LATEST

← Back to GPN · All articles

NVIDIA's SimReady Assets Let Developers Build Robot Pick-and-Place Simulations on Hugging Face

NVIDIA's SimReady Assets Let Developers Build Robot Pick-and-Place Simulations on Hugging Face

OFFICIAL Written by GPN · Nov 14, 2026, 10:17 PM ET · · AI

NVIDIA and Hugging Face have published a hands-on walkthrough showing how developers can build a robot pick-and-place simulation using SimReady assets, the OpenUSD models prepared with the physics properties a simulation needs.

SimReady means an asset arrives ready for a defined simulation use case, with properties such as collisions, mass, joints and joint drives already in place. SimReady Central, an open asset library on Hugging Face, hosts robot and object assets with authored simulation properties and profile-validation evidence, so developers can reuse them instead of preparing every model from scratch.

The walkthrough builds a beverage can pick-and-place task in Isaac Sim 6.2. It combines three assets from the library: a UR17G robot arm, a Robotiq 2F-140 gripper and a beverage can. The developer downloads the three packages, assembles the arm and gripper into a single articulation, and adds task code that defines the work surface, starting positions and controls for moving the arm and opening and closing the gripper.

Before the task runs, the assets are checked against their SimReady profiles. The arm validation report uses the Robot-Body 4.0.0 profile and the gripper report uses Robot-Gripper 3.0.0, with both recording passing PhysX driven-joint checks. The walkthrough also stresses testing the assembled scene first: review the joints, try a small arm movement, open and close the gripper, and confirm the can rests on the work surface without falling through it.

An AI agent can help with much of the work. The example uses an agent with Isaac Sim skills to assemble the robot, validate the physics and generate the pick-and-place controller, which coordinates the arm's movement with the gripper's opening and closing. The full task runs an approach, grasp, lift, hold, transfer and release sequence, and the guide lists checks for judging whether the run behaved correctly, including whether the gripper holds the can while the arm moves.

The guide is upfront about the limits of the example. The setup is fixed and scripted: the can starts at a known position, no vision is used, collision checks are limited, and the grasp follows fixed waypoints without replanning. A 2 N·m drive-effort cap used in the example was chosen through simulation tests, not calibrated to fingertip forces or recommended by the manufacturer.

Developers who want to try it can browse SimReady Central for robots and objects, check what each asset supports against the SimReady Foundation profiles, and follow the linked Isaac Sim installation and agent-skills guides to set up their own scene.

Facts gathered from

More on this