Cooperation for AI-supported 5G-Advanced and 6G tests

The system integrates ray-tracing-based channel models from the NVIDIA Sionna Research Kit into Rohde & Schwarz's measurement technology. This makes it possible to virtually simulate radio conditions as they actually occur at specific locations. This creates a closed test environment for the development and validation of AI-RAN applications that combines simulation and real hardware.

The collaboration builds on previous proofs of concept for neural receiver concepts - including special approaches for pilotless communication. What is new is that the validation is no longer only at connection level, but at system level with the complete 5G NR protocol stack. This allows developers to check complex functions in the interaction of software, AI algorithms and radio hardware.

The NVIDIA Sionna Research Kit runs on a single NVIDIA DGX Spark and operates a software-defined 5G RAN based on OpenAirInterface. In parallel, it runs AI inference workloads that need to meet the real-time requirements of wireless systems. To further open up the platform, an AI/ML-supported link adaptation algorithm was implemented. It dynamically adapts the downlink modulation and coding scheme (MCS) to optimize spectrum efficiency and link reliability.

The AI-based link adaptation learns location-specific propagation characteristics and at the same time takes into account the behavior of the respective end device. This increases the requirements for test systems that can reproducibly capture these effects. This is precisely where the approach comes in.

Test setup with SMW200A and FSW in closed loop

Technically, the solution combines the R&S SMW200A vector signal generator with dynamic channel emulation and the FSW signal and spectrum analyzer. Both devices enable the emulation of complex, site-specific radio channels and can be connected to the NVIDIA Sionna RT software for differentiable ray tracing. The closed-loop system allows AI-supported RAN functions to be evaluated under realistic RF conditions - completely in the laboratory.

Gerald Tietscher, Vice President Signal Generators, Power Supplies and Meters at Rohde & Schwarz, said: "We are pleased to continue our existing collaboration with NVIDIA with this latest proof of concept for testing AI-powered base stations for 5G-Advanced and 6G under realistic propagation conditions. This approach, based on digital twin technology and ray tracing, aims to bridge the gap between AI-powered radio simulations and real-world deployment scenarios to enable more efficient and accurate testing of next-generation receiver architectures."

Soma Velayutham, Global Industry Business Development Lead for Telecommunications at NVIDIA, adds: "Synthetic data generation fundamentally changes how AI-RAN systems are trained and validated, as it ensures accuracy, scalability and privacy. This is particularly crucial when data is difficult to obtain. With the NVIDIA Sionna Research Kit, Rohde & Schwarz is demonstrating how industry-leading expertise and innovative technologies can be brought together to accelerate progress in this important area."

Live demonstration at the MWC 2026

Visitors to MWC 2026 can experience the hardware-in-the-loop validation of novel AI-RAN functions and exchange ideas with experts at stand 5A80 in hall 5 of the Fira Gran Via in Barcelona from March 2 to 5, 2026.