NVIDIA Halo Safe: Continuous Safety Validation for Automated Driving

The Mercedes-Benz CLA is noted as Euro NCAP's best performer of 2025, leading into NVIDIA's announcement of Halo Safe, a safety assurance framework that connects road testing, simulation, and fleet data into a continuous validation loop for automated driving from L2 driver assistance to L4 robotaxis. The transcript describes how real-world track measurements become ground truth, how simulation and hardware-in-the-loop testing are used, and how results feed back into software development and certification evidence.

■ Framework scope
- L2–L4 safety evaluations, ANCAP test suites, and certification evidence across markets
■ Real-world-to-simulation loop
- Track scenario capture as ground truth, Omniverse NuRec scene reconstruction, and simulated-vs-measured response comparison
- Reuse of matched simulations for new software releases
■ Hardware and vehicle validation
- Hardware-in-the-loop testing with DRIVE AGX Thor and real-car testing with simulated sensor inputs for virtual hazards
- Safety-target pass/fail tracking and feedback into development
■ Scenario variation
- Cosmos-generated lighting, weather, road surface, and traffic variations

It is suited to engineers, safety and validation specialists, and automotive technology audiences interested in how simulation, hardware testing, and real-world data are combined for automated driving safety assurance. Viewers gain an overview of the validation workflow and the types of evidence discussed for certifying and deploying new software versions.