无线传感器网络中基于神经网络的移动汇聚路径优化(IJWMT-V7-N4-1)

I.J.WirelessandMicrowaveTechnologies,2017,4,1-13PublishedOnlineJuly2017inMECS()DOI:10.5815/ijwmt.2017.04.01Availableonlineat(numberofrounds)isincreased.Increasedlifetimealsohandlestheproblemofhotspots.IndexTerms:Wirelesssensornetwork,trainedneuralnetwork,stoppoints,clusterhead.©2017PublishedbyMECSPublisher.Selectionand/orpeerreviewunderresponsibilityoftheResearchAssociationofModernEducationandComputerScience1.IntroductionAwirelesssensornetworks(WSN)comprisesofsmallsensorswhicharesuppliedpowerthroughbatteries,whichhelpinwirelesscommunicationintheareasevenwherethehumanscan’treach.Sensorsaredeployedinthefieldtobesensed,tosenseeventslikefire,volcaniceruptionsetc.Thesensorscontinuouslykeeptrackontheeventsandrecordthemandkeepsendingthedatatospecificnodeknownassink.Thesinkthenpassestheinformationtotheobserverviainternet.[4]Sensor,memory,actuatorandaprocessortogetherformasensornode.Communicationofthenodesisthroughwirelessmediumthatisinfrared,radiofrequencies.Deploymentofnodesisrandom.Nodescancommunicatedirectlywitheachotherandiftheycannotdosothentheycommunicateviaothernodes.Thisisknownashoping.Oneofthemajorproblemsfacedbythesensornodesisitsbatterylifetime.Powerprovidedtothesensorsisbythedcpowerbatterieswhichdieoutafterthecertain*Correspondingauthor.E-mailaddress:2MobileSinkPathOptimizationforDataGatheringUsingNeuralNetworksinWSNusage.Thislimitstheuseofsensors.Toachievethemaximumuseofsensorsvariousschemeshavebeenadopted.Wirelesssensornetworksarecategorizedintwotypeslistedbelow:HomogenousNetworks:Thesensornetworkswhichhavesimilartraitsthatispower,processing.Eachnodehasequalload.HeterogeneousNetworks:Thesenetworksdonothavesamecharacteristics.Somenodesarestaticwhileothersareadvancedhavingdifferentproperties.[1]Networkofneuronsisknownasneuralnetworke.g.Humanbrain.Artificialneuronsaredevelopedonthebasisofrealhumanneurons.Thesecanbeeitheraphysicaldeviceorpurelyamathematicalfunction.Inaneuralnetworktheprocessingiscarriedoutinparallel.Itconsistsofmanysimplerprocessingelementsthatareconnectedinaparticularmannersoastoperformaparticulartask.Neuralnetworksweredevelopedbecauseofhighlypowerfulcomputations,noiseandfaulttolerant,highdegreeofparallelism,lowenergyconsumption,easytotrain.Aneuralnetworkisexplainedbythreeparameters.Firstoneistheinterconnectionbetweenthedifferentlayers.Secondistheupdatingoftheweightsandfinally,thefunctionwhichisusedtoconverttheinputtooutput.Inaneuralnetwork,numberofinputsarefedtoinputlayerwhichafterbeingsolvedwithweightsarefedtoahiddenlayer.Inthishiddenlayertheactivationfunctionactsontheinputandproducestheoutput.Thisisalsoknownasneuralnetworklearningorneuralnetworktraining,whichisexplainedlaterinthetextTherearethreetypesoflearning’sclassifiedbelow:SupervisedLearning:Inthisinputandtargetsarealreadyknownandarelationshipisdevelopedbetweentheinputandoutput.UnsupervisedLearning:Itisfasterthansupervised.Inthistypeoutputisnotlinkedwiththeinput.ReinforcementLearning:Inthistypeoflearningweightsadjustmentisnotdirectlyrelatedtoerrorvalue.Forshufflingofweights,errorvalueisusedrandomly.Intheworkcarriedout,theclusterheadsareidentifiedconsideringnodedensity,minimumdistance,andcountofhops.Then,othernodesjointheclusterheadsnearbybycalculatingminimumdistancebetweenthemandclusterheads.Atrainedneuralnetworkidentifiesthebestpathtobefollowedbythemobilesinkforcollectingdata.SectionIIincludesbriefinterpretationofreviewprocess.Itgivesinformationaboutvarioustechniquesusedfordatacollection.ExperimentalsetupisexplainedinsectionIII.SectionIVgivesaccountofproposedwork.SimulationandanalysisarementionedinsectionV,followedbyconclusioninsectionVI.2.RelatedWorksC.Zhu[1]developedtreeshavingrootnodesidentifiedasrendezvouspointsandsomespecificnodes,tolowerloadonrootnodes,knownassub-rendezvouspoints.TheseRP’sandSRP’sactasstoppointforthesinkwhilecollectionofdata.Thisalgorithmnotonlybalancedtheloadbutalsoreducedenergyconsumptionincreasingthelifetimeandeliminatingthehotspotproblem.C.Wang[4]carriedoutworkonsinkrelocation.Thenodesnearthesinkexperiencetoomuchloadsoastoreduceburdenonthemsinkisrelocatedtodifferentplaces.Inthismechanismtheinformationusedisabo