[Market Pulse] Gwangju Autonomous Driving AI Pilot Must Establish a National Standard

Data Reliability and Local Industry Collaboration Are Critical
From Operational Design Domain to Accident Liability
500 km² Pilot Project Aims for Level 4 by 2027

The entire city of Gwangju is becoming a new proving ground for autonomous driving technology in Korea. The government is deploying about 200 autonomous vehicles across actual residential and commercial areas spanning 500.97 square kilometers, repeatedly collecting driving data, conducting AI learning, and re-evaluating to achieve Level 4 end-to-end (E2E) autonomous driving by 2027. Unlike conventional demonstration projects that are limited to specific routes or controlled test beds, this initiative targets the entire city, representing a fundamentally different scale and approach.

Gwangju encompasses everything from the structured roads of new towns to narrow alleys in old downtown districts, traditional markets, school zones, complex intersections, and construction areas. Realistic risk factors coexist, such as unexpected pedestrian crossings, motorcycle lane encroachment, illegal parking, and difficulties in traffic signal recognition. From a technology development perspective, this complexity becomes a ‘valuable asset’ because it allows for systematic acquisition of so-called ‘long-tail scenarios’ that autonomous driving technology will actually face.

However, deploying 200 vehicles and simply driving them extensively is not enough to claim success. The purpose of demonstration is not the distance traveled itself but rather the creation of evidence that proves safety. Evaluation should focus not on how many kilometers were driven, but on what hazards were encountered, how the system responded, how causes of failure were improved, and how the improvements were re-verified under the same conditions. For this reason, the government must establish ‘safety management standards for autonomous driving demonstration cities’ based on the Gwangju demonstration project.

First, the operational design domain (ODD) for each vehicle must be clearly defined. This includes specifying the range and limitations for autonomous operations, such as operating roads, speed, weather, time of day, communication conditions, and construction sections. Whether a vehicle successfully executes a minimum risk maneuver—safely slowing down and stopping when it goes beyond its operational limits—should also be a key assessment item. Recording only accidents is insufficient; information on near-accidents such as emergency braking, lane departure, operational halts, sensor recognition errors, manual intervention by safety officers, and assistance from remote control centers must also be tracked. It would be necessary to establish standard indicators, such as the number of system interventions per 1,000 km, success rates of responses by risk scenario, success rates of minimum risk maneuvers, and recurrence rates of identical scenarios after software improvements.

Public disclosure of information is also necessary to earn citizens’ trust. In this regard, Gwangju City should consider publishing a quarterly ‘Autonomous Driving Safety and Transparency Report.’ This report should include data on cumulative driving distance, operational areas, accidents and near-accidents, emergency braking and operational halts, system intervention status, and accident causes and corrective actions, while ensuring the protection of personal information and corporate trade secrets. When a problem occurs, the causes should be disclosed, along with an explanation of improvement measures taken.

A balance must be maintained between data utilization and personal privacy protection. The revised ‘Autonomous Vehicle Act,’ which took effect in June, allows businesses with temporary operation permits to utilize original video data containing personal information to enhance system performance and safety. Simultaneously, it prohibits personal identification and purposes beyond the original intent, while mandating technical and administrative safeguards and obligating data destruction. Gwangju must not interpret these legal exceptions as blanket permission for unlimited data usage.

The institutional pathway linking demonstration to commercialization must also be clarified. Up to now, autonomous driving policy has focused solely on opportunities for testing through regulatory exemptions and temporary licenses. Going forward, processes must be established so that demonstration results lead to performance certification, approval of suitability for planned operation areas, insurance enrollment, paid transport service permits, and adoption in public services.

An unmanned autonomous vehicle undergoing test operation in Sangam, Seoul. Rideflux

Issues of accident liability and insurance systems must not be relegated to after-the-fact discussions. The connection with local industries in Gwangju should be demonstrated through contracts and sales, and achievements should be measured based on local companies’ testing and certification, joint patents, supply contracts, new sales, and hiring of specialized personnel. It is only when demonstration vehicles not only operate in Gwangju but also incorporate the technologies and components of Gwangju-based companies that the project acquires true meaning as a regional industrial policy.

Alignment with international standards is essential. In June, the United Nations Economic Commission for Europe ‘WP.29’ adopted global regulations for fully autonomous driving systems, covering safety management frameworks, reliable testing, safety validation data, operational monitoring, and autonomous driving information recording. Gwangju's demonstration scenarios, data formats, cybersecurity, and software update management systems must all comply with these standards.

The Gwangju autonomous driving demonstration city is a ‘national project’ capable of shifting Korea's autonomous driving policy from test operations to safety verification and commercialization. What must remain are not just the number of vehicles or kilometers driven, but rather data that proves safety, transparency, growth of local companies, sustainable services, and a verification system recognized internationally. When the accumulated standards spread to other cities and facilitate overseas expansion for Korean companies, Gwangju will truly become the national standard for autonomous driving commercialization.

Seongyong Ha, Professor of Smart Mobility Engineering at Joongbu University (President of the Korean Society for Automobile Mobility Safety)

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