Information Fusion Nonlinear Filter And Application Research In Passive Location

The nonlinear problem of the measurement equations or state equations exists in many areas, such as moving object tracking, modern signal processing, image processing, and automatic control. So the state estimation of a nonlinear system has its significance in both theoretic and engineering applications. However, the current nonlinear filtering schemes in practice are approximation algorithms which normally have the disadvantages of low precision and easy divergence so that these methods have limited performance to solve the nonlinear problems.A practical way to improve the nonlinear filtering precision is to use the multi-sensor information fusion theory in the state estimation of a nonlinear system. Thus, the more accurate and steadier estimation results can be obtained measuring the same target by diversity sensors without adding any measurement base stations. In order to increase the accuracy of estimation and enhance the system stabilization, this thesis proposed a series of information fusion nonlinear filters for nonlinear estimation problems within applying information fusion algorithms. And a part of information fusion filtering methods has been utilized into the passive location and tracking applications.First, under the linear minimum variance information fusion criterion, multi-sensor information fusion extended Kalman filter (EKF) has been proposed for nonlinear system. And the local filtering error variance and co-variance equations have been given for computing optimal weights.Second, under the linear minimum variance information fusion criterion, multi-sensor information fusion unscented Kalman filter (UKF) has been proposed for nonlinear system. And the local filtering error variance and co-variance equations have been given for computing optimal weights.Third, under the linear minimum variance information fusion criterion, a series of multi-sensor information fusion particle filter have been proposed for nonlinear system including information fusion particle filter (PF), information fusion extended Kalman particle filter (EKPF), information fusion unscented particle filter (UPF), information fusion Gaussian particle filter (GPF) and information fusion Gaussian sum particle filter (GSPF). And the local filtering error variance and co-variance equations have been given for computing optimal weights.Finally, applying linearization measurement equation of extended Kalman filtering theory, the solving equations of passive location filter, passive location predictor and passive location smoother have been proposed for passive localization system. With applying the information fusion principles, under the linear minimum variance information fusion criterion, the different forms of multi-sensor information fusion filter, multi-sensor information fusion predictor and multi-sensor information fusion smoother have been acquired for passive localization system. And the local estimation error variance and co-variance equations have been given for computing optimal weights.Many simulation examples prove the efficiency of the above information fusion nonlinear filters and passive location algorithms.