一种基于差分演化的粒子滤波算法

摘要

针对粒子滤波(Particle Filter,PF)存在的粒子退化和贫化问题,该文提出一种基于差分演化(Differential Evolution,DE)的PF算法.首先,为了充分利用最新的观测信息,采用无迹卡尔曼滤波(Unscented Kalman Filter,UKF)来产生重要性分布,对重要性分布产生的采样粒子不再做传统重采样操作,而是直接把采样粒子当作DE中的种群样本,粒子权重作为样本的适应函数,对粒子做差分变异、交叉、选择等迭代优化,最后得到最优的粒子点集.试验结果表明,该算法有效缓解了传统PF算法中的粒子退化和贫化,提高了粒子的利用率,具有较好的估计精度.%The main problems of the Particle Filter (PF) are the sample degeneracy and impoverishment phenomenon. To deal with the problems, a new PF based on Differential Evolution (DE) is proposed. Firstly, the Importance Distribution (ID) which contains the newest measurements is produced with the Unscented Kalman Filter (UKF). Secondly, the particles sampling from the ID are no longer resampled by the conventional algorithm, however, they are regarded as the sample of the current population and their weights as the fitness function. Finally, a process of mutation, recombination and section is repeated until the optimum particles are found. The simulation result shows that the proposed method relieves effectively the sample degradation and poverty problems, improves the efficiency of particles and achieves preferable precision on estimation.