Rohde & Schwarz y Nvidia crean un receptor neural con IA/ML 6G

The Summit Brooklyn 6G is the framework in which Rohde & Schwarz y Nvidia present an AI/ML-based neural receiver with custom modulation, a further step in exploring native AI air interfaces for 6G.

In collaboration with Nvidia, Rohde & Schwarz has presented an improvement to its recent demo hardware-in-the-loop by one receptor neuronal, demonstrating the performance improvements achievable when using trained ML models compared to the traditional signal processing, while for the first time the transmitter side is also optimized.

With this demonstration, framed at the Brooklyn 6G summit, visitors can see for the first time how a neural receptor works combined with a custom modulation learned in a 5G NR multiple-input multiple-output (UL MIMO) uplink scenario. The facility combines several testing solutions to signal generation and analysis de Rohde & Schwarz y la librería de código abierto acelerada en la GPU Nvidia Sionna para permitir 5G and 6G link simulations. Nvidia's Sionna enables rapid prototyping of complex communications system architectures and provides native support for integrating machine learning into 6G signal processing.

This demonstration is based on the success obtained at the Mobile World Congress 2023 in Barcelona, where both companies demonstrated for the first time the concept of receptor neuronal in a loop hardware experiment involving two independent companies. Academia, leading research institutes and leading industry experts around the world predict that a Future 6G standard will use AI/ML for signal processing tasks, such as channel estimation, channel equalization and demapping. This concept is understood as receptor neuronal, a concept that results in improving overall link performance, and therefore throughput, compared to current high-performance deterministic software algorithms used in 5G NR, while keeping computational complexity at a manageable level.

Rohde & Schwarz actively supports 6G research activities in Europe, Asia and the US, while contributing to research projects, industrial alliance work and collaborating with research institutes and universities. The company's test and measurement solutions and expertise are helping pave the way for the next generation of wireless communications, expected to be commercially deployed around 2030.

Claves técnicas de la demo 5G/6G de Rohde & Schwarz y Nvidia

Testbed setup takes advantage of the vector signal generator R&S SMW200, which emulates a single user transmitting a 100 MHz (273 PRB) wide signal in the upward direction at a 2.14 GHz carrier frequency in a 2x4 MIMO configuration. Fading and noise are applied to the transmission to emulate realistic radio channel conditions.

The multipurpose satellite receiver R&S MSR4 captura la señal de RF instantáneamente con sus canales independientes 4x 200 MHz. La señal digitalizada se transmite en tiempo real y se transfiere al marco de Pruebas Basadas en Servidor (SBT) de R&S, que incluye microservicios del explorador de señales vectoriales (VSE) R&S VSE, which performs synchronization and fast Fourier transform (FFT) calculation. The post-FFT data serves as input to a neural receiver implemented and trained with NVIDIA Sionna.

As part of the demonstration, he receptor neuronal trained with a custom constellation is compared to the classical concept of linear minimum mean square error (LMMSE) receiver architecture, which applies traditional signal processing techniques based on deterministically developed software algorithms. These high-performance algorithms are widely adopted in current 4G and 5G mobile networks.

To verify the performance of this improved neural receptor, including custom modulation, the capabilities of the vector signal generator were expanded for this demonstration. R&S SMW200A to allow the user to freely configure each individual constellation point by magnitude and phase in the IQ domain based on the selected QPSK or QAM modulation. The signal and spectrum analyzer R&S FSW and the software R&S Vector Signal Explorer (VSE) They also support custom modulation, enabling users to evaluate transmitter performance using well-known key performance indicators such as error vector magnitude (EVM), even when using learned custom constellations.