LiDAR vs Camera... Which Technology Will Win in Autonomous Driving? [Im Juhyung's Tech Talk]

Sensors as the 'Eyes' of Autonomous Driving Technology
Lidar vs Camera...Development Companies Divided
Precise Lidar...High Cost and Technical Complexity
Using AI but...Camera Still Insufficient

LiDAR product installed in the vehicle / Photo by Yonhap News

[Asia Economy Reporter Lim Juhyung] From automobile manufacturers like Tesla, Volvo, and Hyundai to tech startups, numerous companies are fiercely competing to perfect autonomous vehicle technology. Although the goal is the same, the methods vary significantly. The biggest difference lies in the hardware, which can be considered the 'eyes' of autonomous vehicles. This is largely divided between LiDAR and cameras.

◆Automakers Trusting the Precision of 'LiDAR'

The sensors act as the eyes of autonomous vehicles. There are various types of sensors, but they can be broadly categorized into RADAR, LiDAR, and cameras. These sensors collect spatial information surrounding the vehicle, enabling it to avoid obstacles and find routes.

RADAR emits radio waves and measures the time it takes for them to bounce back from objects to determine distances, shapes, and sizes. LiDAR operates similarly but uses high-powered laser pulses instead of radio waves.

Advanced LiDAR splits laser pulses into multiple beams, allowing it to recognize not only the shape and distance of objects but also their height and width by scanning the surroundings with multiple lasers simultaneously. This enables LiDAR to create a detailed 'map' of its environment, known as '3D mapping.'

Google's autonomous driving technology research company 'Waymo' unveiled 3D mapping / Photo by Waymo official website capture

Because LiDAR can identify objects much more precisely than RADAR, it is considered an 'essential sensor' in autonomous vehicle technology, where safety is paramount.

Global automaker Volvo is developing autonomous driving technology using LiDAR, and Hyundai, a leading domestic automaker, is doing the same. Google's autonomous driving subsidiary Waymo also supports LiDAR technology.

◆Tesla's Approach: Cameras and AI to Find a Workaround

On the other hand, some companies argue that LiDAR is 'overhyped' and have introduced alternative sensors. Tesla, founded by CEO Elon Musk, is a prime example. Tesla uses cameras, which are much simpler and more common sensors than LiDAR. The system recognizes the surroundings by arranging eight cameras around the vehicle.

Of course, cameras cannot achieve the precise object recognition or advanced technologies like 3D mapping that LiDAR can. Instead, Tesla compensates for the lower accuracy with 'computer vision.' Powerful machine learning artificial intelligence (AI) analyzes the visual information collected by the cameras to identify objects. Tesla has developed a massive supercomputer called 'Dojo,' capable of processing 10 quadrillion operations per second, to train its autonomous driving AI.

Not only Tesla but also some startups are betting on technology that combines cameras with AI computer vision rather than LiDAR. Notable examples include 'Arrival,' an electric vehicle manufacturer invested in by Hyundai, and 'Wave,' which specializes in autonomous driving software development.

◆Expensive and Slow LiDAR vs. Still Unreliable Cameras

Why do some companies like Tesla reject ultra-precise sensors like LiDAR? Especially, CEO Musk publicly criticized LiDAR autonomous driving companies in 2019, calling LiDAR "a device for fools."

Musk's skepticism toward LiDAR stems from the following reasons: △the cost per LiDAR unit is too high, △its use of lasers makes it vulnerable to weather conditions, and △implementing complex technologies like 3D mapping takes too long. In contrast, he argues that the method of having AI analyze camera-collected information is much cheaper and faster in the long run.

An autonomous driving AI of a Tesla vehicle mistook the full moon in the sky for a "yellow light" in an unusual incident. / Photo by Twitter capture

However, the camera-based autonomous driving technology that Musk is heavily investing in also has weaknesses. Tesla's Full Self-Driving (FSD) feature, currently in beta service, is known to occasionally make erroneous decisions.

In July, it was reported that Tesla's system mistook a full moon in the evening sky for a yellow traffic light and attempted to slow down, drawing ridicule from netizens. Additionally, there were controversies when the system was seen charging toward parked cars or bushes.

Regarding this, according to U.S. economic media 'CNBC' and others, the U.S. federal independent agency 'National Transportation Safety Board (NTSB)' sent a letter to CEO Musk last month, pointing out that "Tesla provided beta services to customers before resolving issues with FSD's 'design defects.'"

◆"Two Different Solutions to One Very Complex Challenge"

So, which will complete autonomous driving technology first: LiDAR or cameras? In fact, many companies are investing in 'both.'

For example, the UK e-commerce giant 'Ocado' invested ?10 million (about 16 billion KRW) each in LiDAR-based and camera-based autonomous driving companies. While it remains to be seen in the distant future which is more suitable for autonomous driving, the strategy is not to miss out on either.

Regarding this, Ocado's technology developer Alex Harvey explained in an interview with a media outlet, "It's like the relationship between Android smartphones and iPhones," adding, "The two technologies are different solutions to one very complex challenge."