Nvidia and Haging Face merge SimReady assets in simulation: Warehouse automation moves to virtual environments relevant to regional logistics
Listen to this article
Read by Anchor
The Haging Face platform, in collaboration with engineers from Nvidia, launched a practical guide for assembling and operating industrial robot cells within the Isaac Sim 6.2 simulation environment, based on an open library of physics-simulation assets called SimReady Central. The project focuses on the task of picking up and moving metal boxes by combining a UR 17-g arm with a Robotic 2 F 140 gripper, and automatically generating motion-control code using an AI agent that relies on a GPT-6-Sol model. This direct link between open-asset repositories and simulation engines provides automation engineers in Gulf and Egyptian supply chains with a digital pathway to evaluate robot behavior before committing to costly industrial hardware purchase contracts.
The SimReady philosophy is based on defining standard OpenUSD files so that they include not only the visual 3-D structure but also essential physical properties such as body masses, collision points, joints and drive motors.The SimReady Foundation framework connects three technical layers: the generic usage file, the required operational capabilities, and precise verification criteria for the joint-driving engines.The robot arm passed verification against the robot-structure specifications, while the gripper met industrial-grip specifications through tests with the physics engine PhysX, providing a standardized baseline that prevents simulation collapse or object drop on work surfaces during motion.
The AI agent managed the assembly steps by launching a remote Python server inside the simulation engine, using a custom skill package to interact with joints, computational physics and inverse-kinematics calculations.The agent then linked the arm to the gripper at a unified motion joint, created an independent motion path, and executed a full cycle that started with approaching, picking up the box, lifting it, transporting it and releasing it precisely at the target point.This shift toward tasking advanced language models with writing and tuning mathematical control code in simulation environments frees programming teams from manually performing complex calculations for every pivot point.
The experiment also reveals geometric limits that must be considered, as the simulation ran in a fixed coordinate space with the robot base locked and without computer-vision systems. The gripper’s torque limit was set to a maximum of 2 Nm, chosen through virtual testing without calibrating real-finger pressure, and the box was treated as a simplified solid mass that ignored metal deformation or internal fluid motion. These assumptions confirm that successful virtual box transport demonstrates the digital asset configuration’s integrity, but it does not constitute a ready-to-deploy solution for live packing lines without additional sensory calibration.
What does this mean for the region
Making physical robot assets available through open libraries offers a direct opportunity for developers of logistics distribution centers and smart warehouses in Saudi Arabia and the United Arab Emirates, where automation investments in sorting and handling are expanding in industrial cities and ports. This model lets local engineering teams test the compatibility of commercially available mechanical arms and grippers inside an integrated virtual environment before importing and installing them, reducing field-trial costs and the risk of physical failures during pilot operation.
This shift requires engineering and programming departments in Egypt and the Gulf to redirect technical skills toward controlling OpenUSD environments and operating AI agents capable of generating code via simulation engines. Realising the full benefit of these tools also demands a bridge between simulation and field reality, by integrating pressure sensors and 3-D vision systems to handle real-world variations in cargo weight and shape that simplified virtual models alone cannot manage.