NVIDIA Expands Isaac Robotics Stack With New Runtime, Training Assets and Humanoid Platform
NVIDIA announced an Isaac platform expansion spanning an open runtime, simulation assets, robot-learning tools and an Agile Robots humanoid reference system.
NVIDIA announced an expansion of its Isaac robotics platform on October 21, adding an open-source runtime, new training assets, updated robot-learning tools and a humanoid reference platform from Agile Robots. The package is designed to connect data generation, simulation, policy training and deployment for general-purpose robots.
Isaac Lab 3.0 is now generally available, according to NVIDIA, with multi-GPU workflows and support for Newton and PhysX physics engines. NVIDIA also introduced Isaac OS, an open-source runtime based on Holoscan for connecting AI, sensors, controls and robot hardware. New assets include a library of more than 2,000 simulation-ready robotics and factory objects and the Human 1M dataset for training and evaluation.
Agile Robots separately announced Agile ONE SN, a humanoid reference system built on Isaac GR00T and Jetson Thor. NVIDIA says the 1.4-meter platform is scheduled for availability in early 2027. That timing is a vendor roadmap, not a current commercial deployment, and planned safety-stack integration remains prospective.
Official source: NVIDIA newsroom release, October 21, 2026. Source reviewed October 24, 2026.