Evolutionary Search for Cellular Automata with Self-Organizing Properties toward Controlling Decentralized Pervasive Systems and Its Applications

Abstract

Cellular Automata (CAs) have been investigated extensively as abstract models of the decentralized systems composed of autonomous entities characterized by local interactions. However, it is poorly understood how CAs can interact with their external environment, which would be useful for implementing pervasive systems that consist of billions of components (nodes, sensors, etc.). This paper focuses on the emergent properties of CAs induced by external perturbations toward controlling pervasive systems. The authors assumed a minimum task in which a CA has to change its global state drastically after every occurrence of a perturbation period. By conducting evolutionary searches for rules of CAs, they obtained interesting behaviors of CAs in which their global state cyclically transited among different stable states in either ascending or descending order. They analyze the emergent behavior in detail and also introduce applications of the evolved CA for controlling pervasive robots and an interactive art.