NVIDIA introduces new physical AI models and frameworks at CES 2026 to advance humanoid and autonomous vehicle development. OpenUSD provides a common framework for sharing 3D data across tools, while NVIDIA Omniverse serves as a source of ground-truth simulation for the stack.

At CES 2026, NVIDIA showcased physical AI tools ranging from heavy equipment to social robots, tapping into NVIDIA Cosmos world models and Isaac technologies. Caterpillar’s Cat AI Assistant offers natural language interaction in heavy vehicles, while LEM Surgical’s Dynamis Robotic Surgical System uses NVIDIA Jetson AGX Thor for enhanced precision in spinal procedures.

NEURA Robotics is building cognitive robots on a full NVIDIA stack, training in OpenUSD-based digital twins before deployment. AgiBot uses NVIDIA Cosmos Predict 2 for action-conditioned videos, while Intbot utilizes Cosmos Reason 2 for social robots to identify social cues beyond scripted tasks.

NVIDIA introduces Agile, an Isaac Lab-based engine for humanoid loco-manipulation, enabling reinforcement learning policy training on platforms like Unitree G1 and LimX Dynamics TRON. Hugging Face integrates NVIDIA Isaac GR00T N models and simulation frameworks into the LeRobot ecosystem for streamlined policy training and evaluation.

ROBOTIS develops an open source sim-to-real pipeline using NVIDIA Isaac technologies for smart servos and industrial actuators. Resources are available to learn more about OpenUSD and robotics development through technical blogs, video tutorials, and self-paced learning paths. Join the Cosmos Cookoff challenge to explore hands-on physical AI applications.

Read more at NVIDIA: Into the Omniverse: Physical AI Open Models and Frameworks Advance Robots and Autonomous Systems