Research on ATO Control Method for Urban Rail Based on Deep Reinforcement Learning

Aiming at the problems of punctuality, parking accuracy and energy saving urban rail train operation, an intelligent control method for automatic operation (ATO) based on deep Q network (DQN) is proposed. The dynamics model established under condition satisfying safety principle various constraints driving train. Considering transformation rules sequences working conditions between stations, agent in DQN algorithm used as controller to adjust strategy real time according operating state environment, optimizes generation curve. Taking Beijing Yizhuang Subway line example, simulation test results show that reduces consumption by 12.32% compared with traditional PID method, improves running punctuality accuracy; same time, automatically can dynamically, has good adaptability robustness change environment parameters.