Nvidia unveils open humanoid robot for university research

NVIDIA has introduced the Isaac GR00T Reference Humanoid Robot for academic research, positioning it as an open reference platform for humanoid robotics.

The system combines a Unitree H2 Plus humanoid body, Sharpa five-fingered hands, Jetson Thor onboard computing, and NVIDIA's Isaac GR00T software stack. It is intended to give research teams a single platform for hardware integration, data capture, simulation, training, evaluation, and deployment.

Several research institutions plan to use the system, including Ai2, ETH Zurich, Stanford Robotics Centre, and the Advanced Robotics and Controls Labouratory at the University of California San Diego. That places the design directly in university and non-profit research environments focused on general-purpose robotics.

The reference robot is built around a human-scale chassis. The Unitree H2 body stands nearly 6ft tall, weighs 150lb, and has 31 degrees of freedom across the body.

Dexterous manipulation is a central part of the design. Dual Sharpa Wave tactile five-finger hands bring the machine to 75 degrees of freedom across the body and hands, giving researchers a platform to test fine manipulation alongside locomotion.

Sensors include a head-mounted stereo camera, wrist cameras, and an inertial measurement unit for motion tracking. The robot also includes microphones and speakers for voice interaction, along with Ethernet, Wi-Fi 6, Bluetooth 5.2, and USB connectivity.

Onboard computing comes from the Jetson AGX Thor T5000, based on NVIDIA's Blackwell architecture. The module includes a 14-core Arm CPU, 128GB of unified memory, and a configurable power range of 40 to 130 watts for on-robot inference and control.

The machine also includes a battery rated at 15Ah and 0.972kWh, with about three hours of operating life, as well as a remote emergency stop. Arm torque reaches 120 Newton-metres and leg torque reaches 360 Newton-metres, with a rated arm payload of 7kg and a peak payload of 15kg.

Open stack

The wider strategy behind the launch centres on NVIDIA's Isaac GR00T software environment. The platform spans teleoperation for collecting demonstration data, foundation models for humanoid reasoning and learning, simulation tools through Isaac Sim and Isaac Lab, and Isaac ROS middleware for transferring trained policies to physical robots.

Researchers using the software retain control over robot data, training data, telemetry, and logs. Teams can also adopt either the full development stack or selected parts of it within existing workflows.

NVIDIA is also extending the software approach beyond the new reference robot. The Isaac GR00T developer platform will support Unitree's G1 humanoid robot, which is already used in research settings.

The move reflects a broader effort to push NVIDIA's computing hardware and robotics software deeper into physical AI research. By offering a defined hardware and software combination rather than separate components, the company aims to reduce the custom integration work that often slows robotics development.

Jensen Huang, Founder and Chief Executive Officer of NVIDIA, framed the launch as a way to broaden access for researchers. "Humanoid robots will bring physical AI to the world's largest industries, opening a multitrillion-dollar economic opportunity," said Huang. "The NVIDIA Isaac GR00T Reference Humanoid Robot gives researchers a single, open platform to make breakthrough discoveries toward general-purpose physical intelligence."

Research users

The institutions named by NVIDIA pointed to the value of a common platform that can be shared across research groups.

"Robotics moves fastest when researchers can build on open platforms, share code, and test ideas on real machines," said Steve Cousins, Executive Director of Stanford Robotics Centre. "The NVIDIA Isaac GR00T Reference Robot gives our students and collaborators an open humanoid reference design with dexterous hands, onboard AI compute, and the NVIDIA Isaac GR00T development platform for creating, comparing, and sharing robot behaviors on physical hardware."

ETH Zurich also highlighted the role of physical testing in robotics research. "ETH Zurich's robotics research aims to advance machines that can move, perceive, and manipulate reliably in the real world," said Marco Hutter, Professor at ETH Zurich's Robotic Systems Lab. "The NVIDIA Isaac GR00T reference design gives our teams a state-of-the-art humanoid platform for collecting data, testing algorithms, and validating robot behaviors with the NVIDIA Isaac GR00T development platform."

Other researchers focused on accessibility and open science. "To make progress toward general-purpose robots, researchers need platforms that are both capable and broadly accessible," said Deepak Pathak, Cofounder and Chief Executive Officer of Skild AI. "A reference design lets more researchers participate in frontier humanoid research and move from ideas to experiments faster. This helps push the whole robotics research ecosystem forward."

"At Ai2, our mission is to accelerate robotics through open science," said Dieter Fox, Senior Research Director at Ai2 and Professor at the University of Washington. "The NVIDIA Isaac GR00T Reference Robot, built on NVIDIA's open technologies, provides our researchers with the hardware and software components necessary to continue our work in broadly competent robotics."

The University of California San Diego linked the design to work on mobile manipulation. "Advancing robotics research for real-world problems requires humanoids that can move, interact, and manipulate with precision in dynamic environments," said Michael Yip, Professor at UC San Diego and Director of the Advanced Robotics and Controls Labouratory. "An integrated platform that connects robot hardware, data capture, policy learning, and physical evaluation can help researchers accelerate loco-manipulation research and develop more useful real-world systems."