基于径向基-Galerkin解的反馈粒子滤波器

发布时间：2018-03-24 02:09

本文选题：非线性滤波　切入点：贝叶斯滤波　出处：《电子学报》2016年01期

【摘要】：反馈粒子滤波器(FPF)是一种连续时间最优贝叶斯估计器.针对FPF对系统采样率要求较高的问题,提出一种基于径向基函数-Galerkin法求解的反馈粒子滤波器.该算法推导了反馈增益势函数所满足偏微分方程的弱形式,结合径向基函数对势函数进行近似,利用Galerkin法和蒙特卡罗积分得到了反馈增益近似解,并给出了一种径向基参数选取方法,从数值上分析了径向基函数参数选取对于滤波精度的影响.仿真算例表明反馈粒子滤波器在低系统采样率下会严重发散,而本文算法能够避免这一问题,且提高了FPF在低系统采样率下的滤波精度和稳定性.
[Abstract]:Feedback particle filter (FPF) is a continuous-time optimal Bayesian estimator. Aiming at the problem that FPF requires high sampling rate, A feedback particle filter based on Radial basis Function-Galerkin method is proposed, in which the weak form of partial differential equation satisfied by feedback gain potential function is derived, and the potential function is approximated with radial basis function. The approximate solution of feedback gain is obtained by using Galerkin method and Monte Carlo integral, and a radial basis function parameter selection method is given. The influence of radial basis function parameter selection on filtering accuracy is analyzed numerically. The simulation example shows that the feedback particle filter will diverge seriously at low sampling rate, but the algorithm in this paper can avoid this problem. The filtering accuracy and stability of FPF at low sampling rate are improved.
【作者单位】：海军工程大学兵器工程系;
【分类号】：TN713

【相似文献】