STEAM: A Training-Free Congestion-Aware Enhancement Framework for Decentralized Multi-Agent Path Finding
Mingyang Feng, Mengnuo Zhang, Shaoyuan Li, Xiang Yin · 2026
Read STEAM: A Training-Free Congestion-Aware Enhancement Framework for Decentralized Multi-Agent Path Finding by Mingyang Feng, Mengnuo Zhang, Shaoyuan Li, Xiang Yin (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-20929.txt.
Description
Mingyang Feng and three colleagues propose STEAM, a training-free congestion-aware boost for decentralized multi-agent path finding. Given a pretrained local policy, STEAM needs no retraining, design change, or replacement by a centralized planner. It injects lightweight crowding-aware guidance into the original policy execution. STEAM first rolls out shortest paths induced by current cost-to-go maps to spot future crowding hotspots. Spatially avoidable crowding is eased by tuning agent-specific cost-to-go information, while unavoidable bottlenecks get temporal logit correction.
Catalog record
| CREATOR | Mingyang Feng, Mengnuo Zhang, Shaoyuan Li, Xiang Yin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40511 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.20929