Hyundai Motor Group said it plans to work with NVIDIA to speed the rollout of mass-produced Autonomous Driving cars while advancing its in-house Autonomous Driving artificial intelligence (AI), Atria AI. It also said that within five years of starting mass production of Autonomous Driving vehicles, it could overtake rivals in the volume of accumulated data and gain a competitive edge.
Hyundai Motor Group held the HMG Autonomous Driving Media Day at 42dot's headquarters in Sujeong District, Seongnam, on the 11th and unveiled this dual-track plan for Autonomous Driving development. It released a concrete Autonomous Driving strategy and roadmap about six months after recruiting Park Min-woo, Hyundai Motor Group Advanced Vehicle Platform (AVP) Deputy Minister and 42dot president, to lead Autonomous Driving development.
Hyundai Motor Group first plans to launch a mass-produced Autonomous Driving car at the level of Level 2+ that can drive on highways without hands on the wheel in the first half of 2028. This will be based on NVIDIA's Autonomous Driving car platform (a system bundling sensors, driving computers, and software), Drive Hyperion 10. It is also considering adopting NVIDIA's Autonomous Driving AI model, Alpha-Mayo.
Through this, it aims to speed initial market entry and accelerate the build-out of the virtuous data cycle (data flywheel) that is core to Autonomous Driving capabilities. As the number of cars grows, more data accumulates, and more data can rapidly improve AI performance. Rather than insisting on loading its own Autonomous Driving AI, Atria AI, it will first expand data-collection vehicles by leveraging external partnerships.
Kwon Jeong-hyeon, head of the HMG Autonomous Driving Development Center and lead of 42dot's AD Division, said, "Hyundai Motor Group sells more than 7 million vehicles a year in over 190 countries and regions worldwide," adding, "Right now we have around 40 data collection vehicles, but once mass production of vehicles equipped with Autonomous Driving systems gets into full swing, we expect to surpass competitors in the volume of accumulated data within five years."
Kwon also said, "The side that first completes the data loop from collection to learning, and from model training to model deployment, will have a competitive edge in automotive software and AI," adding, "That's why we are building a virtuous data cycle." Given that Hyundai Motor plans to start mass production of Autonomous Driving in 2028, the idea is to catch up with rivals' data volume and secure a competitive edge by 2033.
Securing data collection vehicles is directly tied to Hyundai Motor Group's in-house technology development, including Atria AI. Within Hyundai Motor Group, the AVP headquarters and 42dot are jointly developing Autonomous Driving technology. Initially, the technology was developed in a modular way by independently developing and then integrating each unit of driving, but two years ago it switched to an end-to-end (E2E) approach like Tesla, and it has now established a certain level of Autonomous Driving foundation.
On the day, Hyundai Motor Group also released urban driving footage of a software-defined vehicle (SDV) equipped with Atria AI. The video showed an Ioniq 6 with Atria AI continuing Autonomous Driving smoothly even in downtown Gangnam during rush hour and in downtown Pangyo in the rain. It also avoided a car stopped on the shoulder and handled sudden cut-ins and oncoming vehicles on narrow roads.
Hyundai Motor Group plans to advance Atria AI and roll out Level 2++ Autonomous Driving that does not require holding the steering wheel even in urban areas in the second half of 2029. President Park Min-woo said, "We could release Atria AI sooner, but we strategically delayed the timeline," adding, "Rather than rushing mass production to roll out a half-baked Level 2++, we decided to proceed with mass production using NVIDIA's technology while focusing resources on the final goal (Atria)."
To respond to a rapidly changing technology environment, Hyundai Motor Group is also developing a VLA model in which the car visually recognizes its surroundings, infers and understands situations in human language, and ultimately generates physical actions such as driving control directly. Because the VLA approach learns from language-described content, it has the advantage of being easier to improve than E2E models, which can learn only through visual recognition.
Park said, "By combining the capability to directly develop software and AI with the capability to implement that technology in vehicles and customer experiences at global scale, Hyundai Motor Group can build differentiated Autonomous Driving competitiveness," adding, "We will secure Autonomous Driving safety and completeness at a level customers can trust while learning and improving quickly."