NVIDIA opens a free library of more than 2,000 3D assets for simulating robots on Hugging Face
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
SimReady Central is NVIDIA's new open collection on Hugging Face: more than 2,000 free 3D models that already include mass, friction, collisions and joints. Comau, FANUC, Universal Robots, Vertiv or LG Electronics contribute their arms, grippers, racks, cooling and home appliances. It is meant for training robots in virtual environments and recreating AI factories, and it can be searched by describing in words what is needed.
NVIDIA has unveiled SimReady Central, an open library on Hugging Face with more than 2,000 3D objects prepared for simulation. It is free and can be downloaded without requesting access. The company announced it on Wednesday, October 21, on its Hugging Face blog.
The library brings together robot arms and grippers, warehouse equipment, racks and cooling equipment. It is intended for building environments in which to train and evaluate robots, and for setting up digital twins (virtual copies) of "AI factories," NVIDIA's term for facilities with server racks and cooling for artificial intelligence.
What sets these objects apart from an ordinary 3D model is that they already carry what a simulation needs: collision geometry, mass and friction, joints with their limits, labels to identify each object, and materials and attributes for simulating sensors. According to NVIDIA, developers often had to add those properties by hand until now. They all use OpenUSD, an open format for describing 3D scenes.
The manufacturers contributing equipment include Comau, FANUC, Franka Robotics, Robotiq, Standard Bots and Universal Robots, with robot arms and grippers; Vertiv, with racks and cooling; and LG Electronics, with refrigerators and dishwashers for simulating homes where robots work. There are also objects from Palatial, which includes a dataset made with NVIDIA Robotics, and from ForgeCAD. Manufacturers can validate their own models and send them in to be added to the library.
The objects can be searched by describing them in words, and Hugging Face has built that search into its own. NVIDIA gives the example of asking OpenAI's GPT-6 Astra to set up a warehouse scene with a FANUC arm and a Robotiq gripper that picks up and puts down objects, and to write the code to simulate it. In another example, the model designs two layouts for an AI factory with racks and cooling and checks that the equipment does not overlap and that the aisles have the required clearance.
The company plans to add more manufacturers and more objects, and is asking developers to tell it what they need. The announcement does not specify the license under which the objects can be used.
Why it matters · analysis and opinion
Setting up a robotics simulation used to start with an unglamorous chore: taking a 3D model and telling it how much it weighs, how it collides and how it moves. Having that work done in advance, and done by the manufacturers themselves for their own arms and equipment, makes the testing stage cheaper for integrators and small teams without a simulation department. It also says something about the market: manufacturers are starting to see their virtual catalog as a way for customers to try before they buy. It remains to be seen how good these objects really are outside NVIDIA's tools and under what terms of use they are offered, because the announcement does not spell them out. Anyone considering automating a warehouse can already check whether their suppliers are in.
Official source: NVIDIA on Hugging Face · Written with the help of AI: how we make the news
