Implementation Approach of Affective Interaction for Caregiver Support Robot

Abstract

This paper describes a series of our studies for developing functions for robots to better interact with humans, especially in the welfare field. The caregiver support robot is proposed to help caregivers in the welfare field and functions related to realizing affective behavior were studied. We believe such robotmust understand human emotion state, have own virtual emotion state and be able to express emotion in order to behave affectively. The Virtual Kansei (VK) was proposed to answer this set of requirements and various elements of VK were developed. The VK consists of three parts; the Kansei detector, the Kansei generator and the Kansei expressive regulator. The Kansei detector detects human partner’s emotion state using facial images, voice sounds and body movements. The Kansei generator generates human-like virtual emotion for robots. We devised a mimicking approach in developing the generator where emotion distances are defined and are used in learning and evaluating the generator. The Kansei expressive regulator makes the robot behave emotionally in executing everyday tasks. It modulates the basic robot motion according to the generated virtual emotion. This paper focuses on the concept and the relationship of these elements.