Robot behavior evolution using knowledge array network

Abstract

In this article proposed is a method of robot behavior evolution using a knowledge array network. The knowledge array network is formed by organically connecting knowledge arrays which are used in the Intelligent Composite Motion Control (ICMC) to store and utilize the empirical knowledge and adapt to inexperienced situations. The ICMC is a methodology to build up robot systems in which robots can realize complex and dexterous behaviors autonomously and adaptively based upon parameter optimization and utilization of empirical knowledge only if the motion control for fundamental element motions is given. After reviewing the ICMC and the array operations, presented is the framework of behavior evolution. It is shown how the behavior evolution from primitive motions to complicated tasks is realized, that is, it is shown how the dexterous ball throwing with robotic manipulator, for example, is coordinated from simple joint rotations.