粒子滤波和正则粒子滤波算法性能研究

Comparative Performance Study of Particle Filter and Regularized Particle Filter

主要介绍了解决系统状态估计问题的滤波算法。在提出非线性高斯系统模型的基础上着重阐述了扩展卡尔曼滤波(EKF)、粒子滤波(PF)和正则粒子滤波(RPF)算法。对这三种算法在不同的噪声条件下的估计性能进行了仿真分析。结果表明,在非线性高斯系统中,PF和RPF的估计性能远比EKF的估计性能要好,由于RPF是从离散分布中重构其近似连续分布,再从该连续分布中采样粒子,估计性能比PF要好,尤其在小噪声的环境下,估计性能更加稳定。

The filtering algorithms are introduced using to deal with the problems of system state estimation. Firstly, the nonlinear Gauss system model is put forward on Extended Kalman Filter （EKF） ,Particle Filter （PF） and Regularized Parti- cle Filter （RPF）. Then,the performance of these three algorithms is analyzed in different noise systems. Simulation results reveal that the estimation performance of PF and RPF is much better than EKF in the nonlinear Gauss system. Because RPF reconstitutes the approximately continuous distribution from the discrete distribution and re-samples particles from this con- tinuous distribution. The performance of RPF is much better than PF, especially in the low noise environment.