A Reinforcement Learning-Based AGV Scheduling for Automated Container Terminals With Resilient Charging Strategies

Abstract

Automated guided vehicles (AGVs) serve as pivotal equipment for horizontal transportation in automated container terminals (ACTs), necessitating the optimization of AGV scheduling. The dynamic nature of port operations introduces uncertainties in AGV energy consumption, while battery constraints pose significant operational challenges. However, limited research has integrated charging and discharging behaviors into AGV operations. This study innovatively proposes an AGV scheduling model that incorporates a resilient and adaptive charging strategy, adjusting the balance between vehicle charging and the completion of transportation tasks, enabling AGVs to complete fixed container transportation tasks in the shortest time. Differing from most existing research primarily based on OR-typed algorithms, this study proposes a reinforcement learning-based AGV scheduling method. Finally, a series of numerical experiments, which is based on a real large-scale automated terminal in the Pearl River Delta (PRD) region of Southern China, are conducted to verify the effectiveness and efficiency of the model and the algorithm. Some beneficial management insights are obtained from sensitivity analysis for practitioners. Notably, the paramount observation is that the operational efficacy of AGVs does not necessarily correlate positively with their number. Instead, it follows a “U-shaped” curve trend, indicating an optimal range beyond which performance diminishes.