NVIDIA unveils Project GR00T & Jetson Thor for humanoid robots

NVIDIA has unveiled Project GR00T, a pivotal foundation model for humanoid robots aimed at spurring advancements in robotics and embodied artificial intelligence (AI). The tech giant also debuted Jetson Thor, a state-of-the-art computer specifically designed for humanoid robots. These developments form part of a comprehensive plan to enhance the NVIDIA Isaac robotics platform, bringing forth generative AI foundation models as well as vital equipment for simulation and AI workflow infrastructure.

"Building foundation models for general humanoid robots is one of the most exciting problems to solve in AI today," said Jensen Huang, founder and CEO of NVIDIA. "The enabling technologies are coming together for leading roboticists around the world to take giant leaps towards artificial general robotics."

The GR00T-empowered robots have been programmed to understand natural language and mimic human actions, learning skills from coordination to dexterity effectively. By observing human interactions, these robots are quickly becoming adept at negotiating, adapting to, and interacting with the real world. Huang gave a demonstration of several GR00T robots executing various tasks in his GTC keynote address.

Jetson Thor, engineered as a new computing platform, is capable of complex task performances and natural and safe interaction with humans and other machines. It has a modular architecture optimised for power, performance and size. Its impressive features include an 800 teraflops 8-bit floating point AI performance for running multimodal generative AI models like GR00T, thanks to the next-gen GPU based on the NVIDIA Blackwell architecture and a transformative engine.

Companies including but not limited to 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Figure AI and XPENG Robotics are among the myriad entities for which NVIDIA is creating a comprehensive AI platform. "We are at an inflection point in history, with human-centric robots like Digit poised to change labour forever," said Jonathan Hurst, co-founder and chief robot officer at Agility Robotics. "We're excited to partner with NVIDIA to invest in the computing, simulation tools, machine learning environments and other necessary infrastructure to enable the dream of robots being a part of daily life."

The Isaac tools used by GR00T have been able to create fresh foundation models for a multitude of robot embodiments in an array of environments. With advanced models requiring vast volumes of real and synthetic data, the newly announced Isaac Lab, an application built on Isaac Sim, facilitates thousands of parallel simulations for robot learning. The NVIDIA updates also announced the launch of the Isaac Manipulator and Isaac Perceptor, an ensemble of pre-trained robot model libraries, and reference hardware.

Aiming to better manufacturing and operations efficiency, safety, and reducing error rates and costs, NVIDIA is collaborating with companies like ArcBest, BYD, and KION Group to develop advanced visual AI capabilities. The improvements to the Isaac platform are anticipated to be available in the upcoming quarter.