ClusterOptimizationinMobileAdHocNetworksBasedon …downloads.hindawi.com/journals/complexity/2020/2528189.pdf · 2020. 9. 28. · perform well when the MANET size becomes larger compared

Research ArticleCluster Optimization in Mobile Ad Hoc Networks Based onMemetic Algorithm memeHoc

Masood Ahmad1 Abdul Hameed2 Fasee Ullah 34 Atif Khan 5 Hashem Alyami6

M Irfan Uddin7 and Abdullah ALharbi8

1Department of Computer Science Abdul Wali Khan University Mardan Mardan Pakistan2Department of Computing and Technology Iqra University Islamabad Pakistan3Department of Computer and Information Science Faculty of Science and Technology University of MacauMacau China4School of Software Northwestern Polytechnical University China5Department of Computer Science Islamia College Peshawar Peshawar 25120 Pakistan6Department of Computer Science College of Computers and Information Technology Taif University Taif 21944 Saudi Arabia7Institute of Computing Kohat University of Science and Technology Kohat Pakistan8Department of Information Technology College of Computers and Information Technology Taif UniversityTaif 21944 Saudi Arabia

Correspondence should be addressed to Fasee Ullah faseekhangmailcom

Received 1 July 2020 Accepted 30 July 2020 Published 28 September 2020

Guest Editor Furqan Aziz

Copyright copy 2020 Masood Ahmad et al is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

In mobile ad hoc networks (MANETs) the topology differs very often due to mobile nodes (MNs) e flat network organizationhas high topology maintenance messages overload To reduce this message overload in MANET clustering organizations arerecommended GroupingMANET into MNs has the advantage of controlling congestion and easily repairing the topology Whenthe MANET size is large clustered MN partitioning is a multiobjective optimization problem Several evolutionary algorithmssuch as genetic algorithms (GAs) are used to divide MANET into clusters GAs suffer from premature convergence In this articlea clustering algorithm based on a memetic algorithm (MA) is proposed MA uses local exploration techniques to reduce thelikelihood of early convergence e local search function in MA is to find the optimal local solution before other evolutionaryalgorithms e optimal clusters in MANET can be achieved using MA for dynamic load balancing In this work the network isconsidered a graph G (V E) where V represents MN and E represent the communication links of the neighboring MNse aimof this study is to find the cluster headset (CH) as early as possible when needed High-quality individuals are selected for the newpopulation in the next generation New individuals are generated using the crossover mechanism on the chromosome once thetwo parents have been selected Data are communicated via CHs between other clusterse proposed technique is compared withexisting techniques such as DGAC MobHiD and EMPSO e proposed technique overcomes the state-of-the-art clusteringschemes in terms of cluster counting reaffiliation rate cluster life and overload of control messages

1 Introduction

A wireless sensor network for the Internet of thingsMANET is the set of MNs capable to share data with theirneighbors e MNs may generate their own data or theymay be received from other neighbors Several standardtechnologies for example IEEE 802154 ultrawideband

IEEE 802153 [1] IEEE 80211 [2] and Bluetooth [3]support MANET MANET permits to construct a short-term MANET for instant communication deprived ofsome fixed structures MANETmay be used for managingdifferent applications for example rescue flood moni-toring border monitoring disaster management andbattle field communication e clustering algorithms

HindawiComplexityVolume 2020 Article ID 2528189 12 pageshttpsdoiorg10115520202528189

perform well when the MANET size becomes largercompared to flat MANET regardless of routing methodadopted [4] e scalability issue in flat MANET is verycritical with a large number of node and the nodes aremoving in some directions Once the number of MNs inMANET with flat routing structure is x the complexity ofproactive routing scheme will be O (x2) [5] When thenumber of MNs in the MANET grow the routing over-head grows accordingly which can be calculated as thesquare the number of MNs e re-active routing algo-rithms also cause route setup delay when we increase thenumber of MNs in a MANET e flooding route request(RREQ) packets issue may also arise Hence to accom-plish elementary performance assurance in sized MANETa hierarchal organization is mandatory [6] e classicimplementation of hierarchal design is the clusteringstructure e selection of optimal CHs is an NP-hardproblem [7]

Designing a clustering algorithm able to route infor-mation with little effort is a demand of the day in MANETresearch Clustering is an important paradigm and itsimportance can be stated in two ways Firstly networkmanagement can be carried out effectively via the clusteringalgorithm A typical MANETcomprises more than hundredor even thousand MNs In flat MANET structure needlesspackets are initiated [8] e scalability issue may arise withflat-based MANETwhen we want to increase the number ofMNs in MANETand may saturate the network e MNs inMANETmay be static or mobile and handling the scalabilityis more challenging as compared to other networkserefore managing the MANET effectively is more im-portant To manage the MANET effectively utilizing theclustering mechanism becomes essential Secondly theproblems like controlling the topology construction ofvirtual network intrusion detection and routing can besolved with the help of clustering [9] All the subjects statedpreviously are committed based on well-structured MANETclustering

One of the imperative design consideration of a clus-tering algorithm is the discovery of an optimal CH set thatshould cover all parts of the MANET At one instant MNwill be the member of one cluster e existence of a CH ineach cluster is not mandatory However the existence of aCH in a cluster has the advantage of managing the MANETeasily and most of the protocols discussed in the literatureassume the selection of CHs e cluster formation shouldbe carried out in a way that the control overhead messagesmay decrease Otherwise the clustering algorithm will bemore energy consuming than flat architecture To findoptimal CH set optimization schemes such as particleswarm optimization evolutionary algorithms and neuralnetworks may be used [10]

e contributions are listed as follows

(i) In this paper the MANET is distributed intoclusters using an evolutionary memetic algorithm

(ii) e problem is formulated as a graph and thefitness function is tested according to networkrequirements

(iii) e proactive cluster-based routing scheme namedldquooptimal clustering in MANET based on memeticalgorithm memeHocrdquo is proposed

(iv) e algorithm works by selecting a set of CHswhere a cluster headset denotes a chromosome(solution)

(v) e quality of the chromosome is improved withlocal search method e (CH set) result is evalu-ated by a fitness function e parents are selectedfor reproduction based on the fitness value ecrossover and mutation are applied to generate theoffsprings e new solutions are generated untilthe optimal solution is found

(vi) e efficiency of the clustering algorithm is in-creased with memetic algorithm

(vii) e test outcomes demonstrate that the suggestedmethod has notable performance when comparedto existing techniques

e rest of the article is planned as follows Section 2discusses the existing work In Section 3 a detail descriptionof the basic memetic algorithm is presented in Section 4network formulation using the memetic algorithm is pre-sented and in Section 5 the performance evaluation isdescribed and finally the paper is concluded

2 Literature Review

e recent research on the cluster-based algorithm inMANET is presented in this section e clustering proce-dures are divided into the following classes

21 Clustering Based on Energy Efficiency e main goal ofthis technique is to minimize the power dissipationthroughout the cluster creation e paper [11] presents aclustering algorithm that works in a distributed mannernamed distributed cluster head scheduling (DCHS) algo-rithme purpose of the proposed scheme is to enhance thelifetime of MANET In this algorithm the MANET isgrouped into two tiers ie primary tire and secondary tiere tier formation is based on the received signal strength ofMN in MANET from the base station e algorithm per-forms the cluster head selection for the primary tier as well assecondary tier e authors claim that the proposed clus-tering scheme selects the CHs in a way that load balancingfactor may not compromise e benefit is to avoid orminimize calling reclustering procedure again and againe CHs are mainly selected based on the received signalstrength and remaining energy of mobile nodes

Findings the mobility of node which is a key parameteris not considered in this scheme e MNs with a largenumber of neighbors are the most suitable candidates forCH selection and it is also ignored in this paper Similarlythe communication load trust and reputation are alsoignored

e CHs are selected randomly in the first round in [12]e CHs in the next round are selected on the basis ofresidual energy

2 Complexity

Findings the output of the random selection of CHnodes in the first round may be unbalance partition eCHs may be selected from one part of the network and mayincrease the messages overhead e mobility and degree ofnode during the CH selection process is also ignored eprotocol may not perform well when we have a mobile basestation

22Mobility-Based Clustering Mobility of sensor nodes is akey to consider during the selection of CHs so as to increasethe lifetime of MANET e movement may be random orthe node may use other mobility patterns In leader-basedgroup routing the overhead of routing throughout routingpractice and resource requests are reduced significantly byassuming each set as a different element [13] To reduce theimpact of group variations on the performance of MANETthe resource assignments to every set separately is assumede contact opportunities of mobile nodes in delay tolerantnetworks are very less and to better utilize the rare contactoptions any node of a group within the transmission rangeof different group member exchanges data with other groupinstead of direct communication between group leaders

In the work proposed in [14] the CH nodes are selectedon the basis of node mobility e nodes having leastmobility are the best candidates for the role of CH eweighted clustering algorithm is adopted to perform theelection of CHse connectivity of CHs and their membersare also checked e authors claim that the proposedscheme outperforms in terms of energy and flexibilityagainst mobility

Findings e nodes with least mobility are the bestcandidates for CHs and the scheme may not perform wellwhen the overall speed of the network is high e nodeswith low speed will be disconnected after a while and thereclustering procedure will be called very frequentlyresulting in decreasing network lifetime Similarly the di-rection of nodes also plays an important role during the CHselection e CH with low mobility but different directionsto their neighbors may cause the disconnection of CH withtheir members e degree of node is also ignored in the CHselection process

23 Stable Clustering e CHs should be selected in such away to get more stable and balanced clusters in order todecrease the ripple effect of reclustering For this we need toconsider the parameters like the relative mobility degreeand the current energy of MNs By clustering in MANETsthrough neighborhood stability-based mobility prediction(MobHiD) the clusters that are extremely resilient to var-iations of topology are formed with right prediction of MHsmobility [15] Mobility measurement is based on the locationof its neighbors over time and does not require any specifichardware e correct estimation of future mobility isachieved by the strong correlation between MHs movementIt combines the features of highest degree and node mobilityprediction to form long life In the stable k-hop clusteringalgorithm for routing in MANET the clustering topologyoverhead by forming k-hop clusters is more stable than that

forming by one-hop clustering structure [16] For particularsituations two cluster heads are allowed in the k-hopneighborhood and this will be achieved with maintenancefunction In the k-hop neighborhood the diffusion of thecluster information is reduced by two round cluster headselection mechanisms

24 Optimization and Swarm Intelligence-Based ClusteringIn order to efficiently perform the parameters used for theselection of CHs in clustering algorithms play a very im-portant role In proficient bee colony clustering protocol(PBC-CP) the bee intelligence is used to select the CHs inMANET [17] e CHs are selected based on the currentenergy of node nodes degree and distance from the BS ealgorithm is further improved by choosing the shortest pathfor sending from and to the BS e path which requires lessenergy to transmit a packet is assumed as the shortest path inthis approach

Findings mobility is the main source of energy con-sumption in MANET because the topology changes veryfrequently when the nodes are highly mobile e mobilitymetric during the selection of CH is not assumed in thisprotocol e trust reputation and communication load ofnodes need careful attention running the clustering selectionalgorithm and are ignored in this paper e quality ofneighbors plays a very vital role when dealing with CHsselection e neighborrsquos response is not noted in theproposed scheme

Because of the unequal distribution of CHs in MANETthe dynamic particle swarm optimization-based fuzzyclustering algorithm was proposed [18] e authors claimthat the fuzzy cmeans in addition to PSO results more stableand balanced clusters in MANET An inference systemnamedMamdani fuzzy inference system is used to nominatethe CHse node current energy number of neighbors anddistance to BS are the parameters for the selection of a nodeas CH e parameters are the inputs in the inferencescheme described above e fuzzy instruction is optimizedusing the particle swarm optimizatione proposed schemeresult increases the lifetime of the network

Findings In latest technology we may deal with eitherstatic nodes or mobile nodes in MANET e proposedscheme does not assume the mobility of sensors whileelecting the CHs e factors assumed above are not suffi-cient to form balance clusters e ripple effect of reclus-tering will be high in the proposed scheme e reclusteringburden may decrease the lifetime of MANET

We have proposed a clustering scheme based on honey beeswarm optimization and genetic algorithm in [19] e pro-posed scheme forms stable clusters compared to the existingschemes An energy-efficient clustering in MANETs usingmultiobjective particle swarm optimization (EMPSO) whichmanages effectively the resources of MANET by finding theoptimum number of clusters in multiobjective manner [20]

25 Load Balancing through Evolutionary AlgorithmsClustering eproblem of dynamic load balancing (DLB) isfirst framed to dynamic optimization problem (DOP) [21]

Complexity 3

e use of different dynamic genetic algorithms developedfor DOPs is proposed to solve the balance clustering issue inMANET e fitness of a solution is examined on DLBmetric and every individual denotes a CH set with membernodes associated with each CH To control the environ-mental changing aspects numerous multipopulation algo-rithms colonists memory and permutations of all or someof these are incorporated in the SGA Cluster-based LBP ismodeled precisely to DOP and will be used by different AItechniques like artificial immune system and ant colonyoptimization Dynamic GAs are developed for load bal-ancing and used in this paper Dynamic test and networkenvironments are created and tested

In this approach [22] the compactness separation andoptimal number of clusters are optimized using the geneticalgorithm e weight function is developed based on someempirical analyses e weights between 0 and 1 can beassigned to each parameter e compact value should beminimized and the cluster separation should be maximizede optimal clusters are achieved

Findings the nodesrsquo distance and compactness areconsidered in CHs selection e mobility degree andcommunication load of sensor nodes are ignored in thisapproach

26 Hybrid Clustering To reduce the burden on the nodesnear base station (BS) the clustering algorithm that assumesthe energy constraint of MNs in MANETwas proposed [23]e clustering scheme focuses on the global optimization ofMNs energy e authors claim that the overloaded burdenon the MNs near to BS is reduced to the optimal level escheme uses timer-oriented competition approach for theselection of CH set e scheme topology overheads mes-sages of MNs In this way the network traffic may reducee distance of MNs and BS is taken into consideration toequally distribute the CH throughout the networkis waythe MNs may consume energy in a uniform fashion emember nodes join a cluster on the basis of communicationradius and CH set in order to guarantee the balancepartitioning

Findings CHs are rotated based on time interval and itis hard to know that the reelection is required based on timee nodes may be selected from one part of the networkemobility of nodes is not considered during the CH selection

To improve the productivity and ease the process ofcultivation the precision agriculture needs the services ofMANET is will improve farming techniques and pro-duction cost (labor) may decrease e MANET has widelybeen used for the purpose [24] e technical nature ofwireless sensor nodes makes it difficult to transmit thesensing data in a timely manner without delay e routingissues and short network lifetime may also arise due tolimited energy of nodes e sensor nodes near to basestation manage heavy burden of data coming from otherneighborsemain goal of this proposal is to select the CHsfrom centroid locations and the gateway nodes are selectedfrom each cluster based on their location e purpose ofgateway nodes is to reduce the burden on CHs e gateway

nodes forward data to and from other clusters e authorsclaim that by this way the load on the CHs will be equalecoverage of CH will also increase because the gateway nodescommunicate with other clusters on behalf of CH

3 Proposed Memetic Algorithm for MANETClustering memeHoc

To formulate the problem of dividing the network inclusters we assume N number of MNs in the network andthe task is to divide the MANET to k clusters In the pro-posed algorithm the network model is first presented andthen the dynamic and optimal clustering in MANET isformulated e problem of balanced cluster formation of nMNs is actually solved by constructing a mapping or findinga set of MNs (cluster heads) which shows the allocation of nMNs into k nonoverlapping clusters (C1 C2 Ck) androughly have equal size which the cluster head serves iscan be achieved bymodelingassuming theMANETas graphG (V E) and the edge E represents the communication linksbetween the MNs in the graph e wireless nodes (routers)are represented by the vertices V

A set of vertices (CHs) is selected optimally based on thefitness value of MNs for cluster heads role

e equation described below can be used to computethe weighting value of a MN

WFx WFEx + WFRMx + WFDx (1)

where WFEx is the weighting factor with respect toremaining energy of a MN x and computed as

AEx

1113944

n

x1REx

n

(2)

where REnrNodeiis the residual energy of a MN n is MNs in

MANET (total nodes) and AEx is the average energy of MNsat current A weight value with respect to energy can beextracted as follows the value ofWFEx will be 1 if the REnrxisgreater thanAEx its value will negative if the REnrx is less thanAEx if both cases are false the value of AEx will be zero

WFDx is the weight factor wrt a MNrsquos inner and outerdegreese weight value toWFDx can be assigned based onthe outer and inner degrees of a MN such as

AFDx

1113944

n

x1(OD + ID)x

n

(3)

where OD and ID are the outer and inner degrees of MN xAFDx is the average number of neighbors in MANET eweight value with respect to MN neighbors can be assignedin the following way if the degree of a MN x is greater thanAFDx its value will be 1 if the degree of a MN x is less thanAFDx its value will be minus1 otherwise its value will be zero

Correspondingly the weighting factorvalue of MN wrtmobility WFRMx can be computed by keeping in mind that

4 Complexity

MN has relative mobility or the stationary MNs are idealnominees for the role of CHe value ofWMobx will be 1 ifthe relative mobility of a MN x is nearly equal to that of itsneighbors otherwise its value will be minus1

e number of clusters k will be computed priorchoosing the cluster headset by the equality belowe valueof k is obtained as

k

1113944

n

x1(OD + ID)x

n+ 1

(4)

where the total number of clusters in MANET is k the totalnumber of MNs in MANET is represented by n and(OD + ID)x represents the neighborsrsquo information of a MNx

e cluster headset covers nodes as at least three hopsaway from other CHs When the weighting values of thewhole MANETnodes are computed the function describedin equation (5) will be applied to compute the suitability ofcluster headset

e proposal of the memetic algorithm (MA) for thedynamic and optimal clustering in MANET is illustrated inthis section e CH is denoted by chromosome (individ-ual) e cluster headsets are selected at random initiallye cluster headset is optimized using the objective functione fittest individuals are chosen for the next generation todevelop a new solution e parents for reproduction areselected from the population similar to the classical GAAfter the selection of two parents for the new population themutation and crossover are applied to generate new off-springs e new population is improved by applying thelocal search procedure In the memetic algorithm localsearch is performed to efficiently find the local optimumsolution and proceed for global optimum

e algorithm starts its operation by calculating theweights of every node Nodes with the higher weights areselected for initial population e fitness values of thepopulation are calculated e nodes to become the clusterhead are selected e probability of selection is calculatede local search is applied e fitness value of the node tosurvive in the population is calculated

e process of MA comprises numerous significantmodules such as (1) memetic representation (2)

initialization (3) fitness function (FF) assessment (4) se-lection methods (5) crossover and (6) mutation eprocess of the memetic algorithm to form balanced clustersis shown in Algorithm 1 Table 1 presents notations used inAlgorithm 1

31 Genetic Representation e conventional evolutionaryalgorithms such as genetic algorithm fail to search severalsolutions for the problem domain owing to its inherentquality of premature convergence A memetic algorithmuses a local search to reach its final destination withoutstopping on local maxima e first step in the memeticalgorithm is encoding the chromosome

311 Encoding and Population Initialization e set SCHi ofnodes are randomly selected from all the networks as clusterheads where i 1 2 k Each solution of the problem isthe set of cluster heads (SChi) In this way the cluster headsetcan be generated by a random combination of node IDsAnd the chromosome can be represented by a random set ofcluster head IDs Repetition of node IDs in a chromosome isstrictly prohibited e gene of a chromosome can berepresented by a single node ID Suppose we have a networkof 8 nodes e IDs of the nodes in the network would befrom 1 to 8 In this scenario the chromosome may berepresented by a random permutation 54673128 e nexttask is to extract a set of cluster heads Assume that the firstgene is added to the cluster headset en the second ID isalso addedeweights of the nodes are calculated by addingthe third node to the cluster headset and so on At each stepof adding nodes to the cluster headset the weights of thenode are updatede node having higher weight is replacedwith the node that has lower weight e discarded nodefrom the cluster headset is nomore considered to be a clusterhead in the current round After encoding the objectivefunction is evaluated for the fitness of the solution epopulation is initialized using the following procedure

32 Objective Function e quality of a solution or chro-mosome is evaluated by calculating the standard deviation ofthe fitness values of all nodes considered for the cluster headrole e fitness value of a solution can be calculated by

Minimize Ftn(Wf AFV) 1113944n

x11113944

k

y1RelWeixy WFx minus AFVy1113872 1113873

2

Subject to 1113944k

y1RelWeiy 11113872 1113873 for (y 1 k)

RelWeiy 1 or 0 for (x 1 n y 1 k)

(5)

e equation stated above is a minimization functionwhere the MNs in MANET are n k number of known or

unknown clusters will be designed WFx (i 1 2 3 n) isthe weight of nodei and AFVy is the average fitness value of

Complexity 5

MN to perform the role of CH e following equation canbe used to compute AFVy for a node x

AFVy

1113944

k

y1Re lWeixy lowastWFy

k

(6)

where the clusters in the MANET is k (solution) andRe lWeixy is the relationship weight of MN x and cluster ywhen the MN x is allocated to the cluster y (member node)the Re lWeixy will be assigned 1 else its value will be 0

After all the fitness function values are computed theprobability of choice Px for each CH can be computed by

Px WFx

1113944

k

y1WFy

(7)

where k is equal to total CHs WFx be the weight value ofnode y and local search is applied for optimal CH set in theregion of y by using

nodey(x + 1) nodey(x) + αxy lowast rz (8)

where αxy is the radius of node x with cluster y and rz is therandom variable within the range [minus1 1] to compute thefitness value by using equation (8) above Moving outside theradius of the target node is not permissible in the newpopulation

e cluster headset is selected based on the fitnessob-jective function e neighbor MNs join the nearest clusterheads to form clusters Once the clusters are formed variousMNs change their state to sleep mode and other nodes wakeup from time to time due to limited energy As a result thenetwork topology changes periodically e purpose of thisstudy is finding the set of cluster heads as soon as possiblewhen a topology change occurs

33 Selection is step is very important for improving thequality of the population With good selection high-qualityindividuals are passed to the next population for repro-duction e chromosome is selected on the basis of itsfitness value e selection method used in this work is the

pairwise tournament selection e selection is performedwithout replacement is method of selection is efficientand very simple e size of the tournament is set to 2 Alocal search is applied to extract high-quality genes from thepopulation

34 Local Improver In this step a local search is appliedwhen the chromosome is evaluated and selectedeweightsof the genes in the population are calculated e genes withhigher weights are searched in the population e pop-ulation is searched and the higher weight genes are replacedwith low weight genes Algorithm 2 improves the solutionlocally

35 Crossover and Mutation e significant memetic op-erators are crossover and mutation e offspring generatedfrom two predecessor chromosomes is known as a crossovere properties of the new generated chromosome areextractedinherited from all parts of its parents For thepurpose of this research the well-known method ldquox-order1rdquois used e offspring generated from only one chromosomeby changing one gene is known as mutation e mutationused in this research is substituting one gene in thechromosome

36 Accepting After the mutation and crossover the newchromosomes are placed in the population e chromo-somes with high quality are replaced with low-qualitychromosomes

37 Replace e new generated population is used for thefurther runs of the algorithm is process can be explainedby using Algorithm 3

38 Test e algorithm is iterated until the stopping cri-terion is met e best solution is returned after the lastiteration of the memetic algorithm

4 Experimentation and Results

To evaluate the performance of our proposed memeHoccluster configuration algorithm several simulation experi-ments were conducted on Eastnet In a square simulationarea of 1000mtimes 1000m the mobile nodes (MN) (50ndash500)are distributed randomly e simulation parameters weretaken from [19] e speed of the MN varies from 1 kmh to80 kmh Each MN is equipped with deaf omnidirectionalantennas e transmission range of each antenna differs inradius from 100m to 300m Each node has an archive(queuing mechanism) for incoming and outgoing packetsand keeps information about the mobility of their neighborse continuous bit rate was used to generate the origins ofthe traffic e threshold used for generating traffic was 20packets per second e simulations were run for 50 min-utes e average of 100 simulation runs is represented asgraphs Table 2 shows the parameters used during thesimulation as in [19]

Table 1 Notations used in Algorithm 1

Symbol DefinitionN Total number of nodesK Total number of clustersA Average value of network nodesrsquo degreev [n] Node IDs vectorFv Fitness value of cluster headsetTD Summation of all nodesrsquo degreeDegi Degree of node iCHs Cluster headsPop PopulationNew_pop New populationNew-CHs New cluster heads

6 Complexity

(1) Procedure memhoc clustering(2) Input n k(3) Output CHs(4) TD 0(5) for (i 1 ile n i + +) do(6) TDTD+Degi(7) end for(8) A TDn(9) k A

(10) for(j 1 jle k j + +) do initializes a random cluster headset permutation(11) CHs[j] rand(v[n])

(12) end for(13) pop[j] CHs[j]

(14) pop[j] Local Improver(CHs)(15) pop new[j] call function_replace(pop)

(16) fv [pop_new] calculate fitness (pop_new)(17) if(fv[pop_new]lt fv[pop])

(18) CHs pop_new

(19) Return CHs(20) end procedure

ALGORITHM 1 Psuedocode of memeHoc

(1) Procedure Local Improver(2) Input v [n] CHs k(3) Output CHs(4) for (i 1 i le n i++) do(5) w[i] xi(6) end for(7) for (i 1 i le n i++)(8) for (j 1 j le K j++)(9) if (w [i]gtw [j])(10) CHs [j] v [i](11) end if(12) end for(13) end for(14) return CHs(15) end procedure

ALGORITHM 2 Pseudocode of local improver

(1) Procedure Replace(2) Input pop CHs k(3) Output CHs(4) for (j 1 j le k J++) do Generate new population(5) New_CHs [j] rand (v [n])(6) end for(7) New_pop [j]new_CHs [j](8) for (p 1 j le k p++) do apply crossover and mutation(9) Temp pop [p](10) Pop [p]New_pop [j](11) New_pop [j]Temp(12) end for(13) m k2(14) New_pop [m] pop [m](15) return new_pop(16) end procedure

ALGORITHM 3 Pseudocode for function replace

Complexity 7

In Table 2 the first column represents the parameter andthe adjacent column shows the value ese are kept con-stant throughout all the experiments e number of nodesthe transmission interval and the speed have variedaccordingly

e performance of the memeHoc algorithm is com-pared with dynamic genetic algorithms for the DLB clus-tering problem (DGAC) [21] and energy-efficient clusteringin MANETs using multiobjective PSO (EMPSO) controloverhead computation [20] and a neighborhood stability-based mobility prediction algorithm for clustering theMANETs (MobHiD) [cluster lifetime re-affiliation] [15] Toevaluate memeHocrsquos performance the following metrics aretaken into account

Number of clusters the network is divided into anumber of virtual groups called clusters in cluster-based MANETs In this metric the number of clustersrepresents the number of virtual groups obtained afterrunning the cluster formation algorithm e leastnumber of clusters represents its stability Problemssuch as channel access planning frequency reusepower consumption and latency complexity can arisewhen we increase the number of clustersCluster life it is the period of time that a node isassigned the role of a cluster head (CH)e duration ofthe cluster represents its stability Mobility can affectthe life span of a cluster when neighbors have a differentmobility A clustering algorithm should select the CHsto increase the duration of the clustersAffiliation rate the number of affiliate and disaffiliatenodes as a CH or member during a given time intervalrepresents the affiliate rate In cluster-based routingone node joins the nearest CH and leaves the other Anode can reconnect when its CH no longer exists andanother node plays the CH role in the jurisdictionReaffiliation can occur when the CH is outside thetransmission range of a member node e duration ofa cluster will increase when the reaffiliation rate is lowand vice versaControl message overhead (CMO) exchanges ofpackets of numbers to keep information about topologychanges represent the control overhead Numerousexchanges of control messages (CMs) occur during thecluster building phase To measure the CMO metric

the number of CM sentreceived per unit of time iscounted

41Number ofClusters (NOC) Numerous experiments havebeen conducted to calculate the NOC metric when we in-crease the number of nodes from 50 to 500 e incrementalstep 50 is used to increase the number of nodes in the in-dividual simulation experiment e mobility modeladopted is RWP e speed of the MN varies in the rangefrom 1 kmh to 80 kmh ie from 1 kmh to 5 kmh forwalking from 5 kmh to 20 kmh for running and from20 kmh to 80 kmh for vehicle movement Likewise thetransmission range is 100 meters and 200 for different ex-periments To measure the NOCmetric the results obtainedduring the simulation are presented in the form of a graph inFigures 1 and 2 e curves presented in the figure show thenumber of clusters for networks of different sizes (ie50ndash500 knots) e number of clusters increases as thenumber of nodes increases as shown in Figure 1 e graphshows that memeHoc forms fewer clusters than MobHiDDGAC and EMPSO In memeHoc the number of clusters iscalculated based on the degree of the node Average degreesof nodes are calculated to calculate the number of clusters kerefore memeHoc has fewer clusters and the stability ofthe clusters can be guaranteed even if the degree of the nodeis taken into account After memeHoc DGAC performs wellcompared to other algorithms as shown in Figure 1 An-alyzing the arcs presented in Figure 1 we see that EMPSOhas full performance plus the number of clusters created byMobHiD is considerably less than EMPSOModHiD focuseson cluster stability by predicting future nodemobility duringcluster formation EMPSOrsquos attention to optimization andthe number of calculations become low e results showthat memeHoc outperforms the leading clustering algo-rithms in terms of counting clusters DGAC performs wellafter memeHoc To evaluate memeHocrsquos performance inmore detail a series of simulation experiments are con-ducted for a different transmission interval for example 200meters e results of subsequent experiments are illustratedin Figure 2 e curves in Figure 2 show that the radiotransmission intervals of the mobile nodes influence theperformance of the clustering algorithms e NOC metricdecreases considerably with high transmission intervalsWith a large transmission interval the CH will cover a large

Table 2 Simulation parameters

Parameters ValueData packet size 1400 bytesMobility speed 0ndash80 kmhFrequency (MHz) 2400Link bandwidth 11MbpsFrequency channel 3Simulation time VaryMax (x) 1000mMax (y) 1000mNode space VaryNumber of nodes 50ndash500

8 Complexity

area and the number of member nodes that a CH serviceswill increase therefore fewer clusters result e size of thebackbone network will decrease when we increase the radiotransmission radius of the nodes All algorithms work thesame way as in Figure 1

42 Cluster Lifetime In this section we evaluate the effect ofnode speed on the duration of CH Nodes with a highmobility ratio can reduce the duration of CH Emptying 100nodes in the network performs the simulation e trans-mission interval of each node is set to 200 in the first ex-periment e nodes move at a speed between 1 kmh and80 kmh e movable features are adopted as in [15] eduration of the CH is calculated in seconds Figure 3 showsthe results obtained during the simulation As the graphshows the proposed memeHoc aggregation algorithm se-lects the CH for a long time Stable (permanent) brackets areformed using the memeHoc algorithm as the relativemobility of the nodes and their neighbors are consideredduring CH selection As shown in Figure 3 the proposedmemeHoc exceeds the performance of DGAC MobHiD

and EMPSO in terms of screening stability MobHiD workswell after memeHoc because it assumes future mobility ofnodes during cluster formation e behavior of NMscannot be projected realistically in the long run In mem-eHoc group members remain associated with CH for a longtime e relative mobility of MNs is measured during theselection of CHs and MNs with high degree and relativemobility are the ideal candidates to become CHs

By evaluating the arcs shown in Figure 3 we examinethat memeHoc forms stable and sustained groups comparedto MobHiD EMPSO and DGAC select unstable CHEMPSO is better than DGAC erefore DGAC chooses anunstable CH when we increase the speed of MN Nodemobility is not considered during the CH selection processe radio transmission range of MN was increased to 300meters in the following experiment and the results areshown in Figure 4 All other parameters are constant (as inFigure 3) e graph shows that the radio transmissioninterval of MN significantly influences the performance of allclustered algorithms under consideration including ourproposed memeHoc scheme More stable ramps are ob-tained once by increasing the MN communication rangeClass and online life can increase e proposed memeHocworks well when we increase the delivery With a hightransmission interval the reaffiliation of the nodes decreasesand CH will cover a large number of nodes

43 Reaffiliation Rate (RR) A series of simulation experi-ments were performed to evaluate the performance of thebinding rate for the node speed and the radio transmissionrange e results are presented in the form of graphs egFigures 5 and 6 A new join may occur when the MN leavesthe current CH and joins another cluster or when the CHmoves across a member nodersquos radio transmission intervalNM can no longer communicate with its CH As the speed ofMNs increases new affiliation is more common when MNsleave their CH more often e mobility model chosen is arandom waypoint as in Section 42

e nodes move at a speed of 1 kmh to 80 kmh Severalexperiments are taken into account as an average of the

DGACMobHiD memeHoc

EMPSO

50 100 150 200 250 300 350 400 450 500Number of nodes

510152025303540455055

Num

ber o

f clu

sters

(clu

ster c

ount

)

Figure 1 Arithmetic mean of clusters vs size of MANET (trx range100m)

50 100 150 200 250 300 350 400 450 500Number of nodes

0

5

10

15

20

25

30

35

40

Num

ber o

f clu

sters

(clu

ster c

ount

)

DGACMobHiD memeHoc

EMPSO

Figure 2 Average number of clusters vs network size (trx range200m)

0 10 20 30 40 50 60 70 80Node speed (kmh)

10

20

30

40

50

60

70

Clus

ter l

ifetim

e (se

c)

DGACMobHiD memeHoc

EMPSO

Figure 3 Node speed vs cluster duration (trx range 200m)

Complexity 9

results e average connection rate over 100 differentexperiments is shown in Figure 5 As shown in the graphthe increase in reconnection rate is shown for memeHocMobHiD DGAC and EMPSO with an increase in node

speed e performance of memeHoc in arcs is betterbecause they have the lowest RR compared to otherweighted schemes e duration of CHs is long when thememeHoc clustering algorithm is started and this indi-cates that rejoining is small because the cluster mecha-nism is restarted less frequently when stable CHs areselected at is why CH is selected based on the residualenergy degree and relative mobility of the nodeerefore the duration of CHs increases and membernodes leave the existing cluster less frequently MobHiDperformance is better after memeHoc because MobHiDtakes into account future node mobility in the CH se-lection process MobHiD performance is better among allthe modern clustering algorithms studied as the selectedCH neighbors remain long and therefore the probabilityof reconnection is reduced e worst performancecompared to reconnection is in DGAC compared toMobHiD and EMPSO Node mobility is not taken intoaccount in DGAC when forming clusters ereforeunstable clusters may result e same simulation tests arerepeated for a network of different sizes ranging from50MN to 500MN e results are shown in Figure 6 as agraph An incremental phase 50 was used to evaluate theperformance of memeHoc in a network with differentnumbers of NMs e graph in Figure 6 shows that thereconnection rate decreases as the size of the mobilenetwork becomes large In a large network the CH serverhas many nodes and topology changes are transmittedless frequently e simulation area was the same as in theprevious experiment e curve at the bottom of the graphin Figure 6 shows that the reconnection rate decreases inmemeHoc as the network size increases in line with otherstate-of-the-art clustering algorithms In memeHoc CH isselected based on relative mobility the remaining energyand node degree are obtained and stable clusters areobtained If we have stable clusters the reunification rateis low

44 Control Message Overhead A number of packet ex-changes are required during the CHsrsquo selection process inMANET In this section the number of messages inter-changed throughout the cluster construction and mainte-nance phase is noted and presented e performance ofmemeHoc is compared with MobHiD EMPSO and DGAC

In this series of simulation experiments 50 MNs arerandomly distributed e simulation area is set to1000mtimes 1000m Randommobility of waypoints is acceptedin simulation tests e coverage area of MN radio trans-mission is set at 200 meters e nodes move at a speed of0ndash5 kmh for walking 5ndash20 kmh for running and20ndash80 kmh for a vehicle e results are shown in a lineargraph in Figure 7 e tests were performed for a further300m transmission range and the results are shown inFigure 8 As shown in the graph the DGAC control is toolow when the MNs are moving at low speed ie up to40 kmh and become high as MNs move faster e geneticalgorithm suffers from local maxima and unstable clusterscan occur when the node speed becomes high e same

0 10 20 30 40 50 60 70 80Node speed (kmh)

10

20

30

40

50

60

70

80

90

Clus

ter l

ifetim

e (se

c)

DGACMobHiD memeHoc

EMPSO

Figure 4 Node speed vs lifetime of clusters (trx range 300m)

0 10 20 30 40 50 60 70 80Node speed (kmh)

0

001

002

003

004

005

006007

008

009

01

Reaf

filia

tion

rate

DGACMobHiD memeHoc

EMPSO

Figure 5 Reaffiliation rate with different speeds

50 100 150 200 300 350300 400 450 500Number of nodes

002

0025

003

0035

004

0045

005

0055

006

Reaf

filia

tion

rate

DGACMobHiD memeHoc

EMPSO

Figure 6 Reaffiliation rate vs radio transmission range of nodes

10 Complexity

effect can be observed in all other schemes under consid-eration including memeHoc as the clustering procedure isoften called for high mobility Control overload may de-crease if the MN radio transmission range is increased Longtransmission distances may consume more power duringnetwork operations Prolonged network life may not beguaranteed if the MN transmission interval is high Once theoptimal CH has been found and the optimization techniquehas high overall control values more communication isrequired as shown in Figure 7 and 8 EMPSO and MobHiDprovide uniform performance when the MN radio beam is200 meters If the transmission range is large EMPSO andMobHiD have minimal environmental control e mem-eHoc algorithm we offer requires several messages duringcluster formation Metrics such as energy degree and rel-ative mobility are taken into account and accurate calcu-lation can increase the number of control messages eoptimal CH is calculated based on the higher congestioncost but once the optimal CH is determined the clusteringprocedure is called less frequently

5 Conclusions and Future Work

In this research the memetic algorithm is used to solve thecluster formation problem in MANET known as memeHocDue to its well-organized local search mechanism thememetic algorithm forms balance clusters efficiently A set ofsolutions (chromosomes) is randomly selected as initialpopulation A chromosome is a set of possible cluster headse population is optimized by using the local searchmechanism e fitness of the population is evaluated usingthe fitness function e parents are selected for reproductionif the solution is not optimal Mutation and crossover areapplied for diversity in the population resulting in newpopulation generation e process continues until the op-timal clusters are found e memeHoc is validated via detailsimulation study e performance is compared with otherrelated protocols such as DGAC MobHiD and EMPSOalgorithms e proposed solution gives satisfactory resultscompared to protocols under consideration e proposedmemeHoc performs well in almost every experiment

In future this work can be extended to vehicular ad hocnetworks e good performance with high-speed MANETsgives an indication that it is well suited for vehicular ad hocnetwork cluster formation It can also be used inMANETs ina distributed fashion for the same purpose

Data Availability

e data used to support the findings of this study are in-cluded within the article

Conflicts of Interest

e authors declare no conflicts of interest

Acknowledgments

is work was supported by Islamia College Peshawar andHigher Education Commission (HEC) of Pakistan

References

[1] IEEE 80215 working group for wpan 2011 httpwwwieee802org15

[2] IEEEComputer Society ldquoIEEE 80211wireless lanmedium accesscontrol (mac) and physical layer (phy) specificationsrdquo 2007

[3] Bluetooth official website 2011 httpwwwbluetoothcom[4] P Gupta and P R Kumar ldquoe capacity of wireless net-

worksrdquo IEEE Transactions on Information Aeory vol 46no 2 pp 388ndash404 2000

[5] E M Belding-Royer ldquoHierarchical routing in ad-hoc mobilenetworksrdquo Wireless Communications and Mobile Computingvol 2 no 5 pp 515ndash532 2002

[6] C E Perkins Ad-Hoc Networking Addison-Wesley BostonMA USA 2001

[7] S Basagni ldquoA generalized clustering algorithm for peer-to-peer networksrdquo in Proceedings of the Workshop on Algo-rithmic Aspects of Communication Bologna Italy July 1997

[8] N Shah S A Abid D Qian and W Mehmood ldquoA survey ofP2P content sharing in manetsrdquo Computers amp ElectricalEngineering vol 57 pp 55ndash68 2017

0 10 20 30 40 50 60 70 80Node speed (kmh)

0004

0006

0008

001

0012

0014

0016

0018

002

Con

trol m

essa

ges o

verh

ead

DGACMobHiD memeHoc

EMPSO

Figure 7 MANET MN speed vs message overhead (trx range 200meter)

0 10 20 30 40 50 60 70 80Node speed (kmh)

2

25

3

35

4

Con

trol m

essa

ges o

verh

ead

15

times10ndash3

DGACMobHiD memeHoc

EMPSO

Figure 8 MANET MN speed vs message overhead (trx range300m)

Complexity 11

[9] A H Azni ldquoCorrelated topology control algorithm for sur-vival network in manetsrdquo Lecture Notes in ElectricalEngineering pp 93ndash102 Springer Berlin Germany 2016

[10] K Deb K Sindhya and J Hakanen ldquoMulti-objective opti-mizationrdquo Decision Sciences Aeory and Practice pp 145ndash184 CRC Press Boca Raton FL USA 2016

[11] G Kannan and T Sree Renga Raja ldquoEnergy efficient dis-tributed cluster head scheduling scheme for two tieredwireless sensor networkrdquo Egyptian Informatics Journalvol 16 no 2 pp 167ndash174 2015

[12] T M Behera S K Mohapatra U C Samal M S KhanM Daneshmand and A H Gandomi ldquoResidual energy-basedcluster-head selection in wsns for iot applicationrdquo IEEE In-ternet of Aings Journal vol 6 no 3 pp 5132ndash5139 2019

[13] L F Xie P H J Chong and Y L Guan ldquoLeader based grouprouting in disconnected mobile ad hoc networks with groupmobilityrdquo Wireless Personal Communications vol 71 no 3pp 2003ndash2021 2013

[14] U Xie and R Leela Velusamy ldquoTEA-CBRP distributedcluster head election in MANET by using AHPrdquo Peer-to-PeerNetworking and Applications vol 9 no 1 pp 159ndash170 2016

[15] C Konstantopoulos D Gavalas and G Pantziou ldquoClusteringin mobile ad hoc networks through neighborhood stability-based mobility predictionrdquo Computer Networks vol 52 no 9pp 1797ndash1824 2008

[16] B Guizani B Ayeb and A Koukam ldquoA stable k-hop clus-tering algorithm for routing in mobile ad hoc networksrdquo inProceedings of the InternationalWireless Communications andMobile Computing Conference pp 659ndash664 DubrovnikCroatia August 2015

[17] A Pathak ldquoA proficient bee colony-clustering protocol toprolong lifetime of wireless sensor networksrdquo Journal ofComputer Networks and Communications vol 2020 ArticleID 1236187 9 pages 2020

[18] H Asmat F Ullah M Zareei A Khan and E M MohamedldquoEnergy-efficient centrally controlled caching contents forinformation-centric internet of thingsrdquo IEEE Access vol 8pp 126358ndash126369 2020

[19] M Ahmad A Hameed F Ullah et al ldquoA bio-inspiredclustering in mobile adhoc networks for internet of thingsbased on honey bee and genetic algorithmrdquo Journal of Am-bient Intelligence and Humanized Computing vol 2018pp 1ndash15 2018

[20] H AliW Shahzad and F A Khan ldquoEnergy-efficient clusteringin mobile ad-hoc networks using multi-objective particleswarm optimizationficient clustering in mobile ad-hoc net-works using multi-objectiveparticle swarm optimizationrdquoApplied Soft Computing vol 12 no 7 pp 1913ndash1928 2012

[21] H Cheng S Yang and J Cao ldquoDynamic genetic algorithmsfor the dynamic load balanced clustering problem in mobilead hoc networksrdquo Expert Systems with Applications vol 40no 4 pp 1381ndash1392 2013

[22] R Pal S Yadav R Karnwal et al ldquoEEWC energy-efficientweighted clustering method based on genetic algorithm forHWSNsrdquo Complex amp Intelligent Systems vol 6 no 2 2020

[23] X Wang Y Peng and L Huang ldquoAn improved unequalcluster-based routing protocol for energy efficient wirelesssensor networksrdquo in Proceedings of the International Con-ference on Intelligent Transportation Big Data amp Smart City(ICITBS) pp 165ndash169 Changsha China January 2019

[24] M Ahmad ldquoHoney bee algorithms based-clustering inmobilead hoc networkrdquo International Journal of Distributed SensorNetworks vol 13 no 6 pp 1ndash12 2017

12 Complexity