Multiagent team formation performed by operant learning: an animat approach

An animat approach to dynamic team formation in a group of distributed robots is studied. The goal is that robots learn to align with the others in order to form a row or a column without having communication among them, just local sensing and a reinforcement signal. The action of the robot is controlled by a biologically plausible neural network model of operant learning. The remarkable performance achieved by the proposed model allows the building of new artificial intelligence agents based on neurobiology, psychology and ethology research.