The Mia Hand is a smart device that can sense the environment and adjust its grip accordingly. It can perform five different types of grasps, such as pinch, power, precision, lateral and extension. Imagine losing your hand in an accident and having to live with a prosthetic limb that is clumsy, uncomfortable and limited in functionality. That is the reality for millions of people around the world who suffer from amputation or congenital limb deficiency. But what if there was a way to restore the natural sensation and movement of your hand using a bionic device that is connected to your nervous system and adapts to your needs?

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.