Research map / Detection and segmentation / Open-set and few-shot detection
Learning Object-Centric Motion Priors from Human for Robotic Dexterous Manipulation
Zhengdong Hong, Guofeng Zhang
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 |
|---|---|---|---|
| replace predicted-object following reward with a simple reaching reward and no robot following reward | trajectory-guided RL · YCB grasping, simulation (avg) | success rate: 0.35 -> 0.84 | Improved |
| add object following reward Ro (ablation) | trajectory-guided RL · YCB grasping, 5 DexYCB objects | sample efficiency | Improved |
| add contact reward Rcontact (ablation) | trajectory-guided RL · YCB grasping, 5 DexYCB objects | training stability and success rate | Improved |
| use predicted hand-object trajectories instead of dataset reference trajectories | ours vs ViviDex · YCB grasping, non-DexYCB objects (sim) | success rate: 0.37 -> 0.86 | Improved |
| use object-centric predicted priors rather than residual learning on retargeted trajectories | ours vs ManipTrans · YCB grasping, DexYCB novel poses (sim) | success rate: 0.03 -> 0.75 | Improved |
| use predicted trajectories instead of fixed dataset trajectories for daily articulated objects | ours vs Obj-Dex · Articulated manipulation, daily objects (sim) | success rate: 0.05 -> 0.53 | Improved |
| remove following reward in articulated manipulation | PPO w/o follow. · Articulated manipulation (sim) | success rate: 0.66 -> 0.00 | Hurt |
| swap robot hand (XHand1, Inspire, Ability) with same rewards | Ours · Real robot, 3 tasks | success rate | No clear effect |
| add object-following guidance on grasp with obstacle | Ours vs PPO w/o foll. · Grasp with obstacle, DexYCB obstacle (sim) | success rate: 0.58 -> 0.57 | No clear effect |
Closest papers by meaning
- World Models for Learning Dexterous Hand-Object Interactions from Human Videos (ECCV 2026 · similarity 0.90)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- Dexterous Skill Discovery via Topology-Aware Wasserstein Dependency (NeurIPS 2026 · similarity 0.90)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.90)
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.88)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 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.
