裂变优选重采样粒子滤波算法

A Study of Bootstrap Particle Filtering with Fission and Selection

重采样在缓解粒子退化的同时带来了粒子贫化问题,针对这种情况,提出一种裂变优选重采样方法。在需要进行重采样时刻,提取出有效粒子,该时刻有效粒子对应于上一时刻的粒子确定为裂变父代粒子,裂变子代粒子数正比于各有效粒子权值,将子代粒子进行一次滤波迭代,根据该时刻量测值和权值公式计算每个子代粒子的权值,选择权值大的粒子覆盖该时刻的无效粒子。蒙特卡罗仿真表明:与裂变自举粒子滤波、随机重采样粒子滤波相比,滤波精度更高,有效粒子数增加,而且重采样后粒子退化速度变缓。

Aimed at solving the problem of particle impoverishment introduced by resampling while relie- ving degeneracy , a method of fission with selection bootstrap particle filtering （FSBPF） is proposed. At the moment of resampling, effective particles are picked up.The effective particles at this moment corre-sponding to the particles at last moment are called fission elder generation particles. The number of fission filial generation particles is in direct proportion to the weight of each effective particle, and the filial generation particles are subjected to a filtering and a generation selection. After an iteration, the weights of the new particles can be got according to the measures and the weight formula, and particles with larger weight are chosen to replace the ineffective particles at this time. The results of Monte Carlo simulations show： comparing with the fission bootstrap particle filtering （FBPF） and the random resampling particle filtering （RRPF）, FSBPF are more precise in filtering results, increased in effective particles and lower in the speed of degeneracy.