As Elon Musk made clear Monday, the technology most of his competitors in the self-driving car space use to help vehicles detect what's around them is lame.
And his option is way better.
"LiDAR is a fool's errand," he quipped about the laser-emitting tool that, in the simplest terms, acts as eyes for autonomous cars. "Anyone who is relying on LiDAR is doomed."
That's pretty much most of the businesses testing self-driving cars, including Waymo and Uber who went to court over LiDAR technology last year.
Musk let his strong opinions about LiDAR fly at Tesla's Autonomy Day, an event that gave investors at the company's Palo Alto headquarters and 50,000-plus livestream viewers an inside look at self-driving tech.
It's not just LiDAR that Musk's scrapped for Tesla's self-driving mode. In a conversation about deploying Tesla robotaxis, he added that HD maps -- or highly detailed maps that provide another layer of data for the machine-driven vehicles -- were just as useless as LiDAR. Again, most self-driving car trials use HD maps.
Instead, he finds cameras and their vision abilities best to see and interpret the world around cars -- much like human drivers. A major portion of the event was devoted to computer vision and how the driving world from streetlights to road signs is based on sight.
"You were not shooting lasers out of your eyes to get here," Tesla's head of AI Andrej Karpathy added, before calling LiDAR a "shortcut." Mimicking human driving as closely as possible is a major goal at Tesla. A radar sensor at the front of a Tesla does the work that the cameras can't do, such as seeing through fog, dust, or snow.
Chinese autonomous vehicle company AutoX tried to go all in on a camera-based self-driving system and like Musk had nothing kind to say about LiDAR. But that was 2017, and now its vehicles have at least one LiDAR sensor for redundancy and extra input. One of its self-driving vehicles picked up a burger for me at CES earlier this year; its LiDAR sensor noticeable on top of the car.
Meanwhile, British startup Wayve believes autonomy should be based on machine learning with data training the computer. Instead of multiple data sources like real-time object detection with LiDAR, the company says that information can come from viewing the road and other driving experiences through a camera. Although still early, the company says it only needs a GPS system, camera, and a powerful computer to teach cars to drive like humans. Wayve was slammed for saying so by experts who specialize in sensing technologies. One even told me it was "lunacy" earlier this month.
Musk went on to explain that he's not wholly anti-LiDAR, just when it comes to cars. His SpaceX team, for example, developed its own LiDAR equipment. But for cars, he finds Tesla's eight cameras, 12 ultrasonic sensors, and front-facing radar a price-friendly sensor suite suitable for autonomous driving. He called other LiDAR devices on cars, "expensive sensors that are unnecessary."
He also humble-bragged at one point about Autopilot's 100,000 automated lane changes per day without a human driver involved and zero accidents from those computer-driven moments.
As Tesla vehicles add more and more autonomy without LiDAR sensors, it's almost as though Musk sees it as proof it's "foolish" tech.
