NVIDIA Advances Humanoid Robotics with New Isaac GR00T Updates and Blackwell Systems
NVIDIA has announced significant updates to its Isaac platform, aimed at accelerating the development and deployment of humanoid robots. The updates include NVIDIA Isaac GR00T N1.5, a foundation model for humanoid reasoning and skills, and NVIDIA Isaac GR00T-Dreams, a blueprint for generating synthetic motion data to train robots.
Enhancements to Isaac GR00T
The newly released GR00T N1.5 model represents a significant improvement over its predecessor, GR00T N1. Developed using synthetic training data generated by the GR00T-Dreams blueprint, GR00T N1.5 can better adapt to new environments and workspace configurations. It also demonstrates enhanced object recognition through user instructions, significantly improving its success rate in common material handling and manufacturing tasks.
Synthetic Data Generation with GR00T-Dreams
GR00T-Dreams is a novel blueprint that generates vast amounts of synthetic motion data, or neural trajectories, to teach robots new behaviors. By post-training Cosmos Predict world foundation models (WFMs) and using a single image as input, GR00T-Dreams can generate videos of robots performing new tasks in various environments. The blueprint then extracts action tokens that are used to train robots to perform these tasks.
Accelerating Robot Development with Blackwell Systems
NVIDIA is also introducing its Blackwell systems, designed to accelerate humanoid robot development. The NVIDIA RTX PRO 6000 Blackwell workstations and RTX PRO Servers offer a unified architecture for running various robot development workloads, including training, synthetic data generation, robot learning, and simulation.
Industry Adoption
Several leading robot manufacturers, including Agility Robotics, Boston Dynamics, Foxconn, Lightwheel, NEURA Robotics, and XPENG Robotics, are adopting NVIDIA’s Isaac platform technologies. These companies are leveraging the updates to advance their humanoid robot development and deployment capabilities.
Simulation and Data Generation Frameworks
To address the challenge of generating diverse data for robot training, NVIDIA has unveiled several simulation technologies. These include NVIDIA Cosmos Reason, a new WFM for curating high-quality synthetic data, and NVIDIA Isaac Sim 5.0, a simulation and synthetic data generation framework that will soon be available on GitHub. Additionally, NVIDIA Isaac Lab 2.2, an open-source robot learning framework, will support new evaluation environments for testing GR00T N models.
Conclusion
NVIDIA’s latest advancements in its Isaac platform and the introduction of Blackwell systems mark a significant step forward in humanoid robotics. By providing developers with powerful tools for robot development, simulation, and training, NVIDIA is poised to accelerate the deployment of sophisticated humanoid robots across various industries.
