Evolving Priority Rules for Online Yard Crane Scheduling with Incomplete Tasks Data

In the last decade, the surge in global container port throughput has heightened the need for terminal efficiency. The loading process plays a crucial role in overall port performance. However, the unpredictable arrival of external trucks poses challenges for yard cranes in scheduling both internal loading tasks and external truck tasks simultaneously. Existing approaches on yard crane scheduling, considering uncertain arrivals, typically rely on prior knowledge, which often fails to fully capture the nature of the real-life uncertainties. In response, we propose an online scheduling approach guided by a two-stage decision model, eliminating the need for prior knowledge of uncertain arrival and has the ability to dynamically adapt to different scenarios. In the look-ahead stage, future tasks are filtered dynamically to eliminate undesired tasks, followed by a priority rule guided selection stage, where the task with the highest priority is selected. Genetic Programming (GP) is employed for automated evolution of priority rules without human intervention. Realistic experiments showcase the effectiveness of the proposed dynamic look-ahead method compared to static minimum and maximum look-ahead, as well as the superiority of GP-evolved priority rules compared to manually crafted priority rules in terms of both performance and simplicity. A comprehensive analysis of GP-evolved rules highlights GP's proficiency in problem understanding and rule extraction, comparable to human experts.