The journey toward 6G is not just about faster speeds or greater connectivity. It’s about creating smarter, self-optimizing networks that can think, learn and adapt in real time. One of the most exciting innovations making this possible, is the concept of the Network Digital Twin (NDT) – a virtual mirror of a physical network that allows researchers to model, test and improve its performance safely and intelligently.
A new paper by the NCSR Demokritos team, developed under the MARE project, presents a major step forwards in this area. Titled “Design and Evaluation of a Network Digital Twin Framework for 5G/6G Architectures”, the publication demonstrates how digital twin technology can help design and manage intelligent, AI-driven networks that will be the foundation of the 6G era.
Network Digital Twin framework for native-AI support in a 6G system
What is a network Digital Twin?
Imagine having a complete digital replica of a network – one that updates in real time, mimicking how data flows, how devices connect and how services behave. This digital twin allows engineers to experiment safely with new technologies, predict problems before they happen and test solutions without disrupting real-world services.
For example, an operator could use a digital twin to simulate a surge in user demand or a cyber-attack and immediately see how the network would react. The lessons learned can then be applied to strengthen the actual deployed network.
Bridging the physical and the virtual
The paper presents how the research team designed and implemented a full Network Digital Twin framework, integrating it with a real 5G testbed. Using open-source tools such as Open5GS, Prometheus and Grafana, the team created a virtual environment that accurately reflects real network conditions.
This environment was used to compare the performance of the digital twin with the physical network under identical scenarios – such as different user loads or varying quality-of-service requirements. The results showed that the digital twin could reproduce real-world behavior, confirming its value as a safe and reliable testing ground.
Network Digital Twin framework for native-AI support in a 6G system
AI/ML framework for native-AI support in a 6G system
Why this matters for 6G and MARE
In the 6G vision, AI will play a central role in managing complex networks. However, AI models need vast amounts of accurate data to learn effectively – and collecting that data directly from live networks is often too costly or risky.
That’s where digital twins come in. They can generate realistic data and simulate network conditions, enabling AI systems to be trained and optimized before being deployed in the real world. This is particularly important for security, resilience and efficiency purposes. These are also three pillars of the MARE project’s mission to create trustworthy 6G ecosystems.
A foundation for the Future
This work marks a milestone in building the foundations of AI-driven, self-aware networks. The Network Digital Twin framework developed by NCSR Demokritos demonstrates how European research is leading the way in designing secure and intelligent 6G systems.
By merging physical and virtual worlds, MARE and its partners are shaping a future where our networks don’t just connect us – they also learn, adapt and protect us too.
Publication Details
Design and Evaluation of a Network Digital Twin Framework for 5G/6G Architectures
Authors: I. Vasalos, M. Christopoulou, A. Vasalos, M. A. Kourtis, N. Dimitriou and G. Xylouris.
MARE Partner: National Centre For Scientific Research Demokritos (NCSRD)
Publication: 2025 21st International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT), Lucca, Italy, 2025, pp. 915-920. Conference website.
Abstract: Network Digital Twin (NDT) systems serve as a real-time, virtualized replica of physical networks, enabling closed-loop automation and AI/ML driven network intelligence. In this paper, we present a comprehensive approach to implementing an NDT framework tailored for B5G/6G networks. Through the proposed methodology, we demonstrate the generation of the NDT model, the accurate modeling of network topology, and the seamless synchronization with real world network operations. Our research delves into the deployment_and validation of an NDT prototype platform, seamlessly integrated with a 5G physical network testbed. We also present the end-to-end deployment of the framework and showcase its integration with a physical 5G testbed, effectively demonstrating its capability to support AI-native functionalities in next-generation mobile networks.
