Evolution in ecological agent systems

Abstract

In this paper, we describe an ecologically inspired technique for the development of interaction centred agent systems which facilitates the emergence of organisational patterns that are often found in ecological networks between mutualistic species. We then extend this approach with the introduction of a genetic algorithm that explores the space of parameters these agents employ for making decisions about the interactions they are willing to embark upon and argue that this approximation lays the ground for coevolutionary change between agents in the system, enhancing in this way their adaptive ability and the persistence of the society of agents as a whole.