Minimum Mixture Error Entropy-Based Robust Cubature Kalman Filter for Outlier-Contaminated Measurements

Abstract

This letter investigates the robust state estimation of the nonlinear systems with outlier-contaminated measurements. Due to the advantage of mixture error entropy with two kernel bandwidths in handling non-Gaussian noise caused by the outliers, a novel minimum mixture error entropy (MMEE) criterion-based robust cubature Kalman filter is proposed, in which the cost function is constructed by MMEE criterion, and the nonlinear measurement model is linearized by the statistical linear regression method. By a benchmark target tracking scenario with non-Gaussian measurement noise and INS/GNSS loose combination vehicle tracking experiment, the effectiveness of the proposed filter is demonstrated.