NVIDIA Highlights New Tools and Models for Physical AI and Robotics

This National Robotics Week, NVIDIA is shining a spotlight on significant breakthroughs in physical AI that are rapidly transforming the world of robotics. These advancements are fueling a new generation of intelligent machines capable of perceiving, reasoning, and acting in complex real-world environments. From agriculture and manufacturing to energy and healthcare, these innovations are accelerating robot development and deployment across various industries.

At the recent NVIDIA GTC conference, a suite of powerful new technologies was unveiled, designed to streamline the entire robot development workflow. These tools span simulation, robot learning, and foundation models, working together to enable developers to build and deploy sophisticated AI-powered robots faster than ever before. For a deeper dive into these announcements, you can visit the National Robotics Week Blog.

Unleashing Advanced Robot Intelligence

A cornerstone of NVIDIA's latest offerings includes groundbreaking foundation models that empower robots with more intuitive intelligence. The new NVIDIA Isaac GR00T open models allow robots to understand natural language instructions, enabling them to perform complex, multistep tasks through advanced vision-language-action reasoning. This means robots can go beyond pre-programmed routines to interpret human commands more flexibly.

Complementing this, the new NVIDIA Cosmos world models are designed to generate high-quality synthetic data, crucial for training robots at scale. These models significantly improve learning efficiency and generalization, helping robots adapt to diverse environments and situations more effectively. Further enhancing robot interaction, NVIDIA NemoClaw has been integrated with NVIDIA Isaac Sim, allowing developers to control autonomous robots like the Nova Carter using natural language commands, eliminating the need for manual coding. Simply instruct the robot with phrases like "move two meters forward," and NemoClaw translates it into executable Python scripts in real time within the simulation.

Enhanced Simulation and Real-World Readiness

To ensure robots are truly ready for deployment, NVIDIA has expanded its simulation capabilities. The general availability of NVIDIA Isaac Sim 6.0, alongside Isaac Lab 3.0 and Omniverse NuRec technologies, provides developers with robust tools to model real-world scenarios and rigorously validate robotic systems in a virtual environment before they ever touch the physical world. This comprehensive simulation suite allows for safer testing and faster iteration.

Adding to this, the open-source physics engine Newton 1.0 is now generally available. Newton 1.0 offers a fast and reliable foundation for dexterous robot manipulation, featuring accurate collision detection, realistic object contact, and stable simulation for complex systems involving both rigid and flexible parts. The impact of these tools is already visible: PeritasAI is integrating physical AI into surgical robotics using NVIDIA Isaac for Healthcare and the Rheo blueprint in collaboration with leading hospitals, while the University of Michigan's OceanSim, a GPU-accelerated, high-fidelity simulator, is using NVIDIA Isaac Sim and Omniverse libraries to develop advanced perception systems for underwater robots.

Read more: https://blogs.nvidia.com/blog/national-robotics-week-2026 to explore how these innovations are shaping the future of robotics.