200 Autonomous Vehicles to Roam All of Gwangju... Alleyways and Commute Routes Included in Testing

First Citywide Opening in Korea
Company Applications Invited Next Month
"Opportunity to Narrow the Gap with the US and China"

The government has designated the entire city of Gwangju as an autonomous driving demonstration zone and will deploy 200 dedicated vehicles. This is the first time in Korea that an entire city, rather than specific routes or districts, will be opened up to autonomous vehicles.

On January 22, the Ministry of Land, Infrastructure and Transport announced the "Plan for Promoting Autonomous Driving Demonstration Cities." The initiative aims to enhance Korea's competitiveness in autonomous driving technology, which currently lags behind the United States and China, by conducting large-scale demonstrations on public roads.

Lagging Behind the US and China: "At an Elementary School Level"

Cumulative Autonomous Vehicle Test Distance as of July Last Year (Left), 2024 Autonomous Driving Competitiveness Rankings. Ministry of Land, Infrastructure and Transport

This plan stems from a sense of crisis that Korea's autonomous driving technology is falling behind in global competition. According to the Ministry, in the autonomous driving competitiveness rankings, Waymo of the United States ranks first, Baidu of China ranks second, while Korean companies are in 11th place. In terms of cumulative test mileage, Waymo has reached 160 million kilometers, Baidu 100 million kilometers, while Korea's total is only 13.06 million kilometers.

The background of this technological gap lies in differences in capital strength. Baidu has invested 36 trillion won in autonomous driving, Waymo 16.2 trillion won, but Korea's leading company, Autonomous A2Z, has invested only about 82 billion won. The Ministry of Land, Infrastructure and Transport analyzed that since Korea's autonomous driving industry is centered on startups, there are limitations in securing expensive graphics processing units (GPUs), attracting artificial intelligence (AI) talent, and increasing the number of vehicles.

Minister Kim Yoonduk of the Ministry of Land, Infrastructure and Transport stated, "If the US and China are at an adult level in autonomous driving technology, we are at an elementary school level," adding, "This is our last chance to overcome the technology gap."

Alleyways and Commute Routes Also Included in Demonstrations

Overview of Autonomous Driving Demonstration City. Ministry of Land, Infrastructure and Transport

Gwangju was selected as the first demonstration city because, as a metropolitan city with a population of over 1.3 million and a mix of urban and rural characteristics, it enables large-scale demonstrations in diverse environments. While 55 pilot operation zones have been designated nationwide, most are focused on specific routes, making it difficult to learn from various situations such as intersections, alleyways, and rush hours.

The demonstrations will be expanded in phases. This year, they will begin in new towns and suburban areas of Gwangsan-gu, Buk-gu, and parts of Seo-gu, where traffic is light and road infrastructure is well developed. Next year, the coverage will expand to all areas of Nam-gu, Dong-gu, and Seo-gu, including major destinations such as university hospitals and the old downtown area.

Regulatory exemptions will also be granted so that autonomous driving companies can focus on technology development. The city of Gwangju plans to apply for a regulatory sandbox to comprehensively resolve regulations related to AI learning using original video data, school zone demonstrations, and remote control.

Three Companies to Be Selected, 200 Vehicles to Be Distributed According to Capability

Step-by-step demonstration goal management. Ministry of Land, Infrastructure and Transport

The Ministry of Land, Infrastructure and Transport will invite applications from companies for about a month starting early next month and select around three companies by April. The target is domestic autonomous driving companies capable of large-scale demonstrations at Level 3 or above. After document screening and real vehicle driving evaluations at K-City, 200 dedicated vehicles will be distributed according to each company's technical and operational capabilities.

Unlike the previous approach of reverse engineering and modifying finished vehicles, these dedicated vehicles will be provided as software-defined vehicles (SDVs) optimized for autonomous driving technology. Steering, braking, power, and communications are designed with redundancy to prepare for unmanned autonomous driving.

In the first phase of the demonstration, a test driver will be seated in the driver's seat to develop full-route autonomous driving and remote control technologies. In the second phase, the driver will move to the passenger seat, and in the third phase, the demonstration will transition to fully unmanned autonomous driving to verify service commercialization. If the same type of accident is repeated due to technical issues or there is no improvement over a certain period, the number of vehicles operated by the company in question will be reduced.

Support Provided for GPUs, Insurance, and Personnel

The government will also support the necessary infrastructure for data learning. In cooperation with the Gwangju National AI Data Center, 200 dedicated H100 GPUs for end-to-end AI learning will be provided, and the government is considering allocating more than 2,000 out of 5,000 B200 GPUs from the national project to autonomous driving.

An operational system for 24-hour operation will also be established. The 200 vehicles will be managed in fleets, and in the early stages of the demonstration, more than 500 personnel will be deployed, including 400 test drivers working in two shifts, as well as staff for control, emergency response, and vehicle management.

A dedicated insurance product is also being developed with the goal of launching in May to reduce the burden on companies in the event of an accident. This general insurance will combine automobile, product liability, AI, control, and cybersecurity coverage, providing compensation to victims first and then allocating liability based on the cause of the accident.