A situation-aware action selection based on individual's preference using emotion estimation

Abstract

Long-term relationships between humans and robots will expand in the future. We therefore consider that an affinity to robots is one of the important components of a stable relationship. In this paper, we propose an action selection system for robots. We expect the system to select an appropriate action that matches human's individual preference. We used an emotion estimation system to achieve this. This is an approach to create an affinity to robots. As a human's preference will be affected by their emotions, we used emotion estimation to learn these preferences. In this paper, we outline the proposed system, experiments, and evaluation of its efficacy. We performed two types of experiments using a simulated robot over one week and four weeks. In each of the experiments, the robot was varying its behaviors, based on learning of individual's preferences.