粒子滤波综述  

A Review of Particle Filters

作　　者：吕德潮[1,2] 范江涛[1,2] 韩刚瓮[3] 马冠一[1] 

机构地区：[1]中国科学院国家天文台,北京100012 [2]中国科学院大学物理科学学院,北京100049 [3]北京航天飞行控制中心,北京100094

出　　处：《天文研究与技术》2013年第4期397-409,共13页Astronomical Research & Technology

摘　　要：主要对粒子滤波算法进行综述。首先详细描述了递归贝叶斯估计的基本原理和基于蒙特卡罗方法的重要性采样/重采样技术,在此基础上引出了粒子滤波标准算法——序贯重要性采样算法和序贯重要性重采样算法。针对这两个算法在应用中存在的问题,从提高算法的有效性和实时性两个方面,对近年来国内外在粒子滤波理论及应用研究方面开展的工作进行了介绍、分析归纳了改进粒子滤波算法及其主要改进思想。最后,对粒子滤波算法的研究方向进行了展望。This paper reviews the algorithms for particle filtering. We first present the basic knowledge of the Recursive Bayesian Estimation, and the Importance Sampling/Resampling Techniques which are based on the Monte Carlo Method. We then describe in detail the two standard algorithms, Sequential Importance Sampling and Sequential Importance Resampling. We subsequently introduce recent progress in the research of particle-filtering algorithms to tackle issues encountered by the two standard algorithms in practical applications. We analyze key ideas of the new algorithms. We classify the new algorithms according to their improvements in the validity and real-time performance. We finally give some suggestions of future research directions.

关 键 词：数据处理 粒子滤波算法 综述 非线性滤波 递归贝叶斯估计 重要性采样 重采样 

分 类 号：TN911.72[电子电信—通信与信息系统]