Improved IMM Algorithm for Nonlinear Maneuvering Target Tracking
Liang Gao, Jianping Xing, Zhenliang Ma, Junchen Sha, Xiangzhan Meng · Procedia Engineering · 2012
Devoted to the problem of state estimation of discrete-time stochastic systems, SIMM (Scalar-Weight Interacting Multiple Model) and MIMM (Matrix-Weight Interacting Multiple Model) methods are proposed by X. Fu, in which the filter outputs are combined based on two optimal multi-model fusion criterions weighted by scalars and general matrices, respectively. In this paper, four improved IMM algorithms (EKF-SIMM, EKF-MIMM, UKF-SIMM and UKF-MIMM) are presented for nonlinear maneuvering target tracking based on SIMM and MIMM. The proposed improved algorithms can receive the optimal state estimations of target in the nonlinear minimum variance sense. Experiments results verify the effectiveness of the proposed algorithms by comparing with EKF-IMM and UKFIMM. And the proposed algorithms have an absolute advantage in the velocity estimation. In particular, UKF-MIMM is obviously better than EKF-IMM and UKF-IMM in accuracy while EKF-SIMM is superior in elapsed time. Therefore, the proposed algorithms can be competitive alternatives to the classical IMM–based filter algorithms for nonlinear maneuvering target tracking.