NVIDIA Aerial Expands With New Tools for Building AI-Native Wireless Networks

The telecom industry is increasingly embracing AI to deliver seamless connections — even in conditions of poor signal strength — while maximizing sustainability and spectral efficiency, the amount of information that can be transmitted per unit of bandwidth.

Advancements in AI-RAN technology have set the course toward AI-native wireless networks for 6G, built using AI and accelerated computing from the start, to meet the demands of billions of AI-enabled connected devices, sensors, robots, cameras and autonomous vehicles.

To help developers and telecom leaders pioneer these networks, NVIDIA today unveiled new tools in the NVIDIA Aerial Research portfolio.

The expanded portfolio of solutions include the Aerial Omniverse Digital Twin on NVIDIA DGX Cloud, the Aerial Commercial Test Bed on NVIDIA MGX, the NVIDIA Sionna 1.0 open-source library and the Sionna Research Kit on NVIDIA Jetson — helping accelerate AI-RAN and 6G research.

Industry leaders like Amdocs, Ansys, Capgemini, DeepSig, Fujitsu, Keysight, Kyocera, MathWorks, Mediatek, Samsung Research, SoftBank and VIAVI Solutions and more than 150 higher education and research institutions from U.S. and around the world — including Northeastern University, Rice University, The University of Texas at Austin, ETH Zurich, Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institut, HHI, Singapore University of Technology and Design, and University of Oulu — are harnessing the NVIDIA Aerial Research portfolio to develop, train, simulate and deploy groundbreaking AI-native wireless innovations.

New Tools for Research and Development

The Aerial Research portfolio provides exceptional flexibility and ease of use for developers at every stage of their research — from early experimentation to commercial deployment. Its offerings include:

• Aerial Omniverse Digital Twin (AODT): A simulation platform to test and fine-tune algorithms in physically precise digital replicas of entire wireless systems, now available on NVIDIA DGX Cloud. Developers can now access AODT everywhere, whether on premises, on laptops, via the public cloud or on an NVIDIA cloud service.
• Aerial Commercial Test Bed (aka ARC-OTA): A full-stack AI-RAN deployment system that enables developers to deploy new AI models over the air and test them in real time, now available on NVIDIA MGX and available through manufacturers including Supermicro or as a managed offering via Sterling Skywave. ARC-OTA integrates commercial-grade Aerial CUDA-accelerated RAN software with open-source L2+ and 5G core from OpenAirInterface (OAI) and O-RAN-compliant 7.2 split open radio units from WNC and LITEON Technology to enable an end-to-end system for AI-RAN commercial testing.
• Sionna 1.0: The most widely used GPU-accelerated open-source library for research in communication systems, with more than 135,000 downloads. The latest release of Sionna features a lightning-fast ray tracer for radio propagation, a versatile link-level simulator and new system-level simulation capabilities.
• Sionna Research Kit: Powered by the NVIDIA Jetson platform, it integrates accelerated computing for AI and machine learning workloads and a software-defined RAN built on OAI. With the kit, researchers can connect 5G equipment and begin prototyping AI-RAN algorithms for next-generation wireless networks in just a few hours.

NVIDIA Aerial Research Ecosystem for AI-RAN and 6G

The NVIDIA Aerial Research portfolio includes the NVIDIA 6G Developer Program, an open community that serves more than 2,000 members, representing leading technology companies, academia, research institutions and telecom operators using NVIDIA technologies to complement their AI-RAN and 6G research.

Testing and simulation will play an essential role in developing AI-native wireless networks. Companies such as Amdocs, Ansys, Keysight, MathWorks and VIAVI are enhancing their simulation solutions with NVIDIA AODT, while operators have created digital twins of their radio access networks to optimize performance with changing traffic scenarios.

Nine out of 10 demonstrations chosen by the AI-RAN Alliance for Mobile World Congress were developed using the NVIDIA Aerial Research portfolio, leading to breakthrough results.

SoftBank and Fujitsu demonstrated an up to 50% throughput gain in poor radio environments using AI-based uplink channel interpolation.

DeepSig developed OmniPHY, an AI-native air interface that eliminates traditional pilot overhead, harnessing neural networks to achieve up to 70% throughput gains in certain scenarios. Using the NVIDIA AI Aerial platform, OmniPHY integrates machine learning into modulation, reception and demodulation to optimize spectral efficiency, reduce power consumption and enhance wireless network performance.

“AI-native signal processing is transforming wireless networks, delivering real-world results,” said Jim Shea, cofounder and CEO of DeepSig. “By integrating deep learning to the air interface and leveraging NVIDIA’s tools, we’re redefining how AI-native wireless networks are designed and built.”

In addition to the Aerial Research portfolio, using the open ecosystem of NVIDIA CUDA-X libraries, built on CUDA, developers can build applications that deliver dramatically higher performance.

Join the NVIDIA 6G Developer Program to access NVIDIA Aerial Research platform tools.

See notice regarding software product information.