Nvidia Ushers in the Era of Physical AI
With the new Omniverse libraries announced for Agent Toolkit, NVIDIA enables artificial intelligence spies to create 3D simulation worlds and train physical AI systems.
Complex systems such as autonomous robots, smart factories and driverless vehicles are trained and tested in fully GPU-accelerated virtual environments before being put into production in the real world.
-
NVIDIA Agent Toolkit; It expands with open source Omniverse libraries such as ovrtx, ovphysx and CAD-to-SimReady, providing 3D simulation tools for AI spies.
-
Leading software manufacturers such as SideFX and PTC are initiating artificial intelligence-focused design processes by integrating new libraries into their platforms such as Houdini and Onshape.
-
Newly tools announced and Blender integration guide; It can run at full performance on both local and cloud-based hardware such as NVIDIA RTX Spark and DGX Station.
Simulation-Oriented Artificial Intelligence Development Process
It is of great importance that advanced robotic systems and industrial automation tools work correctly before they land on the field. NVIDIA CEO Jensen Huang emphasizes that the age of physical artificial intelligence will be built primarily in simulation environments. In line with this vision, Omniverse libraries added to the Agent Toolkit give AI spies the ability to inspect 3D scenes, detect disruptions and make objects ready for simulation.
Libraries offered open access on GitHub enable objects to be modeled not only visually but also in accordance with physical rules. For example, a 3D asset’s mass, friction coefficient, material structure and sensor images can be automatically scanned and edited by artificial intelligence spies.
Featured New Libraries and Technical Details
The software components offered with the new package focus on the different simulation layers required by developers:
-
NVIDIA RTX Sensor Simulation (ovrtx):It enables artificial intelligence spies and developers to generate camera, lidar and radar data from virtual scenes. In this way, how autonomous vehicles and robots perceive the environment is tested in a virtual environment.
-
GPU Accelerated Physics (ovphysx):It brings physical properties such as collision, mass, friction and movement to the stage. The interaction of objects with each other is simulated with high accuracy.
-
CAD-to-SimReady Skills:It converts computer-based design (CAD) information into OpenUSD-based SimReady standards. This process enables engineering drawings to be used directly in artificial intelligence simulations.
Industrial Giants Started Integration
Industry-leading 3D and CAD software manufacturers are beginning to incorporate NVIDIA Omniverse libraries into their ecosystems. SideFX, which developed the procedural 3D content creation tool Houdini, uses the ovrtx and ovphysx libraries to make it easier for technical artists to test complex simulation content.
PTC Onshape, a cloud-based CAD and work data management platform, brings together work design processes with direct physical simulation, thanks to OpenUSD and OVRTX integration.
Designers can now perform artificial intelligence-based tests while the work is at the drawing stage. Additionally, startups such as Palatial, Lightwheel, ForgeCAD and Moonlake AI automate SimReady asset production using artificial intelligence spies.
Hardware Support and Blender Integration
NVIDIA is announcing its new integration guide for its open source 3D software Blender at the SIGGRAPH event. The “SimReady Blender” example shows how NVIDIA NeMo-based AI spies can work well with existing graphics applications.
These workflows can run on a wide range of hardware, from local workstations to massive data centers. The systems offer uninterrupted performance on NVIDIA DGX Station computers with GB300 processors, as well as on NVIDIA RTX Spark systems that will be launched by ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI in the autumn.
