Humanoid Robots
Accelerate the development of advanced AI robotics.
Workloads
Robotics
Simulation / Modeling / Design
Industries
Hardware/Semiconductor
Healthcare and Life Sciences
Manufacturing
Retail/ Consumer Packaged Goods
Smart Cities/Spaces
Business Goal
Innovation
Return on Investment
Products
NVIDIA Isaac Lab
NVIDIA Isaac Sim
NVIDIA Isaac GROOT
NVIDIA Jetson Thor
NVIDIA Omniverse
Overview
The Next Era of Physical AI
General-purpose humanoid robots are built to quickly adapt to existing human-centric urban and industrial work spaces, helping tackle tedious, repetitive, or physically demanding tasks.
These robots are finding their way from factory floors to healthcare facilities, where they’re assisting humans and alleviating labor shortages with automation.
However, building humanoid robots presents layers of complexities and engineering challenges. These include replicating human-like perception, degrees of freedom, dexterity, mobility, cognition, and whole-body control.
This demands accelerated progress in robotics research fields and technologies, including artificial intelligence, machine learning, physics-based simulation, sensor technologies, embedded computing, and mechatronics.
Figure
Quick Links
Technical Implementation
Advancing Humanoid Robot Development
NVIDIA is developing accelerated systems, blueprints, tools, services, algorithms, and other robot technologies that can be used to build general-purpose, human-form-factor robots.
Three-Computer Solution
Humanoid robots need to sense, plan, and act autonomously within a given environment, which involves processing large amounts of data in real time. This requires training foundation models that power the robot brain, simulating and validating the robot brain, and finally deploying these brains and associated software onto the actual robot.
The three AI systems are:
- An NVIDIA AI supercomputer—an NVIDIA DGX™ with NVIDIA H100 or B100 processors—for training powerful, generative physical AI models and robotic foundation models.
- An NVIDIA OVX™ computer with NVIDIA RTX™ GPUs for synthetic data generation, robot learning, and testing in simulation, with simulation frameworks such as NVIDIA Isaac Sim™ and Isaac™ Lab built on top of NVIDIA Omniverse™.
- An on-robot runtime computer, such as NVIDIA Jetson Thor™—built on NVIDIA’s Blackwell architecture—that can run low-latency, high-throughput inference.
NVIDIA Isaac GR00T
GR00T is a research initiative and development platform for general-purpose robot foundation models and data pipelines, to accelerate humanoid robotics.
Robot Foundation Models
Humanoid robots need diverse skills for varied tasks, traditionally requiring separate, costly AI models. Robot foundation models solve this by training on broad data, developing generalizable skills. This allows robots to adapt to different tasks and environments.
NVIDIA Isaac GR00T N is the world’s first open foundation model for generalized humanoid robot reasoning and skills. This cross-embodiment model takes multimodal input, including language and images, to perform manipulation tasks in diverse environments. GR00T N was trained on an expansive humanoid dataset, consisting of real captured data, synthetic data generated using the components of the NVIDIA Isaac GR00T-Mimic blueprint and internet-scale video data. It is adaptable through post-training for specific embodiments, tasks and environments.
Robot Learning and Simulation Frameworks
Simulation is key for developers to train humanoid robots across a variety of physically accurate environments and conditions, before deploying them in the real world.
Robot learning and simulation frameworks like NVIDIA Isaac Sim and Isaac Lab—built on the Omniverse platform—enable physically accurate simulations for training and validating multiple humanoid robot agents in parallel.
Isaac Lab is an open-source unified robot learning framework built on Isaac Sim that can be used to apply these learning techniques to train a robot policy. The trained robot policies can then be validated in Isaac Sim, a reference application for building, simulating, and testing humanoids in physically based virtual environments.
GR00T Workflows
Data Generation and Processing
Gathering extensive, high-quality, real-world datasets for this can be challenging, costly, and time-consuming. Synthetic data, generated from physically accurate simulations, addresses this challenge by accelerating data collection and providing the diversity needed to generalize robot learning models.
The NVIDIA Isaac GR00T-Dreams blueprint generates vast amounts of synthetic trajectory data using NVIDIA Cosmos™ world foundation models, prompted by a single image and language instructions. This enables robots to learn new tasks in unfamiliar environments without needing specific teleoperation data.
The GR00T-Dreams blueprint complements the GR00T-Mimic blueprint. While GR00T-Mimic uses Omniverse and Cosmos to augment existing data, GR00T-Dreams uses Cosmos to generate entirely new data.
Dextrous Manipulation
Humanoid robot grasping functionality requires human-like dexterous object manipulation skills, capable of performing both gross and fine-grained manipulation tasks. GR00T-Dexterity is a comprehensive suite of models and policies built using a reinforcement learning-based approach, combined with reference workflows, to enable the development of these advanced capabilities.
Mobility
General-purpose navigation in complex and dynamic environments requires extensive tuning. With the GR00T-Mobility reference workflow, you can create a mobility generalist for navigating across varied settings and robot embodiments.
Whole-Body Control
Achieving whole-body control in humanoid robots is challenging, demanding both stable manipulation and robust locomotion. GR00T-Control addresses this with a suite of advanced motion planning and control models, policies, and reference workflows, streamlining the development of effective control systems.
By using imitation learning and teleoperated datasets, GR00T-Control facilitates training for robust, whole-body motion policies, enabling humanoid robots to learn dexterous manipulation and locomotion skills.
AI-Based Robot Perception
To improve situational awareness and interaction efficiency, humanoid robots require long-term memory for events, spaces, personalized settings, and context-aware responses.
GR00T-Perception enables this with a robust suite of perception libraries, foundation models, and reference workflows built on Isaac Sim and Isaac ROS. These tools integrate advanced technologies like vision-language models and retrieval-augmented memory to enhance perception, cognition, and adaptability in humanoid robots.
The Next-Generation, On-Robot Computing Platform
Robot hardware is also crucial for running an ensemble of multimodal AI models that power humanoids with the right performance, latency, and functional safety in diverse conditions.
NVIDIA Jetson AGX Thor, based on NVIDIA’s Blackwell GPU architecture, delivers ultra-high-performance AI compute and a new transformer engine. This delivers the necessary AI superpower at the edge to enable the new generation of humanoids.
Quick Links
Ecosystem
Get started with our humanoid robotics partners.
Get Started
Develop Humanoid Robots
Advance your humanoid robot development with GR00T foundational technologies by accessing tutorials, forums, release notes, and comprehensive documentation.
Accelerate Physical AI Workloads With NVIDIA RTX PRO 6000 Blackwell Series GPUs
NVIDIA RTX PRO™ 6000 Blackwell Series GPUs accelerate physical AI by running every robot development workload across training, synthetic data generation, robot learning, and simulation.