AKF与EFRLS在动态目标跟踪性能上的比较

Performance Comparison of AKF and EFRLS in Tracking Dynamic Targets

卡尔曼滤波是具有递推估计形式的最优滤波,但最优性的获得是在过程噪声和观测噪声统计特性已知的前提下得到的。然而,在大量的动态目标跟踪实际问题中噪声具有不确定性,因而有必要研究在噪声不确定下动态目标的跟踪算法以满足实际问题的需要。文中介绍自适应Kalman滤波对过程噪声方差的估计以及推广的遗忘因子最小二乘法对状态估计的递推公式,并且在平均误差最小准则下通过计算机仿真比较两种方法对动态目标的跟踪性能.仿真结果表明,在不确定噪声下自适应Kalman滤波能够取得比推广的遗忘因子递推最小二乘法更好的跟踪性能。

Kalman filtering is the optimal state estimation with recursive form, and the optimal performance of Kalman filtering could be obtained only by knowing the information of the process noise and measurement noise. However, in many practial problems for tracking dynamic targets the noise is uncertain. Thus, it is necessary to develop new algorithms and adapt the new problems. This paper describes the estimation of process noise covariance by adaptive Kalman filtering（AKF） and the estimation of the unknown state by recursive formula of the extended forgetting factor recursive least squares（EFRLS）. Through some simulations, the performances of these two methods in tracking dynamic targets are compared under the criterion of the minimum average error. The simulation results show that EFRls could satisfactorily track the dynamic targets in condition of correlated noises, and this adaptive Kalman filtering method could aquire better performance than EFRLS in condition of uncertain noises.