Fact Checked by 
Nvidia has revealed a range of advanced workflows to accelerate its humanoid development aspirations at the Conference for Robot Learning (CoRL) in Munich, Germany.
The new tools showcased at the industry event have been tipped to significantly assist robotics developers in their work on AI-enabled robots.
The line-up of features includes the general rollout of the Nvidia Isaac Lab robot learning frameworks, six new humanoid learning workflows for Project GR00T, and development tools for video data curation and processing, including the Nvidia Cosmos tokenizer and Nvidia NeMo Curator.
Robotics developers are provided with enhanced visual tokenization through the open-source Cosmos tokenizer which is able to break down images and videos into high-quality tokens with advanced compression rates, up to 12 times faster than current tokenizers on the market.
A similar uplift is available with NeMo Creator, with its offering providing video processing curation around seven times faster than competitors.
Nvidia in collaboration with Hugging Face
Meanwhile, Project GR00T is Nvidia’s program to make forward leaps in robot development. It has introduced six fresh workflows, including GR00T-Gen for constructing AI-driven 3D environments and GR00T-Mimic for trajectory generation. It has also delivered GR00T-Dexterity for dexterous manipulation, GR00T-Control for body control, GR00T-Mobility for navigation, and GR00T-Perception for sensory processing.
At CoRL, there was also an announcement of a collaboration between Nvidia and Hugging Face – which recently reached one million downloadable AI models – that will see both entities combine their open-source robotics communities. Nvidia’s AI, Omniverse, and Isaac robotics tech will be leveraged with Hugging Face’s open AI platform LeRobot to further advance the work of researchers and developers across many industries, particularly in manufacturing, logistics, and healthcare.
Image credit: Via Midjourney
