From Humanoid Embodiment to Theory of Mind

We propose to investigate the foundations of communication and symbolic behavior by means or a robotics approach, i.e. by studying how these behaviors might emerge from the physical dynamics of an agent and its sensory-motor interactions with the real world. In this perspective, the human-robot interface problem can be viewed as one of coupling the interaction dynamics of all agents. Through a number of case studies we will show that within this interaction dynamics there is sparse global structure, i.e. a structure that can be characterized by only a small number of points in phase space, and that it is best to interact with the agent, i.e. interfere with its dynamics, at these points. We introduce a humanoid robot with the capability for dynamic full-body movement. The preliminary results of two experiments, sitting and standing up, are presented. Lastly, experiments with self exploratory learning of embodiment and visual motor learning of neonatal imitation abilities are introduced.