基于进化采样的粒子滤波算法

The particle filter algorithm based on evolution sampling

在粒子滤波算法中,重采样的引入有效地改善粒子退化现象,但同时也导致了粒子多样性减弱问题的产生.本文给出了一种基于进化采样的改进粒子滤波算法.该算法在重采样过程后,首先根据马尔可夫链蒙特卡罗(Markov-Chain-Monte-Carlo,MCMC)技术和遗传算法中的模拟二进制交叉原理生成候选粒子,并利用适应度函数完成对于其权重的度量.然后结合当前时刻的重采样粒子构建候选粒子集,进而提升了重采样后粒子的多样性,最终依据粒子自身的权重实现粒子的优选.仿真结果表明:该算法可有效地提高对于非线性系统状态的估计精度.

In particle filter algorithm, the re-sampling step effectively solves the problem of particles degeneracy, however, it reduces the particle variety. An improved particle filtering algorithm is given based on the evolution sampling. In the process of re-sampling, this algorithm generates candidate particles based on the Markov-Chain-Monte-Carlo（MCMC） technique and the analog binary crossover principle, and then, weighs the sampling particles against their importance according to the fitness function. The current re-sampling particles are then associated in constructing the candidate particle set to enhance the variety of re-sampling particles. Finally, the optimizing selection of particles is realized based on the particle weigh. Simulation results show the method can effectively improve the state estimation precision.