Semiconductor company NVIDIA has introduced new open-source technologies to advance robotics research and development.
NVIDIA announced the release of the Newton Physics Engine in NVIDIA Isaac Lab, along with the open Isaac GR00T N1.6 model, which combines reasoning, vision, and language capabilities for robot skills. These tools are supported by new AI infrastructure designed to help developers test, train, and transfer robot skills from simulation to the real world.
“Humanoids are the next frontier of physical AI, requiring the ability to reason, adapt and act safely in an unpredictable world,” said Rev Lebaredian, VP of Omniverse and simulation technology at NVIDIA.
Lebaredian said the updates give developers “the three computers to bring robots from research into everyday life, with Isaac GR00T serving as the robot’s brains, Newton simulating their body and NVIDIA Omniverse as their training ground.”
Robotics development often starts in simulation, where robots can be trained in controlled and repeatable environments. However, humanoid robots, with their joints, balance, and complex movements, place heavy demands on physics engines. NVIDIA said Newton, a GPU-accelerated physics engine now in beta release, is built to handle those challenges.
Developed with contributions from Google DeepMind, Disney Research, and NVIDIA, and managed by the Linux Foundation, Newton is based on NVIDIA Warp and OpenUSD frameworks. Its design allows for detailed simulations of robot actions such as walking on uneven terrain or handling delicate objects. The engine has already been adopted by ETH Zurich’s Robotic Systems Lab, the Technical University of Munich, Peking University, robotics company Lightwheel, and simulation platform Style3D.
Alongside Newton, NVIDIA introduced Isaac GR00T N1.6, a foundation model designed to improve robot reasoning. The model integrates NVIDIA Cosmos Reason, a reasoning vision-language model built for physical AI. Cosmos Reason turns vague human instructions into step-by-step actions by applying prior knowledge, common sense, and physics. It can also organize and label large datasets for training AI models.
NVIDIA said Isaac GR00T N1.6 now allows humanoid robots to move their torso and arms more freely, enabling them to perform more complex tasks such as opening heavy doors. Developers can further refine GR00T models using the open-source NVIDIA Physical AI Dataset on Hugging Face, which contains millions of real and synthetic examples.