A Modular Approach to the Embodiment of Hand Motions from Human Demonstrations

Although manipulation of known objects is a well-studied field, handling deformable or small, fragile objects with II. BACKGROUND AND RELATED WORK human-level skill is a challenge. Behaviors for robotic hands A. Motion Capture of Human Hands can be generated through various approaches, e.g., planning, Capturing human hand motions as fully articulated 3D reinforcement learning, or imitation learning. We are interested hand poses is demanding due to the dexterity of hands and in leveraging intuitive human knowledge to generate high angular velocities. Nevertheless, the task is well studied data for imitation learning with a complex hand. Dataset and there are numerous solutions, including optical, nonoptical, generation is difficult in this case.