搜索能力自适应增强的群智能粒子滤波

Swarm intelligence particle filtering based on adaptive enhancing search ability

针对传统粒子滤波的退化、样本枯竭现象及其导致的状态推理精度差的问题,提出了一种新型粒子滤波算法。利用群智能优化算法中的粒子群优化算法作为优化手段,改进粒子的先验分布。通过自适应地调节粒子的惯性权值增强粒子群的探索和开发能力,减少粒子群优化算法的早熟现象,使得采样后的粒子朝着高似然区域移动,从而有效地提高系统状态推理精度。利用Crame′r-Raolowerbound定义了算法有效性的度量。通过仿真实验证明该算法是有效和稳定的。

For addressing poor inference precision with canonical particle filtering resulting from weight degeneracy and sample impoverish,a new particle filtering algorithm is proposed,which utilizes the improved particle swarm optimization for improving priori particles distribution.Through adaptively adjusting inertia weight,particles exploration ability and exploitation ability are both enhanced so that premature phenomenon with particle swarm optimization is weakened.As a result,particles can move toward high likelihood areas,which can effectively increase status inference precision.The proposed algorithm validity is measured by Crame′r-Rao lower bound.Simulation results show that the proposed particle filtering is valid and stable.