NVIDIA Enables the Next Era Of Physical AI Research With Agent Skills For Autonomous Vehicles, Robotics And Vision AI
NVIDIA, Wednesday, June 3rd, 2026
NVIDIA unveils physical AI agent skills and Cosmos 3 to accelerate autonomous vehicle, robotics, and vision AI research.
At CVPR, NVIDIA introduced new physical AI agent skills designed to streamline development workflows for autonomous vehicles, robots, and vision AI systems. The company announced Cosmos 3, an open frontier model that unifies vision reasoning, world and action generation across physical AI applications.
For autonomous vehicles, new skills enable neural reconstruction of 3D scenes from fleet data and synthetic scenario generation, while AlpaGym and OmniDreams facilitate high-fidelity simulation and photorealistic rendering.
Vision AI researchers benefit from Metropolis skills for synthetic anomaly generation and visual inspection, while robotics developers can leverage Isaac Sim and Isaac Lab frameworks with agent-ready automation for policy training and evaluation across navigation and manipulation tasks.