Robust state estimation for uncertain linear systems with random parametric uncertainties.pdf

《Robust state estimation for uncertain linear systems with random parametric uncertainties.pdf》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、SCIENCECHINAInformationSciences.RESEARCHPAPER.January2017,Vol.60012202:1–012202:13doi:10.1007/s11432-015-0327-xRobuststateestimationforuncertainlinearsystemswithrandomparametricuncertaintiesHuaboLIU1,2*&TongZHOU1,31DepartmentofAutomation,TsinghuaUniversity,Beij

2、ing100084,China;2CollegeofAutomationandElectricalEngineering,QingdaoUniversity,Qingdao266071,China;3TsinghuaNationalLaboratoryforInformationScienceandTechnology(TNList),TsinghuaUniversity,Beijing100084,ChinaReceivedMarch1,2016;acceptedJune28,2016;publishedonlin

3、eNovember22,2016AbstractInthispaper,weinvestigatestateestimationsofadynamicalsystemwithrandomparametricuncertaintieswhichmayarbitrarilyaffectaplantstate-spacemodel.Arobustestimatorisderivedbasedonexpectationminimizationofestimationerrors.Ananalyticsolutionsimila

4、rtothatofthewell-knownKalmanfilterisderivedforthisnewrobustestimatorwhichcanberealizedrecursivelywithacomparablecomputa-tionalcomplexity.Undersomeweakassumptions,itisprovedthatthisestimatorconvergestoastablesystem,thecovariancematrixofestimationerrorsisbounded,a

5、ndtheestimationisasymptoticallyunbiased.Numer-icalsimulationsshowthattheobtainedrobustfilterhasanestimationaccuracycomparabletootherrobustestimatorsandcanbeappliedinawiderrange.Keywordsrobustness,stateestimation,recursiveestimation,parametricuncertainty,regulari

6、zedleast-squaresCitationLiuHB,ZhouT.Robuststateestimationforuncertainlinearsystemswithrandomparametricuncertainties.SciChinaInfSci,2017,60(1):012202,doi:10.1007/s11432-015-0327-x1IntroductionStateestimationhasbeenextensivelyutilizedinsystemcontrolandsignalproce

7、ssing.TheKalmanfilteristheoptimalestimatorunderthecriterionofmean-squaresforlinearsystemswithnormalexternaldisturbancesandwidelyappliedinnumerousfieldssuchascontrol,finance,communicationandsoon[1–3].Moreover,somevariantsofKalmanfilterareputforwardandapplied,suchase

8、xtendedKalmanfilterandunscentedKalmanfilter(see[1,4]andthereferencestherein).Sincemodellingerrorsaregenerallyunavoidableonaccountofcomplexmanufactureprocessandmodellinginaccur