结合粒子滤波和Kalman滤波的边缘粒子滤波(MPF)是一种新的混合线性/非线性系统的状态估计方法,但是粒子滤波在计算上的复杂使得MPF难以兼顾系统实时性和精度的要求。针对此问题,提出一种基于相似采样粒子滤波算法的MPF滤波框架。算法从系统观测值中采样粒子,并通过一个计算相邻时刻粒子转移概率的步骤,提高了粒子使用率,使得算法能以少量粒子实现对非线性状态量的估计,进而提高Kalman滤波的精度和实时性。给出了算法原理分析和实现流程。以混合坐标系下的目标跟踪为对象,利用蒙特卡罗仿真研究了ILLH_MPF算法的应用,并与常规MPF方法进行了对比。

The Marginalized Particle Filter(MPF),which combines Particle Filter(PF) with Kalman filter,is an effective algorithm for the estimation of mixed linear/nonlinear system,but the computational complexity of PF make it difficult to meet the requirement of real-time and precision in estimation. A likelihood-sampling PF was introduced to join in the MPF algorithm. The ILLH_MPF algorithm sampled the particles from the observation,and increased the utilization factor of particles by calculating the transition probability of adjacent times particles. Compared with MPF,this algorithm was expected to get a good estimation precision with fewer particles,and could improve the Kalman filters precision and efficiency. The principle and detailed procedure of this algorithm was introduced. The Monte-Carlo simulation was designed to show the ILLH_MPF algorithms application for a problem of target-tracking in the hybrid coordinate.