wsn lifetime and reliability analysis from the death ...akhattab/files/nouh2017.pdf ·...

DOI: 10.4018/IJHCR.2017070103

International Journal of Handheld Computing ResearchVolume 8 • Issue 3 • July-September 2017

Copyright©2017,IGIGlobal.CopyingordistributinginprintorelectronicformswithoutwrittenpermissionofIGIGlobalisprohibited.

WSN Lifetime and Reliability Analysis From the Death Criterion PerspectiveSara Nouh, The American University in Cairo, New Cairo, Egypt

Nada Elgaml, Cairo University, Giza, Egypt

Ahmed Khattab, Cairo University, Giza, Egypt

Samy S. Soliman, Cairo University and Zewail City of Science and Technology, Giza, Egypt

Ramez M. Daoud, The American University in Cairo, New Cairo, Egypt

Hassanein H. Amer, The American University in Cairo, New Cairo, Egypt

ABSTRACT

WirelessSensorNetworks(WSNs)arewidelyusedinnumerouscriticalapplications,andrequirethenetworktohaveaprolongedlifetimeandhightolerancetofailures.However,thebattery-operatedsensornodesusedinWSNscausethenetworktoberesource-constrained.Ontheonehand,thereisacontinuousurgetoefficientlyexploittheWSNenergy,andhence,prolongthenetworklifetime.Ontheotherhand,WSNnodefailuresarenotonlyattributedtobatterydrain.Nodefailurescanbecausedbyhardwareorsoftwaremalfunctioning.Inthisarticle,theauthorsassesstheimpactofthedeathcriteriononthenetworklifetimeandreliability.Itisrelatedhowthedatafromthedifferentsensorsareaggregatedtothedeathcriterion.Additionally,theimpactofthenumberofsensingcyclespernetworkmasteronthenetworklifetimeandenergyefficiencyforthedifferentconsidereddeathcriteria.Theeffectofthenetworkmasterselectionprocessontheenergyefficiencyisalsoexamined.Finally,theimpactofthedeathcriteriononthereliabilityoftheWSNisevaluated.

KeyWoRDSDeath Criteria, Electromagnetic Pollution, Energy Efficiency, Network Lifetime, Network Reliability, Wireless Sensor Network

1. INTRoDUCTIoN

WirelessSensorNetworks(WSNs)arebasedonsensornodesthatarebattery-operated.Accordingto theapplication forwhich theWSN isdesigned, sensornodes reportmeasurementsofcertainphenomenatoasinknode.WSNshavedifferentapplicationsincludingRADARdetection,agriculturemonitoring,smartcitiesandmanymore(Waghmare,Chatur,&Mathurkar,2016).Thispaperfocusesmainlyonmonitoringelectromagnetic(EM)pollution(Viani,Donelli,Oliveri,Massa&Trinchero,2011);however,thefindingsofthepaperareapplicabletosimilareventdetectionWSNapplications.

OneofthemainchallengesfacedbytheWSNtechnologyistheenergyconsumptionandtheenergyefficiencydue to theuseofbattery-operatedsensornodes. Itdirectlyaffects thenetworklifetime, which is typically defined as the time until the first WSN node fails due to battery

37


38

outage (Mahfoudh & Minet, 2008; Heinzelman, Chandrakasan, & Balakrishnan, 2000; Mamun,Ramakrishnan,&Srinivasan,2010).SuchadefinitionoftheWSNdeathimpliesthatwhenoneofthenode’senergyfallsbelowaspecificthresholdthatallowsittosendandreceivedata,thenthewholenetworkwillbeconsidereddead.Thisdefinitionofthenetworklifetimehadahugedisadvantageonthenetwork’senergyefficiencyaswellasthenetworklifetime.Thisisbecausethedeathofonenodewithinthenetworkdoesnotimplythattheremainingnodesarealsoincapableofcorrectlyperformingtheirtaskofdetectingthemonitoredevent.Inotherwords,theothernodesintheWSNhaveanamountofenergythatishighenoughtoallowthenetworktoperformtherequiredfunctions.

Inthispaper,ourgoalistoevaluatetheenergyefficiencyandreliabilityofthedifferentdefinitionsofthenetworklifetime.Morespecifically,weconsiderthecasesinwhichthenetworklifetimeisdefinedasatleastonesensorisstillfunctioning,atleasthalfthesensorsarefunctioning,andthelegacycasewhichrequiresallthenodestobefunctioninginordertoconsiderthenetworkalive.ThesethreedifferentdeathcriteriacoverthedifferentWSNapplications.OtherrelatedWSNlifetimedefinitionswerediscussedin(Kaur&Singh,2016).However,theyrelyonmobilesensors,gridoptimizationandenergyproficientclusteringtechniques.Moreover,severalclusterheadsexistinsuchnetworks,which are based on the LEACH algorithm (Heinzelman, Chandrakasan, & Balakrishnan, 2000;Heinzelman,Chandrakasan&Balakrishnan,2002).Incontrast,ourworkconsidersthewholenetworkasoneclusterwhichreliesonasinglenetworkmasterperround.Thishasalreadybeenproventoresultinaprolongednetworklifetimein(Botros,Elsayed,Amer,&El-Soudani,2009;Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010).Inthispaper,weevaluatethenetworklifetimecriteriaassumingapredefinednumberofcyclespernetworkmasterasopposedto(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010).Wealsostudytheimpactofrandomlychoosingthesensornodesthatserveasnetworkmasterduringtheoperationcycles,versus(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010;Nouh,Abbas,Seoud,Ali,Daoud,Amer,&Elsayed,2010)whichhadanorderedcircularselectionofthenetworkmasters.Thereasonforthatistoinvestigatewhetherthechoiceofthenetworkmasterhasasignificanteffectonthenetworkbehavior,andaccordingly,onthesensorsenergyornot.Unlike(Nouh,Khattab,Soliman,Daoud,&Amer2017),weanalyticallyevaluatetheimpactoftheWSNdeathcriteriononthereliabilityofthenetworksagainstothertypesofhardwareandsoftwarefailures.

Theremainderofthepaperisorganizedasfollows.InSection2,wedescribethenetworkmodel.Section3presentsthedifferentevaluateddeathcriteria.WeevaluatetheirenergyefficiencyunderdifferentsensingcyclelengthsanddifferentnetworkmasterselectionapproachesinSection4andSection5, respectively. InSection6,weanalyze the impactof thedeathcriteriaon thenetworkreliabilityagainsthardwareandsoftwarefailures.Section7concludesthepaper.

2. WSN SySTeM MoDeL

2.1. Network ArchitectureThedifferentdefinitionsofthenetworklifetimewillbeappliedonthenetworkarchitectureusedin(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010).Morespecifically,weconsidera100×100m2areathatisaffectedbyfourfrequencypollutersF1,F2,F3andF4.Eachofthesefrequencypollutersisplacedononesideofthearea(Ioriatti,Martinelli,Viani,Benedetti&Massa,2009).Atotalof100sensorsareuniformlydistributedovertheareainordertomonitorthelevelsofelectromagneticradiationsofthefourpolluters.Weassociate25sensorswitheachfrequencypolluterasillustratedinFigure1.However,only11sensorsthatareclosesttothepolluteroutofeach25sensorswillbeabletoreporttheviolation.Thisisbecausethepolluter’sradiationisonlycapableofcoveringhalfofthetotalarea.Furthermore,thesinkthataggregatesandanalyzesthedatacollectedbythenetworkmasters(NMs)islocatedinthecenterofthearea.Thispositionwasprovenin(Nouh,Abbas,Seoud,Ali,Daoud,Amer,&Elsayed,2010)toutilizethenetwork’senergyinthemostefficientway,andhence,increasesthenetworklifetime.Therestoftheparametersarelistedasfollows:


39

• Networksize:100×100m2

• NumberofSensors(N):100Sensors• InitialEnergypersensor:2J• Transmitter/ReceiverElectronics(Eelec):50nJ/bit• TransmitterAmplifier(Eamp):100pJ/bit/m2

• PathLossfactor(n):2• AggregationEnergy(Eagg):5nJ/bit/Signal• DatapacketsizesentbyactivenodestoNM(K):64bits• DatapacketsizesentbytheNMtothesink(K1):512bits• Datapacketsizeofsensingpowerlevels(K2):1bit• Sinklocation:fieldcenter• Distribution:HomogeneousDensity(Figure1)

2.2. electromagnetic Pollution MonitoringInordertoanalyzetheeffectofchangingthenetworkdeathcriteria,thesamemonitoringprocessassumedin(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010)willbeadoptedhere.Everyday,oneofthefrequencypollutersFi(startingwithpolluterF1)causespollutionduringthelastsixhoursofday.Onthenextday,F2sendsitsviolatingradiationsduringthesametimeoftheday.F3andF4followthesamemanneronthefollowingdays.ThisprocessrepeatsitselfeveryfourdaysstartingwithF1.Eachhourofthedayisconsideredtobeonecycle.Duringeachcycle,anetworkmaster(NM)ischosentocollectthedatafromthesensorsandsendittothesink.ThecriterionofchoosingtheNMduringacycleisacquiredfrom(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010).Itstartsbytheclosestsensortothesinkandkeepsmovinginacircularpathwayaroundthesink.EachsensorischeckedifitissuitabletoactasanNMusingapre-calculatedthresholdforeach

Figure 1. Uniformly distributed 100 sensors in an area of 100 x 100 m2 and surrounded by the four frequency polluters


40

NM.ThethresholdcomputestherequiredenergyforeachsensortoactasanNM,foronecycle,accordingtoitsdistancefromthesinkanditsdistancetotheremaining99sensors.AsimilarmethodofcalculatingthethresholdfortheNMisusedin(Botros,Elsayed,Amer&El-Soudani,2009).Atthebeginningoftheprocess,thethresholdsarecalculatedonlyonce,atthesink,andhence,representnorunningoverhead.Suchcalculationsrelyontheinformationgatheredaboutthesensors’locations.Laterduringtheprocessofthemonitoringsystem,thethresholdforeachNMisusedtocalculatethenumberofcyclesduringwhicheachsensorwillactasNM.Thethresholdiscalculatedasfollows:

E E N E K N E Ethreshold NMi rx s agg s prot tx_

= × + × × + + (1)

fori=1,2,…,100,where:

E E Krx elec= × (2)

and:

E E K Dtx amp NM to sink

n= × ×1

(3)

InEquation(1),theNsparameterisequaltothenumberofsensingnodes,whichis99inthiscase,becausethe100thistheNM.Furthermore,inEquation(3),theDNM to sinkisthecalculateddistancebetweentheithNMandthesink.

Oncethesensornode’senergyreachestheNMthreshold,itwillactasanordinaryactivenodeandanothernodewillbeelectedtobetheNMofthefollowingcyclesandsoon.Whenanyoftheactivenodesreachtheactivenodethreshold,whichisequaltotheenergythatallowsasensortosenseandsendpacketstoanNM,thenodewillbeconsidereddead.ThefunctionalityoftheWSNdependsonthepercentageornumberofactivenodes.Hence,thenetworklifetimecanbegenerallydefinedintermsofthenumberofcyclesduringwhichaminimumpercentageofthesensorsareactive.Inmanypreviousworks,thatpercentagewasconsidered100%(Heinzelman,Chandrakasan,&Balakrishnan,2000;Botros,Elsayed,Amer&El-Soudani,2009;Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010;Nouh,Abbas,Seoud,Ali,Daoud,Amer,&Elsayed,2010).Inthispaper,weassessvariousconsiderationsfortheWSN’slifetimeandproviderecommendationsaccordingtotheunderlyingapplicationoftheWSN.

3. NeTWoRK DeATH CRITeRIA

Thedeathofasinglenodeiscommonlyconsideredasanindicatorthatthewholenetworkisincapableoffunctioning.Thedisadvantageofsuchanassumptionisthatitunderestimatesthenetworklifetimebecauseevenifonlyonesensornodedies,thereisstillremainingenergypossessedbytherestofthenodes.Suchremainingenergycouldenablethenetworktosustainitsfunctionalityforalongertime.Hence,itismorepracticaltoviewthenetworkasfunctioningevenifsomenodeshavealreadydied.Accordingtothenumberofdeadnodesthatcanbetoleratedwithoutaffectingthefunctionalityofthenetwork,multiplenetworkdeathcriteriacanbedefined.Thedifferentdefinitionsofthedeathcriteriaaredrivenfromtheamountofinformationneededintheaggregationprocessofthedifferentreadingsofthenodes,whichsensethesamephenomenon(Chen,Shu,Zhang,Liu&Sun,2009).Forinstance,ifaggregationisdonebasedonANDingallthemeasurementsofallthenodes,thenetworklifetimeisdefinedasthetimetothefirstnodefailuresinceonenodefailureviolatestheANDrule.Ontheotherhand,ifaggregationisbasedontheORrule,itmeansthatthenetworkisconsideredalive


41

whenatleastonesensorisstillaliveandcorrectlyreportingthesensedphenomenon.Acompromisebetweenthe“AND”and“OR”rulesisthemajorityrule.Themajorityruleimpliesthatthenetworkisnotconsidereddeadifatleasthalfthenodessensingthetargetedphenomenonarestillfunctioning.

3.1. AND Aggregation Rule (All Nodes Must Be Alive)Thefirstdeathcriterionisthelegacyoneinwhichthenetworkisconsidereddeadwhenthefirstnodethatsensesthetargetedphenomenon,EMviolationinourcase,dies.Thiscriterionfollowsthedecisionofthelogical“AND”rule.Figure1illustratesthenetworkareadividedintofourzones,eachwithaspecificpolluter.Foreachpolluter,thereare11sensorswithinitspollutionrangethatcansensetheviolationandsendthepacketstothenetworkmaster.Ifoneofthe11sensorswhicharesensingtheEMviolationisdead,becauseithasreachedtheactivenodethreshold,thewholearea,andaccordinglythewholenetwork,willbeconsidereddead.However,itishighlyprobablethatthe10othersensorsmighthaveremainingenergythatcouldenablethemtoprolongthenetworkfunctionality.Insomecriticalapplicationsthatcannottoleratethedeathofonenodewithinthenetwork,thiscriterionistheoptimum,andsolutionshavetobesoughttoreplacethedeadnode.

3.2. oR Aggregation Rule (At Least one Node Is Alive)TheORruleisdefinedashavingonenodeinthezoneofeachpollutertoreporttheviolationeveniftherestofthenodesintheareaareconsidereddead.Therefore,theruleofthelogical“OR”isapplied.Thistechniqueisthemostenergyefficientonethatisexpectedtoprolongthenetworklifetime,sinceitconsumesthesensor’senergyatmost.ItmightbeneededinsomeWSNapplicationssuchasmonitoringunderwaterpipelines(Benhaddou&Al-Fuqaha,2015),wherethenetworkishardlyaccessibleandtheurgeofprolongingitslifetimeishighlyneeded.

3.3. Majority Aggregation Rule (Half the Nodes Are Alive)Theotherdeathcriterionthatweconsiderinthispaperistheonebasedonthemajorityaggregationrule.InordertobeabletoreportanEMviolation,halformoreofthesensornodesinagivenareaarerequiredtobeactive;otherwisethenetworkwillbeincapableofdetectingtheviolationsinsuchanarea.

SinceWSNsmightbeusedincertainapplications,wherethenetworkisplacedunderharshconditions,thefailureofonenodeormorecouldhighlyoccur.Thus,theORandmajoritycriteriaare advantageous as theymake thenetwork fault tolerant, because the reporting functionof thenetworkisnotaffectedbythedeathofoneorfewnodes(Manjunatha,Verma,&Srividya,2008).Furthermore,theybetterexploitthetotalenergyavailableinthenetworkascomparedtotheANDruleinwhichthenetworkisconsidereddeadwhilealotofresidualenergyisstillavailable.Therefore,thedifferentdeathcriteriaareveryapplication-dependent,andtheirsignificancediffersfromoneapplicationtotheother.

3.4. WSN Lifetime Under Different Death CriteriaFirst,weevaluatetheperformanceoftheaforementioneddeathcriteria.ThethreedeathcriteriaweresimulatedusingMATLAB(MATLAB,2011)inthesystemmodeldescribedinSection2.Figure2showsthedeathofeachareaaccordingtoeachdeathcriteria,andconsequently,thecorrespondinglifetimeofthenetwork.The1stgroupofpoints(atthebottom-leftofFigure2)showsthedeathofthe1stnodeineacharea.Thenthe2ndgroupofpoints(atthecenterofFigure2)displaysthedeathof6nodes,i.e.,themajority,ineacharea.Finally,thelastgroupshowsthedeathofthelastnodeineacharea,usingtheORrule.Notethat,forthelastgroup,thefirstpolluterF1dieslast,whichmeansitcansustainalongerlifetimecomparedtotheotherareas.However,beingtheareawiththehighestlifetimedoesnotmeanthatunderallthreedeathcriteria,thisareawillnecessarilyhavethehighestlifetime.AsshowninFigure2,theF1areacomesatthesecondplace,whentheANDruleisconsideredandcomesatthe3rdplacewhenthemajorityruleisused.Thisisattributedtolocation


42

ofthenodesandthenetworkmasterlocationineachcycle,whichaffecttheenergyconsumptionofeachnodedifferently.

4. IMPACT oF THe NUMBeR oF CyCLeS PeR NeTWoRK MASTeR

4.1. Selecting a Fixed Number of Cycles per NMInthepreviousscenario,anysensorselectedasanNMoperatesasanNMforanumberofCNMicyclesuntilitcompletelydepletesitsenergybyreachingtheNMthreshold,Ethreshold_NMi.Atthebeginningofthesubsequentround,thenextavailablesensorischosenastheactingNM.Consequently,thenumberofcyclesperNMisdependentonthesensor’senergyanddiffersfromoneNMtotheother.Theoptimumnumberofcyclesperroundwasinvestigatedin(Botros,Elsayed,Amer,&El-Soudani,2009)toelongatethenetworklifetime.However,theresultsobtainedin(Botros,Elsayed,Amer,&El-Soudani,2009)aimedtosolvethedrawbacksin(Heinzelman,Chandrakasan,&Balakrishnan,2000;Heinzelman,Chandrakasan&Balakrishnan,2002),andhence,arenotapplicableinthecontextofourmodel.Inwhatfollows,westudytheeffectofsettingapredefinedNMcyclecountsuchthateachsensoractsasanNMforacertainnumberofcyclesirrespectiveofitsresidualenergy.Wefocusonthefollowingnumbersofcyclesperround:

• 100cyclesperNMround• 1000cyclesperNMround• 10000cyclesperNMround

Notethatthesenumbersarechosenbasedonresultsfromtheprevioussection.ItcanbeseenthattheaveragenumberofcyclesperNM,CNMiisaround10000cyclesasshowninFigure2.Hence,thisnumberischosenasthehighestcyclecount.Moreover,twootherpossibilitieswillbeinvestigatedbyreducingthecyclecountsizebyafactorof10and100cycles.Therationalebehindsuchachange,ascomparedtothepreviousscenario,isthatwithsmallercyclecounts,thenodeactingasanNMwill

Figure 2. The different death criteria are illustrated by showing the lifetime versus the number of dead nodes


43

notdepletethemajorityofitsenergywhileactingasanNMandwillhaveenoughresidualenergytoactasanon-NMnodeforalongernumberofcycles.ThiswillalsoenabletherotationoftheNMrolemorefrequently,resultinginanevenenergydissipationprofileforallthesensornodes.

Figure3 illustrates the lifetimecurvesof eachof the aforementionedpredefinednumberofcyclesperround.Forcomparison,thelifetimecurveobtainedinFigure2isshownandislabeledwithMaxCycles/NM.

Figure3impliesthatusingthemaximumnumberofcyclesperNM,whichiscalculatedaccordingtoeachNM’senergyconsumption,resultsinahigherlifetimeinmostofthedeathcriteria.However,ifotherlifetimedefinitionsareadopted,therelativelifetimebehaviorwillvary.Forexample,ifanapplicationrequirestheuseoftheANDaggregationapproachforthenetworkdeathdefinition,thenthemaximumCycle/NMschemeisnotthebestschemeintermsofnetworklifetime.Onthecontrary,alltheotherpredefinedcyclesperNMcurvesachievehigherlifetimesduringthedeathofthe1stFiarea.Hence,onecanpredictfromFigure3,whichisthebestcurvethatprolongsthelifetime,thattherestillexistotheraspectsthatshouldbetakenintoconsideration,suchas:

• Identifyingtherelevantdeathcriteriaaccordingtotheapplicationrequirement;• AspecificdeadareaFiareacouldbedeterminedaheadalongsidethedefinitionofthenetwork

deathcriteria;• Finally,thereisatradeoffbetweenmaximizingthelifetimeandexploitingthesensor’senergy

efficientlyaccordingtotheapplication.Whenmaximizingthelifetimeisneeded,thenFigure3willbesufficienttoidentifythat.Otherwise,thecurvethatmostlyconsumesthenetwork’senergyefficientlyistobeidentified.

Therefore,theenergyconsumedbythefourdifferentlifetimecurvesshouldbeinvestigatedfurther.

4.2. energy Consumption CharacteristicsInorder togetadeeper lookinto theenergyconsumptionof thenetwork in theconsideredfourscenarios,twoaspectswillbeinvestigated.

Figure 3. Different lifetime curves that illustrate the different cycle number per NM


44

4.2.1. The Average Remaining EnergyFigure4shows thatusinghighernumberofcyclesperNMround,either through themaximumtechniqueorthroughusingapredefinednumberofcyclesperNMsuchas10000cycles,willresultinaninefficientuseofthenetwork’senergy.Onthecontrary,using100cyclesperroundand1000cyclesperroundachievealowerremainingenergylevelcomparedtotheothertwoscenarios.Thereasonforthatisthatwhenalowernumberofcyclesperroundisadopted,thelocationoftheNMischangedmorefrequentlywhichreducesthepossibilityofstarvingthenodesfarfromtheNMforalongperiodoftime.Avoidingthiswillpreventthenodesfromexploitingtheirenergyallatonce.

4.2.2. The Variance of the Remaining EnergyFigure5showsthestandarddeviationoftheremainingenergyofthesensornodesversusthecyclesofthenetwork.ThecurvesinFigure5emphasizetheobservationsfromFigure4.Thesecurvescanbedescribedasfollows.Atthebeginningofnetworkoperation,allthenodeshavethesameinitialenergy.Whentheprocessstarts,someofthenodesstarttolosetheirinitialenergyfasterthantheothernodes.Hence,theyreachapointwherethedifferenceintheremainingenergyistoohigh,assomenodesarealreadydead,withaverylowremainingenergy,whileothernodesstillcontainhighremainingenergyandareactingasnetworkmasters.Afterwards,thenodeswithhighremainingenergystarttolosetheirenergies.Thisiswhenthestandarddeviationdecreasesagain.Whenthenodesbecomedeadbyreachingthespecifiedthreshold,thetotalremainingenergyinthewholenetworkwillbealmostthesame.Atthispointthestandarddeviationcurvewillbeapproachingthezerolevel.

5. IMPACT oF NM SeLeCTIoN APPRoACH

AsmentionedintheSection4,thechoiceoftheNMisinacircularorderstartingfromthesensorsneartothesinktothoseawayfromthesink.InordertomakesurethatsuchNMselectionproceduredoesnotcontributetoourfindings,arandomselectionoftheNMisinvestigated.Itisexpectedthat

Figure 4. Average remaining energy for the four scenarios using the ordered


45

thebehaviorofthescenariosdescribedinSection4wouldbeindependentoftheNMselection,astheyarenottiedspecificallytothissystemmodelandshouldbeappliedonanyotherapplication.

Figure6illustratesthefourNMcountschemesdiscussedintheprevioussectionusingrandomselectionofNM.Eachofthe100sensorsisrandomlyselectedtoserveasanNMduringaperiodofroundswithoutaspecificorder.DespitetheuseoftherandomselectionoftheNM,thefourschemeshavesimilarbehaviortothatshowninFigure3.Forexample,themaximumschemeremainsthehighestwithrespecttototallifetimewhilethe10000cyclesperroundcomesnext.Also,theothertwocurvesof1000and100cycles/NMfollowthesamebehaviorasinFigure3.TheonlydifferencebetweenFigure3andFigure6isthelifetimevalueforeachcurve.Figure6impliesthatarandomselectionofthenetworkmasterresultsinahigherlifetimeingeneral.ThereasonforthatisthatchangingthelocationoftheNMmorefrequentlyaspreviouslymentionedinSection4causesthenetworknottoexploitonespecificareaatatimeandinsteaditaveragesoverthewholenetwork.

Figures7and8alsoshowsimilarresultscompared toFigures4and5.TheonlydifferencebetweenFigure4andFigure7isthatinFigure7theremainingenergyisconsumedmoreefficientlythaninFigure4.Likewise,Figure8showsasimilarbehaviorofthestandarddeviation.Nevertheless,themainsignificanceofthisexperimentistoshowthatthepreviouslydiscussedschemesbehaveindependentlyofthesystemmodelassumptions,andhenceforth,thewholesystemdoesnotrelyonaspecificcaseandworksforvariousapplicationsandscenarios.

6. ReLIABILITy ANALySIS

Intheabovediscussion,allsensornodeswereassumedtobefullyoperationalaslongastheirbatterieshadenoughenergytocompletetherequiredtasks.Ifasensornodeexperiencesahardware/softwarefailure,theabovecalculationswillnolongerbevalid.Thefollowinganalysisobtainstheprobabilitythat theWSNsystemwill succeed inoperatingcorrectly for theentire lifetimecalculated in the

Figure 5. Standard deviation of the remaining energy for the four scenarios using the ordered choice of NMs


46

Figure 6. Different lifetime curves illustrating the different cycle number per NM using the random selection of NMs

Figure 7. Average remaining energy for the four scenarios using the random choice of NMs


47

previoussections.WelimitouranalysistothedeterministicNMselectioncasedevelopedin(Seoud,Nouh,Abbass,Ali,Daoud,Amer&Elsayed,2010)inwhichthesensornodeswillfailinacertainordergovernedbythenodelocations.

WeassumethatallnodesintheWSNsystemareidentical.Furthermore,letthefailurerateofanyofthesensorsbeλs.Thetimetofailureistypicallyassumedtofollowtheexponentialdistribution(Siewiorek&Swarz,1998);therefore,thefailurerateλsisconstant.

6.1. Reliability Analysis of AND RuleThesystemwillceasetofunctionproperlyassoonasthefirstsensornode(outofthe100nodes)fails.ForthesystemlifetimetobeequaltotheonecalculatedinSection4,allsensornodesmustnothavefailed.Inotherwords,ifthislifetimeistand,thenall100sensornodesmustbeoperatingcorrectlytilltimet = tand.Let:

R t es

ts( ) = −λ (4)

whereRs(t) is thesensornodereliability, i.e., theprobability that thissensornodeisperformingcorrectlyattimetgiventhatitwasoperationalattimet=0(Siewiorek&Swarz,1998).ItisimportantheretonotethatRs(t)hasnorelationtotheamountofenergyremaininginthenode’sbattery.AsensornodefailureinthecontextofRs(t)isduetoahardware/softwarefailure.Fromareliabilitystandpoint,theWSNsystemisa“series”system;ifanyofthe100sensorsfails(hardware/softwarefailure),thelifetimecalculationsinSection4abovewillnolongerbevalid.Hence:

R t R twsn and s and( ) = ( )100 (5)

Figure 8. Standard deviation of the remaining energy for the four scenarios using the random choice of NMs


48

Rwsn(tand)istheprobabilitythattheWSNwillfunctioncorrectlyuptoitsexpectedlifetimetand.ForapredeterminedvalueRwsn(tand),thevalueofλscanbedeterminedandtheappropriatetype(fromareliabilitypointofview)ofsensornodescanbeusedinthesystem.

6.2. Reliability Analysis of oR RuleIntheORaggregationrule,theWSNsystemreachesitsexpectedlifetimetor(ascomputedinSection4)whenall11sensornodesmeasuringanyoneofthefrequencypollutersdonothaveenoughbatteryenergytocarryouttheirfunctions.The11thsensor(foroneofthefourpolluters)willnothaveenoughenergyatt=tor.Theother10sensornodesmeasuringthissamefrequencypolluterwillhavedepletedtheirenergyatt<tor.Forthesensornodesmeasuringtheotherpolluters,somewillstillbeoperatingwhileotherswillnot,asshowninFigure9inwhichthenodeswith0lifetimesarethosewhicharestilloperating,whilethenodeswithnon-zerolifetimearethosewhichstoppedfunctioning.Lettor_ibethetimewhensensori(foranyofthefourpolluters)doesnothaveenoughenergytooperatecorrectly.ForthecalculationsinSection4abovetobevalid,eachsensorimustnothavefailed(hardware/software)beforetor_i.Ifitfailsafterthattime(evenbeforet=tor),thelifetimecalculatedabovewillbevalid.Letusdividethe100sensorsintotwogroups:(1)Operating:sensorswhichstillhaveenoughenergyatt=tor,2)LowBattery:sensorswhichhavedepletedtheirenergyatt<tor.FortheWSNtooperateaspredictedinSection4,the“operating”sensornodesmustnotfail(hardware/software)beforet=tor.Anysensoriinthe“LowBattery”groupofsensorsmustsurviveatleasttillt=tor_i:

R t e ewsn or

operating

t

LowBattery

ts or s o( ) =

∏ ∏− −λ λ rr i_

(6)

Again, for apredeterminedRwsn(tor),λs canbeobtained and the appropriate sensors (fromareliabilitypointofview)canbeusedinthesystem.

Figure 9. Time at which the individual nodes in Figure 2 die when OR aggregation is used


49

ForthecaseoftheMajorityrule,thesamereasoningandanalysiscanbeused,butinsteadofwaitingforthe11thsensortofail,thesystemwillceasetofunctionassoonasoneofthefourpolluterslosesits6thsensor.

7. CoNCLUSIoN

ThispaperhasevaluatedthreedifferentdefinitionsofthenetworkdeathcriterionthatarerelatedtodifferentWSNmeasurementaggregationtechniques.Choosingbetweenallthesetechniquesisapplication-dependent,sinceeverytechniquebestfitscertainWSNapplications.WeexaminedtheimpactofthedifferentdeathcriteriaontheWSNlifetime.Furthermore,theimpactofthenumberofsensingcyclesisexaminedtoshowthedifferencebetweenexploitingtheNMenergytoitsmaximumallatonce,versusactingasNMseveraltimesandconsumingtheNMenergyonseparateintervals.OurresultsshowthatusingapredefinedlownumberofcyclesperNMwillresultinamoreefficientuseofthenetwork’senergy.However,thereistradeoffhere,sincethenetworklifetimedecreaseswhenthepredefinednumberofcyclesislow.

WehavestudiedthesignificanceoftheprocessforselectingtheNM.Insteadofthecircularorderedpathselection,wehaveconsideredrandomNMselection.ThishasproventhatdespitethechangeoftheNMselection,thepreviousconclusionswerenotchanged,whichindicatesthattheproposedscenariosarenotalignedwithaspecificsystemmodel,howevercanbeimplementedonotherWSNapplications.

Furthermore,weanalyticallyevaluatedthereliabilityofWSNforthedifferentdeathcriteriaconsideredinthepaper.SuchanalysisallowstheWSNdesignertochoosetheappropriatesetofsensornodesthatcanachieveacertainlifetimewithatargetlevelofreliability.

Inconclusion,itisveryimportanttoidentifythetargetedWSNapplicationfirstandthendecidewhethertheaimistohaveaprolongedlifetime,toconsumethenetwork’senergyefficiently,ortobemorereliabletofailures.Accordingly,thevariousdeathcriteriaandNMcount/selectionapproachillustrated in this paper couldbeveryhelpful inobtaining themost adequate conditions for theapplication,meaningchoosingthenumberofcyclesperNMmasterandthebestdeathcriteriatoachievethebestfitforthedesiredapplication.


50

ReFeReNCeS

Benhaddou,D.,&Al-Fuqaha,A.(2015).Wireless Sensor and Mobile Ad-Hoc Networks: Vehicular and Space Applications.NewYork,NY:Springer.doi:10.1007/978-1-4939-2468-4

Botros,S.S.,Elsayed,H.M.,Amer,H.H.,&El-Soudani,M.S.(2009).LifetimeOptimizationinHierarchicalWireless Sensor Networks. IEEE 14th International Conference on Emerging Technologies and Factory Automation (ETFA),Mallorca,Spain.doi:10.1109/etfa.2009.5347016

Chen,Y.,Shu,J.,Zhang,S.,Liu,L.,&Sun,L.(2009).DataFusioninWirelessSensorNetworks.InProceedings of theIEEE Second International Symposium on Electronic Commerce and Security,NanchangCity,China.doi:10.1109/ISECS.2009.170

Heinzelman,W.,Chandrakasan,A.,&Balakrishnan,H.(2000).Energy-EfficientRoutingProtocolsforWirelessMicrosensorNetworks.InProceedings of the33rd Hawaii International Conference on System Sciences (HICSS),Maui,HI.doi:10.1109/HICSS.2000.926982

Heinzelman,W.,Chandrakasan,A.,&Balakrishnan,H.(2002).AnApplicationSpecificProtocolArchitectureforWirelessMicrosensorNetworks.IEEE Transactions on Wireless Communications,1(4),660–670.doi:10.1109/TWC.2002.804190

Ioriatti,L.,Martinelli,M.,Viani,F.,Benedetti,M.,&Massa,A.(2009).Real-timeDistributedMonitoringofElectromagneticPollutioninUrbanEnvironments.InProceedings of theIEEE International Geoscience and Remote Sensing Symposium (IGARSS),CapeTown,SouthAfrica.doi:10.1109/IGARSS.2009.5417724

Kaur,T.,&Singh,M. (2016).LifetimeEnhancementbyOptimizationofGridUsingMobilityandEnergyProficientClusteringTechniqueofWirelessSensorNetwork.InProceedings of theInternational Conference on Computing for Sustainable Global Development(pp.1550-1555).

Mahfoudh, S., & Minet, P. (2008). Survey of Energy Efficient Strategies in Wireless Ad Hoc and SensorNetworks.InProceedings of theSeventh International Conference on Networking (ICN 2008),Cancun,Mexico.doi:10.1109/ICN.2008.55

Mamun,Q.,Ramakrishnan,S.,&Srinivasan,B.(2010).SelectingMemberNodesinaChainOrientedWSN.InProceedings of theIEEE Wireless Communications and Networking Conference (WCNC’03).doi:10.1109/WCNC.2010.5506377

Manjunatha,P.,Verma,A.,&Srividya,A.(2008).Multi-sensorDataFusioninClusterBasedWirelessSensorNetworks Using Fuzzy Logic Method. In Proceedings of the IEEE Region 10 and the Third International Conference on Industrial and Information Systems,Kharagpur,India.doi:10.1109/iciinfs.2008.4798453

MathWorks,Inc.(2011).MATLABandStatisticsToolboxRelease(2011b).Natick,MA,USA.

Nouh,S.,Abbas,R.A.,Seoud,D.A.,Ali,N.A.,Daoud,R.M.,Amer,H.H.,&Elsayed,H.M.(2010).EffectofNodeDistributionsonLifetimeofWirelessSensorNetworks. InProceedings of the IEEE International Symposium on Industrial Electronics (ISIE),Bari,Italy.doi:10.1109/ISIE.2010.5637472

Nouh,S.,Khattab,A.,Soliman,S.S.,Daoud,R.M.,&Amer,H.H.(2017).OntheImpactoftheDeathCriterionontheWSNLifetime.InProceedings of theIEEE International Conference on Electro Information Technology (EIT2017),Lincoln,NE.doi:10.1109/EIT.2017.8053328

Seoud,D.A.,Nouh,S.,Abbass,R.A.,Ali,N.A.,Daoud,R.M.,Amer,H.H.,&Elsayed,H.M. (2010).Monitoring Electromagnetic Pollution Using Wireless Sensor Networks. In Proceedings of the IEEE 15th International Conference on Emerging Technologies and Factory Automation (ETFA),Bilbao,Spain.doi:10.1109/etfa.2010.5641288

Siewiorek,D.P.,&Swarz,R.S.(1998).ReliableComputerSystemsDesignandEvaluation.Natick,MA:AKPeters.

Viani,F.,Donelli,M.,Oliveri,G.,Massa,A.,&Trinchero,D.(2011).AWSN-BasedSystemforReal-timeElectromagnetic Monitoring. In Proceedings of the 2011 IEEE International Symposium on Antennas and Propagation (APSURSI),Spokane,WA.doi:10.1109/APS.2011.5997195

Waghmare, K. A., Chatur, P. N., & Mathurkar, S. S. (2016). Efficient Data Aggregation Methodology forWirelessSensorNetwork.InProceedings of theInternational Conference on Wireless Communications, Signal Processing and Networking (WiSPNET)(pp.137-139).doi:10.1109/WiSPNET.2016.7566107

http://dx.doi.org/10.1007/978-1-4939-2468-4

http://dx.doi.org/10.1109/etfa.2009.5347016

http://dx.doi.org/10.1109/ISECS.2009.170

http://dx.doi.org/10.1109/HICSS.2000.926982

http://dx.doi.org/10.1109/TWC.2002.804190

http://dx.doi.org/10.1109/TWC.2002.804190

http://dx.doi.org/10.1109/IGARSS.2009.5417724

http://dx.doi.org/10.1109/ICN.2008.55

http://dx.doi.org/10.1109/WCNC.2010.5506377

http://dx.doi.org/10.1109/iciinfs.2008.4798453

http://dx.doi.org/10.1109/ISIE.2010.5637472

http://dx.doi.org/10.1109/EIT.2017.8053328



http://dx.doi.org/10.1109/APS.2011.5997195

http://dx.doi.org/10.1109/WiSPNET.2016.7566107


51

Sara Nouh received her B.Sc. (honors) and M.Sc. in Electronics and Communications Engineering from the American University in Cairo (AUC), Egypt, in 2010 and 2017, respectively. She is currently employed as a Senior Sales Engineer Analyst at Dell EMC, which is one of the biggest computer data storage companies in the world. Her research interests are in Wireless Sensor Networks (WSN) and Network Functions Virtualization (NFV). She received several certificates of academic honor at AUC for her high performance. Similarly, she was often awarded at DELL EMC for her great work achievements. She received the Lady of the Year Award and the DELL Technologies Customer Experience (CX) Champ Award in 2016. Additionally, she received the Presales Pinnacle Award, one of the most important awards at DELL EMC, for her great community services.

Nada Elgaml received her B.Sc. in Electronics and Communications Engineering from Cairo University, Egypt, in 2014 and she is working towards her master’s degree at Cairo university. She is currently a research assistant at the EECE Department in Cairo University and the American University in Cairo. Her research interests are in cognitive radios, vehicular networks and wireless sensor networks.

Ahmed Khattab received his B.Sc. (honors) and M.Sc. in Electronics and Communications Engineering from Cairo University, Egypt, in 2002 and 2004, respectively. He received the Master of Electrical Engineering degree from Rice University, and his Ph.D. in Computer Engineering from the University of Louisiana at Lafayette, USA, in 2009 and 2011, respectively. He is currently an Associate Professor at the EECE Department in Cairo University. He is a senior member of the IEEE. His research interests are in wireless communication networks, the Internet of Things and embedded systems. He has authored/co-authored 3 books, over 50 journal and conference publications, and a patent application. He won the best student paper award of the UL IEEE Computer Society in 2010 and 2011.

Samy S. Soliman received his B.Sc. and M.Sc. (with honor) degrees from the Electronics and Electrical Communications Engineering Department, Faculty of Engineering, Cairo University, Egypt, in 2007 and 2009, respectively. He received his Ph.D. degree from the Electrical and Computer Engineering Department, University of Alberta in 2014. He was awarded the Cairo University prize of Engineer Nabil El-Gebaly as the highest standing student in Telecommunications and the Cairo University prize of Engineer Reda Hamza as the highest standing student in Electronics, both in 2008. Dr. Soliman worked as an Assistant Lecturer at the Electronics and Electrical Communications Department, Cairo University, as well as a Research and Teaching Assistant at the Electronics Engineering Department, American University in Cairo (AUC). He joined the AITF/iCORE Wireless Communications Laboratory (iWCL) in 2009 as a Research Assistant, and worked as a Post-Secondary Teaching Assistant at the Electrical and Computer Engineering Department, University of Alberta, Canada. He held a Post-Doctoral Research Fellowship at the University of British Columbia, Canada. He is currently an Assistant Professor at the Electronics and Electrical Communications Department, Cairo University as well at Zewail City University of Science and Technology, Egypt. Dr. Soliman’s research interests include 5G Communication Systems, Wireless Sensor Networks, Wireless Cooperative Networks, Multiple Input Multiple Output (MIMO) Systems, and Ultra-wide Bandwidth Wireless Systems.

Ramèz M. Daoud received his B.Sc. and M.Sc. in Control Engineering in 2000 and 2004 respectively from Cairo University, Egypt and his D.Sc. in Control and Industrial Informatics with High Honors from Universite de Valenciennes, France in 2008. He received the “Best Industrial Application Certificate” in 2005 from the IEEE SMC. He is a founding member of the SEAD research group. He is a member of the IEEE IES NBCS, the IEEE IES TCFA-FT. His research interests include digital circuit design, wireless communications, automotive control, vehicular technology, car automation, real-time systems, networked control systems, and industrial informatics.

Hassanein H. Amer is a professor in the Department of Electronics and Communications Engineering at The American University of Cairo. He was the founding chair of the department, serving from September 2004 through August 2008. Amer received his BSc in electronics engineering from Cairo University in 1978, and his MSc and PhD degrees in electrical engineering from Stanford University, California, United States in 1983 and 1987 respectively. He founded the SEAD research group in 2003. He is a member of the IEEE Reliability Society since 1985, a member of the Industrial Relations Committee for the IEEE SMC Society and an honorary member of the International Society for Advanced Research.

wsn lifetime and reliability analysis from the death ...akhattab/files/nouh2017.pdf ·...

Documents