Optimizing AGV utilization and battery life in automated container terminals: focus on a novel charging strategy and reinforcement learning algorithm

In response to the initiative for environmental protection and low-carbon ports, most automated container terminals (ACTs) primarily use Automated Guided Vehicles (AGVs) as their horizontal transportation tool. AGVs are powered by electricity, and their charging process affects utilization and battery life. This paper designs a Shallow Charge and Shallow Discharge Charging Strategy Based on Idle Time and Thresholds (SCSD-ITT) to optimize the charging scheduling of AGVs. Additionally, considering the differences in travel speeds and power consumption rates of AGVs under various conditions, as well as constraints on battery life degradation, a bi-objective optimization model was established. The model aims to minimize the final task completion time of AGVs and the cumulative cost of battery life loss. By comparing the solution results of the Win or Learn Fast rule and Policy Hill-Climbing (Wolf-PHC) reinforcement learning algorithm, the Non-dominated Sorting Genetic Algorithm-II (NSGA-II), and other related algorithms, the effectiveness of the established model and the superiority of the Wolf-PHC algorithm were verified. Moreover, statistical analyses via Friedman and Wilcoxon signed-rank tests further validate the advantages of Wolf-PHC. Case analysis indicates that the SCSD-ITT charging strategy can enhance AGV utilization and reduce battery life loss. This strategy improves the overall operational efficiency and sustainability of ACTs.