Smart Library: A Multi-Agent Path Planning Logistics Operation using Reinforcement Learning

Robotics navigation holds significant importance in the transportation, manufacturing, and exploration sectors. The prioritization of path planning is essential for efficient navigation in environments where multiple robots coexist. In the realm of logistic libraries, characterized by repetitive tasks with a need to mitigate human errors, this study concentrates on achieving feasible and optimal path planning for multiple agents tasked with transporting books within a two-floor building. The proposed algorithm, based on Q-learning, incorporates transfer and curriculum learning, enabling cooperative and decentralized behavior among the agents. Numerical experiments conducted through a software-in-the-loop approach validate the effectiveness of the method. The results indicate a success rate of 94% with 9 agents, accompanied by a 73.36% reduction in task completion steps compared to a scenario with a single agent.This research serves to showcase the algorithm's capability in enhancing navigation and task efficiency in multi-agent logistics settings, particularly within smart library environments.