Concept Learning By Human Tutelage For Social Robots

Learning by human tutelage means that a human being guides the attention of a robot or agent in order to teach it a given concept. This kind of learning is very important to developing a robotic architecture for social robots. Social robots are embodied agents that are part of a heterogeneous society of robots and humans. The use of a robotic architecture may strongly reduce the time and eort required to construct a social robot. Such architecture must have structures and mechanisms to allow perception and attention, to enable a social robot to realize the richness of the human behavior and of the environment, and to learn from social interactions. In this paper, a robotic architecture inspired on Behavior Analysis is presented with two dierent learning algorithms, learning by Contingency and by Economic TG, for controlling a robotic head. For interaction between human and head robotic, it has also been developed a multimodal interface. The multimodal interface employs mechanisms of face detection, pose estimation, object saliency and speech recog- nition. The experimental results show that the robotic head, with the proposed tutelage mechanism, is able to learn concepts about objects of the environment successfully.