Constructing Communication Networks with Evolved Digital Organisms

Abstract

This paper describes a study where small computer programs called digital organisms are evolved to construct communication networks in a virtual environment. Through the application of various selection pressures, we show how these organisms first evolve to build, and subsequently use, these networks for a collective communication task, specifically the efficient distribution of data among organisms. The results of our experiments demonstrate that evolutionary algorithms can discover behaviors for constructing networks that solve communication tasks, potentially leading to a new family of distributed algorithm where the underlying logical structure of the network automatically self-adapts to suit the application.