NVIDIA has announced new developments in generative AI models and blueprints aimed at expanding the integration of NVIDIA Omniverse into physical AI applications. This includes areas such as robotics, autonomous vehicles, and vision AI. The announcement was made during the CES trade show, where leading software developers and professional services firms revealed their adoption of Omniverse to create new products and services.

Companies such as Accenture, Altair, Ansys, Cadence, Foretellix, Microsoft, Neural Concept, and Siemens are among the first to integrate Omniverse into their offerings. Siemens introduced the Teamcenter Digital Reality Viewer at CES as part of its Xcelerator application powered by NVIDIA Omniverse libraries.

Jensen Huang, founder and CEO of NVIDIA, stated: “Physical AI will revolutionize the $50 trillion manufacturing and logistics industries. Everything that moves — from cars and trucks to factories and warehouses — will be robotic and embodied by AI.”

The creation of 3D worlds for physical AI simulation involves three steps: world building, labeling with physical attributes, and making it photorealistic. NVIDIA's generative AI models aim to streamline these processes. The company has launched USD Code and USD Search microservices to allow developers to generate or search OpenUSD assets using text prompts. Additionally, a new Edify SimReady model can automatically label 3D assets with physics or material attributes.

NVIDIA Omniverse combined with Cosmosworld foundation models creates a synthetic data multiplication engine for generating large volumes of photoreal synthetic data quickly. This facilitates the composition of 3D scenarios in Omniverse which can be rendered as images or videos for use in training physical AI.

During CES, NVIDIA also unveiled four blueprints designed to simplify building Universal Scene Description (OpenUSD)-based digital twins for physical AI:

- Mega: For developing robot fleets at scale within industrial factory or warehouse digital twins.

- Autonomous Vehicle (AV) Simulation: To enable AV developers to replay driving data for closed-loop testing.

- Omniverse Spatial Streaming to Apple Vision Pro: For creating immersive streaming applications.

- Real-Time Digital Twins for Computer Aided Engineering (CAE): For real-time physics visualization using NVIDIA CUDA-X acceleration.

Market leaders are leveraging these advancements in industrial AI development through NVIDIA Omniverse. Cadence is integrating further into Allegro for electronic design applications; Altair is adopting blueprints for real-time CAE digital twins; Ansys is incorporating them into its Fluent application; while Neural Concept enhances engineering workflows with real-time CFD capabilities.

Accenture is utilizing Mega in collaboration with KION Group AG to build autonomous warehouses. Foretellix employs AV simulation blueprints for optimized testing validation alongside MITRE's collaboration with Mcity on an industry-wide AV validation platform.

Katana Studio applies spatial streaming workflows in custom car configurators for Nissan and Volkswagen enhancing customer experiences while Innoactive adds support enabling high-resolution project reviews via Apple Vision Pro involving collaborations with Syntegon Technologies GmbH among others.