Design and Simulation of a Multi-AGV Warehouse System Using a Multi-Agent Reinforcement Learning Model

The proposed project revolves around the development of a multi-Agent Guided Vehicle warehouse system based on the multi-Agent Reinforcement Learning technique. The aim of the project is to create an intelligent coordination algorithm for the simultaneous operation of multiple AGVs within the context of the smart warehouse environment for effective movement and task allocation. The research involves mechanical, electrical, and algorithmic designs, such as the development of an AGV robot prototype, modeling for motion control, and algorithms for decision-making using reinforcement learning (QMIX). A simulation platform was implemented to assess the performance of the system regarding navigation precision, object manipulation, and joint task accomplishment. The experiment outcomes prove the implemented MARL control approach could greatly enhance the efffficacy and flexibility of the warehouse operation processes, thus setting the foundation for fully autonomous logistics solutions in the future.