自动移动无线自组织网络的状态估计

WSN 无线网络 GPS ZIGBEE PLC 智能 控制

JournalofComputerandSystemSciences77(2011)

305–321

ContentslistsavailableatScienceDirect

JournalofComputerandSystemSciences

doc.guandang.net

Automaticmobilitystatusestimationinwirelessself-organisednetworksRichardAnthony ,MonaGhassemian

DepartmentofComputerScience,TheUniversityofGreenwich,Greenwich,London,UK

articleinfoabstract

Wirelesssensornetworksandubiquitouscomputingarerapidlyincreasinginpopularity

anddiversity.Formanyapplicationsofthesesystemsthemobilitystatusofdevicesforms

partoftheoperatingcontextonwhichself-organisationisbased.

Thispaperdescribesanoveltechniquebywhichwirelessdevicessuchassensornodes

candeducetheirownmobilitystatus,basedonanalysisofpatternsintheirlocalneigh-

bourhood.TheSelf-DetectionofMobilityStatusalgorithm(SDMS)usesareinforcement

learninginspiredmechanismtocombinetheindicationsfrom vemobilitymetrics.For

manysystemsinwhichaneighbourtableismaintainedthroughregularstatusmessages

orotherinteraction,thetechniqueincursnoadditionalcommunicationoverhead.Thetech-

niquedoesnotrequirethatnodeshaveadditionalinformationsuchasabsoluteorrelative

locations,orneighbourhoodnodedensity.

Theworkconsiderssystemswithheterogeneoustime-variantmobilitymodels,inwhich

asubsetofnodesfollowsarandomwalkmobilitymodel,anothersubsetfollowsaran-

domwaypointmobilitymodel(i.e.hasintermittentmovement),somenodeshavegroup

mobilityandthereisastaticsubset.

WesimulatetheseheterogeneousmobilitysystemsandevaluatetheperformanceofSDMS

againstanumberofmetricsandinawidevarietyofsystemcon gurations.

©2010ElsevierInc.Allrightsreserved.Articlehistory:Received1June2009Receivedinrevisedform18September2009Availableonline1February2010Keywords:WirelesssensornetworksSelf-detectionofmobilityMobilitymodelsandmetricsReinforcementlearningLévyFlights

1.Introduction

Wirelesssensornetworks(WSN)andubiquitouscomputingarerapidlyincreasinginpopularityanddiversity.Bothtypesofsystemcompriseamixtureof xedandmobile(ormovable)devices.Whilstthetwotypesofsystemhavedifferentcorefunctionalities;theformerbeingprimarilyconcernedwithsensingsomecharacteristicsoftheenvironmentinwhichitisdeployed,andthelatterbeingconcernedwithusingtheenvironmentasameansofcontextualisation;theybothareincreasinglyconcernedwithmobiledevices.Thereisalsoacleartrendinincreasingnumber,varietyandcapabilitiesofmobilecomputingdevicesand,asaresult,thereisincreasingdiversityinthewaythesedevicesareused.Formanyapplicationsofthesesystemsthemobilitystatusofdevicesformspartoftheoperatingcontextonwhichself-organisationisbased.Detectingandclassifyingmobilityiscomplexasthereareawidevarietyofmobilitycharacteristics,andseveralmodelsfordescribingpatternsofmobilityarepopular,althoughnotnecessarilysu cientlyrepresentative.

TraditionallysensornodesinWSNareconsideredtohave xedmobilitycharacteristicsforthegreatmajorityofimple-mentations.Inapplicationswheresensorsarephysicallyattachedtotheinfrastructure(e.g.monitoringvibrationinabridge)itisadequatetoassumenomobility.Thestaticmodelmaybeexpandedbyintroducingmobilityinthreelevels[1]ofthesensornetworksystem.Itcanbethecasewherethesensorsthemselvesmoveandcollectinformation,theeventmonitored*Correspondingauthor.

E-mailaddresses:R.J.Anthony@gre.ac.uk(R.Anthony),M.Ghassemian@gre.ac.uk(M.Ghassemian).

0022-0000/$–seefrontmatter

doi:10.1016/j.jcss.2010.01.009©2010ElsevierInc.Allrightsreserved.