At the European Conference on Networks and Communications -EuCNC 2023, a testbed from the University of Pisa based on a network simulator was presented. The activity was developed within the Hexa-X project, led by Nokia and Ericsson, which brought together companies and research for the development of the 6G network.
The Internet network for 2023 and beyond will have to have stringent requirements for resilience, robustness and digital inclusion. This was addressed by the European Hexa-X project (https://hexa-x.eu/), led by Nokia and Ericsson in a consortium of 25 organizations from 9 countries, and whose results were presented at the European Conference on Networks and Communications (European Conference on Networks and Communications -EuCNC 2023). At that conference, the important results achieved by the research team earned the project the award for the best exhibition space.
from left Giovanni Nardini and Pietro Ducange, professors of Computer Engineering at DII.
Image
Among these, a testbed created by the research unit of the Department of Information Engineering of the University of Pisa (DII) together with Intel and TIM, which emulates a network beyond 5G integrated with innovative and reliable artificial intelligence systems (https://www.youtube.com/watch?v=azuTyB-LdmQ).
“Hexa-x - explains Giovanni Nardini, professor of computer engineering at the University of Pisa - has addressed challenges such as increasing the machine learning capabilities of artificial intelligence, sustainability to optimize network consumption from an energy point of view, the study of efficient solutions for global coverage of the service that is reliable and can provide top-level experiences while guaranteeing extreme transmission speeds and extremely low latencies.
Some of the keywords to address these challenges are federated learning and “explainable” artificial intelligence models, which combined are called Fed-XAI (Federated Learning of eXplainable AI).
On the one hand, Federated Learning allows training AI models collaboratively between different users, ensuring better performance than a traditional approach and at the same time preserving the privacy of the users' data. On the other hand, “explainable” AI models allow producing information on the functioning of the model itself, and therefore allow the results obtained from these models to be more interpretable by human users, contributing to increasing the robustness and reliability of AI techniques”.
“During the conference - he continues - we used our Simu5G network simulator, the first Open-Source network simulator that works in real time, to show that the integration of Federated Learning techniques of explainable AI models within future mobile networks is a promising approach towards reliable and high-quality innovative services to users, while maintaining the privacy of the users themselves.
The simulator itself is an important result, given that since its publication it has been downloaded more than 6000 times from the site we have made available (http://simu5g.org), demonstrating its potential to support the network innovation and digitalization processes necessary for our country”.
Research on artificial intelligence capable of guaranteeing robustness, privacy in data management and transparency and, at the same time, its application to build the network of the future, is today central for those who work in ICT.
This is one of the research lines of FoReLab, the laboratory of the University of Pisa that deals with research on technologies for 2030 and beyond, for the construction of what is now called "society 5.0", and which includes not only innovation in production methods, but in which technologies, in all their declinations, must guarantee resilience, inclusion, and centrality of people and the environment.
