Abstract:
Many HRI researchers try to make guidelines for an appearance of the robot. However, these guidelines are insufficient to design and to evaluate a robot's appearance separately. We propose new evaluation framework called "variable body image" to solve this problem. This framework changes the appearance of a robot with divided human-like robotic parts and makes it possible to evaluate the appearance during interaction. We designed and implemented hardware and software of movable human-like parts to achieve our framework. Using our framework, we evaluated the types of robot appearance that is effective in human-robot interaction by comparing interactions between a participant and a cart equipped with different human-like parts. We used user preference ratings for the SRT (system response time) to evaluate the effectiveness of the appearances. Our result suggests that a user treats a robot as an anthropomorphic agent even if there are no anthropomorphic parts. The result also suggests that the eyes are more effective in an interaction as compared to the mouth.