Reinforcement Learning Optimized MPC for Vehicle Stability with Independent Four-Wheel Drive and Steering

To enhance vehicle maneuverability and stability under extreme conditions, this paper proposes a Reinforcement Learning Optimized Model Predictive Controller (RL-MPC) for vehicles equipped with four-wheel independent driving and steering (4WID-4WIS). Through a 7DoF vehicle dynamics model combined with driver inputs, control target values for side slip angle and yaw rate are established. An MPC is designed using a linear 2DoF model to approximate these target values while minimizing the rate of change of the rear-wheel steering angle and the additional yaw moment. Reinforcement learning is then utilized to automatically adjust the weight coefficient matrix of the MPC to adapt to varying conditions. Finally, the effectiveness of the proposed algorithm is validated on the benchmark challenge platform by comparing its performance with traditional PID and MPC controllers under three typical conditions.