A New Mixture Bootstrap Filter for State Estimation of Maneuvering Target Tracking

摘要

Bootstrap filter (BF) offers a general numerical tool to approximate the posterior density function for the state in nonlinear and non-Gaussian filtering problems. However, it suffers from that it is quite computer intensive, with the computational complexity increasing quickly with the state dimension. One remedy to this problem is to marginalize out the states appearing linearly in the dynamics and then carry out the analytical marginalization using standard algorithms, such as Kalman filter. In this study, a new mixture bootstrap filter (MBF) is presented to track maneuvering target. The main manipulation of proposed method is that we tackle the tracking problem using a modified variable structure multiple model and an efficient Rao-Blackwellized particle filtering. Meanwhile, to adaptive to different cases of target's maneuverability, the covariance matching technique is also employed. Computer simulation indicates that this new method has better performance than conventional IMM (Interactive Multiple Model) method based on a standard Cartesian extended Kalman filter (EKF).