Design of Multi-Intelligent Body AGV Path Planning Algorithm under Cooperative Stochastic Game Based on Reinforcement Learning Improvement

The advent of Automated Guided Vehicles (AGVs) in industrial automation and intelligent logistics has underscored the importance of efficient path planning to augment production efficacy and reduce operational costs. Traditional path planning techniques, while effective in static or less complex environments, often fail to address the dynamic and uncertain conditions present in facilities utilizing multiple AGVs. This paper investigates the application of Multi-Agent Reinforcement Learning (MARL) as a novel approach to overcome these challenges, offering a promising avenue for optimizing multi-AGV path planning. By treating AGVs as agents that learn optimal strategies through environmental interactions, MARL facilitates a nuanced consideration of both individual actions and the interplay between multiple agents. This study synthesizes recent advancements in MARL applications for AGV path planning, highlighting significant contributions, such as the development of enhanced algorithms for various operational scenarios and the introduction of the RFM-AGV algorithm. The latter focuses on crucial aspects of MARL design—convergence, learning speed, and scalability—demonstrating its effectiveness in cooperative tasks through theoretical analysis and comparative experiments. This work not only showcases the potential of MARL in revolutionizing AGV path planning but also sets the stage for future research directions in the logistics and manufacturing sectors.