isspatiallyreliedonrightriversideofSava,oneofthepossiblesolutionstotheproblemistheuseoftherivertransport.Bychoosingthisdeliverymethod,itispossibletoenabletheunloadingofstreetnetworks,withoutvehicleaccumulationandcongestionoccurrence,aswellasthetimelydelivery of large amounts of the material by reduced number of transport means (one pushbarge‐barge system replaces dozens of road vehicles). In addition, the transport cost willdecreaseandtheenvironmentalpollutionwillbesignificantlyreduced.

The use of river transportwas also justified by analysing the potential directions ofmaterialdelivery: Smederevo Ironworks (steel materials), factories from Pančevo producing concreteandglass,brickproductionplant fromNoviSadandsheetmaterialmanufacturer fromŠabac.Forthedeliveryoftheconcrete,itisalsopossibletouserivertransportbychoosingthevesselsfortransportofthewholeroadvehicles(mixersandtruckmixers).

It is possible to establish the connection between the riverside and consolidation centre byinstallation of an adequate type of belt conveyor. For the heavy‐weight and irregular‐shapematerials, it ispossible tousecranes for reloading intoroad transportmeans forbulky loads.Nevertheless, the choice and use of transport and transshipment systems, as well as theopportunitiesforvesseldocking,requireadetailanalysis.

4.5Reverselogistics

The reverse logistics has a very important role in the construction project realization. Largeconstructionsitesgeneratealargeamountofwastewhichcouldcausemanyproblems(Nunesetal.,2009).Thewastematerialcouldbeclassifiedintothreemaingroups:theinertconstructionwaste,which isdisposedat landfills, the recyclingwaste (glass, steel, ironetc.) and thewastewhich is the overage in some facility construction, but it could be temporary reused in someotherfacilityconstruction(marbleslabs,windows,parquet,rodmaterialsetc.).Wastecollectionfromtheactiveconstructionsitescouldberealizedperiodically,byspecialvehicles.Thewastewould be disposed at temporary landfill near the consolidation centre. With the aim of thevehicle movement optimization when delivering material from the consolidation centre andcollectingthewaste,itispossibletoapplythetravellingsalesmanmethodinordertominimizethe distance and, consequently, the collection time and transport costs, and tomaximize thecapacity utilization of vehicles. After collecting a priori determined amount of waste, it isperformed its ad‐hoc removal, i.e. dial a removal. This method of waste removal from thelogisticscentreisgoodbecausethetransportcapacityofvehiclesusedfordeliveryismaximallyexploited.Sincetherearevariousgroupsofmaterialat theconstructionsite, it isnecessarytosort them, in terms of separating inert and recycling materials. It is probably better to sortmaterialsattheconstructionsitebyplacingcontainersfordifferenttypesofwaste,inordertoreduce time interval until the later further sorting. Thematerialwhich could be used for the

constructionofotherfacilityisstoredintheconsolidationcentreuntilthemomentofrequestingits re‐delivery to theembedding site.Besides the reverse logisticsof thematerials, the returnlogisticsoftheconstructionmechanizationmustnotbeforgottenafterthecompletionofwork.

5.CONCLUSIONS

The construction logistics faces numerous technical and organizational problems, as well asenvironmentalproblems,whichinfluencethecostsandconstructiontimeandquality.Inlargerconstruction projects, such as “BelgradeWaterfront”, which includes a broad scope of work,numerous contractors and extensive and complex material flows, an inadequate logisticsconceptandalackofcoordinationresultinseriousdisturbances.Thelogisticscentralizationandsupply chain introduction implies the reduction in total costs of project, quality improvementand shorter time for completion of the construction.Numerous studies have showed that themost of the problems in the supply chain occur in the previous phases of the chain. Theinformation exchange among the participants is often very poor, resulting in poorsynchronization of construction project’s processes and activities. The solution to theseproblems lies in cooperation among all participants in the construction project and incoordinationofallactivitiesandprocessesinitsrealization.Oneofthemainstrategieswouldbethe integration of basic logistics function in the supply chainwithmarketing, production andfinance.

ACKNOWLEDGMENT

ThispaperwaswrittenonthebasisoftheseminarworkforthesubjectCitylogistics.Wewouldlike to thankProf.SlobodanZečevićandAsst.Prof.SnežanaTadić, lecturers, for their supportduring the entire paper development, for sharing their professional experiences and forassistanceinsolvingproblemsthathavearisenduringthepaperdevelopment,thusenablingtherealizationoftheproject.

REFERENCES

[1] Ala‐Risku,T.,Karkkainen,M.(2006).Materialdeliveryproblemsinconstructionprojects:Apossiblesolution.Int.J.ProductionEconomics,104,19‐29

[2] Duiyong,C.,Shidong,J.,Mingshan,S.(2014).Engineeringconstructionprojectsitelogisticsmanagement.JournalofChemicalandPharmaceuticalResearch,6(7),353‐360

[3] Miljuš, M. (2015). Lanci snabdevanja, Faculty of Transport and Traffic Engineering,UniversityofBelgrade,Belgrade.

[4] Nunes, K.R.A., Mahler, C.F., Valle, R.A. (2009). Reverse logistics in the Brazilianconstructionindustry.JournalofEnvironmentalManagement90,3717‐3720

[5] RepublicAgencyforSpatialPlanning.http://www.rapp.gov.rs/(Accessed:20.02.2015.).

[6] Rogers, P. (2005). Improving Construction Logistics. Report of the Strategic Forum forConstructionLogisticsGroup.GreatBritain.

[7] Vrijhoef,R.,Koskela,L.(2000).Thefourrolesofsupplychainmanagementinconstruction.EuropeanJournalofPurchasing&SupplyManagement,6,169‐178

[8] Wegelius‐Lehtonen,T.(2001).Performancemeasurement inconstruction logistics. Int. J.ProductionEconomics,69,107‐116

HOSPITALLOGISTICS

AnđelaĐapića,ŽeljkoNovakovića,PredragMilenkova,*

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Hospitaldecisionmakersarefacingvariousdifficultiesarisingfromunpredictabilityofpatientsandtheirarrivaltimemix.Thecomplexityofflows‐specificonlyforthisindustrybranch–alsocontributes that.Moreefficientbusinesswith lowerstocksandcosts,aswellaswithqualityservices could beachieved through organizedanddeveloped logistics,which could bemanagedfromseparatesector.Properlymanagementofthissystemcouldbeachievedonly ifallflowsandactivitiesareanalyzed together.Thispaper covers someof thehospital logisticsdifficultiesandmeasurestoovercomethem,withspecialemphasisonthesituationandproblemsinClinicalCenterofSerbia.

Keywords:ClinicalCenterofSerbia,logistics,hospitalflows

1.INTRODUCTION

The basic function of a hospital, as one of the more important systems in every country, isprovidinghealthcaretothecitizens.Taking intoaccountallavailableresourcesandaimingtomoreefficientbasicfunctionrealization,everyhospitalneeds logistics.Thenumerouskindsofflows are present in a hospital every day, but unlike manufacturing industry, here it is notpossibletopredictthepatientmixorthedemandforparticularmaterial,implyingverycomplexlogistics. The complexity of activities, flows and participants in hospital instituions’ logisticsrequires extensive research for service improving and cost reduction. The literature containsdifferenthospitallogisticsdefinitionsandexplanationsbywhicheitherthetraditionallogisticsdefinitionisjustmappedonhospitalsystemsorthedifferentformsofhospitallogisticsareseenasseparatemanagementvaluesandareas.

While reviewing a literature, the different definitions of hospital logistics are observed. Thus,Aptel and Pourjalali (2001) suggest that logistics activities in hospitals include purchase,receiving, stock management, information system management, food service, transport andhomecare.

Logisticsisavitalpartofahospitalthatisinchargeofpurchase,receiving,stockmanagement,informationsystemmanagement,telemedicine,food‐relatedservices,transportandhomecareservices(Kriegeletal.,2013).

In its comprehensive meaning, hospital logistics includes purchase management and allplanning‐relatedtasks,implementationandadministrationofagreementsandmethodsleadingtogoal‐orientedflowsofobjects,valuesandinformationconcerninggoodsandservicesrequiredwithinahospital(PieperandMichael,2008).

* predrag.milenkov@yahoo.com

This paper covers hospital system characteristics, its inherent flows and factors that areimportant for its functioning. Certain fields, which are commonly considered in researchconductedinseveralcountries,areextracted.TheseparatechapterisdedicatedtothestateoflogisticswithinClinicalCenterofSerbia(CCS),distributionorganization,stockmanagement,aswell as to the significance of certain systemswithin this complex, for whose elaboration theinformation were collected by conducting interviews with Central Pharmacy and OrthopedicInstituteemployees.

2.HOSPITALLOGISTICS–FLOWSANDPROBLEMS

Hospitallogisticsischaracterizedbyhigh‐leveldivisionoflabour,non‐standardprocessesandalackofrelevantinformation.Duetotheseproblemsdecisionmakersinhospitalmanagementarefacingthechallengeofensuringresourceavailabilityeveryday,atanytreatmentplace,andofconstantimprovingofhospitalservicesconsideringthecapital,efficiency,costsandhealthcarequality.Oneofthemainsectorsforensuringresourceavailabilityisthehospitalsupplysector.Threefactorsarecrucialforserviceredesigningandimproving:costs,customerneedsandthequalityofserviceprovided.Consideringthathospitalsshouldbefocusedontheirbasicactivity,the secondary and tertiary services should be outsourced and this is the trend in otherindustriestoo.

Hospital complex flows are characterized by extreme complexity and they could be classifiedintogoodsflowsandpeopleflows(Fig.1).Besidesthepatients,peopleflowsincludeemployeeflowsandvisitorflows.Withinthisgroupitispossibletosingleouttheoperatingroomlogistics,emergencylogistics,logisticsofpatientadmissionanddischargeandhealthcarelogistics.Goodsflows include flows of medical (pharmaceutical products, medical material, instruments anddevices, blood and organs for transplantation, laboratory samples) and non‐medicalmaterial(food,hygieneitems,clothingandlaundry,bedsandfurniture,administrativematerials,variouswastecategories).Thepatientflowsarethemainflows,whicharethedriversofallotherflowsandactivities.Thearrivalvariabilitydeterminesmaximumandminimumpointsofdemandforhospital resources, medicines, operating supplies etc., creating queues, delays, as well as thestresstohospitalstaff(Noonetal.,2003;HaradenandResar,2004).

Figure4.Divisionoflogisticsbyfields(Kriegel,2012)

Patientflowlogisticsdealwiththeirmovementthroughthedifferentpartsofahospital(suchasoperatingrooms,emergency,dispensaryorambulance)fromthemomentoftheadmissionintothehospital to theirdischarge.Vissers (1998) suggested that following fourproblemsare themain causes of failure in patient flow management: (1) a lack of coordination of differentdepartments within hospital, (2) variability, (3) a lack of, and (4) inadequately allocatedcapacities.Inaddition,Butler(1995)andHaradenandResar(2004)consideredqueues,delaysandreferringpatientstoinappropriateplace,whileVillaetal.(2013)consideredthebottlenecksalongthewholechainwithinahospital,which interferewiththepatient flowandtheelectiveandemergencycaseoverlapping.

3.REVIEWINGTHERESERACHONHOSPITALLOGISTICS

Besides the patient flow analysis, the research on stock, material management and theirdistribution,partnershipsandstrategicalliancesofhospitalandsuppliersalsocouldbefoundintheliterature.Thevariousauthors’opinionsonthesesubjectswillbepresentedinthefollowing.

PanandPokharel(2007)weredealingwithstockmanagementproblems.Theyidentifiedthreemethods forstockmanagement:orderingmethod,periodic fillingmethodandperiodicreviewandfillingmethod.Twomainapproachesaredefinedforlogisticsactivitiesplanninginhospitalsin Singapore, one stock‐oriented and other schedule‐oriented. The stock‐oriented approachimpliesthathospitalsormedicaldepartmentssendtheirorderstotheirsuppliersinthemomentwhentheirstocksreachthere‐orderinglevel.Thesecondapproachisfocusedonmakinggoodsdelivery schedule, which defines the times and quantities for every delivery. In the stock‐oriented approach, Lapierre and Ruiz (2005) added that the supplying of a department isperformedthroughthecentralwarehouse(CW).

Aptel and Pourjalali (2001) provided three basic models: a delivery to medical departmentsthroughthecentralwarehouse,asemidirectdeliverythroughdailyfillingofsmalldepartmentalwarehouse.Thefirstmodelrepresentsthesystemwithlargestockquantitywherethehospitalbearsthecostsofstorage.Thestocksofcommonlyusedmedicinesarestoredinadepartmentalpharmacy, while those being unavailable are requested from central pharmacies. The secondmodel suggests supplier’s direct delivery of necessary quantities to themedical departments,withoutinvolvingcentralpharmacies.Theirapplicationleadstostockreductionand,inaddition,the timeneeded todeliverymedicines to thedepartments is reducing.The thirdmodel,beingthemostsimilartoJIT(JustInTime), ischaracterisedbyverycloserelationsofahospitalandsuppliers who take hospital stock management upon themselves. Pan and Pokharel (2007)upgradedthepreviousauthors’models,i.e.thefirstmodelisdividedontwomodels:(1)directdeliverytocentralwarehouseandthendeliverytomedicaldepartmentforfurtherusing,and(2)directdeliverytocentralwarehouseandthendeliverytodepartmentalwarehouse.

Kim and Schniederjans (1993) identified three types of material management systems inhospitals: conventional, JIT and stockless. Heinbuch (1995) and Jarrett (2006) noted in theirpapersthateffectivematerialmanagementandJITdeliveriescouldreducethehealthcarecosts.Lapierre and Ruiz (2005) found two approaches upon which most hospitals organize theirsupplyactivitiesandthesearetwo‐orthree‐echelonstocksystems.Inadditiontothequestionswhether tousetwo‐or three‐echelonsystem, therearealsoquestionsofwhatgoodstoorderandwhen,howmuchstockstostoreetc.Fig.2showsatwo‐echelonsupplysystem.Itcouldbeseenthatthekeydecisioniswhethertoclassifyaproductintothestocksordeliveryitdirectly.IftheproductisstoredinCW,thefrequencyoforderingfromasupplierisreducing,thestocksincareunits(CU)arereducingtoo,butCW’sstocksareincreasing.AvoidingCWwillreducethehandlingtimeandtheneedforspace inCW,but thisrequiresbettercoordinationofreceivinganddeliverytoCU.

Figure5.Keydecisionsinasupplysystem(LapierreandRuiz,2005)

Aptel and Pourjalali (2001) and Kriegel et al. (2013) highlighted the importance and theadvantages of the partnership with other hospitals and material/service suppliers. Theyconsidered that it is also possible to reduce the stocks in more complex industries such asmedicine, as shown in the hospitals that were the subjects of their research. Improvedcooperation among supply chain participants has a positive impact on total costs and serviceperformanceimproving.PanandPokharel(2007)indicatedthathospitalsinSingaporedon’tseetheassociatingwithsuppliersasastrategicoption,sotheyratherdecidetooutsourcelogisticsservicesandaddthattheirgoodsmainlycomefromlocaldistributors.

4.THESITUATIONINBELGRADE–CLINICALCENTEROFSERBIA

CCScomplex, locatedon34hectaresarea,has41organizationalunits in total:23clinics,ninecenters, polyclinic and nine offices for service activities. Annually, in CCS dispensary units90,000patientsaretreated,50,000operationsareperformed,morethan7,000childbirthsaremadeandmorethan950,000hospitaldaysoftreatmentarerealized.Annually,25,000patientsaretreatedandmorethan5,000operationsareperformedindayhospitals.EmergencyCenterhas 298 beds at disposal, where 167 of them are in intensive care unit. This number is notsufficientregardingthattheiraverageoccupancyis99%.

ForeverycategoryofgoodsCCSannounceatenderbasedonannualplans,whicharemadebyeach clinicmanager on the basis of the consumption in the previous periods. The contract issignedwith each supplier and after that the goods are ordered according to demand of eachclinic.CCSmightrequestagreaterquantitythancontractedoneonlyifthesuppliersareabletomeetthisrequest.Thegoodsareshippedtothecentralpharmacywarehouse(CP),fromwherethedistributionisperformedbyCPfleettothedepartmentalpharmacies(DPs),locatedoneveryclinic, and then they are forwarded to the departments for further use. This supply systemgeneratesthestocksinthreelevels: inCP, inDPsandindepartments.WithinCPisthecentralserverwhichprovidesinformationonastockstatus(foreverytypeofgoodsandforeveryclinic,atanymoment),thatarenecessaryfordeterminingquantitiesandmomentsofsendingordersto the suppliers. CCS tends to reduce the stocks in clinics, i.e. inDPs and indepartments. Forsometypesofgoodsitisnecessarytostoregreaterquantitiesatanymoment,especiallywhenitcomestothefunctioningofEmergencyCenter.Thestocklevelofthesegoodsmustnotdroptothezero,becausethecostsofoutofstock(lossoflife)aremuchgreaterthancostsofadditionalstockstorage.However,duetoaprioricontractedquantities,thereisadecreaseinstocksastheyearcomes to theendand insomecases thiscouldcausethe lackofrequiredgoods. In thesesituations the problems are solving by the cooperation of clinics within a complex, by thecooperationofCCSwithotherhospitalsystems(MilitaryMedicalAcademy)orbytheuseofthe

goodsconsideredadequatesubstituteandbeingavailable ina sufficientquantity (e.g.agauzecanbereplacedbyacompress).

4.1Centralpharmacy

CPbuildingissituatedontheedgeofCCScomplexandconsistsoftwolevels(groundandfirstfloor). It consists of four departments: finished drugs (antibiotics, analgesics, etc.), operatingsupplies (cotton, gauzes, syringes etc.), solutions (sterilization) andoficina (e.g. production ofeyedrops).Thedepartmentsarespatiallydisorganized,consistofseveralareasthatcouldbeondifferentlevels(e.g.operatingsuppliesarestoredonbothgroundandfirstfloor,whilesolutionsonly on first floorbut atmany locations).The space for expensive operating supplies storage(surgicalsuture)andmedicinesrequiringaspecialtemperaturemodeisalsolocatedonthefirstfloor.Themainproblemofmulti‐floorwarehouseorganization is the vertical transportwhilethe main problem of the admission and dispatch of goods is the only one loading ramp,whereforethedispatchtime(7‐9a.m.)anddeliverytime(after9a.m.)aredefined.

4.2Blood,foodandwasteflows

Institute for Blood Transfusion is located within Emergency Center (EC), fromwhere all theclinicswithinCCScomplexaresupplied.Duetoitsspecificity,ECshouldhaveitatanymoment.Thebloodisdeliveredtotheclinicsonlyincasesofscheduledoperations,butifthereisalackofit, it comes to their cancellation. Since its lack is a big problem during summer months, thevoluntarydonorflowsarearisingasanadditionalproblem.

Thespecializedcompany,whichcooperateswithCCS technicalservice, is responsible for foodsupply.ThekitchenissituatednexttoCP,fromwherethedistributiontotheclinicsisstarting.The special smaller elevators are used for vertical transport. The technical service workersdeliverythefoodthreetimesadayandthendistributeditwithineveryclinicandtoallpatientsbyusinghandcarts.

The communal service removes non‐hazardous waste generated by CCS and ecology andtechnical service removehazardouswaste.Thedepartmentsusedifferently colouredbags foreverytypesofwaste(blackoneforcommunalwaste,yellowoneforinfectivewaste,brownonefor pathoanatomic waste etc.). Waste disposal is performed up to several times a day incontainerswithcapacityof220lindepartmentalwarehousesofclinics,whichthenthetechnicalservicetakesover,transportandstoreitinahazardouswastewarehouse,situatednexttoCP.Afteraccumulationthewasteisbeingdispatchedonfurthertreatment.

5.CONCLUSIONS

Althoughtherearemanymodelsofdistribution,stockmanagement,materialmanagement,flowmanagementetc,itisnotpossibletomaptheirapplicationfromonetoanotherhospitalsystem,butitisnecessarytoconductacomprehensiveresearchinordertoadjustthemodeltospecificcase. However, this is amajor challenge for logistics sector, considering all participants, flowdiversity and problems that could arise and that are partially covered by this paper.Reconfiguration of such a complex system, like a hospital, requires major infrastructure,financial and staff training investments. These changes of entire or a part of the system aresimilar to those in other industries, accompanied by participants’ repulsive attitude. Severalstudies showed thathospital systemswithmoredevelopedandorganized logistics, aswell aswithseparatesectorformanagingtheseactivities,havemoreefficientbusinesswithlessstocks,lowercostsandbetterqualityofserviceprovidedtopatients.

CCS complex is characterised by a bad spatial layout, which in some situation leads to theincompatible flows crossing. The cause of this is that goods flows and supply system are not

considering when planning and the consequences involve difficulty and inefficiency in flowrealizationwithinthecomplex,congestionoccurrenceanddisruptionofpatientflowsandbasicactivity.Thecurrentfunctionsofsomefacilitiesbeinginappropriatefortheirprimarypurposearealsoaggravatingcircumstances.ThisobservationisconfirmedinCPfacilitywherethelarge‐turnover goods are stored at not so suitable places, causing unnecessarymanipulations, costincreasing and storage in/off time increasing. In addition, the lack of cooperation of varioussectors responsible for different flows’ optimization results in inefficient functioning of theentiresystem.

ACKNOWLEDGMENT

Thispaperwaswrittenon thebasisof theseminarwork for thesubjectCity logistics.WearegratefultoProf.SlobodanZečevićandAsst.Prof.SnežanaTadić,lecturers,forsharingexpertise,andsincereandvaluableguidanceandencouragementextendedtous.

REFERENCES

[1] Aptel, O., Pourjalali, H. (2001). Improving activites and decreasing costs of logistics inhospitals: A comparison of U.S. and French hospitals. The International Journal ofAccounting,36,65‐90.

[2] Butler,TW.(1995).Managementscience/operationsresearchprojectsinhealthcare:theadministrator’sperspective.HealthCareManagementReview,20(1),19–25.

[3] Haraden,C.,Resar,R. (2004).Patient flow inhospitals:understandingandcontrolling itbetter.FrontiersofHealthServicesManagement,20(4),3–15.

[4] Heinbuch,S.E.(1995).Acaseofsuccessfultechnologytransfertohealthcare:totalqualitymaterialsmanagementandjust‐in‐time.JournalofManagementinmedicine.9(2),48‐56.

[5] Jarrett, P.G. (2006). An analysis of international health care logistics: the benefits andimplicationsofimplementingjust‐in‐timesystemsinthehealthcareindustry.LeadershipinHealthServices.19(1),i‐x.

[6] Kim,G.C., Schniederjans,M.J. (1993).Empirical comparisonof just‐in‐timeand stocklessmaterialmanagement.HospitalMaterialManagementQuarterly.14(4),65‐75.

[7] Kriegel,J.,Jehle,F.,Dieck,M.,Mallory,P.(2013).AdvancedservicesinhospitallogisticsintheGermanhealthservicesector.Springer.

[8] Kriegel,J.(2012).Krankenhauslogistik,SpringerGabler,Wiesbaden.

[9] Lapierre,S.D.,Ruiz,A.B. (2005).Scheduling logisticactivities to improvehospital supplysystems.Computers&OperationsResearch,34,624–641.

[10] Noon, CE., Hankins CT., Côté MJ. (2003). Understanding the impact of variation in thedeliveryofhealthcareservices.JournalofHealthcareManagement,48(2),82–97.

[11] Pan, Z.X., Pokharel, S. (2007). Logistics in hospitals: a case study of some Singaporehospitals.LeadershipinHealthServices,20(3),195‐207.

[12] Pieper,U.,Michael,M.(2008).Healthcarelogistik.Gablerlogistiklexikon,214–220.

[13] Villa,S.,Prenestini,A.,Giusepic, I. (2013).A framework toanalyzehospital‐widepatientflowlogistics:EvidencefromanItaliancomparativestudy.HealthPolicy,115,196‐205.

[14] Vissers, J. (1998). Patient flow‐based allocation of inpatient resources: a case study.EuropeanJournalofOperationalResearch,105,356–370.

UNDERGROUNDSYSTEMSINSERVICEOFCITYLOGISTICS

LjiljanaMilinkovića,JovanaPantelića,*

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Thetopicofthispaperconcernsthemodelsofundergroundlogisticsystemsinordertosolve the logistics problems of urban areas. These systems contribute to the protection ofenvironmentandqualityof life in thecity,displacing logisticactivities,primarily transportation,underground.However,theinvestmentriskislarge,profitsuncertainandcanbeexpectedonlyinthelongterm.Itseemsthattheseprojectswillberealizedonlyasalastoption,whenthepossibilityofexpansionoftheexistingroadsystemisdepleted.

Keywords:undergroundlogistics,urbanfreighttransport,theadvantages,disadvantages

1.INTRODUCTION

Urban freight transportexperiencesacontinuousgrowth,andexpectationsare that this trendwill be continued in the future. The main causes of this are the trends in production anddistributionwhicharebasedonthelowlevelofsuppliesandpreciselydefineddeliveries(JustInTime)aswellasthegrowingtrendofe‐commerceandhomedelivery(Zečević&Tadić,2006).Inordertosolvetheproblemandsecuremoreefficient implementationof logisticprocessesandactivities, various initiatives and conceptual solutions of city logistics have been defined andtested. Underground logistic systems belong to the group of the most radical and mostdemanding financial initiatives. In addition, theabovementionedgroup seemsvery innovativeand takes into account the system of underground networks, the amount of investments andhigh level of automation (Tadić, 2014). Underground systems have been intensively analyzedsincetheendofthe20thcentury,eventhoughtheundergroundtransportsystemhasa200yearslonghistory.ThefirstsystemforundergroundtransportoftelegramsandpostalmailfromthecentertothebranchofficeinLondonstarteditsworkin1853.Fewyearslateritwasintroducedin other European cities as well: Berlin (1865), Paris (1867), Vienna, Munich, Rome, Naples,Marseille etc. These systems can be very effective for solving the problems of urban freighttransport.Theycanalsoreducenegativeimpactsontheenvironmentandimprovethequalityoflife inthecity.Thispapergivesvariousmodelsofundergroundlogisticsystemsanddescribesmaincharacteristics,advantagesanddisadvantagesofsomeofthedesignedsolutions.

* Jovanapan90@gmail.com

2.CHARACTERISTICSOFUNDERGROUNDLOGISTICSYSTEMS

Allundergroundlogisticsystemsarecharacterizedbysomeofthefollowingfactors(Chenglinetal,2014):

1) Municipal structures and the aspects of land use – reducing the use of land for roadfreightstations,parkinglotsandotherfacilities.

2) Freight efficiency – reducing urban freight costs and improving transport services,avoiding congestion on the highways during storms, rain, snow and other weatherconditionsaswellasprovidingahigherlevelofservicethatismorereliable,saferandmoreaccuratethanexistingtransportbasedontheurbanfreightsystem.

3) Thesafetyofroadtransport–reducingtrafficaccidentscausedbytrucks.4) The city’s social environment – reducing energy consumption, emissions of harmful

gases,trafficnoiseandvibrationscausedbytrucks.Undergroundlogisticsystemsappearin(Visser&vanBinsbergen,2000):

1) Urbanareas inorder todelivermail,merchandise,catering, stationery.Theunitof thecargoisthesizeofapallet.TestingofthesystemwascarriedoutintheGermancitiesofUtrecht,LeidenandTilburg,theninJapanandLondon.

2) Withinandbetweentheindustrialcomplexes,logisticcentersandmultimodalterminalssuchasairportsandportcomplexes.Unitsofthecargoarepallets,shippingcontainersandaircraftpallets.OLSSchipholintheNetherlandsisanexampleofsuchaproject.

3) Long‐range transport of agricultural products, ore and solid waste. For this purpose,capsule‐typepiping systemshavebeendevelopedand implemented in Japan,USAandRussia.

4) Transportofmaritimecontainers to inlandports.The researchhasbeenconducted intheUnitedStates.

Byanalyzingtheprojectsandstudiesontheundergroundlogisticsystems,Chenglinetal(2014)havedistinguished threedevelopmentmodels: themetromodel, themodel of pipeline cabinsandthevehiclemodel.

2.1Modelofundergroundlogisticsystembasedonmetro

This logistic system is based on the use of passenger subway system. The final part of thetransportofgoodstothecustomerandthegeneratorflowisrealizedintheusualway.PielageandRijsenbri(2002),Liuetal(2008)investigatedthepossibilityofusingthepassengersubwaysystemforthetransportofgoodsinthecityandanalyzeditscargocapacity.Zhangetal(2005)analyzedtheexperienceofundergroundlogisticssystemsindifferentcountriesandsuggestedconstruction of the metro for “dual‐use”. The functioning of this system is affected by therequirementsofpassengersandcargoflows.

2.2Modelofundergroundlogisticsystemsbasedonpipelinecapsules

This form of underground logistics is based on pipeline technology that involves the use ofround or square capsules, cabins for transportation. The transport capsule has a function ofcargo transfer only, whereas orientation, the control function or any other are excluded. Itscontrol functionisapassiveoneandit isexecutedovertherailsandthewallofthetube.Therateofwearoffacilitiesandequipmentisverylarge,andtheconstructionofthepipelineisverydemanding.QianandGuo(2007)suggestedthatthismodeloftheundergroundlogisticsystemshouldbedivided into three categories:pneumatic, hydraulic andelectromagnetic.Pneumaticandhydraulicpipelinecapsulessystemsareusedto transportmail,package, fruits,vegetables

andothertime‐sensitivegoodsandmaterials.Hydraulicsystemsarealsousedforthetransportofground,sand,oreandothermaterialsthatcanendurealongtransportation(Chenglinetal.,2014).

2.3Modelofundergroundlogisticsystemsbasedonvehicles

This concept uses special vehicles that use mainly batteries as their source of energy forundergroundtransportrealization,e.g.AGV(automaticguidedvehicles)intheNetherlandsandDMTinJapan(Chenglinetal.,2014).

3.EXAMPLESOFUNDERGROUNDLOGISTICSYSTEMS

All around the world, different systems for underground transportation of goods have beenproposedanddesigned.Thischaptergivesbasiccharacteristics,advantagesanddisadvantagesofsomeofthedesignedandtestedsolutionsoftheundergroundlogistics.

In the Japanese city of Sapporo, attempted solution for theproblemsof distributionof goods,suchas: trafficcongestion,negative impactson theenvironment,delaysofdelivery,especiallyduringthewinterwhenthesnowmakestransportationdifficult,consistedofasystemoflogisticintegration of public service station in order to transport goods from the suburbs to the citycenterefficiently.Sapporo(1.9million inhabitants) innorthern Japansuffersserious logisticalproblems inwinterwhen snowdrastically reduces transport efficiencyon thedirt‐roads.Thepossibilityofintegratingthecity’ssubwaysystemwithaconventionaltrucktransporthasbeentested inorder to facilitate thedistributionof goodsbetween the suburbsand the city center(Kikutaetal.,2012).

Carts(sized500x900x700mm,withagrossweightof60kg)thataredescendedtotheplatformofthemetrostationandtransportedbytheurbanmetrosystemincombinationwithpassengerswereused for the transportationofgoods.On thestationclosest to therecipient, i.e. the finaldestinationofgoods,thetrolleyisraisedabovethegroundand,ifnecessary,furthertransportedbyusual roaddelivery vehicle to the recipient. Thepilot project has demonstratednumerousbenefits for residents, senders, recipients and carriers, logistics providers, out of which thefollowingcanbesingledout:(Kikutaetal.,2012)

1) Thosewho aredelivering can avoid traffic jams and reduce thedelayby changing themeansoftransportfromtrucktothesubway

2) The traffic density ismitigated and the urban environment has a lower CO2 emissionbecausethenumberofoperatingtrucksandunloadingofvehiclesdecreases

3) Tradersinthecentralundergroundcomplexcanovercomeshortagesofgoodsthroughfastandfrequentdelivery

4) Theofficeofpublictransportcanincreaserevenuebyprovidingcargoservicesoutsidethecityrush‐hour

The main disadvantage of the project is the fact that passengers and cargo are not spatiallyseparated.

PlannedundergroundlogisticsysteminTokyoinvolvesapplicationofhybridvehiclesthathavetheability tomoveontheregularroad infrastructure,aswellasaseparaterail infrastructurewithintheundergroundsystem.Researchonthe impactofbuildingasystemincentralTokyowitha300kmlongtunnelwith150depotshasshownthatthissystemcouldattractabout30%oftheexistingroadfreighttraffic,whichwouldreduceNO2andCO2emissionsby10%and18%respectively, and power consumption by 18%, while the average speed of traffic in the citywouldbe increasedbyabout24%(Ooishi&Taniguchi,1999).Althoughtheuseof thesystemsignificantly reduces the number of environmental and social problems caused by logistic

activities,analysisoftheeconomicfeasibilityofitsimplementationresultedintheinertrateofreturnwhichamountedto10%inthecasethattheinfrastructureisconstructedbythepublicsector. The results confirm that the system is not self‐sustainable and can be consideredreasonableonlyifwetakeintoaccountallthesocialandenvironmentalbenefitsfortheentireexploitationperiod.

Underground logistic systemshavebeen the subjectof researches in someEuropean cities aswell.Germanytestedaprototypesystembasedoncapsulesthatmoveundergroundtunnelsortubesandthatcancarrytwopallets(Beckmann,2007).AsimilarconcepthasbeendevelopedinItaly. A prototype designed in 2006 involved a capsule transport capacity of one euro pallet.Since then, numerous researches and feasibility studies of this system have been conducted(Cotanaetal.,2008).UndergroundlogisticsystemshavebeenconsideredinseveralcitiesintheNetherlands.Thepipelinetransportationsystem(capsulessystem)forsupplyingmarketchainsofbooksandnutritionalproductshasbeenconsideredinGroningen.Also, feasibilitystudyfortheuseofundergroundsupplysystemsthroughtheimplementationofAGVs(automaticguidedvehicles)wasconductedforUtrecht,LeidenandTilburg(Tadić,2014).Theconceptofautomaticguidedvehiclesmoving through theunderground tunnels at a speedof 20‐40km/hhasbeentested for the transportof flowersbetweenairports inAmsterdam, theworld’s largest flowerauctioninAalsmeerandtherailwayterminalHoofddorp(vanderHeijdenetal.,2002).

The first ideas of development of the underground system OLS (Ondergronds LogistiekSysteeem) appeared in theNetherlands in the early 1990s. Deteriorating accessibility, due totraffic congestion, started to threaten the position of Schiphol airport in Amsterdam andAalsmeerFlowerAution.Inordertosolvethisproblemandensuresufficienttransportcapacity,withminimalimpactontheenvironment,theconceptoftheundergroundtransportsystemhasbeendiscussed.Shortlyafter,itbecameclearthatthethirdconnectionwasneededtoo,andthatit implied linking international railway terminal nearHoofddorp. By using high‐speed freighttrains and connecting the airport, auctions and international railway with the help of anundergroundfreighttransportsystem,atrustedconnectionwascreatednotonlybetweentheauctionandtheairportbutalsowiththerestoftheEurope.In1995,privatecompaniesstartedtoworkonafeasibilitystudyoftheundergroundtransportsystemofflowers,expensivegoods,computerpartsandnewspapersbetweenSchipholairportandAalsmeerflowerstockexchangewithanewrailterminalinHoofddorp(Wiegmansetal.,2010).Theseareasrepresenttheinitialandfinalpointofcargoflowsandtheyconsistofseveralterminals:Schipolairport–twotofiveterminals,Aalsmeerflowerauction–onetothreeterminals,RailterminalHoofddorp–asingleterminal.

Consumers deliver and pick up goods through terminals, and the OLS vehicles performtransport.Theterminalshavetheroleofabufferfortemporarystorageofgoodsandtheyareconnected with tunnels, 5 meters in diameter, which are positioned at a depth of 0‐20mundergroundandtheycanbeunidirectionalorbidirectional.Terminalsareplacedatthelevelofground’s surface in order to reduce construction costs and facilitate interactionwith regionalclients. The system is fully automated and uses AGVs for transporting goods between theterminalandfullyautomatedtransferstation.Around400AGVs,sizedfromtwentytotwenty‐six feet and weighing about 10 tons circulate the system. The usage of AGV has certainadvantages: less vehicles leads to lower transport costs; there is less chance forhumanerrorwhich improvessystemreliability.Acomplexsystemwithseveralhundred independentAGVswhichmovementneeds tobeguidedandmanaged isverycomplicated toorganize.There isapossibilityofdelaysandahighspeedandreliabilityofthesystemarerequired(Wiegmansetal.,2010).

Thisconceptcombinesthesocialbenefitsofplacingtrafficundergroundandtheuseofelectricdrive with the economic advantages of the unobstructed automated transport throughinfrastructure that is separated from the passenger transport. Economic benefits aredemonstratedinalmostdirectdelivery,24‐hourservice,lowoperatingcostsandshortfeedback

time. Social benefits take into account legal limits, which results in a reduction of noise,pollution, emissions, reduction of the energy consumption, reduction of CO2 emissions, betteruseofavailablespaceandincreasedtrafficsafety.

The main disadvantages of this project are a big investment, the private sector and theauthoritiesnotshowingtheinterestinitsdevelopmentandimplementation,aswellasthefactthat several different technologies have been used in its automation, even though the systemitself was easy to understand. After several years of development and testing of variousconcepts,theprojectwaspausedin2002duetoeconomicreasons(Wiegmansetal.,2010).

4.CONCLUSION

Inthispaper,aninitiativeforundergroundlogisticsystemshasbeendemonstratedasasolutionto the problems of urban environments logistics. The concept combines social benefits bydisplacingtrafficundergroundandtheapplicationofelectricpropulsionwiththeadvantagesoftheunobstructedautomatictransportthroughdedicatedinfrastructurewhichisseparatedfromthepassengertraffic.Theideawasto,throughvariousformsofundergroundtransport,presentbasiccharacteristics,advantagesanddisadvantagesaswellastheresultsoftheirapplication.

The economic advantages of this system include almost direct delivery, 24‐hour service, lowoperating costs and short feedback time. Social benefits include reduction of noise, visualpollution, physical interferences, gas emissions, reducing congestion and traffic jams, moreintensiveuseofavailablespace,relievingthestreetnetworkandtheincreaseofoveralltrafficsafety. Investment costs arehigh and technologywhich is used is new,which leads to lackofexperience in automated mass transport systems. The application of these systems requiresbuildingtheentireinfrastructurewhichmeansthatitsrealizationrequiresalongperiodoftime.

ACKNOWLEDGMENT

ThispaperwaswrittenonthebasisoftheseminarworkforthesubjectCitylogistics.Wewouldliketothankourlecturers,Prof.SlobodanZečevićandAsst.Prof.SnežanaTadić,forsharingtheirknowledge and expertise generously, and provide genuine encouragement and guidancethroughouttheproject.

REFERENCES

[1] Beckmann,H.(2007).CargoCap:anewwaytotransportfreight.In:ProceedingsofSchillerInstitute Conference, The Eurasian Land‐Bridge Becomes A Reality! Schiller InstituteConference,Kendrich,Germany.

[2] Chenglin,M.,Haijun,M.,Xuechun,Y.,Jingfeng,D.,Chao,M.,Lihua,A.(2014).StudyontheDevelopment Mode of Urban Underground Logistics System, Service Science andManagementResearch,3(1),7‐12.

[3] Cotana, F., Rossi, F., Marri, A., Amantini, M. (2008). Pipe§net: innovation in transportthrough high rate small volume payloads. In: Proceedings from the 5th InternationalSymposium on Underground Freight Transportation by Capsule Pipelines and OtherTube/TunnelSystems,TheUniversityofTexas,Arlington.

[4] Kikuta, J., Ito,T.,Tomiyama,I.,Yamamoto,S.,Yamada,T.(2012).Newsubway‐integratedcitylogisticssystem,Procedia‐SocialandBehavioralSciences,39,476‐489.

[5] Liu,Y.,He,K.,Liu,J.,Xu,Y.(2008),AnalysisoftheConceptofUrbanRailTransitBasedCityLogistics System, International Conference on Smart Manufacturing Application, ICSMA2008,Gyeonggi‐do,288‐292.

[6] Ooishi, R., Taniguchi, E. (1999). Effects and profitability of constructing the newunderground freight transport system. In: Taniguchi, E., Thompson, R.G. (eds.), CityLogisticsI,InstituteofSystemsScienceResearch,Kyoto,303‐316.

[7] Pielage,B.A.,Rijsenbrij, J.C. (2002).ThePotentialofMetroSystems forCityLogistics,3rdInternational Symposium on Underground Freight Transportation by Capsule PipelinesandOtherTube/TunnelSystems,Ruhr‐UniversityBochum,Germany.

[8] Qian, Q., Guo, D. (2007). Urban Underground Logistics System Introduction. Beijing,People'sCommunicationsPress

[9] Tadić,S. (2014).Modeliranjeperformansi integrisanihcity logističkihrešenja,FacultyofTransportandTrafficEngineering,UniversityofBelgrade,Belgrade.(PhD)

[10] vanderHeijden,M.C.,vanHarten,A.,Ebben,M.J.R.,Saanen,Y.A.,Valentin,E.C.,Verbraeck,A.(2002).UsingsimulationtodesignanautomatedundergroundsystemfortransportingfreightaroundSchipholAirport.Interfaces,32,1‐19.

[11] Visser, J.G.S.N., van Binsbergen, A.J. (eds.) (2000) Proceedings 2nd InternationalSymposiumUndergroundFreightTransportationbyCapsulePipelinesandOtherTube/TunnelSystems(ISUFT2000),TRAILConferenceProceedingsSeriesno.P2000/2,TRAILResearchSchool,Delft.

[12] Wiegmans, B.W., Visser, J., Konings, R., Pielage, B‐J.A. (2010) Review of undergroundlogisticsystemsintheNetherlands:Anex‐postevaluationofbarriers,enablersandspin‐offs,EuropeanTransport,45,1–16.

[13] Zečević, S., Tadić, S. (2014). City logistika, Faculty of Transport andTraffic Engineering,UniversityofBelgrade,Belgrade.

[14] Zhang,M.,Yang,C.,Yang, J. (2005).DevelopedULSComparisonandReference.LogisticsTechnology,03,81‐91.

URBANLOGISTICSSYSTEMSANDNIGHTGOODSDELIVERY

HanaLjubičića,JovanaPavlovića,*

aUniversityofBelgrade,FacultyofTransportandTrafficEngineering,Serbia

Abstract:Thispaperpresentsthereasonsforrelocatinglogisticssystemsintoperipheralareaofacity and the possibilities of their return into urban areas. The implementation of efficient andarchitecturallymoderncityterminalsandtheirconnectiontologisticssystemsinthefringeareasbymoreecologicaltransportmodesarecrucialforsuccessfulreturnoflogisticsintocentralurbanareas. Also, this paper states themain advantages, disadvantages and types of night delivery.Consideringrecentlysetaimtoreducefreighttransportduringpeakhours,thereisatendencytoperformlogisticsactivitiesatnight.Thepurposeistoreducetrafficcongestionandenvironmentaleffect.

Keywords:logisticssystem,cityterminal,nightdelivery.

1.INTRODUCTION

Theevolutionofurbanareascausedchangesofbothformsandphysicalcomponentsofgoodsprocurement, storage and distribution. By spatial city spreading, transport infrastructuredevelopment and urban land price increase, the macro distribution flow ending is movingtowards peripheral areas, in cargo terminals and warehouses dislocated to the urban fringe.Logisticsserviceprovidersandcustomersdonotwaitforpreciseplanningdocuments,buttheysolvelocation‐relatedandlogisticssystemconstructionproblemsbythemselvesinaccordancewith their abilities and requests. These systems are being dislocated in a quite disorganizedmanner,withoutpossibility for concentrationandconsolidationof the flows theyare running(Tadić, 2014; Tadić et al., 2014a). Distancing logistics from urban areas has reinforced theadverse effects of logistics flow realizationwithin a city. The number of commercial vehiclesservingurbanareasisincreasing,butthereisalsoanincreaseinfreightvehicle‐kilometresandinalladverseimpactsontheenvironmentandthequalityoflifeinacity.Inordertostopthesetrendsandincreasetheefficiencyoflogisticsactivities,thedifferentinitiativesandconceptionsof city logistics are defined and analysed. This paper presents the causes and possibilities toreturn logistics into urban areas. Furthermore, it also presents the advantages, disadvantagesandmodalitiesofnightdeliveryasoneoftheinitiativeforsolvingurbandistributionproblem.

2.REASONSFORMOVINGTHELOGISTICSOUTFROMURBANCENTERS

Logistics is very important for city development and residents’ life and it also presents thesignificant source of workplaces (Christopherson & Belzer, 2009; Diziain et al., 2012). Largelogistics systems supplying national and internationalmarket have become a key element of

* jokapavlovic92@yahoo.com

urbaneconomics(Christopherson&Belzer,2009)and tend tobeconcentrated in largeurbanareas(Cidell,2010).Highlandpriceisoneofthereasonstodistancelogisticsfromurbanareas.Moreover, road freightvehiclesandvanspollute theair,makeanoiseandvisually impair theenvironment,sothereisevenlessspaceforlogisticsincityhearts,resultinginitsdislocatingtoother locations (Deblanc & Rakotonarivo, 2010). This relocation is called unplanned logisticsexpansionanditrepresentstheglobalphenomenon(Cidell,2010;Woudsmaetal.,2008).

The logisticsexpansionhas serious consequences for theenvironment.Dislocating logistics toperipheralurbanareasincreasesthedistancesthatneedtobeovercomebyusingroadtransportmodes, thenumberof kilometersmadeby freight vehicles, implying additional harmful gasesemissions, increasing highway congestion and additional cost for a community. As the roadcongestion is increasing, logistics providers and carriers show more interest in locationsdistancedfromthecitycore(Diziainetal.,2012).Whenitcomestothegoodsdeliverywithinacity, it is obvious that road transport doesn’t have a competitive alternative compared to theother transportmodes because of its price, flexibility and very dense network. Limiting roadtransportvolumeincreasebypublicauthoritiescouldmitigateitsenvironmentaleffect,butthisrequiresdefiningofrailorfluvialtransportsolution.

The thingwhich should be emphasised is the reduction of road transport use for “lastmile”deliveryandlogisticsexpansionsuppression.Inordertoachievethat,itisnecessarytodevelopmultimodaltechnologies,bothonperipheryandwithinacity(Diziainetal.,2012).Thegoalistoenablethegoodconnectionofurbandistributioncentres,cityterminalsandlogisticscentresaturban fringe by dense transport network. Ideally, they should be connected bywaterways orrailway.Without these connections, the logisticswould be able to return to the cities only informsofsmaller facilitieswhichcouldbe incorporated intosmallspaces(Diziainetal.,2012).Cargotramwayimplementation,asalinkbetweenacityterminalandlogisticscentreatanotherlocationwithin a city, is a concept tested inmany European cities (Tadić, 2014; Tadić et al.,2014a).Thegoodsdistributionfromthecityterminaltothegeneratorinthegravitationalareacould be performed by alternative electric vehicles in order to mitigate the adverseenvironmental effect (Munuzuri et al., 2005; Russo & Comi, 2012; Tadić, 2014; Tadić et al.,2014a;Tadićetal.,2014b).

Duetotheirlowstatusandbadimage,thelogisticscompaniescannotaffordanexpensiveland,compared to ones who buy it with the residential or commercial purpose. However, variousauthors (e.g.,Diziainet al., 2012;Zečević,2006)have investigated the justificationofbuildingmodern and efficient city terminals ‐ multi‐storey facilities (“logistics hotels”) – representingvery attractive solutions and occupying significantly less space. The systems increase theconstructioncoststoalargeextentbecauseoftechnicalelementssuchasstrengthenedwallsorfloors,whichmustwithstandhighloads.Inaddition, itmustpayattentiontoarchitecturalandtechnical construction requirements defined by local authorities. Designing of such a hotelrequiresfacingwithnoiseandotherdisturbancesincurredbylogisticsactivities’realization,butitalsorequiresthearchitecturaleffortsuchasthisinofficeorresidentialfacilityconstruction.Another aspect that cannotbe ignored is careful introductionof logistics activities intourbanenvironments, where besides good building design it is necessary to ensure a minimum ofinterruptionswhentransportingthegoods,especiallyinroadtransport(Diziainetal.,2012).

3.NIGHTDELIVERY

The night delivery concept is one of the city logistics initiatives. In order to reduce freighttransport in peak hours during a day, the concept suggests the goods delivery during nighthours,most commonly inperiod10:00p.m. ‐7:00a.m.Thismeans shifting logistics activitiesfrompeaktonighthours,whenmostcityactivitiesarereducedtotheminimum,thusavoidingtrafficjamsandreducingcongestiononthestreetscausedbyfreighttransportduringaday.

The idea of freight transport realizationduring off‐peakhours emerged longbefore thenightdelivery concept implementation. The first data on the night delivery implementation wererecorded in the time of Julius Cesar, who prohibited deliveries during a day (Dessau, 1892).Nowadaysthesolutionisimplementedincitiesaroundtheworld(Holguín‐Veras,2008;Niches,2009).

3.1Nightdeliveryadvantages

The night delivery concept implementation could provide a numerous advantages for allparticipantsinlogisticsactivitiesandforthecityitself.Theseadvantagescouldalsobeperceivedonglobal level, intermsofenvironmentaleffectandsustainabledevelopment.Sincethegoodsdeliveryisperformedatnight,whentheroadsareempty,thehighertransportspeedcouldbeachieved, thus decreasing the total delivery time and eliminating delivery delays. Thiscontributes to the quality of service improvement; therefore the service providers achievehighermarket competitiveness. Also, shifting transport activities into off‐peak hours leads totraffic congestion reduction during a day, allowing better quality of life in the city. Anotheradvantage is reflected in the fact that, insteadof greaternumberofvehiclesduringaday, thesmallernumberofvehiclesduringanightisusedforfacilitysupply, i.e.thereisareductioninharmful gases emissions and energy consumption. All these improve the logistic systemsefficiencyandallowtheimplementationofmoreefficientstrategiesandconceptions,especiallyin thedomainofvehicle routingand labour forcescheduling,butalso in thecaseofshipmentconsolidation and grouping. Road safety is improved by congestion and overall transportreduction(Holguín‐Verasetal.,2014;Niches,2009).

From the point of logistics flow generators, the advantages of night delivery include lowerdistributioncosts(althoughtheinitialinvestmentsinlower‐noiseequipmentareemerging)andhigher reliability of a delivery. Higher reliability and quality of service contribute to greaterconfidence in suppliers and better cooperation of facility and its customers as well, becausethere is no customer disturbance due to daily goods delivery while desired products areavailable in earlymorning hours. The citizens could benefit frommore efficient traffic flows,lowertrafficcongestioninthecitycentreandlowerharmfulgasesemissions.Fromthepointofthe city, it could lead to economic and business development due to the lower delivery costs(Niches,2009).Thestudiesshowthatnightdeliveryconceptionscouldbemoreefficient fromtheaspectofenvironmentalexternalities fromthepolicybasedon fleetrenewal(Filippietal.,2010)orrestrictivelocalgovernment’measures(Comietal.,2011).

3.2Nightdeliverydisadvantages

Besidesthenumerousadvantages,thenightdeliveryimpliessomedisadvantages.Althoughitisreasonably toconsider thatnightdeliveryresults in thereductionofharmfulgasesemissions,thus reducing the adverse environmental effect, the exploration of this concept found theopposite.Namely, ifwetakeintoaccountfreightvehiclespeedandclimateconditions, itcouldbeconcludedthatoff‐peaktraffic,regardlessthereductionoftotalharmfulgasesemission,couldhaveanadverseenvironmentaleffectdependingontwomentionedfactors.Thenegativeeffectis more prominent in central than in peripheral urban area because of greater temperaturevariations,leadingtogreatersharesofpollutantsintheair.Thisdisadvantagecouldbeavoidedif the freight transport is realized in specific time periods during a night, thus reducing theadverseeffectsofharmfulgases.Also,theeffectsaremoreprominentathigher freightvehiclespeeds(Sathayeetal.,2010).

Innightdeliveryconceptthereisagreatervarietyofparticipantshavingdifferentrolesandeachof themtends toachieveagreaterbenefitsandprofitbysolution implementation.On theonehand, it can be observed the logistics service providers, which could experience the nightdeliveryadvantagesonthefastestwayduetothereductionoftotaldeliverycosts,fasteractivity

performing,parkingcostseliminationetc.Naturally,allthisdirectlyinfluencebothaconsigneeandhis/hercosts.Nevertheless,theconsigneesarelesswillingtoacceptnightdeliveriesandthereasonsinvolveincreasedrisksandcriminalexposureofbothstaffandgoods.Furthermore,bytransition tonightdeliveries theconsigneesexpect the increase inoperatingcosts,equipmentcosts and salaries (Holguín‐Veras, 2008). They have a task to provide either staff that willreceivethedeliveryorsomefacilitiesforgoodsstoragethatgoalongwithaprotectivesystem.While providers have savings in business, the consigneesmight face the increased operatingcosts arising from additional auxiliary facility costs, overtime costs of employees and higherpriceofnightwork.Thissituation,whereoneparthasmorebenefitscomparedtootherone,ispopularly called BoS (Battle of Sexes). Therefore, it is necessary to balance between servicebearerandcustomerandadjusttheconceptinsomewaytotheconsignees,whichinthiscasestillhavehigherexpensesthanearnings(Holguín‐Verasetal.,2012;Niches,2009).

From the point of population, the problem of the night delivery is the disturbance caused bynoise and delivery vehicle lights, thus decreasing the quality of life in the city. This problemcouldbeinfluencedbyusingspecialvehiclesandequipment.Theemphasisisputontheuseoflow‐noiseorelectricvehicles,infrastructureimprovementandothertechnicalsolutionsaimingto reduce the noise. Although these solutions have double positive effect – reduce the urbannoise and are more ecological solutions – they require large investments which might be aproblem(Holguín‐Verasetal.,2012;Niches,2009).Inaddition,thenightdeliveryconceptislessacceptablesolutionincitieswithhighercrimerate.

3.3Thetypesofnightdelivery

According to themodeof the acceptance of goods, twomain types of night delivery could bedistinguished: SOHD (StaffedOff‐HoursDelivery), implying additional staff engagement for theshipment acceptance, and UOHD (Unassisted Off‐Hours Delivery), with no staff for assistancewhenacceptingthegoods.Thefirstprinciplerequiresadditionalcostoflabourandnightwork,butitisasafeoption,whilesecondprinciplemaybelesssafe.Theoperatingcostsandrisksratio(Fig.1)representsthemaindifferenceofthesetwoconcepts(Holguín‐Verasetal.,2012).

Figure6.ComparisonofUOHDnightdeliveryconceptalternatives(Holguín‐Verasetal.,2012)

SOHDconceptcouldbefoundintwovariants(Holguín‐Verasetal.,2012):

1. Conventional OHD system – one ofmoremembers of the staff are present at the sitewhennightdeliveryisperformed;

2. Two‐stagesystem–itisusedforsupplyinglargegeneratorsandinvolvestwostage,thefirst onewith appropriatedelivery room for shipment acceptance and sending, and inthe second stage the goods are sent to actual consignees as a part of regular dailytransport.

High

Low

OPERATIONALCOSTS

StaffedOHD

Low High

Two‐stagesystem

Electronicdoorman

Double‐doorsystem

KeysystemRisk

UOHDconceptimpliesmorealternativesforacceptanceofgoods(Holguín‐Verasetal.,2012):

1. Double‐door system – the driver performing delivery is provided with a key to anoutsidedoorthatleadstoasmallstoragearea,separatedfromtherestofthebusinessarea,sothedriverdoesn’thaveafullaccesstothefacility;

2. Keysystem–thedriverisprovidedwithakeywhichenableshimtodepositthegoodsataspecificlocation;

3. Electronickeysystem–thedriver isprovidedapasswordorsecuritycode,definedbyfacilityowner,toaccessthefacilityorthestoragebox;

4. Systemwithmanual/electronickeyandsecuritycameras–theprincipleisthesameasinpreviouscase,excepttheinstallationofmonitoringcameras;

5. Electronic doorman system – a remote operator, assisted by security cameras andradio/phone,grantsaccesstothefacilityandrequiresidentificationchecks.

4.CONCLUSIONS

Severeroadcongestionandlandpricearethemainreasonsfordislocatinglogisticssystemsinperipheralareasof thecity.Fromtheotherhand, thereturnof logisticssystem into thecitiesrequires facing the noise and other disturbances arising from logistics operations. The mainproblem is the road transport which attracts a special attention due to its dominant role intraffic flowsandnegativeeffectsthat implies.Thisproblemhas initiatedthedevelopmentandtheuseofothertransportmodes,especiallyintermodaltransportsolutionofcitylogistics.

Inordertosolvetheproblemofgoodsdistributioninurbanareas,thenightdeliveryconceptsare analysed. The deliveries during peak hours are ineffective, unpredictable and implyincreasedfuelconsumption(Vilain&Wolfrom,2001).Fromtheotherhand,thenightdeliveriesare faster,more reliable and do not require the change of vehicles, thus reducing delays andimprovingtheefficiency.Byanalysingtheproblemofnightdeliveryconceptimplementationitcanbeconcludedthattheefficienttransport‐distributionsystemcannotbeachievedwithoutajointactionofallparticipants.Itisnecessarythatallcitylogisticsparticipantscooperatecloselytomakethesystemfunctionefficiently(Tadić,2014).

ACKNOWLEDGMENT

Thispaperwaswrittenon thebasisof theseminarwork for thesubjectCity logistics.Wearegrateful to Prof. Slobodan Zečević and Asst. Prof. Snežana Tadić, lecturers, for greatunderstanding,generoussupportandassistancethattheyprovidedusduringtheelaborationofthis paper. Their ideas, well‐intended advice and suggestions have contributed to theimprovementofthispaper.

REFERENCES

[1] Christopherson, S., Belzer, M. (2009). The next move: metro‐regions and thetransformationof the freight transportanddistributionsystem. In:N.Pindu,H.Wial,H.Wolman (eds.), Urban regional policy and its effects, Brookings Institution Press,Washington,DC,USA(2),194‐222.

[2] Cidell, J. (2010). Concentration and decentralisation: The new geography of freightdistributioninU.S.metropolitanareas.JournalofTransportGeography,18,363‐371.

[3] Comi,A.,DelleSite,P.,Filippi,F.,Nuzzolo,A. (2011).Ex‐postassessmentofcity logisticsmeasures:ThecaseofRome.In:L.,Mussone,U.,Crisalli(eds.),Transportmanagementandland‐useeffectsinpresenceofunusualdemand,FrancoAngeli,Milan,Italy,235‐252.

[4] Dablanc,L.,Rakotonarivo,D.(2010).Theimpactsoflogisticsprawl:Howdoesthelocationofparceltransportterminalsaffecttheenergyefficiencyofgoods’movementsinParisandwhatcanwedoaboutit?Procedia‐SocialandBehavioralSciences,2(3),6087‐6096.

[5] Dessau,H.(1892).InscriptionesLatinaeSelectae,Berolini,apudWeidmannos

[6] Diziain,D.,Ripert,C.,Dablanc,L.(2012).Howcanwebringlogisticsbackintocities?ThecaseofParismetropolitanarea.Procedia‐SocialandBehavioralSciences,39,267‐281.

[7] Filippi,F.,Nuzzolo,A.,Comi,A.,DelleSite,P.(2010).Ex‐anteassessmentofurbanfreighttransportpolicies.Procedia‐SocialandBehavioralSciences,2(3),6332‐6342.

[8] Holguín‐Veras,J.(2008).Necessaryconditionsforoff‐hourdeliveriesandtheeffectivenessof urban freight roadpricing and alternative financial policies. TransportationResearchPartA,42(2),392‐413.

[9] Holguín‐Veras, J., Marquis, R., Brom, M. (2012). Economic impacts of staffed andunassistedoff‐hourdeliveriesinNewYorkCity.Procedia‐SocialandBehavioralSciences,39,34‐46.

[10] Holguín‐Veras, J.,Wang,C.,Browne,M.,DarvilleHodge,S.,Wojtowicz, J. (2014).TheNewYork City Off‐hour Delivery Project: Lessons for City Logistics. Procedia ‐ Social andBehavioralSciences,125,36‐48.

[11] Munuzuri, J., Larraneta J., Onieva L., Cortes, P. (2005). Solutions applicable by localadministrationsforurbanlogisticsimprovement.Cities,22(1),15‐28.

[12] Niches (2009). Polis guide on inner city night delivery. Niches, European Commission,Brussels.

[13] Russo, F., Comi, A. (2010). A classification of city logistics measures and connectedimpacts.Procedia‐SocialandBehavioralSciences,2,6355‐6365.

[14] Russo, F., Comi, A. (2012). City characteristics and urban goods movements: A way toenvironmental transportation system in a sustainable city. Procedia ‐ Social andBehavioralSciences,39,61‐73.

[15] Sathaye,N.,Li,Y.,Horvath,A.,Madanat,S. (2010).Unintendedenvironmental impactsofnighttimefreightlogisticsactivities.TransportationResearchPartA,44,642‐659.

[16] Tadić,S. (2014).Modeliranjeperformansi integrisanihcity logističkihrešenja,FacultyofTransportandTrafficEngineering,UniversityofBelgrade,Belgrade.(PhD)

[17] Tadić,S.,Zečević,S.,Krstić,M.(2014a).RankingofLogisticsSystemScenariosforCentralBusinessDistrict.Promet–Traffic&Transportation,26(2),159‐167.

[18] Tadić, S., Zečević, S., Krstić, M. (2014b). A novel hybrid MCDM model based on fuzzyDEMATEL,fuzzyANPandfuzzyVIKORforcitylogisticsconceptselection.ExpertSystemswithApplications,41(18),8112‐8128.

[19] Vilain, P., Wolfrom, P. (2001). Value pricing and freight traffic: Issues and industryconstraintsinshiftingfrompeaktooff‐peakmovements.TransportationResearchRecord,1707,64‐70.

[20] Woudsma,C., Jensen, J.,Karoglou,P.,Maoh,H. (2008). Logistics landuseand the city:Aspatial‐temporal modeling approach. Transportation Research Part E: Logistics, 44 (2),277‐297.

[21] Zečević, S. (2006). Razvojni koncept logističkog sistema na području Ada Huje. ExpertResearch,UrbanPlanningInstituteofBelgrade,Belgrade.