Hardware
The Machines that Make the Machines
NVIDIA’s Seattle Robotics Lab and Isaac engineering team created reusable robotic systems to automate GB300 tester‑tray assembly, achieving over 95 % success on busbar tasks with a 160‑second cycle ti
NVIDIA’s Seattle Robotics Lab and Isaac engineering team created reusable robotic systems to automate GB300 tester‑tray assembly, achieving over 95 % success on busbar tasks with a 160‑second cycle time through FoundationPose perception, Lissajous‑curve planning, and impedance control. For multi‑connector insertion they combined SAM3 segmentation for cable grasping, a custom DOPER pose‑estimation framework trained on synthetic/real data, and 3‑D‑printed grippers that mechanically constrain motion, then deployed reinforcement‑learning policies pretrained in Isaac Lab and refined with real‑world residual training (SPARR) using force–torque feedback. These pipelines are packaged into TALOS containerized modules for rapid perception, planning, and control deployment across robot platforms.
Related
- How to Use AI Agents to Prepare 3D Scenes for Simulation
- Accelerating a ROS 2 Node with an AI Agent and NVIDIA Isaac ROS
Source: NVIDIA Developer | 2026-10-07