基于鲁棒容积卡尔曼滤波的自适应目标跟踪算法

目标跟踪系统的观测野值将大大降低滤波算法对目标状态的估计精度.为了解决这个问题,提出了一种基于鲁棒容积卡尔曼滤波的自适应目标跟踪算法.借鉴Huber等价权函数的思想,构造了基于平方根平滑逼近函数的修正因子以抑制观测野值的影响,并结合容积卡尔曼滤波器求解框架推导出该算法.区别于Huber方法对观测残差的每个维度分别进行处理,提出的算法能够对观测残差进行综合评判.理论分析证明所提算法具有更好的数值稳定性.仿真实验表明,所提算法能够自适应地减少异常值的不利影响,与现有算法相比具有更优的滤波性能.在仿真实验中还对几种滤波算法的计算花费进行了比较,发现所提算法未大幅增加计算成本.

Adaptive target tracking algorithm based on robust cubature Kalman filter

The outliers in observations will greatly reduce the estimation accuracy of the filtering algorithm. In order to address this problem, an adaptive CKF algorithm based on robust M-estimation is proposed. Inspired by the idea of the Huber equivalent weight function, a correction factor based on the square root smooth approximation function is constructed to suppress the influence of outliers, and the proposed algorithm is derived combined with the cubature Kalman filter solution framework. Theoretical analysis proves that the algorithm has better numerical stability. Simulation experiments show that the proposed algorithm can adaptively reduce the adverse effects of the outlier and exhibit superior filter performance compared to the existing algorithms without greatly increasing the computational cost.