Kakao Mobility is accelerating efforts to secure its own autonomous driving technology by upgrading performance through data learning. The company plans to secure quality data and advance its artificial intelligence (AI) through high-difficulty urban pilots and an expanded cooperation ecosystem, while building data and service infrastructure to lead future cooperation with global companies based on its proven platform technology and service operation know-how.
According to the information technology (IT) industry Tuesday, Kakao Mobility has recently been intensively securing urban driving data in the Gangnam-gu area, which has high traffic congestion, through the "Seoul Autonomous Vehicle Pilot Service."
For an autonomous driving model to respond to unexpected situations, it is essential to learn "edge case" data—such as jaywalking or sudden illegal parking—that has a low probability of occurrence but induces misjudgment. Given the nature of End-to-End (E2E) based autonomous driving, in which the vehicle perceives and judges the environment on its own, learning various unexpected situations determines the AI's reliability and safety more than long-distance driving in plain environments.
Accordingly, Kakao Mobility chose the area around Gangnam Station, which has the highest driving difficulty in Korea, as its pilot site. Real-time data collected by vehicle sensors (AV-Kit) is immediately reflected in the perception and judgment system through an "AI data pipeline." Such an integrated data management system is expected to become the key foundation of a "data flywheel (virtuous cycle structure)," in which large-scale real-world data leads to advanced AI technology, which in turn draws in quality data.
"We have monitored the technology status of various companies at home and abroad, but Kakao Mobility is one of the companies with unrivaled technology capable of stably implementing actual passenger transport services in the complex Gangnam urban area," stressed Kim Jin-kyu, head of Kakao Mobility's Physical AI division.
The autonomous driving solution co-development agreement recently signed with LG Innotek is also expected to enhance the completeness of this data circulation structure. The two companies will conduct joint research and development (R&D) to effectively secure the large-scale real-driving data that is key to advancing autonomous driving technology.
Specifically, LG Innotek will receive real-driving data from Kakao Mobility to develop an "autonomous driving sensing solution" that integrates camera, radar, and lidar modules. Kakao Mobility plans to use this solution to advance its "AI data pipeline," which automates the entire process of collecting, learning, and distributing autonomous driving data.
The industry sees that the advanced hailing-matching system and real-time navigation infrastructure that Kakao Mobility has built through "Kakao T" and "Kakao Navi" will become a strong competitive edge in future autonomous driving service stages. Even global top-tier big tech firms or automakers would need abundant domestic operation know-how to provide stable services in the Korean urban environment.
Based on these platform capabilities, Kakao Mobility aims to draw out organic cooperation with technology companies, automakers, and academia to lead the growth of Korea's overall autonomous driving industry ecosystem.
"The core of the E2E autonomous driving algorithm lies not in simple driving distance, but in how many edge case data points are secured and learned in the actual urban environment," a Kakao Mobility official said. "The data pipeline being advanced based on the Gangnam pilot and the expansion of the cooperation ecosystem will become a key technological competitiveness that leads future collaboration with domestic and overseas manufacturing and technology companies."
