基于观测预测粒子滤波的雷达组网目标状态估计

雷达组网系统跟踪目标时，观测数据与目标跟踪状态成严重非线性关系，难以用卡尔曼滤波最优估计方法，处理非高斯非线性系统滤波估计问题的粒子滤波算法容易产生粒子退化问题。因此，使用观测预测粒子滤波算法解决这个问题，该算法基于观测似然进行重要性采样，结合一步预测信息计算粒子权值，保证了采样粒子处于高观测似然区，并充分利用了一步预测信息。仿真验证表明，将观测预测粒子滤波算法应用于目标状态估计，避免了粒子退化，收敛快，估计精度高。

Study on target state estimation in radar networking based on observation-prediction particle filter

Tracking target with radar networking system(RNS), observation data has the nonlinear relation with state value. So it is hard to use Kalman filter(KF) algorithm, one of optimal estimating algorithm. Particle filter algorithm can be used to estimate state in any Non-Gaussian and nonlinear system in theory. But it is prone to particle degradation problem. Observation-prediction particle filter(OPPF) algorithm is proposed in this paper to solve the problem. Based on observation likelihood importance sampling and prediction information weight calculating, OPPF can make sampling particles in region of high observation likelihood and make full use of prediction information. The simulating verification demonstrates that using OPPF in target state estimating has fast convergence and high precision of estimating and solves particle degradation problem.