Towards a unified control framework for humanoid robots and their virtual avatars in physical and virtual reality-based interactions

Abstract

We introduce a unified modular control architecture and virtual-reality evaluation framework to extend the use of socially capable humanoid robots. Our methodology is based on a unique gesture and facial expression representation module, called cascaded Temporal Disc Controllers (TDCs), that represents all emotional expressions, gestures and time-variant actions in a normalized mathematical space designed to accept high level control commands, while being independent of actual physical robot implementations. At the same time it also provides an underlying mechanism for avoiding repetitive behaviors and increases the “humanness” by minor random perturbations during the interaction process. Our implementation combines two available physical robots (Zeno, FACE) and their virtual representations with active perception in order to drive low- and high level reactive behaviors in support of interactive educational and pedagogical goals. Our photo-realistic representations were used in comparative evaluation studies and a low-cost Augmented-Reality interface was also developed to support seamless interaction in virtual space.