粒子滤波算法改进策略研究*

为了改进粒子滤波算法的性能,这里研究了一种粒子滤波算法改进策略。该粒子滤波算法改进策略包括四部分:首先,采用了结合退火参数的混合建议分布,以考虑当前观测测量值的最新信息;接着,基于有效样本大小确定自适应重采样的阈值,以保证有合适的重采样次数;然后,基于权重优化思想提出了一种改进的部分系统重采样算法,在利用算法执行速度快的同时优化部分系统重采样算法;最后,在重采样后执行粒子变异操作,以保证样本的多样性。通过仿真实验,粒子滤波改进策略的性能和有效性均得以验证。

Research on improved strategy for particle filter algorithm

In order to improve the algorithm performance, this paper studied the improved strategy for particle filter algorithm. The improved strategy for particle filter algorithm mainly included four steps. Firstly, it utilized a hybrid proposal distribution with annealing parameter to consider current information of the latest observed measurement. Moreover, the algorithm determined adaptive resampling threshold by effect sample size in order to assure the appropriate resampling number. Furthermore, it presented an improved partial stratified resampling (PSR) algorithm based on weight optimization, which not only used the implementation advantage of PSR algorithm but also optimized the PSR algorithm. Lastly, particle mutation operation after resampling was implemented to obtain the sample diversity. With the simulation program, the performance of the proposed strategy is evaluated and its validity is verified.