Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Bounded Suboptimal Search with Learned Heuristics for Multi-Agent Systems
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 |
|---|---|---|---|
| learn one joint multi-robot policy (MPMW) instead of per-robot single policy (SPMW) (ablation) | learned policy · AVP validation, 2 robots | action accuracy: 0.25 -> 0.64 | Helped |
| train single-robot policy on two-robot worlds (SPMW) rather than one-robot worlds (SPSW) (ablation) | single-robot policy · AVP validation, 2 robots | action accuracy: 0.045 -> 0.25 | Helped |
| use learned policy as focal heuristic in focal A* instead of plain A* | A*_inf,log with MPMW · AVP 2 robots, 100 instances | success rate: 0.9 -> all of the 100 | Helped |
| use learned heuristic inside focal search rather than directly as A* heuristic | MPMW · AVP 2 robots, 100 instances | success rate: 0.8 -> all of the 100 | Helped |
| continue search with anytime focal A* for 5 min | A*_inf,log · AVP 2 robots | suboptimality factor: 3.1 -> 2.0 | Helped |
| use log-likelihood focal heuristic instead of value-function estimate (ablation) | focal A* · AVP 2 robots | performance | Helped |
Closest papers by meaning
- Optimize Planning Heuristics to Rank, not to Estimate Cost-to-Goal (NeurIPS 2023 · similarity 0.89)
- Imitating Graph-Based Planning with Goal-Conditioned Policies (ICLR 2023 · similarity 0.87)
- Active Fine-Tuning of Multi-Task Policies (ICML 2025 · similarity 0.86)
- Going Beyond Heuristics by Imposing Policy Improvement as a Constraint (NeurIPS 2024 · similarity 0.86)
- Learning Domain-Independent Heuristics for Grounded and Lifted Planning (AAAI 2024 · similarity 0.86)
- Learn to Follow: Decentralized Lifelong Multi-Agent Pathfinding via Planning and Learning (AAAI 2024 · 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.