A new prediction unscented Kalman filter based on robust model and its application

With the increase of state dimension, the calculation of UKF algorithm increases rapidly, and UKF is more sensitive to model error, and it is not suitable for the system model with noise as non-Gaussian distribution. Aiming at this problem, this paper proposes a robust model predictive Unscented Kalman filter based on the study of robust estimation, model predictive filtering and UKF. The algorithm uses the expansion method to add the driving noise to the system state, which increases the state information of the system. The model predictive filter (MPF) is used to suppress the model error, and the robustness of the system is improved by using the robust estimation. The algorithm is sensitive to the error of the model. The proposed algorithm is applied to the SINS / BNTS / CNS integrated navigation system and compared with the adaptive EKF and the robust adaptive UKF algorithm. The results show that the proposed algorithm can effectively suppress the position error and velocity error, the filtering performance Significantly better than adaptive EKF and robust adaptive UKF.