The developed learning-based MPC has been implemented on an all-wheel drive (AWD) Chevrolet Equinox electric vehicle featuring four independent electric motors for each corner wheel. This study focuses on torque vectoring on the front axle, while propulsion torque is applied through the two motors located on the rear axle. The developed learning MPC controller is evaluated through three repeated DLC maneuvers on a wet surface.
The results showed that the yaw rate tracking error and side slip angle progressively diminished over the repeated DLC maneuvers as more data was collected for learning. Specifically, the tracking error is significantly reduced in the third DLC maneuver compared to the first and second DLC maneuvers.
The results demonstrate the superiority of the hybrid learning MPC over conventional MPC. The system exhibits enhanced yaw rate tracking, better sideslip control, and improved overall vehicle stability during aggressive maneuvers. The integration of learning techniques significantly boosts the prediction accuracy of vehicle states, enabling more precise and proactive control actions through torque adjustment.
