NVIDIA Open Source 6G Tools See 10x Adoption Growth and 1 Million Downloads

NVIDIA open sourced its Aerial wireless software a year ago, joining Sionna to provide a foundation for 6G research; adoption of these tools has grown 10x, surpassing 1 million downloads across more than 8,000 institutions.
King’s College London researchers used NVIDIA Jetson AGX Orin and Sionna to implement goal-oriented semantic communications, reducing robot task completion time by up to 52.6% in fruit-packing and cube-stacking experiments.
SoftBank integrated a neural receiver into NVIDIA Aerial CUDA-Accelerated RAN (ACAR), achieving up to 60% higher uplink throughput than conventional receivers in interference tests, while ETH Zurich researchers reduced block error rates by more than one-third using local measurements.
The University of Leeds developed a demand-prediction system using NVIDIA DGX Spark that reduced RAN power consumption by 45.6% while maintaining connectivity, and CTTC accelerated 5G-LENA simulations from three hours to 10-20 minutes using GPU-accelerated models.

Technical Implementation

Goal-oriented semantic communications transmit compact object descriptions instead of full images to lower latency for robots.
Neural receivers in ACAR learn to recover information from incoming signals to improve decoding under interference.
GPU-accelerated radio channel models in ACAR calculate signal travel between antennas and devices to speed up network simulation.

NVIDIA AI Aerial and NVIDIA Sionna are available on GitHub.
Further innovations will be showcased at 6G Developer Sessions during NVIDIA GTC Washington, D.C.