Self-tuning Measurement Fusion Kalman Estimator And Its Applications For Typical Tracking System

Among the development and application of modern science and technology, information widely exist in Communications, control, signal processing, data mining, artificial intelligence, bioinformatics, aerospace technology, economic forecasts and controls, etc. and plays an important role. With the rapid development of the information age, information tends to huge date, extensive sources and more correlated arrangement and so on. Therefore, the simply transmission and collection can not meet the needs of the information age. The needs of Modern information process for precision and fault tolerance became more and more higher. Information fusion technology as one way to solve this problem comes into being. As a new technology in the multi-source information integrated treatment, information fusion can synthesize the multi-source information intelligently, and generate more accurate, complete, reliable estimates and judgments than single source of information.In this paper, the optimal and self-tuning weighted measurement fusion Kalman filter and UKF were studied, using the weighted measurement fusion and various identification methods, for the linear and nonlinear multi-sensor system. Then the prediction control was studied based on the optimal and self-tuning weighted measurement fusion Kalman filter, for a class of controlled linear discrete multi-sensor system. In this paper, it is mainly about the theoretical analysis and computer simulation research, and the specific contents and consequences include the following:Firstly, for a class of linear multisensor systems with unknown noise statistics, several self-tuning weighted measurement fusion Kalman filters were proposed, using the least squares identification, identification of correlation function and multisensor cooperation identification. In three different situations, the self-tuning weighted measurement fusion Kalman filters based on the multisensor cooperation identification were concretely described and analyzed by simulation. And it is proved by the ergodicity of the random sequences that the system noise statistics identified by multisensor cooperation identification converge to the true values of the systems with probability 1. Then the self-tuning weighted measurement fusion Kalman filters obtained have the asymptotic global optimality.Secondly, for a class of nonlinear multisensor systems, the weighted measurement fusion UKF filter was proposed. It is proved that the weighted measurement fusion UKF filter and the centralized measurement fusion UKF filter have the numerical equivalence. So the weighted measurement fusion UKF filter has the asymptotic global optimality, but it reduces the computational burden greatly compared with the centralized measurement fusion UKF filter. For a class of nonlinear multisensor systems with unknown noise statistics, the self-tuning weighted measurement fusion UKF filter was proposed. This method combines the multisensor cooperation identification and the adaptive filter method, and improves the stability, convergence speed and accuracy of the simple adaptive UKF filter. Thirdly, the weighted measurement fusion predictive control algorithm was proposed by combining the weighted measurement fusion algorithm and the predictive control algorithm. This method avoids the complex Diophantine equations, and reduces the computational burden by using the Kalman filter algorithm to predict the states of the systems. The introduction of weighted measurement information fusion improves the stability and accuracy of the control system.Fourthly, for a class of controlled multisensor systems with unknown noise statistics, the self-tuning weighted measurement fusion predictive control algorithm was proposed. This method uses the multisensor cooperation identification proposed in this paper, and improves the stability and accuracy of the identifier. Therefor it improves the stability and accuracy of the controlled system.This paper uses a number of simulation examples to demonstrate the effectiveness of these methods.