ETRI to Commercialize 'AI Autonomous Driving' Learned from Real-World Driving Data

Domestic research institutes and private companies have agreed to mutually cooperate on the development and commercialization of next-generation end-to-end (E2E) autonomous driving technology. The core of E2E autonomous driving technology lies in enabling artificial intelligence (AI) to learn driving strategies from real-world road driving data, allowing vehicles to drive themselves.

The Electronics and Telecommunications Research Institute (ETRI) announced on March 17 that it will collaborate with KG Mobility and autonomous driving specialist Sotis Co. to develop AI-based autonomous driving technology. This effort will utilize large-scale real-world driving data jointly collected by the three parties, aiming to enable integrated learning across the entire process from sensor perception to driving judgment and vehicle control.

(From left) Chanho Kang, CEO of Sotis Co., Ltd., Yongil Kwon, Head of Division at KG Mobility, and Jeongdan Choi, Director of ETRI, are taking a commemorative photo after signing a memorandum of understanding to mutually cooperate in the development of 'E2E Autonomous Driving AI Technology.' Photo by Korea Electronics and Telecommunications Research Institute

Ultimately, the goal is to commercialize the technology by applying it to actual vehicles. The intent is to use real road driving data, enabling AI to comprehensively understand the road environment in a manner similar to humans, autonomously controlling vehicle steering, acceleration, and deceleration. This “Driving Intelligence” will be expanded to a level suitable for application in industrial settings.

Conventional autonomous driving technologies generally process perception, judgment, and control functions sequentially through separate systems. However, ETRI, together with KG Mobility and Sotis Co., plans to develop an E2E autonomous driving technology where a single integrated AI model understands road situations and simultaneously determines vehicle steering, acceleration, and deceleration.

In particular, ETRI expects this research to serve as a milestone, advancing core AI technologies—secured through national research and development projects supported by the Ministry of Science and ICT, the Institute for Information & Communications Technology Planning & Evaluation (IITP), and the Korea Autonomous Driving Innovation Foundation (KADIF)—by combining them with real vehicle data to meet industry demands.

The research will incorporate ETRI’s previously developed ▲ “reinforcement learning-based autonomous driving AI software,” which predicts hazardous situations and determines optimal driving behavior in congested road environments, and ▲ “perception and judgment AI software,” which enables stable driving in adverse or unpredictable road conditions.

Additionally, ETRI explained that it has devised a new approach that combines camera-centric visual information with AI’s logical reasoning capabilities, aiming to develop an intelligent driving model capable of stable autonomous driving with a minimal sensor setup.

In this research, KG Mobility will provide real vehicle driving data and test infrastructure, while ETRI and autonomous driving specialist Sotis Co. will be responsible for developing next-generation end-to-end autonomous driving software and building integrated systems. ETRI plans to maximize technological synergy by collaborating with the two companies throughout the entire process—from technology development and demonstration to industrial application.

Jeongdan Choi, Director of the AI Robotics Research Division at ETRI, stated, “Recently, global competition in this field has intensified, with Tesla expanding its E2E-based autonomous driving software (FSD) and NVIDIA announcing its autonomous driving AI foundation model. Under these circumstances, this research is meaningful in that a national research institute and a complete vehicle manufacturer are cooperating to strengthen global competitiveness in autonomous driving technology.”

He added, “The joint research team plans to gradually advance through data collection in real road environments, AI model training, and actual vehicle demonstrations. Going forward, we will also work to ensure that this technology evolves into ‘general-purpose mobility intelligence AI technology’ applicable to a variety of mobility platforms, including automobiles, robots, and drones.”

This content was produced with the assistance of AI translation services.