一种基于状态空间离散化的粒子滤波器

粒子滤波是一种基于蒙特卡罗仿真的最优回归贝叶斯滤波算法。这种方法不受线性化误差和高斯噪声假定的限制,适用于任何状态转换或测量模型,因此能够很好地解决非线性、非高斯环境下系统的状态估计问题。在它的设计中最重要的一步就是建议分布的选取。传统的算法需要在整个状态空间中进行计算,这浪费了大量的计算时间。该文提出一种新的建议分布的构造方法,它基于状态空间离散化的思想来构造建议分布。仿真结果表明,相对传统的算法这种粒子滤波器能用更少的粒子产生更准确的估计值。

A particle filter based on state space discretization method

Particle filter is a computer-based method for implementing an optimal recursive Bayesian filter by Monte Carlo simulations.The method may cope with any nonlinear model without any limitations of linearization error and Gaussian noises assumption,so it can be used for the state estimation problem of non-Gaussian nonlinear systems.A crucial step in the design of a particle filter is the choice of the proposal distribution.Traditional algorithm requires the computation over the entire state space.It wastes a lot of computation time.We present a particle filter based on the idea of state space discretization to construct proposal distribution.The simulation results show this algorithm can gets more accuracy estimate even with less particles compared with conventional algorithm.