Affective Robot Learner: Implementation of Artificial Emotion System Inspired by Educational Psychology

In this paper, we propose an affective robot learner system with artificial emotions. We define a generic teaching framework which has a planner for the phases of a regular teaching process. A single episode consists of interaction phase, teaching phase where the concepts are introduced to the robot by the human teacher, and testing phase where the learning performance of the robot is questioned by testing of the taught concepts. We frame the teaching process as a collaborative task between human and robot. In order to make the teaching process rewarding for the human teacher, we postulate to utilize academic emotions in our teaching framework. We base on educational psychology to generate education-related emotions in our computational emotion model. A simplified version of Control-Value Theory is implemented using fuzzy logic and presented within the framework. As an illustration of the framework, object teaching scenario is implemented on Nao robot, where a person can teach a number of daily used items to the robot.