Hugging Face and Nvidia have published a workflow for building a robot pick-and-place simulation from reusable SimReady assets. The example combines a UR17G arm, a Robotiq 2F-140 gripper and a beverage can in Isaac Sim, then uses an AI agent to help assemble the scene and generate controls.
SimReady assets are OpenUSD files prepared for a defined simulation use case. They include properties such as collision shapes, mass, joints and joint drives, plus validation evidence against a profile. That shared baseline can reduce repeated setup work when developers combine robots and objects from SimReady Central, an open library hosted on Hugging Face.
The walkthrough still requires developers to test joint limits, movement and contact behavior. A robot can look correct while stationary but fail once physics starts. The example asks the arm to approach, grasp, lift, transfer and release the can, with developers checking whether the object remains in the gripper.
This is a controlled demonstration, not proof of real-world reliability. The can begins at a known location, no vision system is used, collision checks are limited and the motion follows fixed waypoints without replanning. Liquid and deformation are not modeled, and the simulated grip force was not calibrated to physical hardware. The useful result is a repeatable starting point for simulation, not a validated deployment recipe.