基于微分进化的组合重采样粒子滤波算法

摘要

针对粒子滤波重采样带来的样本贫化问题,设计了微分进化组合重采样算法.通过引入微分进化算法的交叉变异操作保持了粒子多样性；同时将残差重采样和微分进化重采样算法组合起来,克服了单纯微分进化重采样的时滞性问题.蒙特卡罗仿真表明,所研究的算法在精度上有很大的提高,并且实时性较好.算法效率提高了1倍,只需300个粒子就可以达到残差重采样粒子滤波1000个粒子的效果,而且实时性提高了很多.%To solve the sample impoverishment problem caused by the resample scheme of conventional particle filter, an evolutionary particle filter was proposed, in which Differential Evolutionary ( DE ) programming was introduced. It maintains the diversity of the particles by using DE, and can improve the target tracking ability via using the crossover and mutation operators. The time delay problem of the DE resample is solved by combining the residual resample scheme with DE resample scheme. Monte Carlo simulation results demonstrate that this new method can evaluate the state and track the target more accurately, and has a better real-time performance. The precision of our algorithm is two times of the standard particle filter, and the time cost is much less.