汽车电子研究中心

Research Direction

Based on the development of the intelligent driving vehicles and the related key technologies, systematic interdisciplinary research has been carried out from the communications security, laser radar, machine vision, active safety, energy saving such aspects.
1.Perception
Research on the information fusion technology of the vehicle radar and machine vision and new technology of space 3D information construction in visual detection and measurement technology
2.Autonomous driving decision
Based on the framework of deep reinforcement learning, to construct the automatic driving strategy learning model and to improve the training efficiency using parallel computing.
3.Car network communication security
Research on the encryption and decryption, identity authentication and key distribution of vehicle network access, safe driving, billing management, traffic management and data communication.
4.Energy-saving driving
Through the path planning and speed adjustment to achieve energy-saving, combined with large traffic data, so as to achieve energy conservation from the city level.

Reinforcement Learning in Autonomous Driving

Automatic driving technology is regarded as a powerful tool to solve traffic problems such as traffic accidents, congestion, and exhaust pollution. In this paper, we propose a new control strategy training method for self-driving vehicles, based on the deep reinforcement learning model. The method involves a Q-learning algorithm with filtered experience replay and pre-training with experiences from professional drivers, which accelerates the training process due to reduced exploration spaces. The method also involves resampling the state input after clustering, which improves the generalization ability of the strategy effectively, due to the individual and independent distribution of the samples. Experimental results demonstrate that the proposed model could reduce the time consumption of training by 92%, and the control stability increases by about 34%, compared with the existing neural fitted Q-iteration algorithm. In addition, procedure on resampling after clustering could increases the average travel distance by 73.4%, compared with the Q-learning algorithm with filtered experience replay that is based on the testing track which slightly more complicated than training set.

Intellectual Properties

Selected Papers

1.Junta Wu, Huiyun Li*, Deep Ensemble Reinforcement Learning with Multiple Deep Deterministic Policy Gradient Algorithm, Mathematical Problems in Engineering, vol. 2020, Article ID 4275623, 12 pages, 2020 (IF= 1.179, SCI, JCR Q3)
2.夏伟，李慧云*，基于深度强化学习的自动驾驶策略学习方法，集成技术，第6卷第3期，2017 年5 月

Selected Patents

1.发明专利，一种汽车的自动驾驶方法与装置，申请号：201710156331.8，申请日：2017.3.16
2.发明专利，一种用于车辆自动驾驶的决策网络模型的生成方法及装置，申请号：201710201086.8，申请日：2017.3.30

Shenzhen Institutes of Advanced Technology , Chinese Academy of Sciences —— Center for Automotive Electronics

Shenzhen Institutes of Advanced Technology , Chinese Academy of Sciences
—— Center for Automotive Electronics