摘要
本文提出一种采用粒子群(PSO)优化扩展卡尔曼粒子滤波(EPF)的新算法.由于上一时刻的目标解对当前时刻目标的影响最大,提出粒子群中的粒子不考虑其自身最佳经历和群体最佳经历,而只考虑前一时刻的全局最优解;取上一时刻的目标解代表粒子集中全局最优解.采用粒子群优化扩展卡尔曼粒子滤波(EPF)的状态转移方程,使得粒子集在权值更新前趋向于高似然区域,从而更加逼近真实状态的后验概率密度分布,克服了粒子退化问题,提高了预估精度,并极大地降低了所需的粒子数.仿真实验结果表明,该算法预估性能优于传统的粒子滤波方法.
A new particle swarm optimized (PSO) extended kalman particle filter (EPF) algorithm is proposed in this paper. due to the target solution of the last time have the greatest impaction on the current target, the particles of PSO dont consider the best experience of themselves and the best experience of the whole group, and only consider the global optimum of the last time ; using the target solution of last time as the global optimum of the particles, particle swarm optimization is introduced into extended kalman particle filter to optimize its state transfer equation, which makes the particle set incline to the high area be- fore its the weights upsetting and the particle of the far from the real state more likely to incline to the area which the real state generate the big probability, so it can match the true posterior more closely , overcome the question of particle degeneration, improve the precision of the estimation and reduce the needed numbers of the particles. The simulation experiment results show that estimation performance of the arithmetic which we proposed is superior to the standard particle filter and the other filters such as the extended kalman particle filter, the particle swarm optimized particle filter.
出处
《天津理工大学学报》
2009年第5期50-53,共4页
Journal of Tianjin University of Technology
基金
天津市高等学校科技发展基金(20071308)
关键词
扩展卡尔曼粒子滤波
粒子群优化
后验概率密度分布
extended kalman particle filter
particle swarm optimization
posterior probability density distribution