NVIDIA Brings Agile Robots Humanoid System Into Isaac Platform

NVIDIA expanded its Isaac robotics platform Oct. 21 with an Agile Robots humanoid reference system and new learning, simulation and runtime tools, giving developers resources to build and test software before deploying it on physical robots.

The reference machine, Agile ONE SN, pairs NVIDIA’s Isaac GR00T system with its Jetson Thor computer. The platform is scheduled to become available in early 2027. NVIDIA also plans to integrate Halos, its robotics safety technology, into the system. The planned integration builds on the company’s Halos safety work for humanoid and industrial robots.

Agile ONE SN is based on the Agile ONE S humanoid, which stands 1.40 meters tall and weighs 40 kilograms. It has 31 degrees of freedom, and each arm can carry up to 3 kilograms. Two five-finger WUJI Hand 2 units provide 20 degrees of freedom apiece, tactile sensing and force control at 1 kHz. The system also includes head and wrist cameras, lidar and inertial measurement units. Its Jetson Thor computer provides 2,070 FP4 teraflops of compute and 128 GB of unified memory.

The software additions cover learning and operation. Isaac Lab 3.0 is generally available as an open-source framework for robot learning. Isaac OS uses Holoscan as its foundation and is designed to support deterministic, real-time behavior as workloads move from simulation to physical robots.

NVIDIA also added SimReady Central Library, a catalog of more than 2,000 digital assets for simulating robotics and factory environments. Human 1M Dataset includes diverse human activities, video synchronized across three views and 3D tracking of hand movements. The library and dataset are available on Hugging Face, while Cosmos Coalition partners can access the complete dataset.