Research map / Reinforcement learning / Visual and representation RL
Sim2Real Viewpoint Invariant Visual Servoing by Recurrent Control
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 reactive feedforward control with recurrent control (ablation) | visual servoing controller · Three objects, novel VP, unseen T | average distance to target (m): 0.1130 -> 0.0802 | Helped |
| add one iteration of on-policy data collection (ablation) | recurrent controller with MC value prediction · Three objects, novel VP, unseen T | average distance to target (m): 0.0802 -> 0.0685 | Helped |
| add one iteration of on-policy data collection (ablation) | recurrent controller · simulation test settings | average distance to target (m): 0.0805 -> 0.0844; 0.0576 -> 0.0346; 0.0729 -> 0.0747; 0.0437 -> 0.0235 | Helped |
| add Monte-Carlo value prediction (ablation) | recurrent controller with on-policy data · Three objects, novel VP, unseen T | average distance to target (m): 0.0844 -> 0.0685 | Helped |
| add Monte-Carlo value prediction (ablation) | recurrent controller · simulation test settings | average distance to target (m): 0.0805 -> 0.0802; 0.0773 -> 0.0769; 0.0576 -> 0.0546; 0.0729 -> 0.0730; 0.0819 -> 0.0757; 0.0437 -> 0.0461 | Mixed |
| increase on-policy retraining to two iterations (ablation) | recurrent controller · novel viewpoints versus seen viewpoints, unseen textures | average distance to target (m): 0.0685 -> 0.0844; 0.0307 -> 0.0265 | Mixed |
| adapt only visual layers using labeled real-world images (ablation) | simulation-trained controller · real-world one-object reaching | success rate (%): 88.9 -> 94.4 | Helped |
| adapt only visual layers using labeled real-world images (ablation) | simulation-trained controller · real-world two-object reaching | success rate (%): 54.1 -> 70.8 | Helped |
Closest papers by meaning
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.88)
- OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation (NeurIPS 2025 · similarity 0.87)
- Robotic Visual Instruction (CVPR 2025 · similarity 0.87)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.87)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 2025 · similarity 0.86)
- VICtoR: Learning Hierarchical Vision-Instruction Correlation Rewards for Long-horizon Manipulation (ICLR 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.