Humanoid integrated UI system for supervised autonomy with massive data visualization over narrow and unreliable network communication for DRC competition

摘要

This paper presents a high level system architecture of the work done by Team NEDO-JSK for DRC (DARPA Robotics Challenge), which was held on 5th-6th, June, 2015, California, USA. DRC was a competition which included driving vehicle, manipulating objects and locomotion over rough terrain. Communication between robot and operators were limited in a similar way to teleoperating situation. The system introduced in this paper was strongly motivated by DRC competition, however system architecture is useful for generic teleoperating tasks. User interface and sensor visualization is the central element of our teleoperating system and operators' performance greatly depends on them. The goal of our system is to provide system to operate robot with less effort of operators and to be able to handle unexpected situation. The key features of the system are: 1) supervised system architecture based on simple command from operators. 2) low-level sensor visualization and high-level teleoperating user interface capable to deal with emergencies. 3) system integration over unreliable and narrow network. This paper summarizes our technical approach and implementation focusing on communication over narrow and unreliable network and user interface for operators. As evaluation of our system, this paper summarizes actual performance through DRC competition.