中文摘要：

针对粒子滤波（PF）计算量大、粒子退化以及缺乏对多传感器量测系统状态估计的适用性等问题的综合处理，提出了一种基于多传感器粒子权重优化的两级Rao．Black．wellized粒子滤波（RBPF）算法。该算法首先采用Rao—Blackwellized建模技术实现对被估计系统状态演化过程的建模，并结合加权融合策略完成多传感器量测对于粒子权重的优化。其次，通过两级预测更新机制的构建和引入，实现最新量测信息对于当前时刻粒子估计结果的修正。另外，考虑到重采样后粒子多样性枯竭问题，在滤波结果中蕴含冗余和互补信息的提取和利用的基础上，给出了一种粒子多样性增强方法。理论分析和仿真实验验证了此算法的可行性和有效性。

英文摘要：

Aiming at the comprehensive treatment of particle filters＇ problems such as large amount of calculations, particle degeneration, and lack of applicability to state estimation for multi-sensor measurement systems, a novel two- stage Rao-Blackwellised particle filter algorithm based on multi-sensor particle weight optimization is proposed. The new algorithm functions as below： Firstly, the Rao-Blackwellised modeling technology is adopted to model the state evolution process of a system needing to estimate, and the weighted fusion strategy is used to optimize the particles＇ weights by combining with multi-sensor measurement information. Next, through the construction and introduction of a two-stage prediction update mechanism, the estimated result of a current particle is modified by the latest meas- urement information. In addition, in view of the diversity exhaustion of particles from re-sampling, a new method for enhancement of particle diversity is given by means of the extraction and utilization of redundancy and comple- mentary information from the current filter result. The theoretical analysis and experimental results show the feasi- bility and efficiency of the proposed algorithm.