Will Commercial Cars See With Laser-Powered Eyes?

Today only Google's car uses LIDAR; with this new system, one day, all cars may well have it

2 min read
Will Commercial Cars See With Laser-Powered Eyes?
Images: Niels Quack & Behnam Behroozpour

Laser-based radar offers a lot to smart cars, but it's too bulky and expensive for anything but Google's cost-be-damned robocar. Things would be different if you could make it small and cheap.

That's what Behnam Behroozpour, a graduate student, and his colleagues at the University of California at Berkeley propose to do. They are trying to pack the entire 3-D imaging system, known as LIDAR, into a solid-state package that could easily fit inside a smart phone or a game-playing wand. 

"We have in mind a 3 millimeter by 3 millimeter silicon chip, with the photonics integrated; a CMOS chip for the electronics; and a VCSEL [vertical cavity] laser, also as a chip," he tells Spectrum. He adds that he and his colleagues have tested the individual parts but are only now beginning to put them all together. Right now, the experimental system is about as big as a Microsoft Kinect box. 

He doesn't yet know the precise specs of the integrated system, but in its first iteration the beam should be able to see about 10 meters distance. Although that's not enough for some automotive applications, the Berkeley team is already thinking about how to reach out 30 or even 100 meters.

LIDAR works by sending out a beam and timing the return of its reflection. To guard against mistaking ambient light for part of that reflection, you can systematically vary the power of the beam; better yet, you can vary the frequency. Such frequency modulation is managed by tiny vibrating MEMS mirrors fabricated right in the silicon chip. 

The researchers electronically tune the frequency to exploit the natural vibration, or resonance, of the mirrors. This trick saves a lot of power. "It's more complicated, but you are gaining in the signal-to-noise ratio," Behroozpour says. Noise, in this case, comes from stray light rays that might be mistaken for reflections.  

Today's robotic cars depend on a lot of overlapping sensors—lasers, radar, ultrasound, standard cameras, stereocameras, GPS, inertial guidance, even WiFi.  Radar is now the key element, in part because it can work in any light and in any weather, in part because the size and cost has come down a lot, thanks to the aerospace industry. But if you could scatter cheap LIDAR sets around the car, in inconspicuous places, you could take advantage of its high resolution and capacity for orderly, focused scanning.

To do that, you have to cut the cost. Google Car's set is reported to cost US $70 000. Behroozpour says he expects his first integrated system to cost a few hundred dollars. Even that might be a bit pricey for smartphones and game controllers; then again, it isn't bad for the first-ever price tag of a solid-state system.

And in the solid-state world, cost follows a familiar trajectory: whatever goes down, goes way down.

The Conversation (0)

Chinese Joint Venture Will Begin Mass-Producing an Autonomous Electric Car

With the Robo-01, Baidu and Chinese carmaker Geely aim for a fully self-driving car

4 min read
A black car sits against a white backdrop decorated with Chinese writing. The car’s doors are open, like a butterfly’s wings. Two charging stations are on the car’s left; two men stand on the right.

The Robo-01 autonomous electric car shows off its butterfly doors at a reveal to the media in Beijing, in June 2022.

Tingshu Wang/Reuters/Alamy
Purple

In October, a startup called Jidu Automotive, backed by Chinese AI giant Baidu and Chinese carmaker Geely, officially released an autonomous electric car, the Robo-01 Lunar Edition. In 2023, the car will go on sale.

At roughly US $55,000, the Robo-01 Lunar Edition is a limited edition, cobranded with China’s Lunar Exploration Project. It has two lidars, a 5-millimeter-wave radars, 12 ultrasonic sensors, and 12 high-definition cameras. It is the first vehicle to offer on-board, AI-assisted voice recognition, with voice response speeds within 700 milliseconds, thanks to the Qualcomm Snapdragon 8295 chip.

Keep Reading ↓Show less