基于小生境遗传算法的粒子滤波算法  

Particle filter algorithm based on niching genetic algorithm

作　　者：张航[1] 李梦丽[2] 杨清波[2] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083 [2]中南大学信息科学与工程学院控制工程系,长沙410083

出　　处：《计算机工程与应用》2013年第18期191-194,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.50808025);2010年中南大学硕士研究生学位论文创新资助项目(No.2010ssxt209)

摘　　要：重采样是解决粒子滤波退化问题的主要方法,重采样的基本思想是采取复制保留权值较高的粒子,删除权值较低的粒子,而这导致了粒子多样性的减弱,特别是在样本受限条件下,甚至导致滤波发散。针对上述问题,提出改进的粒子滤波算法,将Mean Shift与粒子滤波融合,在重采样部分引入小生境遗传算法,提高粒子的多样性,避免粒子退化。实验表明,改进后的算法状态估计精度更高,效果更好。Resampling is a critical operation to solve degeneracy problem with particle filters generally.The basic idea of resampling is to discard particles which have small weights and concentrate on particles with large weights.But resampling often introduces sample impoverishment problem,especially the sample is limited under the condition,even causes the filter to disperse.This paper proposes improved particle filter algorithm.Mean Shift integrates with particle filter,and then the niching genetic algorithm is used in resampling in order to improve the variety of particles and remove the degeneracy phenomenon.The simulation results prove the proposed algorithm reduces the tracking error,and has better precision.

关 键 词：粒子滤波 Mean SHIFT 小生境遗传算法 重采样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]