NVIDIA Launches SimReady Central: A Comprehensive Library for Simulation-Ready 3D Assets

NVIDIA has launched SimReady Central, a library that offers more than 2,000 simulation-ready (SimReady) 3D assets on Hugging Face. This initiative aims to assist developers in creating robotics simulations and AI factory digital twins by providing assets that come pre-equipped with essential properties such as collision geometry, mass, and friction.

SimReady Central serves as a free resource for developers, allowing them to access a diverse range of infrastructure and equipment assets, including robot arms, grippers, warehouse equipment, racks, and cooling systems. Developers can check the capabilities of each asset and combine them from different suppliers to create comprehensive simulation environments.

Original Equipment Manufacturers (OEMs) can utilize the SimReady Central Library to prepare and validate their assets against SimReady requirements. This enables them to reach developers and facilitate faster testing and simulation scenarios, ultimately aiding industries in building simulations around their products while managing their own Hugging Face datasets.

The SimReady assets are based on OpenUSD, an open framework for describing and composing 3D scenes. These assets are prepared with the necessary properties for simulation workflows, which include details like mass, collision geometry, and friction properties for robotic manipulation tasks, as well as electrical or thermal metadata for AI factory applications.

The SimReady Foundation provides specifications and validation tools to ensure that assets meet the requirements for various simulation use cases. Developers are encouraged to select assets from SimReady Central that align with their workflow needs and compose them into OpenUSD scenes.

OEMs can convert their CAD equipment models into SimReady assets, making them discoverable on Hugging Face. This helps manufacturers reach teams focused on building robotics simulations and AI factory digital twins. For example, AI factory OEMs like Vertiv can distribute assets such as warehouse equipment and cooling systems, while robotics OEMs like FANUC and Universal Robots can make their robot arms and grippers available for developers.

Simulation asset providers, such as Palatial, can also contribute to the library, offering datasets like the NVIDIA Robotics 1K Dataset. This collaboration allows developers to build assets and scenes for robot training and evaluation.

To contribute to SimReady Central, OEMs can use the submission skill to validate their assets against the relevant SimReady profiles, package them, and submit them for inclusion in the library.

SimReady Central provides developers with a reusable starting point for creating robot workcells, industrial environments, and AI factory digital twins. They can find assets that meet their workflow profiles and integrate them into OpenUSD environments.

Developers can populate training environments with SimReady assets, allowing for the evaluation of robot behaviors by varying object placements and configurations. This flexibility supports projects like Microduck, an open-source robot developed by Hugging Face and Pollen Robotics.

NVIDIA USD Search enhances the discoverability of assets within SimReady Central, enabling natural-language searches across datasets. Hugging Face has integrated this search functionality into its platform, allowing agents to find SimReady assets suited for their physical AI simulation tasks.

AI-assisted authoring further streamlines the development process. Developers can leverage frontier models like OpenAI’s GPT-6 Astra to assist in finding suitable assets, writing code to assemble scenes, and creating scenario variations based on task requirements.

For instance, a developer creating a robotic pick-and-place scenario could prompt Astra to generate Python code for assembling a warehouse scene using SimReady assets. This process includes querying the SimReady Search API, referencing assets into an OpenUSD stage, and rendering a preview for review.

Similarly, developers can use the same approach to create AI factory digital twins by arranging equipment layouts while checking for overlaps and clearance violations. Astra would facilitate the assembly of selected SimReady assets into OpenUSD layouts, generating scripts for compliance checks.

SimReady Central aims to expand its library with more OEM partners and assets, enhancing the resources available to developers. Community feedback is encouraged to identify which additions are most needed, and OEMs are invited to submit their assets for inclusion.

Developers can start building their robotics simulations or AI factory digital twins by selecting assets from SimReady Central Library and utilizing the available resources to create their scenes.

Source: Hugging Face Blog