一种改进的裂变自举粒子滤波算法

A modified fission bootstrap particle filtering algorithm

退化现象是粒子滤波应用中存在的主要问题,自举粒子滤波虽然可解决退化问题,但易导致粒子耗尽。裂变自举粒子滤波算法虽能增加粒子的多样性,但其裂变过程对权值信息的利用不充分,且权值计算过于简单,滤波精度改进不大。采用高权值粒子裂变小、低权值粒子裂变大的裂变繁殖方法和权值递推计算方法,对裂变自举粒子滤波算法进行了改进,估计精度有所提高。

The degeneracy phenomenon,one problem of the particle filter in application,can be resolved by the bootstrap particle filter,but it causes the sample impoverishment.To overcome the above problem,the fission bootstrap particle filter is proposed,which can improve the diversity of samples,but the particle reproduction makes little use of the particle weight information and the weight of the reproducted particle is inaccurate.The estimation accuracy is improved little.In this paper,the fission bootstrap particle filter is modified by adopting the recursion of weight and improving the particle reproduction,in which particles with high weight change little and those with low weight change much.The estimation accuracy is improved.The simulation results show the effectivity of the modification.