What is the technology or concept?
Artificial Intelligence (AI) is expected to play a central role in future 6G networks, enabling systems to automatically optimize performance, manage resources, and adapt to changing conditions. In the context of Open RAN (O-RAN), this is achieved through intelligent applications (xApps) that rely on data to make decisions.
However, for AI to be effective, it must be trained and tested on realistic data. This is where experimental datasets collected from real network environments become essential.
Why is it relevant for 6G or future networks?
Future networks will be highly dynamic, with users moving across cells, varying traffic demands, and complex radio conditions. AI models trained only on simulations may fail to capture this complexity, leading to poor performance in real deployments. Real-world datasets provide a more accurate representation of network behavior, including mobility, interference, and time-varying conditions. This makes them critical for developing reliable and deployable AI solutions for 6G.
How is 6G-LEADER addressing or using it?
Within the 6G-LEADER project, we developed an experimental dataset using an O-RAN-compliant testbed built by Four Dot Infinity (FDI). The initial version focused on evaluating scheduling strategies in a controlled single-cell setup.
Building on this, we extended the dataset to include multiple cells, user mobility, and realistic radio conditions (click here). The system collects measurements not only from user devices but also from the network itself, providing deeper visibility into how the system operates. This approach allows us to move from simplified evaluations to more realistic scenarios that better reflect real network behavior.
What are the future benefits or applications?
Such datasets enable the development of AI models that can handle mobility, optimize resource allocation across cells, and predict network events such as handovers. More broadly, they support research on cross-layer optimization, where decisions are made by considering multiple parts of the network together.
Ultimately, this contributes to building AI-native RAN systems that are more efficient, adaptive, and reliable, supporting the vision of next-generation 6G networks.
