Korea must urgently secure data and collect edge cases to catch up with U.S. and Chinese autonomous driving technology as it targets Level 4 commercialization by 2027, experts said at a parliamentary forum.
The discussion came as Gwangju pursues Korea's first city-wide autonomous driving demonstration project. Participants emphasized that building operational and service models in advance is crucial for accelerating commercialization.
"Edge cases refer to difficult situations that are hard to predict under normal circumstances," said Choi Jun-won, professor at Seoul National University's Department of Electrical and Computer Engineering, at a policy forum held at the National Assembly on the 11th. "Ultimately, the core of autonomous driving technology depends on how well it responds to edge cases."
The forum, titled "AI Autonomous Driving Demonstration City: From Technology to Service," was hosted by Democratic Party lawmaker Jung Jin-ho and Kakao Mobility. Representatives from Hyundai Motor and Nvidia also participated as panelists.
Choi noted that acquiring "as much data as possible" is essential for improving edge case response capabilities. The government has designated autonomous driving pilot zones in 55 cities and provinces nationwide, but their limited scope and few participating domestic companies have yielded minimal data collection benefits.
"For AI-based autonomous driving models, scaling laws apply—performance continuously improves as data volume increases," Choi explained.
Kim Kun-woo, head of Kakao Mobility's Future Platform Economy Research Institute, said remote monitoring, field response, dynamic routing, and passenger communication will become critical in the Level 4 robotaxi market.
"Kakao Mobility is exploring ways to automate remote monitoring using AI, based on our experience managing affiliated taxi fleets," Kim said.
Lim Wol-si, director of the Autonomous Driving Policy Division at the Ministry of Land, Infrastructure and Transport, said the government plans to develop service models in the final stage of the demonstration city project.
"We are also conducting research to launch the world's first insurance product exclusively for autonomous vehicles," Lim said.
The government plans to operate a regulatory sandbox in Gwangju to ease restrictions hindering demonstration projects. The Personal Information Protection Commission is also preparing measures to relax video data de-identification requirements, which have been considered obstacles to autonomous driving tests.
Industry officials said collaboration among government, academia, and businesses—combined with public acceptance—could help Korea close the technology gap with the U.S. and China.
"An era is coming when 200 vehicles can achieve the effect of 20,000 vehicles through AI technology," said Jung Suk-won, executive vice president at Nvidia Korea. "Korea has significant potential given its extensive experience in public transportation and taxi operations."
Kim Soo-young, vice president of Hyundai Motor's Mobility Business Division, said the company would provide vehicles and technology suitable for commercialization and called for public cooperation to improve autonomous driving services.
The government plans to select companies for the Gwangju demonstration project by April, divided into technology demonstration firms and cooperation model partners including platform operators, automakers, and insurers. Starting in October, 200 autonomous vehicles will be deployed in Gwangju in phases to begin full-scale testing.