Human hand motion retargeting for dexterous robotic hand

One way to achieve dexterous manipulation autonomously with natural input is through learning by demonstration. Unfortunately, grasping an object with a complex dexterous hand is a complicated task. To facilitate the demo acquisition process, we propose a low-cost framework to map the human hand motion from a single RGB-D camera using inverse kinematics. This framework has been implemented in a CoppeliaSim simulation environment. We evaluate two multi-task handling methods and a low-pass filter using two obtained trajectories. Empirically, the proposed framework can successfully perform grasping task imitations. An exemplary video of the object manipulation is presented on the project website: https://sites.google.com/view/retargeting-for-dexterous-hand