Research map / Reinforcement learning / Visual and representation RL
Learning Cross-hand Policies of High-DOF Reaching and Grasping
Qijin She, Shishun Zhang, Yunfan Ye, Ruizhen Hu, Kai Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use IBS representation instead of object contact map (OCM) (ablation) | unified policy · YCB, Shadow | success rate (%): 50.1 -> 68.0 | Improved |
| use IBS instead of gripper contact map (GCM) for transfer (ablation) | unified policy · YCB, Schunk | success rate (%): 45.5 -> 54.6 | Improved |
| replace feature concatenation with transformer policy (ablation) | UNI+IBS · YCB, Schunk | success rate (%): 54.6 -> 65.3 | Improved |
| use two-stage unified policy instead of single joint-space end-to-end model (ablation) | Shadow hand policy · YCB, Shadow | success rate (%): 72.2 -> 68.0 | Hurt |
| use learned adaptation model instead of optimization-based IK with self-collision (ablation) | adaptation model · YCB, Schunk | time per frame (ms): 279.6 -> 59.4 | Improved |
| drop self-collision term in optimization-based adaptation (ablation) | OB-IK · YCB, Schunk | collision percentage: 0.0% -> 2.0% | Hurt |
| predict key point displacements instead of retargeting joint or key point motions (baselines) | cross-hand transfer · YCB + ContactPose, Schunk | success rate (%): 40.1 -> 61.8 | Improved |
| match key points instead of joints in offline motion retargeting (ablation) | MR baseline · YCB + ContactPose, Mano | success rate (%): 6.4 -> 33.8 | Improved |
| make policy transfer online with key point matching (PT-KM) versus offline (MR-KM) (ablation) | baseline · YCB + ContactPose, Schunk | success rate (%): 35.9 -> 35.6 | No clear effect |
| use partial depth-camera point clouds instead of complete point clouds (ablation) | Ours · YCB, Allegro | success rate (%): 55.0 -> 47.1 | Hurt |
Closest papers by meaning
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.90)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.89)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.88)
- Touch in the Wild: Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper (NeurIPS 2025 · similarity 0.88)
- Contact Map Transfer with Conditional Diffusion Model for Generalizable Dexterous Grasp Generation (NeurIPS 2025 · similarity 0.88)
- UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
