SRMT: Shared Memory for Multi-agent Lifelong Pathfinding
Alsu Sagirova Yuri Kuratov Mikhail Burtsev · 2025
Read SRMT: Shared Memory for Multi-agent Lifelong Pathfinding by Alsu Sagirova Yuri Kuratov Mikhail Burtsev (2025) 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-2501-13200.txt.
Description
Alsu Sagirova, Yuri Kuratov, and Mikhail Burtsev give warehouse-style robot teams a shared memory to coordinate their routes, in a research paper from 2025. In multi-agent pathfinding, many agents must reach goals without blocking each other. Letting them talk freely is hard to scale, and central training fits poorly.
Their transformer design gives all agents one global memory workspace. Each agent posts its working memory and reads others' entries, while training and running stay decentralized. On a bottleneck task that forces negotiation, the method beat communicative baselines, above all when rewards were sparse. It also held up on larger maps with more agents, without hand-built rules.
Catalog record
| CREATOR | Alsu Sagirova Yuri Kuratov Mikhail Burtsev |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-13200