Autonomous Vehicle Motion Control and Energy Optimization Based on Q-Learning for a 4-Wheel Independently Driven Electric Vehicle.

The article focuses on developing a hierarchical control strategy for four-wheel independently driven (4WID) electric vehicles (EVs) to optimize autonomous vehicle motion control and energy consumption. It introduces a longitudinal vehicle dynamics model enhanced by an extended state observer (ESO) to estimate model errors, combined with a linear quadratic regulator (LQR) for precise speed tracking. Additionally, a Q-learning-based torque distribution algorithm dynamically allocates torque among the four wheels to minimize vehicle body motion (heave, pitch, roll) and energy use. Simulation results on benchmark problems from the 62nd IEEE Conference on Decision and Control (CDC) 2023 demonstrate that this combined ESO+LQR and Q-learning approach outperforms conventional PID and fuzzy rule controllers in speed tracking accuracy, vehicle stability, and energy efficiency. The study acknowledges that lateral control was simplified using a proportional controller and suggests future work to enhance lateral dynamics control.