Android as a Receptionist in a Shopping Mall Using Inverse Reinforcement Learning

摘要

For human-robot interaction (HRI), it is difficult to hand-craft all the rules for robots owing to diverse situations. Therefore, inverse reinforcement learning (IRL) is a potential solution that helps transfer human knowledge about interactions to robots. However, the feasibility of practically using IRL for HRI remains unknown. Here, we demonstrate a practical HRI application of IRL. An android was trained using IRL and acted as a receptionist to encourage visitors to practice hand hygiene in a shopping mall. We found that android learning through IRL has a competitive ability to a well-trained human operator on the reception task. Furthermore, we found that the android maintained high performance regardless of customer traffic. Our results demonstrate the potential of IRL in advancing the social HRI field. We anticipate that our work will be a starting point for using IRL in future HRI applications.