资源描述：
《基于一致性的分布式粒子滤波算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、AbstractDistributedParticleFilter(DPF)areapowerfulandversatileapproachtodecentralizedstateestimationinwirelesssensornetwork(WSN),theyareespeciallysuitedtolarge-scale,nonlinearandnon-Gaussiandistributedestimatedsystems.However,ithascertainlimitationsonthebandwidthof
2、communication,theresourceofnodes,thedynamicnetworktopologyandtheabilityofcommunicationandsoon.Inordertoreduceinfluencesofthelinksfailureorintermittentlinkthatexistinthenetworkforestimatingthetarget,thedeepstudyofdistributedparticlefilteralgorithmhasagreatsignifican
3、ce.Inviewofnon-linear,non-Gaussiantrackingapplicationinsensornetworks,thepaperproposesaconsensus/fusionbaseddistributedimplementationoftheparticlefilter(CF/DPF).Itrunstwoparticlefilters(PFs)ateachnode,theyarerespectivelylocalfilterwhichcomesfromthedistributedimplem
4、entationoftheparticlefilterandthefusionfilterwhichcomputestheglobalfilteringdistribution.ThepaperapproximatestheproductofthelocalprobabilitydensityfunctionswithGaussiandistribution,andtheaverageconsensusalgorithmisusedtocalculatetheparametersoftheGaussiandistributi
5、on,thenrealizethetargetstateestimation.thisalgorithmandthetrackingperformanceofcentralizedparticlefilteringwillbefinallycomparedWithMonteCarlosimulation,theyshowthatthealgorithmhasabetterfilteringperformance.Duetotheconvergencerateofconsensusisverycrucialintheabove
6、algorithm,sothepaperfocusesonprobability-basedtheweightoptimizationmethodofconsensus.Themethodintroducestheweightsoptimizationprobleminconsensusalgorithmsforspatiallycorrelatedrandomtopologies,itchoosestheconsensusmean-squareerror(MSE)convergencerateastheoptimizati
7、oncriterionandexpressesthisrateasafunctionofthelinkformationprobabilities,thelinkformationspatialcorrelationsandtheconsensusweights.BecausetheMSEconvergencerateisaconvex,non-smoothfunctionoftheweightforthesymmetricrandomnetworks,thepapergivestheclosedformandsub-gra
8、dientalgorithmsolutiontosolvetheproblem.andtheoptimizationmethodiscomparedwithotherweightselectionmethodbythesimulation,resultsshowthattheweightd
