GENOME

GENOME is a Marie Skłodowska-Curie Action (MSCA) aimed at investigating, developing/fine-tuning, and validating large language models (LLMs) for the autonomous management and orchestration of various network technological domains (O-RAN, edge, cloud) through intent-based networking and task-agnostic network functions. It also targets the establishment of a connected intelligence framework for the coordination between these functions via on-the-fly protocol learning to mitigate conflicting management decisions and maximize utilization. Network configuration will also be generated automatically through innovative neuro-symbolic AI algorithms. Moreover, the project includes research on the transparency and resilience of the AI models. All the developed components will be validated via both simulation and testbed experiments and integrated into a final proof-of-concept. Besides, a detailed training plan is designed, including schools, industrial days, and soft-skills courses. It involves industrial partners to ensure that the career perspectives of Doctoral Candidates will be significantly increased by participation in the project.

i2CAT coordinates this research and training programme, with the participation of 12 European partners, involving a mix of leading academic institutions, specialized research centres, and major industrial players, specifically from the telecommunications and digital innovation sectors. Particularly, the Catalan research and innovation centre coordinates WP6-Project Management, which aims to provide the internal project management and the overall coordination activities, financial and technical, planning, and control. I

i2CAT will also have a significant role in WP1 – Scalable GenAI/MoE-based models for network management and Work package WP4 – AI transparency, resilience and PoC. WP1 focuses on three primary objectives:

1. Develop a task-agnostic network orchestration and deployment framework based on GenAI, which considers contextual, situational, and temporal factors.
2. Construct Intent-Based Networking (IBN) frameworks that utilize LLMs to break down higher-level intents and goals into sequences of lower-level tasks, enabling goal achievement over time.
3. Explore the potential of Intent-Based Networking in enabling MEC federation.

WP4 aims to create AI functions that are both transparent and robust. This will be achieved by developing in hoc eXplainable AI (XAI) methods for the Large Language Models (LLMs) designed in WP1. Additionally, it involves establishing context for AI resilience systems through the utilization of LLMs in conjunction with Human-in-the-Loop (HITL) processes.