Exponential Topology-enabled Scalable Communication in Multi-agent Reinforcement Learning

Xinran Li, Xiaolu Wang, Chenjia Bai, Jun Zhang · 2025

Read Exponential Topology-enabled Scalable Communication in Multi-agent Reinforcement Learning by Xinran Li, Xiaolu Wang, Chenjia Bai, Jun Zhang (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-2502-19717.txt.

Description

Xinran Li and co-authors design ExpoComm, a communication plan that lets many learning agents agree fast. In cooperative multi-agent learning, good messaging lifts team performance. Big teams face worse partial views of the world than small ones.

Pairwise link picking grows unwieldy as team size climbs. The team instead designs the whole message topology at once. Exponential graphs spread news quickly with few total links. Their small diameter and size suit huge agent groups. Memory-based processors ground the messages agents exchange. Extra training tasks help messages carry useful content. Tests check whether the topology keeps its edge at scale. The protocol targets real fleets of robots and sensors.

Catalog record

CREATORXinran Li, Xiaolu Wang, Chenjia Bai, Jun Zhang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-19717

Browse the full library · About the Tower of Babel