Autonomous overtaking decision making of driverless bus based on deep Q-learning method

The autonomous overtaking maneuver is a valuable technology in unmanned vehicle field. However, overtaking is always perplexed by its security and time cost. Now, an autonomous overtaking decision making method based on deep Q-learning network is proposed in this paper, which employs a deep neural n...

Full description

Saved in:
Bibliographic Details
Published in2017 IEEE International Conference on Robotics and Biomimetics (ROBIO) pp. 2267 - 2272
Main Authors Yu, Lingli, Shao, Xuanya, Yan, Xiaoxin
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.12.2017
Subjects
Atmospheric modeling
Decision making
deep learning
deep Q network
Learning (artificial intelligence)
Meters
overtaking decision making
reinforcement learning
Roads
Security
Training
Online AccessGet full text
DOI10.1109/ROBIO.2017.8324756

Cover

More Information
Summary:The autonomous overtaking maneuver is a valuable technology in unmanned vehicle field. However, overtaking is always perplexed by its security and time cost. Now, an autonomous overtaking decision making method based on deep Q-learning network is proposed in this paper, which employs a deep neural network(DNN) to learn Q function from action chosen to state transition. Based on the trained DNN, appropriate action is adopted in different environments for higher reward state. A series of experiments are performed to verify the effectiveness and robustness of our proposed approach for overtaking decision making based on deep Q-learning method. The results support that our approach achieves better security and lower time cost compared with traditional reinforcement learning methods.
DOI:10.1109/ROBIO.2017.8324756