Cooperative Tracking Control Combining with Reinforcement Learning for Heavy-Haul Trains

Le résumé fourni par la source

It is essential to ensure the consistency of the cooperative control of heavy-haul train to improve the operational efficiency and safety. Nevertheless, the dynamic environment of heavy-haul train brings the challenge for the stability of cooperative control. In this paper, a tracking control scheme is proposed by combining cooperative control and deep reinforcement learning. Firstly, a dynamic model of heavy-haul trains is proposed based on the multiple freight car cooperation and the communication topology of freight cars is established. Secondly, the cooperative control with adjustable artificial potential function is designed to eliminate the deviation between the actual distances among freight cars and the expected safety distance. Furthermore, the deep reinforcement learning is employed to compensate the control input of cooperative control, adapting to the dynamic operation environment of heavy-haul train. The simulation results verified that the proposed tracking control scheme can ensure that each freight car of the heavy-train runs at the required speed and the actual distances among freight cars can converge to the expected safety distance, minimizing the coupler force among freight cars of heavy-train.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.