基于MCMC方法的正则粒子滤波算法及其应用

Regularized particle filtering algorithm and its application based on MCMC method

粒子退化现象是一般粒子滤波无法避免的问题。通过分析该现象产生的原因,提出了将MCMC(马尔可夫链蒙特卡罗)方法应用于正则粒子滤波算法(RPF),与采样重要重采样(SIR)粒子虑波算法比较,此算法不仅克服了粒子退化现象,而且解决了重采样带来的采样枯竭的影响,仿真和实验结果表明:该算法在滤波精度和自适应调整粒子个数方面比SIR粒子滤波有很大的提高。

Particle degeneracy phenomenon is avoidless problem in particle filtering application. Based on analyzing the cause of particle degeneracy, the regularized particle filtering with MCMC move step is proposed. The improved approach not only overcomes the effect of particle degeneracy, but also solves the sample impoverishment arisen from resampling. Simulation and experiment results demonstrate that in terms of filtering accuracy and the ability of auto- matically adjusting samples, the presented algorithm has much amelioration over SIR particle filtering.