Du, Guodong (Beijing Institute of Technology), Zou, Yuan (Beijing Institute of Technology), Xudong, Zhang (Beijing institute of technology), Dong, Guoshun (Beijing Institute of Technology), Yin, Xin (Beijing Institute of Technology)

Heuristic Reinforcement Learning Based Overtaking Decision for an Autonomous Vehicle

Scheduled for presentation during the Regular session "Vehicle control technology" (MoAT2), Monday, August 23, 2021, 15:20−15:40, Room T2

6th IFAC Conference on Engine and Powertrain Control, Simulation and Modeling, August 23-25, 2021, Tokyo, Japan

This information is tentative and subject to change. Compiled on April 26, 2024

Abstract

This paper proposes an intelligent overtaking decision based on the heuristic reinforcement learning method for an autonomous vehicle. The proposed overtaking control focuses on the safety and efficiency of the autonomous vehicle driving. Firstly, the overtaking problem is modeled and the adaptive safe driving area is constructed. Then, a heuristic reinforcement learning method called Heu-Dyna is developed to derive the optimal overtaking decision, which introduces the heuristic planning function. Besides, the generalized correlation coefficient is designed to evaluate the training perfection of the control strategy. The simulation results show that the performance of the proposed method on the rapidity and optimality is superior to the Q-learning method and the Dyna method. Furthermore, the adaptability of the proposed method is validated by applying different driving conditions.