Methodology for a Gradual Migration from a Centralized towards a Decentralized Control in AGV Systems

In this paper, a methodology is presented which aims at transforming an existing industrial Automated Guided Vehicle (AGV) control architecture from a centralized to a more decentralized architecture in a gradual way. Modern AGV control systems operate mainly in a centralized way. However, literature shows that centralized systems are not suitable to control a large AGV fleet, especially in future flexible manufacturing systems. Research towards a decentralized control of these systems is already very elaborated and industry shows a clear interest in this type of control for AGV systems. However, in literature, the proposed decentralized architectures are rather radical, causing the implementation of them being a real boundary for industrial manufacturers. Still today, the gap between academia and industry is very large which is the reason why industrial AGV systems still work in a centralized fashion, despite the abundance on decentralized algorithms in literature. To overcome this issue, this paper proposes a methodology which facilitates the implementation of decentralized architectures for AGV control in practice. By separating the core AGV control tasks and modularize them by converting them to intelligent Task Agents, a gradual migration between the two architectures will be feasible in practice. This work can be beneficial for industrial R&D centers which are conducting research towards AGV control as well as for AGV manufacturers and consumers who want to decentralize their AGV fleet in a feasible and manageable way.