Decentralized coordination of automated guided vehicles

This paper approaches the issue of coordination of highly autonomous Automated Guided Vehicles (AGVs) working on an automated factory. These vehicles are used for goods delivery tasks between different points of the production system. The coordination is based on a decentralized architecture where each vehicle broadcasts the information about its state in the working environment, and by combining all these states in a local way, each AGV decides which action to take. The heuristic that allows the decentralized traffic control is based on a priority system, based on the current task, and a set of dangerous zones which are defined to avoid possible deadlocks, where mutual exclusion should be ensured. The process is somehow similar to that used by humans when circulating in cars: a set of rules and a set of signals/places. The interaction of many vehicles working on the same area under different collision conditions has been tested in a real industrial warehouse environment.