Robotaxi services are moving from prototype milestones into commercial operations, with programs expanding across multiple regions on NVIDIA DRIVE Hyperion. Uber and Autobrains are launching a robotaxi program in Munich using Autobrains’ agentic Artificial Intelligence, Foxconn is expanding work with NVIDIA to deploy robotaxi fleets in Taiwan, VinFast is working with Autobrains to bring level 4 vehicles built on DRIVE Hyperion to the Southeast Asia market, and HUMAIN is working to bring DRIVE Hyperion-powered robotaxis to Saudi Arabia.
As the robotaxi industry scales, NVIDIA is emphasizing that safety must be designed into the full autonomous vehicle stack rather than added after deployment. Industry discussion on level 4 autonomy often centers on what the vehicle can perceive and decide. Regulators require something more: proof that the overall system behaves reliably, isolates faults before they escalate and never operates outside the boundaries it was designed for. Robotaxi safety requires solving four distinct challenges simultaneously: a safety-certifiable operating system, safe and standardized hardware and software interfaces, Artificial Intelligence that operates within verifiable guardrails, and validation at scale before vehicles reach public roads.
NVIDIA Halos OS, part of the NVIDIA Halos full-stack safety system, is presented as a unified safety foundation for Artificial Intelligence-driven vehicles built on DRIVE Hyperion. Halos Core is the next generation of NVIDIA DriveOS and certified to automotive safety standards, with a hypervisor that isolates safety-critical functions so failures cannot reach vehicle controls. Halos Core is compliant with ISO 26262 ASIL D, includes safety-certified support for NVIDIA CUDA and TensorRT, and provides the TensorRT Edge-LLM open source framework for high-performance large language model inference.
Halos SDK addresses the integration burden created by cameras, radar, lidar and other sensors that stream data in different formats and at different rates. Its sensor abstraction layer decouples the autonomous driving stack from individual sensor drivers, while its vehicle abstraction layer connects the stack to the rest of the vehicle through a single interface. Halos Applications provides safety guardrails for Artificial Intelligence through deterministic, rule-based functions, including world model perception and NVIDIA DRIVE active safety features such as automatic emergency braking, lane departure warning, blind spot monitoring and collision warning.
Halos Infra is the cloud-side development infrastructure that enables autonomous vehicle training, simulation and validation at scale, and it forms the foundation for the NVIDIA Halos Safety Evaluation Framework. SEF provides the tools and guidelines needed to build a credible safety case, from L2 driver assistance to L4 robotaxis. It draws on more than 330 research papers and 1,000 patents developed within NVIDIA Halos OS. Halos Infra runs on NVIDIA’s three-computer autonomous driving solution: NVIDIA DGX for training, NVIDIA Omniverse on NVIDIA OVX for simulation and synthetic data generation, and NVIDIA AGX for in-vehicle real-time processing and safety.
