Blackmore Lidar Sensors
Normally there are only four companies, plus Carnegie Mellon University, driving cars around Pittsburgh with funny looking racks of sensors on the roof.
But on this sunny Thursday afternoon, a Montana-based company was showing off its fancy sensors to journalists in Pittsburgh — literally dropping off a Post-Gazette team before picking up some television reporters — in a 2019 BMW X5 with tinted windows for added secrecy.
The company, called Blackmore, has a shiny, new toy for the autonomous vehicles industry and beyond: an improved lidar sensor that could improve computer vision.
Judging by this media tour across the country to show off some geeky sensors — plus the fact that an entire lawsuit transpired between Uber and Google’s self-driving unit, Waymo — lidar is a pretty huge deal.
The lidar market is huge. According to Yole Developpement, a French consulting firm, the global market was worth $1.3 billion in 2018 and is expected to rise to $6.4 billion by 2024. That’s amid many of the self-set timelines that self-driving companies have for their vehicles’ public deployment.
Over the next two or three years, Blackmore hopes to bring the cost of its “Autonomous Fleet Doppler Lidar” down from a $20,000 price point when it’s first commercially sold this summer to under $1,000.
Read More
And that would be saying something — market leader, Velodyne, is known to have sold units for upward of $75,000.
Spinning tops
No matter the price, the self-driving industry more or less requires lidar (which stands for light detection and ranging) in addition to cameras and radar.
Lidar units are pretty distinctive: they’re the roof-mounted spinning tops that make Uber, Argo AI, Aurora Innovation and Aptiv vehicles so easy to spot on city streets. It’s why you turn your head.
Even the lidar that these companies use has limitations, though.
Lidar sends out pulses of light in waves and measures how long it takes for that signal to bounce back. It’s a similar idea to how bats use echolocation to see, by waiting for sound waves to bounce back.
Current lidar sensors cannot directly measure the velocity of a bike riding around in front of an autonomous vehicle. The lidar can’t collect speed information in one single pulse of light, which slows down reaction time for the car.
It has to take multiple measurements — and missing even one millisecond could be a safety hazard.
“Traditional lidar certainly scans the objects and sees them in 3D space, but to measure the speed it has to see how it moves from frame to frame and that’s a lot harder problem,” said Philip Wipf, a software engineer at Blackmore.
Looking to Doppler
The young startup, based in Bozeman, Mont. is betting on a new version that has improved range and resolution over competitors in the sensor space.
It depends on the same physics principles that make Doppler radar so popular for weather prediction, said Scott Fosgard, spokesman for Blackmore and the brains behind the Pittsburgh tour.
The key differentiation is in the lidar’s range and resolution, he said.
“Along with measuring range for each point, we’re also measuring the speed, the velocity at which it’s moving away from us and that’s the doppler effect,” Philip Wipf, a software engineer for Blackmore, said from the backseat of the BMW.
So when a laser beam hits a bus in front of the self-driving car, its frequency changes based on the velocity at which that signal returns.
The difference is that this lidar doesn’t just show you where something is, but how quickly it’s actually moving. And the more data, the better, says CEO Randy Reibel.
While Blackmore’s gray BMW is equipped with four lidar sensors, at the moment, the idea is to eventually fuse them all into one component — and that could allow the lidar to take speed measurements with a single pulse of light. That could mean the difference between spotting or not spotting the kid who runs into the street to grab a baseball.
Most lidar systems have a reliable range of about 100 to 200 meters, Mr. Fosgard said, but Blackmore’s lidar can detect objects up to 400 meters away.
Driving the North Shore
During a drive around Pittsburgh, the Blackmore BMW took a stop at the dead end of Mazeroski Way on the North Shore. In the back of the car, the company has two split-screen monitors that show the views of their own lidar versus a common off-the-shelf sensor.
The Bill Mazeroski statue was incredibly detailed in the Blackmore image. On the other side of the screen, a point cloud created a discernible shape that a car definitely couldn’t miss, but it was unclear exactly what it was.
And from that point, looking across the Allegheny River, it was easy to see blips of green across the water — cars driving down Fort Duquesne Boulevard, about 300 meters away, Mr. Wipf said. On the other side of the screen, there was one spot that seemed to look like a car, but it wavered in and out of view.
Blackmore’s lidar unit is still a prototype, but the company will be rolling out some of its sensors for commercial sale in the summer at around $20,000 each. The hope is to eventually bring the cost down to under $1,000.
The company currently has at least $21.5 million in funding from companies like BMW and investment funds like Toyota AI Ventures.
A race to improve
Blackmore isn’t the only lidar player betting on improved lidar to enhance computer vision for customers in robotics, self-driving, the military and more.
Waymo began selling its own lidar sensors at a lower price point than most of the industry — just not to competitors in the autonomous vehicle race.
Velodyne, the industry giant, sold lidar units for $75,000 just a few years ago. The price has already come down due to competition. The company has sold over 300,000 units to almost every self-driving company in the game, generating revenue of $500 million, according to Forbes.
Strobe Inc., based in Pasadena, Calif., has a similar idea, using “one pulse Doppler” to enable a range of about 300 meters and, in turn, better and faster decision-making. General Motors acquired that startup in 2017 for its subsidiary self-driving unit, Cruise Automation.
Headed by two former Apple engineers, Mountain View, Calif.-based Aeva is building a single-chip sensor that allows it to directly measure the velocity of objects in its path, unlike conventional lidar, which stitches together individual pulses of light, milliseconds apart.
Mr. Fosgard remains confident Blackmore will be the best commercial option, hence the trip out to Pittsburgh. He said that as CMU is known for its robotics, Bozeman is known for its optics firms.
“This is the new way of seeing,” he said.
Courtney Linder: clinder@post-gazette.com or 412-263-1707. Twitter: @LinderPG.






Comments
Welcome to Post-Gazette subscriber-only commenting community. Please review our community guidelines.