Hyundai Motor Group is set to lead the autonomous driving technology sector with a focus on artificial intelligence (AI), planning to mass-produce Level 2+ autonomous vehicles by 2028. The company aims to usher in a new era of domestic autonomous driving using its Artria AI program and the necessary hardware technology. It has also begun establishing a 'data flywheel' based on a global production capacity of over 7 million vehicles annually.
Mass Production of Autonomous Vehicles by 2028... Developing In-House AI
On September 11, Hyundai Motor Group held a 'Media Day for Autonomous Driving' at its 42dot headquarters in Seongnam, Gyeonggi Province, where it unveiled its autonomous driving development strategy and production roadmap.
The mass production of autonomous vehicles will utilize a 'two-track' strategy, combining external technologies from NVIDIA with in-house development. This approach aims to accelerate the production timeline while internalizing its own AI technology.
Initially, Hyundai plans to apply NVIDIA's Hyperion platform to produce Level 2+ features in mass-produced vehicles by the first half of 2028, followed by the launch of Level 2++ vehicles in the second half of the same year. By the second half of 2029, the company intends to mass-produce Level 2++ vehicles based on its own Artria AI, gradually enhancing autonomous driving technology using real-world data. Level 2 autonomous driving allows the vehicle to perform steering and acceleration while the driver supervises the driving situation.
Park Min-woo, President of Hyundai Motor and Kia's Autonomous Vehicle Platform Division and CEO of 42dot, stated, "Just as it is hard to imagine a car without wheels, autonomous driving will become an essential competitive advantage for vehicles in the future. While it is difficult to pinpoint whether that moment will be next year or the year after, the current pace of technological advancement and market movements suggest we are much closer than previously expected."
The core of Hyundai's autonomous driving technology is AI. Unlike traditional modular approaches that develop perception, decision-making, and control functions separately, Hyundai is developing an end-to-end (E2E) system that integrates these processes based on AI. The E2E system allows AI to process information from sensors like cameras and radars, understand the surrounding environment, and directly connect this understanding to driving decisions and vehicle control.
Additionally, Hyundai announced the commencement of research on next-generation autonomous driving technology called Vision-Language-Action (VLA). VLA combines visual information captured by the vehicle with language-based reasoning to inform actual driving actions. While the existing E2E system connects driving information directly to vehicle actions, VLA enhances this process by incorporating language-based reasoning, demonstrating higher judgment capabilities in complex environments.
Lee Hee-seok, group leader of the VLA model research area, noted, "By basing actions on language, we can improve accuracy and stability. This will also help build trust in the system by allowing the vehicle to explain its reasoning to users."
Data Flywheel: The Key to Advancing Autonomous Driving
Hyundai presented the data flywheel as a key foundation that connects its two-track strategy into a competitive advantage. The data flywheel is a virtuous cycle where data is collected from actual vehicles, AI learns from this data to improve autonomous driving technology, and the new insights are then applied back to the vehicles to gather more data.
Hyundai operates over 40 data collection vehicles day and night in areas like Pangyo, gathering data on edge cases—exceptional situations that are difficult for AI to learn in advance. This includes various real-world scenarios such as road construction, severe weather, sudden lane changes, and parked vehicles on side streets, all of which contribute to enhancing autonomous driving technology.
During the event, Hyundai showcased footage of an SDV testbed vehicle equipped with Artria AI navigating through Seoul's urban environment. The vehicle demonstrated its technology in complex road conditions, including busy rush hours in Gangnam and heavy vehicle traffic in Jamsil, as well as rainy conditions in Pangyo.
Jung Sung-kyun, group leader of the Artria division at 42dot, emphasized, "Once mass production of the autonomous driving system begins, we expect to surpass the amount of data accumulated by competitors within five years."
Hyundai's global production scale serves as a competitive advantage for the data flywheel. The company sells over 7 million vehicles annually across more than 190 countries and regions. The 'Special Event Recorder' (SER), which automatically logs key situations during the autonomous driving process, is also expected to expedite the standardization of data collection for the data flywheel.
Hyundai is also developing technology to efficiently operate large-scale AI models in vehicles. The company explained that it has secured quantization technology that allows autonomous driving models to be trained using large-scale graphics processing units (GPUs) while also functioning within the limited computing environment of actual vehicles. This aims to create a virtuous cycle that enhances the scale, quality, and operational efficiency of data for autonomous driving.
Park Min-woo concluded, "The competition in autonomous driving is no longer about distinguishing the superiority of specific features. Hyundai Motor Group will develop trustworthy autonomous driving technology based on a virtuous cycle of data, AI, and repeated validation."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.