An RL-Based Path Tracking Control Strategy with Rollover Prevention and Safety Enhancement for the Autonomous Truck

This paper presents an integrated path tracking control framework for an autonomous truck with the reinforcement learning (RL) and safe guarantee. It takes advantages of the modular structure and optimization properties of the control distribution method. The proposed integrated framework consists of three parts. In the upper level, a learning-based path tracking control strategy is proposed to follow the reference path accurately with RL. In the middle level, a new active safety control method is designed to reduce rollover risk and improve lateral stability with the 3 degrees of freedom (DOF) reference model. In the lower level, an anti-rollover torque allocation controller based on the active torque output is designed for the driving smoothness by means of optimization, considering the dynamic load change and the actuator power limits. Finally, a validation is conducted in a simulation platform to test the performance. The results show that the integrated control strategy can help the heavy truck accurately follow the path, avoid rollover risk and enhance driving smoothness.