en.Wedoany.com Reported - NVIDIA has introduced the NVIDIA Isaac GR00T reference humanoid robot, built on Jetson Thor hardware and the Isaac GR00T development platform as an open-source reference design. It aims to provide researchers with a shared hardware and software platform for humanoid robot development, reducing reliance on proprietary systems.
As research into general-purpose humanoid robots increases, the development process typically requires completing multiple independent steps such as hardware integration, data collection, simulation, training, evaluation, and deployment. This reference design uses the Unitree H2 Plus humanoid robot and Sharpa Wave tactile five-fingered hand as the hardware platform, equipped with the NVIDIA Jetson Thor onboard computing unit and Isaac GR00T software as the computing and development stack, aiming to provide a unified platform for robot startup, skill development, and real-world testing.
The NVIDIA Isaac GR00T reference humanoid robot integrates the humanoid robot body, manipulation, perception, control, and onboard AI computing. The reference design adopts the Unitree H2 humanoid chassis, standing nearly 6 feet (approximately 1.83 meters) tall, weighing 150 pounds (approximately 68 kilograms), with 31 degrees of freedom across the body. It is equipped with dual Sharpa Wave tactile five-fingered hands, adding 22 degrees of freedom, bringing the total degrees of freedom for the body and hands to 75. The multi-view perception system includes a head-mounted stereo camera (horizontal field of view 140 degrees, vertical field of view 102 degrees), wrist cameras for close-range manipulation, and an inertial measurement unit for motion tracking. For full-body control, arm torque reaches up to 120 N·m, leg torque up to 360 N·m, with a rated arm payload of 7 kilograms and a peak payload of 15 kilograms.
Onboard computing uses the NVIDIA Jetson AGX Thor T5000, equipped with an NVIDIA Blackwell GPU, delivering 2,070 FP4 TFLOPS of AI performance, a 14-core Arm CPU, and 128GB of unified memory, with a configurable power range of 40 to 130 watts for onboard processing and inference. Connectivity supports Ethernet, Wi-Fi 6, Bluetooth 5.2, and USB, along with a microphone and speaker for voice interaction. A 15 Ah, 0.972 kWh battery provides approximately 3 hours of runtime. The design also includes a remote emergency stop function.
NVIDIA's Isaac GR00T software platform provides tools for simulation, training, evaluation, and deployment, allowing researchers to manage robot data, training data, telemetry data, and logs. The platform includes NVIDIA Isaac Teleop for capturing robot demonstration data; NVIDIA Isaac GR00T open foundation models for reasoning and learning in humanoid robots; NVIDIA Isaac Sim and Isaac Lab for simulation, training, and testing; NVIDIA Isaac ROS middleware for policy transfer; and NVIDIA Jetson Thor for real-time inference and control. This modular design allows teams to use the full stack or integrate parts of it into existing workflows. The NVIDIA Isaac GR00T developer platform will also support the Unitree G1 humanoid robot.
Research institutions including the Allen Institute for AI (Ai2), ETH Zurich, Stanford Robotics Center, and UC San Diego’s Advanced Robotics and Controls Laboratory plan to use this reference design. NVIDIA Research also plans to use the platform in its work. The NVIDIA Isaac GR00T reference humanoid robot is expected to be available from Unitree starting in late 2026. The reference workflow for the Unitree G1 is expected to be made available to developers on GitHub and Hugging Face.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
