Genetic Evolution and Self-Organization of Cellular Robotic System

Abstract

A cellular robotic system (CEBOT) is a system composed of many kinds of units called "cells". A cell has a simple function and limited intelligence. In task executions, many cells are required to carry out a same task cooperatively to supplement their functions each other. If some cells have malfunctions, the total performance of the system can maintained by exchanging the malfunctioning cells. Therefore, CEBOT is expected to be adaptable to any kinds of tasks and environments. It is important to study self-organization and self-evolution of CEBOT, because both the hardware and software of CEBOT consist of many kinds of cells which are autonomous agents as mentioned above. This paper addresses self-organization of the hardware and self-evolution of the software based on the Genetic Algorithm. An outline of cells developed by the authors is reported. A self-evolutional knowledge base addressed in this paper is based on the Genetic Algorithm. The self-evolutional knowledge base is applied to a simple task planning system installed into a manipulator system and is verified. The effectiveness of the knowledge base is demonstrated by showing a learning capability of the manipulator in experiments.