Research map / Reinforcement learning / Visual and representation RL
Towards Real-World Navigation With Deep Differentiable Planners
Check 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 |
|---|---|---|---|
| Use CALVIN instead of GPPN with reweighted loss | Differentiable planning on 2D mazes · Unseen 2D mazes, full observation | navigation success rate: 96.6\pm{4.0} -> 99.7\pm{0.5} | Helped |
| Use CALVIN instead of GPPN with reweighted loss | Differentiable planning on 2D mazes · Unseen 2D mazes, partial observation | navigation success rate: 11.25\pm{3.7} -> 92.2\pm{1.3} | Helped |
| Use CALVIN instead of GPPN with reweighted loss | Differentiable planning on 2D mazes · Unseen 2D mazes, embodied navigation | navigation success rate: 28.5\pm{3.5} -> 93.7\pm{6.2} | Helped |
| Reweight trajectory samples | CALVIN · Unseen 2D mazes, partial observation | navigation success rate: 48.0\pm{5.2} -> 92.2\pm{1.3} | Helped |
| Remove action availability loss L_A (ablation) | CALVIN · Partially observable 2D mazes | navigation success rate: 92.2 -> 84.1 | Hurt |
| Remove transition model loss L_P (ablation) | CALVIN · Partially observable 2D mazes | navigation success rate: 92.2 -> 8.3 | Hurt |
| Replace LPN with a CNN backbone (ablation) | CALVIN · MiniWorld unseen 3D mazes, 3x3 | navigation success rate: 97.7 -> 75.2 | Hurt |
| Replace LPN with a CNN backbone (ablation) | CALVIN · MiniWorld unseen 3D mazes, 8x8 | navigation success rate: 69.2 -> 8.6 | Hurt |
| Replace LPN with a CNN backbone (ablation) | VIN · MiniWorld unseen 3D mazes, 8x8 | navigation success rate: 41.2 -> 0.6 | Hurt |
| Replace LPN with a CNN backbone (ablation) | GPPN · MiniWorld unseen 3D mazes, 8x8 | navigation success rate: 33.3 -> 18.3 | Hurt |
Closest papers by meaning
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.88)
- Versatile Navigation Under Partial Observability via Value-guided Diffusion Policy (CVPR 2024 · similarity 0.88)
- Scaling Value Iteration Networks to 5000 Layers for Extreme Long-Term Planning (ICML 2025 · similarity 0.88)
- Affordances-Oriented Planning Using Foundation Models for Continuous Vision-Language Navigation (AAAI 2025 · similarity 0.88)
- Seeing through Uncertainty: Robust Task-Oriented Optimization in Visual Navigation (NeurIPS 2025 · similarity 0.88)
- D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation (CVPR 2026 · 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.