Real-Time Path-tracking MPC for an Over-Actuated Autonomous Electric Vehicle

Abstract

This paper illustrates the development of a nonlinear constrained predictive path-tracking controller, including realistic vehicle dynamics and multiple actuator inputs and its implementation in real time on an experimental vehicle platform. The controller is formulated for a particular over-actuated vehicle equipped with Torque Vectoring (TV) as well as All-Wheel-Steering (AWS) functionalities, which allow for the enhanced control of vehicle dynamics. The proposed Nonlinear Model Predictive Controller (NMPC) takes into account the nonlinearities in vehicle dynamics across the range of operation up to the limits of handling as dictated by the adhesion limits of the tyres. In addition, crucial constraints regarding the actuators’ physical limits are included in the formulation. The performance of the controller is demonstrated in a high fidelity simulation environment, as well as in real-time on a test vehicle, during the execution of demanding driving scenarios.