改进型粒子滤波算法在多站纯方位被动跟踪中的应用

Application of improved particle filter algorithm to bearings-only passive tracking in multiple stations

针对多站纯方位被动定位与跟踪问题,给出了一种基于均匀重采样和带自适应因子的改进型粒子滤波算法.首先,基于无迹卡尔曼(UKF)粒子滤波器,将参考分布融入最新观测信息,得到符合真实状态的后验概率分布;借助重采样和使用鲁棒估计,改善了粒子滤波的退化问题.其次,引入自适应因子以调整UKF的状态模型协方差与观测模型协方差的比例,得到较高精度的概率分布.仿真结果表明,改进的粒子滤波算法能够实现多站纯方位被动跟踪,比传统非线性滤波器有更高的跟踪精度.

For the problem of bearings-only passive localization and tracking in multiple stations, we propose an improved particle filter algorithm with an adaptive factor based on evenly re-sampling. In the unscented Kalman filter（UKF） particle filter, the posterior probability distribution of true state-values is obtained by integrating the reference distribu- tion with the latest observed information. The degeneracy phenomenon in the particle filter is relieved by re-sampling and robust estimation approaches. By introducing an adaptive factor for adjusting the proportion between the state-model covariance and the observation-model covariance of UKF, we obtain a probability distribution with higher precision. Simulation results show that the proposed particle filter algorithm provides higher precision than the traditional nonlinear filters in bearings-only passive localization and tracking for multiple stations.