Building a self-driving car, it turns out, is a bit like planning a wedding. No matter how much time you give yourself, you risk being overwhelmed by the sheer number of things that need doing. Find a venue. Design the interior. Pick your signature music. Rebuild a global supply chain.

In 2016, when BMW said it would deliver fully self-driving cars, as part of a ride-hailing service, by 2021, it seemed to have plenty of time. But now it has just three years left; in an industry where developing a new car can take seven years, it’s a good thing the automaker has gotten around to picking its lidar supplier.

Today, BMW struck a deal with industry supplier Magna, and Magna’s partner Innoviz, to provide the lidar laser scanners for its self-driving cars. The companies provided scant details on the deal, so it’s not clear how BMW plans to use the sensor, but you can expect to see it in the German giant’s full-on robo-cars, as well as in the semi-autonomous features it’s developing to compete with systems like Tesla’s Autopilot and Cadillac’s Super Cruise.

This announcement underlines the importance of lidar in this space, and adds to a growing list of automaker-lasermaker couplings. General Motors bought a lidar startup called Strobe. Argo AI, which is making an autonomous system for Ford, scooped up Princeton Lightwave. Toyota has signed on with Luminar, run by 23-year-old photonics whiz Austin Russell. And Velodyne, the first company to make lidar for robo-cars, is still a major player, putting out 10,000 units a year. Google’s sister company Waymo has poured years and millions of dollars into developing its own proprietary system.

All lidar systems read the world around them by shooting out pulses of light and measuring how long they take to bounce back after hitting nearby objects. They all aim for a range of at least 200 meters, far enough for a car to spot an obstacle and hit the brakes, even at highway speeds. That’s about it for similarities. Some move those lasers around by spinning, others use no moving parts at all. The number of lasers ranges from one, to four, to 128. Some say their units cost less than $1,000, other units approach six figures.

Any given lidar system has a particular combination of range, reliability, scalability, and cost. And it’s the cost that has proved the hardest obstacle to overcome. Velodyne’s top-of-the-line unit, which uses 128 lasers, costs $75,000. In a robo-taxi service, you might be able to amortize that cost over time. In a car you’re selling to people, that’s impossibly expensive.

Innoviz says its lidar will cost “cost in the hundreds of dollars” thanks to its solid state design, an increasingly popular approach among the dozens of lidar companies that have popped up in recent years.

“It has to be low cost,” says Innoviz founder and CEO Omer Keilaf. “There’s no way around it.” The Innoviz lidar uses a tiny mirror (just a few millimeters across) to move the laser beam this way and that, instead of the more mechanically complicated spinning setup pioneered by Velodyne. That makes it simpler and cheaper to build, and easier to make robust enough for life on a bumpy road.

And where Luminar, for example, uses lasers at the 1550 nanometer wavelength, Innoviz stuck to the more conventional 905 nanometers. That limits the distance it can see, but allows Innoviz to use components made of silicon, rather than the far more expensive indium gallium arsenide. Still, Innoviz can spot items that reflect just 10 percent of light (think someone wearing all black) from 200 meters away. Keilaf says that’s possible thanks to improvements to the detection system, which measures the beams of light as they bounce back to the sensor.

Innoviz also does the work of translating that data into what the car needs to know about the world—that this cluster of points is a cyclist, and that one over there is a tree. Technology companies like Waymo and startup Aurora may have no problem doing that kind of work themselves, but Keilaf says it’s a value add for automakers suddenly dealing with reams of laser data. “They don’t know what to do with it otherwise.”

Innoviz now has a bigtime customer and a validation of a product it has spent two years building. BMW now has just one less item to take care of before the big day.


Pew! Pew!

LEAVE A REPLY

Please enter your comment!
Please enter your name here