6G next-generation mobile networks, with first deployments targeted for 2030, will heavily incorporate AI for smarter operation and as a service to network users. AI models will be embedded throughout the network to drive Radio Access Network (RAN) optimization, mobility decisions, and resource scheduling. To make this AI both trustworthy and energy-aware, novel methods must be tested under varied, realistic conditions. However, the main problem in making AI-in-the-loop models reproducible and comparable across labs and time is that today’s testbeds are hard to govern, require a steep learning curve, and expose countless configuration parameters; they are hard to reproduce across labs due to the vast amount of software versions across components (e.g., RAN, core, kernel, drivers), heterogeneous orchestration choices (e.g., Kubernetes, OpenShift, Docker Compose), and divergent configuration practices (tuning, feature flags, container images) that can cause configuration drift that silently alters testbed behavior. While there are proprietary solutions that aim to address some of these issues, they are often expensive and lead to vendor lock-in, limiting openness and flexibility. AGORA-6G will bring CS-EE researchers together to develop a front-end experimental software (CS), collect experimental data from the existing testbeds at TU/e and UT (EE), and generate open datasets from the testbeds for the community’s use. This project explicitly addresses the integration of AI as an NTS key enabling technologies in the future-generation networks.
Building on this context, the global objective of this project is to promote trustworthy, reproducible AI-native 6G experimentation by standardizing configuration disclosure and providing a transparent framework that turns testbed runs into shareable datasets and auditable models whose behaviour is traceable, retrainable, and comparable across laboratories, operators, and vendors.
- Syllas Rangel Carneiro Magalhaes, Postdoctoral Researcher, Faculty of EEMCS, University of Twente
- Blas Gómez Mora , Postdoctoral Researcher, Faculty of EEMCS, University of Twente
- Kishor Joshi, Assistant Professor, Center for Wireless Technology Eindhoven, TU/Eindhoven
- Georgios Iosifidis, Associate Professor, Department of Software Technology, TU Delft
- June Sallou, Assistant Professor, Information Technology, Wageningen University & Research