Open AccessAction Permissibility Prediction in Autonomous Driving through Deep Reinforcement Learning
Author(s) -
Jiaoyang Wang,
Qi Sun,
Dechen Yao,
Feng Xiong
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/782/3/032062
Subject(s) - reinforcement learning , computer science , task (project management) , artificial intelligence , action (physics) , state space , state (computer science) , algorithm , engineering , mathematics , systems engineering , statistics , physics , quantum mechanics
This paper proposes a new nature based on deep deterministic policy gradient for deep reinforcement learning in continuous state and action space, which can greatly accelerate the emergence of artificial intelligence training and solving problem: state-action permissibility. The proposed method is integrated into the latest DDPG algorithm to guide its training and is applied to solve the lane keeping (steering control) problem in autonomous driving or automatic driving. Finally, the TORCS which is the open racing car simulator simulation software builds various simulation environments, including different tracks to verify the effectiveness of the algorithm. The results show that the proposed method can significantly speed up the algorithm training speed of the lane keeping task.