Computer Engineering and Applications ›› 2019, Vol. 55 ›› Issue (11): 74-79.DOI: 10.3778/j.issn.1002-8331.1803-0470

Previous Articles     Next Articles

Single-Observer Tracking Algorithm Based on M-Estimation Robust Backward-Smoothing CKF

REN Zhen1, LI Jiying1, WU Hao2   

  1. 1.College of Electronic & Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China
    2.College of Information and Navigation, Air Force Engineering University, Xi’an 710077, China
  • Online:2019-06-01 Published:2019-05-30

基于M估计的鲁棒后向平滑CKF单站跟踪算法

任  臻1,李积英1,吴  昊2   

  1. 1.兰州交通大学 电子与信息工程学院,兰州 730070
    2.空军工程大学 信息与导航学院,西安 710077

Abstract: M-estimated based Robust Backward-Smoothing Cubature Kalman Filter(MR-BSCKF) algorithm is proposed. The algorithm introduces the improved M-estimation idea into the Backward-Smoothing Cubature Kalman Filter(BSCKF) algorithm, introduces the Mahalanobis distance to construct the P-Huber equivalent weight function, and the robustness of the filtering algorithm is further increased by reducing outliers misjudgment. The algorithm introduces the backward-smoothing function into the traditional CKF algorithm, and the filtering accuracy is further improved by the secondary filtering combined with backward smoothing and forward filtering, that achieving filtering accuracy and robustness at the same time. The simulation results show that MR-BSCKF, compared with the traditional algorithm, can get more accurate target tracking results in the presence and absence of outliers, and the robustness is more robust.

Key words: single station target tracking, nonlinear filter, cubature Kalman filter, backward-smoothing, outliers, M-estimation

摘要: 提出一种基于M估计的鲁棒后向平滑容积卡尔曼滤波(M-estimated based Robust Backward-Smoothing Cubature Kalman Filter,MR-BSCKF)算法。该算法将改进的M估计思想引入后向平滑容积卡尔曼滤波(BSCKF)算法中,引入Mahalanobis距离构建P-Huber等价权函数,通过降低野值误判概率进一步提高滤波算法的鲁棒性;在传统CKF算法的基础上增加后向平滑函数,通过后向平滑和前向滤波相结合的二次滤波进一步提高滤波的精度,实现了算法精度和抗野值能力的统一。仿真结果表明,与传统算法相比,MR-BSCKF在有野值和无野值的情况下都能够得到更加准确的目标跟踪结果,且鲁棒性更强。

关键词: 单站目标跟踪, 非线性滤波, 容积卡尔曼滤波, 后向平滑, 野值, M估计