递推更新高斯粒子滤波器

传统高斯粒子滤波算法(Gaussian particle Filter,GPF)中,粒子的重要性密度函数是由高斯滤波器结合当前最新量测来构建的.由于传统高斯滤波器在量测更新阶段直接利用量测对状态进行线性更新,在某些条件下会导致所构建的重要性密度函数并不能很好地近似状态真实分布.为了解决这一问题,结合递推更新的思想,本文推导出了递推更新高斯滤波器(recursive update Gaussian filter,RUGF)的一般结构.并在此基础上,选用RUGF来构建粒子滤波的重要性密度函数,从而提出了基于递推更新的高斯粒子滤波算法(recursive update gaussian particle filter,RUGPF).仿真表明,在非线性系统状态估计问题中,递推更新可以很好的利用量测信息,相比于传统的GPF,本文所提出的RUGPF滤波算法可以提供更高精度的估计结果.

Recursive update Gaussian particle filter

In traditional Gaussian particle filter (GPF), sample importance density function is constructed through combining the latest measurements based on Gaussian filter (GF). However, in measurement update of the traditional GF, since the measurement value is assimilated directly based on the linear update rule, the constructed sample importance density function may not be approximate to the real posterior distribution under certain conditions. To solve this problem, we propose a new recursive update GF (RUGF) based on the recursive update idea and give out its general framework. On this basis, a new sample importance density function is constructed by using RUGF, based on which a new recursive update GPF (RUGPF) can be derived. Simulation results demonstrate that recursive update idea can assimilate the measurement information commendably, and compared with traditional GPF, the proposed filter has higher estimation accuracy for state estimation in nonlinear systems.