Cooperative Behavior of Artificial Neural Agents Based on Evolutionary Architectures

Abstract

Artificial agents modeled by evolutionary neural networks have been diffusely described in the specific case of static architectures and synaptic weights coded in genetic strings. At present, more attractive theories devoted to a general theory of mind consider the biological and structural levels as necessary elements for an appropriate natural information processing. In this paper, an evolutionary approach has been taken into account for the selection of neural architectures of agents embedded in an artificial environment. Several correspondences between natural and artificial neural behavior has been detected (perception, multimodal integration, memory). Moreover, a cooperative social behavior emerged among the agents for a suitable exploration of the environment and the exploitation of the resources.