NVIDIA Highlights Research Deploying Open Source AI Tools for Next-Generation 6G Wireless Networks

NVIDIA stated that its open source wireless software suite, including Aerial and Sionna, has reached over 1 million downloads and is being utilized by more than 8,000 research institutions to develop AI-native 6G network capabilities. According to NVIDIA, researchers are deploying these tools across edge computing, signal processing, energy optimization, and network simulation. At King's College London, researchers utilized NVIDIA Jetson AGX Orin with Sionna to implement goal-oriented semantic communications for robotic manipulation, reporting a task completion time reduction of up to 52.6% compared to sending periodic images. In wireless reception tests, SoftBank integrated a neural receiver into NVIDIA Aerial CUDA-Accelerated RAN (ACAR), reporting up to a roughly 60% uplink throughput gain under interference conditions. Additionally, ETH Zurich demonstrated that tuning neural receivers with local data reduced block error rates by over one-third compared to selected pretrained baselines. For network efficiency, the University of Leeds tested an AI demand forecasting framework with an Aerial-enabled testbed, reporting a 45.6% reduction in radio access network power consumption. To address simulation bottlenecks, the Centre Tecnològic de Telecomunicacions de Catalunya integrated GPU-accelerated radio channel models into the 5G-LENA simulator, shortening simulation times from approximately three hours to 10–20 minutes in early evaluations.

Key takeaways

• NVIDIA reported global adoption of its Aerial and Sionna open source wireless tools has grown 10x over the past year, surpassing 1 million downloads across over 8,000 institutions.
• King's College London tested semantic communication with a Jetson-powered robotic arm, reporting up to a 52.6% reduction in task completion time.
• SoftBank and ETH Zurich demonstrated AI-based neural receivers, achieving up to a 60% throughput gain and lower block error rates in wireless reception tests.
• University of Leeds and CTTC showed AI-driven base station power savings of 45.6% and network simulation runtime reductions from three hours to 10–20 minutes.