NVIDIA Launches GR00T 2 and Isaac Lab 3.0 to Advance Robot Learning

NVIDIA announced two major updates for robotics development at the Conference on Robot Learning in Austin, Texas. The company introduced GR00T 2, an updated open frontier world action model, and Isaac Lab 3.0, a new version of its robot learning framework. Numerous partners are already testing these tools for factory automation, laundry handling, and cable swapping.

Training Robots with GR00T 2

GR00T 2 is trained on thousands of hours of robot and human data to help machines learn physical manipulation skills. The model achieves high task success rates with very few human demonstrations. It allows AI agents to run skills in sequence and recover from errors. Companies like Foxconn and Siemens are currently evaluating the technology for industrial applications.

Simulation Upgrades in Isaac Lab 3.0

Isaac Lab 3.0 provides the foundation for training robots inside virtual environments. The updated framework integrates closely with Newton, an open-source physics engine. This integration lets researchers model flexible objects like cables and cloth, as well as complex robotic hands. NVIDIA reported that the new setup delivers significantly higher training speed compared to previous versions.

New Industrial Benchmarks

To help measure robot performance, NVIDIA released new evaluation tools. The Isaac Lab-Arena framework allows developers to test policies across thousands of variations in parallel. Additionally, new benchmarks like DexBench and VABAR help assess robot dexterity and adaptability across retail, manufacturing, and logistics tasks.