Vision-based Perception for Dexterous Hand-arm Teleoperation

Abstract

The dissertation devotes itself to employing markerless vision-based teleoperation using an end-to-end learning scheme for anthropomorphic hands and integrating it into a dexterous hand-arm teleoperation system. Sufficient network evaluation and robot experiments show that the proposed teleoperation systems are accurate, efficient, and robust, and can contribute to solving complex tasks in multiple applications, e.g., space station, medical surgery, and daily life.