An Artiicial Life Model for Investigating the Evolution of Modularity

To investigate the issue of how modularity emerges in nature, we present an Arti cial Life model that allow us to reproduce on the computer both the organisms (i.e., robots that have a genotype, a nervous system, and sensory and motor organs) and the environment in which organisms live, behave and reproduce. In our simulations neural networks are evolutionarily trained to control a mobile robot designed to keep an arena clear by picking up trash objects and releasing them outside the arena. During the evolutionary process modular neural networks, which control the robot's behavior, emerge as a result of genetic duplications. Preliminary simulation results show that duplication-based modular architecture outperforms the nonmodular architecture, which represents the starting architecture in our simulations. Moreover, an interaction between mutation and duplication rate emerges from our results. Our future goal is to use this model in order to explore the relationship between the evolutionary emergence of modularity and the phenomenon of gene duplication.