| This thesis addresses the problem of emergence of intelligence in a multi-robotic environment. Each robot is considered to be very simple in terms of hardware and software capabilities and, consequently, the kind of intelligence we are interested in is a very simple one, based only on few and simple behaviors such as the ones present in the individuals in social insect colonies. This is now known as swarm intelligence and has been applied successfully to many optimization problems. The goal is to have a large number of agents with just a few very simple rules and through their social interactions have them solve complicated tasks due to social emergent behaviors that are not directly derived from the individual behaviors of each agent. In this research we assume that the robots do not communicate directly to each other or, if they do so, it is not guaranteed that the other robot receives the message and decodes it the way intended. The main contribution of the present research to this field is the introduction of new design techniques based on Game Theory and their evaluation through experiments. Moreover, we also prove that when we represent the behaviour of robots by using personality traits, a zero-sum game between two robots will always converge to N ash equilibria. Also, we prove that in Markov games, the multiple pursuer-evader game with adaptation by learning automata converges to Nash equilibria as well. |
