D4.3–AI-based Techniques for Sustainable & Human-friendly RRM

This document summarizes the achievements of the CENTRIC consortium during the two and a half years of project duration. In particular, it describes the research outcomes of Task 4.2 focused on developing AI-driven techniques for sustainable and human-centric radio resource management. Indeed, as emerging mobile services require to collect increasingly large volumes of data through the radio link and process them via AI/ML models at the network edge, it is imperative to limit the demand for
communication and computational resources. To address this challenge, we designed innovative techniques to minimize resource, hence energy, consumption while fulfilling performance requirements and system constraints. Specifically, we created and tested novel solutions for (i) the efficient management of radio interfaces and radio traffic, despite the highly dynamic operational environments, (ii) the energy-aware training and execution of AI/ML models at the network edge, and (iii) the configuration
of innovative network architectures that besides network performance metrics, account for EMF exposure. Our solutions often leverage data-driven approaches, which can tackle the complexity and the scale of new generation network systems more effectively when compared to traditional optimization methods. Overall, the solutions and methods we propose represent an effective set of tools to make the support of AI/ML-based mobile services sustainable.