A new network of sensor-equipped Hyundais will begin traversing urban streets this month to record edge-case scenarios for self-driving software. Uber will deploy up to 500 human-driven Hyundai Ioniq 5s equipped with cameras, lidar, and radar to capture unpredictable road events that are difficult to recreate in computer simulations, such as heavy rain or sudden roadworks. The fleet is expected to collect roughly 2 million miles of driving data every month.

Uber AV Labs has current agreements to provide these real-world datasets to autonomous vehicle developers Nvidia and Wayve, and is discussing partnerships with other firms. Unlike robotaxi projects, these vehicles remain human-operated to ensure a steady stream of diverse city navigation data, a resource the company notes that most AI firms would struggle to build from scratch.