FANUC: open platform for physical AI (ROS 2, NVIDIA)

2025-12-03 · Publisher : FANUC · Domain : Robotics

At iREX 2025, FANUC opens up its robots: an open-source ROS 2 driver, standard Python execution and a collaboration with NVIDIA (Jetson, Isaac Sim) to ease physical AI across its range.

### Background At the iREX International Robot Exhibition in Tokyo (December 3-6, 2025), FANUC, Japan's leading industrial robot maker, announced it is opening its platforms to physical AI. Long known for proprietary environments, the group released an open-source ROS 2 driver and teamed up with NVIDIA. ### Specifications - ROS 2 driver for FANUC robots, released as open source on GitHub and compatible with the ros2_control framework. - Very fast real-time control at 1 ms, and high-speed external command input. - Covers the whole range, from 3 kg payload robots up to 2.3-ton models, including CRX cobots. - Standard Python execution on the robots. - NVIDIA collaboration: onboard Jetson modules, photorealistic simulation in Isaac Sim, Omniverse libraries, OpenUSD SimReady assets of FANUC robots and a link with the ROBOGUIDE simulator. - iREX demos: voice recognition, obstacle avoidance, multi-arm coordination, dynamic part tracking. ### Why it matters Physical AI is developing mostly in the ROS and Python ecosystem. By opening up to it, FANUC avoids being bypassed by startups training manipulation models on more open arms. ### Key takeaways Approximate date: the announcement accompanies iREX, December 3-6, 2025; this entry uses the show's first day.

Tags : Industriel, ROS 2, IA physique, NVIDIA, Japon