Quanergy’s lidar sensors help autonomous cars ‘see’ around them
Two Mercedes-Benz cars sit side by side in a garage at Quanergy Systems’ Sunnyvale headquarters. One, a black sedan, has a scaffold-like structure on its roof holding two canister-shaped sensors. The other, a cherry-red SUV with a stylized “Q” on the hood, also has two sensors — but they are hidden behind a panel on the front bumper.
The car with the awkward canisters is the “before,” while the SUV that has small, hidden sensors is Quanergy’s “after” version of lidar, a laser form of radar that helps self-driving cars perceive what’s around them. Quanergy says its lidar sensors are smaller, cheaper and more reliable than those of rivals. It says it is the first company to produce a solid-state lidar that doesn’t need to rotate, meaning it’s less likely to break down.
“Our sensors are higher performance in range, accuracy and resolution,” said Quanergy CEO and co-founder Louay Eldada, 50, who has a doctorate in optical engineering and founded the company, his fourth startup, in 2012.
Lidar bounces laser pulses off objects and measures how long it takes for them to return, sending out hundreds of thousands of beams per second. Since the system knows the speed of light, the reflection time lets it calculate how far away objects are, creating a detailed 3-D map of surroundings within a 650-foot radius.
Quanergy’s devices are sensitive enough to tell the difference between a tumbleweed and an armadillo, or between a floating plastic bag and a rock.
“We can see not only the shape of an object, but how it behaves from frame to frame,” Eldada said. “We can tell whether someone intends to cross the street, for instance.”
To aid in that, in August Quanergy acquired Raytheon’s Otus People Tracker, software that uses human perception algorithms to figure out intention. It’s used in battlefields to tell friend from foe by analyzing their behavior, Eldada said. Two Quanergy sensors in the company’s reception area capture people’s movements, which are displayed in colorful scribbles on an overhead monitor via Otus.
Lidar, which is either short for “light detection and ranging” or just a mashup of “light” and “radar,” depending on which sensor geek you ask, is poised to be a massive industry as carmakers and tech companies work feverishly to make self-driving cars a mass-market reality. That’s why Quanergy, which has raised $150 million, has a private market valuation of $1.59 billion.
“Lidar is an active sensing technology which makes it a useful complement to cameras,” said Nidhi Kalra, senior information scientist at nonpartisan think tank Rand Corp. and director of Rand San Francisco. “Cameras are passive, because they just receive light coming in through their aperture. They cannot always distinguish between an object and a picture of an object, for instance.”
But every sensor has a weakness. For lidar, rain and fog are problematic. “If the laser hits a raindrop or puddle, it can bounce in the wrong direction,” Kalra said. That’s why most autonomous vehicle makers are taking a “belt and suspenders” approach of multiple sensors — lidar, radar and cameras being the big three.
But unlike cameras, lidar isn’t limited at night. “The light that lidar senses is the light it sends, so we see very well in pitch darkness,” Eldada said.
Until recently, a lidar system carried a $75,000 price tag, making it impractical for widespread use. It’s most commonly seen as spinning bucket-like objects mounted on the roof of a self-driving car, giving them a distinctive if ungainly appearance.
Quanergy’s S3 lidar for transportation, the size of two decks of cards stacked atop each other, will initially sell for $250, and eventually even less, Eldada said. It will be integrated seamlessly into cars, as in the test Mercedes at Quanergy headquarters. Sensata, a former Texas Instruments subsidiary, will start producing them in Massachusetts this year, although large-scale production won’t happen until 2018.
Quanergy also makes a smaller, $100 lidar called the S3-Qi for use in drones, security devices and robots. It will be manufactured by Singapore’s Flex Ltd. (formerly Flextronics) in a Milpitas facility and will be in full-scale production this year.
Since it does not spin, Quanergy’s S3 device only “sees” a 120-degree view, which means a car would need at least three of them, and possibly four for overlap to get full, 360-degree visibility.
Quanergy is mum about just who its customers are, although it says it works closely with virtually every automaker. Partners Eldada said he can mention are Renault-Nissan, Hyundai and Koito as well as auto parts giant Delphi Automotive and Mercedes, which are both investors.
Morgan Hill’s Velodyne LiDar Inc. is a major rival. Ford Motor Co. and China’s Baidu recently invested $150 million in the 220-person company. (Prior investment rounds were not disclosed.)
Last week Velodyne announced a San Jose “Megafactory” that it said will manufacture a million lidar sensors a year by 2018. Velodyne’s lidar sensors now cost about $8,000, but it said the price will be “hundreds of dollars” once the factory ramps up production.
Waymo, the self-driving company spun out of Google, now a part of Alphabet, originally used Velodyne’s lidar but now plans to build its own. Waymo CEO John Krafcik said in a keynote at the Detroit Auto Show this month that Waymo had brought its lidar system’s cost down to about $7,500. “As we look to scale, we will do even better, with the goal of making this technology accessible to millions of people,” he said.
There’s one notable exception to lidar’s fan club among makers of autonomous vehicles. Tesla CEO Elon Musk has rejected lidar as too expensive. “Good thing about radar is that unlike lidar ... it can see through rain, snow, fog and dust,” he tweeted in July.
“Cameras certainly are cheaper, but it’s a gamble to say that all the (autonomous) driving tasks can be accomplished with radar and cameras,” said Rand’s Kalra. The
fatal crash of a Tesla driving on Autopilot in Florida last year “likely could have been prevented if that vehicle had had lidar.” A National Highway Traffic Safety Administration probe of that incident recently closed without issuing a recall or fine.
http://www.sfchronicle.com/business...elp-autonomous-cars-10872896.php?t=eb0d8b2114
