Model predictive control of vehicle roll-over with experimental verification

This paper presents a model predictive approach to directly control untripped vehicle roll-over. First, a novel real-time estimation scheme is designed to provide the vehicle roll angle using an observer on a combined model of vehicle kinematics and roll dynamics. This angle is used in an integrated directional and roll dynamics model for the prediction of vehicle states and roll-over index. The roll-over prevention objective is specified as a soft constraint to ensure persistent feasibility. If the controller foresees impending vehicle roll-over, it intervenes and reduces the roll-over index by torque vectoring. Software simulations are used to assess the effectiveness of the proposed roll-over control method in the industry standard fishhook maneuvers. In addition, the accuracy and performance of the suggested estimation and control algorithms are verified in double-lane change and flick maneuvers on dry asphalt with an instrumented test vehicle. The results show accurate estimation of the vehicle roll angle and excellent control on roll-over index with the proposed predictive roll-over controller.