“Nvidia Introduces AI Safety Framework for Driverless Cars”

Nvidia has introduced a comprehensive safety framework tailored for driverless cars, encompassing training, simulation, and deployment aspects, as General Motors embraces its cutting-edge technology. The framework, known as Halos, covers platform, algorithmic, and ecosystem safety. It incorporates design-time, deployment-time, and validation-time guardrails, utilizing the DGX supercomputer for AI training, Omniverse and Cosmos on NVIDIA OVX for simulation, and NVIDIA DRIVE AGX chips and Drive OS for deployment.

General Motors has recently partnered with Nvidia to leverage its technologies for training, simulation, and AGX chips in their vehicles. Jensen Huang, the CEO of Nvidia, expressed enthusiasm about the collaboration, stating, “The time for autonomous vehicles has arrived, and we will work with AI in manufacturing, enterprise, and the way they simulate and design cars and in the car.” He highlighted the extensive experience Nvidia has in the autonomous driving sector, emphasizing the importance of safety in every aspect of the technology.

Halos plays a crucial role in meeting safety requirements, as Huang emphasized, “Safety requires technology from silicon to software, everything from diversity to monitoring and transparency, explainability, all of these philosophies have to be engrained in the silicon and software.” The framework ensures that safety is prioritized in every line of code, with meticulous safety assessments conducted on the AGX Orin chip.

Halos also encompasses the DriveOS software, a safety-certified operating system that spans from CPU to GPU, providing a secure foundation for various applications. Additionally, DRIVE AGX Hyperion, a hardware platform, integrates SoC, DriveOS, and sensors in an electronic control unit architecture. The framework includes libraries for safety data loading and accelerators, along with APIs for safety data creation and curation to filter out undesirable behaviors and biases before training.

Moreover, Halos offers training, simulation, and validation environments utilizing the Omniverse Blueprint for AV simulation with Cosmos world foundation models. This software stack combines modular components with end-to-end AI models to ensure safety with advanced AI models in the loop. The Nvidia AI Systems Inspection Lab, launched at CES, serves as a gateway for car manufacturers and developers to verify the safe integration of their products with Nvidia technology.

With the introduction of Halos, Nvidia aims to empower partners and developers to leverage state-of-the-art technology elements for creating safe and reliable autonomous vehicles. Riccardo Mariani, vice president of industry safety at Nvidia, highlighted the significance of Halos in complementing existing safety practices and potentially expediting standardization and regulatory compliance. The holistic approach to safety offered by Halos is crucial for companies looking to harness generative AI for developing advanced AV systems.