New sigma point filtering algorithms for nonlinear stochastic systems with correlated noises.pdfVIP

J．Cent．SouthUniv．(2012)19：1010—1020 垒Springer D0I：10．1007／slI77l一0l2一Il04—4 New sigmapointfilteringalgorithmsfor nonlinearstochasticsystemswithcorrelatednoises WANGXiao—xu(~d、旭)，PANQuan(潘泉)，CHENGYong．mei(程咏梅)，ZHAOChun—hui(赵春晖) CollegeofAutomation，NorthwesternPolytechnicalUniversity，Xi’an710072，China ⑥ CentralSouthUniversityPressandSpringer—VerlagBerlinHeidelberg2012 Abstraet：New sigmapointfilteringalgorithms，includingtheunscentedKalmanfiIter(UKF)andthedivideddifferencefilter fDDF1，aredesignedtosolvethenonlinearfilteringproblem undertheconditionofcorrelatednoises．Basedontheminimum mean squareerrorestimationtheory,thenonlinearoptima1predictiveandcorrectionrecursiveofrmulasunderthehypothesisthattheinput noiseiscorrelatedwiththemeasurementnoisearederivedandcanbedescribedinaunifiedframework．Then．UKFandDDFwith correlatednoisesareproposedonthebasisofapproximationoftheposteriormeanandcovaria／iceintheunifiedframeworkbyusing unscentedtransformationandsecondorderStirling’sinterpolation．TheproposedUKFandDDFwithcorrelatednoisesbreakthrough thelimitationthatinputnoiseandmeasurementnoisemustbeassumedtobeuncorrelatedinstandardUKF andDDF．Twosimulation examplesshow theeffectivenessandfeasibilityofnew algorithmsfordealingwithnonlinearfilteringissuewithcorrelatednoises． Keywords：nonlinearsystem；correlatednoise；sigmapoint；unscentedKalmanfilter；divideddifferencefilter and thathtederivration oftheJacobian matrix isvery lIntrOductiOn complicatedinmanycases． In order to avoid these shortcomings of EKF, Moreandmoreattentionhasbeenpaidtosolving recently,the novel and efficient filtering tec