In this paper, we intend to present a theoretical model for the control of the evolution of artificial agents, in multi-robots simulations. We'll study this model from an Artificial Life point of view, we think this research field is one of the most promising to implement adaptive robots in multi-agents systems with evolutionary environment.
To set up this model, we'll define some steps of " intelligence " found as much in an individual level as in a collective level. With these steps, we'll set up a classification of some formal classes of agents. We'll find in this classification great topics in multi-agents systems like reactive or cognitive agents, but it'll allow to control the evolution of an agent, a sub-group of agents, or an entire society of agents.
We'll conclude by explaining an example of application including this classification, using Artificial Life processes, with robots evolving by classical operators, and by a new operator : the macro-mutation operator, a kind of mutation altering the size of artificial chromosoms.
Evolution's Control, Artificial Life, Multi-Agents Systems.