Hyundai Motor Group has activated its Data Flywheel, a continuous learning system designed to rapidly advance autonomous driving technology through automated data collection, AI training, validation and deployment. The announcement came during an industry event at 42dot headquarters in Gyeonggi Province, where the group detailed its strategy to compete in the global autonomous vehicle market.
The Data Flywheel operates as a cycle in which driving data collected from vehicles is used to train and validate AI models, with improved models then deployed back to vehicles to generate new data for further refinement. This creates what the group describes as a “virtuous cycle” enabling faster technology advancement than traditional development approaches.
The Group’s competitive advantage begins with scale. Hyundai Motor and Kia collectively sell more than 7 million vehicles annually across approximately 190 countries, providing a foundational advantage for autonomous driving data collection. The group currently operates approximately 40 dedicated data collection vehicles around the clock to gather real-world driving scenarios, including challenging edge cases that are difficult to reproduce in controlled environments.
Hyundai has adopted a dual-track strategy to balance near-term deployment with long-term independence. The first track involves collaboration with NVIDIA, focusing on accelerated mass production of Level 2+ and Level 2++ autonomous driving capabilities for production vehicles beginning in mid-2028. The second track continues development of Atria AI, the group’s proprietary autonomous driving system, with Level 2++ production vehicles targeted for late 2029.
According to Minwoo Park, president of Hyundai Motor Group’s Advanced Vehicle Platform Division, autonomous driving competitiveness now depends on how rapidly companies can connect data collection to learning, validation and performance improvement. The group has integrated several technologies into its Data Flywheel to accelerate this cycle, including hard example mining, which automatically identifies challenging driving situations; continuous training pipelines that repeatedly incorporate new data; and virtual validation technology that reconstructs real-world scenarios in three-dimensional environments for repeated testing.
The group also operates a “Follow-the-Sun” development model, connecting development centers in South Korea and the United States to enable continuous 24-hour improvement cycles. A newly integrated Special Event Recorder automatically captures and stores significant events during autonomous driving for AI training purposes.
Beyond conventional autonomous driving, 42dot—a subsidiary focused on autonomous vehicle technology—is developing Vision-Language-Action (VLA) technology, which combines visual recognition, language-based reasoning and action generation. The approach is designed to improve decision-making in complex driving scenarios by enabling AI models to reason through situations using language-based understanding. Currently in simulation-based validation, the group plans to begin real-vehicle testing in late 2026 through early 2027.
In partnership with South Korea’s Ministry of Land, Infrastructure and Transport, Hyundai plans to deploy an Atria AI-equipped autonomous vehicle in Jeonnam-Gwangju Special Metropolitan City by year-end to conduct real-world Level 4 autonomous driving validation. This pilot project will operate in actual Korean road environments with complex traffic dynamics, with findings fed directly back into the Data Flywheel.
