Multi-Agent Reinforcement Learning based Warehouse Task Assignment

The rise of e-commerce demands greater efficiency in warehouses, requiring dynamic task allocation among humans and robots. Traditional methods often fail in such complex environments. This research introduces a MultiAgent Reinforcement Learning (MARL) framework to optimize task assignments. The model considers detailed task and device attributes like type, urgency, and availability. Agents use Deep Q-Networks (DQNs) to make informed decisions based on perceived warehouse state. A carefully designed reward system prioritizes timely order completion and efficient task execution. Simulations show significant improvements in efficiency, especially in pick task completion and device use. The results confirm MARL’s potential for optimizing warehouse operations.