一种改进粒子滤波算法分析

Analysis of an improved particle filter algorithm

在非线性、非高斯系统的状态估计研究中,最常用的是EKF和UKF两种方法,但是这两种方法还局限于高斯分布的情况。在之后的研究中出现了不受非线性、非高斯分布问题限制的粒子滤波算法。这种算法的主要问题是粒子退化问题,常规的再采样方法虽然可以解决退化问题,但是容易导致粒子耗尽。针对这种问题,本文提出用辅助变量粒子滤波算法,对标准粒子滤波算法步骤中的再采样部分进行改进,最后对算法进行性能仿真及分析。仿真结果表明,改进的粒子滤波算法性能良好。

In the non-linear,non-Gaussian system state estimation studies,the commonly used methods are EKF and UKF,which is limited to Gaussian distribution.In the later study,the particle filter algorithm is proposed in the research of non-linear,non-Gaussian systematic.Degeneracy phenomenon is a main disadvantage to particle filter application,common resampling methods can resolve degeneracy phenomenon,but the sample impoverishment is deduced.Therefore,this paper is presented auxiliary variable PF to improve the part of resampling.Finally,make the simulation and analysis of the AVPF algorithm.Simulation results show that the improved particle filter algorithm has good performance.