Abstract: Aiming at solving the sample impoverishment phenomenon caused by the re-sample scheme of conventional particle filter, an evolutionary particle filter is proposed, in which differential evolutionary programming is introduced. The improved approach relieves the effect caused by samples impoverishment through ameliorating the diversity of samples set and improves the ability of tracking the target via using the crossover and mutation operators. Simulation results demonstrate that compared with the traditional particle filter, this improved method can evaluate the state and track target more accurately, and the precision of this algorithm increases more than 100 percent than standard particle filter. It needs fewer particles to achieve tracking problems.

Key words: differential evolution, particle filter, resample, samples impoverishment

摘要： 针对传统粒子滤波重采样算法带来的样本贫化问题，提出了一种利用微分进化算法进行重采样的粒子滤波改进方法，新方法通过引入交叉变异操作，保持了粒子的多样性并抑制了粒子退化现象，提高了目标状态的估计与跟踪能力。仿真结果表明，相对于普通粒子滤波，新算法的估计精度提高了一倍，使用较少的粒子数即可完成跟踪任务。

关键词: 微分进化, 粒子滤波, 重采样, 粒子退化