Robust Gaussian Filtering based on M-estimate with Adaptive Measurement Noise Covariance

In this paper, a robust Gaussian filtering is proposed based on M-estimate with adaptive measurement noise covariance. In the proposed method, the M-estimate is incorporated into the Gaussian filtering framework through modifying the measurements residues to introduce robustness. The modified measurements residues are also stored to identify the measurement noise covariance based on Myers-Tapley method. The proposed method can handle both the non-Gaussian measurement noise and inaccurate prior knowledge of the measurement noise covariance.