A Deep Reinforcement Learning Method for Multi-Load AGVs Scheduling with Task Priority

This paper presents an approach to optimize task scheduling for multi-load Automated Guided Vehicles (m-AGVs) in logistics environments using Deep Reinforcement Learning (DRL). For this, the task scheduling problem has been modeled as a Markov Decision Process (MDP) and utilized a Deep Q-Network (DQN) to learn optimal dispatch with mixed-rule policy frameworks. The proposed DQN approach significantly reduces task delays, especially for higher-priority tasks, and outperforms traditional scheduling rules with increases of 73% in reward and performs better at least 25% compared to mix rules. This work highlights the effectiveness of DRL in solving complex logistics problems and improving the efficiency of m-AGVs in real-time operations.