김은정
| 2026-09-13 09:00:06
By Kim Eun-jung
SEONGNAM, South Korea, Sept. 13 (Yonhap) -- In the fast-evolving autonomous driving industry, real-world driving data has become a critical resource for training artificial intelligence (AI) models, and Hyundai Motor Group is betting that the scale of its global fleet can give it a distinct edge.
With Hyundai Motor and Kia selling more than 7 million vehicles annually, the world's third-largest automotive group by sales is accelerating efforts to build a "data flywheel" that continuously collects driving data, trains AI models and feeds improvements back into vehicles.
At the center of the push is 42dot, the group's mobility software affiliate and a key pillar of its autonomous driving and software-defined vehicle (SDV) strategy.
During a media tour of 42dot's headquarters in Seongnam, south of Seoul, the company offered a glimpse of how Hyundai is building the data cycle needed to advance its technology.
At its vehicle workshop, an IONIQ 6-based SDV testbed equipped with eight cameras and one radar showed how the vehicle perceives its surroundings.
Parked next to it was an IONIQ 5-based data collection vehicle that looked markedly different from those on the road, with roof-mounted lidar and an array of other sensors, designed to collect high-quality real-world data used to train and evaluate 42dot's Atria AI model. The company currently operates around 40 such data collection vehicles day and night across a range of driving environments and conditions.
Park Min-woo, head of Hyundai Motor Group's Advanced Vehicle Platform Division and CEO of 42dot, said the ability to collect, learn from and quickly apply data will increasingly determine who gains an edge in autonomous driving.
"The key to competition in autonomous driving lies in how much quality data you can secure, how quickly you can learn from it and how fast you can reflect the results in real-world products and services," said Park, who previously worked on autonomous driving technologies at Tesla and Nvidia.
The challenge for Hyundai is turning the potential scale of its fleet into usable data quickly enough to catch up with rivals that have had a head start.
U.S. electric vehicle maker Tesla has spent years collecting real-world driving data through its customer fleet, while Waymo has focused on Level 4 autonomous driving through driverless robotaxi operations in designated areas. Chinese automakers and technology companies are also rapidly rolling out increasingly sophisticated driver assistance systems in the world's largest auto market.
To speed up development, Hyundai is pursuing a two-track strategy.
On one track, the group is developing its own autonomous driving technology centered on Atria AI, an end-to-end AI model designed to perceive its surroundings and make driving decisions.
On the other, Hyundai is working with U.S. AI chip giant Nvidia to accelerate the commercialization of advanced driver assistance and autonomous driving systems.
Under its latest road map, Hyundai plans to introduce Level 2+ autonomous driving technology developed in collaboration with Nvidia in its first mass-produced SDV in 2028.
In parallel, the group plans to continue training and improving its in-house Atria AI using real-world driving data and gradually expand the technology across production vehicles, with the longer-term goal of advancing toward Level 4 autonomous driving.
At Level 2, a vehicle can control steering and acceleration or braking simultaneously, but the driver must continue to monitor the road. More advanced Level 2+ systems can handle a wider range of driving tasks but still require driver supervision.
By comparison, at Level 3, a vehicle can drive itself under specified conditions but may require the driver to take over. At Level 4, the vehicle can operate without human intervention within its designated operating conditions.
In parallel with the Nvidia collaboration, Hyundai plans to continue training and improving its in-house Atria AI with real-world driving data, gradually expanding the technology across production vehicles with the longer-term goal of advancing toward Level 4 autonomy.
During the media tour, 42dot showed footage of its autonomous vehicles being tested in challenging real-world conditions, including morning rush-hour traffic in Seoul's Gangnam district, bus- and commercial vehicle-heavy roads in Jamsil and rainy conditions in Pangyo.
Hyundai also plans to deploy Atria AI in Gwangju's citywide autonomous driving test program later this year, using data collected on public roads to further train and refine the technology.
Beyond its in-house development, Hyundai is tapping partnerships with global technology companies to gain experience and accelerate commercialization.
Hyundai Motor is working with U.S. autonomous driving company Waymo, combining Hyundai's vehicle manufacturing capabilities with Waymo's Level 4 autonomous driving technology. Hyundai plans to begin delivering specially modified IONIQ 5 robotaxis equipped with Waymo's autonomous driving system in the fourth quarter of 2026.
Motional, Hyundai Motor Group's autonomous driving joint venture with Aptiv, has also launched a pilot robotaxi service with Uber in Las Vegas using the IONIQ 5, with plans to expand into Europe and the Asia-Pacific.
Ultimately, Hyundai believes its biggest advantage could come from scale.
"Hyundai Motor Group is well positioned to scale up its data collection by tapping its large fleet of mass-produced vehicles and expects to surpass rivals in data volume within five years," said Jung Sung-kyun, who is in charge of 42dot's Atria AI model development.
(END)
[ⓒ K-VIBE. 무단전재-재배포 금지]