Designing an Affective Cognitive Architecture for Human-Humanoid Interaction

Abstract

Robots involved in HRI should be able to adapt to their partners by learning to select autonomously the behaviors that maximize the pleasantness of the interaction for them. To this aim, affect could play two important roles: serve as perceptual input to infer the emotional status and reactions of the human partner; and act as internal motivation system for the robot, supporting reasoning and action selection. In this perspective, we propose to develop an affect-based architecture for the humanoid robot iCub with the purpose of fully autonomous personalized HRI. This base framework can be generalized to fit many different contexts -social, educational, collaborative and assistive - allowing for natural, long-term, and adaptive interaction.