With GR00T, NVIDIA wants to make it easier to create humanoid robots

During the opening of the GTC 2024 in San Jose, United States, the NVIDIA made an announcement that promises to be a watershed in the world of humanoid robotics: the launch of Project GR00T.

With the aim of boosting the development of these areas, Project GR00T represents an ambitious effort by the company to integrate and improve the capabilities of humanoid machines, allowing them to perform tasks with sophistication and intelligence.

How the GR00t project works

With Project GR00T, humanoid robots will be able to understand natural language and imitate human movements, acquiring coordination skills, dexterity and adaptation to the real world quickly and effectively.

In support of this development, the NVIDIA is collaborating with robotics industry leaders such as 1X Technologies, Agility Robotics, Apptronic, Boston Dynamics, between others. Jonathan Hurst, co-founder and chief robotics officer at Agility Robotics, expressed his excitement about the partnership, highlighting the potential of modern AI to accelerate the development of robots useful in everyday life.

Together with Project GR00T, the Jetson Thor, a new computer developed specifically for humanoid robots, based on the NVIDIA Thor system-on-a-chip (SoC). This device stands out for its modular architecture, which combines improved performance, energy efficiency and compact design.

Main updates on the Isaac Platform

The NVIDIA Isaac™ robotics platform also received major updates, introducing generative AI models and advanced tools for simulation and AI workflow infrastructure. O Isaac Manipulator and the Isaac Perceptor, two new solutions comprised of pre-trained models, libraries, and reference hardware, were revealed as part of the innovations.

Jensen Huang, founder and CEO of NVIDIA, highlighted the importance of this advancement: “We are tackling one of the most exciting challenges in modern AI by building foundation models for general humanoid robots.” According to him, emerging technologies are enabling major advances in general artificial robotics.

The Isaac platform, recognized for its flexibility and power, now includes the Isaac Lab for reinforcement learning and the OSMO, an advanced compute orchestration service. Isaac Lab, a GPU-accelerated application optimized for Isaac Sim, is capable of running thousands of parallel simulations for training robots using real and synthetic data.

The OSMO service plays a crucial role in scaling robot development by coordinating data generation, model training, and software and hardware workflows in distributed environments.

Main features and resources

O Isaac Manipulator is a solution for robotic arms, offering advanced dexterity and modular AI capabilities, promising to accelerate trajectory planning by up to 80 times and increase efficiency in robotic tasks. Companies like Yaskawa and Universal Robots are some of the first to adopt this technology.

O Isaac Perceptor, focused on multi-camera 3D surround vision, is increasingly used in autonomous mobile robots to improve worker efficiency and safety, as well as reduce costs and error rates in manufacturing and fulfillment operations. ArcBest, BYD and KION Group are some of the companies that are at the forefront of adopting this technology.

NVIDIA expects new features on the Isaac platform to be available in the next quarter, marking a revolutionary step in the field of artificial intelligence and robotics.

These updates and new releases position NVIDIA at the forefront of robotics development, combining multimodal AI with practical robotics capabilities. These advances are expected to not only transform the robotics industry but also open up new possibilities for practical applications in a variety of sectors, from home care to complex industrial operations.

There is clear excitement around this innovation, both in the technology industry and in the scientific and academic communities. NVIDIA's collaboration with industry leaders promises to further accelerate the development and implementation of advanced robotic solutions, leading to a future where human-robot interaction becomes increasingly integrated and beneficial to all aspects of society.

See also other features

Sources: NVIDIA, Venture Beat, Techcrunch

reviewed by Glaucon Vital in 19 / 3 / 24.