IntelligentHQ
IntelligentHQ
Skip to content

Robotics & Physical AI

The stack for robots that perceive and act: Isaac Sim and Isaac Lab for simulation and robot learning, Isaac GR00T for humanoids, Isaac ROS for ROS 2, Cosmos world foundation models and Jetson computers on the robot.

Technology profiles for this category are in research.

Overview

Physical AI means models that sense and act in the real world: mobile robots, robot arms and humanoids. Training them only on real hardware is slow and risky, so teams combine simulation, generated data and onboard computing.

NVIDIA Isaac is a robotics platform. Isaac Sim, open source under Apache 2.0 and built on Omniverse libraries, simulates robots and sensors and generates synthetic data. Isaac Lab, built on Isaac Sim, targets robot learning at scale. Isaac ROS brings accelerated packages to ROS 2, and Isaac GR00T is NVIDIA's research initiative and platform for humanoid robots. Cosmos 3 is an open world foundation model under the OpenMDW 1.1 license, used for reasoning, synthetic video and robot policy backbones. On the robot, Jetson Thor and Jetson Orin modules run models with the JetPack 7 SDK.

The audience is robotics engineers, research labs and manufacturers adding autonomy.

Fixed automation with well-defined motions often needs no learned model; classical motion planning and a standard controller may be enough.1234

Problems it addresses

  • Collecting data is expensive2

    Isaac Sim generates synthetic data, and Cosmos models can add variation to it.

  • Training policies safely2

    Reinforcement and imitation learning need many trials. Isaac Lab runs robot learning in simulation.

  • Gap between simulation and reality2

    Isaac Sim supports software-in-the-loop and hardware-in-the-loop testing before deployment.

  • Compute on the robot4

    Robots need onboard AI within a power budget. Jetson modules target robotics and edge AI.

A typical workflow

  1. Model the robot2

    Import CAD or URDF into Isaac Sim, which converts it to USD.

  2. Generate data1

    Create labeled synthetic data and add variation with Cosmos.

  3. Train the policy1

    Train in Isaac Lab, or start from an Isaac GR00T model for humanoids.

  4. Test in the loop2

    Validate with software-in-the-loop and hardware-in-the-loop runs.

  5. Deploy on the robot4

    Run perception and control on Jetson with JetPack and Isaac ROS.

Digital Twin Opportunity Lab

Find the first digital twin worth building for your site.

Open the lab

Next steps

  1. Start with one task, such as pick-and-place or navigation, and a robot model that already exists in URDF or USD.

  2. Install Isaac Sim from GitHub and run a sample scene with your sensor set.

  3. Use the Digital Twin Opportunity Lab to estimate the value of simulation for your site.

  4. Check Jetson module power and I/O against your robot's battery and sensors.

Sources

  1. NVIDIA Isaac (developer page) (opens in a new tab)NVIDIA · Vendor-reported
  2. NVIDIA Isaac Sim (developer page and FAQ) (opens in a new tab)NVIDIA · Vendor-reported
  3. NVIDIA Cosmos (product page and FAQ) (opens in a new tab)NVIDIA · Vendor-reported
  4. NVIDIA Jetson embedded systems (product page) (opens in a new tab)NVIDIA · Vendor-reported

Fill out the form below to request your copy.

Name(Required)