How NVIDIA Open Source Tools Are Helping Researchers Build Next-Generation 6G Networks

A year after open-sourcing its Aerial wireless software, NVIDIA says its open source 5G and 6G tools have passed one million downloads and are used by more than 8,000 research and academic institutions. In a company blog post, it describes four research projects built on Aerial and its Sionna software.

At King's College London, a robot arm sent compact descriptions of objects and their positions instead of full images. In fruit-packing and cube-stacking tests, that cut task completion time by up to 52.6 percent, according to the post.

SoftBank built a neural receiver into NVIDIA's radio access software and tested it with a commercial 5G phone. Under interference it delivered higher uplink throughput than a conventional receiver in all 15 settings tested, with a peak gain of about 60 percent. ETH Zurich researchers found that tuning such receivers on local measurements cut block error rates by more than a third.

At the University of Leeds, a system that forecasts demand 10 minutes ahead and puts base stations to sleep reduced radio network power use by 45.6 percent against a full-power baseline while keeping devices connected. And Spain's CTTC used GPU-accelerated channel models to shorten a network simulation from about three hours to between 10 and 20 minutes.

These are results reported by NVIDIA and the research groups it features. The company presents them as steps toward 6G networks designed around AI from the start.