Nvidia Showcases GR00T 2 and Isaac Lab 3.0 at CoRL 2026 in Austin

Nvidia showcased GR00T 2, a new foundation model for robots, and Isaac Lab 3.0, an upgraded simulation framework, at the Conference on Robot Learning (CoRL) in Austin, Texas, expanding its software tools for training and deploying robots.

The company also introduced an open benchmark for evaluating robot policies in simulation and presented Isaac Lab-Arena. CoRL 2026 runs Nov. 9-12 in Austin.

Isaac Lab 3.0 has a modular design that lets developers use different physics engines, including Newton and PhysX. It supports training across multiple graphics processing units (GPUs), and includes LEAPP, a tool designed to help transfer a trained policy from simulation to physical hardware.

GR00T 2 is the latest entry in Nvidia's robot foundation-model lineup. A foundation model is a general-purpose AI system that can be adapted for different tasks.

Robotics developers use simulation to train systems before deploying them. Unlike language models that can learn from online text, robots have limited access to real-world practice, making simulated training environments one way to develop their capabilities.