Adaptive Robust Cubature Kalman Filter with Maximum Correntropy Criterion and Variational Bayesian for Vehicle State Estimation

Abstract The Kalman filter can accurately estimate vehicle states under the assumption of Gaussian noise. However, when the measurement system is subject to complex non-Gaussian noise disturbances, the estimation performance of traditional methods degrades significantly. To address this issue, this paper proposes an adaptive and robust nonlinear Kalman filtering method for vehicle state estimation, namely the Cubature Kalman Filter with Maximum Correntropy Criterion and Variational Bayesian. The proposed framework mainly consists of regression model construction, robust state estimation, and adaptive noise variance estimation. Specifically, a batch regression model is first constructed to account for linearization errors, and the MCC is introduced within this framework to derive the MCC-CKF-E algorithm. Then, a variational Bayesian approach is employed to perform online adaptive estimation of the time-varying statistics of the measurement noise covariance. Furthermore, the computational complexity of the VB-MCC-CKF-E is analyzed, demonstrating that it meets real-time application requirements. Finally, simulation experiments for vehicle state estimation under complex non-Gaussian noise scenarios show that the proposed VB-MCC-CKF-E outperforms EKF, CKF, and other MCC-based variants in terms of both robustness and estimation accuracy.