A naukri.com initiative
VentureBeat
1M
159
Nvidia advances robot learning and humanoid development with AI and simulation tools
- Nvidia has announced new AI and simulation tools that can push the advancement of humanoid development and robot learning. The tools will allow developers to significantly boost their work on AI-enabled robots, helping them to accelerate robot and humanoid development. Nvidia has also revealed six new humanoid robot learning workflows for Project GR00T and a new Nvidia Isaac Lab robot learning framework.
- Nvidia Isaac Lab is built using the open-source robot learning framework, Nvidia Omniverse. At scale, developers can use the framework to train robot polices. The software applies to any embodiment robot, including humanoids to quadrupeds and collaborative robots.
- Isaac Lab is set to be used by many top robot makers, including Agility Robotics, Field AI and Swiss-Mile, among others.
- Nvidia has also released 23 papers for the advancement of robot learning, including nine workshops at the Conference for Robot Learning (CoRL) in Germany.
- World models, simulation, robotics application development, and world curation pose significant challenges due to their data-intensive and computational necessity. Nvidia has, therefore, released the Nvidia Cosmos tokenizer and the Nvidia NeMo Curator for video processing to facilitate high compression rates and curation without compromising visual tokenization.
- With humanoid robots, building advanced humanoids is regarded as a challenging multilayered task that demands a collaborative and technological approach. Building such humanoids require robots that can perceive, act and learn human-robot interactions effectively.
- Nvidia’s Project GR00T initiative aims to foster the accelerated development of humanoid robot ecosystems by providing accelerated libraries, foundation models, and data pipelines. The workflows for Project GR00T enable humanoid developers to realise the most challenging humanoid robot capabilities.
- 1X, a humanoid robot company, has updated the 1X World Model Challenge dataset to integrate with Nvidia Cosmos tokenizer. Other developers like Xpeng Robotics and Hillbot are using the tool to maintain high-resolution videos and images.
- The Nvidia Cosmos tokenizer is available on Hugging Face and GitHub, while the Nvidia Isaac Lab is open source on GitHub. NeMo Curator will be available later this month.
- Nvidia aims to reduce processing time for AI development and promote scalability across multi-node GPU systems, hence reducing costs for Robot developers and AI curators.
Read Full Article
9 Likes
For uninterrupted reading, download the app