A hierarchical online planning framework for anticipatory yard allocation optimization in dry bulk terminals
提出一个三层分层在线规划框架，结合强化学习预测、整数规划预分配和启发式实时调度，在宁波舟山港鼠浪湖码头将运营与缺货成本降低69.9%，适用于复杂堆场及仓储物流场景。
Growing throughput in maritime dry bulk ports demands improved yard allocation to reduce operational and task-missing costs. We address this challenge by proposing a Hierarchical Online Planning (HOP) framework with three layers: a forecasting layer leveraging reinforcement learning to capture temporal patterns and estimate future cargo proportions, a pre-planning layer employing integer programming to optimally allocate yard space, and an implementing layer using heuristic rules for real-time operations. Numerical experiments at the Shulanghu terminal in Ningbo-Zhoushan Port demonstrate the framework’s effectiveness in significantly reducing both operational and task-missing costs. Compared to both the business rule (BR) benchmark currently used at the port and a neural-network-based planning (ICT) benchmark we proposed, our framework reduces total costs by 69.9% and 35.3% under normal operation conditions, and by 19.2% and 12.4% under stress-test conditions, respectively. From this research, we demonstrate that integrating forecasting, yard space optimization, and adaptive execution significantly decreases overall costs. This method offers a strategic framework for managing complex stockyards and can be readily adapted to broader warehousing and logistics scenarios.