Robust model predictive dynamics control for electric tracked vehicle combined with disturbance observer

Abstract

Advanced motion controllers have the potential to make automated or remotely operated vehicles less dependent on human operation. Among the different control strategies, model predictive control (MPC) has proven to have good performance in constrained systems. In this study, a combination of disturbance observer and robust model predictive control is proposed as a dynamics controller for tracked vehicles. Two different robust MPC approaches, nominal robust MPC and Tube-MPC, are compared. The latter has the potential to achieve offline computation based only on pre-planned reference states, which makes it possible to achieve real-time control with small sampling intervals. The effect of the reduced sampling interval on the state tracking accuracy is also investigated. The simulation results indicate that the nominal robust MPC has a significant advantage over the Tube-MPC when the control constraints become active and with the same sampling interval. Two model predictive controllers are evaluated on an electric tracked mobile robot. Compared to the nominal robust MPC with a sampling interval of 0.1 s, the Tube-MPC with a sampling interval of 0.03 s reduces vehicle velocity and yaw rate tracking errors by 3.8% and 9.6%, respectively.