Maestro: LLM-Driven Collaborative Automation of Intent-Based 6G Networks

This letter presents M<sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">aestro</small>, a collaborative framework leveraging Large Language Models (LLMs) for automation of shared networks. M<sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">aestro</small> enables conflict resolution and collaboration among stakeholders in a shared intent-based 6G network by abstracting diverse network infrastructures into declarative intents across business, service, and network planes. LLM-based agents negotiate resources, mediated by M<sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">aestro</small> to achieve consensus that aligns multi-party business and network goals. Evaluation on a 5G Open RAN testbed reveals that integrating LLMs with optimization tools and contextual units builds autonomous agents with comparable accuracy to the state-of-the-art algorithms while being flexible to spatio-temporal business and network variability.