Multi-Goal Robot Path Planning Based on Q-Learning for Library Logistics

Abstract

This paper presents a solution for dealing with the organization of books during periods of high demand, such as assessment weeks, at the library of the Federal University of Viçosa. During these periods, library staff face an increased workload due to borrowing, returning, collecting and storing books. In this scenario, we propose a solution based on Q-Learning to visit strategic points in the library in order to improve staff performance. To validate the proposed approach, a comparison was made between the proposed approach in simulation and a greedy method based on Dijkstra’s algorithm. The results showed that the proposed approach outperformed the greedy-Dijkstra algorithm in terms of planning time and number of turns, with a 20% reduction in the number of turns and a planning time at least twice as fast. The success rate was 100% for completing the distribution task, demonstrating the system’s applicability in the proposed scenario.