预览加载中,请您耐心等待几秒...
1/2
2/2

在线预览结束,喜欢就下载吧,查找使用更方便

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

基于KLD采样改进的高斯粒子滤波算法(英文) Introduction: Particlefilter(PF)isapowerfultechniqueforstateestimationproblems.Ithasbeenusedinawiderangeofapplications,includingrobotics,computervision,andsignalprocessing.Oneofthemainchallengesofparticlefilteristohandlethecurseofdimensionality,thatis,whenthestatespaceishighdimensional,thenumberofparticlesrequiredtoachieveacceptableestimationaccuracygrowsexponentially.Inrecentyears,severaltechniqueshavebeenproposedtoaddressthisissue,includingKullback-Leiblerdivergence(KLD)samplingandGaussianparticlefilter(GPF).Inthispaper,wepresentanovelapproachthatcombinesthesetwotechniquestoimprovetheperformanceofparticlefilter. KLDSampling: KLDsamplingisatechniquethatusestheKullback-Leiblerdivergencetoadaptivelydeterminethenumberofparticlesrequiredtoachieveadesiredlevelofaccuracy.ThebasicideaistoestimatethevarianceoftheimportanceweightsanduseittocalculatetheKLDbetweenthecurrentdistributionandthetruedistribution.IftheKLDisbelowathreshold,thealgorithmstops,otherwiseitgeneratesnewparticlesuntiltheKLDisbelowthethreshold. GaussianParticleFilter: GaussianparticlefilterisatechniquethatapproximatestheposteriordistributionbyaGaussianmixturemodel.TheparticlesaresampledfromtheGaussianmixturemodel,andtheweightsarecalculatedbyevaluatingthelikelihoodofeachparticle.ThemeanandcovarianceoftheGaussianmixturemodelareupdatedusingaweightedmeanandcovarianceformula. ProposedAlgorithm: TheproposedalgorithmcombinesKLDsamplingandGaussianparticlefiltertoimprovetheperformanceofparticlefilter.Thealgorithmconsistsofthefollowingsteps: 1.Settheinitialstateestimateandcovariance. 2.GeneratetheinitialsetofparticlesfromtheGaussianmixturemodel. 3.Calculatethelikelihoodofeachparticle. 4.CalculatetheimportanceweightsbasedonthelikelihoodandtheKLDsampling. 5.ComputethemeanandcovarianceoftheGaussianmixturemodelusingtheimportanceweights. 6.Resampletheparticlesusingasamplingschemebasedontheimportanceweights. 7.GeneratenewparticlesiftheKLDisaboveathreshold. 8.Repeatsteps3-7untilconvergenceoramaximumnu