Research on Attractor Design for the Dynamics-based Information Processing System

—On-line Design for Plastic Propety and Hierarchical Design for Large-scale Systems—

Abstract

Because robot motions are generated through the interaction between the robot body and its environments, it is necessary for the robot to modify the motion pattern that is prepared a prior, depending on the situation. We have proposed dynamics-based information processing system in which a dynamical system memorizes and replays the robot whole body motion using the entrainment phenomenon of the nonlinear dynamics. In this paper, we propose the motion pattern modifying method based on the re-design of the dynamics using the on-line least square method, and the hierarchical design method of the dynamical system for the large-scale system to generate the context dependent motion sequence.