Hyundai Motor is ramping up its efforts to secure autonomous driving data. Following a successful bid for the Seoul RoboTaxi project, the company is set to collect road environment data in the city, despite the immediate profitability of the venture being uncertain. The driving data collected is expected to enhance the sophistication of its 'data flywheel' process, which involves iterative AI learning and vehicle application.
According to industry sources, Hyundai plans to launch its autonomous taxi service in Seoul by the end of this year. The company was selected alongside Kakao and Tmoney Mobility as an operator for the city’s autonomous taxi platform, which encompasses everything from ride-hailing to operational management.
The motivation behind Hyundai's entry into the Seoul RoboTaxi market is the need to gather autonomous driving data. In the bustling urban environment of Seoul, where vehicles, pedestrians, and two-wheelers frequently interact and unexpected situations arise, the company aims to capture 'edge cases' that are difficult for AI to learn in advance. This data is directly linked to enhancing autonomous driving competitiveness.
Hyundai is reportedly considering applying the autonomous taxi operation experience and data collected in Seoul to its own data flywheel. The data flywheel is a cyclical structure where data is collected from real vehicles, which is then used by AI to improve autonomous driving technology, and subsequently applied back to the vehicles to gather new data. This approach is similar to Tesla's method of training its autonomous driving AI using vast amounts of driving data from production vehicles.
Last month, Park Min-woo, head of Hyundai Motor Group's Autonomous Vehicle Division, identified the data flywheel process—'data collection → AI learning → vehicle application → data collection'—as a core competitive advantage in autonomous driving during the CEO Investor Day. He emphasized that the speed and quality of data acquisition and the iterative AI learning process are crucial for maintaining competitiveness in the autonomous driving sector.
To establish its data flywheel, Hyundai plans to deploy over 200 autonomous vehicles in Gwangju in the second half of this year. The company will also utilize its end-to-end (E2E) autonomous driving solution, 'Atria AI.' Next year, it intends to expand the testing areas to five districts in Gwangju.
Hyundai is also building a large-scale infrastructure to support its data flywheel. The group plans to establish an AI data center in Saemangeum to process the vast amounts of data collected from vehicles and use it as a foundation for AI model training. The structure will involve collecting data on actual roads in Seoul and Gwangju, training and improving AI at the data center, and then applying the results back to the vehicles.
The demand for data acquisition is expected to grow in the future. Starting in 2028, Hyundai plans to sequentially launch its software-defined vehicles (SDVs) to expand its data collection capabilities to production vehicles. This initiative builds on its previous work with Motional and the development of the Ioniq 5 RoboTaxi, as well as supplying Ioniq 5-based vehicles to Waymo, aiming to create an ecosystem that encompasses its own vehicles, autonomous driving AI, and mobility platforms.
A Hyundai representative stated, "To enhance autonomous driving technology, it is crucial to gather a sufficient amount of diverse data generated on actual roads."
Meanwhile, Hyundai will unveil its autonomous driving strategy, including its software-defined vehicles and data flywheel, at the 'HMG Autonomous Driving Media Day' on the 11th.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.