Engineer News Network The ultimate online news and information resource for today’s engineer
Rohde & Schwarz, in collaboration with NVIDIA, continues to drive AI-RAN innovation for 5G-Advanced and 6G. The latest testbed integrates ray-tracing-based, site-specific channel emulation with the NVIDIA Sionna Research Kit to enable digital twin-based hardware-in-the-loop testing without leaving the lab.
The testbed integrates hardware-in-the-loop site-specific channel emulation using the NVIDIA Sionna Research Kit, enabling testing of AI-RAN applications under realistic channel conditions. The demonstration highlights the long-term collaboration of Rohde & Schwarz and NVIDIA, focusing on prototyping and validation of AI-RAN with cutting-edge test and measurement solutions.
Evolving from prior proof-of-concepts in advanced neural receiver design – including custom constellations for pilotless communication – the new testbed advances from link-level validation to system-level verification using the full 5G NR protocol stack.
Powered by a single NVIDIA DGX Spark, the NVIDIA Sionna Research Kit runs a software-defined 5G RAN based on OpenAirInterface, while supporting AI inference workloads that comply with the strict real-time constraints of wireless systems. To showcase the flexibility of the research platform, a novel AI/ML-enhanced link adaptation algorithm has been integrated into the end-to-end system. It dynamically adjusts the downlink modulation and coding scheme (MCS) to optimise spectral efficiency and link reliability. The AI-driven link adaptation can learn not only site-specific propagation characteristics but also user equipment-specific behavior on the fly, emphasising the need for end-to-end testbeds that capture these effects.
The testbed integrates the R&S SMW200A vector signal generator featuring dynamic channel emulation capabilities and the FSW signal and spectrum analyzer. Jointly, these instruments enable the emulation of complex site-specific radio channels, seamlessly interfacing with the NVIDIA Sionna RT differentiable ray-tracing software. This closed-loop setup enables researchers and developers to evaluate the performance of novel AI-driven RAN features under dynamic, site-specific RF conditions – all without leaving the lab.
