1 trafﬁc-aware efﬁcient mapping of wireless body area...

1536-1233 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMC.2018.2822279, IEEETransactions on Mobile Computing

1

Traffic-Aware Efficient Mapping of Wireless Body AreaNetworks to Health Cloud Service Providers in Critical

Emergency SituationsSudip Misra, Senior Member, IEEE, and Amit Samanta, Student Member, IEEE

Abstract—In a post-disaster situation, increased concentration of patients in an area increases the traffic load of the networksignificantly, which degrades its performance with respect to mapping cost and network throughput. Therefore, to manage theincreased traffic load and to provide ubiquitous medical services, we propose a disease-centric health-care management system usingwireless body area networks (WBAN) in the presence of multiple health-cloud service providers (H-CSP). The theory of Social NetworkAnalysis (SNA) is adopted to optimize the computational complexity and the traffic load of the network in an area, considering differentdisease types and the criticality indices of the WBANs. In such a scenario, Disease-centric Patient Group (DPG) formation amongcoexisting WBANs ensures optimized traffic load and reduced computational complexity. However, the formation of DPG alone is notsufficient to provide Quality-of-Service (QoS) to each WBAN. Therefore, to address these issues, we formulate a pricing model for theefficient mapping of critical WBANs from a DPG to a H-CSP to optimize the expected packet delivery delay and the networkthroughput. Consequently, to identify the critical WBANs from a DPG, we design a decision parameter based on an assortment ofselection parameters. The performance of the Efficient Healthcare Management (HCM) scheme is analyzed based on distinctmeasures such as cost effectiveness, service delay, and throughput. Simulation results exhibit significant improvement in the networkperformance over the existing schemes.

Index Terms—Wireless Body Area Networks, Cloud Computing, Disease-centric Patient Grouping, Heterogeneous Health CloudService Provider, Quality of Service, Energy Efficient, Traffic Aware, Efficient Mapping

F

1 INTRODUCTION

In a post-disaster situation, monitoring of affected patients inan area and providing them with reliable and ubiquitous elec-tronic healthcare services, is a major challenge [1]. Therefore,to manage such situations, a cloud-assisted WBAN architectureprovides cost-effective and real-time services to the affected vic-tims. Cloud-assisted WBAN is an infrastructural and systematicintegration of traditional WBAN with cloud [2]. In a conven-tional WBAN architecture [3]–[5], the body sensors located inthe vicinity of human tissue sense the physiological signals ofthe patients, process them, and then send the sensed informa-tion to the Local Processing Unit (LPU). The LPU sends themedical data to the servers through the local APs for analysis bythe medical experts. In case of cloud-assisted WBANs, the localAPs send the medical data directly to the health cloud serviceprovider (H-CSP). Therefore, WBAN equipped patients can getcost-effective and ubiquitous electronic healthcare services in acritical emergency situation. Further, a cloud-assisted WBANprovides adequate services for wide ambulatory and sportsapplications [6], [7].

Motivation: Increased number of WBANs in a specificarea degrades the performance of each WBAN in terms of end-to-end packet delivery delay and network throughput. There-fore, the management of increased traffic load of WBANs is amajor challenge, as each WBAN carries sensitive medical data.In a conventional WBAN architecture, medical data specific toa particular disease is dedicatedly stored in a particular serverto manage the data efficiently. Therefore, when specific disease-affected patients are not present in that area, then these serversbecome unutilized and the network management cost increases.

S. Misra, and A. Samanta are with the Department of Computer Science andEngineering, Indian Institute of Technology, Kharagpur, 721302, India, Email:smisra, [email protected]

In this context, the integration of cloud services to a WBANplatform provides cost effective, elastic and real-time healthcareservices [8]. Due to the use of cloud services, if the WBANsspecific to particular diseases are not present in that area, cloudservices can be utilized by giving the same to other WBANs.In normal situations, resource demand from each WBAN maydiffer, which may even increase during emergencies. Due tothe fact that different disease-specific WBANs demand differentkinds of services, each such WBAN needs to choose an optimalH-CSP among heterogeneous cloud service providers.

Contribution: Our work attempts to identify the prob-lem of traffic load minimization and selection of an optimalH-CSP pricing policy for heterogeneous WBANs in a cloud-enabled platform. A cloud infrastructure provides resourceson requirement to a WBAN in an ubiquitous manner. EachWBAN can store medical data, depending upon the patients’medical situations, without being deeply concerned about theinfrastructure of the cloud [9]. This means that a WBAN at-tempting access to resources may not deploy its own resourceinfrastructure and even may not be aware of the physicalpresence of the deployed infrastructure. The cloud users onlypay for the amount of resources used by them for a specifiedtime period. Thus, it leads to an overall resource optimizationdue to sharing of resources and reduction in the overall cost ofusage. The major contributions of this manuscript are discussedas follows :

i) Disease-centric patient grouping is considered, depend-ing on the syndrome associated with these diseasessuch as epidemic cholera, glandular fever, and epi-demic parotitis, to minimize the computational com-plexity and traffic load for epidemic emergency situa-tion in a hospital environment.

ii) The work focuses on the optimal mapping of critical



2

WBANs to a particular H-CSP among several heteroge-neous H-CSPs, based on the capacity, robustness, andcapacity in terms of available resources, service delayand pricing policy of the H-CSPs.

iii) The proposed solution also maintains the energy effi-ciency of each WBAN belonging to a particular DPG inthe presence of heterogeneous H-CSP.

The manuscript is organized as follows. In Section 2, we crisplydiscuss the existing works on cloud-assisted WBANs. Section 3describes the system and mathematical model of the proposedscheme. In Section 5, we propose an optimization problem forenergy-efficient social relation grouping to optimize the trafficload of WBANs. Section 6 presents the optimal pricing policyin the presence of heterogeneous H-CSPs. Section 8 presents thesimulation results and Section 9 discuss the future works.

2 RELATED WORK

WBANs are used to monitor physiological parameters of acommunity of people. They produces huge volumes of medicaldata packets [10]. To store, analyze, and process such data,cloud computing provides adjustable storage and processinginfrastructure to analyze the data streams generated in WBANsfor both online and offline algorithms.

In this domain, Giancarlo et al. proposed a SaaS-based ap-proach for building a community of WBANs to support cloud-assisted WBAN applications, named BodyCloud [11]. Body-Cloud is an application-level infrastructure to integrate cloudand medical resources having multi-tier. Similarly, Fortino et al.[2] deployed cloud-assisted WBANs and identified the impor-tant issues, which are required to be solved for advancementand execution in advanced healthcare. This present system isoverviewed and wrapped based on the necessities of creat-ing efficient cloud-assisted WBAN architecture. Consequently,Quwaider et al. [12] proposed a novel cloudlet-based for pro-ductive data aggregation in WBANs. In this work, the authorsfocused on a large-scale data-generating WBANs to be availableto the end-user or to the service provider in a reliable man-ner. Additionally, Zhang et al. proposed adaptive map-reducedframework to scale the capability of cloud resources for real-time applications [13]. In order to address these challenges,Chen at al. proposed the integration of WBANs with Long TermEvolution (LTE) infrastructure, to support high user mobility[14]. They also proposed an efficient scheme, termed as nameddata networking (NDN) to support rich and adaptive mediastreaming for healthcare content size suitability with low-costand bandwidth saving. Due to the energy constrained nature ofthe body sensors, the huge amounts of packets aggregated bythe WBANs necessitates powerful and secure storage, and anefficient query processing mechanism, while considering bothreal-time and energy constraints of WBANs. Diallo et al. [15]proposed a new architecture that integrates statistical modelingtechnologies into cloud-based WBANs, to maintain privacy forstorage infrastructure and to minimize the overhead of real-time user query processing. Due to the huge volume of gen-erated data and their long-term processing, the computationalpower and energy consumption of the data centers increasestremendously. Rachkidi et al. proposed a cooperative approachbetween WBANs and H-CSPs for efficient provisioning ofhealth services in health-cloud to minimize service latency [16].

In a post-disaster scenario, it is important to aggregate phys-iological sensor data in an efficient manner, and channelizedthem to the cloud platform with minimum delay. Abbas etal. proposed personalized healthcare services for disease riskmanagement using cloud services [17]. Additionally, Prasad

et al. [18] proposed an optimal resource management schemein cloud infrastructure using Auction mechanism. Zhou etal. proposed a blind online scheduling approach for mobilecloud environments to assign available severs based on theusers request [19]. Samanta et al. [20] proposed an efficientarchitecture with varying traffic in an epidemic medical emer-gency situation. In this work, they proposed to form RelationalPatient Group (RPG), depending on different disease types andsyndromes. Similarly, Samanta et al. proposed a joint dynamicresource allocation and load balancing scheme for WBANs inthe presence of poor link-quality [21]–[23]. Additionally, Ibarraet al. proposed a joint power and QoS control for energy-harvesting in WBANs [24]. Chen et al. proposed a novel roboticsand cloud-assisted healthcare system to provide pervasivehealthcare services, especially for mental diseases [25]. Kulkarnet al. also proposed optimized mobile cloud service architecturefor real-time healthcare services [26].

Synthesis: Based on the review of existing literature,we infer that the existing works did not consider the problemof disease-centric electronic healthcare services to WBANs inmultiple H-CSP environments, while the utility of each WBANis optimal and the computational complexity decreases. Wepropose an architecture for offering disease-centric helathcareservices using SNA [27] in a cloud-assisted WBAN architecture.Additionally, we consider the criticality index of WBANs toprovide reliable and efficient medical services to medicallyemergent patients in real-time.

3 SYSTEM MODEL

In a cloud-assisted WBAN architecture, N = 1, 2, · · · , n aset of WBANs, B = B1, B2, · · · , Bn, are present in an area,to monitor the physiological condition of the patients, whereeach WBAN is composed of a set M = 1, 2, · · · ,M ofheterogeneous body sensors, B = b1, b2, · · · , bm, which areconnected to an LPU to aggregate the heterogeneous medicaldata from each body sensor [28], [29]. Due to the hetero-geneity of the body sensors, each such sensor has differentbandwidth requirement, B = B1,B2, · · · ,Bm, to transmit itsdata effectively. As each body sensor requires different band-width, each WBAN demands different aggregated bandwidths,BW ∈ BW1, BW2, · · · , BWn, to transmit its data from aspecific H-CSP. Therefore, to fulfill individual requirements,each WBAN Bi connects to K number of heterogeneous H-CSPs, C = C1, C2, · · · , Ck, to achieve ubiquitous healthservices. The communication between sensor nodes to LPUsand LPUs to APs follow the single-hop star topology, in orderto gather the information of WBANs. In this work, we haveconsidered the convergecast model by Zhang et al. [30] tocollect information from the WBANs. The basic elements of theproposed architecture are defined as follows:

• Φ = Φt1,Φt2, · · · ,Φtn: Set of criticality indices, whereΦti represents the ctiticality index of the ith WBAN.

• T = T1, T2, · · · , Tn: Traffic load of each WBAN in aspecific area.

• ψ = ψ1, ψ2, · · · , ψn: Throughput of each WBAN.• Pt = Pt1,Pt2, · · · ,PtK: Price charged by K number of

CSPs at time instant t.• G = G1,G2, · · · ,Gl: groups formed depending on the

relation among coexisting WBANs.• d = d1, d2, · · · , de: e number of disease types present

in a particular area.• D = D1, D2, · · · , DM: Proximity of the WBANs from

Access Points (APs).• SG = SG1 , SG2 , · · · , SGl: Size of the WBAN groups.



3

Figure 1: DPG-based Cloud-assisted WBAN Architecture

• RdG : Disease spreading rate in a WBAN group G.• T and t: Total time duration and a particular time

instant, where T = 1, 2, · · · , t.

As depicted in Figure 1, in a emergency situation, the totalvolume of WBANs in a particular area expendss tremendously,which increases the traffic load of WBANs significantly. In thepresence of increased traffic load of cloud-assisted WBANs,the performance of WBANs decrease notably in terms of map-ping cost and network throughput. Therefore, to optimize theincreased traffic load and maximize the network throughput,WBANs are initiated to form a DPG, depending on distinctsyndromes. In the formation of DPG, the WBANs with critical-ity Φ less than the threshold criticality Φth, are not included inthe same group. For different diseases, the threshold criticalityof WBANs differs. After the formation of relation grouping,each group is mapped with a particular CSP. In the traditionalWBAN architecture corresponding to a disease, a specific serveris dedicated for processing.Therefore, if any disease-specificWBAN is absent in a particular hospital, then the server re-mains unutilized. However, in a WBAN-cloud if a disease-specific WBAN is absent from a hospital, then that can be usedby other disease-based WBANs due to rapid elasticity of thecloud. Therefore, a particular relation based group of WBANs isdedicated to a particular CSP. In the presence of heterogeneousH-CSPs, a specific group-based WBAN selects an optimal H-CSP, depending on the resource available, the total criticality ofthe WBAN, and the proximity of the AP.

4 PRELIMINARIES

The proposed system uses two approaches — (a) disease-centricrelation estimation among WBANs, and (b) cloud computing,for modeling the optimization problem and solution approach,respectively. For easier understanding of the problem formu-lation, we discuss the basics of the proposed system in thisSection. Also, we present the list of symbols used in problemformulation in Table 1.

4.1 Disease-centric Relation Estimation Approach

We elaborate the preliminaries of the disease-centric relationestimation approach for cloud-assisted WBANs.

Definition 1. DPG is defined as the composition of several WBANs,depending on the distinct syndromes, d = d1, d2, · · · , dn, wheren ⊆ N .

G = nr1, nr2, · · · , nrt, (t ⊆ T ) (1)

where nrt denotes the average number of WBANs present in a DPGat time t with a relation r.

Definition 2. Disease-centric relation among WBANs is obtainedbased on similar disease types, which is calculated using the n × nencounter matrix. Due to the mobility of WBANs, a WBANBi comesin contact with another WBAN Bj . During mutual connection,WBANs transfer its CI Φ and syndromes d with each other usingbeacon messages. The encounter matrix n×n is the count of contactseach WBAN encounters. The encounter matrix is defined as,

Wi,j =

1, if Bi and Bj encountered each other0, otherwise

(2)

Definition 3. The disease similarity index (DSI) among WBAN Biand WBAN Bj is expressed as aff(i, j). Mathematically,

aff(i, j) =

1, if dBi ⊆ dBj0, otherwise

(3)

where dBi and dBj denote the syndromes of diseases for WBAN Biand Bj , respectively.

Definition 4. The relational value rij denotes the distinct connectionΥ among ith WBAN Bi and jth WBAN Bj . The edge is dependenton the type of disease of the concerned patient and the criticality Φ ofthe WBAN. Mathematically,

rij = Υ(Φi,Φj) (4)

where Φit and Φjt denotes the criticality indices of ith WBAN andjth WBAN at time t, subsequently.

Definition 5. Relation matrix R presents the relation between twoWBANs, which is expressed as:

R =

r1,1 · · · r1,j · · · r1,n

.... . .

.... . .

...ri,1 · · · ri,j · · · ri,n

.... . .

.... . .

...rm,1 · · · rm,j · · · rm,n

(5)

4.2 Cloud Infrastructure for WBANsCloud computing infrastructure supports ubiquitous and elas-tic resource provision to the real-life applications [31], [32].Therefore, to store and analyze huge data generated from the

Figure 2: Functionality of Cloud-assisted WBAN Architecture

WBANs, H-CSP provides cost-effective and real-time medicalservices to the WBANs in a critical emergency situation [33].Using cloud, it is possible to offer ubiquitous services withelastic demands. In a cloud-assisted WBAN, after sensing the



4

physiological data, the body sensors transmit the packets tothe health cloud. The health cloud service provider sends themedical data to the medical experts, so that real-time health-care services can be provision to subscribers.

5 DISEASE-CENTRIC CLUSTER FORMATION AMONGHETEROGENEOUS WBANS

We consider the mobile WBANs in cloud-assisted WBAN archi-tecture. This section theoretically analyzes the necessity of DPGformation for increased traffic load of cloud-assisted WBANsin an area. Whenever the total number of WBANs increases,then the network throughput decreases, and the traffic load ina particular area increases. We discuss the concept of disease-centric cluster formation among WBANs associated with sim-ilar disease types, based on SNA [27], to optimize the trafficload and computational complexity of cluster formation. Onthe other hand, we explain that only DPG formation is notinvariably satisfactory for providing QoS-services to WBANs.

5.1 Estimation of Traffic Load

To optimize the traffic load of cloud-assisted WBANs, weneed to estimate the effective traffic load, T tA of cloud-assistedWBANs in a critical environment in a area A.

Definition 6. Due to the coexistence of WBANs, the traffic load of aparticular WBAN Bi at area A covered by an AP, Aj , at time tκ, iscalculated as the total number of packets generated Ht =

∑ni=1 Pi ×

tκ divided by the covered area [34].

T(Pi,tκ) =

∑ni=1 Pi × tκN (d)

× 1

1− pi(6)

where Pi is the packet transmission rate of WBAN Bi at time tκ.Ht denotes the amount of accumulated packets send by these N (d)number of coexisting WBANs within the areaA, d denotes the densityof WBANs, and pi denotes the packet loss rate of WBAN Bi.

Theorem 1. Increased traffic load in a location decreases the through-put of the WBANs.

Proof. Let us consider that in a hospital environment a WBANBi transmits K number of successful packets with packet trans-mission rate Pi at time t− 1. Then, the throughput is expressedas:

ψt−1 = K∑ni=1 Pi

T − 1× 1

1− pi(7)

If the traffic load t(L,t) exceeds the threshold traffic tth, i.e.,(t(L,t) >> tth) at time t, then the number of successful packetsbecomes less, as each WBAN wants to send its data, therebyresulting in congestion in the network. Hence, due to channelcapacity constraints and potential congestion problems in thenetwork, WBAN Bi transmits h number of successful packetsat time t, which is less than the number of successful packetsat time t − 1, i.e., K < h. Therefore, the calculated throughputbecomes:

ψt = h

∑ni=1 Pi

T× 1

1− pi(8)

We can observe that the network throughput at time instant t isgreater than the network throughput at time t− 1. Mathemati-cally,

ψt < ψt−1 (9)

Therefore, increased traffic decreases the throughput of WBANsin a particular instant of time. Hence, the proof concludes.

5.2 Necessity of DPG Formation

Medical emergency situation increases the traffic capacity Tof cloud-assisted WBANs in an area A covered by an AP, Ak.Therefore, increased traffic load T in an area A significantlydegrades the performance with respect to mapping cost andservice rate. Consequently, due to the resource constrainednature of the AP, an AP is not able to fulfill the bandwidthrequirements of all WBANs always and is also not able toprovide all the services to the WBANs. In Theorem 2, we provethe necessity of DPG formation.

Theorem 2. In a hospital environment, under an AP, N(A,Bi)

WBANs get access in an area A. In a medical emergency situation,several new WBANs come into the area A, because of which thecriticality index of WBANs increases, Φit << Φit and associatedpacket criticality also increases Ωit << Ωit. Then, the total trafficincreases to T(r,Bi) from T(r,Bi), i.e., T(r,Bi) > T(r,Bi).

Proof. Suppose, in a medical emergency situation, the totalnumber of WBANs in an areaA covered by an AP,Aj , increasestremendously. The area covered by an AP, Aj , is calculated as:

A =πR2 −AR∑n

i=1 πa2i

(10)

where πR2 denotes the total area considered in the proposedmodel with radius R, AR the total area covered by an AP, and∑ni=1 πa

2i the total area covered by WBANs with radius ai.

Therefore, the total traffic load of the WBANs in the area Ais calculated as:

T(Pi,tκ) =

∑∞κ=1

∑ni=1 Pi × tκΩit

(πr2−Ar)∑ni=1 πa

2i× g

× 1

1− pi(11)

where p denotes the packet loss per second, d the density ofWBANs in the area A, and g the average number of WBANspresent in an area. The packet criticality rate Ωit is defined as[35]:

Ωit = f

(dnςBi(t)

dtn,dn−1ςBi(t)

dtn, · · · , dςBi(t)

dtn, ςBi(t)

)(12)

where ςBi(t) denotes the packet generation rate of WBAN Bi. Ina critical emergency situation, the number of WBANs increases,i.e, Ntot(A,Bi) = N(A,Bi) +Nnew(A,Bi), where Nnew(A,Bi) is thenewly arrived WBANs in the area A covered by an AP. Notonly the newly arrived WBANs increase in the area, but alsothe critcality index Φit << Φit, of WBANs and packet criticalityΩit << Ωit increases, with the change in the environment andmedical conditions∗. Therefore, the total traffic load of thenetwork increases. Mathematically,

T(Pi,tκ) =

∑∞κ=1

∑ni=1 Pi × tκΩit


2i× g

× 1

1− pi(13)

From Equations (11) and (13), we can infer that

T(r,Bi) > T(r,Bi) (14)

As the traffic load increases due to increase in the count ofWBANs and their criticality indices, the performance of thetotal network also degrades. Consequently, the critical WBANsmay face medical emergency. Therefore, it is required to de-crease the network traffic load, so that each WBAN can get theservice optimally.

∗. It depends on the patient’s age, sex, height, weight, and location.



5

The traffic load of the network increases, due to increase inthe number of WBANs and their criticality indices, as statedin Theorem 2. Therefore, the WBANs initiate to form DPGusing Algorithm 1. After the formation of DPG, we calculate theindividual load of each WBAN. Then, we formulate a networktraffic load optimization problem for cloud-assisted WBANs, inorder to minimize the increased traffic load, which is expressedas follows:

Minimize T(Pi,tκ) =

[ ∑∞κ=1

∑ni=1 PitκΦit


2i× g(1− pi)

](15)

Subject ton∑i=1

T(Pi,tκ) ≥ Tth, n ∈ N (16)

Φit ≥ Φtht , t ∈ T , n ∈ N (17)Pi ≥ Pth, i ∈ n (18)

The problem is solved using the Lagrangian Optimizationmethod. We have,

LT =1

T th(Pi,tκ)

Li( ∑∞

κ=1

∑ni=1 PitκΦit


2i× g(1− pi)

)

−X1

( n∑i=1

T(Pi,tκ) − Tth)−X2

( n∑i=1

T∑t=1

Φit − Φtht

)−X3

( n∑i=1

Pi − Pth)

where X1, X2 and X3 are the Lagrangian Multipliers con-straints. We aim to maximize T(Pi,tκ) using the approach ofLagrange Multipliers.

Algorithm 1: Algorithm for DPG FormationInput: Number of WBANs (Bi ∈ B), criticality index (Φit), disease

similarity index aff(i, j), relational value rij , waiting time τ∗i,t.Output: Optimized traffic load T ∗(Pi,tκ)

1 Receive information from the present WBANs Bi ∈ B at time t;2 Create neighbor list of each WBAN;3 Create encounter matrix among WBANs using Equation (2);4 Estimate disease similarity matrix using Equation (3);5 Find the relation value between two WBANs using Equation (4);6 for i = 1 to n do7 if Φtht ≤ Φit then8 while aff(i, j) == 1 do9 if τ∗i,t ≤ δ

lowth then

10 Compute τ∗i,t = δlowth ;11 Form Optimal DPG using Equation (1);12 Optimize the traffic load using Equation (15);

13 if τ∗i,t ≥ δhighth then

14 Compute τ∗i,t = δhighth ;15 Wait for random amount of time δt;

16 Compute τ∗i,t−1 = τ∗i,t;

17 Return T ∗(Pi,tκ);

5.3 Algorithm for DPG FormationIn this section, we discuss the proposed disease-centric clusterformation algorithm, which takes care of the total traffic of theWBANs in a particular area by forming DPG. Also, it takescare of the energy efficiency of the WBANs. The computationalcomplexity of the DPG algorithm is O(n). We compared thecomplexity of the proposed scheme with the existing scheme[36], which is relevant to our work. Moulik et al. [36] pro-posed a multi-stage Nash bargaining approach for cost-efficientmapping between WBANs and CSPs. The complexity of thisapproach is VO(n2), where V denotes the number of stages

required for the optimal cost agreement between WBANs andCSPs. Therefore, as the traffic load of WBANs increases, thenumber of stages required to get the optimal agreement alsoincreases, and inherently, the complexity also increases. Hence,the computational complexity of the DPG algorithm is compar-atively lesser than the existing scheme [36].

6 MAPPING OF DPG TO AN OPTIMAL H-CSPThough the process of DPG formation among WBANs de-creases the computational complexity and the traffic load ofthe network, it does not guarantee provisioning of QoS servicesto each WBAN. Therefore, in this section, we prove that DPGformation is not alone sufficient to provide QoS services toWBANs. To manage QoS services among WBANs, the mappingof DPG to an optimal H-CSP is needed and the selection ofcritical WBAN from a DPG is also needed. Therefore, the selec-tion of critical WBANs from a DPG is to be decided based ondifferent selection parameters. Consequently, the mapping ofDPG to an optimal H-CSP is to be decided based on the pricingmodel. We consider an architecture of one WBAN Bi and a setof H-CSPs K = 1, 2, 3, · · · k, where K is the total number ofH-CSPs. As we consider heterogeneous H-CSPs, we have eachH-CSP follow different pricing schemes and accordingly chargedifferent prices to provide the service. The price provided byeach H-CSP is denoted as P = p1, p2, · · · , pm, where pm isthe maximum price charged by H-CSP. The DPG with groupof WBANs is denoted as G = G1,G2, · · · GN, where N denotesthe total number of DPGs.

6.1 Necessity of Mapping between DPG and Optimal H-CSPDue to the mobility of WBANs, each WBAN changes its regionfrom one community to another, which significantly changesthe QoS and bandwidth requirement of each WBAN. In Theo-rem 3, we prove the necessity of optimal mapping between aDPG and H-CSP.

Theorem 3. Due to mobility and underutilization of the localmedical server, the conventional WBAN architecture is inefficient.Mapping to an optimal H-CSP provides a Cloud-assisted WBANmore effective health-care services.

Proof. At time t, the required bandwidth of a WBAN Bi isBW t

i and the available bandwidth of a local server specific to adisease is BWL. The total bandwidth requirement of N numberof WBANs within a DPG is expressed as, BW t

tot =∑Ni=1 BW

ti .

Due to heavy traffic load T and mobility of the WBANs, theavailable bandwidth requirement changes with time. Therefore,the rate of change in bandwidth requirement, ∆t, at time t ismathematically expressed as:

∆t =

∑ni=2 |BW

ti −BW t−1

i |∑Tt=2 |tt − tt−1|

(19)

After t + 1 time, the rate of change in bandwidth requirementof WBANs also increases, as in the presence of different ex-ternal effects (disease syndrome, environment, and mobility)the criticality index of WBANs increases periodically. On theother hand, if the traffic load traffic load T increases in an area,then all WBANs do not get fair amount of resources. Therefore,the criticality indices of WBANs with sensitive data increases.The increase in the criticality indices of WBANs also increasesthe packet generation rate of the body sensor nodes. In sucha scenario, the WBANs need increased bandwidth in orderto transmit their data packets successfully. Therefore, the rate



6

of change in bandwidth requirement, ∆t+1, at time t + 1 ismathematically expressed as:

∆t+1 =

∑ni=2 |BW

t+1i −BW t

i |∑Tt=2 |tt+1 − tt|

(20)

Due to the variation in the criticality indices of WBANs, the rateof change in the bandwidth requirement at time t+ 1 increasesover the rate of change in bandwidth requirement at time t,(∆t+1 >> ∆t)

†. This is because the rate of change in bandwidthrequirement varies with time and criticality index. Therefore,the available bandwidth, BWL, on a local server is not enoughto fulfill its requirements. Mathematically,

BWL < ∆t+1

N∑i=1

BW ti << ∆t+1

N∑i=1

BW t+1i (21)

As the available bandwidth does not fulfill the required band-width, there is a requirement of optimal mapping betweenWBANs and CSPs, in order to provide fair resources amongWBANs with optimal price. However, the bandwidth require-ment converges at a point, when the the bandwidth require-ment reaches the maximum bandwidth availability of a localserver BWmax

L . Mathematically,

BWL < ∆t+1

N∑i=1

BW ti << ∆t+1

N∑i=1

BW t+1i ≥ BWmax

L (22)

Hence, the proof concludes.

Remark 1. In the absence of a particular disease, di, the affectedWBANs in a hospital make the particular disease-centric serverunutilized and inefficient.

Suppose, the capacity of a local server corresponding to a particu-lar disease di in a hospital be Cdi . The utilization factor of the localserver Si is mathematically expresses as, USi =

Cdiβ

. Here, β is thearrival rate of the WBANs of a particular disease type. If the presentarrival rate β of WBANs specific to a disease type decreases, then therevised utilization factor decreases. Mathematically,

USi < USi (23)

Hence, in the absence of a particular disease di, the affected WBANsin a hospital make the particular disease-centric server unutilized andinefficient.

6.2 Selection Parameters

After the formation of different DPGs, we need to identify thecritical WBANs from each DPG to optimize the traffic capacityof cloud-assisted WBANs. The election of critical WBANs is tobe decided depending on different election parameters. Theyare discussed below.

• Criticality Index of WBAN: The criticality index of aWBAN Bi at time t is denoted by Φti [37], and ispresented as:

Φti =∣∣∣ (uc −t)2 − (t −lc)2

(| uc |+ | lc |)2

∣∣∣ (24)

where uc and lc are the upper and lower range of aphysiological specification of a patient. t = (uc+lc)

2

defines the measurement of distinct physiological data.

†. However, this conclusion is only true, if the traffic load of WBANsin an area exceeds the predefined threshold limit.

• Proximity to AP: The utility of the H-CSP UGiSj (t) de-pends on the distance between the WBAN B

Gki in the

group Gk.

D(Bi,AP ) =

(1/√

(Bxi −AP xj )2 + (Byi −APyj )2

)(25)

where (Bxi , Byi ) and (AP xj , AP

yj ) are the coordinates of

ith WBAN and jth AP, respectively.• Size of DPG: Size of the jth relational group is presented

as StGi at time t. Mathematically,

N∑j=1

StGj =

N∑j=1

NGj , 1 ≤ j < N (26)

where NGj denotes the number of WBANs present in agroup Gj , based on the disease based relation r.

• Epidemic Spread Factor (ESF): Epidemic spread factordenotes the change or the deviation in the criticalityindex Φt of the WBANs in a particular group Gj , andis denoted by Υ.

Υ = Φti,Gj − Φt−1i,Gj (27)

where Φti,Gj and Φt−1i,Gj are criticalities of the WBAN refer

to a particular group at a time t and t− 1 respectively.• Residual Energy Level: The residual energy level of

WBANs is presented as ΨtBi

, and is mathematicallyexpressed as, Ψt

i =Eprei

Einii. Here, Eprei and Einii denote

the required and preliminary energy levels of the ith

WBAN Bi, subsequently.• Available Bandwidth of the H-CSP: Available band-

width of the H-CSP is calculated as,

BWava = BWtot −BWuse (28)

where BWtot and BWuse denote the total and usedbandwidths of the H-CSP, respectively.

Definition 7. The utility factor SGji of a WBAN, Bi, affinity to agroup, Gj , to choose an H-CSP among heterogeneous H-CSPs at timet, is expressed as,

SGji =µ

(Eprei

Einii

+StGj|Φti,Gj − Φt−1

i,Gj |Φt

+D(Bi,AP )min

D(Bi,AP )max+BWava

BWtot

)(29)

where µ denotes utility constant, Φt denotes criticality indexand ψ denotes epidemic spreading factor. D(Bi, AP )max andD(Bi, AP )min denote maximum and minimum the distance betweena WBAN and an AP, respectively.

The utility constant, µ, of the utility factor is expressed as,

µ =

1, if EiniBi

≤ EpreBi

0, otherwise(30)

The maximization of the selection rate SGji is defined as follows:

ϑ = Max SGji , ∀i ∈ n,G ∈ G (31)

Theorem 4. The maximum and minimum values of the selectionparameter is expressed as:

fmax(SGji )

=

(2 +

0.5YF

)(32)

fmin(SGji )

=

(1 + 0.5Y +

1

β

)(33)



7

Proof. The selection function, fi, is a linear function of vari-ables RE, D, and BW , and Θ is a constant for choosing thecritical WBANs. The function is derived by subtracting the costfunction from the selection rate function shown in Equation(29). The selection rate in Equation (29) gives the maximumvalue. The selection rate is maximum when Eprei = Einii andD(Bi,AP )min = D(Bi,AP )max . Therefore the maximum selectionrate is expressed as:

fmax(SGji )

=

(2 +

0.5YF

)(34)

where Y is the maximum size of the DPG and 0.5F is the

maximum criticality index, F < 1. Also,

EpreBi

EiniBi

+D(Bi,AP )min

D(Bi,AP )max

= 2 (35)

The selection rate is minimum when 1β

<EpreBi

EiniBi

and the

minimum size of the DPG is Y. The minimum selection rateis expressed as:

fmin(SGji )

=

(1 + 0.5Y +

1

β

)therefore, we get the maximum and minimum values of theselection parameter. Hence the proof concludes.

Lemma 1. Every critical WBAN has its own preference of choosingoptimal H-CSP. Therefore, the preference Pi of the WBAN Bi ofchoosing H-CSP Ci among heterogeneous H-CSPs are the symmetricand asymmetric components of the relation.

Proof. Suppose, the WBAN BGhi belongs to group Gh with

relation rh having a preference of choosing H-CSP Sm andanother WBAN B

Ghj belongs to group GkS with relation rk

having a preference of choosing H-CSP Sn. As the WBANsbelong to the different groups Gh and Gk, each WBAN fromthe different groups has the different preferences Pp and Poof choosing H-CSP. Then, the WBANs follow asymmetricity,which is expressed as,

BGhi rhSmPp 6⇒ B

Gkj rkSnPo (36)

For the asymmetricity statement, it is assumed that the relationsamong WBANs are different. It is also assumed that theybelong to different groups. Therefore, the WBANs do not knowone another’s preferences of choosing optimal H-CSP. For themapping process, WBANs take a cumulative decision in orderto get an optimal solution between WBANs and CSPs, basedon one another’s individual preferences. As the WBANs belongto different groups, they cannot take any cumulative decisionto choose an optimal decision. Consequently, we designed ourscheme in such a way that the WBANs belonging to differ-ent groups and relations must follow an asymmetric relation.Suppose, the WBAN B

Ghi and another WBAN B

Ghj belong to

the same group Gh with relation rk having a preference Pp ofchoosing H-CSP Sm. As the WBAN Bi and Bj belong to thesame group Gh, all the WBANs in the group have the samepreference of choosing Pp the H-CSP. Then, the WBAN followssymmetricity, which is expressed as,

BGhi PpSmrk ⇒ B

Ghj PpSmrk (37)

Similarly, for the symmetricity statement, the relation amongWBANs are the same, and also they belong to the same group.Therefore, the WBANs know the preferences of one anotherfor choosing the optimal H-CSP. For the mapping process,

WBANs take a cumulative decision in order to get an optimalbetween the WBANs and the CSPs, based on the individualpreferences. As the WBANs belong to the same group, they cantake the cumulative decision to choose an optimal CSP. Withthis objective, we designed our scheme in a manner that theWBANs belonging to the same group and relation must followa symmetric relation. Hence, the proof concludes.

6.3 Formulation of Utility Function

The WBAN BGji belonging to group Gj can be model as a

customer and expecting to obtain more benefits with leastpossible payments. The utility function of WBAN is expressedas,

UBGji

= RBi − Ptot (38)

where R(.) defines the revenue function and P (.) denotesthe price charged by the H-CSP or total charge payed by theWBAN. The revenue function R(.) defines the selection rate ofchoosing a H-CSP SGji . It is mathematically expressed as,

RBi = η ×N∑j=1

SGji (39)

where η denotes the gain per unit selection rate. P (.) definesthe total charge payed by the WBAN for using cloud services.It is mathematically expressed as follows:

Ptot =

n∑i=1

Pti ×RCk + PQoS (40)

where Pti represents the price per unit of resource selling fromH-CSP Ck to WBAN, RCk represents the amount of resourcebought by WBAN from H-CSP, and PQoS represents the fixedprice charged by all the H-CSPs to give QoS to the WBAN.

We segregated the QoS satisfying pricing into two cate-gories, as follows:

a) Price charged for interference management among coex-isting WBANs in an area,A, is denoted as, PI . Therefore,the interference management factor is expressed as:

αk,l =

K∑k=1

β(I)ak,l (41)

where β(I) denotes the interference management factorand ak,l denotes the price charged by the H-CSP to man-age the interference among the coexisting WBANs. Theprice charged per WBAN for interference management,PI , is expressed as:

PI =

n∑i=1

(∑Ml=1 αk,l

Mi

)(42)

where Mi denotes the average number of coexistingWBANs and M denotes the total number of coexist-ing WBANs in an area. Hence, the total price chargedper WBAN by H-CSP for interference management isexpressed as:

PI =

n∑i=1

1

Mi

M∑l=1

K∑k=1

β(I)ak,l

(43)

b) Price charged for queue and mobility management, isdenoted as PQ. Therefore, the queue and mobility man-



8

agement factor is expressed as:

Λk,l = Λ

n∑i=1

$i(I) +Qi (44)

where Λ denotes the price charged by H-CSP to managethe problem related to queue overflow and transientconnection due to mobility. $i(I) and Qi denote themobility and queue management factor for WBAN Bi.The queue and mobility management price chargedfrom WBAN is expressed as:

PQ =

n∑i=1

K∑k=1

Vmax(

∑Ml=1 Λk,l)

V t1i + V t2i + · · ·+ V tti

(45)

where Vmax denotes the maximum velocity of a WBANand V 1

i denotes the velocity of WBAN Bi at differenttime instant.

Therefore, the total price charged by H-CSP for QoS manage-ment of WBANs, PQoS , is expressed as:

PQoS = PI + PQ (46)

PQoS =

n∑i=1

[1

Mi

M∑l=1

K∑k=1

β(I)ak,l +Vmax

∑Kk=1(

∑Ml=1 Λk,l)

V t1i + V t2i + · · ·+ V tti

](47)

The optimization problem for optimal mapping betweenWABNs and CSPs is mathematically expressed as:

Maximize∑

i∈n,j∈NUBGji

=[η ×

N∑j=1

SGji − Ptot]

(48)

Subject ton∑i=1

N∑i=1

SGji ≥n∑i=1

Sthi , n ∈ N (49)

Φit ≥ Φtht , t ∈ T , n ∈ N (50)Ptot ≥ Pth) (51)

Therefore, the characteristic of the utility function is expressedas follows:

i) The first order derivative of the utility function is anon-decreasing function, as each WBAN attempts tomap with the optimal H-CSP in a cost effective way.Mathematically, we have:

∂UBGji

∂t≥ 0 (52)

ii) The second order derivative of the utility functionis decreasing function, so that each WBAN gets themaximum utility value. Mathematically we have:

∂2UBGji

∂t2≤ 0 (53)

As the WBAN BGi in the group Gj tries to maximize the utilityUBGji

by buying optimal amount of resources, the utility UBGji

varies with the resource for a WBAN BGi .(∂UBGji

∂t= η × ∂SGji

∂t−

n∑i=1

Pti ×RCk − PQoS

)(54)

If Pti ≤∂SGji∂t

, then from Equation (54), we can conclude,

∂UBGji

∂t> 0 (55)

Theorem 5. Though the size of the relational group does matter, butthe provision to the H-CSP mainly depends on the disease spreadingfactor of a particular group.

Proof. Suppose, in the proposed architecture, the size of the H-CSP is classified into Small-cloud (SGs ), Medium-cloud (SGm )and Large-cloud (SGl ), depending upon their computationalpower (Pcom) and the resource poll (i.e., bandwidth BW , andthe access time Tacc).

Let, the disease spread factor of the ith group be ψGi . There-fore, the disease spread factor of different groups is denotedas Υ = ΥG1 ,ΥG2 ,ΥG3 , · · · ,ΥGl. Based on this, each group isassigned a different cloud. Let ΥG1 ,ΥG2 and ΥG3 be the diseasespread factor of the groups G1, G2 and G3. At a specific time t,the disease spread factor of G1 is higher than the same for G2,and the disease spread factor of the G2 is higher than the samefor G3. This is mathematically expressed as,

ΥG1 > ΥG2 > ΥG3 (56)

Similarly, at time t, the size of the group G1 is higher than thesame for G2, and the size of the group of the G2 is higher thanthe same for G3. Mathematically,

G1 > G2 > G3 (57)

However, we use the incentive compatibility mechanism [38]in order to prove the Theorem. Let, a WBAN with the diseasespreading factor, ΥGi , of a Gi be mapped to the small-cloudSGs . However, if the disease spreading factor, ΥGi of that groupis greater than the threshold disease spreading factor, Υth, i.e.,ΥGi >> Υth, then the WBANs with the increased spreadingfactor in that particular group do not get fair resources fromSGs . In such a condition, the incentive compatibility constraintensures that, instead of mapping the WABNs to SGs , theWBANs are mapped to the medium-cloud SGm for efficientutilization of the resources. Mathematically,

ΥGi >> Υth, SGm > SGs (58)

Similarly, a WBAN with disease spreading factor, ΥGi , of a Giis mapped to the medium-cloud SGm . However, if ΥGi of thatgroup is greater than the threshold disease spreading factor,Υth, i.e., ΥGi >> Υth, then the WBANs with the increaseddisease spreading factor in that particular group do not get fairresources from SGm . In such condition, the incentive compati-bility constraint ensures that instead of mapping the WABNs toSGm , the WBANs are mapped to the large-cloud SGl for efficientutilization of the resources. Mathematically,

ΥGi >> Υth, SGl > SGm (59)

In such a scenario, at the same time t, G1 is assigned to theLarge-cloud (SGl ). As the size of the group is large and also thedisease spreading factor increases with time, therefore G1 cannot mapped with SGs and SGm . It is mathematically expressedas:

(G1 → SGl) at access time (Tacc) (60)

Also, G2 is assigned to the Medium-cloud (SGl ). Mathemati-cally, it is expressed as,

(G2 → SGm) at access time (Tacc) (61)

Also, G3 is assigned to the Small-cloud (SGl ). Mathematically, itis expressed as,

(G3 → SGs) at access time (Tacc) (62)

Therefore, based on the disease spread factors different H-CSPs



9

are provided using the incentive compatibility mechanism.Hence, the proof concludes.

The total number of requested WBANs belonging from agroup G is denoted by R. We get,

R =

n∑i=1

ZGi (63)

where ZGi represents the request of the ith WBAN in a partic-ular group G. Therefore, within a particular group G total Rnumber of requests are sent by WBANs belonging to the samegroup. If the WBAN requests are satisfied by the H-CSPs attime t, then the satisfaction factor for each WBAN is denotedby ΠBi . It is mathematically expressed as,

ΠBi =

[(∑ni=1

∑Nj=1 t− t

)× U

BGji∑n

i=1 ZGi

](64)

Using the Lagrange multiplier, we express the optimizationproblem in Equation (51) as,

5i∈nUBGji

= −λ5i∈n UBGji

(65)

[δUBGji

δt1,δU

BGhi

δt2, · · · ,

δUBGji

δtt

]λt1λt2

...λtt

= 0 (66)

where, 5i∈nUBGji

=(∂δU

BGji

∂t,∂δU

BGji

∂t

)(67)

6.4 Algorithm for Mapping to an Optimal H-CSPIn this section, we propose an algorithm to choose an optimalH-CSP among heterogeneous H-CSPs in a medical emergencysituation. From Algorithm 1, we get the WBANs with thesimilar disease types in a DPG to choose the critical ones withinthe cluster easily and achieve lower computational complexity.Algorithm 2 describes the mapping of DPG to an optimal H-CSP, to provide fair amount of resources to the critical WBANsin a medical emergency situation.

7 COMPLEXITY ANALYSIS

In Section, we analyze the performance of the proposed schemetheoretically. We discuss the complexity of the proposed schemefor cloud-assisted WBANs.

Theorem 6. The worst case asymptotic time complexity of the systemis O(zn2), where n is the number of critical WBANs.

Proof. In the first algorithm, each WBAN related to the samedisease forms a cluster to minimize the computational complex-ity. To form the DPG, the worst case computation complexityis O(qn) from Algorithm 1. After the formation of DPG, thecritical WBANs choose from the DPG. To choose the criticalWBANs from each WBAN, a selection procedure using theselection parameter Si. Therefore, the worst case complexityof selecting critical WBANs is O(pn).

T(X) = C1qT(Xs) + pT(Xk)+ C2T(1) (68)

The combined worst case complexity of DPG formation andselection of critical WBANs is O(zn2), where z = q + p.Hence, we infer that the total computational complexity of theproposed approach, in the worst case, is O(zn2), where n is thecount of critical WBANs. This completes the proof.

Algorithm 2: Algorithm for Mapping to Optimal H-CSPInput: Number of WBANs within a DPG (Bi ∈ B), Disease types

d = d1, d2, · · · , dk, Ctiticality index of each WBAN Φit,waiting time τ∗i,t.

Output: Optimal Mapping Matrix (U∗BGji

)

1 Calculate the criticality index of all WBANs Bi ∈ B at time t;2 Measure the Euclidean distance between WBAN and the gateway ;3 Compute the group size of the DPG, StGi ;4 Compute the disease spread factor within the DPG;5 Calculate the residual energy level of the each WBAN;6 Compute the available bandwidth for the AP;7 Compute the selection parameter Si, ∀Bi ∈ B;8 if SGiSj ≥ S

thSj

then9 The detected WBANs are in critical medical condition in a DPG;

10 Update waiting time τ∗i,t = δlowth ;11 if U

BGhi

≥ UthBGhi

then

12 Optimal Mapping matrix U∗BGji

;

13 QoS is provided by the H-CSP with extra price PQoS ;

14 if UBGhi

≤ UthBGhi

then

15 Standard mapping matrix generated;16 QoS is not provided by the H-CSP;

17 if SGji ≤ Sthi then

18 The detected WBANs are in normal medical condition in a DPG;19 Update waiting time τ∗i,t = δhighth ;

20 Update τ∗i,t−1 = τ∗i,t;21 Return U∗

BGji

;

Table 1: Testbed Information

Parameters ValuesProcessor Used Intel(R) i5-2500 CPU @ 3.30 GHzRAM 4GB, DDR3Disk Space 320 GBOperating System Ubuntu 14.04 LTSApplication Software MATLAB R2013a

8 PERFORMANCE EVALUATION

We discuss the results of the proposed schemes, based on dif-ferent metrics. We have used a modeled and created simulationplatform using MATLAB. We have given the system informa-tion in Table 1 and also enlisted the simulation parameters inTable 2.

Table 2: Simulation Parameters

Parameter ValueSimulation area 1 Km× 1 KmNumber of WBANs 300Number of body sensors in a WBAN 8Number of APs 10Velocity of each WBAN 1.5 m/sInitial energy WBAN 0.5 JPower consumption of Tx-circuit 16.7 nJPower consumption of Rx-circuit 36.1 nJPower consumption of Amplifier-circuit 1.97 nJEnergy of WBANs 0.5 JSINR threshold 5-15 dBSensing range of body sensors 0.5− 1.5 mPacket rate 4 packets/secPacket size 512 Bytes

8.1 BenchmarksCHMS, the benchmark considered in this study, is a cloud-based medical system, proposed by Almashaqbeh et al., [39].The authors proposed a data aggregation and classificationsystem to reduce the traffic load of the network. They alsoproposed a dynamic channel assignment approach to reducethe interference among coexisting WBANs. As the benchmark



10

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 10 20 30 40 50 60 70 80 90 100

Ene

rgy

(J)

Number of WBANs

O-MAP Normal WoP

(a) 100 WBANs, 5 H-CSPs

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 25 50 75 100 125 150

Ene

rgy

(J)

Number of WBANs

O-MAP Normal WoP

(b) 150 WBANs, 10 H-CSPs

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 50 100 150 200 250

Ene

rgy

(J)

Number of WBANs

O-MAP Normal WoP

(c) 200 WBANs, 15 H-CSPs

Figure 3: Fixed Number of WBANs with Varying Number of H-CSPs

0

20

40

60

80

100

1 2 3 4 5

Perc

enta

ge o

f C

onne

cted

W

BA

Ns

H-CSP Id Number

O-MAPNormal

WoP

(a) Number of H-CSP, 5

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10

Perc

enat

ge o

f C

onne

cted

W

BA

Ns

H-CSP Id Number

O-MAP Normal

WoP

(b) Number of H-CSP, 10

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Perc

enat

ge o

f C

onne

cted

W

BA

Ns

H-CSP Id Number

O-MAP Normal

WoP

(c) Number of H-CSP, 15

Figure 4: Fixed Number of WBANs with Varying Number of H-CSPs

algorithm proposed a traffic flow based dynamic channel as-signment approach for cloud-assisted WBANs, therefore wehave considered this approach to compare it with our existingscheme.

8.2 Simulation Settings

To analyze the performance of the two schemes — EfficientHealthcare Management (HCM) and Optimal Mapping (O-MAP), we considered the Group-based mobility of WBANs[40]. Also, we considered the single-hop star topology fordata transmission in WBANs, where each WBAN consists of8 sensor nodes placed on the body, according to [41]–[43]. TheWBANs are distributed randomly in an area of 1 Km× 1 Km.To consider the varying traffic of WBANs in an area, we varythe WBAN density factor g from 5 to 20 per square meter.Also, we vary the traffic load from 500 Kbps to 950 Kbpsto observe the mapping cost between WBANs and H-CSPs.Subsequently, we considered the different data rates of bodysenor nodes, i.e., 100 Kbps to 700 Kbps, to compute the optimalmapping cost of WBANs. For the validation of our proposedapproach, we use the IEEE 802.15.4 standard CSMA/CA accessmechanism [44]. The price function for mapping is defined asf(Ptot) = PtQ(ytbi)

2, where ytbi defines the data transmissionrate of the sensor nodes.

8.3 Performance Metrics

We illustrate the different performance metrics used for perfor-mance evaluation.

• Network Throughput: Network throughput is defined asthe number of packets successfully received per second.

• Traffic Load: Traffic capacity of the network defines thetotal number of packets transmitted from a DPG persecond.

• Residual Energy: Residual energy of a WBAN definesthe initial energy of the WBAN to perform differentoperations.

• Packet Delivery Delay: Packet delivery delay defines thetime duration between the data transmitted from theWBAN to the data received at the AP.

• Packet Loss: Packet loss defines the difference betweenthe number of data packets transmitted from the WBANto the total amount of data packets received at the APper unit time.

σ =Pttran − Ptrec

t(69)

where Pttran and Ptrec denote the amount of data packetstransmitted and the number of packets received at timet, respectively.

• Service Rate: Service rate denotes the ratio of the numberof patients served to the number of patients present inthe area.

S =N tser

N ttot

(70)

where N tser and N t

tot present the amount of patientsserved and the total amount of patients at time t, re-spectively.

• Average Energy Consumption: Energy consumption ofeach WBAN, E, is calculated as [45]:

Ei = E0 −K(Eelec + εamp)D2 (71)

where E0 denotes the initial energy, Eelec the energyconsumption due to electric circuit and εamp the energyconsumption due to amplifier circuit. K denotes thepacket size and D the distance between the sensor nodeand the LPU.

• Path Communication Loss: Path communication loss isdefined as the power loss due to transmission from



11

0

5000

10000

15000

20000

0 10 20 30 40 50 60 70 80 90 100

Net

wor

k T

hrou

ghpu

t (M

bps)

Number of WBANs

CHMS HCM

(a) Network Throughput

0

20

40

60

80

100

50 100 150 200 250 300

Serv

ice

Rat

e (%

)

Number of WBANs

CHMS HCM

(b) Service Rate

0

0.1

0.2

0.3

0.4

0.5

0.6

0 1 2 3 4 5 6 7 8

Res

idua

l Ene

rgy

(J)

Sensor IDs

CHMS HCM

(c) Residual Energy

Figure 5: Analysis of throughput, service rate, and residual energy

0

20000

40000

60000

80000

100000

120000

0 10 20 30 40 50 60 70 80 90 100

Pack

et D

rops

(K

bps)

Number of WBANs

CHMS HCM

(a) Number of Packets Drop

30

40

50

60

70

1 2 3 4 5 6 7 8

Path

Los

s (d

B)

Number of Sensor Nodes

CHMS HCM

(b) Path Loss

0 20000 40000 60000 80000

100000

0 50 100 150 200 250 300Map

ping

Cos

t (un

its)

Number of WBANs

Traffic Load = 950 KbpsTraffic Load = 750 KbpsTraffic Load = 500 Kbps

(c) Mapping Cost

Figure 6: Analysis of packet drop, path loss, and mapping cost

sensor node to the reception in the LPU [45].

PL(dB) = 10Ξ log10(D0

D )+10 log10

(4πD0)2

λ+Xσ (72)

where Ξ is the path loss exponent varying between2− 3.5, D denotes the distance among body sensor andLPU, D0 is the actual distance among body sensor andLPU, σ is the Gaussian variance, and X is the Gaussianrandom variable. The operating frequency is 2.4 GHZ.

8.4 Discussion on ResultsWe considered the proposed algorithm — HCM to analyze theperformance, based on different simulation metrics.

8.4.1 Energy ConsumptionFigure 3 depicts the energy consumption of WBANs for theproposed scheme. From the figure, we can see that with the in-crease in the number of HCSPs, the energy consumption of thecritical WBANs decreases. In our proposed scheme, the criticalWBANs form a DPG and are efficiently mapped to an optimalH-CSP. Therefore, the critical WBANs consume reduced energythan the normal WBANs, while transmitting the packets. Withthe increase in the number of H-CSPs, the energy consumptionof each WBAN decreases. We also compare our existing workO-MAP with the Without Optimal Mapping Scheme (WoP) andexisting scheme CHMS (considered to be normal), in whichHCM outperforms the WoP and normal scheme (CHMS).

8.4.2 Total Number of Connected WBANsFigure 4 depicts the total number of connected WBANs tooptimal H-CSP in a critical emergency situation. We analyzedthe proposed scheme with varying number of WBANs 100,150, and 200 and varying number of H-CSPs 5, 10, and 15.Figures 4(a), 4(b), and 4(c) depict that the total number ofserved WBANs is more than the existing scheme based on

the available bandwidths of the H-CSPs. On the other hand,from the figures, we can observe that with the increase in thenumber of H-CSPs, the total number of served WBANs froman particular H-CSP increases. Therefore, the critical WBANcan get improved services in the presence of multiple H-CSPs.We also compared our scheme with the exiting solution, andobserve that our scheme outperforms the existing one.

8.4.3 Communication Path LossFigure 6(b) depicts the cumulative path loss of sensor nodeswith two schemes — CHMS and HCM. We measured thepath loss using Equation (72), where the path exponent variesfrom 3 − 4.5, and the operating frequency 2.4 GHz. In thepresence of group-based mobility, the distance between thesensor nodes and LPUs increases over time, which increasesthe path loss of body sensor. Therefore, to minimize the powerconsumption, it is very important to minimize the path loss.As the proposed scheme manages the mobility of WBANs bycharging an extra amount for mobility management, thereforeit minimizes the path loss. Hence, we observe that the HCMsignificantly minimizes the path loss of sensor nodes over theexisting schemes.

8.4.4 Packet LossFigure 6(a) shows the mean packet loss of WBANs for differentschemes — CHMS and HCM. Due to group-based mobilityof WBANs, the quality of channel between sensor nodes andLPUs decreases significantly, which increases the packet lossthe exciting system. The proposed scheme provides an optimaand efficient between WBANs and H-CSPs, which inherentlydecreases the packet loss rate of the system and provides fairresources to sensor nodes. Therefore, the proposed scheme,HCM, reduces the packet loss, while maintaining the QoSrequirements of WBANs. The scheme used in HCM, able toprovide fair performance than the existing schemes CHMS, asthey do not consider QoS requirements of the WBANs.



12

0

1

2

3

4

5

1 5 10 15 20 25 26 30 35 40 45 46 50

Del

ay (

s)

Number of WBANs

CHMS HCM

(a) Average Delay

0

20

40

60

80

0 10 20 30 40 50 60 70 80 90 100

Tra

ffic

Loa

d (

pack

ets/

sec)

Number of WBANs

g=(05,10)g=(10,15)

g=(15,20)

(b) Traffic Load with varying in density

0

10

20

30

40

50

60

50 100 150 200 250 300

Tra

ffic

Loa

d (p

acke

ts/s

ec)

Number of WBANs

CHMS HCM

(c) Traffic Load

Figure 7: Analysis of average delay, traffic load with variation in the density and traffic load

0

2000

4000

6000

8000

100 200 300 400 500 600 700

Map

ping

Cos

t (un

its)

Data Rate (Kbps)

WoPNormal

O-MAP

(a) Number of WBANs: 100

0 2000 4000 6000 8000

10000 12000

100 200 300 400 500 600 700

Map

ping

Cos

t (un

its)

Data Rate (Kbps)

WoPNormal

O-MAP

(b) Number of WBANs: 200

0 2500 5000 7500

10000 12500 15000

100 200 300 400 500 600 700

Map

ping

Cos

t (un

its)

Data Rate (Kbps)

WoPNormal

O-MAP

(c) Number of WBANs: 300

Figure 8: Mapping cost with with variation in number of WBANs

8.4.5 Network Throughput

Figure 5(a) presents the mean throughput of WBANs for HCMand CHMS. Throughput is the reception of the packets success-fully at the AP. As previously mentioned, throughput dependson the number of successful reception of packets at the sink.Therefore, due to the increase of alive nodes in the network,the packet reception rate increases. As a result, the networkthroughput increases due to the increase of alive nodes. From5(a), we can observe that using HCM, the WBANs from aDPG are efficiently mapped to the H-CSP, which maximizesthe packet transmission rate. Consequently, as WBANs carrysensitive and important medical data packets, therefore it isnecessary to increase the throughput of WBANs in efficientway, where our proposed approach provides significant im-provement in throughput for critical WBANs. Therefore, theproposed scheme HCM has less packet drop rate compared toCHMS.

8.4.6 Service Rate

Figure 5(c) shows the service rate of the H-CSP for HCM andCHMS. We have calculated the service rate using Equation70. In our proposed scheme, the WBANs are mapped to H-CSP efficiently while considering the traffic load. Therefore, theservice rate of the proposed scheme is 5 − 10% more than thatof the existing scheme. Hence, we observe that the proposedscheme, HCM, outperforms the existing scheme CHMS, as thelatter does not consider the efficient mapping of WBANs toH-CSP.

8.4.7 Residual Energy

Figure 5(c) depicts the energy levels of WBANs for HCM andCHMS. Each WBAN struggles to transmit its data packets. Nev-ertheless, due to mobility of the WBANs, the packet transmis-sion rate decreases and the WBAN consumes more energy. As a

result, the residual energy of the WBAN decreases. In our pro-posed scheme, due to efficient mapping of critical WBANs to H-CSP, each WBAN connects to optimal H-CSP, while preservingthe QoS requirement and providing fair amount of resources.Therefore, for the efficient mapping and data transmission, theHCM scheme consumes less energy. Therefore, the proposedscheme, HCM, outruns the existing scheme CHMS, as the latterdoes not consider the efficient mapping of WBANs to H-CSP.

8.4.8 Packet Delivery DelayFigure 7(a) describes the packet delivery delay from WBAN toAP. In this figure, we observe that using the proposed scheme,the critical WBANs have less packet delivery delay than thenormal WBANs. As, in our scheme, the critical WBANs areefficiently mapped to the H-CSP, therefore the WBANs can sendtheir data immediately. Therefore, efficient mapping incurs lesspacket delivery delay using HCM. However, we observe thatthe HCM scheme, outruns the scheme CHMS, as the latter doesnot consider the efficient mapping of WBANs to H-CSP.

8.4.9 Traffic LoadFig. 7(b) depicts the traffic capacity of cloud-assisted WBANs.From Fig. 7(c), we can observe that with the increase in thedensity g of WBANs, the traffic volume increases, which sig-nificantly decreases the network performance. To cope with thesituation, our proposed algorithm deals with the increase intraffic load and provides QoS to each WBAN. From Fig. 7(b), wecan see that our proposed scheme HCM decreases the increasedtraffic load in a medical emergency situation.

8.4.10 ReliabilityTo calculate the reliability of the proposed scheme, we usedthe elastic reliability optimization technique proposed by Zhaoet al. [46]. Figure 9(a) presents the reliability of the proposedscheme with the variation in the number of WBANs. From the



13

0

20

40

60

80

100

50 100 150 200 250 300

Rel

iabi

lity

(%)

Number of WBANs

Normal WoP

O-MAP

(a) Reliability

0

20

40

60

80

100

50 100 150 200 250 300

QoS

(%

)

Number of WBANs

WoPNormal

O-MAP

(b) QoS

0

20

40

60

80

100

10 20 30 40 50 60 70 80

Rel

iabi

lity

(%)

Extra Price (units)

NormalWoP

O-MAP

(c) Reliability

Figure 9: Analysis of reliability and QoS

figure, we observe that the reliability of the system increaseswith the increase in the number of WBANs. Our proposedscheme uses the optimal mapping technique between WBANsand CSPs, while taking into consideration of interference man-agement price for the minimization of interference amongcoexisting WBANs. Therefore, if any WBAN fails to functionin the presence of other coexisting WBANs and/or other radiotechnologies, the H-CSPs charge an extra price to minimize theeffects of the interference. As a result, our proposed approachprovides higher reliability than the existing schemes. Figure 9(c)shows the reliability with the variation in extra price in thenetwork. Here, we vary the extra price within the range 10−80units. From the figure, we observe that the reliability of WBANsincreases with the increase in the extra price. The mutual andcross technology interference among WBANs is accounted forby the extra charge. Therefore, as the extra price increases,the interference among WBANs decreases, and the reliabilityincreases. Consequently, the proposed scheme achieves higherreliability than the existing approaches.

8.4.11 QoS

Figure 9(b) presents the QoS of the proposed scheme withvarying number of WBANs in a medical emergency situation.From the figure, we observe that QoS increases using theproposed approach – O-MAP. As the O-MAP scheme proposesefficient mapping between WBANs and H-CSPs, the servicedelay of the network decreases. Consequently, the successfulpacket transmission rate increases. As a result, the QoS ofthe network also increases. We also compared our proposedscheme in the normal and WoP scenario. We observe that theproposed scheme outperforms the existing others.

8.4.12 Mapping Cost

Figure 6(c) shows the mapping cost of WBANs for differenttraffic load in a medical emergency situation. From the figure,we observe that with the increase of traffic load the mappingcost of WBANs increases significantly. On the other hand, thetraffic load of 950 Kbps incurs more mapping cost than thetraffic load of 750 Kbps. To cope with the increased mappingcost for optimal mapping between WBANs and H-CSPs, weprovide an optimal mapping algorithm — O-MAP. Figure 8depicts the mapping cost between WBANs and H-CSPs fordifferent data rates of body sensor nodes. From this figure, weobserve that the mapping cost using the proposed approach —O-MAP — is lesser than that without optimal mapping — WoP— and the conventional approach (CHMS).

9 CONCLUSION

Increased traffic load in an area decreases the performance ofthe network with respect to the mapping cost and the meannetwork throughput. Therefore, to minimize the traffic load,we proposed a disease-centric relation estimation approach tooptimize the computational complexity of cluster formation.After estimation of the relation among WBANs, we proposedan algorithm to form DPG, considering several disease typessuch as epidemic cholera, glandular fever, and epidemic paroti-tis. DPG alone is not sufficient to provide QoS services tothe WBANs. Therefore, we also proposed another algorithmfor efficient mapping of WBANs belonging to a DPG withan optimal H-CSP. We performed the complexity analysis andstability of the proposed algorithm. We also compared ourproposed schemes with the existing schemes, from which weare able to show that our approach outperforms the existingapproaches we considered as the benchmarks and achievessignificant results.

In the future, we plan to analyze the dynamic behavior ofWBANs in a critical emergency situation and also propose a dy-namic cache optimization scheme for handling such emergencysituation. Also, we plan to design an optimal and effectiveresource augmentation process in WBANs for edge computingplatform [47].

REFERENCES

[1] T. Hayajneh, G. Almashaqbeh, S. Ullah, and A. V. Vasilakos, “ASurvey of Wireless Technologies Coexistence in WBAN: Analysisand Open Research Issues,” Wireless Networks, vol. 20, no. 8, pp.2165–2199, 2014.

[2] G. Fortino, G. Di Fatta, M. Pathan, and A. Vasilakos, “Cloud-Assisted Body Area Networks: State-of-the-Art and Future Chal-lenges,” Wireless Networks, vol. 20, no. 7, pp. 1925–1938, 2014.

[3] M. Chen, S. Gonzalez, A. Vasilakos, H. Cao, and V. C. Leung,“Body Area Networks: A Survey,” Mobile Networks and Applica-tions, vol. 16, no. 2, pp. 171–193, 2011.

[4] A. Argyriou, A. C. Breva, and M. Aoun, “Optimizing Data For-warding from Body Area Networks in the Presence of Body Shad-owing with Dual Wireless Technology Nodes,” IEEE Transactionson Mobile Computing, vol. 14, no. 3, pp. 632–645, 2015.

[5] G. E. Santagati and T. Melodia, “Experimental Evaluation ofImpulsive Ultrasonic Intra-Body Communications for ImplantableBiomedical Devices,” IEEE Transactions on Mobile Computing,vol. PP, no. 99, pp. 1–1, 2016.

[6] P. Rakovic and B. Lutovac, “A Cloud Computing Architecturewith Wireless Body Area Network for Professional Athletes HealthMonitoring in Sports Organizations – Case Study of Montenegro,”in Proceedings of Mediterranean Conference on Embedded Computing,2015, pp. 387–390.

[7] A. Abbas and S. U. Khan, “A Review on the State-of-the-ArtPrivacy-Preserving Approaches in the e-health Clouds,” IEEEJournal of Biomedical and Health Informatics, vol. 18, no. 4, pp. 1431–1441, 2014.



14

[8] B. E. Reddy, T. S. Kumar, and G. Ramu, “An Efficient CloudFramework for Health Care Monitoring System,” in Proceedingsof International Symposium on Cloud and Services Computing, 2012,pp. 113–117.

[9] G. Ferna’ndez, I. de la Torre-Diez, and J. J. Rodrigues, “Analysisof the Cloud Computing Paradigm on Mobile Health RecordsSystems,” in Proceedings of International Conference on InnovativeMobile and Internet Services in Ubiquitous Computing, 2012, pp. 927–932.

[10] S. Ullah, K. Pervez, N. Ullah, S. Saleem, H. Higgins, K. S. Kwaket al., “A Review of Wireless Body Area Networks for MedicalApplications,” International Journal of Communications, Network andSystem Sciences, vol. 2, no. 08, p. 797, 2009.

[11] G. Fortino, D. Parisi, V. Pirrone, and G. D. Fatta, “BodyCloud: ASaaS Approach for Community Body Sensor Networks,” FutureGeneration Computer Systems, vol. 35, no. 0, pp. 62–79, 2014.

[12] M. Quwaider and Y. Jararweh, “Cloudlet-Based Efficient DataCollection in Wireless Body Area Networks ,” Simulation ModellingPractice and Theory, vol. 50, no. 0, pp. 57–71, 2015.

[13] F. Zhang, J. Cao, S. U. Khan, K. Li, and K. Hwang, “A Task-LevelAdaptive MapReduce Framework for Real-Time Streaming Datain Healthcare Applications,” Future Generation Computer Systems,vol. 43, pp. 149–160, 2015.

[14] M. Chen, “NDNC-BAN: Supporting Rich Media Healthcare Ser-vices via Named Data Networking in Cloud-Assisted WirelessBody Area Networks,” Information Sciences, vol. 284, no. 0, pp. 142–156, 2014.

[15] O. Diallo, J. J. Rodrigues, M. Sene, and J. Niu, “Real-Time QueryProcessing Optimization for Cloud-based Wireless Body AreaNetworks,” Information Sciences, vol. 284, no. 0, pp. 84–94, 2014.

[16] E. Rachkidi, E. H. Cherkaoui, M. Ait-idir, N. Agoulmine, N. C.Taher, M. Santos, and S. Fernandes, “Cooperative DynamiceHealth Service Placement in Mobile Cloud Computing,” in Pro-ceedings of International Conference on E-health Networking, Applica-tion Services, 2015, pp. 627–632.

[17] A. Abbas and S. U. Khan, “e-Health Cloud: Privacy Concerns andMitigation Strategies,” in Medical Data Privacy Handbook. Springer,2015, pp. 389–421.

[18] G. V. Prasad, A. S. Prasad, and S. Rao, “A Combinatorial AuctionMechanism for Multiple Resource Procurement in Cloud Comput-ing,” IEEE Transactions on Cloud Computing, vol. PP, no. 99, pp. 1–1,2016.

[19] L. Zhou, Z. Yang, J. J. P. C. Rodrigues, and M. Guizani, “ExploringBlind Online Scheduling for Mobile Cloud Multimedia Services,”IEEE Wireless Communications, vol. 20, no. 3, pp. 54–61, 2013.

[20] A. Samanta, S. Misra, and M. S. Obidat, “Wireless Body AreaNetworks with Varying Traffic in Epidemic Medical EmergencySitaution,” in Proceedings of IEEE International Conference on Com-munications, 2015, pp. 6929–6934.

[21] A. Samanta, S. Bera, and S. Misra, “Link-Quality-Aware ResourceAllocation With Load Balance in Wireless Body Area Networks,”IEEE Systems Journal, vol. 12, no. 1, pp. 74–81, 2018.

[22] A. Samanta and S. Misra, “Energy-Efficient and Distributed Net-work Management Cost Minimization in Opportunistic WirelessBody Area Networks,” IEEE Transactions on Mobile Computing,vol. 17, no. 2, pp. 376–389, 2018.

[23] ——, “EReM: Energy-Efficient Resource Management in BodyArea Networks with Fault Tolerance,” in Proceedings of IEEE GlobalCommunications Conference, 2017.

[24] E. Ibarra, A. Antonopoulos, E. Kartsakli, J. J. Rodrigues, andC. Verikoukis, “Joint Power-QoS Control Scheme for Energy Har-vesting Body Sensor Nodes,” in Proceedings of IEEE InternationalConference on Communications, 2014, pp. 3511–3516.

[25] M. Chen, Y. Ma, S. Ullah, W. Cai, and E. Song, “ROCHAS: Roboticsand Cloud-assisted Healthcare System for Empty Nester,” in Pro-ceedings of International Conference on Body Area Networks, 2013, pp.217–220.

[26] G. Kulkarni, R. Shelke, P. B. N. Patil, V. Kulkarni, and S. Mo-hite, “Optimization in Mobile Cloud Computing for Cloud BasedHealth Application,” in Proceedings of International Conference onCommunication Systems and Network Technologies, 2014, pp. 569–572.

[27] S. Wasserman, Social Network Analysis: Methods and Applications.Cambridge University Press, 1994, vol. 8.

[28] S. Ullah, H. Higgins, B. Braem, B. Latre, C. Blondia, I. Moerman,S. Saleem, Z. Rahman, and K. S. Kwak, “A Comprehensive Surveyof Wireless Body Area Networks,” Journal of Medical Systems,vol. 36, no. 3, pp. 1065–1094, 2012.

[29] N. Karowski, A. C. Viana, and A. Wolisz, “Optimized Asyn-chronous Multichannel Discovery of IEEE 802.15.4-Based WirelessPersonal Area Networks,” IEEE Transactions on Mobile Computing,vol. 12, no. 10, pp. 1972–1985, 2013.

[30] Y. Zhang, C.-H. Feng, I. Demirkol, and W. B. Heinzelman, “Energy-efficient duty cycle assignment for receiver-based convergecast inwireless sensor networks,” in Proceedings of IEEE Global Telecom-munications Conference, 2010, pp. 1–5.

[31] A. Kamel, A. Al-Fuqaha, and M. Guizani, “Exploiting Client-SideCollected Measurements to Perform QoS Assessment of IaaS,”IEEE Transactions on Mobile Computing, vol. 14, no. 9, pp. 1876–1887, 2015.

[32] S. Ren and M. van der Schaar, “Dynamic Scheduling and Pricingin Wireless Cloud Computing,” IEEE Transactions on Mobile Com-puting, vol. 13, no. 10, pp. 2283–2292, 2014.

[33] N. A. Nuaimi, A. A. Shamsi, N. Mohamed, and J. Al-Jaroodi, “e-Health Cloud Implementation Issues and Efforts,” in Proceedingsof International Conference on Industrial Engineering and OperationsManagement, 2015, pp. 1–10.

[34] Q. Chen, S. Kanhere, and M. Hassan, “Analysis of Per-node TrafficLoad in Multi-hop Wireless Sensor Networks,” IEEE Transactionson Wireless Communications, vol. 8, no. 2, pp. 958–967, Feb 2009.

[35] V. Kantere, D. Dash, G. Francois, S. Kyriakopoulou, and A. Aila-maki, “Optimal Service Pricing for a Cloud Cache,” IEEE Transac-tions on Knowledge and Data Engineering, vol. 23, no. 9, pp. 1345–1358, 2011.

[36] S. Moulik, S. Misra, and A. Gaurav, “Cost-Effective Mapping be-tween Wireless Body Area Networks and Cloud Service Providersbased on Multi-Stage Bargaining,” IEEE Transactions on MobileComputing, vol. 16, no. 6, pp. 1573–1586, 2017.

[37] S. Misra and S. Sarkar, “Priority-Based Time-Slot Allocation inWireless Body Area Networks During Medical Emergency Situa-tions: An Evolutionary Game Theoretic Perspective,” IEEE Journalof Biomedical and Health Informatics, vol. PP, no. 99, pp. 1–1, 2014.

[38] P. Wang and X. Du, “QoS-aware Service Selection Using AnIncentive Mechanism,” IEEE Transactions on Services Computing,vol. PP, no. 99, pp. 1–1, 2016.

[39] G. Almashaqbeh, T. Hayajneh, and A. Vasilakos, “A Cloud-BasedInterference-Aware Remote Health Monitoring System for Non-hospitalized Patients,” in Proceedings of IEEE Global Communica-tions Conference, 2014, pp. 2436–2441.

[40] M. Nabi, M. Geilen, and T. Basten, “MoBAN: A ConfigurableMobility Model for Wireless Body Area Networks,” in Proceedingsof International Conference on Simulation Tools and Techniques, 2011,pp. 168–177.

[41] X. Cai, J. Li, J. Yuan, W. Zhu, and Q. Wu, “Energy-Aware AdaptiveTopology Adjustment in Wireless Body Area Networks,” Telecom-munication Systems, vol. 58, no. 2, pp. 139–152, 2015.

[42] A. Samanta, Y. Li, and S. Chen, “QoS-Aware Heuristic Schedulingwith Delay-Constraint for WBSNs,” in Proceedings of IEEE Interna-tional Conference on Communications, 2018.

[43] A. Samanta and S. Misra, “Dynamic Connectivity Establishmentand Cooperative Scheduling for QoS-Aware Wireless Body AreaNetworks,” IEEE Transactions on Mobile Computing, vol. PP, no. 99,pp. 1–1, 2018.

[44] D. D. Guglielmo, F. Restuccia, G. Anastasi, M. Conti, andS. Das, “Accurate and Efficient Modeling of 802.15.4 UnslottedCSMA/CA through Event Chains Computation,” IEEE Transac-tions on Mobile Computing, vol. PP, no. 99, pp. 1–1, 2016.

[45] M. Sandhu, N. Javaid, M. Akbar, F. Najeeb, U. Qasim, andZ. Khan, “FEEL: Forwarding Data Energy Efficiently with LoadBalancing in Wireless Body Area Networks,” in Proceedings ofIEEE International Conference on Advanced Information Networkingand Applications, 2014, pp. 783–789.

[46] J. Zhao, Y. Xiang, T. Lan, H. H. Huang, and S. Subramaniam, “Elas-tic Reliability Optimization Through Peer-to-Peer Checkpointingin Cloud Computing,” IEEE Transactions on Parallel and DistributedSystems, vol. 28, no. 2, pp. 491–502, 2017.

[47] A. Samanta and Y. Li, “DeServE: Delay-agnostic Service Offload-ing in Mobile Edge Clouds: Poster,” in Proceedings of ACM/IEEESymposium on Edge Computing, 2017, pp. 24:1–24:2.

1 trafﬁc-aware efﬁcient mapping of wireless body area...

Documents