改进粒子滤波算法研究

Research of Improved Particle Filtering Algorithm

重采样思想能解决粒子滤波中的粒子退化问题,但却导致粒子多样性丧失的现象,使描述状态后验概率密度的粒子不够充分。围绕如何增加粒子的多样性,已提出的改进算法包括MCMC移动步骤及正则化粒子滤波(RPF)算法。讨论2种改进算法的基本思想及步骤,通过对一典型标量非线性系统的仿真实验,分析改进算法的性能特点。实验结果表明,2种改进算法都有效增加了粒子的多样性,缓解了粒子匮乏问题。

The resample technique can resolve particle degeneration in particle filtering, but it leads loss of particle diversities. The result is that the particles which used to describe the probabilities of posterior probability density of state are not enough. Concerning advanced ways on how to expand diversities of particles, including algorithms of Monte Carlo Markov Chain （MCMC） movement and Regularized Particle Filter （RPF）. Discuss basic ideas and steps of the two algorithms. Through a simulation on a typical scalar quantity non-linear system, analyze the characteristics of the algorithms. The results show that both the two algorithms can expand diversities of particles, and can ease up the problem of lack of particles.