Korean automakers in the dark on mapless self-driving systems

A Tesla Model 3 vehicle drives using Full Self-Driving (FSD) in Encinitas, California [REUTERS/YONHAP]

 
Self-driving cars don’t navigate the same way as humans do: Most of them, including Hyundai's, use high-tech and expensive sensors paired with accurate map data to steer themselves.
 
But Tesla, with its vision-based autonomous driving technology, has been implementing a more human-like navigation system in its latest vehicles, removing the need for a map and sensors to achieve a cost-effective and possibly more powerful solution. 
 

Related Article

• Hyundai EVs to qualify for U.S. tax credits for the first time

 
Vision-based autonomous driving technology relies on cameras installed on vehicles to detect and interpret the surroundings, instead of relying on expensive components like ultrasonic and lidar sensors. Tesla’s autonomous driving technology, including its supervised full-self-driving feature, also does not rely on high-definition maps.
 
Korean automakers and tech firms have been focusing on map and sensor-based autonomous driving technologies, but they will soon have to decide if and how they will implement a vision-based self-driving system.
 
The key advantage of Tesla’s vision-based autonomous driving technology is that it relies less on maps, therefore cutting costs on drawing and maintaining a high-quality and highly updated map database. The technology also allows vehicles to better react in unexpected circumstances, including road construction, sudden environmental changes and unrecorded environments. Using a system that combines map data and lidar sensor data is prone to data conflicts, as the computer needs to potentially deal with two data sets saying different things.
 
Tesla, confident in its choice, has been completely removing ultrasonic sensors on its Model 3 and Model Y vehicles since 2022. Chinese companies have also quickly joined the trend, with Huawei and Xpeng developing mapless, vision-only autonomous driving systems, while Li Auto is going for a method that combines map data with a vision-based approach.
 
Hyundai Motor Group declined to comment on the automaker’s progress on vision-based autonomous driving technology. Kia and Genesis had previously announced the production of the Highway Driving Pilot, a non-vision-based level 3 autonomous driving system that doesn’t require the driver’s attention in certain circumstances, only to cancel the rollout before the official launch.
 
“Korea’s vision-based autonomous driving technology is far behind that of Tesla,” Prof. Park Chul-wan of Seojeong University’s Department of Car Engineering told the Korea JoongAng Daily.
 
The professor argued that Hyundai and other Korean car manufacturers would have to soon choose between developing vision-based technology from scratch or search for a partnership with Tesla or Nvidia.
 
The government’s privacy-related regulations may also pose a hurdle for camera-based technology, which is required to record and analyze all visual data it sees. Despite the Personal Information Protection Commission releasing an update to its policy allowing delivery robots and autonomous vehicles to utilize recorded videos for AI development, Prof. Park called for further deregulation.
 
“I think mapless autonomous driving technology can be the future,” Professor Park said. “If the technology is up to standards, it can react to changes in the environment much faster than the conventional technology; cameras are also far cheaper to replace than lidar sensors.”
 
“But Korean companies currently lag behind compared to the technologies in the United States and China, and we would have to act fast.”

BY CHO YONG-JUN [cho.yongjun1@joongang.co.kr]