Development of interface for teleoperation of humanoid robot using task model method

Abstract

Teleoperation systems for humanoid robots that replace human tasks in emergency situations have been widely studied. Humanoid robots have the capability of utilizing the equipment designed for human manipulation. When the humanoid robot works on behalf of human beings, rapid deployment can be achieved when a disaster occurs. However, since the humanoid robot and the human have different physical abilities, it is difficult to control the robot by mapping the human motion directly to the robot. Recently, the interfaces that map the human motion, taken by the vision sensors, to the humanoid robot have become popular. Though the vision-based teleoperation system can be rapidly constructed anywhere, the interfaces require some sort of feedback to smoothly control the robot. The tactile and force feedback system has limitations on human motion because of mechanical structures. The teleoperation system also has the potential issue of time delay. In this paper, we propose an interface for teleoperation of the humanoid robot with vision sensors. The inner states of the robot and the recognized human motions are visualized for the feedback to the operator. We also introduce the task model to avoid the time delay. In the visual feedback, the task states are visualized to assist the smooth operations.