Railway Adaptive Dispatching Decision Method Based on Double DQN

Rail transit has the advantages of stability, high efficiency, and no congestion. It is an essential traveling means for people currently. Combined with the real-time flow, the adaptive dispatch scheme can reduce operating costs and passenger waiting time. This paper designs an MDP simulation environment model for rail trains and gives an environment model under regular and occasional passenger flows. We combined the deep reinforcement learning method based on the value function, gave the method of feature extraction, and conducted experiments on the scheme under regular and occasional passenger flow. The results show that the combination of deep reinforcement learning methods can meet the needs of adaptive dispatch.