一种约束扩展卡尔曼粒子滤波器

Constrained Extended Kalman Particle Filter

针对非线性非高斯动态系统中的状态估计问题,提出了一种约束扩展卡尔曼粒子滤波新算法。设计非线性状态约束条件,即修正先验信息将非线性状态约束信息融入贝叶斯估计目标函数的构建,得到修正的似然函数,同时,在滤波更新过程中引入当前观测信息,从而,有效利用观测约束信息修正新息、增益及重要度权值的的预测及更新。所提出的算法概率地选择粒子,采用有效的数值方法来执行一系列约束优化,使得远离可行区域的那些粒子不太可能传播到下一个时间步长。引起的近似误差使用重要性抽样方法进行修正。仿真实验表明该方法能够构成更接近真实的后验分布,滤波性能优于传统粒子滤波器。

A new constrained extended Kalman particle filter(EKPF)algorithm is proposed to solve the state estimation problem in nonlinear non-Gaussian dynamic systems.The nonlinear state constraint condition is designed,that is,the modified priori information is incorporated into the construction of the Bayesian estimation objective function,and the modified likelihood function is obtained.At the same time,the current observation information is introduced in the process of filtering and updating.The prediction and updating of innovation,gain and importance weight are corrected by using observation constraint information.The proposed algorithm selects the particle probability and adopts an effective method to perform a series of constrained optimization,which makes it unlikely that those particles far away from the feasible region will propagate to the next time step.The approximate error caused by this method is corrected by the importance sampling method.The simulation results show that the proposed method can form a more realistic posterior distribution,and the filtering performance is better than that of the traditional particle filter.