基于混合多策略优化的粒子滤波算法

Particle Filter Algorithm Based on Hybrid Multi-Strategy Optimization

标准粒子滤波存在粒子贫化问题,在处理非线性问题时需要大量粒子才能达到所需的估计精度,降低了算法的综合性能。为此,文中提出了一种结合莱维(Levy)飞行策略、差分进化算法与成功历史策略的混合多策略优化的粒子滤波算法。该算法首先采用Levy飞行策略确定样本集的基本框架,并通过差分进化算法优化低权重的无效粒子,然后采用成功历史策略进行参数自适应调整,以动态调节算法寻优步长,将更多的粒子导向高似然区域。仿真结果表明,文中算法有效地提高了粒子多样性与滤波精度,并改善了在低测量噪声下的粒子贫化问题,降低对非线性系统估计所需的粒子数量。

The standard particle filter has the problem of particle impoverishment,while dealing with nonlinear problems requires many particles to achieve the required estimation accuracy,so the standard particle filter reduces the comprehensive performance of algorithm.This paper proposed a hybrid multi-strategy optimization particle filtering algorithm,which combines Levy flight strategy,differential evolution algorithm and success history strategy.The method firstly defines the basic framework of the sample set with Levy flight strategy,and optimizes the low-weight invalid particles with the differential evolution algorithm.Then the successful history strategy was used to adjust the parameters adaptively,to dynamically adjust the algorithm's optimum length,so as to guide more particles to the high likelihood region.Simulation results show that the proposed algorithm can effectively improve the particle diversity and filtering accuracy,enhance the particle impoverish problem under low measurement noise,and reduce the number of particles required for nonlinear system estimation.