摘要
现有的二阶互差分(SOMD)算法能够给出与状态估计误差解耦的观测噪声协方差估计,但是需要满足冗余测量的条件,但这一条件往往难以满足。针对这一问题,提出了一种利用状态预测值构造相邻2个时刻伪观测的方法,将原SOMD算法扩展到具有单测量的系统中。使用目标跟踪问题对该算法的有效性进行验证。仿真结果表明,当采样周期较小时,该算法能够忽略状态估计误差的影响并给出较准确的观测噪声方差,在精度和鲁棒性方面优于其他参考算法。
Existing second-order mutual difference(SOMD) algorithm can give the estimation of measurement noise covariance decoupled with the state estimation error, but it requires redundant measurements which are generally not satisfied. This paper proposes a method for constructing pseudo-measurement of two adjacent moments using state prediction, and then the SOMD algorithm is expanded to the system with a single measurement. The efficiency of the approach is verified via a target tracking problem. Simulation results indicate that the proposed algorithm can ignore the influence of state estimation error when the sampling period is small and provide accurate measurement noise properties, which is superior to other reference algorithms in accuracy and robustness.
作者
蒋刘洋
张海
JIANG Liu-yang;ZHANG Hai(School of Automation Science and Electrical Engineering,Beihang University,Beijing 100191,China)
出处
《导航定位与授时》
2020年第1期73-79,共7页
Navigation Positioning and Timing
基金
国家重点研发计划(2017YFC0821102,2016YFB0502004)
北京市科技计划项目(Z171100000517006)
关键词
自适应滤波
观测噪声方差
冗余测量
伪观测
目标跟踪
Adaptive filters
Measurement noise covariance
Redundant measurement
Pseudo-measurement
Target tracking
