Interactive Reinforcement Learning based Assistive Robot for the Emotional Support of Children

In this work, we challenge the Interactive Reinforcement Learning paradigm by implementing an interactive action-planning module developed with the goal of exploring the feasibility of using a robot to socially engage with children and improve their mood. Facial features of the child are captured and processed, determining their emotional reaction to a behavior performed by the robot. Then, these emotions are classified as affective states in a multi-dimensional model. Leveraging the expertise of a human trainer, the action-planning module interactively learns those actions that are the most appropriate to perform when the child subject is in a specific affective state. To validate the usefulness of the proposed methodology, we evaluated the impact of the robot on elementary school aged children. Our findings show that using this methodology, the robot is able not only to learn in real time from the human trainer through interactions, but also that performing these social actions a robot can improve the mood of children.