一种粒子滤波自适应优化算法

A kind of adaptive optimization algorithm for particle filter

通过对粒子滤波算法中建议分布与重采样2种改进技术分析,提出了一种粒子滤波自适应优化算法.首先,基于退火参数自适应优化混合建议分布,以改进建议分布的采样范围.然后,在基于有效样本大小的自适应重采样技术之上,借助另一多样性测度即种群多样性因子来自适应调整重采样阈值,而且,样本变异操作在重采样之后被引入确保样本的多样性.同时,结合部分分层重采样算法研究并进行改进,改进的部分分层重采样算法具有原算法执行快时间短的优点,同时结合权重优化的思想改进重采样的样本权重计算.通过仿真实验,粒子滤波自适应优化算法的性能和有效性均得以验证.

By analyzing two techniques, namely, proposal distribution and resampling, an adaptive optimiza- tion algorithm for particle filter is presented. Firstly, hybrid proposal distribution is adaptively optimized based on anneal parameter in order to improve the sampling range of proposal distribution, se＇condly, based on the a- daptive resampling techniques on effective sample size, auother diversity measure, namely population factor, is used to adaptively adjust the resampling threshold. Moreover, the particle mutation operation is integrated into PF after resampling so as to ensure the diversity of particle sets. finally, an improved partial stratified re- sampling （PSR） algorithm in PF is studied, which keeps the advantage of PSR in implementation speed and time and improves the pefrmance of PF with weight optimization. Throngh simulation experiments, validity of the proposed method is verified.