Learning Continuous Grasping Function with a Dexterous Hand from Human Demonstrations
This work addresses the challenge of efficient and generalizable grasping for robotic manipulation, though it appears incremental as it builds on existing methods like generative modeling and motion retargeting.
The authors tackled the problem of generating grasping motions for a dexterous hand by learning a continuous grasping function from human demonstrations, achieving significant improvement in success rate when transferred to a real Allegro Hand compared to previous planning algorithms.
We propose to learn to generate grasping motion for manipulation with a dexterous hand using implicit functions. With continuous time inputs, the model can generate a continuous and smooth grasping plan. We name the proposed model Continuous Grasping Function (CGF). CGF is learned via generative modeling with a Conditional Variational Autoencoder using 3D human demonstrations. We will first convert the large-scale human-object interaction trajectories to robot demonstrations via motion retargeting, and then use these demonstrations to train CGF. During inference, we perform sampling with CGF to generate different grasping plans in the simulator and select the successful ones to transfer to the real robot. By training on diverse human data, our CGF allows generalization to manipulate multiple objects. Compared to previous planning algorithms, CGF is more efficient and achieves significant improvement on success rate when transferred to grasping with the real Allegro Hand. Our project page is available at https://jianglongye.com/cgf .