NVIDIA announced a comprehensive suite of services, AI models, and computing platforms aimed at accelerating the development of humanoid robots worldwide. The offerings include NVIDIA NIM Microservices for robot simulation, the NVIDIA OSMO orchestration service for managing robotics workloads, and an AI-enabled teleoperation workflow for training robots with minimal human demonstration data.
“We’re advancing the entire NVIDIA robotics stack, opening access for worldwide humanoid developers and companies to use the platforms, acceleration libraries and AI models best suited for their needs,” said NVIDIA chief Jensen Huang, who believes that the next wave of AI is Physical AI.
NIM and OSMO
NVIDIA NIM microservices offer pre-built containers powered by NVIDIA inference software, reducing deployment times from weeks to minutes. Two new microservices, MimicGen and Robocasa, enhance simulation workflows for generative physical AI in NVIDIA Isaac Sim, a robotics simulation application built on the NVIDIA Omniverse platform.
MimicGen generates synthetic motion data from teleoperated data captured by devices like Apple Vision Pro, while Robocasa creates robot tasks and simulation-ready environments in OpenUSD, a universal 3D development framework.
NVIDIA OSMO, a cloud-native managed service orchestrates complex robotics development workflows, significantly reducing deployment and development cycle times. This service allows users to manage tasks like generating synthetic data, training models, and conducting reinforcement learning and software-in-the-loop testing at scale.
Data Capture for Humanoid Robots
Training humanoid robots requires extensive data, traditionally gathered through time-consuming teleoperation. NVIDIA’s AI and Omniverse-enabled teleoperation workflow, demonstrated at the SIGGRAPH conference, allows developers to generate large amounts of synthetic data from minimal human demonstrations.
This process involves capturing teleoperated demonstrations using Apple Vision Pro, simulating them in NVIDIA Isaac Sim, and using MimicGen to create synthetic datasets. The Project GR00T humanoid foundation model is then trained using both real and synthetic data, significantly reducing development time and costs.
Fourier, a company focused on general-purpose robots, highlighted the benefits of NVIDIA’s technology. “Developing humanoid robots is extremely complex, requiring an incredible amount of real data, tediously captured from the real world,” said Alex Gu, CEO of Fourier. “NVIDIA’s new simulation and generative AI developer tools will help bootstrap and accelerate our model development workflows.”
Powering Humanoids
Through a new NVIDIA Humanoid Robot Developer Program, developers can gain early access to the new offerings as well as the latest releases of NVIDIA Isaac Sim, which is built on Omniverse.
The program has already attracted some of the biggest names in robotics including Boston Dynamics, 1X, ByteDance Research, and Neura Robotics, and others.
“Boston Dynamics and NVIDIA have a long history of close collaboration to push the boundaries of what’s possible in robotics. We’re really excited to see the fruits of this work accelerating the industry at large, and the early-access program is a fantastic way to access best-in-class technology,” said Aaron Saunders, chief technology officer of Boston Dynamics.
With the rise in humanoid projects and major companies investing big on the same, NVIDIA is placing itself as an indispensable part in powering these projects.
