Towards a Novel Interactive Reinforcement Learning Framework for Socially Assistive Robotics

Currently, several industries and the public sector in Germany are facing the problem of the shortage of skilled human resources. Thus, it is needed to support different professions to increase the efficiency of the provided services. Various studies demonstrate that robots are able to support different user groups, e.g. elderly individuals or students, thereby easing the workload of skilled workers, e.g. caregivers or teachers. In this work, we choose to support the education sector to meet the shortage of teachers. We develop an adaptive robot assisting international students in learning the German language and investigate its suitability from the student's point of view. In our approach, the robot learns how to adapt nonverbal responses for the student based on two information sources: (i) implicit student signals (e.g. engagement, emotions, tiredness), and (ii) teacher's explicit feedback (e.g. given through a tablet).