CENTRIC Showcased Cutting-Edge AI-RAN Research at ICMLCN 2025 in Barcelona

🕒 2 mins

The CENTRIC project is pleased to report its participation at the IEEE ICMLCN 2025 in Barcelona, a leading event at the intersection of machine learning and next-generation communication networks. CENTRIC partners demonstrated state-of-the-art research in AI-driven radio access networks (AI-RAN), featuring two live prototypes that have been pushing the boundaries of 6G connectivity.

Photos from the live demonstration area in Barcelona – showcasing from left to right: Jacob Hoydis (NVIDIA, FR), Sebastian Cammerer (NVIDIA, FR), Alejandro Villena Rodriguez (Keysight, ES), Carles Navarro Manchon (Keysight, ES), Asil Koc (InterDigital, US)

Left: Demo 8 – Sionna Research Kit (NVIDIA)
Right: Demo 3 – Network Digital Twin with Hardware-in-the-Loop (Keysight, NVIDIA, InterDigital)

Featured Demos from the CENTRIC Ecosystem

▶ Demo 3 (D3): AI-RAN Evaluation via Network Digital Twin and Hardware-in-the-Loop

Demonstrator: Keysight, NVIDIA, InterDigital

This joint demonstration brings to life a high-fidelity testbed for evaluating AI/ML algorithms in the physical layer of future mobile networks. Leveraging a network digital twin and radio channel emulation with PROPSIM, the system emulates realistic RF conditions to test advanced AI receivers and activity detection solutions. The setup includes:

Neural Receiver (by NVIDIA): A real-time, multi-user MIMO system using deep learning for detection and decoding.
AI Activity Recognition (by InterDigital): Using RF signals to detect and classify real-world activity patterns.
Hardware-in-the-loop emulation powered by Keysight and integrated with commercial O-RAN radio units.

This setup offers a tangible preview of how AI can redefine the performance and adaptability of future mobile networks.

▶ Demo 8 (D8): Sionna Research Kit – A GPU-Accelerated Platform for AI-RAN

Demonstrator: NVIDIA

The Sionna Research Kit is a complete, GPU-accelerated research platform built for developing and validating AI/ML algorithms in 5G and beyond. Using the Jetson AGX Orin and integrated with OpenAirInterface (OAI), the platform enables:

Rapid prototyping of 5G NR and O-RAN-compliant AI solutions
Real-time deployment of neural receivers trained using NVIDIA Sionna
High-throughput signal processing with TensorRT for inference

This powerful yet affordable testbed allows researchers to deploy and test edge AI use cases in real-world settings—marking a significant step forward for scalable AI-RAN experimentation.