粒子滤波和无轨迹粒子滤波算法比较  

Comparison of Particle Filter and Unscented Particle Filter Algorithms

作　　者：郑琛瑶[1] 董真杰 张维全[1] 

机构地区：[1]91388部队93分队

出　　处：《舰船电子工程》2014年第12期47-49,61,共4页Ship Electronic Engineering

基　　金：国家自然科学基金重点项目"基于联合决策与估计的高频超视距雷达信息处理与融合"(编号:61135001)资助

摘　　要：解决水下水声目标的定位跟踪问题,需要建立动态的非线性非高斯模型,粒子滤波直接采用未含有最新量测信息的状态转移先验分布函数作为重要性密度函数来逼近后验概率密度函数,无轨迹粒子滤波是在粒子滤波的过程中引入重采样技术,通过无轨迹变换设计重要性密度函数,使其更加接近系统状态后验概率密度。仿真结果表明粒子滤波和无轨迹粒子滤波算法都可以提高定位跟踪精度,但无轨迹粒子滤波算法的估计精度更高,更适用于工程实践。To resolve the underwater acoustic position precision and tracking ,this paper establishes a dynamic nonlinear non‐Gaussian model .The particle filter directly employs the state transition prior distribution function which does not include the latest measuring information as an importance density function to approximate the posterior density function .The un‐scented paticle filter brings in resampling technique ,develops the importance density function by unscented transformation that is more close to the posterior density .Simulation results demonstrate that paticle filter and unscented paticle filter all can increase the precision of target position and tracking ,but there is higher estimation precision with the latter .The algorithm of unscented paticle filter is more suitable for engineering pactice .

关 键 词：无轨迹粒子滤波 非线性非高斯 重要性密度函数 重采样 定位跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]