Nvidia unveiled a comprehensive hardware and software framework at its annual developer conference aimed at advancing humanoid robotics with the integration of generative artificial intelligence (AI).
Generative AI – the use scenarios from Nvidia.
According to the official information, this new platform utilizes both robust computing hardware to empower the robots and sophisticated software packages, including generative AI, tailored for crafting lifelike robots.
The incorporation of generative artificial intelligence technology equips humanoid robots with the capability to discern and execute actions based on a blend of inputs such as language, video data, human demonstrations, and past experiences.
Dubbed Project GR00T, these components intended for the advancement of robotic control systems complement Nvidia’s existing suite of hardware and software resources dedicated to robotics.
Rev Lebaredian, Vice President of Omniverse and Simulation Technology, highlighted that such enhanced robots, endowed with greater intelligence and agility, in the future will be deployed across heavy industries worldwide. Nvidia aims to collaborate with the global robotics and simulation community to expedite the development and adoption of these advanced technologies.
The backbone of the new generative AI software is a formidable computing system dubbed “Jetson Thor,” which uses Nvidia’s cutting-edge AI chips and complementary hardware components. This powerhouse computer augments the robot with ample computational prowess to execute intricate tasks and seamlessly interact with both humans and machinery, Nvidia emphasized.
Moreover, the suite of software tools, collectively referred to as the “Isaac” platform, offers improved versatility, allowing to adapt the platform to various robotic embodiments in different environments. These tools encompass capabilities such as training the software to enhance decision-making through reinforcement learning, thereby facilitating continuous improvement over time.
Nvidia also revealed plans to roll out pre-trained robotics models and additional software enhancements aimed at refining robotic arm functionalities and bolstering multi-camera sensing capabilities.
Written by Alius Noreika