+ A

Naver to 3-D map Seoul’s major roads this year

June 26,2019
이미지뷰
Naver Labs Ceo Seok Sang-ok shares the “A-City” plan in a press conference held Tuesday in central Seoul. [NAVER]
이미지뷰
Naver Labs’ R1 sensors used for 3-D mapping. [NAVER]
Naver Labs will release high-definition 3-D maps for self-driving cars that cover Seoul’s four-lane roads this year.

Machine-readable maps of Gangnam, southern Seoul, and Sangam-dong, Magok-dong and Yeouido, western Seoul, will be released by August. These areas have major roads with multiple lanes, making them the perfect place to test self-driving cars on their capacity to automatically change lanes.

The system works through a combination of aerial 2-D images taken from above and 3-D images shot by mapping sensors on the ground. The technology was showcased at the Consumer Electronics Show in January, but upgrades have been added that automate the mapping process to speed up the process of retrieving the 3-D images.

“Autonomous cars and artificial intelligence [AI] are at the verge of commercialization - we’re seeing walls break down between physical and virtual spaces,” said Naver Labs CEO Seok Sang-ok in a press conference held Tuesday in central Seoul.

As Korea’s largest portal site, Naver’s business realm until now was limited to online. Seok said that the company’s future direction was to expand its reach to physical spaces. In order to find ways to penetrate real-life places, developing technology to make machine-readable maps was of the utmost importance, he explained.

Since its spin-off from the portal site in 2017, Naver Labs has put its research focus on mapping, self-driving and robotics.

On Tuesday it introduced a plan that requires advancements in all three of those areas: the A City. Machines will run autonomously while AI and robots analyze location data to connect various infrastructure in the city, like logistics and delivery.

A new feature Naver Labs unveiled at the event was “visual localization,” which can pinpoint where a user is standing in an indoor facility with a single photo.

This is based on an older Naver technology that uses self-driving robots with sensors and cameras to draw indoor maps. The company also said it has upgraded its indoor 3-D mapping robot - M1 - to M1X by enabling it to measure ceilings and increasing its accuracy by 30 percent while lowering production costs.

“The essential idea of virtual localization is that a single photo can immediately pinpoint where you stand in indoor spaces that can’t be tracked through GPS,” said Seok.

“This year’s plan is to apply mapping and localization technology to outdoor passenger roads where there are so many more variables.”

Creating a detailed machine-readable map on passenger roads is difficult, Seok explained, because the system needs to perceive the same sidewalk regardless of variables like leaves, snow and the light at different times of day. Naver Labs will use robots developed by MIT through a project it funded for the project.

“Our physical life space is full of new opportunities but only companies with the right technology will grab them in the end,” Seok said.

BY SONG KYOUNG-SON [song.kyoungson@joongang.co.kr]