Decentral task allocation for industrial AGV-systems with resource constraints

Abstract

Automated Guided Vehicles (AGVs) form a large and important part of the logistics transportation systems in today's industry and are widely used, especially in Europe. Today's AGV-systems offered by global manufacturers almost all operate under some form of centralized control where a single central controller coordinates the entire fleet of AGVs. There is a trend towards decentralized control of these systems where AGVs make individual decisions that promote the flexibility, robustness and scalability of transport. However, its practical implementation seems to be in its infancy. In addition to the lack of practical implementation of decentralized control in industrial AGV-systems, we have observed a research gap in intelligent resource management of AGV-systems, which we have tried to address in previous work by proposing a more intelligent resource management approach. In this paper, we have addressed both the perceived lack of practical decentralized AGV control and the lack of intelligent resource management by proposing a decentralized task allocation algorithm based on sequential single-item auctions, taking into account resource constraints, and in which our intelligent resource management approach from previous work is introduced. We have benchmarked our new approach to a genetic algorithm-based task-allocation solver that uses "threshold-100"-charging as a resource management strategy. The result of the proposal is a decentralized task-allocation architecture under resource constraints that could be used in current AGV-systems to add more decentralized features to the fleet.