An Improved Variational Bayesian Adaptive Robust Kalman Filter

Abstract

Inaccurate or vague system noise model (Q and R) of the Kalman filter (KF) will seriously affect the accuracy of filter estimation. In this paper, the noise model is adapted by the variational Bayesian method, and the outliers in the measurement are eliminated by the Mahalanobis distance criterion method. The proposed IVBARKF inherits the excellent adaptability of the variational Bayesian method and the strong robustness of the Mahalanobis distance criterion. The effectiveness is verified by the target tracking experiment. The proposed method has better performance than conventional method when the noise model is inaccurate and contaminated by outliers.