A Guide Path Network Design Method for Automated Guided Vehicle Systems Using Q-Learning

Abstract

In this paper, a guide path design method is suggested for automated guided vehicle systems using Q-learning technique. Numerous manufacturing companies have recognized various advantages of Automated Guided Vehicle System (AGVS) for material handling. These advantages include the flexibility in transportation, the improved space utilization, and the lead-time reduction. With the rapid advance of the state of art technology for AGVS, the application of AGVS to automated manufacturing systems has been more popular than ever before. However, the design of the guide-path network has been considered as one of difficulties in the application of AGVS. By applying the Q-learning technique, it is possible to consider the traffic congestion at intersections or at pickup/delivery stations, and interference among vehicles on bi-directional path segments. It is discussed how the Q-learning technique can be applied to the guide path design problem. A numerical experiment was performed to evaluate the performance of the rules obtained from the learning process for the network design. The result of this research is compared with those by previous studies.