NVIDIA has announced a number of updates for humanoid robotics, including a way of generating synthetic motion data to teach robots new behaviors, and a foundation model for humanoid reasoning. 

NVIDIA's robotics platform can be used to train, develop, and deploy AI-enabled robots. The announcements of the new facilities were made at the Computex 2025 trade show in Taiwan,

The first announcement is a blueprint that provides a way to generate large amounts of synthetic motion data, known as neural trajectories, that can be used to teach robots new behaviors, including how to adapt to changing environments.

The way the blueprint, NVIDIA Isaac GR00T-Dreams, works is that developers first post-train Cosmos Predict world foundation models (WFMs) for their robot. Then, using a single image as the input, GR00T-Dreams generates videos of the robot performing new tasks in new environments. The videos are then used to extract action tokens that are used to teach real robots how to perform these new tasks. The action tokens are compressed, digestible pieces of data.

NVIDIA says its research arm used the GR00T-Dreams blueprint to generate synthetic training data to develop an updated version of its GR00T N1 foundation model in 36 hours, compared with what would have taken nearly three months of manual human data collection.

The blueprint complements the Isaac GR00T-Mimic blueprint, which was released at the NVIDIA GTC conference in March. While GR00T-Mimic uses the NVIDIA Omniverse and NVIDIA Cosmos platforms to augment existing data, GR00T-Dreams uses Cosmos to generate entirely new data.

Alongside the new blueprint, NVIDIA announced a number of new and updated technologies aimed at accelerating training pipelines. There's a new WFM (World Foundation Model) called Cosmos Reason. This is an open reasoning model that is aware of both location and time. It interprets visual inputs, analyzes them in the context of a provided text prompt, runs chain-of-thought reasoning to reward responses, and generates decisions or captions. Cosmos Reason is now available on Hugging Face. 

Another blueprint, NVIDIA Isaac GR00T-Mimic, can generate exponentially large quantities of synthetic motion trajectories for robot manipulation, using just a few human demonstrations; while Open-Source Physical AI Dataset now includes 24,000 high-quality humanoid robot motion trajectories used to develop GR00T N models.

The final elements of this set are NVIDIA Isaac Lab 2.2, an open-source robot learning framework that supports new evaluation environments to help developers test GR00T N models; and NVIDIA Isaac Sim 5.0, a simulation and synthetic data generation framework, which NVIDIA says will soon be openly available on GitHub.

More details of the new models and blueprints are available from NVIDIA. 

More Information

NVIDIA

Related Articles

NVIDIA's Turing - A Big Leap Forward For GPUs

NVIDIA's Neural Network Drives A Car 

A Billion Neuronal Connections On The Cheap 

NVIDA Updates Free Deep Learning Software

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info