Double Deep Q network-based speed trajectory intelligent optimization for high-speed train

Abstract

As a part of the automatic train operation (ATO) system, the recommended speed trajectory is crucial to the safety and efficiency of train operation. This paper proposes a speed trajectory intelligent optimization approach based on a Double Deep Q network (Double DQN). The train dynamics model is established taking into account the traction and braking characteristics of the train, the line conditions, and the train position and speed constraints. The reinforcement learning environment suitable for high-speed train operation is established, including state collection, action collection, and reward function. The agent is responsible for selecting appropriate driving strategies with the aim of achieving energy efficiency, punctuality, and riding comfort under speed and position constraints. Numerical experiments based on the data of Beijing-Shanghai High-speed Railway and high-speed train CRH2 are conducted to illustrate the effectiveness of the proposed Double DQN method under the scene of fixed-line speed restriction conditions. The results show that the proposed method can improve punctuality and reduce energy consumption in case of loss of a small amount of comfort.