Multi-person human-robot interaction system for android robot

Abstract

Communication with multiple people is more common than one-to-one communication. We therefore developed the system for multi-person communication. By extending our already developed real-time gesture planning method, we propose gesture adjustment suitable for human's demand through parameterization and gaze motion planning which can communicate with multiple people and adjust a gesture to the location of talker and/or object. We implemented the proposed motion planning method on the android robot Actorid-SIT. The components of the system (i.e. input/output processes and selection of interaction rule) are connected each other via Key-Value Store, which is an internet technology and has parallelism and scalability. We conducted multi-person HRI experiments for over 500 subjects in total. In our HRI system, the induction rate of communication was over 60% thanks to parameterization. In addition, residence time of communication was also longer thanks to interruptibility.