Research map / Architectures and efficiency / Neural architecture search
Path Planning using Neural A* Search
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 |
|---|---|---|---|
| use differentiable A* rather than black-box A* optimization | Neural A* versus BB-A* · MP, Tiled MP, and CSM | Hmean: 42.1 (40.5, 43.6) -> 52.0 (50.7, 53.3); 31.1 (29.2, 33.0) -> 54.2 (52.6, 55.8); 35.6 (33.8, 37.4) -> 43.6 (41.7, 45.5) | Helped |
| use differentiable A* rather than black-box A* optimization | Neural A* versus BB-A* · SDD intra-scenes and inter-scenes | Chamfer distance: 152.2 (144.9, 159.1) -> 16.1 (13.2, 18.8); 134.3 (132.1, 136.4) -> 37.4 (35.8, 39.0) | Helped |
| add start and goal locations to encoder input (ablation) | Neural A* · MP | Opt: 67.0 (65.1, 68.8) -> 87.7 (86.6, 88.9) | Helped |
| add start and goal locations to encoder input (ablation) | Neural A* · MP | Hmean: 41.5 (40.2, 42.7) -> 52.0 (50.7, 53.3) | Helped |
| use VGG-16 instead of ResNet-18 backbone (ablation) | Neural A* · MP | Hmean: 49.2 (47.9, 50.5) -> 52.0 (50.7, 53.3) | Helped |
| use straight-through Gumbel-softmax instead of discretized activation (ablation) | Neural A* · MP | Opt: 0 | Hurt |
| use straight-through Gumbel-softmax instead of discretized activation (ablation) | Neural A* · MP | Exp: 0 | Hurt |
| use mean squared loss instead of mean L1 loss (ablation) | Neural A* · MP | Performance | No clear effect |
Closest papers by meaning
- Learning to Plan in High Dimensions via Neural Exploration-Exploitation Trees (ICLR 2020 · similarity 0.88)
- Generalizable Motion Planning via Operator Learning (ICLR 2025 · similarity 0.86)
- On the Road to Portability: Compressing End-to-End Motion Planner for Autonomous Driving (CVPR 2024 · similarity 0.84)
- Learning Cognitive Maps from Transformer Representations for Efficient Planning in Partially Observed Environments (ICML 2024 · similarity 0.84)
- System 1.x: Learning to Balance Fast and Slow Planning with Language Models (ICLR 2025 · similarity 0.84)
- Scaling Value Iteration Networks to 5000 Layers for Extreme Long-Term Planning (ICML 2025 · 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.