Emotion Regulation with Markov Decision Process for Human-robot Interaction

Abstract

Emotion Regulation is an essential aspect since it helps people avoid harmful emotions. Shaping emotions and satisfying people are some of the key reasons to value a social robot in human-robot interaction. Thus, in this paper, a multi-weighted Markov Decision Process Emotion Regulation (MDPER) robot is created to maximize the transfer from a negative emotional arousal to positive one while minimizing the robot service spendings in cost and steps. By using an emotion regulation method external stimuli as a robot action set, the MDPER robot generates a series of its actions to help people regulate their emotions. Personality and emotion/intention degrees are weighted by the analytic hierarchy process under a specialized Markov Decision Process framework. The simulation and the experiment are implemented to prove that the MDPER system successfully achieves the goal.