Nvidia has unveiled a groundbreaking set of technologies aimed at accelerating the development of humanoid robots. At the forefront is Isaac GR00T N1, the world’s first fully customisable foundation model for humanoid reasoning and skill execution. Designed to address the increasing demand for automation due to global labour shortages, GR00T N1 is set to transform robotics by enabling more adaptable and intelligent machines.
This model is part of a broader initiative that includes simulation frameworks and an open-source physics engine called Newton, developed in collaboration with Google DeepMind and Disney Research. Newton is engineered to refine robotic movements with greater precision, paving the way for advanced learning in real-world environments.
“The age of generalist robotics is here,” said Jensen Huang, founder and CEO of Nvidia. “With Nvidia Isaac GR00T N1 and new data-generation and robot-learning frameworks, robotics developers everywhere will open the next frontier in the age of AI.”
GR00T N1: Mimicking Human Cognition for Smarter Robots
The GR00T N1 model is built on a dual-system architecture inspired by human cognition. It integrates two core components:
- System 1: A fast-thinking action model that mimics reflexive decision-making, optimising real-time responses.
- System 2: A slow-thinking model designed for deliberate reasoning and methodical decision-making.
Powered by a vision-language model, System 2 interprets environmental inputs and instructions, formulating action plans that System 1 translates into fluid, coordinated movements. These capabilities allow GR00T N1-powered robots to handle various tasks, including grasping, manipulating objects, and performing multi-step operations such as sorting and assembly.
As showcased in Nvidia’s GTC keynote, robotics company 1X Technologies has successfully post-trained GR00T N1 on its NEO Gamma humanoid robot, enabling it to autonomously carry out domestic tasks.
“The future of humanoids is about adaptability and learning,” said Bernt Børnich, CEO of 1X Technologies. “While we develop our own models, Nvidia’s GR00T N1 provides a significant boost to robot reasoning and skills. With minimal post-training data, we fully deployed on NEO Gamma — advancing our mission of creating robots that are not just tools, but companions capable of assisting humans in meaningful, immeasurable ways.”
Other leading robotics firms, including Agility Robotics, Boston Dynamics, Mentee Robotics, and NEURA Robotics, have also gained early access to GR00T N1 for their humanoid research and development.
Newton Physics Engine: A Leap Forward in Robotic Learning
In a move to enhance the physics-based training of humanoid robots, Nvidia, Google DeepMind, and Disney Research have collaborated on Newton, a next-generation open-source physics engine. Designed to integrate with simulation platforms like Google DeepMind’s MuJoCo and Nvidia Isaac Lab, Newton allows robots to interact with digital environments that closely replicate real-world physics.
Google DeepMind and Nvidia have also announced MuJoCo-Warp, a new framework expected to accelerate machine learning for robotics by over 70 times. Meanwhile, Disney Research is set to deploy Newton in its robotic character platform, aiming to create highly expressive, interactive robots for immersive entertainment experiences.
“The BDX droids are just the beginning,” said Kyle Laughlin, senior vice president at Walt Disney Imagineering Research & Development. “We’re committed to bringing more characters to life in ways the world hasn’t seen before, and this collaboration with Disney Research, Nvidia and Google DeepMind is a key part of that vision.”
Advancing Robotics With Synthetic Data
Real-world data collection for humanoid robots presents significant challenges due to time constraints and logistical limitations. To address this, Nvidia introduced the Isaac GR00T Blueprint, a synthetic motion generation framework that enables large-scale data creation for robotic training.
Using this technology, Nvidia generated 780,000 synthetic motion trajectories in just 11 hours, equivalent to 9 months of continuous human demonstration data. This dataset enhanced GR00T N1’s performance by 40 per cent, demonstrating the potential of synthetic data to accelerate humanoid training.
Additionally, Nvidia has made the GR00T N1 dataset publicly available as part of a broader open-source initiative, with datasets now accessible on Hugging Face.
Availability and Future Prospects
Nvidia’s GR00T N1 training data, task evaluation scenarios, and the Isaac GR00T Blueprint are now available for download on Hugging Face and GitHub. The company has also introduced the DGX Spark, a personal AI supercomputer designed to expand the capabilities of GR00T N1, enabling developers to fine-tune the model for customised robotic applications.
Meanwhile, the Newton physics engine is set to launch later this year, marking another step forward in the evolution of intelligent humanoid robots.
