Algorithm of Automatic Train Operation System Based on Reinforcement Learning
Abstract— Achieving good speed control is the key to the smooth operation of the automatic driving system. In this paper, the image processing method in deep learning is introduced, combined with the classic deep neural network algorithm, and the CNN is strengthened through the reinforcement learning method. A speed controller with good tracking effect is designed. This method enhances the universality of the fuzzy control method, meanwhile, it creatively puts forward the idea of using the combination of neural network and reinforcement learning to achieve a good effect of automatic train operation control.
Index Terms— ATO control, reinforcement learning, CNN network.
Yanmei Guo, Ziheng Wu, Xiangxian Chen
Department of Instrumentation Science and Engineering, Zhejiang University, CHINA
Zhejiang Train Intelligent Engineering Technology Research Center Co., Ltd, CHINA
Cite: Yanmei Guo, Ziheng Wu, Xiangxian Chen, Zhujun Ling, "Algorithm of Automatic Train Operation System Based on Reinforcement Learning," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 705-712, Bangkok, 28-30 June, 2018.