A hybrid algorithm combining ant colony optimization and large neighborhood search based on reinforcement learning for integrated container-truck scheduling problem

The integrated container-truck problem is a classical problem in the logistic industry, which combines the vehicle routing problem and container allocation problem. In order to solve this problem efficiently, this paper proposed a hybrid solution approach combining the merits of ant colony optimization (ACO) and large neighborhood search (LNS). In the solution process, deep reinforcement learning is used to dynamically adopt the most suitable algorithm (ACO and GA) to conduct the trial-and-error process. The pheromone trails generated by ACO are used as the space, and the current solution generated by LNS is used as the action space. Then the objective value is used as the reward signal. The reinforcement learning agent is trained such to learn the best policy. The proposed approach is suitable to solving discrete container-truck scheduling problem and improve the convergence speed of ACO algorithm. Our computational study demonstrates the effectiveness of the proposed algorithm.