JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.

ORG

Performance Evaluation of Route Cost for Wireless Sensor Networks with a Mobile Sink
Yarmuhammat Nizamudun, Naoshi Nakaya, Yukari Hagihara, Yuji Koui
ABSTRACT Recently Wireless Sensor Network (WSN) technology has been developing remarkably and its expected to be applied to many applications such as plant management, crime prevention, disaster prevention, medical treatment, traffic systems, and so on. In this article, we propose a Mobile Sink node control method for Wireless Sensor Networks. These WNS Networks are composed from two types of sensor node, one a fixed node which is immovable, and the other one a mobile sink node which is movable. The fixed nodes will form the clusters, and then the Mobile Sink node by using the Nearest Addition Method of TSP (Traveling Salesman Problem) which moves around the cluster center, decides the best rotation of the communication among the clusters, following which it decides the best fixed node to make the shortest distance of communication. The chosen best fixed node will transfer its data to the Mobile Sink node when it reaches them. Moreover the mobile Sink node is also capable to sense the possibility of communication of more than a few clusters in between, from then it would identify the best path for collecting the data. This method has proved its efficiency from the results of simulation and experiments that we have conducted and it would evaluate the performance of route cost for a mobile sink node. Index TermsWireless Sensor Network (WSN), route cost, mobile sink node, cluster, Traveling Salesman Problem (TSP).

1 INTRODUCTION
VER the past few years, these studies of wireless sensor networks are remarkably active, and these technologies have been applied to many methods [1],[2],[3],[4].Anumberofresearchersofmobiletechnol ogyaredevelopingrobotictechnologiesusingthemobile sensor networks where a large number of sensor nodes are difficult to deploy, such as plant management, crime prevention, disaster prevention, medical treatment, and trafficsystems,tomentionafew[2],[3].Suchsensornet worksarecomposedfromalargenumberofsensornodes and mobile sensor nodes. These mobile sensor nodes would move around the network under their own con trol,inordertocollectthedata. Infactthisisoneoftheproblemswhichneedtosolve since the mobile sensor nodes require large power for moving around the network. These sensor nodes have limitedpoweranddependonthemovingdistancewhich consumesitsenergy. So if the distance is longer it will need more energy. Hence, it has been very challenging to accumulate the datafromnetworkinthemobilesensorefficiently. For decades, a number of control methods for mobile

sinknodeinmobilesensornetworkshavebeenproposed. In[1],itwasinvestigatedaspecialofTSPNasthehigh levelrepresentationoftherobotroutingprobleminwire less sensor networks. And, in view of the strengths and weaknessesofexistingapproximationalgorithms,anovel EAbasedTSPNalgorithmwasproposed,whichcanpro duce significantly improved performance. In [2], it was proposed an effective mobile sensor control method DATFM for sparse sensor networks. DATFM uses fixed node and mobile node, and accumulates the acquired dataonafixednodebeforetransferringthemtothesink node. In addition, DATFM transmits the accumulated dataefficientlybyconstructingacommunicationrouteof mobilenodesbetweenfixednodes.In[5],itwasproposed mobilesensornetworkstocopewiththeplacementprob lem. Unlike conventional sensor networks, each sensor needstodynamicallychangeitslocationduringexplora tioninmobilesensornetworks.Inthisstudy,thedynamic sensorplacement(SDSP)isadesignedandimplemented scheme to realize the mobile sensor networks. In [7], the Incremental Deployment algorithm is proposed for mo bile sensor node. In the Incremental deployment algo rithm,theinformationofmoveplaceforthemobilenode is acquired from the information of other mobile nodes Y.Nizamudun is with the Graduate Institute of Engineering, Iwate Uniwhicharemovedatbeforethemobilenode.Andthemo versity,japan, 020-8551 bilesinknodemovestonewplace. N.N is with the Graduate Institute of Enginiring, Iwate University, japan, In this study, we proposed a more efficient control 020-8551. Y.H is with the Graduate Institute of Enginiring, Iwate University, japan, methodthatwehavedeveloped. 020-8551. Recognizing the fact that the Mobile sink node accu Y.K is with the Graduate Institute of Enginiring, Iwate University, japan, 020-8551. mulates the data by moving around the Wireless Sensor Networks,wecomposethetwotypesofsensornode;one isfixednodewhichisimmoveableandtheotheroneis

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

mobilesinknodewhichismovable. The fixed nodes will form the clusters which are al readyabletocommunicatewitheachothermutually,and thentheMobileSinknodedecidesfirst,thebestrotation of the communication among the clusters by using the Nearest Addition Method of TSP (Traveling Salesman Problem) which moves around the cluster center and se cond, it decides the best fixed node to make the shortest distance of communication. The chosen best fixed nodes willtransferthedatawhichhasbeencollectedfromother fixed nodes in the same cluster to the Mobile Sink node when it reaches them. Moreover, the sink node is also capabletosensethepossibilityofcommunicationofmore thanafewclustersinbetween,fromthenitwouldidenti fythebestpathforcollectingthedata. Uponthistheoryweconductedsimulationandexper iments to evaluate the performance of route cost for a MobileSinknode. Theremainingarticleisorganizedasfollows:Section2, explanationofthesystemmodel,Section3,explanationof our method. The results of simulation and experiments arepresentedinSection4,finallyourconclusionisinSec tion5.

nodeinsensingarea.Thesefixednodeshaveparame ter ( x, y ) oftheirsettingplaceandradiusrofthesensing area,asshowninFigure1.

Fig.1.Fixednode

Theclustersareformedbyfixednodeswhichhaveal ready communicated with each other. When these nodes form the cluster, they are judged distance d from other nodes. Forexample, d ij isthedistancefromnode i ( xi , yi ) to node

j ( x j , y j ) , dij

x x )
i j

( yi y j ) 2

AsshowninFigure2.

2 SYSTEM MODEL
Wecameupwiththetheoryofmonitoringavastarea bysmallnumberofnodesinshorttime,whichwouldbe applicable for areas such as dangerous to do activities. Forexample,whenanaccidenthappensinaplacelikea nuclearpowerplantIfthesensornodeshavemalfunction, it is difficult to access the nodes in a potential high radiationarea.Butwecanobtainthedatabycontrolling the mobile nodes toward the nodes located in high radiationareawithoutanyoneenteringthatarea.

Fig.2consistofcluster If the distance d ij < r, node i and node

j are can

communicateandtheyareformedonthesamecluster,as shown in Figure2(a). If the distance d ij > r they did not form on the same cluster, as shown Figure2 (b). In this same way, these N of fixed nodes form k of the clusters aftertheeachfixednodejudgesdistancefromotherfixed nodes. 3.3.2ClusterscenterandClustersorder Inthisproceedingarticle,thesystemdecidestheorder of cluster by using the Nearest Addition Method of TSP (TravelingSalesmanProblem)[8]whichmovesaroundthe clustercenter,decidesthewayofthebestrotationofthe communication among the clusters. Then all of the fixed nodes form clusters, the system decides center of each clusters by using parameters of nodes. Cluster cen ter ( xc , yc ) is:

EXPLANATION OF CONTROL METHOD

3.1 Fixed node

Fixed node is immovable. The fixed node can collect the data from other nodes. In addition, the fixed nodes can form clusters with other nearby nodes, they would communicate with each other so that when the sink node approaches it, it would transmit the accumulated data to the sink node.

3.2 Mobile sink node

A Mobile Sink node moves within the network area, at the same time it is capable of identifying the best routing order Also it chooses the ideal fixed node to collect the data from. In addition, the Mobile Sink node is able to collect the data from the sensing area of the fixed node without going through the center of it.

3.3 Theory Explanation

3.3.1 Cluster Inproceedingwiththisstudy,wearesettingNoffixed

xc n 1 1 xn , yc n 1 1 yn
n n

nisnumberofnodeincluster. When sink node decides the order of cluster, the mo

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

bilesinknodejudgesdistancesfromoneselftoeachclus ter center. First, the mobile sink node chooses one of the clustercenterswhichthedistanceisshortest.Ontheway, theyenteranotheroneoftheclustercenterswhichhasthe shortdistanceandcorrectstheorder.Intheend,theorder ofthemobilesinknodemovingaroundtheclustersinthe networkisdecided.AsshownFigure3(a),(b),(c),(d).

4.Theclusterheadofeachclusteriscorrectedonceagain. When cluster head of number K cluster is decided, the clusterheadofnumber(K+1)clusterisusedasthetarget point and the cluster head of each cluster is selected by usingstep2andstep3. Forexample,clusterC1consistedoffixednode(1,2,3, 4), and cluster C2 consistedof fixed node(5, 6, 7, 8). The clusterorderis1,2,3themobilesinknodechoosesone of fixed nodes for cluster head in cluster C1. When the mobile sink node decides cluster head for cluster C1, the centerofclusterC2ischosenfortargetpoint.

Fig.3clusterorder Fig.4settingclusterhead

3.3.3Clusterhead In this part, we explain that what decides the cluster head? In this study, the mobile sink node is moving aroundallclustersanddecidesthebestfixednodeforthe clusterheadtomaketheshortestdistanceofcommunica tioninnetwork.Itisdecidedthatthefixednodeforclus ter head. The cluster head will be collecting data from other fixed nodes at same cluster, and will transfer its data to the Mobile Sink node when it reaches them. We havealreadydecidedtheorderofclustersforsensing.But the clusters are formed from some fixed nodes; it is a problemofhowtodecidewhichtheclusterheadfromall ofthefixednodesinacluster. Themobilesinknodeselectsclusterheadbytheflow ingsteps: 1. When the mobile sink node decides cluster head of number K cluster, the center of number (K+1) cluster is usedasatargetpoint. 2. The distance from the mobile sink node to each fixed nodesofnumberKclusterandfromeachfixednodesof numberKclustertotargetpointofnumber(K+1)cluster isjudgedandthefixednodeofshortestdistanceischosen forclusterheadofnumberKcluster. 3. The sensing order of cluster head is selected by using theNearestAdditionMethodofTSP(TravelingSalesman Problem).

On the way, the mobile sink node judges the distance

d mio ( d mio d mi dio ). d mio Is the distance from mobile

sink node to each fixed nodes (1,2,3,4) of cluster C1 and from each fixed nodes(1,2,3,4) to center of cluster C2. d mi Is the distance from mobile sink node to fixed node

i of cluster C1. d io Is the distance from fixed node i of

clusterC1tocenterofclusterC2. Ontheway,thedistancefrommobilesinknodetocen terofclusterC2is d m1o , d m 2o , d m 3o , d m 4o :

dm1o dm1 d1o dm2o dm2 d2o dm3o dm3 d3o dm4o dm4 d4o
The mobile sink node judges the distance d m1o , d m 2 o ,

d m 3o , d m 4o ,andchoosesoneofthefixednodeswhichhas
theshortestdistanceinclusterC1.Thefixednode4ischo sen for cluster head in cluster C1, as shown in figure 4. Next step, the cluster head of next cluster is chosen by distancefromclusterheadofclusterC1tothetargetpoint. Atthattime,theclustercenterofclusterC3ischosenfor targetpoint,asshowninfigure5.Theparametersofclus terheadofclusterC1hasusedmobilesinknodesparam eters and the cluster head of cluster C3 is chosen for the targetpoint.

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

sinknodemovestotheclusterhead,itdoesnotgotothe clusterhead,andMobilesinknodecancollectdatafrom theclustercenterwhereitcancommunicatewiththeclus terheadatsensingarea. Whenthemobilesinknodemovestoclusterhead,the clusterheadofnextclusterisusedasatargetpoint,and the point ( x0 , y0 ) is decided that the distance is r from clusterhead.AsshownFigure7.

Fig.5settingclusterhead

After that, the mobile sink node judges the distance fromclusterheadofclusterC1toeachfixednodeofclus terC2andfromeachnodeofclusterC2tothetargetpoint of cluster C3. Finally, fixed node5 is the chosen cluster headforclusterC2.Onsameway,theclusterheadofeach clusterisdecided.AsshowninFigure5.
Fig.7settingthebestpointfordatacollection

MoreoverthemobileSinknodeisalsocapabletosense the possibility of communication of more than few clus tersinmiddlepoint,fromthenitwouldidentifythebest path for collecting the data. First, the mobile sink node judges the distance with each clusters center, when the mobilesinknodecansenseafewclustercentersatsame time, mobile sink node chooses and goes to the point in between the clusters. As shown in Figure 8 (a), (b). The pointsparameteris:

xo k 1 1 xk
k

yo k 1 1 yk
k

k isclusternumber.
Fig.6settingclusterhead

After that the system has corrected the cluster order once again for each cluster head of clusters by using the Nearest Addition Method of TSP (Traveling Salesman Problem).Whenmobilesinknodedecidedclusterheadof numberKcluster,theclusterheadofnumber(K+1)clus terisusedastargetpoint.Forexample,asshowninFig ure6,thetargetpointisnode5. 3.3.4Thebestpointfordatacollection Inthisarticle,aftertheclusterheadisdecided,themo bilesinknodechoosesthebestpointofeachclusterhead bycollectingdatafromeachfixednode.WhentheMobile
Fig.8thebestpointforclusters

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

3.3.5Explanationofthealgorithm The key theory is monitoring a vast area by small numbersofnodes.Figure9presentstheimageofthesys tem based on our theory; here we used ten fixed nodes (shownasfrom1to10)andonemobilesinknode.

InFigure10,thesimulationenvironmentisconsistedof 100offixedsensornodeandonemobilesinknode.Each fixed nodes have parameter of setting place and sensing area and can communicate with each other which is in sensingarea.

Fig.9thesystemimage

These fixed nodes parameters of the setting place are as follows ( x0 , y0 ) ( x1 , y1 ) ( x9 , y9 ) . The clusters from C1 to C5 are formed by fixed nodes which are already communicatedeachother,anditdecidesthecenterofthe clusterbyusingeachonesparameter.Thenthesinknode decidesthebestroutingorderasfollows: C1C2C3 C4C5

Fig.10Sensornodes

Theclustersareformedbyfixednodeswhichhaveal readycommunicatedeachother,anditdecidesthecenter ofclusterbyusingeachonesparameter,asshowninFig

Second the mobile sink node chooses one of the fixed nodesforclusterheadtocollectthedatafromeachcluster. Shownasnode1,3,5,7,9inFigure 9.Then,themobile sinknodechoosesthebestpointofeachclusterheadlike redpointinfigure8.Moreover,themobilesinknodecan communicatewithnode5andnode7atsametime,so,the bestpointischosenbetweennode5andnode7.

EVALUATION OF PERFORMANCE

4.1 Simulation for Control Method

To evaluate the performance of route cost of mobile sinknode,Weveconductedasimulationexperiment. The simulation was conducted using 100of fixed sen sor node and one mobile sink node in MATLAB, as showninFigure10. Settingparameters: Sensingarea 100m100m r f (Radiusofsensingareaforfixednode):6m

rm (Radiusofsensingareaformobilesinknode):6m

ure11.

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

Fig.11 Clusters

In Figure11, each fixed nodes of the same cluster is represented with the same color and different cluster is represented by a different color. Furthermore, these 100 fixed sensor nodes are forming 46 clusters. There are someclusters formedfrom onefixednodeorfromsome otherfixednodes. Moreover,theclustercenterpointofeachclusterisde cidedbyusingparametersofeachfixednodesinclusters. TheblackpointisshowninFigure11. Weve conductedthree types of simulations using 100 of fixed sensor node which are forming 46 clusters, and onemobilesinknodewhichare: One let the mobile sink node go through the network choosingthecenterofclustersonly.

The yellow line indicates simulation 3 and the red pointindicatesthebestpointforcollectingdata,asshown inFigure13.Intypethree,aftertheclusterheadisdecided, the best point is chosen where the mobile sink node can communicatewiththeclusterhead.Then,themobilesink node collecting data from each fixed node at the best pointofclusterheadineachclusterwhenthemobilesink node arrive the cluster. Thered point is shown as in fig ure14.

Fig13Simulation2

Fig.12Simulation1

AsshownInFigure12,theRedlineindicatesthesimu lation 1. After the cluster center is decided, the sensing orderofeachclusterisdecidedbyusingtheNearestAd dition Method of TSP (Traveling Salesman Problem) whichmovesaroundtheclustercenter. The mobile sink node collecting data from each fixed nodeattheclustercenterineachclusterwhenthemobile sinknodearrivedthecluster. Two let the mobile sink node go through the network choosing thebest fixed nodes for clusterhead from each clustertocollectthedata.AsshowninFigure13. Threeletthemobilesinknodegothroughthenetwork collecting the data from the sensing area of cluster head withoutreachingitscenter.AsshowninFigure14.

Fig14Simulation3

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

4.2 EvaluationofroutecostandEnergy
In this part, we have conducted to evaluate the route costofthesimulationexperimentformobilesinknodeby using50,100,150fixednodes. Thesimulationresultisshownasbelowtable1. Thedistanceofmobilesinknodeisshownforthethree types,intable1. Teble1 Themovingdistanceofmobilesinknode

2.Searchforasetofnhittingpoints(correspondedto ourbestpoint)basedonthepermutationfoundin Step1[1]. Aftertheorderofeachfixednodeisdecided,themo bilesinknodegoestothesensingpointofeachfixednode andcollecteddatafromthem.Theroutecostofthesimu lation experiment for mobile sink node is showed in fig ure16.
1000 900 800 700 600 500 400 300 200 100 0 10 50 100 node number 150

distance

EA-based type three

Throughthesimulationexperiments,theresultbyus ing the method of type three is the most efficient. Also, the energy consumption is the most important for WSN. In mobile sensor networks, we can compute the energy consumptionformobilesinknodebyusing:

Fig.16Evaluationofroutecost

E mS

(4.1)

E is energy, is coefficient of kinetic friction, m is mass of mobile sink node, S is the moving distance of mobilesinknode. ,misconstant,sorelationofenergy consumptionandmovingdistanceis:

Fromthesimulationexperiment,weknowthedistance of our algorithm is more efficient than EAbased TSPN algorithm.Ifthenumberofsensornodesisincreased,the movingdistanceofmobile sinknodeisalsoincreasedin EAbasedTSPNalgorithmatsamesettingarea.However, whenthenumberofsensornodesisincreased,therateof movingdistanceisgentleslopeinouralgorithm.

E S (4.2)
If the distance is longer it will need more energy. The energyconsumptionisshowedinfigure15.
800 700 600 energy 500 400 300 200 100 0 50 100 fixed node 150 Type one Type two Type three

CONCLUSION

Fig.15EnergyconsumptionofMobilesinknode

Thesimulationexperimentsshows,theresultbyusing the method of type three is the most efficient for energy consumptionofmobilesinknode. Also,weconductedtoevaluatetheroutecostofmobile sink node with EAbased TSPN algorithm [1]. In EA basedTSPNalgorithm: 1.ConstructaTSPtourbasedon(s,e1...en). eiiscenterofnodei,sissinknode.

In this paperwe have proposed the Mobile Sink node controlmethodforWirelessSensorNetworks.Thesenet worksarecomposedfromtwotypesofsensornode,one isfixednodewhichisimmovableandtheotheroneis mobile sink node which is movable. The fixed nodes will form the clusters; the Mobile Sink node decides the best rotation among the clusters, after it decides the best fixed node to make the shortest distance of communica tion.Thechosenbestfixednodeswilltransferitsdatato the Mobile Sink node when it reaches them. The mobile Sink node is also capable of sensing the possibility of communicationofmorethanfewclustersinbetween,and itwouldidentifythebestpathofcollectingthedata. Wehaveconductedsimulationexperimentstoevaluate theperformanceofroutecostforamobilesinknode. Based on the results, we came to the conclusion that the most efficient way is let the mobile sink node go throughthenetworkcollectingthedatafromthesensing areaoffixednodewithoutreachingitscenter.Thismeth od makes the route cost shorter than let the mobile sink nodegothroughthenetwork,choosingthefixednodein center of clusters only or choosing the best fixed nodes fromeachclustertocollectthedata.

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 2, FEBRUARY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG

In future studies, we would dig deeper into the con trolling method of more than few of mobile sink nodes. We are convinced that our study would lead to a new stageofWSNtechnology. REFERENCES
[1] BoYuan,M.Orlowska,S.Sadiq,OntheOptimalRobotRouting Problem in Wireless Sensor Networks, IEEE, VOL. 19, NO. 9, SEPTEMBER2007 K.Treeprapin, A.Kanzaki, T.Hara and S. Nishio, An Effective Mobile sensor Control Method for Sparse Sensor Networks, Sensors2009,9,327354 H.Nakayama,N.Ansari,A.Jamalipour,N.Kato,Faultresilient sensing in wireless sensor networks, computer Communica tions30(2007)23752384 T.SHINJO, S.KITAJIMA, T.OGAW, A.HARA ,S.NISHIO, Evaluation of a Node Movement Control Method using Push-based Broadcast in Mobile Sensor Networks, Journal of the DBSJ Vol.7, No.1 Self-organizing Dynamic Sensor Placement Scheme For Mobile SensorNetworks,M.Murase,http://www.sfc.wide.ad.jp/thesis/ 2002/master/masana/masana_thesis.pdf Distributed Sensing and Data Collection via Broken Ad hoc WirelessSensorNetworksofMobileRobotsInDistributedAu tonomousRoboticsSystems(2000),L.E.Parker,G.W.Berkey, andJ.Barhen,Eds.,vol.4,pp.273282 HOWARD,A.,MATARIC,M.J.,ANDSUKHATME,G.S.An IncrementalSelfDeploymentAlgorithmforMobileSensor networks.InAutonomousRobotsSpecialIssueonIntelligent EmbeddedSystems(ToAppear)(2002). Nearest Addition Method of TSP http://www-sys.ist.osakau.ac.jp/~umetani/lecture/uec2005/tsp_slide.pdf.

[2]

[3]

[4]

[5]

[6]

[7]

[8]

YarmuhammatNizamudunobtainedaMasterofEngineering(M.E) degree in 2009 from the department of Computer and Information Sciences,IwateUniversity,Japan.Cunrrently,heisaPh.D.candidate in Graduate School of Engineering, Iwate University. His research interests include Mobile Wireless Networks, Wireless sensor Net works,Cooperativecommunications,MultiChanneloperation,pat ternrecognitionanditsapplications. Naoshi Nakaya obtained a Ph.D. in Mathematical Science, Depart ment of Science and Engineering, Saitama University, Japan. Pres entlyheisworkingasanassistantprofessorintheIwateUniversity, Japan. His research interests are computer networks, wireless net workssecurity,computerviruses,andEvolutionaryalgorithms. HagiharaYukarigraduatedfromtheDepartmentofInformationon System Fundamentals, the University of ElectroCommunications, Japan. She is working as a technical staff in the Iwate University, Japan. Her research interests are wireless networks, mobile robot andcomputervision. Yuji Koi worked as Department Chief, Mitsubishi electric corpora tion,InformationTechnologyR&DAudioVisualInformationTech nology.HeobtainedaPh.D.degreeinInformationEngineeringfrom TohokuUniversity,Japan.Presentlyheisworkingasaprofessorin the Iwate University, Japan. His research interests are wireless sen sornetworks,wirelessnetworkssecurity,Unknowncomputervirus es,andRemotecontrolprotocols.