基于Huber的改进鲁棒滤波算法

Huber-based Modified Robust Filter Algorithm

针对在观测噪声为非高斯强噪声的情况下,传统Kalman滤波将会失效,同时基于l1/l2联合范数的Huber法,其估计精度也会降低等问题,提出一种利用新息卡方检测法预判断的鲁棒滤波算法,该算法可以抑制观测连续非高斯强噪声的影响,提高滤波精度及稳定性,具有良好的鲁棒性。仿真实验对比了四种滤波方法在不同混合高斯噪声环境下的性能,结果表明:进行了卡方检验预判断的鲁棒滤波算法具有更高的状态估计精度和稳定性。

A modified robust filtering algorithm using of information detection method prejudges was proposed to establish a problem that Kalman filter and Huber-based robust filter of estimation accuracy reduced or failed for the non-Gaussian intensity observation noise. The modified algorithm could suppress non-Gaussian intensity observation noise and improve the accuracy and stability with excellent robustness. The performance of Kalman filter, detection filter, Huber-based filter and modified robust filter in environment with different Gaussian mixture were compared. The experimental results show that modified robust filtering algorithm has better state estimation of accuracy and stability.