NVIDIA Unveils Breakthroughs in Humanoid Robot Development
NVIDIA, a leader in accelerated computing, has announced significant advancements in humanoid robot development, including the launch of 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. These innovations are expected to revolutionize the robotics industry by enabling faster development and deployment of humanoid robots.
Key Takeaways:
- NVIDIA Isaac GR00T N1.5 is a foundation model that can adapt to new environments and workspace configurations, and can recognize objects through user instructions, significantly improving the model's success rate for common material handling and manufacturing tasks.
- NVIDIA Isaac GR00T-Dreams is a blueprint that generates vast amounts of synthetic motion data, allowing developers to teach robots new behaviors and adapt to changing environments.
- The blueprint uses a single image as input and generates videos of the robot performing new tasks in new environments, extracting action tokens that are used to teach robots how to perform these new tasks.
- NVIDIA Research used GR00T-Dreams to generate synthetic training data to develop GR00TN1.5 in just 36 hours, compared with what would have taken nearly three months of manual human data collection.
- Early adopters of GR00T N models include AeiRobot, Foxlink, Lightwheel, and NEURA Robotics, who are using the models to improve industrial robot manipulator flexibility, accelerate humanoid robot deployment, and develop household automation systems.
- NVIDIA has also unveiled new simulation technologies, including NVIDIA Cosmos Reason, a WFM that uses chain-of-thought reasoning to curate accurate, higher-quality synthetic data for physical AI model training, and NVIDIA Isaac GR00T-Mimic, a blueprint for generating exponentially large quantities of synthetic motion trajectories for robot manipulation.
- Other companies, including Foxconn, Agility Robotics, Boston Dynamics,fourier, Mentee Robotics, and XPENG Robotics, are adopting NVIDIA Isaac platform technologies to advance humanoid robot development and deployment.
- NVIDIA has partnered with global systems manufacturers to build NVIDIA RTX PRO 6000 workstations and servers, offering a single architecture to run every robot development workload across training, synthetic data generation, robot learning, and simulation.
Statistics:
- NVIDIA Isaac GR00T N models have a success rate of over 90% for common material handling and manufacturing tasks.
- The GR00T-Dreams blueprint generates up to 10,000 synthetic motion data samples in under an hour.
- NVIDIA Isaac Sim 5.0 will be openly available on GitHub, allowing developers to simulate and train their humanoid robots more efficiently.
- Over 24,000 high-quality humanoid robot motion trajectories are included in the Open-Source Physical AI Dataset.
- NVIDIA Blackwell systems can achieve up to 18x greater performance for data processing compared to traditional servers.
Sources:
- NVIDIA Corporation
- NVIDIA GTC Taipei
- GlobeNewswire (MIL-OSI)
- COMPUTEX
- NVIDIA RTX PRO 6000 workstations and servers
- NVIDIA Cloud Partners
- NVIDIA DGX Cloud
- NVIDIA Jetson Thor platform
- Agility Robotics
- Boston Dynamics
- Foxlink Group
- Lightwheel
- NEURA Robotics
- General Robotics
- Skild AI
- Universal Blackwell Systems
- Cisco
- Dell Technologies
- Hewlett Packard Enterprise
- Lenovo
- Supermicro