Robot Action Selection based on Relationships between Mood, Situations, and Facial Expression

The goal of our research is comforting a human by a robot action. A robot action which comforts a human should differ according to the human mood. It seems to be important to estimate and predict the human mood accurately to comfort the human by a robot action. In this paper, we performed action selection experiments with Mof-mof robot which has soft visual aspect. A subject played quiz games and the robot selected an action which was expected to comfort the subject. The action was selected based on change of mood prediction. The change of mood was predicted based on the relationship between a mood, situations and facial expressions. Situations of the games and recognized facial expressions of the subjects are expected to be keys to estimate a mood based on the relationship.