本文提出了边缘 Rao-Blackwellized 粒子滤波器(marginal Rao-Blackwellized particle filter, MRBPF)算法,算法融合了 Rao-Blackwellized 粒子滤波器(Rao-Blackwellized particle filter , RBPF)算法和边缘粒子滤波器(marginal particle filter, MPF)算法。算法中状态被分为线形和非线性两部分,分别用 MPF 和卡尔曼滤波器(Kalman Filter)进行估计。地形辅助导航(terrain aided navigation, TAN)的仿真结果表明,与 RBPF 相比,提出算法的非线性状态估计的误差均方根(root mean square error, RMSE)和误差方差分别降低了约 29%和 96%,独立粒子数提高了约80%,获得了更好的收敛结果。分析表明,现有RBPF是MRBPF的一个特例。 In this paper, an edge Rao-Blackwellized particle filter (MRBPF) algorithm is proposed. The algorithm combines Rao-Blackwellized particle filter (RBPF) algorithm and marginal particle filter , MPF) algorithm. The state of the algorithm is divided into linear and nonlinear two parts, respectively, with MPF and Kalman Filter (Kalman Filter) to estimate. The simulation results of terrain aided navigation (TAN) show that compared with RBPF, the root mean square error (RMSE) and error variance of the proposed nonlinear state estimation algorithm are reduced by about 29% and 96%, the number of independent particles increased by about 80%, get better convergence results. Analysis shows that the existing RBPF is a special case of MRBPF.