Kalman Filter Theory Of The Mismatched System Parameters

With the rapid development of modern technology,especially the sensor technology,computer technology and intelligent information processing technology are widely used in many civil and defense fields.Faced with a complex environment,it put forwards higher requirements to traget location,tracking accuracy and algorithm performance.However,we often have problems with the mismatched parameters when in dealing with the real system model,and the influence of various factors makes filtering estimation does not match actual.Then,under the assumption,the performance evaluation can not accurately describe the actual estimation performance.Therefore,it is necessary to study the Kalman filter theory of the mismatched system parameters.The problems caused by the sytem parameters mismatched: 1)The Kalman filtering estimation performance evaluation of linear system is not accuracy.2)For nonlinear system,linearization and introduced fator,which impact the estimation performance.3)The problems of Filter design and performance analysis when the system model is mismatched in the nonlinear system.In view of the above problems,the concrete research work of this thesis is arranged as follows:1)Performance analysis of Kalman filter with mismatched noise covariance.In the practice engineering,the noise covariance cannot completely known.Then,mismatched noise covariance would affect the filter estimation performance.Therefore,the estimation performance of MSEs under different definitions is studied,and the ranking and relative closeness are analysised.In the end,the engineering practice guidance is given,which can provide guideline for the actual system filter estimation performance evaluation.2)Performance analysis of extended strong tracking filter(ESTF)for nonlinear system.The ESTF introduces a time-variant fading factor evaluated based on the current measurement innovation in real time which can be used to forcedly correct one step prediction error covariance.Therefore,the complexity phenomena,such as inaccurate models and sudden change of state,can be dealt with effectively.Due to the subjective fading factor,however,ESTF differs from the linear Kalman filtering and some conventional properties are changed.Therefore,filter parameters between the EKF and the ESTF are compared;the inconsistency between estimation performances of the EKF and ESTF is explicitly revealed;the mismatching analysis of estimation performance expressions,which are the RMSE and the MSE,are briefly presented for the ESTF.The results show that some basic properties existed in the conventional Kalman filtering frame have been changed or broken for the ESTF.Finally,the effectiveness of research contents aforementioned is verified by the intelligent vehicle tracking system platform.3)A new cubature Kalman filter algorithm and its fusion method are studied.A novel stronger nonlinear filter EnFICIF is designed by combing cubature information filter,fifth-degree cubature rule,iterated method and ensemble scheme.The proposed EnFICIF has better nonlinear estimation performance and stability.Moreover,the associated multisensory fusion is deeply studied,which includes comparison of the estimation accuracies among four nonlinear fusion methods as well as demonstration of the exchanging property of measurements update order.Finally,the conclusion is effectively verified by the simulation experiments of target tracking system.