Research map / Reinforcement learning / Visual and representation RL
Grasp-Oriented Non-Prehensile Manipulation via Learning a Graspability Field
Licheng Zhong, Gim Hee Lee
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 target-pose reaching with graspability-field objective (GOMP teacher vs retrained baseline) | non-prehensile manipulation policy · Simulation, seen objects | success rate: 36.1% -> 84.0% | Improved |
| replace target-pose reaching with graspability-field objective (GOMP teacher vs retrained baseline) | non-prehensile manipulation policy · Simulation, unseen objects | success rate: 28.8% -> 77.9% | Improved |
| use graspability objective instead of nearest-reference oracle targets | GOMP teacher · Simulation, unseen objects | success rate: 50.2% -> 77.9% | Improved |
| distill recurrent student on partial point clouds from teacher | GOMP · Simulation, seen objects | success rate: 84.0% -> 75.5% | Hurt |
| add graspability-guided manipulation before grasping | real robot grasp · Real-world challenging initial configurations | total success: 8% -> 62%(31/50) | Improved |
| use 10 reference poses instead of 1 (ablation) | GOMP · Simulation, unseen objects | success rate: 72.8% -> 77.9% | Improved |
| use 10 reference poses instead of 20 (ablation) | GOMP · Simulation, unseen objects | success rate: 68.1% -> 77.9% | Improved |
| use 10 reference poses instead of 5 (ablation) | GOMP · Simulation, seen objects | success rate: 76.2% -> 84.0% | Improved |
| use soft aggregation instead of hard minimum (ablation) | graspability distance · Simulation, seen objects | success rate: 81.1% -> 84.0% | Improved |
| use soft aggregation instead of hard minimum (ablation) | graspability distance · Simulation, unseen objects | success rate: 74.3% -> 77.9% | Improved |
| add temporal anchor term (ablation) | graspability distance · Simulation, seen objects | success rate: 78.1% -> 84.0% | Improved |
| add temporal anchor term (ablation) | graspability distance · Simulation, unseen objects | success rate: 74.1% -> 77.9% | Improved |
Closest papers by meaning
- UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping (CVPR 2025 · similarity 0.88)
- Generalizing 6-DoF Grasp Detection via Domain Prior Knowledge (CVPR 2024 · similarity 0.88)
- Grasp-Then-Plan with Failure Attribution: A Closed Two-Stage Framework for Precise and Generalizable Robotic Manipulation (NeurIPS 2026 · similarity 0.88)
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Contact Map Transfer with Conditional Diffusion Model for Generalizable Dexterous Grasp Generation (NeurIPS 2025 · similarity 0.87)
- Grasp2Grasp: Vision-Based Dexterous Grasp Translation via Schrödinger Bridges (NeurIPS 2025 · similarity 0.87)
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.
