NVIDIA announced NVIDIA Isaac GR00T N1.5, the first update to NVIDIA’s open, generalized and fully customizable foundational model for humanoid reasoning and skills; NVIDIA Isaac GR00T-Dreams, a blueprint for synthetic motion data generation; and a series of NVIDIA Blackwell systems to accelerate humanoid robot development.
Humanoid and robotics developers Agility Robotics, Boston Dynamics, Fourier, Foxlink, Galbot, Mentee Robotics, NEURA Robotics, General Robotics, Skild AI and XPENG Robotics are adopting NVIDIA Isaac™ platform technology to advance the development and deployment of humanoid robots .
“Physical AI and robotics will power the next industrial revolution,” said Jensen Huang, founder and CEO of NVIDIA. “From AI brains for robots to simulated worlds to practice in, to AI supercomputers to train the underlying models, NVIDIA provides the building blocks for every stage of robotics development.”
Also Read: Mitsubishi, NTT & SK C&C Launch Carbon Tracking Pilot
New Data Generation Isaac GR00T Blueprint Fills Data Gaps
Introduced during Huang’s COMPUTEX keynote, NVIDIA Isaac GR00T-Dreams is a blueprint that helps generate vast amounts of synthetic motion data (i.e. neural trajectories) that physical AI developers can use to teach robots new behaviors, including how to adapt to changing environments.
Developers can first post-train the Cosmos Predict World Foundation Model (WFM) for their robot. Then, using a single image as input, GR00T-Dreams generates videos of the robot performing new tasks in new environments. Blueprints are then used to extract action tokens (compressed, digestible pieces of data) and teach the robot how to perform these new tasks.
The GR00T-Dreams blueprint complements the Isaac GR00T-Mimic blueprint released at the NVIDIA GTC conference in March . While GR00T-Mimic leverages the NVIDIA Omniverse ™ and NVIDIA Cosmos ™ platforms to augment existing data, GR00T-Dreams uses Cosmos to generate entirely new data.
New Isaac GR00T Model Advances Humanoid Robot Development
Using GR00T-Dreams Blueprint to generate synthetic training data, NVIDIA Research developed an update to GR00T N1, GR00T N1.5 , in just 36 hours, which would have taken nearly three months of manual data collection by humans.
The GR00T N1.5 is now more adaptable to new environments and workspace configurations, as well as able to recognize objects when instructed by the user. This update significantly improves the success rate of common material handling and manufacturing tasks such as sorting and storing objects.
Early adopters of the GR00T N model include AeiRobot, Foxlink, Lightwheel, and NEURA Robotics. AeiRobot is using the model to enable ALICE4 to understand natural language instructions and perform complex pick-and-place workflows in industrial environments. Foxlink Group is using the model to improve the flexibility and efficiency of industrial robotic manipulators, Lightwheel is using it to validate synthetic data for faster deployment of humanoid robots in factories, and NEURA Robotics is evaluating the model to speed up the development of home automation.
New robot simulation and data generation framework speeds up training pipelines
Developing highly skilled humanoid robots requires large volumes of diverse data that are expensive to acquire and process, and the robots need to be tested in the physical world, which can be costly and risky.
To bridge the data and testing gap, NVIDIA announced the following simulation technologies:
NVIDIA Cosmos Reason , a new WFM that uses thought-chain reasoning to curate accurate, high-quality synthetic data for training physical AI models, is now available on Hugging Face .
Cosmos Predict 2 , used in GR00T Dreams , is also coming soon to Hugging Face, bringing performance enhancements for higher quality world generation and reduced hallucination.
NVIDIA Isaac GR00T-Mimic is a blueprint for generating exponentially larger amounts of synthetic motion trajectories for robotic manipulation based on just a few human demonstrations.
The open-source Physical AI dataset contains 24,000 high-quality humanoid robot motion trajectories that were used to develop the GR00T N model.
NVIDIA Isaac Sim™ 5.0 , a simulation and synthetic data generation framework , will soon be available on GitHub .
NVIDIA Isaac Lab 2.2 , an open source robotics learning framework , includes a new evaluation environment to help developers test GR00T N models.
Foxconn and Foxlink are accelerating their robotics training pipelines with GR00T-Mimic Blueprints for synthetic motion manipulation generation; Agility Robotics, Boston Dynamics, Fourier, Mentee Robotics, NEURA Robotics and XPENG Robotics are using NVIDIA Isaac Sim and Isaac Lab to advance humanoid robot simulation and training; Skild AI is using simulation frameworks to develop general robot intelligence, and General Robotics is integrating them into its robot intelligence platform.
The universal Blackwell system for robot developers
Global system manufacturers are building NVIDIA RTX PRO™ 6000 workstations and servers to provide a single architecture to easily run all robotics development workloads including training, synthetic data generation, robot learning, and simulation.
Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo and Supermicro announced NVIDIA RTX PRO-based servers, while Dell Technologies and Lenovo announced NVIDIA RTX PRO 6000 Blackwell-based workstations .
For those who need more compute power to run large-scale training or data generation workloads, developers can leverage NVIDIA Blackwell systems such as the GB200 NVL72, available with major cloud providers and NVIDIA cloud partner NVIDIA DGX™ Cloud , to achieve up to 18x performance gains in data processing.
Developers can deploy their robotics foundation models to the upcoming NVIDIA Jetson Thor platform to accelerate inference and runtime performance on the robot.
SOURCE: PRTimes
