Development and Evaluation of a Combined Driveline Oscillation and Traction Controller Using Model Predictive Control and Reinforcement Learning: A Comparative Case Study

This paper presents a combined driveline oscillation and traction control system developed and evaluated using two advanced approaches, specifically adaptive model predictive control (MPC) and model-based reinforcement learning (RL). The objective is to suppress driveline oscillations while maintaining optimal traction across varying road conditions such as dry, wet, and snowy surfaces. Both controllers are derived from the same non-linear vehicle model and are validated through pre-defined test scenarios. The MPC approach employs input and output trajectory tracking with soft constraints to ensure feasible control actions even in the presence of constraint violations and is further supported by a Kalman filter for robust state estimation and prediction. In contrast, the RL controller leverages the model-based DreamerV3 algorithm to learn control policies autonomously, adapting to different road conditions without relying on external information. The results indicate that both controllers achieve comparable overall performance although the MPC solution provides more precise input tracking and smoother control actions, while the RL approach inherently adapts to changing road surfaces, eliminating the need for prior friction knowledge or online friction estimation. We discuss the trade-offs between MPC and RL in terms of complexity, adaptability and performance as well as avenues for future work, such as integrating road-condition estimation into the MPC framework and refining the RL controller for smoother, more precise control action.