ACCEPTED TO APPEAR ON IEEE COMMUNICATIONS … · · 2016-09-15tutorial and survey of recent research and development efforts ... cording to the latest developments of WSNs

6ACCEPTED TO APPEAR ON IEEE COMMUNICATIONS SURVEYS amp TUTORIALS SEPT 2016 1

A Survey of Multi-Objective Optimization inWireless Sensor Networks Metrics Algorithms and

Open ProblemsZesong FeiSenior Member IEEEBin Li Shaoshi YangMember IEEE

Chengwen XingMember IEEEHongbin Chen and Lajos HanzoFellow IEEE

AbstractmdashWireless sensor networks (WSNs) have attractedsubstantial research interest especially in the context of perform-ing monitoring and surveillance tasks However it is challengingto strike compelling trade-offs amongst the various conflictingoptimization criteria such as the networkrsquos energy dissipationpacket-loss rate coverage and lifetime This paper provides atutorial and survey of recent research and development effortsaddressing this issue by using the technique of multi-objectiveoptimization (MOO) First we provide an overview of themain optimization objectives used in WSNs Then we elaborateon various prevalent approaches conceived for MOO such asthe family of mathematical programming based scalarizationmethods the family of heuristicsmetaheuristics based optimiza-tion algorithms and a variety of other advanced optimizationtechniques Furthermore we summarize a range of recent studiesof MOO in the context of WSNs which are intended to provideuseful guidelines for researchers to understand the referencedliterature Finally we discuss a range of open problems to betackled by future research

Index TermsmdashWireless sensor networks (WSNs) multi-objective optimization trade-offs Pareto-optimal solution

GLOSSARY

2D two-dimensional3D three-dimensionalABC artificial bee colonyACO ant colony optimizationAHP analytical hierarchy processAI artificial intelligenceANN artificial neural networkAPF artificial potential fieldBER bit-error rateBOA Bayesian optimization algorithmCIVA centralized immune-Voronoi deployment algorithmCR cognitive radioCR-WSN cognitive radio aided WSNDE differential evolutionDoS denial-of-serviceDPAP deployment and power assignment problemDSA dynamic spectrum accessDSC disjoint set cover

The financial support of the National Natural Science Foundation of China(Grant No 61371075 and 61421001) of the 111 Project of China (Grant NoB14010) and of the European Research Council (ERC) Advanced FellowGrant ldquoBeam-Me-Uprdquo is gratefully acknowledged

Z Fei B Li and C Xing are with the School of Information andElectronics Beijing Institute of Technology Beijing 100081 China (e-mailfeizesongbiteducn libinsunbiteducn chengwenxingieeeorg)

S Yang and L Hanzo are with the School of Electronics and ComputerScience University of Southampton Southampton SO17 1BJ UK (e-mailsy7g09 lhecssotonacuk)

H Chen is with the Key Laboratory of Cognitive Radio and InformationProcessing (Ministry of Education) Guilin University of Electronic Technol-ogy Guilin 541004 China (email chbscutgueteducn)

EA evolutionary algorithmEDLA energy-density-latency-accuracyEMOCA evolutionary multi-objective crowding algorithmFA firefly algorithmFL fuzzy logicFRMOO fuzzy random multi-objective optimizationGA genetic algorithmGP goal programmingHBOA hierarchical Bayesian optimization algorithmICA imperialist competitive algorithmIoT Internet of ThingsISM industrial scientific and medicalMA memetic algorithmMAC medium access controlMDP Markov decision processMODA multi-objective deployment algorithmMODE multi-objective differential evolutionMOEA multi-objective evolutionary algorithmMOEAD multi-objective evolutionary algorithm based on decomposi-

tionMOEADFD multi-objective evolutionary algorithm based ondecomposi-

tion with fuzzy dominanceMOGA multi-objective genetic algorithmMOGLS multi-objective genetic local searchMOICA multi-objective imperialist competitive algorithmMOMGA multi-objective messy genetic algorithmMOMGA-II multi-objective messy genetic algorithm-IIMOO multi-objective optimizationMOP multi-objective optimization problemMOSS multi-objective scatter searchMOTS multi-objective tabu searchNPGA niched Pareto genetic algorithmNSGA non-dominated sorting genetic algorithmNSGA-II non-dominated sorting genetic algorithm-IINUM network utility maximizationOSI open systems interconnectionPAES Pareto archive evolution strategyPESA Pareto envelope-based selection algorithmPESA-II Pareto envelope-based selection algorithm-IIPF Pareto frontPHY physical layerPS Pareto setPSO particle swarm optimizationQoS quality-of-serviceRL reinforcement learningSDD subgradient dual decompositionSINR signal-to-interference-plus-noise ratioSIOA swarm intelligence based optimization algorithmSPEA strength Pareto evolutionary algorithmSPEA2 strength Pareto evolutionary algorithm-2WBAN wireless body area networkWLAN wireless local area networkWSN wireless sensor network

I I NTRODUCTION

A Motivation

ACCEPTED TO APPEAR ON IEEE COMMUNICATIONS SURVEYS amp TUTORIALS SEPT 2016 2

W IRELESS sensor networks (WSNs) consist of alarge number of compact low-cost low-power multi-

functional sensor nodes that communicate wirelessly overshort distances [1] [2] In WSNs the sensor nodes aregenerally deployed randomly in the field of interest which areextensively used for performing monitoring and surveillancetasks [3]ndash[5] Depending on the specific application scenariosWSNs may rely on diverse performance metrics to be opti-mized For example the energy efficiency and network lifetimeare among the major concerns in WSNs since the sensor nodesare typically powered by battery whose replacement is oftendifficult Furthermore the network coverage latency and thefairness among sensor nodes are important for maintainingthe quality-of-service (QoS) [6] [7] In practice these metricsoften conflict with each other hence the careful balancing ofthe trade-offs among them is vital in terms of optimizing theoverall performance of WSNs in real applications

In conventional WSN designs typically the most salientperformance metric is chosen as the optimization objectivewhile the remaining performance metrics are normally treatedas the constraints of the optimization problem Such single-objective optimization approaches however may be unfairand unreasonable in real WSN applications since it artificiallyover-emphasizes the importance of one of the metrics to thedetriment of the rest [8] Hence a more realistic optimizationis to simultaneously satisfy multiple objectives such as themaximal energy efficiency the shortest delay the longestnetwork lifetime the highest reliability and the most balanceddistribution of the nodesrsquo residual energy or the trade-offsamong the above objectives [9] [10] Accordingly multi-objective optimization (MOO) can be naturally adopted forsolving the above problem since it may be more consistentwith the realistic scenarios [11]

MOO algorithms have been a subject of intense interestto researchers for solving diverse multi-objective optimizationproblems (MOPs) in which multiple objectives are treatedsimultaneously subject to a set of constraints [12] Howeverit is infeasible for multiple objectives to achieve their re-spective optima at the same time thus there may not exista single globally optimal solution which is the best withrespect to all objectives Nevertheless there exists a setofPareto-optimal or non-dominated solutions generating a setof Pareto-optimal outcomesobjective vectors which is calledPareto frontfrontier (PF) or Pareto boundarycurvesurfaceExplicitly the PF is generated by the specific set of solutionsfor which none of the multiple objectives can be improvedwithout sacrificing the other objectives [13] This set of Pareto-optimal or non-dominated solutions constitutes the focus ofour interest and it is also called the Pareto-efficient set orPareto set (PS) that is mapped to the PF in the objectivefunction space [14]

Diverse approaches such as mathematical programmingbased scalarization methods and nature-inspired metaheuris-tics may be used for finding the PSs of MOPs Scalarizingan MOP means formulating a single-objective optimizationproblem such that optimal solutions to the single-objectiveoptimization problem are Pareto-optimal solutions to the MOP[15] In addition it is often required that every Pareto-optimal

solution can be reached with the aid of specific parametersof the scalarization Representatives of scalarization methodsinclude the linear weighted-sum method theε-constraintsmethod [15] and goal programming (GP) based methodsMOPs are more often solved by bio-inspired metaheuristicssuch as multi-objective evolutionary algorithms (MOEAs)[16] [17] and swarm intelligence based optimization algo-rithms (SIOAs) [18] MOEAs aim for finding a set of represen-tative Pareto-optimal solutions in a single run [14] [19][20]As a subset of MOEAs the multi-objective genetic algorithms(MOGAs) such as the strength Pareto evolutionary algorithm(SPEA) [16] and the non-dominated sorting genetic algorithm-II (NSGA-II) [21] have been particularly widely researched inthe family of MOO algorithms [22] because they are capableof efficiently constructing an approximate PF This is mainlydue to the fact that MOGAs accommodate a diverse varietyof bio-inspired operators to iteratively generate a populationof feasible solutions Compared to genetic algorithms (GAs)that rely on the interplay between genetics and biologicalevolution SIOAs seek to understand the collective behavior ofanimals particularly insects and to use this understanding forsolving complex nonlinear problems One of the most widelyused SIOAs is the ant colony optimization (ACO) algorithm[23] which has indeed been invoked for solving the MOPs inWSNs [24] Several other bio-inspired algorithms related toswarm intelligence will be surveyed in Section IV

B Contributions of This Survey

In this paper we focus our attention on various basic con-cepts conflicting performance criteriaoptimization objectivesas well as the MOO techniques conceived for striking a trade-off in the context of WSNs The contributions of our work arefour-fold which are listed as follows

bull We provide in-depth discussions on the basics metricsand relevant algorithms conceived for MOO in WSNs

bull We present a comprehensive coverage and clear classi-fication of various prevalent MOO algorithms conceivedfor solving MOPs and clarify the strengths and weak-nesses of each MOO algorithm in the context of WSNs

bull We provide an exhaustive review of the up-to-date re-search progress of MOO in WSNs according to differenttrade-off metrics

bull We highlight a variety of open research challenges andidentify possible future trends for MOO in WSNs ac-cording to the latest developments of WSNs

C Paper Organization

The reminder of this paper is organized as follows InSection II we summarize the related surveys of MOO inWSNs In Section III we commence with an overview ofWSNs in terms of their system model and applicationsFurthermore we introduce the main optimization objectivesof interest in WSNs In Section IV we present the family ofMOO techniques that can in principle be used for solving thiskind of problems In Section V we provide an overview ofthe existing studies dedicated to multi-objective methodsinWSNs Finally in Section VI we describe a range of open

Sec II Related Surveys and Tutorials

Sec III Fundamentals of WSNs

System Model

Applications

MOO Metrics

Coverage

Network Connectivity

Network Lifet ime

Energy Consumption

Energy Efficiency

Network Latency

Differentiated Detection Levels

Number of Nodes

Fault Tolerance

Fair Rate Allocation

Detection Accuracy

Network Security

Sec IV Techniques of MOO

Optimization Strategies

MOO Algorithms

Sec V Existing Literature on Using MOO in WSNs

Coverage-versus-Lifet ime Trade-offs

Energy-versus-Latency Trade-offs

Lifet ime-versus-Applicat ion-Performance Trade-offs

Trade-offs Related to the Number of Nodes

Reliabili ty-Related Trade-offs

Trade-offs Related to Other Metrics

Sec VI Open Problems and Discussions

Sec VII Conclusions

Sec I Introduction

Software Tools

Fig 1 The organization of this paper

problems and possible future research directions followed byour conclusions in Section VII For the sake of explicit claritythe organization of this paper is shown in Fig 1

II RELATED SURVEYS AND TUTORIALS

A range of surveys have been dedicated to diverse single-objective research domains in WSNs such as their energyefficiency [25] routing [26] congestion control [27] theirMAC protocols [28] data collection [6] privacy and security[29] localization [30] [31] cross-layer QoS guarantees[32]sink mobility management [33] and network virtualization[34]

In recent years several surveys and tutorials advocatedMOO methods for optimizing the conflicting performanceobjectives of WSNs Specifically the authors of [35] provideda review of recent studies on multi-objective scheduling anddiscussed its future research trends In [36] the MOO criteriaand strategies conceived for node deployment in WSNs weresurveyed Performance trade-off mechanisms of the routingprotocols designed for energy-efficient WSNs were reviewedin [37] where various artificial intelligence techniques andthe related technical features of the routing protocols werediscussed The authors of [38] surveyed the most representa-tive MOEAs and their major applications from a historicalperspective Konaket al [39] presented a comprehensivesurvey and tutorial of MOGAs Furthermore Adnanet al[40] provided a holistic overview of bio-inspired optimizationtechniques such as particle swarm optimization (PSO) ACOand GA In particular the authors of [41] presented a briefsurvey of how to apply PSO in WSN applications whilebearing in mind the peculiar characteristics of sensor nodesThey also presented a state-of-the-art survey of computa-tional intelligence in the context of WSNs and highlightednumerous challenges facing each of the MOPs discussed [42]Additionally the authors of [43] reviewed the optimizationin biological systems and discussed bio-inspired optimizationof non-biological systems In contrast to other surveys theauthors of [44] provided a classification of algorithms pro-posed in the literature for planned deployment of WSNs Theydiscussed and compared diverse WSN deployment algorithmsin terms of their assumptions objectives and performanceAdditionally in a more recent study [45] [46] the authorsreviewed the MOO techniques and simulation tools conceivedfor solving different problems related to the design operationdeployment placement planning and management of WSNsThe above-mentioned surveys related to MOO in WSNs areoutlined at a glance in Table I which allows the readers tocapture the main contributions of each of the existing surveys

III F UNDAMENTALS OF WSNS

A System Model

WSNs generally consist of hundreds or potentially eventhousands of spatially distributed low-cost low-powermulti-functional autonomous sensor nodes and communicate overshort distances [1] Each node is usually equipped with asensor unit a processor a radio transceiver an AD converter

Sensors13

Processor13Memory13A1313D 13

Con13verter13

Radio Transceiver13

Power Supply13

Fig 2 Typical architecture of a WSN node

a memory unit and a power supply (battery) The typical ar-chitecture of a WSN node is illustrated in Fig 2 A WSN nodemay also have additional application-dependent componentsattached such as the location finding system and mobilizerBy combining these different components into a miniaturizeddevice these sensor nodes become multi-functional In otherwords the structure and characteristics of sensor nodes dependboth on their electronic mechanical and communication lim-itations as well as on their application-specific requirementsOne of the great challenges facing WSNs is to use suchresource-constrained sensor nodes to meet certain applicationrequirements including sensing coverage network lifetimeand end-to-end delay

Typically sensor nodes are grouped into clusters and eachcluster has a node that acts as the cluster head which hasmore resources and computational power than the other clusternodes All nodes gather and deliver their sensed informationto the cluster head which in turn forwards it to a specializednode namely the sink node or base station via a hop-by-hopwireless communication link In indoor scenarios a WSN istypically rather small and consists of a single cluster supportedby a single base station Multiple clusters relying on multiplebase stations are possible in a large-scale deployment ofWSNs Fig 3 shows the relationship between WSNs andthe infrastructure-based networks Typically a sink nodeorbase station is responsible for gathering the uplink informationgleaned from sensor nodes through either single-hop or multi-hop communications Then the sink node sends the collectedinformation to the interested users via a gateway often usingthe internet or any other communication path [47] It shouldbe noted that with the development of machine-to-machinecommunications it is possible to have sensors and machinesdirectly connected to cellular network based mobile Internet

At the time of writing the most common way of construct-ing WSNs relies on the ZigBee communications protocolwhich complies with the IEEE 802154 standard outliningthe specifications of both the physical layer (PHY) and themedium access control (MAC) layer This is widely regardedas thede factostandard for WSNs [48] A WSN operates inthe unlicensed industrial scientific and medical (ISM) bandin which it coexists with many other successful communication

TABLE I Existing Surveys Relating to MOO in WSNs

Reference Focus Topics Reference Focus Topics

[35] multi-objective scheduling [41] a brief survey of PSO

[36] sensor node deployment [42] computational intelligence paradigms

[37] artificial intelligence methods for routing pro-tocols

[43] analogy between optimization in biological systemsand bio-inspired optimization in non-biological sys-tems

[38] MOEAs [44] multi-objective node deployment algorithms

[39] MOGAs [45] MOO techniques associated with the design operationdeployment placement planning and management

[40] bio-inspired optimization techniques [46] engineering applications and simulation tools

Proxy Server

Gateway

Sensors

Fig 3 Wireless sensor networks and their relationship to infrastructure-based networks

systems such as the IEEE 80211 standard based wirelesslocal area network (WLAN) and the 802151 standard basedBluetooth communication systems Therefore a WSN mayface the challenge of co-channel interferences imposed byboth other WSNs and other co-existing heterogeneous wirelesssystems This coexistence problem may substantially affect theperformance of WSNs

B Applications

In WSNs sensor nodes are generally deployed randomly inthe majority of application domains When the sensor nodesare deployed in hostile remote environments they may beequipped with high-efficiency energy harvesting devices (egsolar cells) for extending the network lifetime [49] Numerouspractical applications of WSNs have been rolled out with theadvancement of technologies In general the applicationsofWSNs can be classified into two types monitoring and track-ing Monitoring is used for analyzing supervising and care-fully controlling operation of a system in real-time Tracking isgenerally used for following the change of an event a personan animal and so on Existing monitoring applications in-clude indooroutdoor environmental monitoring [50] industrial

monitoring [51] precision agriculture (eg irrigationmanage-ment and crop disease prediction) [52] biomedical or healthmonitoring [53] electrical network monitoring [54] militarylocation monitoring [55] and so forth Tracking applicationsinclude habitat tracking [56] traffic tracking [57] militarytarget tracking [58] etc We summarize their classification inFig 4

A more detailed portrayal of WSN applications is given inFig 5 For instance in environmental monitoring WSNs canhelp us perform forest fire detection flood detection forecastof earthquakes and eruptions pollution monitoring etc [50][59] In industry and agriculture WSNs can sense and detectfarming and wildlife monitor equipment and goods protectcommercial property predict crop disease and production qual-ity control pests and diseases etc [51] [52] In healthcare andbiomedical applications WSNs can be utilized for diagnosticsdistance-monitoring of patients and their physiological dataas well as for tracking the medicine particles inside patientsrsquobody etc [53] In particular WSN based wireless body areanetworks (WBANs) can help monitor human body functionsand characteristics (eg artificial retina vital signsautomaticdrug delivery etc) acting as anin vitro or in vivo diagnosticsystem [60] In the infrastructure WSNs have been widely

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

(sense and detect the environment to forecast impending disasters such

as water quality weather temperature seismic forest fires)

(irrigation management humidity monitoring)

(monitoring animals)

(supply chain industrial processes machinery productivity)

(transport and circulation systems self-identification parking management)

(organ monitoring wellness surgical operation)

(intrusion detection)

(monitoring any addressable device in the house)

(traffic monitoring fault detection)

(tracking the migration of animals)

(monitoring of doctors and patients in a hospital)

(sensor nodes can be deployed on a battlefield or enemy zone to track

monitor and locate enemy troop movements)

Fig 4 Taxonomy of WSN applications

Wireless Sensor Network

Applications

EnvironmentIndustry and Agriculture Military

Body Area Network (Health) Infrastructure

Farming and

WildlifeEquipment and goods

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

devices SportsAssisted

livingRoads and

Railway

buildings

Smart energy

meteringSmart

infrastructure Water

monitoring

Fig 5 Overview of WSN applications in real environments Here SHIMMER represents the Intel digital health grouprsquosldquosensing health with intelligence modularity mobilityand experimental re-usabilityrdquo [47]

used for monitoring the railway systems and their componentssuch as bridges rail tracks track beds track equipment aswell as chassis wheels and wagons that are closely relatedtorolling stock quality [61] WSNs also allow users to managevarious appliances both locally and remotely for buildingautomation applications [62] In military target trackingandsurveillance a WSN can assist in intrusion detection andidentification Specific examples include enemy troop andtank movements battle damage assessment detection andreconnaissance of biological chemical and nuclear attacks etc[58]

Furthermore several novel WSN application scenarios suchas the Internet of Things [63] cyber-physical systems [64]and smart grids [65] among others have adopted new designapproaches that support multiple concurrent applicationsonthe same WSN The applications of WSNs are not limitedto the areas mentioned in this paper The future prospects of

WSN applications are promising in terms of revolutionizingour daily lives

C MOO Metrics

In this subsection a succinct overview of the most popularoptimization objectives of WSNs is provided Over the pastyears a number of research contributions have addressed di-verse aspects of WSNs including their protocols [66] routing[67] energy conservation [25] lifetime [68] and so forthTheQoS as perceived by the users or applications was giveninsufficient attention at the beginning However at the timeof writing how to provide the desired QoS is becoming anincreasingly important topic for researchers Different appli-cations may have their own specific QoS requirements butsome of the more commonly used metrics for characterizingQoS are the coverage area and quality the delay the number

of active nodes the bit-error rate (BER) and the overall WSNlifetime

There are many other QoS metrics worth mentioning anda range of factors affecting the QoS in WSNs are portrayedin Fig 6 which was directly inspired by [69] and reflectsthe application requirements of a WSN It is indeed plausiblethat the networkrsquos performance can be quantified in termsof its energy conservation lifetime and QoS-based metricsin specific applications However multiple metrics usuallyconflict with each other For example when more energy isconsumed by the nodes the operating lifetime of the networkreduces Similarly if more active nodes are deployed in a givenregion a lower per-node power is sufficient for maintainingconnectivity but the overall delay is likely to be increaseddue to the increased number of hops Hence an application-specific compromise has to be struck between having moreshort hops imposing a lower power dissipation but higher delayand having more longer hops which reduces the delay but mayincrease the transmit-power dissipation

1) Coverage ldquoCoveragerdquo is one of the most importantperformance metrics for a sensor network In other wirelesscommunication networks coverage typically means the radiocoverage By contrast coverage in the context of WSNs cor-responds to the sensing range while connectivity correspondsto the communication range WSN coverage can be classifiedinto three typesarea coverage point coverageand barriercoverage[44] In area coverage the coverage quality of anentire two-dimensional (2D) region is considered where eachpoint in the region is observed by at least one sensor nodeIn point coverage the objective is to simply guarantee that afinite set of points in the region are observed by at least onesensor nodeBarrier coverageusually deals with the detectionof movement across a barrier of sensor nodes The mostrichly studied coverage problem in the WSN literature is thearea coverage problem The characterization of the coveragevaries depending both on the underlying models of eachnodersquos field of view and on the metric used for appraising thecollective coverage Several coverage models [70]ndash[72] havebeen proposed for different application scenarios A coveragemodel is normally defined with respect to the sensing rangeof a sensor node The most commonly used node coveragemodel is the so-calledsensing disk model where all pointswithin a disk centered at the node are considered to be coveredby the node [73] More specifically a pointp is regarded tobe coveredmonitored by at least a nodev if their Euclideandistance is less than the sensing rangeRs of nodev

Given a set of nodes finding the optimal positions of thesenodes to achieve maximum coverage is in general an NP-complete problem [11] There are many different ways ofsolving the coverage optimization problem suboptimally Herein order to make the coverage problem more computationallymanageable we consider an areaA represented by a rectan-gular grid which is divided intoG = xy rectangular cells ofidentical size and let(xj yj) denote the coordinate of nodej As a result the network coverageCv(x) is defined as thepercentage of the adequately covered cells over the total cellsof A and it is evaluated as follows [11] [74]

Cv(x) =

xprime=0

yprime=0 g(xprime yprime)

xy (1)

g(xprime

yprime) =

1 if exist j isin 1 N d(xjyj)(xprimeyprime) le Rs

0 otherwise(2)

whereN is the number of nodesg(xprime yprime) is the monitoringstatus of the cell centered at(xprime yprime) Rs is the sensing rangeof a node andd(xj yj)(xprimeyprime) is the distance from the locationof node j to the cell centered at(xprime yprime) Having a bettercoverage also leads to a higher probability of detecting theevent monitored [75]

2) Network ConnectivityAnother issue in WSN designis the network connectivity that is dependent on the selectedcommunication protocol [76] Two sensor nodes are directlyconnected if the distance of the two nodes is smaller thanthe communication rangeRc Connectivity only requires thatthe location of any active nodeis within the communicationrange of one or more active nodes so thatall active nodescan form a connected communication network The mostcommon protocol relies on a cluster-based architecture whereall nodes in the same cluster can directly communicate witheach other via a single hop and all nodes in the same clustercan communicate with all nodes in the neighboring clustersvia the cluster head In a given cluster only a single nodeacts as the cluster head which has to be active in terms ofcollecting all information gleaned by the other nodes for thesake of maintaining connectivity In cluster-based WSNs theconnectivity issues tend to hinge on the number of nodes ineach cluster (because a cluster head can only handle up toa specific number of connected nodes) as well as on thecoverage issues related to the ability of any location to becovered by at least one active sensor node

For an area represented by a rectangular grid of sizextimes ylet Rci andRsi denote the communication range and sensingrange of theith sensor node respectively To guarantee eachsensor node is placed within the communication range of atleast another sensor node and to prevent sensor nodes frombecoming too close to each others the objective functionassociated with the network connectivity can be expressed as[77]

fcon =

xtimesysum

1minus eminus(RciminusRsi

whereRci minusRsi gt 0 has to be satisfied for achieving networkconnectivity

Maintaining the networkrsquos connectivity is essential for en-suring that the messages are indeed propagated to the appropri-ate sink node or base station and the loss of connectivity isoften treated as the end of the networkrsquos lifetime Networkconnectivity is closely related to the coverage and energyefficiency of WSNs To elaborate a little further substantialenergy savings can be achieved by dynamic managementof node duty cycles in WSNs having high node densityIn this method some nodes can be scheduled to sleep (orenter a power-saving mode) while the remaining active nodes

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Fig 6 The interplay of factors affecting the QoS in WSNs

provide continuous service As far as this approach is con-cerned a fundamental problem is to minimize the numberof nodes that remain active while still achieving acceptableQoS In particular maintaining an adequate sensing coverageand network connectivity with the active nodes is a criticalrequirement in WSNs The relationship between coverageand connectivity hinges on the ratio of the communicationrange to the sensing range A connected network may notbe capable of guaranteeing adequate coverage regardless ofthe ranges By contrast in [78] [79] the authors presenteda sufficient condition for guaranteeing network connectivitywhich states that for a set of nodes that cover a convexregion the network remains connected ifRc ge 2Rs Thereexist tighter relationships betweenRc and Rs for achievingnetwork connectivity provided that adequate sensing coverageis guaranteed [80]ndash[82] Intuitively if the communicationsrange of sensor nodes is sufficiently large then maintainingconnectivity is not a problem because in this case there alwaysexists a node to communicate with A more in-depth discussionof the relationship between coverage connectivity and energyefficiency of WSNs can be found in [78]ndash[82]

3) Network LifetimeAnother important performance met-ric in WSNs is their lifetime Tremendous research efforts havebeen invested into solving the problem of prolonging networklifetime by energy conservation in WSNs Indeed the energysource of each node is generally limited while recharging orreplacing the battery at the sensors may be impossible Henceboth the radio transceiver unit and the sensor unit of eachnode have to be energy-efficient and it is vitally importanttomaximize the attainable network lifetime [83] defined as thetime interval between the initialization of the network andthedepletion of the battery of any of the sensor nodes

For the simplicity of exposition typically all sensor nodesare assumed to be of equal importance which is a reasonableassumption since the ldquodeathrdquo of one sensor node may resultin the network becoming partitioned or some area requiringmonitoring to be uncovered Thus the networkrsquos lifetime

is defined as the time duration from the applicationrsquos firstactivation to the time instant when any of the sensor nodesin the cluster fails due to its depleted energy source

More explicitly this objective can be formulated as

Tnet = minTj (4)

wherej = 1 2 middot middot middot N The lifetime of a sensor node is generally inversely pro-

portional both to the average rate of its own informationgenerated and to the information relayed by this node Hencethe networkrsquos lifetime is also partially determined by thesource rates of all the sensor nodes in the network

4) Energy ConsumptionSensor nodes are equipped withlimited battery power and the total energy consumption of theWSN is a critical consideration Each node consumes someenergy during its data acquisition processing and transmissionphases For instance in a heterogeneous WSN different sensornodes might have diverse power and data processing capabil-ities Hence the energy consumption of a WSN depends bothon the Shannon capacity of the channels among the nodesand on these nodesrsquo functionality The energy consumption ofa pathP is the sum of the energy expended at each node alongthe path hence the total energy consumptionE(P ) of a givenpath is given by [84]

E(P ) =

(tai + tpi )times P oi + P t

i times tm (5)

wheretai andtpi indicate the time durations of data acquisitionand data processing taking place at nodei respectivelyL isthe number of nodes on the given path Furthermoretm isthe message transmission time whileP o

i andP ti denote the

operational power and transmission power dissipation of nodei respectively

5) Energy EfficiencyEnergy efficiency is a key concern inWSNs and this metric is closely related to network lifetimein

the particular context of WSNs1 [90] As an example hereinthe energy efficiency of nodei is defined as the ratio of thetransmission rate to the power dissipation Explicitly itisformulated as

ηi =W log2(1 + γi)

pi (6)

where W denotes the communication bandwidthpi is thetransmission power of nodei and γi denotes the signal-to-interference-plus-noise ratio (SINR) at the destination receiverrelative to nodei respectively

Due to the limited energy resources of each node we have toutilize these nodes in an efficient manner so as to increase thelifetime of the network [91] There are at least two approachesto deal with the energy conservation problem in WSNs Thefirst approach is to plan a schedule of active nodes whileenabling the other nodes to enter a sleep mode The secondapproach is to dynamically adjust the sensing range of nodesfor the sake of energy conservation

6) Network LatencyFor a WSN typically a fixed band-width is available for data transfer between nodes Againhaving an increased number of nodes results in more pathsbecoming available for simultaneously routing packets to theirdestinations which is beneficial for reducing the latencyMeanwhile this may also degrade the latency that increasesproportionally to the number of nodes on the invoked pathsThis is due to additional contention for the wireless channelwhen the node density increases as well as owing to routingand buffering delays

The delay between source nodeuso and sink nodeusidenoted asDusousi

is defined as the time elapsed betweenthe departure of a collected data packet fromuso and its arrivalto usi and is given by [92] [93]

Dusousi= (Tq + Tp + Td)timesN(uso usi)

= ctimesN(uso usi)

prop N(uso usi) (7)

whereTq is the queue delay per intermediate forwarderTp

is the propagation delay andTd is the transmission delay Allof them are for the sake of simplicity regarded as constantsand collectively denoted byc FinallyN(uso usi) denotes thetotal number of data disseminators betweenuso andusi Asa consequence the minimization of the delay corresponds tominimizing the number of intermediate forwarders betweenthe source and the sink It is worth noting that Haenggiet al [94] astutely argued that long-hop based routing is avery competitive strategy compared to short-hop aided routingin terms of latency albeit this design dilemma also hasramifications as to the scarce energy resource of the nodes

1Note that the concept of energy efficiency is also widely usedingreen communications [85]ndash[89] The definition of energy efficiency hasseveral variants It is typically defined as the ratio of the spectral efficiency(bitssecondHz) to the power dissipation of the system considered Henceits unit is bitssecondHzWatt or equivalently bitsHzJoule Alternatively itcan be defined as the power-normalized transmission rate and hence its unitbecomes bitssecondWatt or bitsJoule

7) Differentiated Detection LevelsDifferentiated sensornetwork deployment is also an important issue In many realWSN applications such as underwater sensor deploymentsor surveillance applications certain parts of the supervisedregion may require extremely high detection probabilitiesifthese parts constitute safety-critical geographic area Howeverin the less sensitive geographic area relatively low detectionprobabilities have to be maintained for reducing the numberof nodes deployed which corresponds to reducing the costimposed Therefore different geographic areas require differ-ent densities of deployed nodes and the sensing requirementsare not necessarily uniformly distributed within the entiresupervised region

Let us used((mn) (i j)) to denote the Euclidean distancebetween the coordinates(mn) and(i j) A probabilistic nodedetection model can be formulated as [95]ndash[97]

p((mn) (i j)) =

eminusad((mn)(ij)) d((mn) (i j)) le Rs

0 d((mn) (i j)) gt Rs(8)

wherea is a parameter associated with the physical character-istics of the sensing device andRs is the sensing range

8) Number of NodesEach sensor node imposes a certaincost including its production deployment and maintenanceAs a result the total cost of the WSN increases with the num-ber of sensor nodes When deploying a WSN in a battlegroundsensor nodes have to operate as stealthily as possible to avoidbeing detected by the enemy This implies that the numberof nodes has to be kept at a minimum in order to reducethe probability of any of them being discovered [98]ndash[103]are dedicated to optimal node deployment by considering theaccomplishment of the specified goals at a minimum cost

Minimizing the number of active nodes is equivalent tomaximizing the following objective [104]

f(K prime) = 1minus|K prime|

|K| (9)

where|K prime| is the number of active nodes and|K| is the totalnumber of nodes

9) Fault Tolerance Sensor nodes may fail for exampledue to the surrounding physical conditions or when theirenergy runs out It may be difficult to replace the existingnodes hence the network has to be fault tolerant in order toprevent individual failures from reducing the network lifetime[36] [105] In other words fault tolerance can be viewed asanability to maintain the networkrsquos operation without interruptionin the case of a node failure and it is typically implementedin the routing and transport protocols The fault toleranceorreliability Rk(t) of a sensor node can be modeled using thePoisson distribution in order to capture the probability ofnothaving a failure within the time interval(0 t) as [1]

Rk(t) = eminusλkt (10)

whereλk is the failure rate of sensor nodek andt is the timeperiod

Numerous studies have been focused on formingk-connected WSNs [78] [79] [106] Thek-connectivity impliesthat there arek independent paths in the full set of the pairof nodes Fork ge 2 the network can tolerate some node

and link failures Due to the many-to-one interaction patternk-connectivity is a particularly important design factor intheneighborhood of base stations and guarantees maintaining acertain communication capacity among the nodes [106]

10) Fair Rate AllocationIt is important to guarantee thatthe sink node receives information from all sensor nodes ina fair manner when the bandwidth is limited The accuracyof the received source information depends on the allocatedsource rate Simply maximizing the total throughput of thenetwork is insufficient for guaranteeing the specific applica-tionrsquos performance since this objective may only be achievedat the expense of sacrificing the source rate supposed to beallocated to some nodes [107] For example in a sensornetwork that tracks the mobility of certain objects in a largefield of observation lower rates impose a reduced locationtracking accuracy and vice versa By simply maximizingthe total throughput instead of additionally considering theabove fairness issues among sources we may end up with asolution that shuts off many sources in the network and enablesonly those sources whose transport energy-cost to the sink isthe lowest Hence considering the fairness of rate allocationamong different sensor nodes is of high significance

An attractive methodology of achieving this goal is toadopt a network utility maximization (NUM) framework [107]in which a concave non-decreasing and twice differentiableutility function Ui(xi) quantifies the grade of satisfactionof sensor nodei with the assigned ratexi and the goalis to maximize the sum of individual utilities A specificclass of utility functions that has been extensively used forachieving fair resource allocation in economics and distributedcomputing [108] is formulated as

Uα(x) =

logx α = 11

1minusαx1minusα α gt 1

wherex = (xi foralli) and the functional operations are elemen-twise When we haveα = 1 the above utility function leadsto the so-called proportional fairness whereas whenα rarr infinthis utility function leads to maxndashmin fairness2

11) Detection AccuracyHaving a high target detectionaccuracy is also an important design goal for the sake ofachieving accurate inference about the target in WSNs Targetdetection accuracy is directly related to the timely delivery ofthe density and latency information of the WSN Assume thata nodek receives a certain amount of energyek(u) from atarget located at locationu andKo is the energy emitted bythe target Then the signal energyek(u) measured by nodek

2Themax-min criterionconstitutes one of the most commonly used fairnessmetrics [108] in which a feasible flow rate vectorx = (xiforalli) can beinterpreted as being max-min fair if the ratexi cannot be increased withoutdecreasing somexj that is smaller than or equal toxi foralli 6= j The conceptof proportional fairnesswas proposed by Kelly [109] A vector of ratesxlowast =(xlowast

i foralli) is proportionally fair if it is feasible (that isxlowast ge 0 andATxlowast le c)and if for any other feasible vectorx = (xiforalli) the aggregate of proportional

change is non-positive iesum

ximinusxlowast

xlowast

ile 0 Herec = (cmforallm) with each

element denoting the source rates to be allocatedA = (Aimforallim) is amatrix that satisfies if nodei is allocated source ratem Aim = 1 otherwiseAim = 0

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

Fig 7 Two types of privacy attacks in a WSN [29] Dataanalysis attack and traffic analysis attack conducted by amalicious node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

wheredk is the Euclidean distance between the target locationand the location of nodek p is the pathloss exponent thattypically assumes values in the range of[2 4] while α is anadjustable constant

12) Network SecuritySensor nodes may be deployed inan uncontrollable environment such as a battlefield wherean adversary might aim for launching physical attacks inorder to capture sensor nodes or to deploy counterfeit onesAs a result an adversary may retrieve private keys usedfor secure communications by eavesdropping and decrypt thecommunications of the legitimate sensors Recently muchattention has been paid to the security of WSNs There aretwo main types of privacy concerns namely data-oriented andcontext-oriented concerns [29] Data-oriented concerns focuson the privacy of data collected from a WSN while context-oriented concerns concentrate on contextual informationsuchas the location and timing of traffic flows in a WSN A simpleillustration of the two types of security attacks is depicted inFig 7

We can observe from Fig 7 that a malicious node ofthe WSN abuses its ability of decrypting data in order tocompromise the payload being transmitted in the case ofdata analysis attack In traffic-analysis attacks the adversarydoes not have the ability to decrypt data payloads Insteadit eavesdrops to intercept the transmitted data and tracks thetraffic flow on a hop-by-hop basis

For the sake of improving the network security we canminimize the loss of privacy that is calculated based oninformation theory as [110]

ζ = 1minus 2minusI(SX) (13)

whereI(SX) is the mutual information between the randomvariablesS andX More specificallyS represents the currentposition of the node of interestX is the observed variableknown to the attacker and correlated toS while I(SX) =H(S)minusH(S|X) with H(middot) denoting the entropy Additionallywe can also minimize the probability of eavesdropping in aWSN as presented in [111] to improve the network security

Table II summarizes the representative existing contribu-tions to optimizing the particular WSN performance metricsmentioned above

IV T ECHNIQUES OFMOO

In this section we briefly present the MOO techniques pro-posed in the literature for tackling various important problemsin WSNs

A Optimization Strategies

Optimization covers almost all aspects of human life andwork In practice the resources are limited hence optimizationis important Most research activities in computer scienceand engineering involve a certain amount of modeling dataanalysis computer simulations and mathematical optimizationThis branch of applied science aims for finding the particularvalues of associated variables which results in either themin-imum or the maximum values of a single objective functionor multiple objective functions [40] A typical optimizationprocess is composed of three components [43] the modelthe optimizeralgorithm and the evaluatorsimulator as shownin Fig 8 The representation of the physical problem iscarried out by using mathematical formulations to establisha mathematical model

As an important step of solving any optimization probleman efficient optimizer or algorithm has to be designed forensuring that the optimal solution is obtained There is nosingle algorithm that is suitable for all problems Optimizationalgorithms can be classified in many ways depending on thespecific characteristics that we set out to compare In generaloptimization algorithms can be classified as

1) Finitely terminating algorithms such as the family ofsimplex algorithms and their extensions as well as thefamily of combinatorial algorithms

2) Convergent iterative methods thatbull evaluate Hessians (or approximate Hessians using

finite differences) such as Newtonrsquos method andsequential quadratic programming

bull evaluate gradients or approximate gradients usingfinite differences (or even subgradients) such asquasi-Newton methods conjugate gradient meth-ods interior point methods gradient descent (al-ternatively rdquosteepest descentrdquo or rdquosteepest ascentrdquo)methods subgradient methods bundle method ofdescent ellipsoid method reduced gradient methodand simultaneous perturbation based stochastic ap-proximation methods

bull and evaluate only function values such as interpo-lation methods and pattern search methods

3) Heuristicsmetaheuristics that can provide approximatesolutions to some optimization problems

Recently bio-mimetic heuristicsmetaheuristics based strate-gies have been widely used for solving MOPs since they arecapable of obtaining near-optimal solutions to optimizationproblems characterized by non-differential nonlinear objectivefunctions which are particularly hard to deal with usingclassical gradient- or Hessian-based algorithms

A general MOP consists of a number of objectives to besimultaneously optimized and it is associated with a numberofinequality and equality constraints Without loss of generalitya multi-objective minimization problem havingn variables andm (m gt 1) objectives can be formulated as

min f(x) =min[f1(x) f2(x) middot middot middot fm(x)]

st gi(x) le 0 i = 1 2 mie

hj(x) = 0 j = 1 2 meq (14)

where we havex isin Rn with R

n being the decision spaceand f(x) isin R

m with Rm representing the objective space

The objective functions of (14) are typically in conflict witheach other in the real world Explicitly the improvement ofone of the objectives may result in the degradation of otherobjectives thus it is important to achieve the Pareto-optimalitywhich represents the conditions when none of the objectivefunctions can be reduced without increasing at least one ofthe other objective functions [13] For the minimization ofm objectivesf1(x) f2(x) middot middot middot fm(x) we have the followingdefinitions

bull Non-dominated solutions A solution a is said to domi-nate a solutionb if and only if [170](1) fi(a) le fi(b) foralli isin 1 2 middot middot middot m(2) fi(a) lt fi(b) existi isin 1 2 middot middot middot mSolutions that dominate the others but do not dominatethemselves are termed non-dominated solutions

bull Local optimality in the Pareto sense A solution a issaid to be locally optimal in the Pareto sense if thereexists a realǫ gt 0 such that there is no other solutionbdominating the solutiona with b isin R

n capB(a ǫ) whereB(a ǫ) shows a bowl having a centera and a radiusǫ

bull Global optimality in the Pareto sense A solution a isglobally optimal in the Pareto sense if there does notexist any vectorb that dominates the vectora The maindifference between global and local optimality lies inthe fact that for global optimality we no longer have arestriction imposed on the decision spaceR

n bull Pareto-optimality A feasible solution is said to be Pareto-

optimal when it is not dominated by any other solutionsin the feasible space PS which is also often referred toas the efficient set is the collection of all Pareto-optimalsolutions and their corresponding images in the objectivespace are termed the PF

The PF of an MOP is portrayed both with and withoutconstrains in Fig 9 Observe from Fig 9 (a) that the Pareto-optimal solutions of the objective functions in the PF (markedas asterisk) provide better values than any other solution inR

m The ideal solution marked by a square indicates thejoint minimum of the objective valuesf1 and f2 and it isoften difficult to reach The remaining solutions marked assolid circles are all dominated by at least one solution of thePF In contrast to the unconstrained scenario of Fig 9 (a) inFig 9 (b) the curve illustrates the PF of a constrained MOPThe solid circles in the feasible region represent the feasiblesolutions while the remaining points outside the feasibleregion (eg the points marked by triangles) are infeasible[171]

TABLE II Major Existing Approaches for EvaluatingImprovingOptimizing Each Metric

ReferencesMajor EvaluationImprovementOptimization Approaches

Protocol designMathematical programmingEAs SIOAs Hybrid algorithms Theoretical analysisSimulator

ge Area coverage [11] [91] [112]ndash[116] X X X X

Point coverage [117] [118] X

Barrier coverage [119]ndash[121] X X

Network connectivity [115] [122] [123] X

Network lifetime [11] [74] [107] [112][115] [124]ndash[131]

X X X X X X

Energy consumption [92] [116] [132]ndash[147] X X X X X X X

Energy efficiency [8] [91] [148]ndash[152] X X X X

Network latency [91] [92] [135]ndash[147][153] [154]

X X X X X X X

Differentiated detection levels[95] [155] X X

Number of nodes [98]ndash[103] [156]ndash[158] X X X

Fault tolerance [105] [159] [160] X

Fair rate allocation [107] [129] [161][162]

Detection accuracy [84] [163]ndash[167] X X X X

Network security [110] [111] [168][169]

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Mathematical model

middot Numerical model

middot Derivative free

middot Derivative based

middot Bio-mimic

middot Trajectory based

middot Population based

middot Deterministic

middot Stochastic

middotMemory less

middot History based

middot Direct calculation

middot Numerical simulator

middot Experiment or trial-and-error

Fig 8 A simple illustration of optimization process

B MOO Algorithms

Numerous studies have been devoted to the subject ofMOO and a variety of algorithms have been developed forsolving MOPs in WSNs In fact optimization algorithms aremore diverse than the types of objective functions but theright choice of the objective function has a much more graveimpact than the specific choice of the optimization algorithmNevertheless the careful choice of the optimization algorithmis also vital especially when complex MOPs are considered

Solving an MOP means finding the PS of the MOP Asmentioned in Section I there are various classes of methods

designed for generating the PSs of MOPs such as math-ematical programming based scalarization methods nature-inspired metaheuristics and so forth It should be noted thatscalarizing an MOP means formulating a single-objectiveoptimization problem whose optimal solutions are also Pareto-optimal solutions to the MOP [15] Additionally it is oftenrequired that every Pareto-optimal solution can be reachedwith the aid of specific parameters of the scalarization

1) Mathematical Programming Based Scalarization Meth-ods Mathematical programming based classic scalarizationmethods conceived for MOO include the linear weighted-

(a) PF of unconstrained MOP

Feasible region

(b) PF of constrained MOP

Fig 9 The PF of MOP with and without constrainsm = 2

sum method theǫ-constraints method [15] and the goalprogramming (GP) based methods [9] [172]ndash[177] as detailedbelow

a) Linear Weighted-Sum MethodThe linear weighted-sum method scalarizes multiple performance metrics into asingle-objective function by pre-multiplying each performancemetric (ie component objective) with a weight Since differ-ent performance metrics have different properties and eachmetric may have a different unit the normalization mustbe implemented firstly when using the linear weighted-summethod for striking compelling performance trade-offs Thena different weight is assigned to each metric to get an eval-uation function Finally the optimal compromise is obtainedaccording to the Pareto-optimal solutions generated by solvingmultiple single-objective problems each corresponding to aspecific vector of weight values It can be proved that theoptimal solution to each of these single-objective problemsis a Pareto-optimal solution to the original multi-objectiveproblem ie the image of these solutions belong to the PF

The linear weighted-sum method is easy to implement andcan avoid complex computations provided that the weightsare appropriately chosen since only a single optimal valuehas to be calculated for each single-objective problem It isworth pointing out that all the weights are in the range[0 1]and the sum of them is1 However there is noa prioricorrespondence between a weight vector and a solution vectorand the linear weighted-sum method usually uses subjectiveweights which often results in poor objectivity and makes the

objectives to be optimized sensitive to the weights The needto solve multiple single-objective optimization problemswiththe aid of different sets of weight values also implies that asubstantial overall computational complexity may be imposedFurthermore the lack of a reasonable weight allocation methoddegrades its scientific acceptance To elaborate a little furthertypically the decision maker isa priori unaware of whichweights are the most appropriate ones to generate a satisfactorysolution hence heshe does not know in general how toadjust the weights to consistently change the solution Thisalso means that it is not easy to develop heuristic algorithmsthat starting from certain weights are capable of iterativelygenerating weight vectors to reach a certain portion of the PFIn addition the linear weighted-sum method is incapable ofreaching the non-convex parts of the PF Finally a uniformspread of the weight values in general does not produce auniform spread of points on the PF This fact implies thatusually all the points are grouped in certain parts of the PFwhile some (potentially significant) portions of the PF basedtrade-off curve have not been reached

Note that if the decision maker hasa priori preferenceamong the multiple objectives considered or heshe would liketo select the most satisfactory solution from the Pareto-optimalsolutions obtained a powerful multiple criteria decision-making method referred to as the analytical hierarchy process(AHP) [178]ndash[182] can be used to determine the relativeweights Using AHP the weights can be flexibly altered ac-cording to the specific application requirements AHP has beenwidely used in the context of trade-off mechanisms (eg [183]and [184]) where AHP first decomposes a complex probleminto a hierarchy of simple subproblems then synthesizes theirimportance to the original problem and finally chooses thebest solution

b) ε-Constraints MethodTheε-constraints method cre-ates a single-objective function where only one of the orig-inal objective functions is optimized while dealing with theremaining objective functions as constraints This methodcanbe expressed as [185]

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

wherefi(x) i = (1 2 N) is the selected function for op-timization and the remainingNminus1 functions act as constraintsIt was proved by Miettinen [186] that if an objectivefj anda vectorε = (ε1 εjminus1 εj+1 εN ) isin R

Nminus1 exist suchthatxlowast is an optimal solution to the above problem thenx

lowast isa weak Pareto optimum of the original MOP Therefore thismethod is capable of obtainingweak Pareto-optimalpointsby varying theε vector but it is not guaranteed to obtainall of them Under certain stronger conditions it can evenobtain the strict Pareto optimum [186] This method is veryintuitive and the parametersεj used as upper bounds are easyto interpret Another advantage of this method is that it iscapable of finding Pareto-optimal solutions on a non-convexPF Similar to the linear weighted-sum method having to

empirically vary the upper boundεj also implies a drawbackof theε-constraints method and it is not particularly efficientif the number of objective functions is higher than two

c) Goal Programming (GP)Instead of maximizing mul-tiple objectives GP is an analytical approach devised forsolving MOPs where the goal values (or targets) have beenassigned to all the objective measures and where the decision-maker is interested in minimizing the ldquonon-achievementrdquoof the corresponding goals In other words the underlyingassumption of GP is that the decision-maker seeks a satis-factory and sufficient solution with the aid of this strategyGP can be regarded as an extension or generalization oflinear programming to handle multiple conflicting objectivemeasures Each of these measures is given a goal or targetvalue to be achieved The sum of undesirable deviations fromthis set of user-specified target values is then minimized withthe aid of a so-called achievement function There are variousforms of achievement functions which largely determine thespecific GP variant The three oldest and still widely usedforms of achievement functions include the weighted-sum(Archimedean) preemptive (lexicographic) and MINMAX(Chebyshev) [9] [172]ndash[177]

There exist other scalarization methods devised for MOOsuch as the conic scalarization method of [176] [187]

2) Nature-Inspired Metaheuristic AlgorithmsMOPs aremore often solved by nature-inspired metaheuristics suchasmulti-objective evolutionary algorithms (MOEAs) [16] [17]and swarm intelligence based optimization algorithms (SIOAs)[18] This is because most classical optimization methodsare based on a limited number of standard forms whichmeans that they have to comply with the particular structuresof objective functions and constraints However in realisticscenarios it is often impossible to accurately characterize thephysical problem with an ideal standard-form optimizationproblem model Additionally many complicated factors suchas a large number of integer variables non-linearities andso forth may occur Both of them can make the realisticproblems hard to solve Therefore the classical mathematicalprogramming based optimization methods may not be suitablefor solving the MOPs encountered in real-world WSNs

Over the most recent decade metaheuristics have madesubstantial progress in approximate search methods for solvingcomplex optimization problems [188] A metaheuristic tech-nique guides a subordinate heuristic using concepts typicallyderived from the biological chemical physical and evensocial sciences as well as from artificial intelligence toimprove the optimization performance Compared to math-ematical programming based methods metaheuristics basedoptimization algorithms are relatively insensitive to thespecificmathematical form of the optimization problems Howeverthe higher the degree of accuracy required the higher thecomputational cost becomes So far the field of metaheuristicsbased optimization algorithms has been mostly constitutedby the family of evolutionary algorithms (EAs) [17] and thefamily of SIOAs [44] In the following we will review someof their salient representatives

a) Evolutionary Algorithms (EAs)EAs belong to thefamily of stochastic search algorithms inspired by the natural

selection and survival of the fittest in the biological world Thegoal of EAs is to search for the globally near-optimal solutionsby repeatedly evaluating the objective functions or fitnessfunc-tions using exploration and exploitation methods Comparedwith mathematical programming EAs are eminently suitablefor solving MOPs because they simultaneously deal with aset of solutions and find a number of Pareto-optimal solutionsin a single run of the algorithm Additionally they are lesssusceptible to the specific shape or to the continuity of the PFand they are also capable of approximating the discontinuousor non-convex PF [189] The most efficient PF-based MOEAshave been demonstrated to be powerful and robust in terms ofsolving MOPs [17]bull Genetic Algorithms (GAs)GAs constitute the most pop-

ular branch of EAs [190] GAs are based on genetics andevolutionary theory and they have been successfully used forsolving diverse optimization problems including MOPs Theyhave appealing advantages over traditional mathematical pro-gramming based algorithms [44] in terms of handling complexproblems and convenience for parallel implementation GAcan deal with all sorts of objective functions no matter theyare stationary or transient linear or nonlinear continuous ordiscontinuous These advantageous properties of GAs haveinspired their employment in solving MOPs of WSNs GAsrely on the bio-inspired processes of initialization evaluationselection crossover mutation and replacement as portrayedin its simplest form in the flow-chart of Fig 10

The MOGA [191] has attracted particularly extensive re-search attention among all the algorithms of MOO By op-erating on the generation-by-generation basis a number ofPareto-optimal solutions can be found throughout the evolutiongenerations Thus obtaining the Pareto-optimal solutionsetprovides us with a set of flexible trade-offs Several solutionmethods based on MOGAs were presented in the literature[112] [192] for optimizing the layout of a WSN Morespecifically the authors of [112] advocated a MOGA for theoptimal deployment of static sensor nodes in a region ofinterest which simultaneously maximized the coverage areaand the networkrsquos lifetime The authors then extended theirwork to three specific surveillance scenarios in [192] usingthe same MOGA Recently the non-dominated sorting geneticalgorithm (NSGA) [193] the niched Pareto genetic algorithm(NPGA) [194] and the SPEA [16] have been recommended asthe most efficient MOEAsbull Differential Evolution (DE)DE was developed by Storn

and Price [195] It is arguably one of the most powerful real-valued optimization algorithms DE relies on similar compu-tational steps as employed by a standard EA It commencesits operation from randomly initiated parameter vectors eachof which (also calledgenomeor chromosome) forms a can-didate solution to the optimization problem Then a mutantvector (known asdonor vector) is obtained by the differentialmutation operation To enhance the potential diversity of thecandidate solutions a crossover operation comes into playafter generating thedonor vectorthrough mutation The finalstep of the algorithm calls for selection in order to determinewhether the target vector survives to the next generation [196]The main stages of the DE algorithm are illustrated in Fig 11

Initialization of search space

Results output

Calculate the fitnessobjective function

values of candidate solutions

Select a pair of candidate solutions which

generate high fitnessobjective function

values

Crossover mix the Genes of the

candidate solutions selected which have

potentially enhanced child solutions

Mutation randomly perturb the child

solutions to further explore the search

Apply elitism to retain the candidate

solutions having the highest fitness

objective function values

Do the results meet the stopping

criterion

Fig 10 Simplified flow-chart of a GA

Initial vectors SelectionCrossoverMutation with

difference vectors

Fig 11 Simplified illustration of a DE algorithm

However unlike traditional EAs the DE algorithm is muchsimpler to implement with only a few parameters to beset Therefore DE has drawn much attention and has beensuccessfully applied in numerous domains of science andengineering (see eg [197] [198])bull Artificial Immune System (AIS)AIS is a computational

intelligence paradigm inspired by the biological immune sys-tem It has been applied to a variety of optimization problemsand has shown several attractive properties that allow EAsto avoid premature convergence and to enhance local search[199] AIS is capable of recognizing and combating pathogensMolecular patterns expressed on those pathogens are referredto as antigens An antigen is any molecule that can berecognized by the immune system and is capable of provokingthe immune response This immune response is specific toeachantigen The cells calledlymphocyteshave a vital rolein the immune system There are two types oflymphocytesB cells and T cells When anantigen is detected B cellsthat best recognize theantigen will proliferate by cloning

TABLE III Pseudocode of Artificial Immune System

Repeat

1 Select an antigenA from population of antigens

2 TakeR antibodies from population of antibodies

3 For each antibodyr isin R

match it against the selected antigenA

4 Find the antibody with the highest match score

break ties randomly and compute its match score

5 Add match score of winning antibody to its fitness

Until the maximum number of cycles is reached

Some of these new cloned cells will differentiate into plasmacells which are the most activeantibody secretors Thesecloning and mutation processes are termed the clonal selectionprinciple [200] which is one of the inspiring methodologiesemployed in AIS for solving optimization problems Basedon the clonal selection principle an algorithm is developedwhere various immune system aspects are taken into accountsuch as the maintenance of the memory cells selection andcloning of the most stimulated cells death of non-stimulatedcells re-selection of the clones with higher affinity as well asthe generation and maintenance of diversity The steps of theAIS are provided in form of pseudocode in Table III

Similar to the computational frameworks of EA AIS canbe readily incorporated into the evolutionary optimization pro-cess and particularly AIS are often exploited in evolutionarytechniques devised for MOO to avoid premature convergenceOn the other hand the main distinction between the field ofAIS and GAs is the nature of population development [201]Specifically the population of GAs is evolved using crossoverand mutation operations However in the AIS similar toevolutionary strategies where reproduction is a cloning eachchild produced by a cell is an exact copy of its parent Bothalgorithms then use mutation to alter the progeny of the cellsand introduce further genetic variations

bull Imperialist Competitive Algorithm (ICA)Inspired by thesocio-political evolution process of imperialism and imperial-istic competition ICA was originally proposed by Atashpaz-Gargari and Lucas in 2007 [202] Similar to other EAs ICAstarts with an initial population with each of them termed acountry The countries can be viewed as population individualsand are basically divided into two groups based on their powerie imperialists (countries with the least cost functionvalue)and colonies After forming initial empires the colonies startmoving toward their relevant imperialist This movement isasimple manifestation of the assimilation policy which is pur-sued by some of the imperialists and results in improvementsof the socio-political characteristics such as culture languageand economical policy in the colonies Then the imperialisticcompetition starts among all the empires The imperialisticcompetition will gradually result in an increase of the powerof stronger empires and a decrease in the power of weakerones In this process weak empires will lose their coloniesand eventually collapse In the long run ICA converges to astate where only a single powerful empire exists in the world

and all the other countries are colonies of that empire3 In thisstate the best solution of the optimization problem is givenwhen all colonies and the corresponding imperialist have thesame cost

ICA has been successfully applied in numerous single-objective optimization problems [203] [204] where mostresults indicate that it is superior to the GA in terms of bothits accuracy and convergence rate The basic structure of themulti-objective imperialist competitive algorithm (MOICA) isthe same as that of the ICA However new methods are devel-oped to determine the imperialist countries to define the powerof the imperialist countries and to calculate the total powerof empires for imperialistic competition Selecting imperialists(best countries) from a set of Pareto-optimal solutions impactsboth the coverage and the diversity of solutions This impactis more significant when the optimization problem has ahigh number of objectives A novel MOICA was proposedin [157] for handling node deployment where both the fastnon-dominated sorting and the Sigma method were employedfor selecting the best countries as imperialists

b) Swarm Intelligence Optimization Algorithms (SIOAs)Swarm intelligence constitutes a branch of artificial intelli-gence (AI) and it exploits the collective behavior of self-organized decentralized systems that rely on a social structuresuch as bird flocks ant colonies and fish schools These sys-tems consist of low-intelligence interacting agents organizedin small societies (also referred to as swarms) exhibitingtraitsof intelligence such as the ability of reacting to environmentalthreats and the decision making capacities Swarm intelligencehas been utilized in the global optimization framework ofcontrolling robotic swarms in [44] Three main SIOAs havebeen developed namely ACO [24] PSO [205] and artificialbee colony (ABC) [206]bull Ant Colony Optimization (ACO)ACO was inspired by

the foraging behavior of some ant species These ants depositpheromones on the ground in order to mark their nest-to-foodpaths that should be followed by other members of the colonyAdditionally they also deposit a different type of pheromoneto mark dangerous paths for the others to avoid any threat TheACO algorithm is capable of solving discretecombinatorialoptimization problems in various engineering domains It wasinitially proposed by Dorigo in [23] [207] and has since beenwidely researched and diversified to solve a class of numericalproblems To illuminate the basic principle of ACO let usconsider the pathsA andB of Fig 12 between a nest and afood source as an example [40] Furthermore let us denoteby nA(t) and nB(t) the number of ants along the pathsAand B at the time stept respectively and byPA and PB

the probability of choosing pathA and pathB respectivelyThen the probability of an ant choosing pathA at the timestept+ 1 is given by

PA(t+ 1) =[c+ nA(t)]

[c+ nA(t)]α + [c+ nB(t)]α= 1minus PB(t+ 1)

3This might not be the case in realistic world since every empire in thehistory has a limited life cycle

where c is the degree of attraction of a hitherto unexploredbranch andα (α ge 0) is the bias towards using pheromonedeposits in the decision process An ant chooses betweenthe path A or path B using the following decision rule ifU(0 1) le PA(t + 1) then choose pathA otherwise choosepath B Here U is a random number having a uniformdistribution in the range of[0 1] ACO performs well in thedynamic and distributed routing problems of WSNsbull Particle Swarm Optimization (PSO)Similar to the under-

lying philosophy of other swarm intelligence approaches PSOaims for mimicking the social behavior of a flock of birds Itconsists of a swarm ofs candidate solutions termed particleswhich explore ann-dimensional hyperspace in search of theglobal solution In PSO the particles regulate their flyingdirections based both on their own flying experience and ontheir neighborsrsquo flying experience [208] After several im-provements conceived by researchers PSO became an often-used population-based optimizer which is capable of solvingstochastic nonlinear optimization problems at an affordablecomplexity The position of theith particle is represented asXid while its velocity is represented asVid Each particleis evaluated through an objective functionf(x1 x2 xn)where we havef Rn minusrarr R The cost (fitness) of a particleclose to the global solution is lower (higher) than that of aparticle being farther away The best position of particlei isdenoted asPid Then the particles are manipulated accordingto the following two equations [208]

Vid(t+ 1) = wVid(t) + λ1r1(t)[Pid(t)minusXi(t)]

+λ2r2(t)[Pd(t)minusXi(t)] (17)

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

where we have1 le i le s and 1 le d le n Additionally λ1

andλ2 are constantsw is the so-called inertia weight whilePd is the position of the best particle Still referring to (17)r1(t) andr2(t) are random numbers uniformly distributed in[0 1] In thetth iteration the velocityV and the positionX areupdated using (17) and (18) The update process is iterativelyrepeated until either an acceptablePd is achieved or a fixednumber of iterationstmax is reached The general frameworkof the multi-objective PSO is shown in Fig 13 which includessome key operations such as the maintenance of the archiveglobal optimum selection as well as the velocity and positionupdate [209] More explicitly the particle population relies onan archive for storing the Pareto-optimal solutions duringtheiterative process and for selecting the global optimum fromthese solutions This is the key point in which the multi-objective PSO is different from the traditional single-objectiveoptimization

The employment of the PSO as a stochastic global optimiza-tion algorithm in the MOP of WSNs is relatively new andhence there is a paucity of contributions A multi-objectiverouting model based on ACO was proposed in [24] whichoptimizes the networkrsquos delay energy consumption and datapacket loss rate This novel method has been shown to becapable of adapting to different service requirements Theauthors of [115] developed a MOO model based on PSO

Fig 12 A stylized optimization process of ACO

Population13

New population13

Final archive13

New archive13

Maintenance of archive13

Archive13

Multi13-13objective function 13calculation13

Velocity and 13position13 update13

Global optimal selection13

Individual optimal selection13

I13t13e13r13a

13t13i13o13n13

Fig 13 The general framework of the multi-objective PSO

and fuzzy logic (FL) for sensor node deployment aiming formaximizing the networkrsquos coverage connectivity and lifetimeThey have shown that the technique provides efficient andaccurate decisions for node deployment in conjunction withlow estimation errorsbull Artificial Bee Colony (ABC)The ABC algorithm was first

introduced by Karaboga and Basturk [206] and it was derivedfrom the behavior of honey bees in nature Since the structureof the algorithm is simple it has been widely used for solvingoptimization problems In the ABC model the position of afood source represents a possible solution to the optimizationproblem and the amount of nectar in a food source correspondsto the quality (fitness) of the associated solution The honeybee swarm consists of three groups of bees namely theemployed bees onlookers and scouts Correspondingly theABC algorithm has three phases [151] [210] It is assumedthat there is only a single artificial employed bee for each food

source Therefore the number of employed bees in the honeybee swarm is equal to the number of food sources around thehive

i) Employed bee phase At the first step the randomlydistributed initial food sources are produced for all employedbees Then each employed bee flies to a food source inits memory and determines a neighbor source whose nectaramount is then evaluated If the nectar amount of the neighborsource is higher than that of the previous source the employedbee memorizes the new source position and forgets the old oneOtherwise it keeps the position of the one in its memory Inother words an employed bee updates the source position inits memory if it discovers a better food source

ii) Onlooker phase After all employed bees have completedthe above food-source search process they return to the hive toshare the position and nectar amount of their individual foodsource with the onlookers Each onlooker evaluates the nectar

information taken from all employed bees and then chooses afood source depending on the nectar amounts of these sourcesTherefore food sources with high nectar content attract a largenumber of onlooker Similar to the case of the employed beean onlooker then updates the source position in its memory bychecking the nectar amount of a neighbor source If its nectaramount is higher than that of the previous one the onlookermemorizes the new position and forgets the old one

iii) Scout phase As a result the sources abandoned havebeen determined and the employed bee whose food source hasbeen abandoned becomes a scout New sources are randomlyproduced by scouts without considering any experience inorder to replace the abandoned ones

The above foraging behavior can be simulated using anABC algorithm to determine the globally optimal solutionsof optimization problems as shown in [210]

c) Artificial Neural Network (ANN)ANN is a sophisti-cated computational intelligence structure inspired by the neu-robiological system It is used for estimating or approximatingfunctions that depend on a large number of inputs that aregenerally unknown [211] The biological neuron consists ofdendrites an axon and a cell body calledsoma Each neuronmay form a connection to another neuron via the synapsewhich is a junction of an axon and a dendrite The so-called postsynaptic potentials generated within the synapsesare received via dendrites and chemically transformed withinthe soma The axon carries away the action potential sent outby the somato the next synapse The analogy of biologicalneurons to artificial neurons is explained as follows In ar-tificial neurons the incoming signals are weighted which isanalogous to what is done in synapses Then the weightedsignals are further processed The functionf(x) is basicallya weighted-sum of all inputs while the output corresponds tothe axon In the context of WSNs the sensor node convertsthe physical signal to an electronic signal which is filteredor preprocessed using weighting (analogous to synapse) Thesubsequent processing within the processor is representedby the particular functionh(x) which corresponds to thechemical processing accomplished by thesoma Eventually asensor node sends out the modified sensor reading via the radiolink This strong analogy shows that the sensor node itself canbe viewed as a biological or artificial neuron [212] Thereforewe can readily extend our horizon and regard some WSNs aslarge-scale ANNs Having said this we are fully aware ofthe inherent dangers of analogies The shift procedure fromabiological neuron to a sensor node is portrayed in Fig 14

The entire sensor network can be modelled from an ANNperspective For each sensor node within the sensor networkwe can also rely on ANNs to decide the output action Thusit is possible to envision a two-layer ANN architecture forWSNs

d) Reinforcement Learning (RL)RL is a powerful math-ematical framework that enables an agent (sensor node) tolearn via interacting with its environment and to model aproblem as a Markov decision process (MDP) [213] The mostwell-known RL technique isQ-learning The visualization ofQ-learning is shown in Fig 15 where an agent (sensor node)regularly updates its achieved reward based on the action taken

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Fig 14 The shift procedure from a biological neuron to asensor node

(sensor node)Environment

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Fig 15 Visualization of theQ-learning

at a given state The future total reward (ie theQ-value) ofperforming an actionat at statest is calculated using

Q(st+1 at+1) = Q(st at) + λ middot [r(st at)minusQ(st at)] (19)

wherer(st at) represents the immediate reward of performingan actionat at statest and 0 le λ le 1 is the learning ratethat determines how fast learning takes place This algorithmcan be easily implemented in a distributed architecture likeWSNs where each node seeks to choose specific actions thatare expected to maximize its long-term rewards For instanceQ-learning has been efficiently used in WSN routing problems[214] [215]

3) Other Advanced Optimization TechniquesThere areseveral other advanced optimization methods capable ofachieving appealing performance trade-offs such as fuzzylogic game theory and so forth Although these methods areless frequently used in WSNs the trade-offs achieved by themcan be compelling

a) Fuzzy Logic (FL)FL as a mathematical model wasintroduced by Zadeh in the 1960s [216] It is a useful techniquethat can use human language to describe inputs as well asoutputs and it provides a simple method of achieving a conclu-sion based on imprecise or ambiguous input information Sincethen the applications of FL have been expanding especiallyin adaptive control systems and system identification

A fuzzy system comprises four basic elements namelyfuzzifier inference engine fuzzy rule base and defuzzifier[217] as shown in Fig 16 The fuzzifier converts the inputsinto fuzzy variables using membership functions each ofwhich represents for each object adegree of belongingness

Fuzzification Defuzzification

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

Fig 16 Typical structure of the FL

to a specific fuzzy set Fuzzy variables provide a mappingof objects to a continuous membership value which is nor-malized in the range[0 1] Each fuzzy set is represented by alinguistic term such as ldquohighrdquo ldquolowrdquo ldquomediumrdquo ldquosmallrdquo andldquolargerdquo The inference engine is often a collection of if-thenrules by which the fuzzy input is mapped to a linguistic outputvariable according to the fuzzy rule base This output variablehas to be converted into a crisp output by the defuzzificationprocess such as the centroid method averaging method rootsum squared method and mean of maximum

Low-complexity FL is suitable for WSNs and various areasof WSNs have been investigated using the rules of FL Forexample the FL-based routing path search for a maximumnetwork lifetime and minimum delay was investigated in [37]where a fuzzy membership function (edge-weight function)was used for formulating a multi-objective cost aggregationfunction which may reflect the effects of all the objectivescollectively as a scalar value As a beneficial result it offers abeneficial trade-off between maximizing the network lifetimeand minimizing the source-to-sink delay

b) Game TheoryGame theory is a powerful mathe-matical tool that characterizes the phenomenon of conflictand cooperation between rational decision-makers [218] Sincegame theory introduces a series of successful mechanismssuch as the pricing mechanism it has achieved a great successin the design of WSNs In particular the pricing schemescan guide the nodesrsquo behaviors towards an efficient Nashequilibrium by introducing a certain degree of coordinationinto a non-cooperative game In [219] a Nash equilibrium-based game model a cooperative coalition game model andan evolutionary game model were used for solving MOPsrespectively

Indeed a number of MOO approaches have appeared inthe literature over the past decade For the sake of claritysome representative MOO algorithms are summarized in TableIV including the MOGA [191] NPGA [194] NSGA [193]NSGA-II [21] SPEA [16] the strength Pareto evolutionaryalgorithm-2 (SPEA2) [220] the multi-objective messy geneticalgorithm (MOMGA) [221] the multi-objective messy geneticalgorithm-II (MOMGA-II) [222] the Bayesian optimizationalgorithm (BOA) [223] the hierarchical Bayesian optimiza-tion algorithm (HBOA) [224] the Pareto archive evolutionstrategy (PAES) [225] the Pareto envelope-based selectionalgorithm (PESA) [226] the Pareto envelope-based selectionalgorithm-II (PESA-II) [227] multi-objective differential evo-lution (MODE) [196] multi-objective evolutionary algorithm

based on decomposition (MOEAD) [228] Additionally thereare some other methods such as the multi-objective geneticlocal search (MOGLS) [229] the multi-objective Tabu search(MOTS) [230] the multi-objective scatter search (MOSS)[231] ACO [24] PSO [205] ABC [151] FL [37] ANN AISgame theory [219] MOICA [157] memetic algorithm (MA)[232] and centralized immune-Voronoi deployment algorithm(CIVA) [116] just to name a few

C Software Tools

At the time of writing numerous software tools are availablefor solving MOPs These software packages are briefly intro-duced in Table V including BENSOLVE [233] the distributedevolutionary algorithms in Python (DEAP) [234] Decisionar-ium [235] D-Sight [236] the graphical user interface designedfor multi-objective optimization (GUIMOO) [237] the intel-ligent decision support system (IDSS) [238] iSIGHT [239]jMetal [240] the multiple objective metaheuristics library inC++ (MOMHLib++) [241] ParadisEO-MOEO [242] [243]SOLVEX [244] and WWW-NIMBUS [245]

V EXISTING L ITERATURE ON USING MOO IN WSNS

The performance metrics presented in Section II entailconflicting objectives eg the coverage versus lifetime andthe energy consumption versus delay etc Therefore it is nec-essary to balance multiple trade-offs efficiently by employingMOO techniques In this section we present an overview ofthe existing contributions that are focused on using MOO inthe context of WSNs

A Coverage-versus-Lifetime Trade-offs

The reasons why the coverage and the lifetime of a WSNconstitute conflicting objectives are given as follows Op-timizing the coverage represents the maximization of theproportion of the adequately monitored area relative to thetotalarea From another perspective the coverage objective desireshaving a ldquospread-outrdquo network layout where sensor nodes areas far apart from each other as possible in order to minimizethe overlap between their sensing disks This results in a largenumber of relay transmissions taking place at the intermediatesensor nodes especially for those communicating directlywiththe base station Hence the depletion of energy at these sensornodes will happen sooner and the network lifetime will thenbe shorter On the other hand in order to get a longer lifetimeall the sensor nodes tend to communicate using as few hops aspossible (or even communicate directly) with the base stationso that their energy is used for their own data transmissionas much as possible This implies a clustered configurationaround the base station with substantial overlap betweensensing disks and yielding a poor coverage performance TableVI shows a summary of the existing major contributions oncoverage-versus-lifetime trade-offs

More specifically Jourdanet al [112] conceived a MOGAfor optimizing the layout of WSNs ie the locations ofnodes by considering both the sensing and communicationconnectivity requirements The algorithm aims for maximizing

TABLE IV Qualitative Comparison of Representative MOO Algorithms

Approach Complexity Convergence Scalability Optimality

linear weighted-summethod

moderate fast limited mathematically guaranteed optimal

ε-constraints method low fast limited mathematically guaranteed optimal

GP moderate fast good mathematically guaranteed optimal

MOGA moderate fast limited empirically very near-optimal

NSGA high slow limited empirically very near-optimal

NSGA-II moderate fast good empirically very near-optimal

NPGA low slow limited empirically very near-optimal

SPEA high fast good empirically very near-optimal

SPEA2 high fast good empirically very near-optimal

PAES moderate fast limited empirically very near-optimal

PESA moderate moderate moderate empirically very near-optimal

PESA-II low moderate good empirically very near-optimal

MOEAD low fast good empirically very near-optimal

MOGLS moderate fast limited empirically very near-optimal

MOMGA high moderate moderate empirically very near-optimal

MOMGA-II low fast good empirically very near-optimal

MOTS moderate slow good near-optimal

MOSS moderate moderate limited near-optimal

MODE high moderate limited empirically very near-optimal

BOA high slow moderate near-optimal

HBOA low moderate limited near-optimal

PSO low slow limited empirically very near-optimal

ACO high moderate good empirically very near-optimal

ABC low fast good empirically very near-optimal

FL low fast limited empirically very near-optimal

ANN low slow good empirically very near-optimal

AIS moderate moderate good near-optimal

MOICA moderate fast good near-optimal

Game Theory moderate low good empirically very near-optimal

MA moderate fast good near-optimal

CIVA low slow good near-optimal

RL low fast good empirically very near-optimal

TABLE V Representative Software Tools

Software Tools(alphabetically)

License Brief Introduction

BENSOLVE open sourceBENSOLVE is a solver for vector linear programs particularly for the subclass of multiple objective linear programswhich is based on Bensonrsquos algorithm and its extensions [Online] Available httpbensolveorg

DEAP open sourceThe DEAP framework is built with the Python programming language that provides the essential glue for assemblingsophisticated evolutionary computation systems [Online] Available httpdeapreadthedocsioenmaster

Decisionarium open sourceDecisionarium is the first public site for interactive multicriteria decision support with tools for individual decisionmaking as well as for group collaboration and negotiation Also Decisionarium offers access to complete e-learningmodules based on the use of the software [Online] Available wwwdecisionariumhutfi

D-Sight open sourceD-Sight developed by Quantin Hayez at the CoDE-SMG laboratory is a relatively new MOO software It offers multipleinteractive and visual tools that help the decision maker tobetter understand and manage MOPs Compared to theprevious software several functional improvements have been implemented in addition to a modern user interface[Online] Available httpacad-sightcom

GUIMOO open sourceGUIMOO is free software for analyzing results in MOPs It provides visualization of approximative PFs and metricsfor quantitative and qualitative performance evaluation[Online] Available httpguimoogforgeinriafr

IDSS open sourceIDSS is a decision support system that makes the extensive use of AI techniques The development of the IDSS softwarepackage is a primary exploration that puts the decision support method into the context of the real-life world [Online]Available httpidsscsputpoznanplsitesoftwarehtml

iSIGHT commercial iSIGHT is a generic software framework for integration automation and optimization of design processeswhich was developed on the foundation of interdigitation tosolve complex problems [Online] Availablehttpwww3dscomproducts-servicessimuliaproductsisight-simulia-execution-engineportfolio

jMetal open sourcejMetal is a an object-oriented Java-based framework for solving MOPs using metaheuristics It is a flexible extensibleand easy-to-use software package [Online] Available httpjmetalsourceforgenet

MOMHLib++ open sourceMOMHLib++ is a library of C++ classes that implements a number of multiple objective metaheuristics Each methodonly needs the local search operation to be implemented [Online] Available httphomegnaorgmomh

ParadisEO-MOEO

open sourceParadisEO-MOEO is a white-box object-oriented software framework dedicated to the reusable design of metaheuristicsfor MOO Technical details on the implementation of evolutionary MOO algorithms under ParadisEO-MOEO can befound on the ParadisEO website [Online] Available httpparadiseogforgeinriafr

SOLVEX open sourceSOLVEX is a FORTRAN library of more than 20 numerical algorithms for solving MOPs We have both the SOLVEXWindows and the SOLVEX DOC versions [Online] Available httpwwwccasrupmaproducthtm

WWW-NIMBUS

open sourceWWW-NIMBUS has been designed to solve differentiable and non-differentiable MOPs subject to nonlinear andlinear constraints with bounds on the variables and it can also accommodate integer variables [Online] Availablehttpnimbusmitjyufi

TABLE VI Coverage-versus-Lifetime Trade-offs

Ref Technical Tasks Optimization Objectives Algorithms Type of Sensors Topology EvaluationMethodology

Scope of Applications

[112] deployment maximize coverage maxi-mize lifetime

MOGA homogeneous-static flat experimental trial satellite or a high-altitude aircraft

MOEAD homogeneous-static flat simulation general-purpose

[115] deployment maximize coverage max-imize connectivity maxi-mize lifetime

hybrid FL and PSOheterogeneous-staticflat simulation general-purpose

[91] data aggregationmaximize coverage maxi-mize lifetime (via minimiz-ing latency)

recursive algorithm homogeneous-static flat simulation densely deployed environment

CIVA homogeneous-mobileflat simulation general-purpose

both the coverage and the lifetime of the network henceyielding a PF from which the network can dynamically chooseits most desired solution Konstantinidiset al [11] [247]considered optimizing both the locations and the transmitpower levels of sensor nodes ie the so-called deploymentand power assignment problem (DPAP) for maximizing thenetwork coverage and lifetime Using the MOEAD of [228]the multi-objective DPAP was decomposed into a set ofscalar subproblems in [11] [247] By extending [11] theauthors further addressed theK-connected DPAP in WSNsfor maximizing the network coverage and lifetime under theK-connectivity constraints by using again the MOEAD ap-proach [246] Furthermore Raniet al [115] proposed a multi-objective PSO and FL based optimization model for sensornode deployment which is based on the maximization of thenetworkrsquos coverage connectivity and lifetime Choiet al [91]proposed a randomizedk-disjoint-sensor selection scheme thattraded off the coverage against the data reporting latencywhile enhancing the attainable energy efficiency dependingon the specific type of applications Additionally a CIVA wasproposed for mobile WSNs in [116] to strike an improvedtrade-off between coverage and lifetime The CIVA comprisestwo phases in the first phase CIVA controls the locations andthe sensing ranges of mobile nodes to maximize the coveragein the second phase CIVA adjusts the transmit power ofactivesleep mobile nodes to minimize the number of activenodes

B Energy-versus-Latency Trade-offs

Indeed minimizing the energy consumption requires trans-mitting the sensed data over reduced distance in each hop Bycontrast minimizing the delay requires minimizing the numberof intermediate forwarders between a source and the sink Thisgoal may be achieved by maximizing the distance betweenany pair of consecutive forwarders Furthermore a reducedsearch space for candidate forwarders yields an unbalanceddistribution of the data forwarding load among nodes thuscausing a non-uniform depletion of their available energy[249] [250] Therefore it is necessary to jointly optimize thenetworkrsquos energy consumption and delay The energy-versus-latency trade-off related issues have been lavishly documentedin various specific WSN scenarios [92] [135]ndash[147] [248][249] [251]ndash[253] Table VII shows a summary of the exist-ing major contributions to energy-consumption-versus-latencytrade-offs

More specifically in [248] the authors studied the energy-latency-density trade-off of WSNs by proposing a topology-and-energy-management scheme which promptly wakes upnodes from a deep sleep state without the need for an ultra-low-power radio As a result the WSN designer can tradethe energy efficiency of this sleep state against the latencyassociated with waking up the node In addition the authorsintegrated their scheme with the classic geographical adap-tive fidelity algorithm to exploit excess network density In[135] Zorzi et al developed the energy-versus-latency trade-offs based on the geographical location of the nodes andproposed a collision avoidance protocol Then Yanget al

[136] designed a node wake-up scheme namely the so-calledldquopipelined tone wake-uprdquo which struck a balance betweenthe energy savings and the end-to-end delay This node wake-up scheme was based on an asynchronous wake-up pipelinewhere the wake-up procedures overlapped with the packettransmissions It used wake-up tones that allowed a highduty-cycle ratio without imposing a large wake-up delay ateach hop Yuet al [137] studied the energy-versus-latencytrade-offs using the so-calleddata aggregation tree4 [254] ina real-time scenario with a specified latency constraint anddeveloped algorithms for minimizing the overall energy dissi-pation of the sensor nodes The authors of [138] presented thefirst work on energy-balanced task allocation in WSNs whereboth the time and the energy costs of the computation andcommunication activities were considered They explored theenergy-versus-latency trade-offs of communication activitiesover thedata aggregation treefor modelling the packet flowin multiple-source single-sink communications A numericalalgorithm was conceived for obtaining the exact optimalsolution and a dynamic programming based approximationalgorithm was also proposed

In [140] Borghiniet al considered the problem of analyzingthe trade-offs between the energy efficiency and the delay forlarge and dense WSNs They used an analytical model whichfacilitated the comparison of the trade-offs in scenarios em-ploying different deployment-phase protocols and presented apair of novel algorithms (ie latency-orientedenergy-orienteddata aggregation tree construction algorithms) which outper-formed the existing ones In [92] Ammariet al investigatedthe energy-versus-delay trade-offs of a WSN by varying thetransmission range Huynhet al [141] proposed a cluster-and-chain based energy-delay-efficient routing protocol for WSNswhere eachk-hop cluster uses both cluster-based and chain-based5 approaches Each communication round consisted of acluster- and chain-formation phase as well as a data transmis-sion phase

Furthermore Moscibrodaet al [142] analyzed the energy-efficiency versus propagation-delay trade-offs by definingaformal model with a particular emphasis on the deploymentphase Specifically the authors presented two new algorithmsone of which is entirely unstructured while the other isbased on clustering Leowet al [143] provided an asymptoticanalysis of the transmission delay and energy dissipation ofa 2D multi-state WSN where the sensor nodes were equallyspaced in a line or in a square grid They also discussed thetransmission delay-energy trade-offs for the case where theenergy transmitted attenuates according to the inverse second-power pathloss law As a further development the authors

4In general data aggregation tree is interpreted as a tree that aggregatesinformation from multiple sources en route to the sink (or recipient) In a tree-based network sensor nodes are organized into a tree wheredata aggregationis performed at intermediate nodes along the tree and a concise representationof the data is transmitted to the root node Tree-based data aggregation issuitable for applications that involve in-network data aggregation

5Note that sensor nodes are distributed into multiple clusters and eachcluster has a cluster head that aggregates all data sent to itby all its membersAfterwards cluster heads form multiple binary chains in which each nodecommunicates with the closest neighbor and takes turns transmitting to thebase station

TABLE VII Energy-versus-Latency Trade-offs

Ref Technical Tasks Optimization Objectives Algorithms Type of Sensors Topology EvaluationMethodol-ogy

[248] topology and en-ergy management

minimize energy consump-tion minimize latency mini-mize network density

a node wake-up basedtopology-and-energy-management algorithm andthe classic geographicaladaptive fidelity algorithm

homogeneous-staticflat simulation general-purpose

[135] localization minimize energy consump-tion minimize latency

a collision avoidance pro-tocol

[136] clustering minimize energy consump-tion minimize latency

a node wake-up schemebased on an asynchronouswake-up pipeline

homogeneous-staticflat withclustering

simulation large-scale WSNs

[137] data aggregation minimize energy consump-tion minimize latency

dynamic programming heterogeneous-staticflat simulation real-time monitoring ormission-critical applications

[138] task allocation minimize energy consump-tion minimize latency

a three-phase heuristic homogeneous-staticflat simulation real-time application

tabu search and ACO heterogeneous-staticflat simulation large-scale WSNs

re-routing algorithms homogeneous-staticflat simulation large and dense WSNs

[92] routing minimize energy consump-tion minimize latency

a data dissemination proto-col

a cluster-and-chain basedenergy-delay-efficient rout-ing protocol

simulation inhospitable physical environ-ments

[142] deployment minimize energy consump-tion minimize latency

uniform algorithm clusteralgorithm

homogeneous-staticflat simulation harsh environments

[143] scheduling minimize energy consump-tion minimize latency

analytical method homogeneous-staticflat analytical abstract multi-state one- andtwo-dimensional line WSN

[144] scheduling andMAC

minimize energy consump-tion minimize latency

hybrid GA and PSO homogeneous-staticflat simulation general-purpose

FL homogeneous-staticflat simulation delay-sensitive WSNs

NSGA-II heterogeneous-staticflat simulation general-purpose

energy-efficient minimum-latency data aggregationalgorithm

[249] [250] data forwarding minimize energy consump-tion uniform battery powerdepletion minimize latency

weighted scale-uniform-unit sum algorithm

homogeneous-staticflat simulation sensing applications

queue theory heterogeneous-statichierarchical simulation general-purpose

[139] data aggregation minimize energy consump-tion minimize latency max-imize lifetime

centralized and distributedheuristics inspired by tech-niques developed for avariant of the vehicle rout-ing problem

heterogeneous-staticflat simulation general-purpose

[252] data aggregationand processing

integer programming homogeneous-staticflat simulation industrial Internet of Things

queue theory homogenous-static flat withclustering

simulation sensing applications

of [144] presented a new MOO framework conceived forslot scheduling in many-to-one sensor networks Two specificoptimization objectives were considered in [144] The firstonewas to minimize the energy consumption while the other wasto shorten the total delay Minhaset al [145] proposed arouting algorithm based on FL for finding a path that offers adesirable balance between the maximum lifetime (associatedwith energy consumption) and the minimum source-to-sinkdelay

In the same spirit Chenget al [146] proposed a MOOframework for cluster-based WSNs The framework was de-signed to strike attractive trade-offs between the energy con-sumption and the duration of the data collection processThe effectiveness of this framework was evaluated with apair of energy-aware clustering algorithms However cluster-ing techniques typically impose bottlenecks during the datacollection process and cause extra delays Liet al [147]investigated the trade-offs of data aggregation in WSNs inthe presence of interference and they conceived an energy-efficient minimum-latency data aggregation algorithm whichachieved the asymptotically minimal aggregation latency aswell as the desired energy-versus-latency trade-offs Ammari[249] [250] proposed a data forwarding protocol for findingthe best trade-offs among minimum energy consumptionuniform battery power depletion and minimum delay whichrelied on slicing the communication range of the nodes intoconcentric circular bands He also conceived a novel approachtermed theweighted scale-uniform-unit sum which was usedby the source nodes for solving this MOP Shahrakiet al[251] defined a new cost function and developed a new intra-cluster routing scheme for balancing the attainable clusterlifetime against the end-to-end delay between the clustermembers and the cluster head In [139] Yaoet al devel-oped a data collection protocol for balancing the trade-offsbetween energy-efficiency (associated with lifetime) and delayin heterogeneous WSNs in which a centralized heuristic wasdevised for reducing the computational cost and a distributedheuristic was conceived for making the algorithm scalableBoth heuristics were inspired by recent techniques developedfor the so-called ldquoopen vehicle routing problems with timedeadlinesrdquo which is mainly studied in operational research In[253] Donget al investigated the trade-offs between energyconsumption and transport latency minimization under certainreliability constraints in WSNs Based on the analysis strategyconceived for satisfying sensing application requirements theyproposed a data gathering protocol named broadcasting com-bined with multi-NACKACK to strike attractive trade-offsIn [252] the authors proposed an energy-efficient and delay-aware wireless computing system for industrial WSNs basedsmart factories

C Lifetime-versus-Application-Performance Trade-offs

In certain sensor network applications the specific applica-tionrsquos performance strongly depends on the amount of datagathered from each sensor node in the network Howeverhigher data rates result in increased sensing and communica-tion costs across the sensor network as well as in escalating

energy consumption and reduced network lifetime [255] Thusthere is an inherent trade-offs between the network lifetimeand a specific applicationrsquos performance while the latter isoften correlated to the rate at which the application canreliably send its data across sensor networks This problemhas been extensively studied in recent years Table VIII showsa summary of the existing major contributions to lifetime-versus-application-performance trade-offs

In [128] Namaet al investigated the trade-offs betweennetwork utility and network lifetime maximization in a WSNThey proposed a general cross-layer optimization-based frame-work that took into account the associated radio resourceallocation issues and designed a distributed algorithm by rely-ing on the so-called dual decomposition [258] of the originalproblem Similarly in [129] Zhuet al studied the trade-offbetween network lifetime (associated with energy conserva-tion) and rate-allocation by using the gradient projection[259]method However no detailed information was provided abouthow to distributively implement this algorithm in the interestof solving the lifetime-versus-rate-allocation trade-off problemin each layer of the open systems interconnection (OSI) modelZhu et al also studied the trade-offs between the networkrsquoslifetime and fair rate allocation in the context of multi-pathrouting sensor networks [107] where they formulated anMOP subject to a set of convex constraints They invokedthe NUM framework [104] and introduced an adjustablefactor to guarantee rate-allocation fairness amongst all sensornodes Chenet al [74] have addressed the utility-versus-lifetime trade-offs with the aid of an optimal flow control inapractical WSN They formulated the problem as a non-linearMOP subjected to certain constraints and introduced auxiliaryvariables for decoupling the individual objectives embeddedin the scalar-valued multi-objective function The conceptof inconsistent coordination price6 was first introduced forbalancing the energy consumption of the sensor node and thegradient projection method [259] was adopted for designingadistributed algorithm that is capable of finding the optimalrateallocation In [130] Heet al focused on the rate allocationproblem in multi-path routing WSNs subjected to time-varyingchannel conditions with two objectives in mind maximizingthe aggregate utility and prolonging the networkrsquos lifetimerespectively They decomposed the optimization problem withthe aid of the classicLagrange dual decomposition[258] andadopted the stochastic quasi-gradient algorithm [259] forsolv-ing the primal-dual problem in a distributed way Luoet al[131] have also carried out a systematic study of the trade-offs between the networkrsquos throughput and lifetime for WSNshaving stationary nodes where the link transmissions werecarefully coordinated to avoid interference The authors useda realistic interference model based on the SINR for modelingthe conflicts to avoid when scheduling the wireless linksrsquotransmissions Their analytical and numerical results providednovel insights into the interplay among the throughput lifetimeand transmit power Xieet al [256] adopted a specific fairnessconcept to analyze the performance degradation experienced

6Inconsistent coordination price can be interpreted as the auxiliary variable(Lagrange multiplier) for coordinating the energy consumption among thesensor nodes in the constrained MOP formulated

TABLE VIII Lifetime-versus-Application-Performance Trade-offs

[128] routing maximize lifetime maximizenetwork utility

subgradient algorithm heterogeneous-statichierarchical simulation self-regulating WSNs

gradient projection algo-rithm

heterogeneous-statichierarchical simulation cross-layer applications

subgradient algorithm homogeneous-staticflat simulation large-scale WSNs

[74] optimal flow control maximize lifetime maximizenetwork utility

heterogeneous-staticflat simulation video technology WSNs

[130] routing maximize lifetime maximizeaggregate utility

stochastic quasi-gradientalgorithm

heterogeneous-statichierarchical simulation online query applications

[131] scheduling and MACmaximize lifetime maximizethroughput

analytic method homogeneous-staticflat simulation general-purpose

[256] MAC routing maximize lifetime maximizethroughput

improved MAC protocol homogeneous-staticflat with clustering simulation information service orientedsensing

distributively extendedprimal-dual algorithm

homogeneous-staticflat simulation streaming video and audio ap-plications

in multirate WSNs and then took into account the trade-offs between the throughput attained and energy consumptionimposed Eventually a multirate-supportive MAC protocolwas proposed for balancing the throughput versus energy con-sumption Liaoet al [257] generalized the NUM model to amultiutility framework using MOO and applied this frameworkto trade off the network utility against the lifetime in WSNsAn extended Lagrange duality method was proposed whichis capable of converging to a selected Pareto-optimal solution

D Trade-offs Related to the Number of Nodes

Intuitively deploying more sensor nodes would improvethe overall event-detection probability of the system albeitat the expense of increasing both the energy consumption anddeployment cost This indicates the trade-offs among multipleconflicting objectives related to the number of nodes TableIXportrays a number of existing contributions to these trade-offsat a glance

To elaborate in [98] a pair of multi-objective metaheuristicalgorithms (MOEA and NSGA-II) have been used for solvingthe WSNrsquos layout problem determining both the numberand the locations of the sensor nodes that formed a WSNso that reliable full coverage of a given sensor field wasachieved Specifically the authors focused their attention onthe energy efficiency of the network as well as on the numberof nodes while the coverage obtained by the network wasconsidered as a constraint In [101] Jiaet al proposed anew coverage control scheme based on an improved NSGA-IIusing an adjustable sensing radius The objective was to findthe most appropriate balance among the conflicting factors ofthe maximum coverage rate the least energy consumption aswell as the minimum number of active nodes As a furtherdevelopment Woehrleet al [99] have invoked the MOEAto identify attractive trade-offs between low deployment-costand highly reliable wireless transmission ie to minimizetransmission failure probability at as low deployment-costas possible Chenget al [158] investigated the trade-offs

between the maximum affordable number of nodes and theminimum duration of the data collection process in a delay-aware data collection network by exploiting the concepts ofPareto-optimality

In [102] Rajagopalan employed the evolutionary multi-objective crowding algorithm (EMOCA) for solving the sensorplacement problem There were three objectives maximizingthe probability of global target detection minimizing thetotal energy dissipated by the sensor network and minimiz-ing the total number of nodes to be deployed The MOOapproach simultaneously optimized the three objectives andobtained multiple Pareto-optimal solutions In [103] Aitsaadiet al considered a multi-objective combinatorial optimizationproblem where a new multi-objective deployment algorithm(MODA) was proposed The optimization objective was toreduce the number of deployed nodes to satisfy the targetquality of monitoring to guarantee the networkrsquos connectivityand finally to maximize the networkrsquos lifetime In [100] LeBerre et al formulated an MOP of maximizing three objec-tives The first objective was the maximization of the coveragearea in real time the second objective was the maximizationof the networkrsquos lifetime depending on the coverage and thefinal objective was to minimize the number of deployed nodessubject to the connectivity on the network The solutions foundby three different algorithms (ie NSGA-II SPEA2 and ACO)were compared Recently a novel MOICA was proposed forsensor node deployment in [157] where the minimization ofthe number of active sensor nodes and the maximization ofthe coverage were jointly considered The numerical results of[157] demonstrated that the MOICA was capable of providingmore-accurate solutions at a lower computational complexitythan the existing methods

E Reliability-Related Trade-offs

The main objective behind the deployment of WSNs is tocapture and transmit pictures videos and other important datato the sink reliably These applications require us to maintain

TABLE IX Trade-offs Related to the Number of Sensor Nodes

[98] deployment minimize number of nodesminimize energy consumption

MOEA NSGA-II heterogeneous-staticflat experimental trial complex and real WSNs

[101] coverage controlminimize number of nodesminimize energy consumption

NSGA-II heterogeneous-statichierarchical simulation event detection

[99] deployment minimize number of nodesguarantee high transmissionreliability

MOEA homogeneous-staticflat simulation general-purpose

[158] data aggregationmaximize number of nodesminimize latency

analytic method homogeneous-staticflat simulation time-sensitive applications

EMOCA homogeneous-staticflat simulation event detection

[103] deployment minimize number of nodesguarantee network connectiv-ity maximize lifetime

MODA homogeneous-staticflat simulation forest fire detection

[100] deployment minimize number ofnodes maximize coveragemaximize lifetime

NSGA-II SPEA2 ACO homogeneous-staticflat experimental trial general-purpose

[157] deployment minimize number of nodesmaximize coverage

MOICA homogeneous-staticflat simulation densely deployed environment

a strict QoS guarantee [260] [261] However maintaining theQoS during routing hinges on numerous factors such as theenergy status of the nodes in the network the delay the band-width and the reliability requirements Hence sophisticatedrouting protocols have to take into considerations multiplepotentially conflicting factors which makes the problem evenmore challenging Table X shows a summary of the existingcontributions to reliability-related trade-offs

To expound a little further Milleret al [152] studied thetrade-offs amongst the energy latency and reliability Theyconceived a meritorious probability-based broadcast forward-ing scheme for minimizing both the energy usage and thelatency whilst improving the reliability EkbataniFardet al[262] have developed a QoS-based energy-aware routing pro-tocol for a two-tier WSN from the perspective of MOO Theproposed protocol utilizing the NSGA-II efficiently optimizedthe QoS parameters formulated in terms of the reliabilityand end-to-end delay whilst reducing the average powerconsumption of the nodes which substantially extended thelifetime of the network

A high data rate can be maintained by a link at the expenseof a reduced delivery reliability andor increased energyconsumption which in turn reduces the network lifetimeAgain there is an inherent trade-off among the data ratereliability and network lifetime Although numerous treatiseshave extensively studied the data rate reliability and networklifetime in isolation only a few of them have consideredthe trade-offs among them Xuet al [263] jointly consid-ered the rate reliability and network lifetime in a rigorousframework They addressed the optimal rate-reliability-lifetimetrade-offs under a specific link capacity constraint reliabilityconstraint and energy constraint However the optimizationformulation was neither separable nor convex Hence a seriesof transformations have been invoked and then a separa-ble and convex problem was derived Finally an efficient

distributed subgradient dual decomposition (SDD) algorithmwas developed for striking an appealing trade-off In [8] Luet al formulated WSN routing as a fuzzy random multi-objective optimization (FRMOO) problem which simultane-ously considered the multiple objectives of delay reliabilityenergy delay jitter the interference aspects and the energybalance of a path They introduced a fuzzy random variablefor characterizing the link delay link reliability and thenodesrsquoresidual energy with the objective of accurately reflecting therandom characteristics in WSN routing Eventually a hybridrouting algorithm based on FRMOO was designed In [264]Razzaqueet al proposed a QoS-aware routing protocol forbody sensor networks in which a lexicographic optimizationapproach was used for trading off the QoS requirements andenergy costs In [151] Lanza-Gutierrezet al considered thedeployment of energy harvesting relay nodes for resolvingthe conflict among average energy cost average sensing areaand network reliability Two multi-objective metaheuristicsie the ABC algorithm and the firefly algorithm (FA) wereapplied for solving the problem respectively Ansariet al[265] considered the energy consumption reliability coverageintensity and end-to-end delay trade-offs based on the locationof the nodes and a new multi-mode switching protocol wasadopted Liuet al [266] proposed an energy-efficient cooper-ative spectrum sensing scheme for a cognitive WSN by takinginto account the energy consumption and the spectrum sensingperformance both of which were jointly optimized using fastMODE Xiao et al [267] proposed a time-sensitive utilitymodel for low-duty-cycle WSNs where they simultaneouslytook into account the transmission cost utility reliability andlatency Moreover they designed two optimal time-sensitiveutility-based routing algorithms to strike the most appropriatebalance among these four metrics

TABLE X Reliability-Related Trade-offs

energy-latency-reliabilitytrade-off

a probability-based broad-cast forwarding scheme

[262] routing latency-reliability trade-off NSGA-II heterogeneous-statichierarchical simulation real time audio-visual applications

[263] flow control rate-reliability-lifetime trade-off

stochastic subgradient al-gorithm

heterogeneous-statichierarchical simulation WSNs with time-varying channel

[8] routing latency-reliability-energytrade-off

hybrid FRMOO and GA heterogeneous-statichierarchical simulation agriculture surveillance and build-ing monitoring

[264] routing reliability-energy trade-off lexicographic optimizationapproach

homogeneous-staticflat simulation human body location

[151] deployment energy-reliability-sensing areatrade-off

ABC and FA homogeneous-staticflat simulation intensive agriculture

[265] deployment energy-reliability-coverage-latency trade-off

multi-mode switching pro-tocol

[266] spectrum sensingenergy-reliability trade-off fast MODE homogeneous-staticflat simulation cognitive WSNs

[267] routing energy-utility-reliability-latency trade-off

time-sensitive utility-basedrouting algorithm

F Trade-offs Related to Other Metrics

As mentioned in Section I in practice it is unfeasible tojointly satisfy the optimum of several potentially conflict-ing objectives To circumvent this dilemma the concept ofPareto-optimalhas been widely invoked resulting in a PFgenerated by all Pareto-optimal solutions of a MOP whereit is impossible to improve any of the objectives withoutdegrading one or several of the others Therefore accordingto the needs of decision makers and the actual situation ofthe WSN considered efficient routing algorithms are requiredfor finding a satisfactory path in WSNs Table XI summarizesother metrics and their trade-offs

As seen in Table XI Lozano-Garzonet al [268] proposeda distributedN -to-1 multi-path routing scheme for a WSNby taking into account the number of hops the energy con-sumption and the free space loss7 and these three objectiveswere optimized by the SPEA2 [220] with the aim of using theenergy efficiently in the network whilst reducing the packet-loss rate Bandyppadhyayet al [269] proposed a transmissionscheduling scheme using a collision-free protocol for gatheringsensor data Moreover they studied diverse trade-offs amongstthe energy usage the sensor density and the temporalspatialsampling rates As a further advance Rajagopalanet al [84]developed a MOO framework for mobile agent routing inWSNs The multi-objective evolutionary optimization algo-rithms EMOCA and NSGA-II were employed to find the mo-bile agentsrsquo routes aiming for maximizing the total detectedsignal energy while minimizing the energy consumption byreducing the hop-length In [24] Weiet al established amulti-objective routing model that relies on the delay en-ergy consumption data packet-loss rate as its optimizationobjectives By adjusting the specific weight of each functionthe algorithm adapts well to various services having different

7Note that the concept of free space loss is defined as the ratioof the powerradiated by the transmitting antenna over that picked up by the recipient infree space conditions Free space loss is the basic propagation loss

energy cost delay and packet-loss rate requirements Thisprotocol was implemented using an advanced ACO algorithmthat is based on a cloud model8

For a typical WSN the accuracy of an application andthe longevity of the network are inversely proportional toeach other which is partially due to the finite energy re-serves of the nodes and owing to the desire for applicationsto have large volumes of fresh data to process Adlakhaet al [166] created a four-dimensional design space basedon four independent QoS parameters namely the accuracydelay energy consumption and the node density In order toachieve an improved accuracy or lifetime various parametersof the individual techniques can be adjusted The insights andrelationships identified in [166] were not unique to mobilitytracking applications many potential applications of WSNsrequiring a balance amongst the factors of energy consump-tion node density latency and accuracy may also benefitfrom exploiting the results and trends identified in [166] Forinstance the energy-density-latency-accuracy (EDLA) trade-offs have been studied in the context of WSNs in [167]By contrast Armeniaet al [110] introduced a Markov-based modeling of the random routing behavior for evaluatingthe trade-offs between location privacy and energy efficiencyin a WSN Notably their approach used the informationtheoretic concept ofprivacy lossBoth the network securityand lifetime have been studied by Liuet al in [111] wherethey proposed a three-phase routing scheme which is termedsecurity and energy-efficient disjoint routing Based on thesecret-sharing algorithm this routing scheme dispersively andrandomly delivered its source-information to the sink node

8In contrast to the ldquocloudrdquo concept related to cloud computing and cloud-based networking herein the ldquocloud modelrdquo represents an effective tooldesigned for characterizing the uncertain transformationbetween a qualita-tive concept which is expressed by natural language and its quantitativeexpression It mainly reflects two kinds of uncertainty such as fuzziness andrandomness of the qualitative concept As a reflection of therandomness andfuzziness the cloud model constructs a mapping from qualities to quantities

ensuring that the network security was maximized withoutdegrading the lifetime of WSNs As a further developmentTang et al [270] proposed a geography-based cost-awaresecure routing protocol to address the conflicting lifetime-versus-security trade-off issue in multi-hop WSNs The designgoal was achieved by controlling energy deployment balanceand invoking a random walking routing strategy Atteaet al[271] studied the MOP of how to optimally divide sensornodes into multiple disjoint subsets so that two conflictingobjectives namely the network lifetime and coverage prob-ability can be jointly maximized Each subset of sensors isrequired to completely cover a set of targets having knownlocations Hence they formulated a multi-objective disjointset cover (DSC) problem which was tackled by MOEADand NSGA-II In [272] Senguptaet al employed a novelheuristic algorithm termed MOEAD with fuzzy dominance(MOEADFD) for finding the best trade-offs among coverageenergy consumption lifetime and the number of nodes whilemaintaining the connectivity between each sensor node and thesink node In [273] Wanget al quantified the probabilisticperformance trade-offs among the network lifetime end-to-end communication delay and network throughput in real-timeWSNs A heuristic-based multiple-local-search techniquewasemployed for finding the solutions Inspired by the concept ofpotential field from the discipline of physics Zhanget al[274] designed a novel potential-based routing algorithmknown as the integrity and delay differentiated routing forWSNs The objective was to improve the data fidelity forhigh-integrity applications and to reduce the end-to-end delaysimultaneously

VI OPEN PROBLEMS AND DISCUSSIONS

Despite the increasing attention paid to the MOO of WSNsthis research area still has numerous open facets for futurework as discussed below

Most of the studies investigated the MOO of single-hoptransmission whereas only a limited amount of contributionssuch as [275] paid attention to multi-hop WSNs Clearlymulti-hop transmission in energy-limited WSNs is essentialfor conserving transmission energy and thus for prolongingthe networkrsquos lifetime Therefore it is promising to intensifythe research of MOO in the context of multi-hop WSNs

Sensor nodes may move from one place to another asrequired by the application or may be displaced by objects(human animalsetc) Hence the mobility of the nodes has asubstantial impact on the networkrsquos connectivity For exampleif a data packet is long and the node changes its currentlocation during the packetrsquos forwarding part of the data maybe lost at the receiving node Similarly when a node selectsa routing path but the nodes in the routing path changetheir locations the connectivity between the source nodesanddestination nodes will be affected Therefore the deploymentof nodes in highly dynamic scenarios requires a deploymentapproach that equips the network with a self-organizing capa-

bility Artificial potential field (APF) techniques9 have beenapplied to the problems of formation control and obstacleavoidance in multi-robot systems [277] Since these problemsare of similar nature to the deployment problem of sensornodes the APF techniques may also be used to devise adeployment approach for WSNs

Since WSNs are typically deployed in physically open andpossibly hostile environments they will be confronted withsecurity attacks ranging from passive attacks active attacksand denial-of-service (DoS) attacks Therefore their securityis one of the imperative aspects in future research Sincethe security may gravely affect the network performanceespecially during the information exchange phase designing asecure routing protocol for WSNs is a must Most of the knownrouting algorithms assume that the nodes are static and relyon a single path Once the routing is attacked the networkrsquosperformance will be significantly degraded Therefore it isnecessary to study multi-route protocols conceived for mobile-node based routing capable of satisfying the security andQoS requirements in any real-time application The trade-offsbetween the security and QoS requirements will bring aboutfurther new challenges

Existing contributions often assume that the sensor networksare spread across a two-dimensional plane but in practice theyare indeed of three-dimensional (3D) nature The extensionofa 2D network into 3D is both interesting and challengingIn two-tier WSNs multiple objectives have to be satisfiedby the routing algorithms The authors of [8] proposed arouting solution based on thefuzzy random expected valuemodel and the standard deviation model of [278]10 to meet therequirements of different applications of the clustered networkadvocated Since the fuzzy random expected value modelmay become inaccurate in uncertain environments improvedrouting model based on MOO is necessitated Moreover dueto the limitation of GAs the distributed solving methods basedon local information and on the decomposition theory areexpected to be further investigated

Serious natural disasters such as sandstorms tsunamislandslides etc have routinely damaged the natural environ-ment and inflicted the loss of human lives Although WSNsprovide a promising solution to realize real-time environmentmonitoring numerous issues have to be resolved for theirpractical implementation One of the major issues is howto effectively deploy WSNs to guarantee large-area sensingcoverage and reliable communication connectivity in hostilepropagation scenarios

In recent years considering the similarity between multi-objective design and game theory the latter has also been em-

9The APF techniques mainly rely on force vectors associatedwith theobstacles or target positions which may be linear or tangential and aregenerated by a potential functions The concept of APF can beschematicallydescribed as ldquothe manipulator moves in a field of forces the position to bereached is an attractive pole for the end effector and obstacles are repulsivesurfaces for the manipulator partsrdquo It has been widely usedfor mobile robots[276]

10In fact both random uncertainty and fuzzy uncertainty simultaneouslyexist in link quality and nodesrsquo residual energy From the perspective ofstatistics fuzzy random expected value reflects the average value of a fuzzyrandom variable while the standard deviation reflects the degree measure ofdeviating from the expected value

TABLE XI Trade-offs Related to Other Metrics

[268] routing minimize energy consump-tion minimize packet lossminimize hop count

SPEA2 homogeneous-static flat experimentaltrial

general-purpose

[269] scheduling density-energy-throughput-delay-temporal samplingrates-spatial sampling ratestrade-off

analytical method homogeneous-static flat with clustering simulation general-purpose

[84] routing maximize detection accuracyminimize energy consump-tion minimize path loss

EMOCA NSGA-II heterogeneous-mobile hierarchical simulation general-purpose

[24] routing minimize energy consump-tion minimize latency mini-mize packet loss

improved ACO heterogeneous-mobile flat simulation large-scale WSNs

[166] deployment accuracy-delay-energy-density trade-off

analytical method homogeneous-static flat simulation general-purpose

[167] target tracking energy-density-latency-accuracy trade-off

na homogeneous static andmobile

flat simulation adaptive mobility tracking

[110] security privacy loss and energy effi-ciency trade-off

analytical method homogeneous-static flat simulation data mining systems

[111] data aggregationmaximize network securitymaximize lifetime

a security and energy-efficient disjoint routingscheme

homogeneous-static flat simulation densely deployed environment

[270] routing maximize network securitymaximize lifetime

a cost-aware secure rout-ing protocol

homogeneous-static flat simulation general-purpose

[271] scheduling maximize coverage probabil-ity maximize network life-time

the multi-objective DSCproblem formulated wassolved using MOEADand NSGA-II

homogenous-static flat simulation large-scale surveillance applications

[272] deployment maximize coverage minimizeenergy consumption maxi-mize lifetime and minimizethe number of nodes

MOEADFD homogenous-static flat simulation general-purpose

[273] deployment maximize lifetime maximizethroughput and minimize la-tency

heuristic-based multiple-local-search

homogenous-static flat simulation general-purpose

[274] routing maximize data fidelity andminimize latency

an integrity and delaydifferentiated routing al-gorithm

homogenous-static flat simulation integrity-sensitive applications

ployed to solve multi-objective design problems By analogym-objective designs can be regarded asm-player games Theauthors of [279] introduced game theory and the concept ofco-evolution into GAs for the sake of solving the MOPs whichhas been shown to perform well In [219] a Nash-equilibriumbased game model a cooperative coalition game model and anevolutionary game model were used for solving MOP SinceEAs are capable of finding the global solution of MOPs withgood robustness while Nash games can be used for conflictresolution and Stackelberg games for hierarchical designitis promising to solve MOPs of WSNs by combining a Nashgame with EAs or combining a Stackelberg game with EAsAs an adaptive parameter control method based on sensitivityresults the Pascoletti-Serafini scalarization method [280] isa more general formulation relying on an unrestricted searchdirection and an auxiliary vector variable It has been usedfor both linear and nonlinear MOPs [280] Naturally thismethod can also be applied to solve MOPs in WSNs yieldingapproximate solutions of the problems considered Indeedforthe MOP of multicell networks [281] the relationship between

the parameters and the optimal solutions was elucidated by thePascoletti-Serafini scalarization

The mutually interfering networks are ubiquitous hencefinding innovative cross-layer and cross-network solutionsbecomes essential To this end we believe that many hy-brid computational intelligence algorithms which combinethebenefits of two or more algorithms should be given carefulattention such as swarm-FL control neuro-FL control GA-PSO GA-ANN and neuro-immune systems etc

Cognitive radio (CR) is an emerging wireless communica-tion paradigm in which the transceivers are capable of intelli-gently detecting in their vicinity which specific communicationchannels are in use and which are not Then they promptlyswitch to vacant channels while avoiding occupied ones Thisis essentially a form of dynamic spectrum access (DSA)[282] which may substantially improve the exploitation oftheavailable wireless spectrum Typically a transceiver in CR maybe capable of determining its geographic location identifyingand authorizing its users sensing neighboring wireless devicesand automatically adjusting its transmission and reception

parameters to allow more concurrent wireless communicationsin a given spectrum band at a specific location Dependingon which parts of spectrum are available for the operationof CR we have CR operating either in licensed bands orin unlicensed bands On the other hand WSNs often use theunlicensed ISM band for communications but with the rapidlyincreasing demand of the Internet of Things (IoT) basedapplications (eg healthcare and tele-medicine) the currentlyavailable ISM band may become insufficient which can resultin various technical problems such as unreliable transmissionof useful data Therefore in order to alleviate this ldquospectrumcrunchrdquo an emerging trend in WSNs is to equip the wirelesssensor nodes with the CR based DSA capability thus givingbirth to CR aided WSNs (CR-WSNs) [283]ndash[285] Due to itspotential advantages CR-WSN might be a promising solutionfor some specific WSN applications such as indoor sensingmulticlass heterogeneous sensing and real-time surveillance[283] Additionally WBAN which is a promising technologyfor ubiquitous health monitoring systems is also an applica-tion area of CR-WSN In general CR-WSNs constitute an un-explored field with only a handful of studies More specificallythe authors of [286] determined the optimal packet size thatmaximizes the energy-efficiency of a practical realizationofa CR-WSN In [287] the authors proposed a spectrum-awareclustering protocol to address the event-to-sink communicationcoordination issue in mobile CR-WSNs A cross-layer frame-work that employed CR to circumvent the hostile propagationconditions for the smart grid was discussed in [288] and theMAC-layer delay of a cognitive sensor node was modeledin [289] The realization of CR-WSN primarily requires anefficient spectrum management framework for regulating theDSA of densely deployed resource-constrained sensor nodesTherefore MOO techniques invoked for designing CR-WSNsshould be sufficiently intelligent to differentiate between thetraffic types and to satisfy their QoS requirements The currentresearch efforts on MOPs for large-scale CR-WSNs have tobe strengthened

WSNs play a key role in creating a highly reliable andself-healing smart electric power grid that rapidly respondsto online events with appropriate actions However due tothe broadcast nature of radio propagation and varying spectralcharacteristics establishing a secure and robust low-powersmart grid over the WSN must be addressed Other technicalchallenges of WSNs in the smart grid include harsh environ-mental conditions tight reliability and latency requirementsas well as low packet errors and variable link capacity Untilnow there have been only a few approaches available andmore studies are needed in these areas

VII C ONCLUSIONS

In this paper we have provided a tutorial and survey ofthe research of MOO in the context of WSNs We commencewith the rudimentary concepts of WSNs and the optimizationobjectives in WSNs then focus on illuminating the family ofalgorithms for solving MOPs Since having multiple objectivesin a problem gives rise to a set of Pareto-optimal solutionsinstead of a single globally optimal solution none of these

Pareto-optimal solutions can be considered to be better thanthe others on the Pareto front without any further informationThus the MOO algorithms may be invoked for finding as manyPareto-optimal solutions as possible Additionally diverse de-sign trade-offs relying both on classical optimization methodsand on the recent advances of MOO have been reviewed inthe context of WSNs Future research directions on MOOconceived for WSNs include multi-hop transmissions the de-ployment of nodes in highly dynamic scenarios secure multi-path routing protocols and solving optimization problems in3D networks CR-WSNs and smart grid

REFERENCES

[1] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoIEEE Communications Magazine vol 40no 8 pp 102ndash114 Aug 2002

[2] J Yick B Mukherjee and D Ghosal ldquoWireless sensor networksurveyrdquo Computer Networks vol 52 no 12 pp 2292ndash2330 Aug2008

[3] D Bruckner C Picus R Velik W Herzner and G Zucker ldquoHierarchi-cal semantic processing architecture for smart sensors in surveillancenetworksrdquo IEEE Transactions on Industrial Informatics vol 8 no 2pp 291ndash301 May 2012

[4] H Yetgin K T K Cheung M El-Hajjar and L HanzoldquoNetwork-lifetime maximization of wireless sensor networksrdquoIEEE Access vol 3 pp 2191ndash2226 2015 [Online] Availablehttpieeexploreieeeorgstampstampjsptp=amparnumber=7322190

[5] G Han J Jiang N Bao L Wan and M Guizani ldquoRouting proto-cols for underwater wireless sensor networksrdquoIEEE CommunicationsMagazine vol 53 no 11 pp 72ndash78 Nov 2015

[6] F Wang and J Liu ldquoNetworked wireless sensor data collectionIssues challenges and approachesrdquoIEEE Communications Surveysand Tutorials vol 13 no 4 pp 673ndash687 Fourth Quarter 2011

[7] L Cheng J Niu J Cao S K Das and Y Gu ldquoQoS aware geographicopportunistic routing in wireless sensor networksrdquoIEEE Transactionson Parallel and Distributed Systems vol 25 no 7 pp 1864ndash1875Jul 2014

[8] J Lu X Wang L Zhang and X Zhao ldquoFuzzy random multi-objective optimization based routing for wireless sensor networksrdquoSoftComputing vol 18 no 5 pp 981ndash994 May 2014

[9] R Marler and J Arora ldquoSurvey of multi-objective optimization meth-ods for engineeringrdquoStructural and Multidisciplinary Optimizationvol 26 no 6 pp 369ndash395 Apr 2004

[10] R Tharmarasa T Kirubarajan J Peng and T Lang ldquoOptimization-based dynamic sensor management for distributed multitarget track-ingrdquo IEEE Transactions on Systems Man and Cybernetics Part C(Applications and Reviews) vol 39 no 5 pp 534ndash546 Sep 2013

[11] A Konstantinidis K Yang Q Zhang and D Zeinalipour-Yazti ldquoAmulti-objective evolutionary algorithm for the deployment and powerassignment problem in wireless sensor networksrdquoComputer Networksvol 54 no 6 pp 960ndash976 Apr 2010

[12] B S P Reddy and C S P Rao ldquoA hybrid multi-objectiveGA forsimultaneous scheduling of machines and AGVs in FMSrdquoInternationalJournal of Advanced Manufacturing Technology vol 5 no 6 pp 602ndash613 Dec 2006

[13] C A C Coello G Toscano and M Salazar ldquoHandling multipleobjectives with particle swarm optimizationrdquoIEEE Transactions onEvolutionary Computation vol 8 no 3 pp 256ndash279 Jun 2004

[14] H Li and Q Zhang ldquoMultiobjective optimization problems withcomplicated Pareto sets MOEAD and NSGA-IIrdquoIEEE Transactionson Evolutionary Computation vol 13 no 2 pp 284ndash302 Sep 2009

[15] M Ehrgott and M M Wiecek ldquoMultiobjective programmingrdquo inMultiple Criteria Decision Analysis State of the Art Surveys serInternational Series in Operations Research amp Management ScienceJ Figueira S Greco and M Ehrgott Eds Springer New York Oct2005 vol 78 pp 667ndash708

[16] E Zitzler and L Thiele ldquoMultiobjective evolutionary algorithmsA comparative case study and the strength Pareto approachrdquoIEEETransactions on Evolutionary Computation vol 3 no 4 pp 257ndash271Nov 1999

[17] K S N Ripon C-H Sang and S Kwong ldquoMulti-objective evolu-tionary job-shop scheduling using jumping genes genetic algorithmrdquoin Proc IEEE International Joint Conference on Neural Networks(IJCNNrsquo06) Vancouver Canada Jul 2006 pp 3100ndash3107

[18] Z Zhang K Long J Wang and F Dressler ldquoOn swarm intelligenceinspired self-organized networking Its bionic mechanisms designingprinciples and optimization approachesrdquoIEEE Communications Sur-veys and Tutorials vol 16 no 1 pp 513ndash537 First Quarter 2014

[19] A Zhou B Y Qu H Li S Z Zhao P N Suganthan and QZhangldquoMultiobjective evolutionary algorithms A survey of the state of theartrdquo Swarm and Evolutionary Computation vol 1 no 1 pp 32ndash49Mar 2011

[20] K C Tan T H Lee and E F Khor ldquoEvolutionary algorithms formulti-objective optimization Performance assessments and compar-isonsrdquoArtificial Intelligence Review vol 17 no 4 pp 251ndash290 Jun2002

[21] K Deb A Pratap S Agarwal and T Meyarivan ldquoA fast and elitistmultiobjective genetic algorithm NSGA IIrdquoIEEE Transactions onEvolutionary Computation vol 6 no 2 pp 182ndash197 Apr 2002

[22] C A C Coello ldquoAn updated survey of GA-based multiobjectiveoptimization techniquesrdquoACM Computing Surveys vol 32 no 2 pp109ndash143 Jun 2000

[23] M Dorigo and G D Caro ldquoAnt colony optimization A newmeta-heuristicrdquo in Proc IEEE Congress on Evolutionary Computation(CECrsquo99) Washington USA Jul 1999 pp 1470ndash1477

[24] X Wei and L Zhi ldquoThe multi-objective routing optimization of WSNsbased on an improved ant colony algorithmrdquo inProc 6th IEEEInternational Conference on Wireless Communications Networking andMobile Computing (WiCOMrsquo10) Chengdu China Sep 2010 pp 1ndash4

[25] G Anastasi M Conti M D Francesco and A Passarella ldquoEnergyconservation in wireless sensor networks A surveyrdquoAd hoc networksvol 7 no 3 pp 537ndash568 May 2009

[26] S Ehsan and B Hamdaoui ldquoA survey on energy-efficient routing tech-niques with QoS assurances for wireless multimedia sensor networksrdquoIEEE Communications Surveys and Tutorials vol 14 no 2 pp 265ndash278 Second Quarter 2012

[27] C Sergiou P Antoniou and V Vassiliou ldquoA comprehensive surveyof congestion control protocols in wireless sensor networksrdquo IEEECommunications Surveys and Tutorials vol 16 no 4 pp 1839ndash1859Fourth Quarter 2014

[28] P Huang L Xiao S Soltani M W Mutka and X Ning ldquoTheevolution of MAC protocols in wireless sensor networks A surveyrdquoIEEE Communications Surveys and Tutorials vol 15 no 1 pp 101ndash120 First Quarter 2013

[29] N Li N Zhang S K Das and B Thuraisingham ldquoPrivacy preser-vation in wireless sensor networks A state-of-the-art surveyrdquo Ad HocNetworks vol 7 no 8 pp 1501ndash1514 Nov 2009

[30] M Erol-Kantarci H T Mouftah and S Oktug ldquoA surveyof ar-chitectures and localization techniques for underwater acoustic sensornetworksrdquoIEEE Communications Surveys and Tutorials vol 13 no 3pp 487ndash502 Third Quarter 2011

[31] Y Zeng J Cao J Hong S Zhang and L Xie ldquoSecure localizationand location verification in wireless sensor networks A surveyrdquo TheJournal of Supercomputing vol 64 no 3 pp 685ndash701 Jun 2013

[32] I Al-Anbagi M Erol-Kantarci and H T Mouftah ldquoA survey on cross-layer quality-of-service approaches in WSNs for delay and reliability-aware applicationsrdquoIEEE Communications Surveys and Tutorialsvol 18 no 1 pp 525ndash552 First Quarter 2016

[33] Y Gu F Ren Y Ji and J Li ldquoThe evolution of sink mobility manage-ment in wireless sensor networks A surveyrdquoIEEE CommunicationsSurveys and Tutorials vol 18 no 1 pp 507ndash524 First Quarter 2016

[34] I Khan F Belqasmi R Glitho N Crespi M Morrow and P PolakosldquoWireless sensor network virtualization A surveyrdquoIEEE Communica-tions Surveys and Tutorials vol 18 no 1 pp 553ndash576 First Quarter2016

[35] X Wang C Zhang L Gao and P Li ldquoA survey and future trend ofstudy on multi-objective schedulingrdquo inProc 4th IEEE InternationalConference on Natural Computation (ICNCrsquo08) Jinan China Oct2008 pp 382ndash391

[36] M Marks ldquoA survey of multi-objective deployment in wireless sensornetworksrdquoJournal of Telecommunication and Information Technologyvol 3 no 3 pp 36ndash41 Mar 2010

[37] T Gao J Y Song J Y Zou J H Ding D Q Wang and R C JinldquoAn overview of performance trade-off mechanisms in routing protocolfor green wireless sensor networksrdquoWireless Networks vol 22 no 1pp 135ndash157 Jan 2016

[38] C A C Coello ldquoEvolutionary multi-objective optimization A his-torical view of the fieldrdquoIEEE Computational Intelligence Magazinevol 1 no 1 pp 28ndash36 Feb 2006

[39] A Konak D W Coit and A E Smith ldquoMulti-objective optimizationusing genetic algorithms A tutorialrdquoReliability Engineering amp SystemSafety vol 91 no 9 pp 992ndash1007 Sep 2006

[40] M A Adnan M A Razzaque I Ahmed and I F Isnin ldquoBio-mimicoptimization strategies in wireless sensor networks A surveyrdquo Sensorsvol 14 no 1 pp 299ndash345 Dec 2014

[41] R V Kulkarni and G K Venayagamoorthy ldquoParticle swarm optimiza-tion in wireless sensor networks A brief surveyrdquoIEEE Transactionson Systems Man and Cybernetics Part C (Applications and Reviews)vol 41 no 2 pp 262ndash267 Mar 2011

[42] R V Kulkarni A Forster and G K Venayagamoorthy ldquoCompu-tational intelligence in wireless sensor networks A surveyrdquo IEEECommunications Surveys and Tutorials vol 13 no 1 pp 68ndash96 FirstQuarter 2011

[43] S Jabbar R Iram A A Minhas I Shafi S Khalid and M Ah-mad ldquoIntelligent optimization of wireless sensor networks throughbio-inspired computing Survey and future directionsrdquoInternationalJournal of Distributed Sensor Networks Volume 2013 Article ID421084 13 pages httpdxdoiorg1011552013421084

[44] D S Deif and Y Gadallah ldquoClassification of wireless sensor networksdeployment techniquesrdquoIEEE Communications Surveys and Tutorialsvol 16 no 2 pp 834ndash855 Second Quarter 2014

[45] M Iqbal M Naeem A Anpalagan A Ahmed and M Azam ldquoWire-less sensor network optimization Multi-objective paradigmrdquo Sensorsvol 15 no 7 pp 17 572ndash17 620 Jul 2015

[46] M Iqbal M Naeem A Anpalagan N N Qadri and M Imran ldquoMulti-objective optimization in sensor networks Optimization classificationapplications and solution approachesrdquoComputer Networks vol 99no 22 pp 134ndash161 Apr 2016

[47] P Rawat K D Singh H Chaouchi and J M Bonnin ldquoWireless sensornetworks A survey on recent developments and potential synergiesrdquoThe Journal of supercomputing vol 68 no 1 pp 1ndash48 Apr 2014

[48] M Khanafer M Guennoun and H T Mouftah ldquoA survey ofbeacon-enabled IEEE 802154 MAC protocols in wireless sensor networksrdquoIEEE Communications Surveys and Tutorials vol 16 no 2 pp 856ndash876 Second Quarter 2014

[49] R J M Vullers R van Schaijk H J Visser J Penders and C VHoof ldquoEnergy harvesting for autonomous wireless sensor networksrdquoIEEE Solid-State Circuits Magazine vol 2 no 2 pp 29ndash38 Spring2010

[50] T Arampatzis J Lygeros and S Manesis ldquoA survey of applicationsof wireless sensors and wireless sensor networksrdquo inProc IEEE Inter-national Symposium on Intelligent Control Mediterrean Conference onControl and Automation Limassol Cyprus Jun 2005 pp 719ndash724

[51] A A K Somappa K Oslashvsthus and L M Kristensen ldquoAn industrialperspective on wireless sensor networks-A survey of requirements pro-tocols and challengesrdquoIEEE Communications Surveys and Tutorialsvol 16 no 3 pp 1391ndash1412 Third Quarter 2014

[52] Y Zhu J Song and F Dong ldquoApplications of wireless sensor networkin the agriculture environment monitoringrdquoProcedia Engineeringvol 16 no 3 pp 608ndash614 Jan 2014

[53] A Milenkovic C Otto and E Jovanov ldquoWireless sensor networks forpersonal health monitoring Issues and an implementationrdquo ComputerNetworks vol 29 no 13 pp 2521ndash2533 Aug 2006

[54] S J Isaac G P Hancke H Madhoo and A Khatri ldquoA survey ofwireless sensor network applications from a power utilityrsquos distributionperspectiverdquo in Proc 9th IEEE Africon Conference LivingstoneZambia Sep 2011 pp 1ndash5

[55] A Boukerche H A B F Oliveira E F Nakamura and AA FLoureiro ldquoSecure localization algorithms for wireless sensor net-worksrdquo IEEE Communications Magazine vol 46 no 4 pp 96ndash101Apr 2008

[56] Prachi and S Sharma ldquoTarget tracking technique in wireless sensornetworkrdquo in Proc IEEE International Conference on ComputingCommunication and Automation (ICCCA) Noida India May 2015pp 486ndash491

[57] B Rashid and M H Rehmani ldquoApplications of wireless sensornetworks for urban areas A surveyrdquoJournal of Network and ComputerApplications vol 60 no C pp 192ndash219 Jan 2016

[58] I Butun S D Morgera and R Sankar ldquoA survey of intrusion detectionsystems in wireless sensor networksrdquoIEEE Communications Surveysand Tutorials vol 16 no 1 pp 266ndash282 First Quarter 2014

[59] P Spachos and D Hatzinakos ldquoReal-time indoor carbondioxidemonitoring through cognitive wireless sensor networksrdquoIEEE SensorsJournal vol 16 no 2 pp 506ndash514 Jan 2016

[60] S Movassaghi M Abolhasan J Lipman D Smith and AJamalipourldquoWireless body area networks A surveyrdquoIEEE CommunicationsSurveys and Tutorials vol 16 no 3 pp 856ndash876 Third Quarter 2014

[61] J Chinrungrueng U Sununtachaikul and S Triamlumlerd ldquoA vehic-ular monitoring system with power-efficient wireless sensor networksrdquoin Proc 6th International Conference on ITS TelecommunicationsChengdu China Jun 2006 pp 951ndash954

[62] M M Alam and H E Ben ldquoSurveying wearable human assistive tech-nology for life and safety critical applications Standards challengesand opportunitiesrdquoSensors vol 14 no 5 pp 9153ndash9209 May 2014

[63] L Mainetti L Patrono and A Vilei ldquoEvolution of wireless sensornetworks towards the internet of things A surveyrdquo inProc 19thIEEE International Conference on Software Telecommunications andComputer Networks(SoftCOM) Split Croatia Sep 2011 pp 1ndash6

[64] F J Wu Y F Kao and Y C Tseng ldquoFrom wireless sensornetworkstowards cyber physical systemsrdquoPervasive and Mobile Computingvol 7 no 4 pp 397ndash413 Aug 2011

[65] E Fadela V C Gungorb L Nassefa N Akkaria M G AMalikaS Almasria and I F Akyildiz ldquoA survey on wireless sensornetworksfor smart gridrdquoComputer Communications vol 71 no 1 pp 22ndash33Nov 2015

[66] K Sohrabi J Gao V Ailawadhi and G J Pottie ldquoProtocols for self-organization of a wireless sensor networkrdquoIEEE Personal Communi-cations vol 7 no 5 pp 16ndash27 Oct 2000

[67] J N Al-Karaki and A E Kamal ldquoRouting techniques in wirelesssensor networks A surveyrdquoIEEE Wireless Communications vol 11no 6 pp 6ndash28 Dec 2004

[68] D Wang B Xie and D P Agrawal ldquoCoverage and lifetime opti-mization of wireless sensor networks with Gaussian distributionrdquo IEEETransactions on Mobile Computing vol 7 no 12 pp 1444ndash1458 Dec2008

[69] B Bhuyan H K D Sarma N Sarma A Kar and R Mall ldquoQualityof service (QoS) provisions in wireless sensor networks andrelatedchallengesrdquoWireless Sensor Network vol 2 no 11 pp 861ndash868Nov 2010

[70] S Meguerdichian F Koushanfar M Potkonjak and M B SrivastavaldquoCoverage problems in wireless ad-hoc sensor networksrdquo inProc20th Annual Joint Conference of the IEEE Computer and Commu-nications Societies ndash The Conference on Computer Communications(INFOCOMrsquo01) vol 3 Anchorage USA Apr 2001 pp 1380ndash1387

[71] S Meguerdichian F Koushanfar G Qu and M Potkonjak ldquoExposurein wireless ad-hoc sensor networksrdquo inProc 7th Annual InternationalConference on Mobile Computing and Networking (MobiComrsquo01)Rome Italy 2001 pp 139ndash150

[72] C-F Huang and Y-C Tseng ldquoThe coverage problem in a wirelesssensor networkrdquoMobile Networks and Applications vol 10 no 4pp 519ndash528 Aug 2005

[73] X Bai Z Yun D Xuan B Chen and W Zhao ldquoOptimal multiple-coverage of sensor networksrdquo inProc IEEE International Conferenceon Computer Communications (INFOCOMrsquo11) Shanghai China Apr2011 pp 2498ndash2506

[74] J Chen S He Y Sun P Thulasiraman and X S Shen ldquoOptimalflow control for utility-lifetime tradeoff in wireless sensor networksrdquoComputer Networks vol 53 no 18 pp 3031ndash3041 Dec 2009

[75] M A Alsheikh D T Hoang D Niyato H-P Tan and S LinldquoMarkov decision processes with applications in wireless sensornetworks A surveyrdquoIEEE Communications Surveys and Tutorialsvol 17 no 3 pp 1239ndash1267 Third Quarter 2015

[76] B Krishnamachari and F Ordonez ldquoAnalysis of energy-efficient fairrouting in wireless sensor networks through non-linear optimizationrdquoin Proc 58th IEEE Vehicular Technology Conference (VTCrsquo03-Fall)Orlando USA Oct 2003 pp 2844ndash2848

[77] A Syarif I Benyahia A Abouaissa L Idoumghar R F Sari andP Lorenz ldquoEvolutionary multi-objective based approach for wirelesssensor network deploymentrdquo inProc IEEE International Conferenceon Communications (ICC) Sydney Australia Jun 2014 pp 1831ndash1836

[78] X Wang G Xing Y Zhang C Lu R Pless and C Gill ldquoIntegratedcoverage and connectivity configuration in wireless sensornetworksrdquoin Proc 1st ACM International Conference on Embedded NetworkedSensor Systems (SenSysrsquo03) Los Angeles USA Nov 2003 pp 28ndash39

[79] G Xing X Wang Y Zhang C Lu R Pless and C Gill ldquoIntegratedcoverage and connectivity configuration for energy conservation in

sensor networksrdquoACM Transactions on Sensor Networks vol 1 no 1pp 36ndash72 Aug 2005

[80] H M Ammari and S K Das ldquoCoverage connectivity andfault toler-ance measures of wireless sensor networksrdquo inProc 8th InternationalSymposium on Stabilization Safety and Security of Distributed Systems(SSSrsquo06) Dallas TX USA Nov 2006 pp 35ndash49

[81] mdashmdash ldquoIntegrated coverage and connectivity in wirelesssensor net-works A two-dimensional percolation problemrdquoIEEE Transactionson Computers vol 57 no 10 pp 1423ndash1434 Oct 2008

[82] Y Li Y Q Song Y h Zhu and R Schott ldquoDeploying wirelesssensors for differentiated coverage and probabilistic connectivityrdquo inProc IEEE Wireless Communications and Networking Conference(WCNCrsquo10) Sydney Australia Apr 2010 pp 1ndash6

[83] J H Chang and L Tassiulas ldquoMaximum lifetime routingin wirelesssensor networksrdquoIEEEACM Transactions on Networking vol 12no 4 pp 609ndash619 Aug 2004

[84] R Rajagopalan C K Mohan P Varshney and K Mehrotra ldquoMulti-objective mobile agent routing in wireless sensor networksrdquo in ProcIEEE Congress on Evolutionary Computation (CECrsquo05) EdinburghUK Sep 2005 pp 1730ndash1737

[85] K T K Cheung S Yang and L Hanzo ldquoAchieving maximum energy-efficiency in multi-relay OFDMA cellular networks A fractional pro-gramming approachrdquoIEEE Transactions on Communications vol 61no 7 pp 2746ndash2757 May 2013

[86] mdashmdash ldquoSpectral and energy spectral efficiency optimization of jointtransmit and receive beamforming based multi-relay MIMO-OFDMAcellular networksrdquoIEEE Transactions on Wireless Communicationsvol 13 no 11 pp 6147ndash6165 Aug 2014

[87] W Jing Z Lu X Wen Z Hu and S Yang ldquoFlexible resourceallocation for joint optimization of energy and spectral efficiency inOFDMA multi-cell networksrdquoIEEE Communications Letters vol 19no 3 pp 451ndash454 Jan 2015

[88] K T K Cheung S Yang and L Hanzo ldquoDistributed energy spectralefficiency optimization for partialfull interference alignment in multi-user multi-relay multi-cell MIMO systemsrdquoIEEE Transactions onSignal Processing vol 64 no 4 pp 882ndash896 Feb 2016

[89] T Abrao L D H Sampaio S Yang K T K Cheung P JE Jeszen-sky and L Hanzo ldquoEnergy efficient OFDMA networks maintainingstatistical QoS guarantees for delay-sensitive trafficrdquoIEEE Accessvol 4 pp 774ndash791 Mar 2016

[90] X Mao S Tang X Xu X Li and H Ma ldquoEnergy efficientopportunistic routing in wireless sensor networksrdquoIEEE Transactionson Parallel and Distributed Systems vol 22 no 11 pp 1934ndash1942Nov 2011

[91] W Choi and S Das ldquoA novel framework for energy-conserving datagathering in wireless sensor networksrdquo inProc IEEE InternationalConference on Computer Communications (INFOCOMrsquo05) MiamiUSA Mar 2005 pp 1985ndash1996

[92] H M Ammari and S K Das ldquoTrade-off between energy savingsand source-to-sink delay in data dissemination for wireless sensornetworksrdquo in Proc 8th ACM International Symposium on ModelingAnalysis and Simulation of Wireless and Mobile Systems (MSWiMrsquo05)Montreal Canada Oct 2005 pp 126ndash133

[93] mdashmdash ldquoA trade-off between energy and delay in data dissemination forwireless sensor networks using transmission range slicingrdquo ComputerCommunications vol 31 no 9 pp 1687ndash1704 Jun 2008

[94] M Haenggi and D Puccinelli ldquoRouting in ad hoc networks A casefor long hopsrdquo IEEE Communication Magazine vol 43 no 10 pp93ndash101 Dec 2005

[95] J Zhang T Yan and S H Son ldquoDeployment strategies for differ-entiated detection in wireless sensor networksrdquo inProc 3rd AnnualIEEE Communications Society Conference on Sensor Mesh andAdHoc Communications and Networks (SECONrsquo06) Reston USA Sep2006 pp 316ndash325

[96] Y Zou and K Chakrabarty ldquoUncertainty-aware and coverage-orienteddeployment for sensor networksrdquoJournal of Parallel and DistributedComputing vol 64 no 7 pp 788ndash798 Jul 2004

[97] mdashmdash ldquoSensor deployment and target localization in distributed sensornetworksrdquoACM Transactions on Embedded Computing Systems vol 3no 1 pp 61ndash91 Feb 2004

[98] G Molina E Alba and E-G Talbi ldquoOptimal sensor network layoutusing multi-objective metaheuristicsrdquoJournal of Universal ComputerScience vol 14 no 15 pp 2549ndash2565 Aug 2008

[99] M Woehrle D Brockhoff T Hohm and S Bleuler ldquoInvestigatingcoverage and connectivity trade-offs in wireless sensor networksThe benefits of MOEAsrdquo inMultiple Criteria Decision Making forSustainable Energy and Transportation Systems ser Lecture Notes in

Economics and Mathematical Systems M Ehrgott B Naujoks T JStewart and J Wallenius Eds Springer Berlin Heidelberg Oct 2010vol 634 pp 211ndash221

[100] M L Berre F Hnaien and H Snoussi ldquoMulti-objective optimizationin wireless sensors networksrdquo inProc IEEE International Conferenceon Microelectronics (ICMrsquo11) Hammamet Tunisia Dec 2011 pp1ndash4

[101] J Jia J Chen G Chang Y Wen and J Song ldquoMulti-objectiveoptimization for coverage control in wireless sensor network withadjustable sensing radiusrdquoComputers and Mathematics with Appli-cations vol 57 no 11-12 pp 1767ndash1775 Jun 2009

[102] R Rajagopalan ldquoMulti-objective optimization algorithms for sensornetwork designrdquo inProc 11th IEEE Annual Wireless and MicrowaveTechnology Conference (WAMICONrsquo10) Melbourne USA Apr 2010pp 1ndash4

[103] N Aitsaadi N Achir K Boussetta and G Pujolle ldquoMulti-objectiveWSN deployment Quality of monitoring connectivity and lifetimerdquoin Proc IEEE International Conference on Communications (ICCrsquo10)Cape Town South Africa May 2010 pp 1ndash6

[104] S M Jameii and S M Jameii ldquoMulti-objective energyefficientoptimization algorithm for coverage control in wireless sensor net-worksrdquo International Journal of Computer Science Engineering andInformation Technology vol 3 no 4 pp 25ndash33 Aug 2013

[105] M Z A Bhuiyan G Wang J Cao and J Wu ldquoDeploying wirelesssensor networks with fault-tolerance for structural health monitoringrdquoIEEE Transactions on Computers vol 64 no 2 pp 382ndash395 Feb2015

[106] R R Patra and P K Patra ldquoAnalysis ofk-coverage in wireless sensornetworksrdquo International Journal of Advanced Computer Science andApplications vol 2 no 9 pp 91ndash96 2011

[107] J H Zhu K L Hung B Bensaou and F Nait-Abdesselam ldquoRate-lifetime tradeoff for reliable communication in wireless sensor net-worksrdquo Computer Networks vol 52 no 1 pp 25ndash43 Jan 2008

[108] J Mo and J Walrand ldquoFair end-to-end window-based congestioncontrolrdquo IEEEACM Transactions on Networking vol 8 no 5 pp556ndash567 Dec 2000

[109] F Kelly ldquoCharging and rate control for elastic trafficrdquo EuropeanTransactions on Telecommunications vol 8 no 1 pp 33ndash37 JanFeb1997

[110] S Armenia G Morabito and S Palazzo ldquoAnalysis of locationprivacyenergy efficiency tradeoffs in wireless sensor networksrdquo inProc 6th International IFIP-TC6 Networking Conference AtlantaUSA May 2007 pp 215ndash226

[111] A Liu Z Zheng C Zhang and Z Chen ldquoSecure and energy-efficientdisjoint multipath routing for WSNsrdquoIEEE Transactions on VehicularTechnology vol 61 no 7 pp 3255ndash3265 Jun 2012

[112] D B Jourdan and O L de Weck ldquoLayout optimization for a wirelesssensor network using a multi-objective genetic algorithmrdquo in Proc59th IEEE Vehicular Technology Conference (VTCrsquo04-Spring) MilanItaly May 2004 pp 2466ndash2470

[113] Z Yun X Bai D Xuan T H Lai and W Jia ldquoOptimal deploymentpatterns for full coverage andk-connectivity (k le 6) wireless sensornetworksrdquo IEEEACM Transactions on Networking vol 18 no 3 pp934ndash947 Jun 2010

[114] J S Li and H C Kao ldquoDistributedk-coverage self-location estimationscheme based on Voronoi diagramrdquoIET Communications vol 4 no 2pp 167ndash177 Jan 2010

[115] K S S Rani and N Devarajan ldquoMultiobjective sensornode deploye-ment in wireless sensor networksrdquoInternational Journal of Engineer-ing Science vol 4 no 4 pp 1262ndash1266 Apr 2012

[116] M Abo-Zahhad N Sabor S Sasaki and S M Ahmed ldquoA centralizedimmune-Voronoi deployment algorithm for coverage maximization andenergy conservation in mobile wireless sensor networksrdquoInformationFusion vol 30 pp 36ndash51 Jul 2016

[117] Q Zhao and M Gurusamy ldquoLifetime maximization for connectedtarget coverage in wireless sensor networksrdquoIEEEACM Transactionson Networking vol 16 no 6 pp 1378ndash1391 Dec 2008

[118] K P Shih H C Chen C M Chou and B J Liu ldquoOn target coveragein wireless heterogeneous sensor networks with multiple sensing unitsrdquoJournal of Network and Computer Applications vol 32 no 4 pp866ndash877 Jul 2009

[119] S S Ram D Manjunath S K Iyer and D YogeshwaranldquoOn the pathcoverage properties of random sensor networksrdquoIEEE Transactions onMobile Computing vol 6 no 5 pp 1536ndash1233 May 2007

[120] A Chen S Kumar and T H Lai ldquoLocal barrier coverage in wirelesssensor networksrdquoIEEE Transactions on Mobile Computing vol 9no 4 pp 491ndash504 Apr 2010

[121] K Sakai M T Sun W S Ku T H Lai and A V Vasilakos ldquoAframework for the optimalk-coverage deployment patterns of wirelesssensorsrdquoIEEE Sensors Journal vol 15 no 12 pp 7273ndash7283 Dec2015

[122] C Ma W Liang M Zheng and H Sharif ldquoA connectivity-awareapproximation algorithm for relay node placement in wireless sensornetworksrdquo IEEE Sensors Journal vol 16 no 2 pp 515ndash528 Jan2016

[123] C Zhang X Bai J Teng D Xuan and W Jia ldquoConstructing low-connectivity and full-coverage three dimensional sensor networksrdquoIEEE Journal on Selected Areas in Communications vol 28 no 7pp 984ndash993 Sep 2010

[124] Y Zhao J Wu F Li and S Lu ldquoOn maximizing the lifetime ofwireless sensor networks using virtual backbone schedulingrdquo IEEETransactions on Parallel and Distributed Systems vol 23 no 8 pp1528ndash1535 Aug 2012

[125] V Shah-Mansouri and V W S Wong ldquoLifetime-resource tradeoff formulticast traffic in wireless sensor networksrdquoIEEE Transactions onWireless Communications vol 9 no 6 pp 1924ndash1934 Jun 2010

[126] R R Rout and S K Ghosh ldquoEnhancement of lifetime using duty cycleand network coding in wireless sensor networksrdquoIEEE Transactionson Wireless Communications vol 12 no 2 pp 656ndash667 Feb 2013

[127] C F Wang J D Shih B H Pan and T Y Wu ldquoA network lifetimeenhancement method for sink relocation and its analysis in wirelesssensor networksrdquoIEEE Sensors Journal vol 14 no 6 pp 1932ndash1943 Jun 2014

[128] H Nama M Chiang and N Mandayam ldquoUtility-lifetime trade-off in self-regulating wireless sensor networks A cross-layer designapproachrdquo inProc IEEE International Conference on Communications(ICCrsquo06) Istanbul Turkey Jun 2006 pp 3511ndash3516

[129] J H Zhu S Chen B Bensaou and K-L Hung ldquoTradeoff betweenlifetime and rate allocation in wireless sensor networks Across layerapproachrdquo inProc 26th IEEE International Conference on ComputerCommunications (INFOCOMrsquo07) Anchorage USA May 2007 pp267ndash275

[130] S He J Chen W Xu Y Sun and P Thulasiraman ldquoA stochasticmultiobjective optimization framework for wireless sensor networksrdquoEURASIP Journal of Wireless Communication and Networking vol2010 May 2010

[131] J Luo A Iyer and C Rosenberg ldquoThroughput-lifetime trade-offs inmultihop wireless networks under an SINR-based interference modelrdquoIEEE Transactions on Mobile Computing vol 10 no 3 pp 419ndash433Sep 2011

[132] M J Miller and N H Vaidya ldquoA MAC protocol to reduce sensornetwork energy consumption using a wakeup radiordquoIEEE Transactionson Mobile Computing vol 4 no 3 pp 228ndash242 MayJun 2005

[133] S J Baek G de Veciana and X Su ldquoMinimizing energyconsumptionin large-scale sensor networks through distributed data compressionand hierarchical aggregationrdquoIEEE Journal on Selected Areas inCommunications vol 22 no 6 pp 1130ndash1140 Agu 2004

[134] M Elhoseny X Yuan Z Yu C Mao H K El-Minir and A MRiad ldquoBalancing energy consumption in heterogeneous wireless sensornetworks using genetic algorithmrdquoIEEE Communications Lettersvol 19 no 12 pp 2194ndash2197 Dec 2015

[135] M Zorzi and R R Rao ldquoGeographic random forwarding (GeRaF) forad hoc and sensor networks Energy and latency performancerdquo IEEETransactions on Mobile Computing vol 2 no 4 pp 349ndash365 Oct2003

[136] X Yang and N Vaidya ldquoA wakeup scheme for sensor networksAchieving balance between energy saving and end-to-end delayrdquo inProc 10th IEEE Real-Time and Embedded Technology and Applica-tions Symposium (RTASrsquo04) Toronto Canada May 2004 pp 19ndash26

[137] Y Yu B Krishnamachari and V K Prasanna ldquoEnergy-latencytradeoffs for data gathering in wireless sensor networksrdquoin ProcIEEE International Conference on Computer Communications(INFO-COMrsquo04) Hong Kong China Mar 2004 pp 244ndash255

[138] Y Yu and V K Prasanna ldquoEnergy-balanced task allocation forcollaborative processing in wireless sensor networksrdquoMobile Networksand Applications vol 10 no 1-2 pp 115ndash131 Feb 2005

[139] Y Yao Q Cao and A V Vasilakos ldquoEDAL An energy-efficientdelay-aware and lifetime-balancing data collection protocol for het-erogeneous wireless sensor networksrdquoIEEEACM Transactions onNetworking vol 23 no 3 pp 810ndash823 Jun 2015

[140] M Borghini F Cuomo T Melodia U Monaco and F RicciatoldquoOptimal data delivery in wireless sensor networks in the energyand latency domainsrdquo inProc 1st IEEE International Conference on

Wireless Internet (WICONrsquo05) Budapest Hungary Jul 2005 pp 138ndash145

[141] T T Huynh and C S Hong ldquoAn energydelay efficient multi-hoprouting scheme for wireless sensor networksrdquoIEICE Transactions onInformation and Systems vol E89-D no 5 pp 1654ndash1661 May 2006

[142] T Moscibroda P V Rickenbach and R Wattenhofer ldquoAnalyzing theenergy-latency trade-off during the deployment of sensor networksrdquo inProc IEEE International Conference on Computer Communications(INFOCOMrsquo06) Barcelona Spain Apr 2006 pp 1ndash13

[143] W L Leow and H Pishro-Nik ldquoDelay and energy tradeoff in multi-state wireless sensor networksrdquo inProc IEEE Global Telecommuni-cations Conference (GLOBECOMrsquo07) Washington USA Nov 2007pp 1028ndash1032

[144] J Mao Z Wu and X Wu ldquoA TDMA scheduling scheme formany-to-one communications in wireless sensor networksrdquoComputerCommunications vol 30 no 4 pp 863ndash872 Feb 2007

[145] M R Minhas S Gopalakrishnan and V Leung ldquoMultiobjective rout-ing for simultaneously optimizing system lifetime and source-to-sinkdelay in wireless sensor networksrdquo inProc 29th IEEE InternationalConference on Distributed Computing Systems Workshops (ICDCSWorkshopsrsquo09) Montreal Canada Jun 2009 pp 123ndash129

[146] C T Cheng and H Leung ldquoA multi-objective optimization frameworkfor cluster-based wireless sensor networksrdquo inProc IEEE Inter-national Conference on Cyber-Enabled Distributed Computing andKnowledge Discovery (CyberCrsquo12) Sanya China Oct 2012 pp 341ndash347

[147] H Li C Wu D Yu Q-S Hua and F Lau ldquoAggregation latency-energy tradeoff in wireless sensor networks with successive interferencecancellationrdquoIEEE Transactions on Parallel and Distributed Systemsvol 24 no 1 pp 2160ndash2170 Nov 2013

[148] Q Gao Y Zuo J Zhang and X H Peng ldquoImproving energy efficiencyin a wireless sensor network by combining cooperative MIMO withdata aggregationrdquoIEEE Transactions on Vehicular Technology vol 59no 8 pp 3956ndash3965 Oct 2010

[149] W Fang F Liu F Yang L Shu and S Nishio ldquoEnergy-efficientcooperative communication for data transmission in wireless sensornetworksrdquoIEEE Transactions on Consumer Electronics vol 59 no 8pp 2185ndash2192 Nov 2010

[150] S D Muruganathan D C F Ma R I Bhasin and A O FapojuwoldquoA centralized energy-efficient routing protocol for wireless sensornetworksrdquo IEEE Communications Magazine vol 43 no 3 pp s8ndashs13 Mar 2005

[151] J M Lanza-Gutierrez and J A Gomez-Pulido ldquoAssuming multiob-jective metaheuristics to solve a three-objective optimisation problemfor relay node deployment in wireless sensor networksrdquoApplied SoftComputing vol 30 no C pp 675ndash687 May 2015

[152] M J Miller C Sengul and I Gupta ldquoExploring the energy-latencytrade-off for broadcasts in energy-saving sensor networksrdquo in Proc25th IEEE International Conference on Distributed Computing Systems(ICDCSrsquo05) Columbus USA Jun 2005 pp 17ndash26

[153] A Sahoo and S Chilukuri ldquoDGRAM A delay guaranteed routingand MAC protocol for wireless sensor networksrdquoIEEE Transactionson Mobile Computing vol 9 no 10 pp 1407ndash1423 Oct 2010

[154] C T Cheng C K Tse and F C M Lau ldquoA delay-aware data col-lection network structure for wireless sensor networksrdquoIEEE SensorsJournal vol 11 no 3 pp 699ndash710 Mar 2010

[155] N Aitsaadi N Achir K Boussetta and G Pujolle ldquoTarget track-ing technique in wireless sensor networkrdquo inProc IEEE VehicularTechnology Conference (VTCrsquo08-Spring) Singapore May 2008 pp123ndash127

[156] Y Wang I G Guardiola and X Wu ldquoRSSI and LQI dataclustering techniques to determine the number of nodes in wire-less sensor networksrdquoInternational Journal of Distributed Sen-sor Networks Volume 2014 Article ID 380526 11 pageshttpdxdoiorg1011552014380526

[157] R Enayatifar M Yousefi A H Abdullah and A N Darus ldquoAnovel sensor deployment approach using multi-objective imperialistcompetitive algorithm in wireless sensor networksrdquoArabian Journalfor Science and Engineering vol 39 no 6 pp 4637ndash4650 Jun 2014

[158] C T Cheng and C K Tse ldquoAn analysis on the delay-aware datacollection network structure using Pareto optimalityrdquo inProc IEEEInternational Conference on Cyber-Enabled Distributed Computingand Knowledge Discovery (CyberCrsquo12) Sanya China Oct 2012 pp348ndash352

[159] X Han X Cao E L Lloyd and C C Shen ldquoFault-tolerant relaynode placement in heterogeneous wireless sensor networksrdquo IEEE

Transactions on Mobile Computing vol 9 no 5 pp 643ndash656 May2010

[160] T Y Wang Y S Han P K Varshney and P N Chen ldquoDistributedfault-tolerant classification in wireless sensor networksrdquo IEEE Journalon Selected Areas in Communications vol 23 no 4 pp 724ndash734Apr 2005

[161] M Yao C Lin P Zhang Y Tian and S Xu ldquoTDMA schedulingwith maximum throughput and fair rate allocation in wireless sensornetworksrdquo inProc IEEE International Conference on Communications(ICC) Budapest Hungary Jun 2013 pp 1576ndash1581

[162] S Narayanan J H Jun V Pandit and D P Agrawal ldquoProportionallyfair rate allocation in regular wireless sensor networksrdquoin Proc IEEEConference on Computer Communications Workshops (INFOCOMWKSHPS) Shanghai China Apr 2011 pp 549ndash554

[163] Y-H C M-H Lee ldquoFault detection of wireless sensor networksrdquoComputer Communications vol 31 no 14 pp 3469ndash3475 Sep 2008

[164] R Tan G Xing J Wang and H C So ldquoExploiting reactive mobilityfor collaborative target detection in wireless sensor networksrdquo IEEETransactions on Wireless Communications vol 9 no 3 pp 317ndash332Mar 2010

[165] F Yan P Martins and L Decreusefond ldquoAccuracy of homology basedcoverage hole detection for wireless sensor networks on sphererdquo IEEETransactions on Wireless Communications vol 13 no 7 pp 3583ndash3595 Jul 2014

[166] S Adlakha S Ganeriwal C Schurgers and M Srivastava ldquoPosterabstract Density accuracy delay and lifetime tradeoffsin wirelesssensor networks-a multidimensional design perspectiverdquoin Proc 1stACM International Conference on Embedded Networked SensorSys-tems (SenSysrsquo03) Los Angeles USA Nov 2003 pp 296ndash297

[167] R Tynan G M OrsquoHare M J OrsquoGrady and C Muldoon ldquoEDLAtradeoffs for wireless sensor network target trackingrdquo inProc 29thIEEE International Conference on Distributed Computing SystemsWorkshops (ICDCS Workshopsrsquo09) Montreal Canada Jun 2009 pp440ndash446

[168] K Ren W Lou and Y Zhang ldquoLEDS Providing location-aware end-to-end data security in wireless sensor networksrdquoIEEE Transactionson Mobile Computing vol 7 no 5 pp 585ndash598 May 2008

[169] X Chen K Makki K Yen and N Pissinou ldquoSensor network securityA surveyrdquo IEEE Communications Surveys and Tutorials vol 11 no 2pp 52ndash73 Second Quarter 2009

[170] A R Rahimi-Vahed and S M Mirghorbani ldquoA multi-objective particleswarm for a flow shop scheduling problemrdquoJournal of combinatorialoptimization vol 13 no 1 pp 79ndash102 Jan 2007

[171] S M Jameii K Faez and M Dehghan ldquoMultiobjectiveoptimizationfor topology and coverage control in wireless sensor networksrdquo Inter-national Journal of Distributed Sensor Networks Article ID 36381511 pages 2015 httpdxdoiorg1011552015363815

[172] A P Wierzbicki Multiple Criteria Decision Making Theory andApplication HagenKonigswinter West Germany Springer BerlinHeidelberg Aug 1979 ch The Use of Reference Objectives inMultiobjective Optimization pp 468ndash486

[173] C Romero ldquoA general structure of achievement function for a goalprogramming modelrdquoEuropean Journal of Operational Research vol153 no 3 pp 675ndash686 Mar 2004

[174] U Ozcan and B Toklu ldquoMultiple-criteria decision-making intwo-sided assembly line balancing A goal programming and a fuzzy goalprogramming modelsrdquoComputers amp Operations Research vol 36no 6 pp 1955ndash1965 Jun 2009

[175] G S Antonio-Javier G S Felipe R H David and G H JoanldquoOn the optimization of wireless multimedia sensor networks A goalprogramming approachrdquoSensors vol 12 no 9 pp 12 634ndash12 660Sep 2012

[176] O Ustun ldquoMulti-choice goal programming formulation based on theconic scalarizing functionrdquoApplied Mathematical Modelling vol 36no 3 pp 974ndash988 Mar 2012

[177] Y-C Tang and C-T Chang ldquoMulticriteria decision-making based ongoal programming and fuzzy analytic hierarchy process An applicationto capital budgeting problemrdquoKnowledge-Based Systems vol 26 pp288ndash293 Feb 2012

[178] T L Saaty ldquoRelative measurement and its generalization in decisionmaking Why pairwise comparisons are central in mathematics forthe measurement of intangible factors the analytic hierarchynetworkprocessrdquoReview of the Royal Academy of Exact Physical and NaturalSciences Series A Mathematics (RACSAM) vol 102 no 2 pp 251ndash318 Jun 2008

[179] mdashmdash Decision Making for Leaders The Analytic Hierarchy Processfor Decisions in a Complex World Pittsburgh Pennsylvania RWSPublications 2010

[180] A Ray B Sarkar and S Sanyal ldquoThe TOC-based algorithm for solv-ing multiple constraint resourcesrdquoIEEE Transactions on EngineeringManagement vol 57 no 2 pp 301ndash309 May 2010

[181] J Q Wang Z T Zhang J Chen Y Z Guo S Wang S D SunT Qu and G Q Huang ldquoThe TOC-based algorithm for solvingmultiple constraint resources A re-examinationrdquoIEEE Transactionson Engineering Management vol 61 no 1 pp 138ndash146 Feb 2014

[182] C Li S G Anavatti and T Ray ldquoAnalytical hierarchy process usingfuzzy inference technique for real-time route guidance systemrdquo IEEETransactions on Intelligent Transportation Systems vol 15 no 1 pp84ndash93 Feb 2014

[183] M Wang and S N Li ldquoAn energy-efficient load-balanceable multipathrouting algorithm based on AHP for wireless sensor networksrdquo in ProcIEEE international conference on intelligent computing and intelligentsystems (ICIS) Xiamen China Oct 2010 pp 251ndash256

[184] T Gao R C Jin J Y Qin and L D Wang ldquoA novel node-disjointmultipath routing protocol for wireless multimedia sensornetworksrdquo inProc 2nd IEEE international conference on signal processing systems(ICSPS) Dalian China Jul 2010 pp 790ndash794

[185] G Eichfelder ldquoScalarizations for adaptively solving multi-objectiveoptimization problemsrdquoComputational Optimization and Applicationsvol 44 no 44 pp 249ndash273 Nov 2009

[186] K Miettinen Nonlinear Multiobjective Optimization Kluwer Aca-demic Publishers 1998

[187] R Kasimbeyli ldquoA conic scalarization method in multi-objective opti-mizationrdquo Journal of Global Optimization vol 56 no 2 pp 279ndash297Jun 2013

[188] D F Jones S K Mirrazavi and M Tamiz ldquoMulti-objective meta-heuristics An overview of the current state-of-the-artrdquoEuropeanJournal of Operational Research vol 137 no 1 pp 1ndash9 Feb 2002

[189] D Lei and Z Wu ldquoCrowding-measure-based multiobjective evolu-tionary algorithm for job shop schedulingrdquoInternational Journal ofAdvanced Manufacturing Technology vol 30 no 1 pp 112ndash117 Aug2006

[190] S K Pal S Bandyopadhyay and C A Murthy ldquoGeneticalgorithmsfor generation of class boundariesrdquoIEEE Transaction on Systems Manand Cybernetics-Part B Cybernetics vol 28 no 6 pp 816ndash828 Oct1998

[191] C M Fonseca and P J Fleming ldquoGenetic algorithms for multiob-jective optimization Formulation discussion and generalizationrdquo inProc 5th International Conference on Genetic Algorithms (ICGArsquo93)San Mateo USA Jun 1993 pp 416ndash423

[192] D B Jourdan and O L de Weck ldquoMulti-objective genetic algorithmfor the automated planning of a wireless sensor network to monitor acritical facilityrdquo in Proc SPIE 5403 Sensors and Command ControlCommunications and Intelligence (C3I) Technologies for HomelandSecurity and Homeland Defense III 565 Sep 2004 pp 565ndash575

[193] N Srinivas and K Deb ldquoMultiobjective optimizationusing nondomi-nated sorting in genetic algorithmsrdquoEvolutionary Computation vol 2no 3 pp 221ndash248 Dec 1994

[194] J Horn N Nafpliotis and D E Goldberg ldquoA niched Pareto genetic al-gorithm for multiobjective optimizationrdquo inProc 1st IEEE Conferenceon Evolutionary Computation IEEE World Congress on ComputationalIntelligence Orlando USA Jun 1994 pp 82ndash87

[195] R Storn and K Price ldquoDifferential evolution-a simple and efficientheuristic for global optimization over continuous spacesrdquo Journal ofGlobal Optimization vol 11 no 4 pp 341ndash359 Dec 1997

[196] S Das and P N Suganthan ldquoDifferential evolution Asurvey ofthe state-of-the-artrdquoIEEE Transactions on Evolutionary Computationvol 15 no 1 pp 4ndash31 Feb 2011

[197] A Rubio-Largo and M A Vega-Rodrguez ldquoApplying MOEAs to solvethe static routing and wavelength assignment problem in optical WDMnetworksrdquoEngineering Applications of Artificial Intelligence vol 26no 5-6 pp 1602ndash1619 MayJun 2013

[198] R Murugeswari and S Radhakrishnan ldquoDiscrete multi-objective differential evolution algorithm for routing inwirelessmesh networkrdquo Soft Computing [Online] 2015 Availablehttplinkspringercomarticle101007s00500-015-1730-5

[199] D Dasgupta S Yu and F Nino ldquoRecent advances in artificial immunesystems Models and applicationsrdquoApplied Soft Computing vol 11no 2 pp 1574ndash1587 Mar 2011

[200] C A C Coello and N C Cortes ldquoSolving multiobjective optimizationproblems using an artificial immune systemrdquoGenetic Programmingand Evolvable Machines vol 6 no 2 pp 163ndash190 Jun 2005

[201] S N Omkar R Khandelwal S Yathindra G N Naik andS Gopalakrishnan ldquoArtificial immune system for multi-objective de-sign optimization of composite structuresrdquoEngineering Applicationsof Artificial Intelligence vol 21 no 8 pp 1416ndash1429 Dec 2008

[202] E Atashpaz-Gargari and C Lucas ldquoImperialist competitive algorithmAn algorithm for optimization inspired by imperialistic competitionrdquo inProc IEEE Congress on Evolutionary Computation (CEC) SingaporeSep 2007 pp 4661ndash4667

[203] M Yousefi M Yousefi and A N Darus ldquoA modified imperialistcompetitive algorithm for constrained optimization of plate-fin heatexchangersrdquoJournal of Power and Energy vol 226 no 8 pp 1050ndash1059 Nov 2012

[204] B Mohammadi-Ivatloo A Rabiee A Soroudi and M Ehsan ldquoImpe-rialist competitive algorithm for solving non-convex dynamic economicpower dispatchrdquoEnergy vol 44 no 1 pp 228ndash240 Aug 2012

[205] K E Parsopoulos and M NVrahatis ldquoParticle swarm optimizationmethod in multiobjective problemsrdquo inProc ACM Symposium onApplied Computing (SACrsquo02) Madrid Spain Mar 2002 pp 603ndash607

[206] D Karaboga and B Basturk ldquoA powerful and efficient algorithm fornumerical func-tion optimization artificial bee colony (abc) algorithmrdquoJournal of global optimization vol 39 no 3 pp 459ndash471 Nov 2007

[207] M Dorigo M Birattari and T Stutzle ldquoAnt colony optimizationrdquoIEEE Computational Intelligence Magazine vol 1 no 4 pp 28ndash39Nov 2006

[208] Y D Valle G K Venayagamoorthy S Mohagheghi J C Hernandezand R G Harley ldquoParticle swarm optimization Basic conceptsvariants and applications in power systemsrdquoIEEE Transactions onEvolutionary Computation vol 12 no 2 pp 171ndash195 Apr 2008

[209] Z Sun L Tao X Wang and Z Zhou ldquoLocalization algorithmin wireless sensor networks based on multiobjective particle swarmoptimizationrdquo International Journal of Distributed Sensor Networks2015 httpdxdoiorg1011552015716291

[210] R H Liang C Y Wu Y T Chen and W T Tseng ldquoMulti-objective dynamic optimal power flow using improved artificial beecolony algorithm based on Pareto optimizationrdquoInternational Trans-actions on Electrical Energy Systems [Online] 2015 Availablehttponlinelibrarywileycomdoi101002etep2101epdf

[211] J Barbancho C Leon F J Molina and A BarbancholdquoUsing artificialintelligence in routing schemes for wireless networksrdquoComputerCommunications vol 30 no 14-15 pp 2802ndash2811 Oct 2008

[212] F Oldewurtel and P Mahonen ldquoNeural wireless sensornetworksrdquoin Proc IEEE International Conference on Systems and NetworksCommunications (ICSNCrsquo06) Tahiti Oct 2006 pp 1ndash8

[213] M Rovcanin E D Poorter D van den Akker I Moerman P De-meester and C Blondia ldquoExperimental validation of a reinforcementlearning based approach for a service-wise optimisation ofheteroge-neous wireless sensor networksrdquoWireless Networks vol 21 no 3 pp931ndash948 Apr 2015

[214] S Dong P Agrawal and K Sivalingam ldquoReinforcement learningbased geographic routing protocol for UWB wireless sensor networkrdquoin Proc IEEE Global Telecommunications Conference (GLOBE-COMrsquo07) Washington USA Jun 2007 pp 652ndash656

[215] M Rovcanin E D Poorter I Moerman and P Demeester ldquoA rein-forcement learning based solution for cognitive network cooperationbetween co-located heterogeneous wireless sensor networksrdquo Ad HocNetworks vol 17 pp 98ndash113 Jun 2014

[216] L A Zadeh ldquoSoft computing and fuzzy logicrdquoIEEE Software vol 11no 6 pp 48ndash56 Nov 1994

[217] I S AlShawi L Yan W Pan and B Luo ldquoLifetime enhancement inwireless sensor networks using fuzzy approach and A-star algorithmrdquoIEEE Sensors Journal vol 12 no 10 pp 3010ndash3018 Oct 2012

[218] H Y Shi W L Wang N M Kwok and S Y Chen ldquoGame theoryfor wireless sensor networks A surveyrdquoSensors vol 12 no 7 pp9055ndash9097 Jul 2012

[219] R Meng Y Ye and N Xie ldquoMulti-objective optimization designmethods based on game theoryrdquo inProc 8th IEEE World Congresson Intelligent Control and Automation (WCICArsquo10) Jinan China Jul2010 pp 2220ndash2227

[220] E Zitzler M Laumanns and L Thiele ldquoSPEA2 Improving thestrength Pareto evolutionary algorithm for multiobjective optimizationrdquoin Proc Evolutionary Methods for Design Optimization and ControlWith Applications to Industrial Problems (EUROGENrsquo01) AthensGreece Sep 2001 pp 95ndash100

[221] D A Van Veldhuizen and G B Lamont ldquoMultiobjectiveoptimizationwith messy genetic algorithmsrdquo inProc ACM Symposium on AppliedComputing (SACrsquo00) Villa Olmo Italy Mar 2000 pp 470ndash476

[222] J B Zydallis and G B Lamont ldquoExplicit building-block multiob-jective evolutionary algorithms for NPC problemsrdquo inProc IEEECongress on Evolutionary Computation (CECrsquo03) Canberra AustraliaDec 2003 pp 2685ndash2695

[223] M Pelikan D E Goldberg and E Cantu-Paz ldquoHierarchical problemsolving and the Bayesian optimization algorithmrdquo inProc Genetic andEvolutionary Computation Conference (GECCOrsquo00) Las Vegas USAJul 2000 pp 267ndash274

[224] M Laumanns and J Ocenasek ldquoBayesian optimization algorithms formulti-objective optimizationrdquo inParallel Problem Solving from Naturendash PPSN VII ser Lecture Notes in Computer Science J GuervosP Adamidis H-G Beyer H-P Schwefel and J-L Fernandez-Villacanas Eds Springer Berlin Heidelberg Oct 2002 vol 2439pp 298ndash307

[225] J D Knowles and D W Corne ldquoApproximating the nondominatedfront using the Pareto archived evolution strategyrdquoEvolutionary Com-putation vol 8 no 2 pp 149ndash172 Mar 2000

[226] D W Corne J D Knowles and M J Oates ldquoThe Pareto envelope-based selection algorithm for multiobjective optimizationrdquo in ParallelProblem Solving from Nature ndash PPSN VI ser Lecture Notes inComputer Science M Schoenauer K Deb G Rudolph X YaoE Lutton J Merelo and H-P Schwefel Eds Springer BerlinHeidelberg Sep 2000 vol 1917 pp 839ndash848

[227] D W Corne N R Jerram J D Knowles and M J OatesldquoPESA-II Region-based selection in evolutionary multiobjective optimiza-tionrdquo in Proc Genetic and Evolutionary Computation Conference(GECCOrsquo01) San Francisco USA Jul 2001 pp 283ndash290

[228] Q Zhang and H Li ldquoMOEAD A multi-objective evolutionary al-gorithm based on decompositionrdquoIEEE Transactions on EvolutionaryComputation vol 11 no 6 pp 712ndash731 Dec 2007

[229] A Jaszkiewicz ldquoGenetic local search for multi-objective combinatorialoptimizationrdquo European Journal of Operational Research vol 137no 1 pp 50ndash71 Feb 2002

[230] V A Armentano and J E Claudio ldquoAn application of a multi-objectivetabu search algorithm to a bicriteria flowshop problemrdquoJournal ofHeuristics vol 10 no 5 pp 463ndash481 Sep 2004

[231] R P Beausoleil ldquoMOSS multiobjective scatter search applied to non-linear multiple criteria optimizationrdquoAdvanced Engineering Informat-ics vol 169 no 2 pp 426ndash449 Mar 2006

[232] Z Chen S Li and W Yue ldquoMemetic algorithm-based multi-objectivecoverage optimization for wireless sensor networksrdquoSensors vol 14no 11 pp 20 500ndash20 518 Oct 2014

[233] A Lohne and B Weiszliging ldquoThe vector linear programsolver Bensolve ndash notes on theoretical backgroundrdquoEuropeanJournal of Operational Research Mar 2016 [Online] Availablehttpdoi101016jejor201602039

[234] F A Fortin F M D Rainville M A Gardner M Parizeau andC Gagne ldquoDEAP Evolutionary algorithms made easyrdquoJournal ofMachine Learning Research vol 13 pp 2171ndash2175 Jul 2012

[235] R P Hamalainen ldquoDecisionarium ndash aiding decisions negotiating andcollecting opinions on the webrdquoJournal of Multi-Criteria DecisionMaking vol 12 no 2-3 pp 101ndash110 Mar-Jun 2003

[236] Q Hayez Y D Smet and J Bonney ldquoD-Sight A new decision makingsoftware to address multi-criteria problemsrdquoInternational Journal ofDecision Support System Technology vol 4 no 4 pp 1ndash23 Oct-Dec2012

[237] E-G Talbi E Tantar and U V D Hekke ldquoGUIMOO A graphicaluser interface for multi objective optimizationrdquo 2005 [Online]Available httpguimoogforgeinriafr

[238] R Lin Q Wang J Hu L Gao and L Lu ldquoAn intelligentdecisionsupport system applied to the investment of real estaterdquo inProcIEEE International Conference on Industrial Technology (ICITrsquo96)Shanghai China Dec 1996 pp 801ndash805

[239] P N Koch J P Evans and D Powell ldquoInterdigitation for effective de-sign space exploration using iSIGHTrdquoStructural and MultidisciplinaryOptimization vol 23 no 2 pp 111ndash126 Mar 2002

[240] J J Durillo A J Nebro and E Alba ldquoThe jMetal framework formulti-objective optimization Design and architecturerdquoin Proc IEEECongress on Evolutionary Computation (CECrsquo10) Barcelona SpainJul 2010 pp 1ndash8

[241] M Ziadloo S S Ghamsary and N Mozayani ldquoA framework toevaluate multi-objective optimization algorithms in multi-agent nego-tiationsrdquo in Proc IEEE International Conference on ComputationalIntelligence for Measurement Systems and Applications (CIMSArsquo09)Hong Kong May 2009 pp 264ndash267

[242] A Liefooghe M Basseur L Jourdan and E G Talbi ldquoParadisEO-MOEO A framework for evolutionary multi-objective optimizationrdquo

in Evolutionary Multi-Criterion Optimization ser Lecture Notes inComputer Science S Obayashi K Deb C Poloni T Hiroyasu andT Murata Eds Springer Berlin Heidelberg Mar 2007 vol4403pp 386ndash400

[243] A Liefooghe L Jourdan and E G Talbi ldquoA software frameworkbased on a conceptual unified model for evolutionary multiobjectiveoptimization ParadisEO-MOEOrdquoEuropean Journal of OperationalResearch vol 209 no 2 pp 104ndash112 Mar 2011

[244] M A Potapov and P N Kabanov ldquoSOLVEX ndash system for solvingnonlinear global and multicriteria problemsrdquo inProc 3rd IFIP WG-76 Working Conference on Optimization-Based Computer-Aided Mod-elling and Design Prague Czech Republic 1994 pp 343ndash347

[245] K Miettinen and M M Makela ldquoInteractive multiobjective opti-mization system WWW-NIMBUS on the InternetrdquoComputers ampOperations Research vol 27 no 7-8 pp 709ndash723 Jun 2000

[246] A Konstantinidis and K Yang ldquoMulti-objectivek-connected de-ployment and power assignment in WSNs using a problem-specificconstrained evolutionary algorithm based on decompositionrdquo ComputerCommunications vol 34 no 7 pp 83ndash98 Jul 2011

[247] A Konstantinidis K Yang and Q Zhang ldquoAn evolutionary algorithmto a multi-objective deployment and power assignment problem inwireless sensor networksrdquo inProc IEEE Global CommunicationsConference (GLOBECOMrsquo08) New Orleans USA Dec 2008 pp475ndash481

[248] C Schurgers V Tsiatsis S Ganeriwal and M Srivastava ldquoOptimizingsensor networks in the energy-latency-density design spacerdquo IEEETransactions on Mobile Computing vol 1 no 1 pp 70ndash80 Jan-Mar2002

[249] H M AmmariChallenges and Opportunities of Connected k-CoveredWireless Sensor Networks From Sensor Deployment to Data Gather-ing Berlin Heidelberg Springer Berlin Heidelberg 2009 ch Trade-Off between Energy and Delay in Geographic Forwarding on Always-On Sensors pp 201ndash240

[250] mdashmdash ldquoOn the energy-delay trade-off in geographic forwarding inalways-on wireless sensor networks A multi-objective optimizationproblemrdquo Computer Networks vol 57 no 9 pp 1913ndash1935 Jun2013

[251] A Shahraki M K Rafsanjani and A B Saeid ldquoA new approachfor energy and delay trade-off intra-clustering routing inWSNsrdquoComputers and Mathematics with Applications vol 62 no 4 pp1670ndash1676 Aug 2011

[252] K Suto H Nishiyama N Kato and C W HuangldquoAn energy-efficient and delay-aware wireless computingsystem for industrial wireless sensor networksrdquoIEEEAccess vol 3 pp 1026ndash1035 2015 [Online] Availablehttpieeexploreieeeorgstampstampjsptp=amparnumber=7123559

[253] M Dong K Ota A Liu and M Guo ldquoJoint optimizationoflifetime and transport delay under reliability constraintwireless sensornetworksrdquo IEEE Transactions on Parallel and Distributed Systemsvol 27 no 1 pp 225ndash236 Jan 2016

[254] R Rajagopalan and P K Varshney ldquoData aggregation techniquesin sensor networks A surveyrdquoIEEE Communications Surveys andTutorials vol 8 no 4 pp 48ndash63 Fourth Quarter 2006

[255] S K A Imon A Khan M D Francesco and S K Das ldquoEnergy-efficient randomized switching for maximizing lifetime in tree-basedwireless sensor networksrdquoIEEEACM Transactions on Networkingvol 23 no 5 pp 1401ndash1415 Oct 2015

[256] D Xie W Wei Y Wang and H Zhu ldquoTradeoff between throughputand energy consumption in multirate wireless sensor networksrdquo IEEESensors Journal vol 13 no 10 pp 3667ndash3676 Oct 2013

[257] S Liao and Q Zhang ldquoA multiutility framework with applicationfor studying tradeoff between utility and lifetime in wireless sensornetworksrdquoIEEE Transactions on Vehicular Technology vol 64 no 10pp 4701ndash4711 Oct 2015

[258] S Boyd and L VandenbergheConvex Optimization CambridgeUK Cambridge University Press 2004

[259] D P BertsekasNonlinear Programming Athena Scientific 1999[260] M R Senouci A Mellouk K Asnoune and F Y Bouhidel

ldquoMovement-assisted sensor deployment algorithms A survey and tax-onomyrdquo IEEE Communications Surveys and Tutorials vol 17 no 4pp 2493ndash2510 Fourth Quarter 2015

[261] F Dobslaw T Zhang and M Gidlund ldquoEnd-to-end reliability-awarescheduling for wireless sensor networksrdquoIEEE Transactions on Indus-trial Informatics vol 12 no 2 pp 758ndash767 Apr 2016

[262] G H EkbataniFard R Monsefi M R Akbarzadeh-T andM HYaghmaee ldquoA multi-objective genetic algorithm based approach forenergy efficient QoS-routing in two-tiered wireless sensornetworksrdquo

in Proc 5th IEEE Wireless Pervasive Computing (ISWPCrsquo10) ModenaItaly May 2010 pp 80ndash85

[263] W Xu Q Shi X Wei M Zheng Z Xu and Y Wang ldquoDistributedoptimal rate-reliability-lifetime tradeoff in time-varying wireless sensornetworksrdquo IEEE Transactions on Wireless Communications vol 13no 9 pp 4836ndash4847 Jul 2014

[264] M A Razzaque C S Hong and S Lee ldquoData-centric multiobjec-tive QoS-aware routing protocol for body sensor networksrdquoSensorsvol 11 no 1 pp 917ndash937 Jan 2011

[265] M S Ansari A Mahani and Y S Kavian ldquoEnergy-efficient networkdesign via modelling Optimal designing point for energy reliabilitycoverage and end-to-end delayrdquoIET Networks vol 2 no 1 pp 11ndash18Mar 2013

[266] W Liu G Qin S Li J He and X Zhang ldquoA multiobjectiveevolutionary algorithm for energy-efficient cooperative spectrum sens-ing in cognitive radio sensor networkrdquoInternational Journal of Dis-tributed Sensor Networks Volume 2015 Article ID 581589 13 pageshttpdxdoiorg1011552015581589

[267] M Xiao J Wu and L Huang ldquoTime-sensitive utility-based single-copy routing in low-duty-cycle wireless sensor networksrdquoIEEE Trans-actions on Parallel and Distributed System vol 26 no 5 pp 1452ndash1465 May 2015

[268] C Lozano-Garzon and Y Donoso ldquoA multi-objective routing protocolfor a wireless sensor network using a SPEA2 approachrdquo inProcInternational Conference on Applied Numerical and ComputationalMathematics (ICANCMrsquo11) International Conference on ComputersDigital Communications and Computing (ICDCCCrsquo11) BarcelonaSpain Sep 2011 pp 39ndash44

[269] S Bandyopadhyay Q Tian and E J Coyle ldquoSpatio-temporal samplingrates and energy efficiency in wireless sensor networksrdquoIEEEACMTransactions on Networking vol 13 no 6 pp 1339ndash1352 Dec 2005

[270] D Tang T Li J Ren and J Wu ldquoCost-aware secure routing (CASER)protocol design for wireless sensor networksrdquoIEEE Transactions onParallel and Distributed Systems vol 26 no 4 pp 960ndash973 Apr2015

[271] B A Attea E A Khalil S Ozdemir and O Yıldız ldquoA multi-objectivedisjoint set covers for reliable lifetime maximization of wireless sensornetworksrdquoWireless Personal Communications vol 81 no 2 pp 819ndash838 Mar 2015

[272] S Sengupta S Das M D Nasir and B K Panigrahi ldquoMulti-objectivenode deployment in WSNs In search of an optimal trade-off amongcoverage lifetime energy consumption and connectivityrdquo EngineeringApplications of Artificial Intelligence vol 26 no 1 pp 405ndash416 Jan2013

[273] Y Wang M C Vuran and S Goddard ldquoStochastic performancetrade-offs in the design of real-time wireless sensor networksrdquo inProc IEEE International Conference on Computing Networking andCommunications (ICNC) Garden Grove USA Feb 2015 pp 931ndash937

[274] J Zhang F Ren S Gao H Yang and C Lin ldquoDynamic routingfor data integrity and delay differentiated services in wireless sensornetworksrdquo IEEE Transactions on Mobile Computing vol 14 no 2pp 328ndash343 Feb 2016

[275] J Li and G AlRegib ldquoNetwork lifetime maximization for estimationin multihop wireless sensor networksrdquoIEEE Transactions on SignalProcessing vol 57 no 7 pp 2456ndash2466 Mar 2009

[276] O Khatib ldquoReal-time obstacle avoidance for manipulators and mobilerobotsrdquo The International Journal of Robotics Research vol 5 no 1pp 90ndash98 Mar 1986

[277] M Zhang Y Shen Q Wang and Y Wang ldquoDynamic artificialpotential field based multi-robot formation controlrdquo inProc IEEEInstrumentation and Measurement Technology Conference (I2MTCrsquo10)Austin USA May 2010 pp 1530ndash1534

[278] Y K Liu and B Liu ldquoA class of fuzzy random optimization Expectedvalue modelsrdquoInformation Sciences vol 155 no 1-2 pp 89ndash102Oct 2003

[279] J Chi and Y Liu ldquoMulti-objective genetic algorithmbased on gametheory and its applicationrdquo inProc 2nd International Conference onElectronic and Mechanical Engineering and Information TechnologyShengyang China 2012 pp 2341ndash2344

[280] G EichfelderAdaptive Scalarization Methods in Multiobjective Opti-mization Springer-Verlag Berlin Heidelberg 2008

[281] Z Fei C Xing N Li and J Kuang ldquoAdaptive multiobjectiveoptimisation for energy efficient interference coordination in multicellnetworksrdquo IET Communications vol 8 no 8 pp 1374ndash1383 May2014

[282] G Hattab and M Ibnkahla ldquoMultiband spectrum access Greatpromises for future cognitive radio networksrdquoProceedings of the IEEEvol 102 no 3 pp 282ndash306 Mar 2014

[283] O B Akan O B Karli and O Ergul ldquoCognitive radio sensornetworksrdquo IEEE Network vol 23 no 4 pp 34ndash40 Jul-Aug 2009

[284] A Ahmad S Ahmad M H Rehmani and N U Hassan ldquoA surveyon radio resource allocation in cognitive radio sensor networksrdquo IEEECommunications Surveys and Tutorials vol 17 no 2 pp 888ndash917Second Quarter 2015

[285] S H R Bukhari M H Rehmani and S Siraj ldquoA survey of channelbonding for wireless networks and guidelines of channel bondingfor futuristic cognitive radio sensor networksrdquoIEEE CommunicationsSurveys and Tutorials vol 18 no 2 pp 924ndash948 Second Quarter2016

[286] M C Oto and O B Akan ldquoEnergy-efficient packet size optimizationfor cognitive radio sensor networksrdquoIEEE Transactions on WirelessCommunications vol 11 no 4 pp 1544ndash1553 Apr 2012

[287] M Ozger E A Fadel and O B Akan ldquoEvent-to-sink spectrum-aware clustering in mobile cognitive radio sensor networksrdquo IEEETransactions on Mobile Computing 2015 [Online] Availablehttpieeexploreieeeorgstampstampjsptp=amparnumber=7303936

[288] G A Shah V C Gungor and O B Akan ldquoA cross-layer QoS-awarecommunication framework in cognitive radio sensor networks for smartgrid applicationsrdquoIEEE Transactions on Industrial Informatics vol 9no 3 pp 1477ndash1485 Aug 2013

[289] V Esmaeelzadeh E S Hosseini R Berangi and O B AkanldquoModeling of rate-based congestion control schemes in cognitive radiosensor networksrdquoAd Hoc Networks vol 36 no 1 pp 177ndash188 Jan2016

Zesong Fei (Mrsquo07-SMrsquo16) received the PhD de-gree in Electronic Engineering in 2004 from BeijingInstitute of Technology (BIT) He is now a Professorwith the Research Institute of Communication Tech-nology of BIT where he is involved in the design ofthe next generation high-speed wireless communica-tion systems His research interests include wirelesscommunications and multimedia signal processingHe is a principal investigator of projects fundedby National Natural Science Foundation of ChinaHe is also a senior member of Chinese Institute of

Electronics and China Institute of Communications

Bin Li received the MS degree in Communicationand Information Systems from Guilin University ofElectronic Technology Guilin China in 2013 FromOct 2013 to Jun 2014 he was a research assistantin the Department of Electronic and Information En-gineering Hong Kong Polytechnic University HongKong He is currently pursuing the PhD degree inthe School of Information and Electronics BeijingInstitute of Technology Beijing China His researchinterests include wireless sensor networks cognitiveradio wireless cooperative networks physical layer

security and MIMO techniques

Shaoshi Yang (Srsquo09-Mrsquo13) received his BEngdegree in Information Engineering from Beijing Uni-versity of Posts and Telecommunications (BUPT)China in 2006 his first PhD degree in Electron-ics and Electrical Engineering from University ofSouthampton UK in 2013 and his second PhDdegree in Signal and Information Processing fromBUPT in 2014 He is now a Research Fellow inUniversity of Southampton From 2008 to 2009he was an Intern Research Fellow with Intel LabsChina working on the mobile WiMAX standard-

ization His research interests include high-dimensionalsignal processingfor communications green radio heterogeneous networkscross-layer in-terference management mathematical optimization and itsapplications Hereceived the prestigious Deanrsquos Award for Early Career Research Excellence atUniversity of Southampton the PMC-Sierra Telecommunications TechnologyPaper Award at BUPT and the Best PhD Thesis Award of BUPT He is a juniormember of Isaac Newton Institute for Mathematical Sciences CambridgeUniversity and a Guest Associate Editor of IEEE Journal on Selected Areasin Communications (httpsitesgooglecomsiteshaoshiyang)

Chengwen Xing (Srsquo08-Mrsquo10) received the BEngdegree from Xidian University Xirsquoan China in2005 and the PhD degree from University of HongKong Hong Kong in 2010 Since Sept 2010he has been with the School of Information andElectronics Beijing Institute of Technology BeijingChina where he is currently an Associate ProfessorFrom Sept 2012 to Dec 2012 he was a visitingscholar at University of Macau His current researchinterests include statistical signal processing convexoptimization multivariate statistics combinatorial

optimization massive MIMO systems and high frequency-band communi-cation systems Dr Xing is currently serving as an Associate Editor forIEEE Transactions on Vehicular Technology KSII Transactions on Internetand Information Systems Transactions on Emerging TelecommunicationsTechnologies and China Communications

Hongbin Chen was born in Hunan Province Chinain 1981 He received the BEng degree in Electronicand Information Engineering from Nanjing Univer-sity of Posts and Telecommunications China in2004 and the PhD degree in Circuits and Systemsfrom South China University of Technology Chinain 2009 He is currently a Professor in the School ofInformation and Communication Guilin Universityof Electronic Technology China From Oct 2006 toMay 2008 he was a Research Assistant in the De-partment of Electronic and Information Engineering

Hong Kong Polytechnic University China From Mar 2014 to Apr 2014 hewas a Research Associate in the same department His research interests liein energy-efficient wireless communications

Lajos Hanzo (Mrsquo91-SMrsquo92-Frsquo04) (F-REng F-IEEE F-IET F-EURASIP DSc) received his degreein electronics in 1976 and his doctorate in 1983In 2009 he was awarded an honorary doctorate bythe Technical University of Budapest and in 2015by the University of Edinburgh In 2016 he wasadmitted to the Hungarian Academy of ScienceDuring his 40-year career in telecommunicationshe has held various research and academic postsin Hungary Germany and UK Since 1986 he hasbeen with the School of Electronics and Computer

Science University of Southampton UK where he holds the chair intelecommunications He has successfully supervised 111 PhD students co-authored 20 John WileyIEEE Press books on mobile radio communicationstotalling in excess of 10 000 pages published 1100+ research contributionsat IEEE Xplore acted both as TPC and General Chair of IEEE conferencespresented keynote lectures and has been awarded a number of distinctionsCurrently he is directing a 60-strong academic research team working on arange of research projects in the field of wireless multimedia communicationssponsored by industry the Engineering and Physical Sciences ResearchCouncil (EPSRC) UK the European Research Councilrsquos Advanced FellowGrant and the Royal Societyrsquos Wolfson Research Merit AwardHe is anenthusiastic supporter of industrial and academic liaisonand he offers a rangeof industrial courses He is also a Governor of the IEEE VTS During 2008- 2012 he was the Editor-in-Chief of the IEEE Press and also a ChairedProfessor at Tsinghua University Beijing Lajos has 25 000+ citations Forfurther information on research in progress and associatedpublications pleaserefer to httpwww-mobileecssotonacuk

I Introduction

I-A Motivation

I-B Contributions of This Survey

I-C Paper Organization

II Related Surveys and Tutorials

III Fundamentals of WSNs

III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage

III-C2 Network Connectivity

III-C3 Network Lifetime

III-C4 Energy Consumption

III-C5 Energy Efficiency

III-C6 Network Latency

III-C7 Differentiated Detection Levels

III-C8 Number of Nodes

III-C9 Fault Tolerance

III-C10 Fair Rate Allocation

III-C11 Detection Accuracy

III-C12 Network Security

IV Techniques of MOO

IV-A Optimization Strategies

IV-B MOO Algorithms

IV-B1 Mathematical Programming Based Scalarization Methods

IV-B2 Nature-Inspired Metaheuristic Algorithms

IV-B3 Other Advanced Optimization Techniques

IV-C Software Tools

V Existing Literature on Using MOO in WSNs

V-A Coverage-versus-Lifetime Trade-offs

V-B Energy-versus-Latency Trade-offs

V-C Lifetime-versus-Application-Performance Trade-offs

V-D Trade-offs Related to the Number of Nodes

V-E Reliability-Related Trade-offs

V-F Trade-offs Related to Other Metrics

VI Open Problems and Discussions

VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

W IRELESS sensor networks (WSNs) consist of alarge number of compact low-cost low-power multi-

functional sensor nodes that communicate wirelessly overshort distances [1] [2] In WSNs the sensor nodes aregenerally deployed randomly in the field of interest which areextensively used for performing monitoring and surveillancetasks [3]ndash[5] Depending on the specific application scenariosWSNs may rely on diverse performance metrics to be opti-mized For example the energy efficiency and network lifetimeare among the major concerns in WSNs since the sensor nodesare typically powered by battery whose replacement is oftendifficult Furthermore the network coverage latency and thefairness among sensor nodes are important for maintainingthe quality-of-service (QoS) [6] [7] In practice these metricsoften conflict with each other hence the careful balancing ofthe trade-offs among them is vital in terms of optimizing theoverall performance of WSNs in real applications

In conventional WSN designs typically the most salientperformance metric is chosen as the optimization objectivewhile the remaining performance metrics are normally treatedas the constraints of the optimization problem Such single-objective optimization approaches however may be unfairand unreasonable in real WSN applications since it artificiallyover-emphasizes the importance of one of the metrics to thedetriment of the rest [8] Hence a more realistic optimizationis to simultaneously satisfy multiple objectives such as themaximal energy efficiency the shortest delay the longestnetwork lifetime the highest reliability and the most balanceddistribution of the nodesrsquo residual energy or the trade-offsamong the above objectives [9] [10] Accordingly multi-objective optimization (MOO) can be naturally adopted forsolving the above problem since it may be more consistentwith the realistic scenarios [11]

MOO algorithms have been a subject of intense interestto researchers for solving diverse multi-objective optimizationproblems (MOPs) in which multiple objectives are treatedsimultaneously subject to a set of constraints [12] Howeverit is infeasible for multiple objectives to achieve their re-spective optima at the same time thus there may not exista single globally optimal solution which is the best withrespect to all objectives Nevertheless there exists a setofPareto-optimal or non-dominated solutions generating a setof Pareto-optimal outcomesobjective vectors which is calledPareto frontfrontier (PF) or Pareto boundarycurvesurfaceExplicitly the PF is generated by the specific set of solutionsfor which none of the multiple objectives can be improvedwithout sacrificing the other objectives [13] This set of Pareto-optimal or non-dominated solutions constitutes the focus ofour interest and it is also called the Pareto-efficient set orPareto set (PS) that is mapped to the PF in the objectivefunction space [14]

Diverse approaches such as mathematical programmingbased scalarization methods and nature-inspired metaheuris-tics may be used for finding the PSs of MOPs Scalarizingan MOP means formulating a single-objective optimizationproblem such that optimal solutions to the single-objectiveoptimization problem are Pareto-optimal solutions to the MOP[15] In addition it is often required that every Pareto-optimal

solution can be reached with the aid of specific parametersof the scalarization Representatives of scalarization methodsinclude the linear weighted-sum method theε-constraintsmethod [15] and goal programming (GP) based methodsMOPs are more often solved by bio-inspired metaheuristicssuch as multi-objective evolutionary algorithms (MOEAs)[16] [17] and swarm intelligence based optimization algo-rithms (SIOAs) [18] MOEAs aim for finding a set of represen-tative Pareto-optimal solutions in a single run [14] [19][20]As a subset of MOEAs the multi-objective genetic algorithms(MOGAs) such as the strength Pareto evolutionary algorithm(SPEA) [16] and the non-dominated sorting genetic algorithm-II (NSGA-II) [21] have been particularly widely researched inthe family of MOO algorithms [22] because they are capableof efficiently constructing an approximate PF This is mainlydue to the fact that MOGAs accommodate a diverse varietyof bio-inspired operators to iteratively generate a populationof feasible solutions Compared to genetic algorithms (GAs)that rely on the interplay between genetics and biologicalevolution SIOAs seek to understand the collective behavior ofanimals particularly insects and to use this understanding forsolving complex nonlinear problems One of the most widelyused SIOAs is the ant colony optimization (ACO) algorithm[23] which has indeed been invoked for solving the MOPs inWSNs [24] Several other bio-inspired algorithms related toswarm intelligence will be surveyed in Section IV

B Contributions of This Survey

In this paper we focus our attention on various basic con-cepts conflicting performance criteriaoptimization objectivesas well as the MOO techniques conceived for striking a trade-off in the context of WSNs The contributions of our work arefour-fold which are listed as follows

bull We provide in-depth discussions on the basics metricsand relevant algorithms conceived for MOO in WSNs

bull We present a comprehensive coverage and clear classi-fication of various prevalent MOO algorithms conceivedfor solving MOPs and clarify the strengths and weak-nesses of each MOO algorithm in the context of WSNs

bull We provide an exhaustive review of the up-to-date re-search progress of MOO in WSNs according to differenttrade-off metrics

bull We highlight a variety of open research challenges andidentify possible future trends for MOO in WSNs ac-cording to the latest developments of WSNs

C Paper Organization

The reminder of this paper is organized as follows InSection II we summarize the related surveys of MOO inWSNs In Section III we commence with an overview ofWSNs in terms of their system model and applicationsFurthermore we introduce the main optimization objectivesof interest in WSNs In Section IV we present the family ofMOO techniques that can in principle be used for solving thiskind of problems In Section V we provide an overview ofthe existing studies dedicated to multi-objective methodsinWSNs Finally in Section VI we describe a range of open

System Model

Applications

MOO Metrics

Coverage

Network Lifet ime

Energy Consumption

Energy Efficiency

Network Latency

Number of Nodes

Fault Tolerance

Detection Accuracy

Network Security

MOO Algorithms

Sec VII Conclusions

Sec I Introduction

Software Tools

A System Model

Sensors13

Con13verter13

Radio Transceiver13

Power Supply13

Proxy Server

Gateway

Sensors

B Applications

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

Applications

Farming and

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

livingRoads and

Railway

buildings

Smart energy

meteringSmart

monitoring

C MOO Metrics

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

System Model

Applications

MOO Metrics

Coverage

Network Lifet ime

Energy Consumption

Energy Efficiency

Network Latency

Number of Nodes

Fault Tolerance

Detection Accuracy

Network Security

MOO Algorithms

Sec VII Conclusions

Sec I Introduction

Software Tools

A System Model

Sensors13

Con13verter13

Radio Transceiver13

Power Supply13

Proxy Server

Gateway

Sensors

B Applications

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

Applications

Farming and

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

livingRoads and

Railway

buildings

Smart energy

meteringSmart

monitoring

C MOO Metrics

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

A System Model

Sensors13

Con13verter13

Radio Transceiver13

Power Supply13

Proxy Server

Gateway

Sensors

B Applications

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

Applications

Farming and

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

livingRoads and

Railway

buildings

Smart energy

meteringSmart

monitoring

C MOO Metrics

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Proxy Server

Gateway

Sensors

B Applications

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

Applications

Farming and

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

livingRoads and

Railway

buildings

Smart energy

meteringSmart

monitoring

C MOO Metrics

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Applications

Monitoring

Tracking

Environment

Agriculture

Industry

Health care

Ecology

Smart house

Military

Industry

Public health

Ecology

Military

Applications

Farming and

monitoring

Commercial

property protection

and recovery

Self-healing

minefield

PinPtr-Sniper

detection

CenWits-search

and rescue

VigiNet

detection

ZebraNetLandslide

monitoring

Tsunami

monitoringPollution

monitoring

Volcanic

monitoring

Vineyard

monitoring

Artificial

retina

Patient

monitoringEmergency

response

SHIMMER

livingRoads and

Railway

buildings

Smart energy

meteringSmart

monitoring

C MOO Metrics

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Cv(x) =

xprime=0

xy (1)

g(xprime

yprime) =

0 otherwise(2)

fcon =

xtimesysum

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

QoS in WSN

Data Aggregation

and Fusion

Security

Cross-Layer

Design

Synchronization

Node SleepWake-Up

Scheduling

Controlled

Mobility

Network

Topology

Localization

Tnet = minTj (4)

E(P ) =

i times tm (5)

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

pi (6)

= ctimesN(uso usi)

prop N(uso usi) (7)

p((mn) (i j)) =

0 d((mn) (i j)) gt Rs(8)

|K| (9)

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Uα(x) =

logx α = 11

ximinusxlowast

xlowast

Source Node

Routing Path

Analysis

Base StationTraffic

Analysis

Compromised Node

is given by [84]

ek(u) = Ko(1 + αdpk) (12)

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

hj(x) = 0 j = 1 2 meq (14)

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

X X X X X X

X X X X X X X

Optimization

Optimizers

Algorithms

Evaluators

Simulators

middot Bio-mimic

middot Stochastic

middotMemory less

B MOO Algorithms

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Feasible region

minfi(x)

st fj le εj j 6= i

H(x) = 0

G(x) le 0 (15)

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Results output

values

criterion

difference vectors

Repeat

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

PA(t+ 1) =[c+ nA(t)]

Xid(t+ 1) = Xid(t) + Vid(t+ 1) (18)

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Population13

New population13

Final archive13

New archive13

Archive13

I13t13e13r13a

13t13i13o13n13

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

soma axon 1axon 0

biological neuron

f(x) outputsignal

artificial neuron

synapse

dendrites

weight

processingh(x)radio

filteringnode

sensor node

Initial state

Achieved reward

Taken action

New state

( )t t

r a s1( )ts

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Inference

Engine

Rule Base

Inputs

Outputs

Inputs Outputs

FuzzyFuzzy

C Software Tools

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

ParadisEO-MOEO

WWW-NIMBUS

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

general-purpose

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

VII C ONCLUSIONS

REFERENCES

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

I Introduction

I-A Motivation





III-A System Model

III-B Applications

III-C MOO Metrics

III-C1 Coverage














IV-B MOO Algorithms




IV-C Software Tools









VII Conclusions

References

Biographies

Zesong Fei

Bin Li

Shaoshi Yang

Chengwen Xing

Hongbin Chen

Lajos Hanzo

Documents

ACCEPTED TO APPEAR ON IEEE COMMUNICATIONS … · · 2016-09-15tutorial and survey of recent research and development efforts ... cording to the latest developments of WSNs