Comfortability Detection for Adaptive Human-Robot Interactions

Recognizing emotional states from nonverbal cues is basic for any kind of social interaction. Extrapolating this capability to robots would definitely attribute them skills which might enhance their interactions with people. This thesis looks to achieve two main goals. The first one is to unravel the Comfortability concept, which we define as the persons internal agreement-acceptance to the situation that arises as a result of an interaction. The second and main goal is to build a robot-embedded system capable of recognizing this internal state, adapting its behavior accordingly. The recognition model will be developed by applying artificial intelligence techniques for temporal modeling data through visual information (body movements and facial expressions). Then, the adaptation model will take into account both the Comfortability perceived, as well as contextual information (concretely, the previous task performed) in order to decide the consecutive action that the robot will perform.