A New Robust Centered Error Entropy Cubature Kalman Filter

Abstract

The heavy-tailed non-Gaussian noise often appears in the actual system and the classical cubature Kalman filter (CKF) algorithm will have reduced filtering accuracy or even filtering divergence in this condition. To make the CKF algorithm more robust, the centered error entropy cubature Kalman filter (CEECKF) algorithm is derived by combining the Spherical-Radial cubature rule and the centered error entropy (CEE) criterion. The proposed algorithm uses the cubature rule to obtain the one-step prediction state mean and covariance and then uses the CEE criterion to update the posterior state. The application in attitude determination shows the effectiveness of the algorithm.