Evolution of communication in a genetic based multi-agent system: use wise resources

We studied how communication evolves in a genetic based multi-agent system using an adapted Pittsburgh style classifier system. This work is an extension of a minimal model of communication which consisted in making two agents communicate through a medium of communication and play naming game with a limited number of situations to recognize. We propose to make that model more complex by increasing both the number of agents within the multi-agent system and the number of words that agents can use. In this paper, we study how confusion may emerge through communication and how agents use their available "cognitive" resources in order to learn to communicate. We point out that further study with new measures is needed in order to better understand how communication evolves in a genetic based multi-agent system.