Research map / Reinforcement learning / Visual and representation RL
Parameterized Quasi-Physical Simulators for Dexterous Manipulations Transfer
Xueyi Liu, Kangbo Lyu, jieqiong zhang, Tao Du, Li Yi
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 |
|---|---|---|---|
| add the physics curriculum to DGrasp-Tracking | DGrasp-Tracking · Bullet | success rate (%): 0/15.69/15.69 -> 7.84/23.53/37.25 | Improved |
| add the physics curriculum to DGrasp-Tracking | DGrasp-Tracking · Isaac Gym | success rate (%): 0/0/7.84 -> 0/23.53/31.37 | Improved |
| add the curriculum of simulators instead of optimizing directly in the stiffest model (ablation) | QuasiSim · Bullet | success rate (%): 0/17.64/29.41 -> 27.45/37.25/58.82 | Improved |
| use the full curriculum rather than a version with some stages moved out (ablation) | QuasiSim · Bullet | success rate (%): 11.76/27.45/50.98 -> 27.45/37.25/58.82 | Improved |
| add the analytical simulator base instead of pure neural approximation (ablation) | QuasiSim · Bullet | success rate (%): 0/13.73/25.49 -> 27.45/37.25/58.82 | Improved |
| add residual physics networks (ablation) | QuasiSim · Bullet | success rate (%): 5.88/23.53/41.18 -> 27.45/37.25/58.82 | Improved |
| add the local residual force network (ablation) | QuasiSim · Bullet | success rate (%): 0/19.61/35.29 -> 27.45/37.25/58.82 | Improved |
| add MPC closed-loop control on the optimized simulator versus open-loop trajectory | QuasiSim · qualitative example (mouse) | tracking success | Improved |
| use QuasiSim instead of best baselines for Shadow hand tracking | Shadow hand tracking · Bullet and Isaac Gym | success rate (%): 27.45/37.25/58.82 vs 7.84/23.53/37.25 (Bullet) | Improved |
| use QuasiSim instead of MPC (w/ base sim.) on hand tracking error | Shadow hand tracking · Isaac Gym | MPJPE (mm): 23.19 -> 25.33 | Hurt |
Closest papers by meaning
- Position: Scaling Simulation is Neither Necessary Nor Sufficient for In-the-Wild Robot Manipulation (ICML 2024 · similarity 0.88)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.87)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- DexHandDiff: Interaction-aware Diffusion Planning for Adaptive Dexterous Manipulation (CVPR 2025 · similarity 0.86)
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.86)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 2025 · similarity 0.86)
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.
