Variational Bayesian-Based Robust Weighted Cubature Kalman Filter for State Estimation

The stable operation of the system requires reliable state estimation, but the existence of measurement outliers will lead to inaccurate estimation results. As a state estimation algorithm, the cubature Kalman filter (CKF) is not robust to outliers, so some variants of CKF are proposed to overcome this problem. We propose a novel robust weighted cubature Kalman filter, which employs the variational Bayesian method to simultaneously estimate the state and weight parameters. For the current time step’s state, observations with low weight will have a less contribution. To avoid the issue of excessive convergence of state values resulting from large weight parameters, we propose an improved adaptive dynamics approach that utilizes the innovation vector and the measurement prediction error matrix to match the appropriate parameter prior expression. The simulation results demonstrate the effectiveness and robustness of the new filtering.