particle-filters-豆柴文库

您所在位置：网站首页 / particle-filters.ppt / 文档详情

免费试读已结束，剩余 61 页请下载文档后查看

16 金币

下载文档

/ 66

下载提示

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

文本预览

ProbabilisticRoboticsSample-basedLocalization(sonar)Representbeliefbyrandomsamples
Estimationofnon-Gaussian,nonlinearprocesses

MonteCarlofilter,Survivalofthefittest,Condensation,Bootstrapfilter,Particlefilter

Filtering:[Rubin,88],[Gordonetal.,93],[Kitagawa96]
Computervision:[IsardandBlake96,98]
DynamicBayesianNetworks:[Kanazawaetal.,95]dWeightsamples:w=f/gImportanceSamplingwithResampling:LandmarkDetectionExampleDistributionsDistributionsThisisEasy!ImportanceSamplingwithResamplingImportanceSamplingwithResamplingParticleFiltersSensorInformation:ImportanceSampling
SensorInformation:ImportanceSamplingRobotMotionAlgorithmparticle_filter(St-1,ut-1zt):

ForGeneratenewsamples
	Sampleindexj(i)fromthediscretedistributiongivenbywt-1
	Samplefromusingand
					Computeimportanceweight
					Updatenormalizationfactor
					Insert
For
					Normalizeweights
drawxit-1fromBel(xt-1)Resamplingw2Algorithmsystematic_resampling(S,n):


For				Generatecdf
		
				Initializethreshold

For				Drawsamples…
	While()	Skipuntilnextthresholdreached
	
				Insert
Incrementthreshold

ReturnS’StartProximitySensorModelReminder23242526272829303132333435363738394041Sample-basedLocalization(sonar)InitialDistributionAfterIncorporatingTenUltrasoundScansAfterIncorporating65UltrasoundScansEstimatedPathUsingCeilingMapsforLocalizationVision-basedLocalizationUnderaLightNexttoaLightElsewhereGlobalLocalizationUsingVisionRobotsinAction:AlbertApplication:RhinoandAlbertSynchronizedinMunichandBonnLocalizationforAIBOrobotsLimitationsApproachesRandomSamplesVision-BasedLocalizationOdometryInformationImageSequenceResultingTrajectoriesResultingTrajectoriesGlobalLocalizationKidnappingtheRobotRecoveryfromFailureSummary