Summary
NVIDIA says its open-source 5G/6G tools have passed 1 million downloads and are being used in research on robot communications, AI receivers, network energy use and faster simulations. The examples are research tests using 5G equipment, testbeds and simulators.
NVIDIA says its open-source 5G/6G tools have surpassed 1 million downloads, with more than 8,000 research and academic institutions worldwide using the platforms. The company also reports that developer adoption of its AI Aerial and Sionna software grew tenfold over the past year.
The tools are intended to help researchers move between wireless-network simulation, AI training and experiments with real radio equipment. The projects NVIDIA highlights apply them to four practical challenges: keeping robots responsive, improving signal reception, reducing network energy use and speeding up simulations.
From open-source tools to research platforms
NVIDIA says it open-sourced its Aerial wireless software a year ago, joining Sionna and other tools for developing next-generation networks. AI Aerial and Sionna are available on GitHub, according to the company.
The work described is research into future network capabilities, but the examples use 5G equipment, testbeds and simulations. They offer evidence of approaches being explored for 6G rather than demonstrations of a deployed 6G service.
Four experiments target network bottlenecks
Sending robots task-relevant information
At King’s College London, researchers linked a camera-equipped robot arm to an NVIDIA Jetson AGX Orin running Sionna. They tested fruit-packing and cube-stacking tasks with deliberately introduced faults, comparing a communications approach called goal-oriented semantic communications with sending periodic images.
Rather than transmit full images, the system sent compact descriptions of objects, their positions and their relationships. The researchers reported that this reduced task-completion time by up to 52.6% and improved task success compared with the periodic-image approach. The system also detected dropped objects and recovered from unsuccessful actions. In some configurations, a digital twin—a virtual model used to test planned actions—checked movements for possible collisions before execution.
Learning to receive signals more reliably
SoftBank integrated a neural receiver into NVIDIA Aerial CUDA-Accelerated RAN (ACAR) and tested it with a commercial 5G smartphone. In controlled tests with interference, it delivered higher uplink throughput than a conventional receiver at all 15 tested settings, with a maximum gain of roughly 60%. Uplink refers to data sent from a device towards the network.
In a separate project, ETH Zurich researchers used local measurements to tune neural receivers. They reported block error rates more than one-third lower than with selected pretrained versions. Block error rate measures how often received blocks of data are decoded incorrectly.
Matching network capacity to demand
University of Leeds researchers developed a system that forecasts network demand 10 minutes ahead and uses the predictions to guide decisions about putting base stations into a lower-power sleep state. It checks whether neighbouring active stations can serve affected devices and monitors for changes that call for capacity to return.
Tested with a portable 5G network-in-a-box and a digital twin, the approach reduced radio access network (RAN) power consumption by 45.6% relative to a full-power baseline while maintaining device connectivity, the researchers reported.
Speeding up network simulations
The Centre Tecnològic de Telecomunicacions de Catalunya integrated GPU-accelerated radio-channel models running on ACAR into the open-source 5G-LENA network simulator. In an early evaluation, simulation runtime fell from approximately three hours to 10–20 minutes, depending on the integration, while preserving expected statistical behaviour and network performance.
Taken together, the projects show how AI and accelerated computing could help researchers test network designs, improve how devices exchange information and manage radio resources. Their results come from distinct experiments, so the reported gains describe those specific setups rather than a single performance measure for future 6G networks.