Research map / Reinforcement learning / Visual and representation RL
Visual-RRT: Finding Paths toward Visual-Goals via Differentiable Rendering
Sebin Lee, Jumin Lee, Taeyeon Kim, Youngju Na, Woobin Im, Sung-Eui Yoon
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 tree-based exploration (vRRT) on top of visual-gradient optimization | Dr.Robot · Visual-goal motion planning, Franka (avg) | success rate (%): 19.3 -> 75.2 | Improved |
| use joint planning with vRRT instead of two-stage Dr.Robot goal estimate then RRT* | visual-goal planning · Visual-goal motion planning, UR5e (avg) | success rate (%): 28.1 -> 79.8 | Improved |
| use vRRT instead of Dr.Robot + RRT* on path length | visual-goal planning · Visual-goal motion planning, Franka (avg) | path length (rad): 1.46 -> 1.61 | Hurt |
| use vRRT instead of Dr.Robot for pose reconstruction | visual-goal pose reconstruction · Franka (avg) | mean joint error (rad): 0.588 -> 0.142 | Improved |
| use vRRT instead of Dr.Robot and regressors on real images | pose reconstruction · Panda-3CAM-Azure | avg per-joint error (rad): 0.164 -> 0.083 (Dr.Robot) | Improved |
| add frontier-based parent sampling (eta above 0) instead of uniform node selection (ablation) | vRRT · UR5e motion planning | success rate (%): near-zero at eta=0.0; 50-80% at eta=0.2 | Improved |
| use exclusive frontier sampling (eta=1.0) instead of eta 0.6-0.8 (ablation) | vRRT · UR5e motion planning | success rate (%) | Hurt |
| raise exploration ratio r to 0.9 (ablation) | vRRT · UR5e motion planning | success rate (%) | Hurt |
| lower exploration ratio r to 0.1 instead of 0.3 (ablation) | vRRT · UR5e motion planning | success rate (%) | Hurt |
| lower Adam beta1 from 0.9 to 0.5 in inertial gradient expansion (ablation) | vRRT · UR5e motion planning (avg) | success rate (%): 79.8 -> 29.6 | Hurt |
| raise Adam beta1 from 0.9 to 0.99 (ablation) | vRRT · UR5e motion planning (avg) | success rate (%): 79.8 -> 65.2 | Hurt |
| vary Adam beta2 from 0.9 to 0.99 or 0.999 (ablation) | vRRT · UR5e motion planning (avg) | success rate (%): 79.8 -> 78.9 / 79.1 | No clear effect |
Closest papers by meaning
- Provably Efficient Long-Horizon Exploration in Monte Carlo Tree Search through State Occupancy Regularization (ICML 2024 · similarity 0.86)
- FLIP: Flow-Centric Generative Planning as General-Purpose Manipulation World Model (ICLR 2025 · similarity 0.86)
- Tree-Guided Diffusion Planner (NeurIPS 2025 · similarity 0.85)
- Graph Sparse Sampling: Breaking the Curse of the Horizon in Continuous MDP Planning (NeurIPS 2026 · similarity 0.85)
- Efficient Robotic Policy Learning via Latent Space Backward Planning (ICML 2025 · similarity 0.84)
- Video Language Planning (ICLR 2024 · similarity 0.84)
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.
