Research map / Learning theory and algorithms / Combinatorial and search optimization
TransPath: Learning Heuristics for Grid-Based Pathfinding via Transformers
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 a learned correction factor in WA* | WA*+CF · TMP test | Optimal Found Ratio (%): 79.11 -> 85.40 | Helped |
| use a learned path probability map as Focal Search's secondary heuristic | FS+PPM · TMP test | Expansions Ratio (%): 44.43 ±25.92 -> 26.36 ±21.08 | Helped |
| use a learned path probability map with greedy best-first search | GBFS+PPM · TMP test | Expansions Ratio (%): 100 -> 23.60 ±18.34 | Helped |
| add a Transformer block to the convolutional network (ablation) | FS+PPM · Hard TMP tasks (hardness > 1.5) | Optimal Found Ratio (%): 61.74 -> 85.22 | Helped |
| add a Transformer block to the convolutional network (ablation) | FS+PPM · Hard TMP tasks (hardness > 1.5) | MSE × 10⁻³: 5.3 -> 3.2 | Helped |
| use direct ground-truth correction-factor supervision instead of differentiable A* search-error training (ablation) | WA*+CF · TMP test | Optimal Found Ratio (%): 29.15 -> 85.40 | Helped |
| use direct ground-truth correction-factor supervision instead of differentiable A* search-error training (ablation) | WA*+CF · TMP test | Average Expansions Ratio (%): 48.47 ± 28.50 -> 36.98 ± 21.18 | Helped |
| use Focal Search with path probability maps rather than WA* with correction factors | bounded-suboptimal planners · TMP test | Cost Ratio (%): 100.25 ±1.13 -> 100.24 ±0.74 | Helped |
Closest papers by meaning
- Grid Pathfinding on the 2k Neighborhoods (AAAI 2017 · similarity 0.89)
- Learning Domain-Independent Heuristics for Grounded and Lifted Planning (AAAI 2024 · similarity 0.86)
- Breadth-First Exploration on Adaptive Grid for Reinforcement Learning (ICML 2024 · similarity 0.85)
- Learn to Follow: Decentralized Lifelong Multi-Agent Pathfinding via Planning and Learning (AAAI 2024 · similarity 0.85)
- Large-Scale Multi-Robot Coverage Path Planning via Local Search (AAAI 2024 · similarity 0.84)
- MAPF-GPT: Imitation Learning for Multi-Agent Pathfinding at Scale (AAAI 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.