Robust Derivative Unscented Kalman Filter Under Non-Gaussian Noise

Abstract

A robust derivative unscented Kalman filter is proposed for a nonlinear system with non-Gaussian noise and outliers based on Huber function. In this paper, the time update process can be performed using a Kalman filter (KF), and measurement update process can be carried out utilizing an unscented transform. This novel filtering algorithm differs from the traditional unscented Kalman filter (UKF) which is sensitive to non-Gaussian noise and outliers. Besides, for a nonlinear system with linear state equation, the presented filtering has the merits of both KF and UKF, and it has better performance under non-Gaussian noise and large outliers. Furthermore, the proposed filtering algorithm can not only track the target effectively but also suppress the effects of the non-Gaussian noise and large outliers. Finally, numerical simulations are conducted to verify estimation accuracy and effectiveness of the proposed filtering algorithm.