智能卡尔曼滤波概述

《智能卡尔曼滤波概述》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、IntelligentKalmanfilterfortrackingamanoeuvringtargetKalmanfilterisanidealrandomlinearadaptivefilter.ItneedstoclearlydefinetheMathematicalmodeltodescribetherelationshipbetweenoutputandinput,itisthemodelofthecontrollerofMathematicalmodelwhichworkswellwhilethefunctionbetweenoutputandinputisdete

2、rmined.Totheunknownenvironment,controllerofMathematicalmodelusuallydealwithitbyProbabilitystatistics(Conditionalmean,covariance),thelinearorMarkovmodelareoftenusedinpractice.buttherearesomeproblems:（1）ItishardtodescribetheUncertaintyclearlybyclassicaldatemodel.（2）Itishardtoaddtheknowledgefr

3、omdomainexpertstothesystem,itcanonlyuseittoestimateconditionsoftheinitialstateandcovariance.So,whenmanoeuvringtoturnortakingevasiveaction,thestandardKFcannotbeapplied,becausetheunknowntargetaccelerationduringthemanoeuvreappearsasextensiveprocessnoiseonthetargetmodel,andtheoriginalprocessnoi

4、sevariancecannotcoverit.Thetraditionalresearchmethodanditsshortcomingsatpresentareasfollows:1：Detectingthemanoeuvreandthentocopewithiteffectively.Examplesofthisapproachincludeinputestimation(IE)techniques,thevariabledimension(VD)filter,thetwo-stageKalmanestimator.shortcoming：thesetechniquesr

5、equireadditionaleffortsuchastheestimationanddetectionofacceleration,andthecompensationofthestateestimateorthetransitionbetweenthenon-manoeuvringfilterandthemanoeuvringfilterinordertodealwiththeunknowntargetmanoeuvres.2：describeingthemotionofatargetusingmultiplesub-filters.Thegeneralisedpseudo-

6、Bayesian(GPB)method,theinteractingmultiplemodel(IMM)method,andtheadaptiveinteractingmultiplemodel(AIMM)method[10]areincludedinthisapproach.shortcoming：Thesetechniquesalsoneedextraeffortsuchaspredefiningmultiplesub-filtersandupdatingthemodeltransitionprobability,inadditiontothelargecomputation

7、alloadimposedbyusingmultiplesub-filters.TheauthorofthispaperproposeaIKFofoptimizedfuzzysystembasedonGAmethod.Advantagesareasfollows：（1）UnliketheIEtechniqueandthesimilarmethods,theproposedIKFrequiresnoadditionaleffortforestimationanddetectionofthetargetma