Acquisition of communication protocol for autonomous multi-AGVs driving

Abstract

In order to realize autonomous AGV driving, two problems arise: the autonomous vehicle navigation problem; and the acquisition of communication protocol. The communication protocol means that an AGV can understand other AGV intention by its behavior. This paper proposes an acquisition method for communication protocol using Q-learning. Multi-AGVs are defined by a set of agents and each agent is defined by a learning automaton with Q-learning function. Numerical experiments are executed to examine what kind of communication protocol is acquired by Q-learning by the multi-AGVs operating together. By comparing the acquired communication protocol with other options, the AGVs that adopt the correct communication protocol process the given workpieces slightly faster. This result shows that an AGV can acquire a suitable communication protocol by learning in any environment.