Several Types Of Models For Uncertain Dynamic Systems Filter Design And Simulation

With the development of science and technology,especially the rise of modern industrial electronics,aerospace,image processing and other fields,filter technology has become one of the most popular research topics in academic and engineering fields.According to the different requirements of scene,there are also differences in the filtering framework.Among the various filters of linear systems,Kalman filtering requires the model to be accurate and the noise to be Gaussian,which is difficult to meet the requirements in the actual environment and affects the better development.Therefore,in this paper,the filter design method is studied for several kinds of systems whose models are uncertain or whose noise is non-gaussian.The main innovation contributions are as follows:(1)An online iterative solution method for the fading factor of strong tracking filters is proposed.Firstly,the scalar performance evaluation function is established based on the matrix identity of the strong tracking filter;Secondly,the scalar performance evaluation function is established based on the relevant information of the residual information matrix;Thirdly,solving the fading factor step by step based on the traversal range of the fading factor.Finally,the feasibility of the method is verified by simulation experiments.(2)A Double layer cubature Kalman filter is established.Firstly,the Sigma point sampling is performed based on the prediction error covariance matrix,and the second layer of Sigma point sampling is further centered on each Sigma point to enhance the statistical characteristic representation ability of samples.Secondly,the estimated value and covariance of the first layer samples are established on the second layer,and the overall predicted value and covariance matrix are established on this basis.Thirdly,sampling the measurement system and building the corresponding filter estimator;Finally,the feasibility of this method is verified by simulation experiments.(3)For a class of linear dynamic systems with non-Gaussian observation noise,a method of Maximum correntropy Kalman filter based on sample mean estimator is proposed.Firstly,the estimated value is regarded as a new observation equation and an extended measurement system based on the original measurement equation is established.Secondly,the measurement error covariance matrix is established based on the resampling of the measurement,and combined with the prediction error covariance matrix,the mutually independent decomposition method of the dimension vector of the estimation value and the measurement value is established.Thirdly,an objective function isestablished to estimate the state of the system.Then,the gain matrix is modeled as a fixed point equation,and the online iterative optimization is performed.Finally,the validity of the method was verified by simulation.