Multi-Objective Torque Coordination for Four-Wheel Independently Driven Electric Vehicles Based on Sliding-Mode-Guided Deep Reinforcement Learning

To address the multi-objective torque optimization and coordinated control requirements of distributed electric vehicles under coupled longitudinal-lateral dynamics, a deep reinforcement learning-based torque control strategy is proposed. A vehicle dynamics model is first established, and a non-singular fast terminal sliding mode controller is designed. To overcome typical challenges of deep reinforcement learning-such as instability, poor convergence, and limited generalization in complex dynamic systems-a yaw moment error-guided mechanism is introduced. With lateral stability, energy efficiency, tire load balance, and torque fluctuation suppression as optimization objectives, a sliding-mode-guided deep reinforcement learning strategy for coordinated torque distribution is developed. Finally, the proposed strategy is evaluated through simulation and compared with conventional methods, with in-depth discussion and analysis of the results.