🚀 The AI for Industry Challenge Toolkit is LIVE!

Open Robotics has officially launched the AI for Industry Challenge Toolkit, providing researchers and developers with a production-ready simulation environment for training AI-powered industrial robots. The toolkit centers around a detailed robotic workcell featuring a Universal Robots UR5e manipulator equipped with an ATI Industrial Automation force-torque sensor, three Basler AG wrist cameras, and a Robotiq Hand-E gripper. This high-fidelity environment includes realistic assets for electronics assembly tasks like network cards and flexible cables, running on ROS with Zenoh for efficient data transport. The release represents a significant step toward bridging the simulation-to-reality gap in industrial automation.

The technical backbone includes a specialized impedance controller operating at 500Hz for smooth motion control while accepting targets at 10-30Hz frequencies suitable for vision-language-action (VLA) policies. Crucially, the toolkit offers multi-simulator flexibility, supporting Gazebo as the official evaluation environment while also working with MuJoCo (via Google DeepMind partnership) and Isaac Lab for massive parallel training. Integration with Hugging Face's Lerobot framework enables immediate teleoperation, data collection, and policy training. The cross-platform pixi package manager ensures accessibility across different systems, lowering barriers for widespread participation in the AI for Industry Challenge competition.