适于无源阵列跟踪的粒子滤波交互多模型算法

针对无源阵列被动跟踪效果较差的问题,融合交互式多模型和粒子滤波方法,提出了一种基于粒子滤波的交互多模型(IMM-PF)算法。该算法采用多模型结构跟踪目标的任意机动;各模型采用粒子滤波算法处理非线性、非高斯问题。各模型中相对固定数目的粒子群经过相互交互、粒子滤波后再进行重抽样以减少滤波退化现象。在交互阶段,对各模型的相应粒子进行输入交互;在滤波阶段,抽取N个采样点,得到估计采样,从而求得估计输出和有关函数;在混合阶段,获得状态向量的后验条件概率密度函数,通过这个后验概率密度便可获得状态向量的估计量。与典型的交互式多模型算法(IMM-KF)进行了比较,计算机仿真结果证实了本文新算法的正确性和有效性。

Algorithm of Particle Filter Interacting Multiple Model Suitable for Passive Array Tracking

To solve the ineffective performance of passive array tracking,this paper presents an interacting multiple model particle filter algorithm(IMM-PF) by combining the interacting multiple model with the particle filter method together.In using this algorithm,the structure of multiple models is adopted to track arbitrary maneuvering of the target,and at the same time particle filter method is employed in each model to deal with the nonlinear/non-Gaussian problems.After interaction and particle filtering,particles in each model with the fixed number are re-sampled to reduce the degeneracy of filtering.First,in the interaction stage,the particles corresponding to each model are input and interacting.Then,estimation resample is obtained by picking out N sampling points in the filtering stage,thereby the estimation output and the related function are gained.In the combination stage,the posteriori probability density functions of the state vectors are obtained,by combining the probability density functions of the different modes taking into account the mode probabilities.In the simulations,by comparison with the general interacting multiple model,the results demonstrate the correctness and efficiency of this new filtering method.