In a groundbreaking announcement at the CES trade show, NVIDIA unveiled enhancements to its three-computer solution for autonomous mobility through the integration of its Cosmos platform. This development is poised to revolutionize the transportation industry, with numerous industry leaders among the early adopters of the Cosmos foundation models.
Autonomous vehicle (AV) technologies have traditionally relied on three critical components: NVIDIA DGX systems for AI training within data centers, NVIDIA Omniverse powered by NVIDIA OVX systems for simulation and synthetic data generation, and NVIDIA AGX in-vehicle computers that process real-time sensor data to ensure safety. Together, these systems are engineered to enable continuous development cycles, accelerating improvements in performance and safety.
Adding to this sophisticated framework, the newly introduced NVIDIA Cosmos platform incorporates state-of-the-art generative world foundation models (WFMs), advanced tokenizers, guardrails, and an accelerated video processing pipeline. This enhanced system aims to dramatically push the boundaries of physical AI systems such as AVs and robotics.
With the Cosmos platform, developers are now equipped with a data "flywheel" capable of transforming thousands of miles driven by humans into billions of virtual test miles. This new dimension dramatically elevates training data quality and accelerates the process.
"The AV data factory flywheel consists of fleet data collection, accurate 4D reconstruction, and AI to generate scenes and traffic variations for training and closed-loop evaluation," said Sanja Fidler, Vice President of AI Research at NVIDIA. "Using the NVIDIA Omniverse platform, along with Cosmos and supporting AI models, developers can generate synthetic driving scenarios to amplify training data by orders of magnitude."
NVIDIA's Cosmos is already making significant inroads in the transportation sector as industry leaders leverage its capabilities to refine physical AI models for autonomous vehicles. Companies like Waabi are harnessing Cosmos for video data curation and simulation in AV software development, while Wayve explores Cosmos for searching complex driving scenarios to optimize safety validation processes.
Moreover, renowned AV toolchain provider Foretellix plans to integrate Cosmos with NVIDIA Omniverse Sensor RTX APIs to scale high-fidelity testing scenarios and training data generation.
"Developing physical AI models has traditionally been resource-intensive and costly for developers," remarked Norm Marks, Vice President of Automotive at NVIDIA. "Cosmos speeds up this process with generative AI, fostering smarter, faster, and more precise development for AVs and robotics."
Furthermore, NVIDIA's partnership with ridesharing giant Uber is set to expedite the future of autonomous mobility. The collaboration utilizes Uber's extensive driving datasets combined with Cosmos platform features and NVIDIA DGX Cloud, enabling AV partners to build robust AI models with newfound efficiency.
Cosmos WFMs are available under an open model license on Hugging Face and the NVIDIA NGC catalog, with forthcoming releases anticipated as fully optimized NVIDIA NIM microservices. This accessibility presents developers with the opportunity to embark on transformative projects, offering a significant leap forward in the realms of AI, automation, and process mapping.
As NVIDIA propels into this new era of autonomous mobility, Jengu.ai, a leader in automation, AI, and process mapping, observes the extraordinary potential in these advancements. They stand ready to guide and influence the adaptive tech landscape, empowering organizations to harness these innovations to craft the future of transportation and robotic systems.
```
