基于马尔可夫切换系统的固定时滞平滑算法

Fixed-Lag Smoothing Algorithm for Markov Switching Systems

提出了一种基于马尔可夫切换系统的固定时滞平滑算法,在交互式多模型结构框架内引入状态增广矩阵,将目标状态的当前值和过去值相结合,使状态预测与平滑算法同步进行,算法还应用当前观测数据估算时滞模型概率,方法简便易行。通过仿真对比跟踪性能可以得出:固定时滞平滑算法的性能要优于标准IMM-EKF滤波算法,并随着时滞长度的增加,性能趋向更优。

A suboptimal approach to the fixed-lag smoothing problem was proposed based on Markovswitching system. By introducing state-augmented system in the interacting multiple model framework, whichcombined the current state with the past one, the state prediction was made synchronously with the smoothingfor the original system. The fixed-lag mode probabilities were estimated by using the current observationdata, which is a simple but effective method. The proposed smoothing algorithm was compared with that ofthe normal IMM-EKF filtering algorithm through the trajectory of a maneuvering target. The result shows thatthe performance of the fixed-lag IMM smoothing algorithm is great better than that of later, and theperformance tends to get better with the increasing of the lag.