Using Learning Automata in Coordination Among Heterogeneous Agents in a Complex Multi-Agent Domain

This paper describes our use of Learning Automata as a reinforcement learning method incoordination among three heterogeneous teams of agents acting in RoboCup Rescue Simulationenvironment. We provide a brief introduction to Learning Automata and Cellular LearningAutomata, the reinforcement machine learning methods that we have used in lots of parts of ouragents’ development. Then we will describe the major challenges each team of agents should beconcerned about in such a complex domain and for each challenge, we propose our approachesto develop cooperative teams. Finally, some results of using Learning Automata in coordinatingthese heterogeneous teams of agents that cooperate to mitigate the disastrous damages in asimulated city are evaluated.