Nvidia charges ahead with humanoid robotics aided by the cloud

Nvidia said he had competed in advance Humanoid Robotics More black color systems to prepare a plan to create a specific foundation model, synthetic movement data to create humanoid fashion data and accelerate the development of humanoid robot.

2025 Trade Shows in Taiwan in Taiwan 2025 Trade Shows, NVIDIA ISAAC Gr00t N1.5, NVIDIA’s first update for open, generalized, customizable, humanoid thinking and skills; NVIDIA ISAAC GR00T-DREAMS, prepare a plan to create synthetic movement information; And NVIDIA Blackwell systems, to accelerate the development of humanoid robot.

The creators of humanoid and robotics are agility Robotics, Boston Dynamics, Foxlink, Galbot, Mentee Robotototics, Galbot, Mentee Robotics, Skill AI and XPeng Robot, SKILID AI and XPeng Robot, Humanoid Robot Development and Placement Technologies.

“Physical AI and robototics will inform the next industrial revolution,” he said. “AI brain gives construction blocks for each stage of the NVIDIA, robotics development of NVIDIA, to practice the simulated worlds for robots or for AI supercomputers.”

New Isaac Gr00t The Information Generation Closes BluePrint data gap

The Huang’s Compex was displayed in the Compex Rice, a plan that helps Nvidia create a large amount of synthetic movement information in Dreams – Aka Neural trajectors, physical AI developers will be able to teach new behaviors to adapt to variable environments.

Developers can first predict world background fondown models (WFMS) for robots. Then, using an image such as input, GR00T-Dreams creates videos of the robot performing new tasks in new environments. Blueprint removes pieces of information that is used, compressed, digested, digested data, which are used to teach these new tasks.

GR00T -R’s dreams complement the ISAAC Gr00t-MIMIIC Blueprint, published in the NVIDIA GTC conference in March. To enhance existing data from the GR00T-Mimic Nvidia Omniverse and Nvidia Cosmos platforms, GR00T-Dreams uses space to create completely new data.

The director of the EU and Distinguished, Jim Fanati, in the press briefing, “NVIDIA has a very strong robot strategy, and Jensen is centered around three computer problems.”

The firm’s simulation and graphical simulation physics is used to make and create information that is designed to make a engine engine and synthesize the data. And this information is consumed by the DGX computer used to prepare basic models. Then the Humoid robots are placed on the HX computer, which is time to work outside for platforms.

The GR00T said the fan of physical AI and robot-based working streams.

“This is an urgent in three computer problems,” he said.

The project pointed out GR00T, GR00T’s dreams and GR00T N1.5. (He said he was very proud of these names).

GR00T said fan for his dreams, a model that could create a video to grow robots. He showed a bunch of videos and said that all videos were created by Nvidia Cosmos.

“We did not find a way to recognize the developed video generation models such as space. Thus, we can use this technically adjustable model using the robots robots, principle of robots with robots of robots,” he said. “And now real robot becomes synthetic information to enhance our data sets, it also takes a lot of time to collect information about the actual robot and costs more.

The result is that the robots will be able to properly take the objects. You can say a cucumber or to pour a little orange juice or open the laptop. The robot has never trained in these special actions, fan, but because it is taught with video models, the robot is capable of “understanding the physics and meaning of these verbs.”

Thus, it learns how to fulfill the actions.

New Isaac Gr00T Models Advance Humanoid Robot Development

The NVIDIA research has used the GR00T -R Dreams Blueprint to create synthetic learning information to develop GR00T N1.5 – a update in Gr00t N1 – for about 36 hours, for about 36 hours, about 36 hours, without about 36 hours, without an update, without a plan.

The GR00T can better adapt to new environments and jobs, as well as the user’s guidelines. This update increases the success speed of the model for general material processing and production assignments, such as sorting or putting objects. The GR00T can be placed in Jetson Thor, launched at N1.5, at the end of this year.

The GR00T N1.5 Foundation model combines GR00T dreams as part of a synthetic information generation pipeline. NVIDIA has improved a visual language spell, so the GR00T N1.5 will be better adaptation and language training.

The GR00T N1.5 will be announced on this Computex and then open source until June 9. As for the GR00T’s dreams, Nvidia is still working on schedule. Hope is to open as many sources as possible.

The early children of the GR00T N include Aeirobot, Foxlink, Lightwheel and Neura Robotics. Aeirobot uses the model to understand Alice4 natural language teachings and carry out complex selection work in industry parameters. Foxlink Group uses the industrial robot manipulator to develop the ability to develop the ability and weakness, light, faster in factories to confirm synthetic data for humanoid robot locations. Neura evaluates a model to accelerate the development of robotic household automation.

The new robot simulation and data generation frames accelerate training pipelines

Highly skilled humanoid requires a high-ending mass to develop, capture and process robots. Robots should be tested in a physical world that can provide costs and risk.

In order to help close the data and testing cavity, NVIDIA presented the following simulation technologies:
● NVIDIA Cosmos is a new WFM that uses the thinking chain to help accurate, higher quality synthetic data for physical AI model education
● 2 space forecast used in GR00T’s dreams will come soon soon to face performance accessories for high quality world generation and hallucination.
● NVIDIA ISAAC Gr00T-MIMIC, a plan to create a large amount of synthetic movement trajectories for robotic manipulation using only a few human demonstrations.
● Now the open source physical AI database containing 24,000 high quality humanoid robot trajectories used to develop Gr00T N models.
● NVIDIA ISAAC SIM 5.0, a simulation and synthetic information generation framework, is now available in Open Github.
● NVIDIA ISAAC Laboratory 2.2, GR00T NO Open Source Robot Learning Frame that will include new evaluation environments to test the models of n models.

Foxconn and Foxlink use GR00T-MIMIC BluePrint for synthetic movement manipulation generation to speed up robotic training pipelines. Agility robototics, Boston Dynamics, Fourier, Mentee Robotics, Neura Robotics and XPeng Robots imitate and teach humanoid robots using NVIDIA ISAAC SIMA and ISAAC. Skild AI uses simulation frames to develop a common robot
Integrates into intelligence and common robotics to the robot exploration platform.

Universal Blackwell Systems for Robot Developers

Global Systems Manufacturers NVIDIA RTX Pro 6000 Workstations and Servers, Each Robot Development Workload is a single architectural oinger to easily manage the workload of every robot throughout the learning and simulation.

Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo and Supermicro announced NVIDIA RTX Pro 6000 Blackwell-powered servers and Dell Technologies and Lenovo Nvidia RTX Pro 6000 announced working stations working with Blackwell.

When required to create large-scale training or information, developers can hit NVIDIA Blackwell systems like GB200 NVL72 – NVIDIA Blackwell systems and NVIDIA Cloud Partners available with NVIDIA DGX clouds and NVIDIA Cloud Partners – Information To get 18x ​​greater performance for processing.