无线传感器网络节点的智能分布式定位(IJWMT-V5-N2-3)

I.J.WirelessandMicrowaveTechnologies,2015,2,22-36PublishedOnlineApril2015inMECS()DOI:10.5815/ijwmt.2015.02.03Availableonlineat*,AminaY.Al-SallutaaComputerEngineeringDepartment,IslamicUniversity,P.O.Box108,Gaza,Palestine.AbstractInwirelesssensornetworks,theissueofnodeslocalizationhastakenawideareaofresearch.Mostapplicationsneedtoknowthepositionofsensornodesforreasonsofoptimalandfastdatarouting.Inthispaper,anewdistributedlocalizationalgorithmbasedonSelfOrganizingMaps(SOMs)isproposedtodeterminethelocationofanodeinawirelesssensornetwork.Theproposedalgorithmisclassifiedasarange-freealgorithmwhichusesonlytheconnectivityinformationbetweennodeswithouttheneedtomeasurethetimeofarrivalorsignalstrengthasrange-basedalgorithmsrequire.Itutilizestheneighborhoodinformationandthewell-knownanchors'positionstocalculatetheestimatedlocationsofnodes.Ouralgorithmismadeupoftwomainstages.Theinitialestimatedlocationsofnodesarecalculatedintheinitializationstage,andfedtothelearningstageinwhichaSOMisusedtocalculatethefinalestimatedlocationsofnodes.Byusingtheneighborhoodinformationatthefirststage,thealgorithmhassignificantlyreducedtheSOMlearningtimeandthenumberofiterationstoconverge.Ontheotherhand,startingwithrealdataratherthanrandomdatamaximizedtheaccuracyoftheresultedlocations.Furthermore,thedistributedimplementationofthealgorithmhighlyalleviatedthepressureonthewirelessnodeswhicharecharacterizedwithlowpowerandlimitedcapabilities.TheproposedalgorithmhasbeenimplementedusingMATLABsoftwareandexperimentedbydeployingdifferentnumberofnodesinaspecificareawithdifferentcommunicationradioranges.Extensivesimulationsevidentlyverifiedtheperformanceofthealgorithmandachievedaverygoodaccuracy.Moreover,thealgorithmproveditseffectivenesswithaloweraverageerrorandlowernumberofiterationscomparedtootherrelatedalgorithms.IndexTerms:Wirelesssensornetworks,Localization,Anchornodes,Selforganizingmaps.©2015PublishedbyMECSPublisher.Selectionand/orpeerreviewunderresponsibilityoftheResearchAssociationofModernEducationandComputerScience*Correspondingauthor.E-mailaddress:[e-mail]utedLocalizationofNodesinWirelessSensorNetworks231.IntroductionAwirelesssensornetwork(WSN)isusuallyarelativelylarge-scalenetworkofinexpensiveenergyefficientdevices[1,2].ForanodeinaWSN,awarenessofitslocationandmaybethelocationofsomeothernodesiscrucialforasuccessfuloperation.Asacaseinpoint,routingdatainsensornetworksrequiresafinecooperationamongnodesinordertousesmallamountofenergyandtodeliverdataasfastaspossible.Anodecanchooseaproperwaytothedestination,usuallyasinknode,ifitknowsgeographiclocationofitselfanditsneighbornodes[3].Furthermore,mostapplicationsofsensornetworksneedtoknowthepositionofsensornodes.Forinstance,ajunglewatchingWSNmustfindoutandreportthelocationofaprobablefire[4].UsingGlobalPositioningSystem(GPS)devicesisthesimplestwaytodeterminethelocationofasensornode[5].Nevertheless,becauseofsometraitsofGPSdeviceswhichareincontrarywithsensornetworksdemands,usingtheminallsensornodesisnotjustifiable.Thesetraitsincluderelativelyhighcost,highweight,anddebatableaccuracyofGPSequipmentinsomesituations.ToovercomeGPSlimitations,manylocalizationtechniqueshavebeendevelopedforsensornetworkswhichdonotdependonGPSdevicesmerely.Intheselocalizationmethods,afewnodes,calledanchorsorseeds,areequippedwithGPSdevicesandhelpothernodestodeterminetheirpositions.ManyalgorithmshavebeenproposedforlocalizationofstaticWSNs[6-8].NodesinstaticWSNsdonothavemovement;inconsequences,ifanodeofthesenetworkscouldestimateitslocationonce,itwouldnothavetorepeatthelocalizationprocessagain.NodesinmobileWSNsmaymovebyanexternalagentlikewind,animal’smovements,streamofariver,orbyinternalmovementagentslikewheelsandcontinuoustrack[9,10].Mobilityhastwocontraryeffectsonlocalizationprocess.Inonehand,aspreviousworksindicate,mobilitycanhelplocalizationofstaticsensornetworks[11].Inthat,morenodescangetinformationfrommobileanchornodes.Ontheotherhand,mobilesensornetworksmaysufferfromrapidlychangingsituationswhichleadtolessvalidationtimeforobservedinformation.Ingeneral,thepreviouslocalizationalgorithmsconcentratedonstaticWSNsduetothehighimportanceofthisissue.Undermobilityconditions,astaticlocalizationalgorithmissupposedtobeapplicablewithsomesuperficialmodificationsandperiodicmobilityparameterstracking.Inthispaper,weconsidertheproblemoflocatingandorientingawirelessad-hocnetworkofunattendedsensornodesthathavebeendeployedinanareaatunknownlocations.Thebasicconceptistodeployalargenumberoflow-costandself-poweredsensornodesthatacquireandprocessdata.Thesensornodesmayincludeoneormoreacousticmicrophonesaswellasseismic,magnetic,orimagingsensors.Weconsiderlocationestimationinnetworkswhereasmallproportionofdevices,calledreferencedevicesoranchors,haveaprioriinformationabouttheircoordinates.Alldevices,regardlessoftheirabsolutecoordinateknowledge,estimatetherangebetweenthemselvesandtheirnei