NVIDIA Advances Robot Learning with Open-Source Tools

NVIDIA has introduced new open-source tools designed to accelerate robot learning through simulation, focusing on a "sim-first" approach. The announcement, made at the Conference on Robot Learning, highlights key advancements, including the Newton Physics Engine, Isaac GR00T N1.6, and Isaac Lab 2.3.

Newton Physics Engine

The Newton Physics Engine is a GPU-accelerated, open-source tool co-developed with Google DeepMind, Disney Research, and managed by the Linux Foundation. Built on NVIDIA Warp and OpenUSD, it enables robots to learn complex tasks with greater precision through advanced simulation.

Isaac GR00T N1.6

Isaac GR00T N1.6 is the latest open robot foundation model, set to be available on Hugging Face. It integrates NVIDIA Cosmos Reason, allowing robots to understand ambiguous instructions and navigate unforeseen scenarios using a reasoning vision language model.

Isaac Lab 2.3

The latest version of Isaac Lab, built on NVIDIA Isaac Sim and OpenUSD, is now available as an early developer release. Version 2.3 introduces features like advanced whole-body control, enhancing capabilities for robotics researchers.

Adoption and Impact

Leading robotics developers are adopting these technologies to accelerate physical AI development and deployment. The tools are being used to train robots for various real-world applications, demonstrating their potential to revolutionize the field.