A reinforcement learning method for container terminal storage space allocation problem

Storage space allocation is a critical and complex decision-making task in container terminal operations, characterized by uncertainty, multiple objectives, and large-scale problem sizes. Traditional methods often rely on two-stage modeling or heuristic algorithms, which suffer from limited generalization and poor scalability. To address these limitations, this paper proposes a deep reinforcement learning (DRL) approach for the Storage Space Allocation Problem (SSAP). A comprehensive SSAP model is constructed, and its objectives and constraints are transformed into reward functions and environmental rules within the DRL framework. Instead of solving the problem directly, the DQN is trained to learn optimal allocation strategies through interaction with the simulated terminal environment. Extensive numerical experiments are conducted to evaluate the proposed method against genetic algorithms, CPLEX, and DQN variants. Results demonstrate that the DRL approach achieves superior computational efficiency, scalability, and solution quality, especially in large-scale instances. Furthermore, terminal KPIs such as reshuffle rate, crane travel distance, and yard utilization imbalance are analyzed to validate the practical effectiveness of the method. The proposed framework provides a general and efficient solution for intelligent storage space allocation in container terminals.