Nvidia Unveils New Robotics Tools to Master Delicate Tasks in Simulation

Teaching Robots Fine Motor Skills

Simulation has long helped researchers teach machines how to walk and roll. Now, virtual training environments are helping robots learn tasks that require a finer touch, such as connecting cables, aligning gears, and assembling small components. Nvidia revealed new open robotics technologies at the Conference on Robot Learning in Austin, Texas, aiming to help machines move from lab tests to real-world industrial work.

World Action Models and Training Tools

At the event, Nvidia introduced GR00T 2, an open frontier world action model built on its Cosmos platform. Trained on extensive human and robot data, the model helps machines learn dexterous manipulation skills. Companies are already testing it for tasks ranging from robotic laundry handling to factory automation. To support such training, Nvidia released Isaac Lab 3.0, a framework that lets researchers set up simulation environments quickly.

Accurate Physics and Performance Evaluation

Isaac Lab integrates tightly with an open source physics engine called Newton. This allows the software to model complex systems, such as robotic hands with many joints, and deformable objects like cables and cloth. By combining accurate physics with large-scale parallel processing, researchers can train thousands of simulated robots at once. New frameworks also allow developers to evaluate robot performance across thousands of variations and identify conditions that cause failures.