文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 自然科学 > 天文/地理 > An optimal adaptive Kalman filter
侵权投诉

An optimal adaptive Kalman filter

自适应卡尔曼滤波算法

178

whereαrangevector,k(0<¯αk≤1)isanadaptivefactor,λkisaLag-1997;kisanequivalentcovariancematrixofl(Yang1991,

Yangetal.2001b).Theestimatorfollowsk(Yangetal.2001b)

x k=(I K¯kAk)x¯k+K¯klk,(8)with

x k=(I K¯kAk) x

¯k/αk,(9)

whereK

¯kisagain matrix.K¯11 1k=k x¯TkTkAAk x¯kAk+

¯k(10)

k

Anequivalentestimatoroftheadaptivelyrobust lteris(Yangetal.2001a)

x k=(ATk

P¯kAk+αkPx¯k) 1(ATk

P¯klk+αkPx

¯kx¯k),(8a)whereP

¯kisanequivalentweightmatrixoflkPx¯k= x¯1

andkistheweightmatrixofthepredictedstatevector(Yang1997;Yangetal.2001a).

Afteradaptiverobust lteringwasdeveloped,threelearn-ingstatisticsandthreeadaptivefactorshavebeensetupbasedonexperience,andhaveproveneffectiveinpracticalappli-cations.Anaccompanyingadaptivefactorwithathree-seg-mentdescendingfunctionandalearningstatistic,constructedbyusingthediscrepancybetweenthepredictedstatefromthekinematicmodelandthestateestimatesfromthemea-surements,havebeenestablished.Twootherkindsofadap-tivefactorshavebeendeveloped,whichareatwo-segmentdescendingfunction(Yangetal.2001b)andanexponentialfunction(YangandGao2005).Twoadditionallearningsta-tisticshavealsobeensetup,whichareapredictedresidualstatistic(XuandYang2000;Y.YangandW.Gao2005,sub-mitted)andavariancecomponentratiostatisticfromboththemeasurementsandthepredictedstates(YangandXu2003).Someotheradaptivefactorsmaybedeveloped.Thequestioniswhatrulesshouldbefollowedorwhetherthereexistsomeoptimaladaptivefactors.

Ingeneral,anoptimaladaptivefactorshouldsatisfytheconditionsthatthetheoreticaluncertaintyofthepredictedstateoutputfromtheadaptive lteringequalsornearlyequalsitsactualestimateduncertainty,and/orthetheoreticaluncer-taintyofthepredictedresidualvectorequalsornearlyequalsitsactualestimateduncertainty.Theapproximateuncertain-tiesofthepredictedstatevectorx¯ualvectorv¯kandthepredictedresid-kcanbederivedbyusingactualmeasurements,especiallyinGPSnavigationorintegratednavigation,whereprecisemeasurementscanbeobtained.Thus,toobtainanoptimaladaptivefactorbecomesthedeterminationoftheactualuncertaintyofthe lteroutputs.

Todeterminetheactualuncertaintyofthe lteroutputs,weshouldhavetheknowledgeofthedegreeofactualnoisedispersionornoisemagnitude.Theproblemis,however,theestimatedstatesofamovingvehiclecannotbecomparedwithitstruemovementtrace,becausethetruetrajectoryofthemovingvehicleisunknown.Thus,theactualuncertaintiesofthecorrespondingnoisescannotbeobtained.Additionally,

theestimatedstatesareaffectedbythemeasurementerrorsandthekinematicmodeldisturbances.Theoptimaladap-tivefactorsbasedontheconditionsmentionedabovecannotbeobtainedintheory.Inpractice,however,wecangettheapproximateestimatedcovariancematricesofthepredictedresidualvectorandthepredictedstatevector,fromwhichthecorrespondingoptimaladaptivefactorscanbederived.

Assumethatthemeasurementsareaccurate.Ifnot,thenthedetectionandidenti cationandadaptationprocedurecanbeemployed(Teunissen1990)orarobustKalman ltercanbeapplied(KochandYang1998;MasreliezandMartin1977;Schaffrin1991,p.32–34;TsaiandKurz1983;Yang1991,1997;Zhouetal.1997).Inthiscase,thepredictedresidualvectormainlyre ectsthemagnitudeofkinematicmodelerrors,andthe ltercanprovideitstheoreticalcovari-ancematrixbythecovariancepropagationlaw.Anapproxi-matecovariancematrixforthepredictedresidualvectorcanbeevaluatedfromitscorrespondingvalues.Comparingthedifferencebetweenthecalculatedcovariancematrixandtherelatedtheoreticaloneofthepredictedresidualvector,wecandetermineanadaptivefactor.

Ifthemeasurementsthemselvescanprovidetheesti-matedstatevector,thenbycomparingthepredictedstatefromthekinematicmodelandtheestimatedstatefromthemeasurements,wecanobtainthedifferencesbetweenthem,andinturn,estimatetheapproximatecovariancematrixofthepredictedstatevectorfromthedifferences.Byapplyingthecovariancepropagationlaw,weobtainthetheoreticalcovari-ancematrixofthestate.Thus,anotheroptimaladaptivefac-torisderivedbycomparingthetheoreticalcovariancematrixofthepredictedstatevectorandtheevaluatedone.Matchingtheestimatedcovariancematricesofthepredictedstateandthepredictedresidualvectorsandtheirtheoreticalmatrices,wecanderivetheadaptivefactors.Intheactualderivations,wematchthetracesofthesecovariancematricestogettheuni edadaptivefactorssinceitismuchsimplerandeasierforcomputation.

2Optimaladaptivefactorexpressedbyestimatedcovariancematrixofpredictedresiduals

Ifthemeasurementvectorlkatepochtkisreliable,thenthepredictedresidualvectorv

¯k=Akx¯k lk,(11)

re ectstheerrormagnitudeofthekinematicmodelinforma-tion.FromEq.(11)andthecovariancepropagationlaw,wegetthetheoreticalcovariancematrixofthepredictedresidualvector

v¯AT

k=Ak x¯kk+ k.

(12)

Inthecaseofadaptive ltering,ifthekinematicmodelerrorisx¯large,theweightmatrixPx¯kofthepredictedstatevectorkshouldbedecreased,i.e.,theweightmatrixPx¯kofthepredictedstatevectorshouldbecomeαetal.2001a).FromEqs.(7)or(10),thekPxcovariance¯k(αk≤1)matrix

(Yang

上一页第2页

你可能喜欢

热门文档

相关文档

  • kalman_filter_for_INS_Gps

    The purpose of an adaptive filter is to reduce or bound the gaps by ...Another approach is to estimate the optimal Kalman gain directly without ...

  • An Adaptive Full-Feedback IIR Filter

    An active filter based o... 4页 1财富值 An Adaptive Color-Based ... 22页 免费 An optimal adaptive Kalm... 7页 5财富值 Adaptive Fading Kalman ...

  • [95.06]Steering an adaptive antenna array by the si...

    An optimal adaptive Kalm... 7页 5财富值 Adaptive Fading Kalman F... ... [95.06]Steering an adaptive antenna array by the simplified Kalman filter...

  • ...Alignment Based on Unscented Kalman Filter

    Adaptive Kalman Filter o... 暂无评价 5页 2.00元 Blind Channel ...In addition, particle filtering as an optimal nonlinear/non-Gaussian filtering...

  • 优化卡尔曼滤波算法中的目标函数选择

    (or in probability)an optimal estimator for H{z.in an oKFs.Then,these ...(optimized multiple model adaptive Kalman filter, oMM—AKF);由于oMM—AKF算法...

  • AHRS中一种自适应Kalman滤波算法

    文章编号:1001‐2400(2011)06‐0119‐07 An Algorithm of Adaptive Kalman Filter in AHRS TIAN Yi,SUN Jinhai,LI Jinhai,YAN Yuepeng (Institute of ...

  • An Adaptive Color-Based Particle Filter

    An Adaptive Color-Based Particle Filter_专业资料...mean shift or the Kalman/mean shift tracker the...Ramesh, Mean Shift and Optimal Prediction for E?...

  • federated_kalman_filter

    The federated filter yields estimates that are globally optimal or ...Neural network-aided adaptive unscented Kalman filter for nonlinear state ...

  • 基于Kalman滤波的MEMS陀螺仪滤波算法

    an adaptive Kalman filter. The simulation demonstrated that both methods could improve the performance of the filter. However,considering of practical limits ...

  • Kalman滤波和MATLAB实现

    Kalman滤波和MATLAB实现_信息与通信_工程科技_专业资料。卡尔曼滤波器(Kalman Filter)是一个最优化自回归数据处理算法(optimal recursive data processing algorithm)...

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email