Research Area

6G hyper-distributed edge-to-cloud continuum

hyper-distributed edge-to-cloud continuum, co2-aware, disaggregated ran, intelligence workload placement, digital twinning

The world economy is increasingly reliant on digital systems, leading to the emergence of new global digital services. To meet this demand, distributed and decentralized clouds are becoming the norm. Mobile network operators play a vital role in delivering services across various domains. The 6G hyper-distributed edge-to-cloud continuum research aims to address these challenges using AI-based techniques. The goal is to create an energy-efficient and fully distributed AI-powered platform that can manage secure edge nodes across multiple domains. This research activity has three main sub-directions.

• Energy-aware disaggregated radio access platform. The first sub-direction aims to develop an energy-aware disaggregated radio access platform. It leverages strategies like network-based decision making, resource allocation, and energy-efficient hardware to reduce energy consumption in wireless networks. This leads to cost savings and environmental benefits
• Digital twinning for 6G. The research focuses on digital twinning for 6G networks, investigating data collection, computational needs, interoperability, security, and standardization for application orchestration. Real-world data and synthetic traffic traces will be used to create accurate digital twins, leveraging the CINECA cluster for computational requirements. Standardized approaches will be explored to enhance scalability and reliability, while the hyper-distributed edge-to-cloud continuum will ensure security and privacy of digital twin data
• Energy-efficient provision of AI-driven applications. The research focuses on AI-enabled application service orchestration, aiming to maximize its potential across multiple domains. It utilizes established standards and innovative solutions to develop end-to-end orchestration, enabling simulation and monitoring of Quality of Service metrics. The process also addresses environmental impact by reducing carbon footprint through CO2-aware deployment. Additionally, a distributed learning approach called hybrid federated-split learning will be used to train a global machine learning model without sharing raw data across multiple silos.

Roberto Riggio

Research centers: University of Antwerp (BE), University of Ghent (BE), TU Berlin (DE), TU Munich (DE), University of Avignone (FR), Politecnico di Milano (IT), TU Wien (AS), IMEC (BE), TU delft (NE), University of Waterloo (CA), University of Campinas (BR), CNAM (FR), INRIA (FR), RISE (SK) – Compannies: Safran (DE), Italtel S.p.A. (IT), TIM S.p.A  (IT), 8BELLS (CY), ATOS (ES), TurkCell (TU), Deutsche Telekom (DE), Ericsson (SK)

H2020 AI@EDGE