Unified Form for the Robust Gaussian Information Filtering Based on M-Estimate

Abstract

In this paper, a unified form for robust Gaussian information filtering based on M-estimate is proposed, which can incorporate robust weight functions with zero weight for large residues. The proposed information filtering framework can avoid the numerical problem introduced by the zero weight in the Kalman filtering framework. Simulation results show the efficiency and superiority of the proposed robust filters over the nonrobust filter against heavy-tailed measurement noises.