Self-supervised 6-DoF Robot Grasping by Demonstration via Augmented Reality Teleoperation System
This addresses the need for efficient robot grasping in restricted areas without manual annotations, though it is incremental as it builds on existing teleoperation and self-supervised methods.
The paper tackles the problem of laborious supervision for 6-DoF robot grasping by proposing a self-supervised framework using an AR teleoperation system to learn from human demonstrations, achieving satisfactory grasping abilities and learning to grasp unknown objects within three demonstrations.
Most existing 6-DoF robot grasping solutions depend on strong supervision on grasp pose to ensure satisfactory performance, which could be laborious and impractical when the robot works in some restricted area. To this end, we propose a self-supervised 6-DoF grasp pose detection framework via an Augmented Reality (AR) teleoperation system that can efficiently learn human demonstrations and provide 6-DoF grasp poses without grasp pose annotations. Specifically, the system collects the human demonstration from the AR environment and contrastively learns the grasping strategy from the demonstration. For the real-world experiment, the proposed system leads to satisfactory grasping abilities and learning to grasp unknown objects within three demonstrations.