一种基于GHF的高斯粒子滤波算法

An Gaussian particle filter based on the Gaussian-Hermite filter

高斯粒子滤波算法重要性权值方差不会随迭代次数的增加而增加,能够较好地解决粒子退化问题,但其重要性密度函数没有考虑最新的量测信息,导致有效粒子数减少,算法滤波性能下降.针对该问题,提出一种基于Gaussian-Hermite滤波(GHF)的高斯粒子滤波算法,采用GHF构造高斯粒子滤波的重要性密度函数,考虑最新的量测信息,增加有效粒子数,提高算法的滤波精度.仿真结果表明,所提出算法的滤波精度明显优于高斯粒子滤波算法.

Excluding the latest measuring information, the number of the effective particles reduces, so the performance of the Gaussian particle filter descends, which can conquer the particle degeneracy problem well for the variance of the important sampling weights not getting larger with time. Therefore, an improved Gaussian particle filter method based on the Gaussian-Hermite filter is proposed, and the importance density function is structured by using GHF. Including the latest measuring information, the number of effective particles are increased and filtering accuracy is improved significantly. The experimental results show that the proposed method is superior to the Gaussian particle filter.