Complex behavior by means of dynamical systems for an anthropomorphic robot

We present an architecture to generate behavior for an anthropomorphic robo t. The goal is to equip the robot with the capacity to interact with a huma n. Motivated by the research on biological systems, our basic assumption is that the behavior to perform determines the external and internal structur e of the behaving system. We describe the anthropomorphic design of our rob ot and present a distributed control system that generates human-like navig ation and manipulation behavior. As the mathematical framework for this pur pose we have developed a control system which is entirely based on dynamica l systems in the form of instantiated dynamics and neural fields. We also p resent a dynamic scheme for the behavioral organization based on competitiv e dynamics. (C) 1999 Elsevier Science Ltd. All rights reserved.