Physical AI Learning#

Humanoid robots are uniquely suited to operate in environments built for people, including reaching across tables and grasping everyday objects. This course introduces the GR00T reference workflow: a validated, open, and reproducible sim-first to deployment workflow for Unitree G1 manipulation policy learning with Jetson Thor.

Level: Intermediate
Time: 4-12 hours
Key Skills: NVIDIA Isaac GR00T, Isaac Lab-Arena, Isaac Teleop, Isaac ROS, Jetson Thor, Unitree G1

Start Learning

Explore a complete sim-to-real workflow using the SO-101 arm and an affordable physical workspace you can build yourself. Configure and calibrate the robot, then perform demonstrations to post-train a GR00T policy for centrifuge vial pick-and-place. Evaluate with Isaac Lab and deploy to real hardware, applying four strategies to help close the sim-to-real gap.

Level: Intermediate
Time: 6-10 hours
Key Skills: NVIDIA Isaac Lab, NVIDIA GR00T, LeRobot by Hugging Face, Cosmos, sim-to-real transfer principles

Start Learning

Learn to assemble complex industrial scenes using NVIDIA Omniverse and OpenUSD standards for factories and warehouses. Apply scene composition, asset organization, and collaborative workflows.

Level: Intermediate
Time: 3-4 hours
Key Skills: OpenUSD, scene composition, asset organization, pipeline development

Start Learning

Explore the NVIDIA Isaac Sim interface. Build a robot from scratch, configure physics properties, add sensors, and run your first simulation.

Level: Beginner
Time: 2-3 hours
Key Skills: Physics simulation, sensor integration, environment setup

Start Learning

Dive into reinforcement learning and GPU-accelerated training. Understand how NVIDIA Isaac Lab trains thousands of robots in parallel, achieving convergence in hours instead of days.

Level: Intermediate
Time: 3-4 hours
Key Skills: Reinforcement learning, GPU acceleration, policy training

Start Learning

Accelerate ROS 2 development with NVIDIA Isaac ROS GEMs and NITROS for production-grade perception and navigation systems. Deploy AI policies on real hardware with real-time constraints.

Level: Intermediate
Time: 2-3 hours
Key Skills: ROS 2, NITROS acceleration, production robotics

Start Learning

Develop OpenUSD interoperability skills. Convert URDF to USD, optimize robotic assets, and build enterprise-grade digital twins that scale to complex multi-robot systems.

Level: Advanced
Time: 4-5 hours
Key Skills: USD interoperability, asset optimization, enterprise deployment

Start Learning

Explore robotics applications in healthcare environments. Learn safety-critical system design, regulatory considerations, and healthcare-specific sensor integration approaches.

Level: Intermediate
Time: 3-4 hours
Key Skills: Healthcare-specific robotics, safety protocols, domain adaptation

Start Learning

Explore Universal Scene Description (USD) for complex 3D workflows. This free curriculum covers asset structure, composition arcs, and pipeline development—preparing you for OpenUSD certification.

Level: Beginner to Advanced
Format: Self-paced, open-source
Key Skills: USD fundamentals, asset modularity, data exchange

Start Learning