A New Reinforcement Learning for Multi-Train Marshaling with Time Evaluation

Abstract

In this paper, a new reinforcement learning method is proposed to solve a train marshaling problem for assembling several outgoing trains simultaneously. In the addressed problem, the order of the incoming freight cars is assumed to be random. Then, the freight cars are classified into several sub-tracks. The cars on sub-tracks are rearranged to the main track by a certain desirable order. In the proposed method, each set of freight cars that have the same destination make a group, and the desirable group layout constitutes the best outgoing trains. When a rearrangement operation is conducted, the best number of sub-tracks used in the operation is obtained by a reinforcement learning system, as well as the best layout of groups in the trains, the best order to rearrange cars by the desirable order, and the best sub-track for the car to be removed. The marshaling plan that consists of series of removal and rearrangement operations are generated based on the processing time of movements of freight cars. The total processing time required to assemble outgoing trains can be minimized by the proposed method.