▲Min Kyung-wook, Director of ETRI



"Narrowing the technological gap with advanced autonomous driving countries within five years."

AI algorithms process camera and lidar recognition information in real time to generate a route.
"Korea lags the US by five years in the cognitive, judgment, and control software technologies essential for autonomous driving."

[Editor's Note] As the movement toward commercializing fully autonomous vehicles by 2027 gains momentum, domestic autonomous driving technology development is also actively underway. Among many companies and institutions, the Electronics and Telecommunications Research Institute (ETRI) attracted attention last year by unveiling the driverless autonomous vehicle "Otobi." ETRI, which took the initiative in research and development to provide autonomous driving services to the general public with a vehicle without a driver's seat, developed core AI technologies that recognize and predict driving environments and situations based on the fusion of cameras and lidar to enhance the accuracy of autonomous driving sensors. It also provides AI learning and inference models to implement recognition and prediction functions. Accordingly, this magazine spoke with Min Kyung-wook, head of ETRI's Intelligent Robotics Research Center, about ETRI's unmanned autonomous vehicle technology and Autobee.

■ What are Otobi's core technologies or differentiating ideas?

The core technology of Autobee is the application of autonomous driving AI technology by integrating camera and lidar sensors. Autobee processes the information received from the camera and lidar in real time using AI algorithms applied to Autobee, recognizing the surrounding environment and objects and creating an autonomous driving path.

A distinctive idea is that, from a service perspective, the absence of a driver's seat allows the general public to experience autonomous driving more deeply. Finally, the vehicle is equipped with features such as voice recognition and VR displays in addition to autonomous driving, providing convenience while traveling.

I heard you're researching autonomous driving on irregular roads. Is there a difference between this technology and driving on regular roads?

Atypical environments are those where lanes and road markings are unclear. This includes unpaved environments like country roads and back roads where the road and sidewalk are not clearly defined and pedestrians and illegally parked vehicles are constantly present. These environments present challenges in implementing autonomous driving location recognition technology and in creating driving paths. In this environment, we are conducting advanced research focusing on artificial intelligence technology that can effectively perform autonomous driving.

■ Please explain the sensors installed in the autobi.

The Autovi vehicle is equipped with two cameras in the front and two lidars each in the front, side, and rear. Equipped with six lidars, it eliminates blind spots and enhances safety. The camera sensors detect road markings, such as lane markings, and traffic lights, and recognize whether the taillights of a stationary vehicle are hazard lights or reverse lights. The lidar is used to recognize dynamic objects (people, vehicles, etc.) in 3D using artificial intelligence.

The reason for the lack of radar is that Autobees don't drive at high speeds, and radar can't clearly distinguish between cars and people. Current law stipulates a maximum speed of 25 km/h for driverless autonomous vehicles for safety reasons, meaning they must maintain this speed even on public roads. Therefore, radar is unnecessary.

■ When the evasive command was issued, the vehicle was observed crossing the center line. While I understand it's designed to strictly adhere to road traffic laws, can this be interpreted as including flexibility depending on the situation?

Crossing the center line is generally illegal. However, in campus environments, back roads, and two-lane roads, it is implicitly permitted, and drivers are encouraged to avoid crossing the center line in these situations. Although it is technically possible to flexibly respond to each situation and environment by distinguishing between them, it is avoided on major roads because of the constant risk and strong legal sanctions.

■ As AI technology becomes central, we often hear talk of training data. What exactly does training data mean?

For example, AI learning is performed to better recognize objects ahead while driving, whether they are people, cars, buses, or trucks. Furthermore, to better detect not only the type of object but also its three-dimensional size, speed, and direction, a large amount of data must be built and trained to achieve safe autonomous driving technology.

Research and standardization efforts in autonomous driving technology are actively underway around the world. Where does Korea stand and what are our thoughts on the direction we should take?

The scope of autonomous driving technology is so broad that it's difficult to quantify the technological gap in a single word. However, the most crucial technology in autonomous driving is the perception/judgment/control software that enables safe vehicle operation. Artificial intelligence software is particularly crucial. I believe Korea is approximately five years behind the advanced US in this area of perception/judgment/control software technology. This gap is expected to narrow significantly by 2027, when the current Autonomous Driving Technology Development Innovation Project (2021-2027) concludes.

■ Were there any difficulties during research and development?

Since there is no accelerator, steering wheel or brake, functional testing is several times more difficult. While standard autonomous vehicles can simply apply the brakes in an emergency, the lack of brakes in Autobees makes them slower to respond to dangerous situations, a major challenge during research and development. Furthermore, the lack of a driver's seat also necessitates manual steering, requiring a joystick for maneuvering, which presents significant inconvenience.

■ A word to the readers

Self-driving services that citizens can experience are still a story of the future. Currently, some local governments are conducting pilot services for self-driving shuttle services, and in the medium to long term, I believe that many industries will be revitalized, such as robotaxis that citizens can experience, self-driving services linked to public transportation, and self-driving services in transportation-deprived areas. I ask for your continued encouragement and support.