![nvidia]()
What’s New:
At SIGGRAPH today, NVIDIA unveiled a major leap in robotics innovation with three key offerings:
-
Omniverse Libraries & SDKs
-
These tools now support Universal Scene Description (OpenUSD) and MuJoCo (MJCF) formats, unlocking seamless simulation for over 250,000 robot learning developers.
-
NuRec libraries and neural rendering models enable realistic 3D world reconstruction from sensor data using RTX ray-traced Gaussian splatting.
-
Isaac Sim 5.0 and Isaac Lab 2.2 open-source frameworks now feature the above enhancements plus OpenUSD-based schemas, easing the sim-to-real gap. These tools are fully available on GitHub.
![robotics and omniverse]()
-
Cosmos Foundation Models for Robotics
-
Cosmos Transfer-2 streamlines synthetic data creation from simulation scenes using rich spatial controls like depth maps and segmentation. A distilled variant simplifies this down to a single-step generation on RTX PRO servers.
-
Cosmos Reason is a 7-billion-parameter vision-language model that blends physics understanding and common sense, enabling robots to decompose complex commands, reason in unfamiliar settings, and support data annotation, planning, and analytics tasks.
-
Powerful AI Compute Tools
-
RTX PRO Blackwell Servers handle the full spectrum—from training and simulation to synthetic data generation and inference.
-
DGX Cloud, now on Microsoft Azure Marketplace, provides a fully managed platform for streaming OpenUSD- and RTX-based applications. Early adopters include Accenture and Hexagon.
Watch the NVIDIA Research special address at SIGGRAPH.
Why It Matters for Developers:
Together, these tools propel physical-AI and robotics development forward, enabling realistic world modeling, streamlined synthetic data workflows, advanced reasoning, and scalable compute access. It’s a powerful ecosystem opening new possibilities for autonomous systems and robotics innovation.
Read more here.