CAD-to-Control Robot Simulation with NVIDIA Omniverse (Sim-to-Real)
In this OpenUSD Insiders livestream, host and guest Kamate Richards (founder of 10 Things Inc., former Amazon) dive into CAD-to-control robotics workflows using OpenUSD and NVIDIA Omniverse. This session focuses on building higher-fidelity robot simulation pipelines and improving sim-to-real outcomes by: - Preserving rich CAD and material data with OpenUSD layering - Understanding why simulations break when moving between environments - Using sim-to-sim-to-real workflows to validate physics and behavior before deployment - Bringing real-world effects back into simulation (noise, slip, lighting, environment variation) What you will learn: - Why simplified robot models (URDF-era workflows) often fail in real environments - How OpenUSD enables layered, tool-agnostic robotics assets and physics metadata - Where Isaac Sim fits in the robotics pipeline (sensors, collisions, training workflows) - When to use faster physics-focused workflows (Warp + Newton Viewer) for rapid iteration - How custom physics and differentiable components can extend simulation capability Resources: - OpenUSD: https://www.nvidia.com/en-us/glossary/openusd/ - Isaac Sim: https://github.com/isaac-sim ---------------------------------------------------------------------------- Key Moments: 0:00 Intro music 0:19 Welcome to OpenUSD Insiders and guest introduction 1:13 Session overview: CAD-to-control, sim-to-real, and common failure modes 2:13 Kamate Richards background and robotics focus 3:47 Why simulation breaks and what “good” robotics deployment looks like 8:14 Continuous discovery, learning, and deployment cycle 9:05 NVIDIA robotics ecosystem overview (CAD, training, deployment) 11:35 Toolchain gaps and why robotics is hard 16:12 Why OpenUSD for robotics simulation 18:24 Layered USD approach for higher-fidelity simulation 19:35 Sim-to-sim before sim-to-real: when and why 21:25 Demo setup: upgrading a USD asset into a physics-ready object 22:12 Chair asset walkthrough (Sketchfab) and articulation discovery 25:41 Compute discussion: DGX Spark, Jetson Thor, portability 28:58 Q&A: “Fully autonomous” and top challenges 33:34 Demo continuation: rigid bodies, collisions, and joint testing 36:00 Material physics and deformable behavior considerations 37:43 Newton Viewer overview and physics-focused iteration 40:10 Warp example: custom aerodynamic forces for a paper airplane 42:22 Q&A: engineering analysis integration, sim-to-real gaps, domain randomization 49:47 Warp highlights: custom physics, differentiable components 51:15 10 Things Inc. overview and robotics platform direction 53:07 GTC mention: networking, sessions, labs, and community meetups 1:07:19 Cosmos Cookoff and community competitions 1:10:05 Closing thoughts and community resources ---------------------------------------------------------------------------- 📆 Check out the full calendar for all of our upcoming events → https://nvda.ws/3JqaWnA ---------------------------------------------------------------------------- ⬇️Get Started → https://nvda.ws/4cZAZO1 👀Explore OpenUSD → https://nvda.ws/3CeozBQ 👥Join the Community → https://nvda.ws/3ZMfc6e #OpenUSD #NVIDIAOmniverse #Robotics #RobotSimulation #SimToReal #Simulation #IsaacSim #PhysicsSimulation #ReinforcementLearning #DomainRandomization #URDF #USD #DeveloperLivestream #RobotLearning #DigitalTwins