Engineering Intelligent Kalman Filtering Method

Kalman filtering is a basic method in the field of state estimation.When the model parameters are accurate,Kalman filtering can achieve the optimal estimation under the minimum mean square error.However,in most practical engineering applications,the limitations of parameter selection methods make it difficult to guarantee the accurate model.At the same time,the approximate representation of nonlinear system or non Gaussian system will inevitably lead to the mismatch between the model and the actual system.In view of the above problems,the main innovative research work is as follows:(1)An improved strong tracking filtering method is proposed,in which the fading factor acts on the variance of process noise.The traditional strong tracking filter uses the fading factor to calculate the covariance of prediction and estimation error.Its principle of action and the theory of generating utility are not well explained.In order to solve this problem,a method of applying the fading factor to the variance of process noise is proposed.Compared with the traditional strong tracking filter,the new method is equivalent to the real-time adaptive estimation of process noise variance,and has a good principle interpretability.At the same time,simulation results show that the new method has better estimation performance than the traditional method.(2)An intelligent Kalman filter design method based on credibility theory is proposed which solve he problem that the self-evaluation mechanism of Kalman filter is damaged.In the framework of Kalman filter performance analysis of existing mismatched systems,the filter calculated MSE(FMSE)and true MSE(TMSE)are used,a reliability analysis method for performance measurement of filter estimation is proposed,and the optimization condition of filter is successfully transformed from estimation mean square error frame to innovation mean square error frame,so as to realize the engineering design idea of Kalman filter.Finally,the calculation process of confidence factor is modeled as an optimization problem,which is solved by particle swarm optimization.The joint estimation of process and measurement noise variance is realized successfully.(3)In view of the cognitive limitations of the traditional optimal Kalman filtering on the uniqueness of the model parameter values,a design method of Engineering Kalman filters based on Model Parameter Ratios(MPR)is proposed.On the basis of the conclusion that the estimation error of Kalman filter reaches the minimum when the innovation mean square error is minimized,the estimation of the variance of process and measurement noise is transformed into the problem of extremum optimization.This work breaks through the traditional recognition of the uniqueness of the optimal model parameters of Kalman method,and provides a new and different perspective and thinking for the innovative design of adaptive filtering.