Speed Regulation of Overhead Catenary System Inspection Robot for High-Speed Railway through Reinforcement Learning

High-speed railway has developed rapidly in recent years. The overhead catenary system (OCS) that transmits electrical energy to the train has to be regularly inspected due to catenary-related defects might directly threaten the safe operation of high-speed railway. In this paper, we present a novel method for autonomous speed regulation of OCS inspection robot using reinforcement learning. To train the robot, we first build a simulation platform based on Unity3D for accurate reconstruction of the railway environment including the OCS, the rail track and the inspection robot. Next, we utilize the range data recorded by the LiDAR mounted on the robot to detect OCS components. Then we leverage the detection results to train the robot to learn speed regulation for shortening the inspection time through reinforcement learning. Experimental results show the validity of the proposed method, which can greatly improve inspection efficiency. Our work has immediate practical significance for high-speed railway automation and informatization.