“The native AI architectural model proposed by DAEMON has been adopted by the 5G PPP and 6G AI architecture working groups, from where it has the potential to gain further traction in major standardization bodies such as 3GPP,” says Marco Fiore, research professor at IMDEA Networks and coordinator of the DAEMON project. He adds: “Furthermore, the project has generated two pilot projects now deployed on Telefónica’s production network, five solutions included in the European Commission’s Innovation Radar initiative, nine patents, and over 100 scientific articles. All of this together constitutes DAEMON’s significant legacy.”.
The IMDEA Networks research team, particularly those in the Network Data Science group, has contributed to research on customizing loss functions for network-adapted AI, introducing concepts that are also highly innovative for the machine learning community, such as loss meta-learning. Furthermore, they have demonstrated how AI models can be directly integrated into programmable network hardware for classifying traffic flows with delays on the order of tens of nanoseconds.
Challenges
The project addressed several challenges, including:
- Understanding the limits of AI for mobile networks: They conducted a systematic and critical analysis of which network management tasks can be adequately solved with AI models rather than other techniques and provided a robust set of guidelines for using machine learning in next-generation mobile network automation.
- Designing highly customized AI-powered network management solutions: Based on the above guidelines, they developed and implemented a series of algorithms tailored to the unique requirements and constraints of mobile network environments, advancing the state of the art in automating a precise set of core network management tasks.
- Designing an end-to-end AI-native architecture: They outlined the first AI-native mobile network architecture model, which, derived from current standardization trends, enables the coordination of the numerous and varied AI instances deployed across the network.
High Scientific Output and Future of Research:
The research conducted at DAEMON has undoubtedly been pioneering in nature. The solutions produced have been implemented and demonstrated in production-relevant network environments. At the same time, the project's innovative and high-quality research output is evidenced by numerous publications (more than 30) in top-tier conferences and journals, such as ACM SIGCOMM, ACM MobiCom, IEEE INFOCOM, ACM/IEEE Transactions on Networking, and IEEE Transactions on Mobile Computing, among others.
Although the DAEMON project has concluded, the research continues with ORIGAMI, a European project launched in January 2024. The search for mobile networks with lower latency, higher performance, and greater reliability is progressing, promising improvements in various sectors and social activities.
