Project R-CASTLE: Robotic-Cognitive Adaptive System for Teaching and Learning

Abstract

Robots are already present in people's lives as receptionists, caregivers, and tutors. In human-robot interaction, social behavior is not only expected but often associated with users' confidence. Although several studies have been researching in this direction, the robot adaptation and the existing gap between the system and nonprogramming designers still need more effort to achieve success. In this article, a cognitive architecture is proposed and implemented into a humanoid robot. The aim is to offer a framework programmable for controlling the robot's resources, approaching previous knowledge, and new content in educational interactive activities. Furthermore, the system adapts the robot's behavior according to objective measures of users, attention and engagement during the activity. After the interactive sessions, these measures are provided in a graphical interface for students, skills evaluation. Functions of visual classification, speech processing, autonomous web search for new content, and attention detectors were tested and analyzed separately. This approach shows effectiveness in basic and medium condition levels from a set of sceneries for each module.