# NVIDIA Shows AI Agents Building Omniverse Simulation Apps By Simon Yoon Canonical URL: https://www.tokenpost.com/news/technology/43939 Published: 2026-10-29T22:04:56.000Z Updated: 2026-10-29T22:04:56.000Z Section: Technology > A warehouse test ran at 30 frames per second for two minutes with 800 boxes, though performance fell when the scene doubled in size. NVIDIA is showing developers how AI agents can assemble Omniverse simulation applications, using virtual scenes to explore robotics, industrial equipment and autonomous-driving workflows before physical changes. Developers give agents natural-language instructions, inspect the generated work and steer revisions. Omniverse libraries provide GPU-powered physics, rendering and sensor simulation. The workflows are aimed at building and refining digital twins, robot tests and scenario analysis tools. David Wehr, part of NVIDIA’s simulation technology team, used Astra to turn panoramic photos and point-cloud measurements into an editable building model. The model is stored in OpenUSD, where developers can reposition shelving and machinery in a virtual layout before making changes at a site. Pradeep Kaushik, from NVIDIA’s Omniverse developer relations team, built a what-if application around a simulated gas turbine and generator. Users can adjust load, ambient temperature and fuel composition, then compare power, efficiency, fuel use and equipment response. Six procedures include scenarios with a stuck fuel valve and an exhaust-sensor error. One warehouse simulation ran at 30 frames per second in full HD for two minutes with 80 conveyors, 40 arms and 800 boxes. Performance fell when the scene was doubled in size, limiting the reported frame rate to the smaller configuration. A separate humanoid warehouse simulator was also demonstrated. Other demonstrations include flexible cable routing for robot assembly, converting computer-aided design models into SimReady assets, building an International Space Station application in a browser, and comparing real and simulated camera and LiDAR outputs. The examples also cover robotic disassembly and autonomous-driving workflows. Separately, NVIDIA and Mistral AI plan to connect Mistral models with Omniverse libraries for engineering simulation in their [planned engineering simulation partnership](). ## Links in this article - [planned engineering simulation partnership](https://www.tokenpost.com/news/technology/37088)