scorecardresearch
Clear all
Search

COMPANIES

No Data Found

NEWS

No Data Found
Sign in Subscribe
NVIDIA Project GR00T: CEO Jensen Huang unveils tools to power $38 billion humanoid robot market at CES 2025

NVIDIA Project GR00T: CEO Jensen Huang unveils tools to power $38 billion humanoid robot market at CES 2025

NVIDIA is revolutionising humanoid robotics with cutting-edge tools for faster development.

NVIDIA NVIDIA

As the market for humanoid robots is projected to soar to $38 billion over the next two decades, NVIDIA is stepping up to accelerate innovation in this transformative field. At CES 2025, NVIDIA CEO Jensen Huang unveiled a suite of tools, including robot foundation models, data pipelines, and simulation frameworks, to advance humanoid robot development, particularly for industrial and manufacturing applications.

At the core of NVIDIA’s announcement is the Isaac GR00T Blueprint, designed to simplify and scale the creation of synthetic motion datasets for humanoid robots. This groundbreaking tool enables robots to learn through imitation learning, a technique where machines acquire skills by mimicking expert human behaviour.

“Collecting high-quality datasets in the real world can be time-intensive, expensive, and tedious. Isaac GR00T bridges that gap by exponentially expanding datasets with minimal input,” said Huang.

The blueprint is built around three workflows:

    1.    GR00T-Teleop: Utilises Apple Vision Pro to record human actions, creating a digital twin for simulations.
    2.    GR00T-Mimic: Multiplies captured actions into larger datasets for training.
    3.    GR00T-Gen: Expands datasets further using NVIDIA Omniverse and NVIDIA Cosmos, incorporating domain randomisation and 3D upscaling.

These datasets train robots using NVIDIA’s Isaac Lab, an open-source modular framework for developing safe and effective robot policies.

NVIDIA also launched Cosmos, a platform featuring pretrained world foundation models designed to generate physics-aware simulations and world states for physical AI. These models, trained on 18 quadrillion tokens, integrate data from autonomous driving, robotics, drone footage, and synthetic sources.

Cosmos minimises the “simulation-to-reality gap” by upscaling 3D simulations into realistic visuals while reducing errors commonly associated with hallucinations in world models. Its integration with Omniverse, NVIDIA’s 3D development platform, ensures accuracy and control in simulation environments.

NVIDIA’s innovations are already making waves, with robotics leaders like Boston Dynamics and Figure adopting the Isaac GR00T framework. NVIDIA also announced an early access program for humanoid robot software and hardware manufacturers to leverage its advanced tools.

For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine

Published on: Jan 07, 2025, 11:02 AM IST
×
Advertisement