粒子滤波及其在导航系统中的应用综述

传统的扩展卡尔曼滤波方法要求对非线性系统近似线性化,有可能会引入较大的模型误差.应用粒子滤波解决了这一问题.该算法可以直接应用于原系统的非线性模型当中,并且不需考虑系统噪声和量测噪声是否为高斯白噪声,都能得到很好的滤波效果.文中介绍了粒子滤波的理论基础-贝叶斯估计及具体的实现方式-蒙特卡罗方法;指出粒子滤波存在的退化问题,并从减小退化现象入手将重要性采样和再采样方法引入到算法之中;最后阐述了粒子滤波在导航系统中的一些应用.

Overview of particle filter and its applications in integrated navigation system

Traditional Extended Kalman Filter(EKF) may bring serious model errors.To overcome this problem,an algorithm of particle filtering was developed.It could be directly applied to the nonlinear model of the initial system,and could get good filtering result whether the system noise or measured noise was Gaussian or not.The paper introduced the filtering algorithm and its applications in INS,including Bayesian state estimation and Sequential Mote Carlo sampling.To overcome the degeneracy in particle filtering,an importance sampling and re-sampling method was introduced into the algorithm.The applications of particle filtering in navigation system were presented at last.