LIDAR point clouds including velocity from Doppler shift
Blackmore/Aurora
Aurora, the high-flying robocar startup founded by Chris Urmson of Waymo/Google, Sterling Anderson of Tesla and Drew Bagnell of Uber has used its super-high valuation money to purchase Blackmore, an FMCW LIDAR company based in Bozeman, Montana. That’s a strong expression of support for the value of LIDAR, in opposition to Elon Musk’s public declarations that it’s lame. Earlier this year, I rode in a car with the Blackmore sensor and was impressed.
ADVERTISEMENT
Blackmore’s technology is of a special new class that is the most capable form of LIDAR out there. First, it operates in the 1500nm thermal infrared band. This type of light does not transmit through water, so when it strikes your eye, it is not focused by the lens into a point on your retina. The result is you can emit much more power without risk of burning spots on people’s eyes. (Though, in rare situations, you can burn out a camera sensor as my colleague, Jit Ray Chowdhury, discovered.
Because you can put out more power, you can see much further. Regular near-infrared light, at much lower powers, just can’t be made bright enough to see things many hundreds of meters away. But almost all LIDARs use that because silicon chips can detect that form of light, and the world is very, very good at making low-cost silicon chips. Seeing longer wave light needs different materials that today cost a lot more because they are not made by the billions the way silicon chips are. Less power is also less power, which has some virtues.
The Blackmore LIDAR, unlike almost all LIDARs, uses an approach called FMCW to detect the “Doppler” shift of returning light, the way radar does with radio waves. From this, you can measure three things about the target:
- How far it is away from you
- How fast it is moving towards or away from you
- How reflective it is to this light — a deep red, low-resolution version of grayscale photography.
The second is the gift of Doppler and FMCW. This approach also enables very long range. I saw the Blackmore unit see large objects over 500m away from the car. Arguably this is as far as you need to see, though there is value in noticing problems even a mile away on the highway, particularly for trucking. Usually, LIDAR units like this are mounted only forward facing, to spot obstacles far ahead on the highway when going at fast highway speeds. Some teams, however, such as Toyota, have been buying them even to look to the sides, for the best visibility when trying to cross a busy road.
ADVERTISEMENT
LIDAR is already superhuman vision because it sees distance natively, and does it day or night. Doppler LIDAR sees speed, which is an immediate aid to figuring out what’s what, and very importantly, where it’s going. All robocar systems try to figure that out, but they do it the way humans do — watch something move over multiple scans. (They also have radars which also offer Doppler, but don’t tell you very accurately where the target is. The normal approach is to try to match radar targets to lidar and vision targets to learn this important speed information.)
Doppler’s not perfect. It doesn’t measure motion from side to side. That’s why Tesla’s LIDARs did not see the transport trucks moving across the road in front of them, resulting in the deaths of two Tesla owners. But it’s still a great boon. While you can figure out the speed of things if you are doing a good perception system, it still can take 300 to 400ms to do so with most LIDAR systems. Doppler gives you the speed in one axis with no delay. You could look at a single frame where you had two cyclists passing one another close. To regular LIDAR, if they are far away so you don’t see much detail, you will just see a blob at one distance. With FMCW you would know there were two things, going at different speeds, and know it immediately.
The price was not disclosed, but last year Blackmore raised $18M from BMW, Toyota and others.
Blackmore FMCW 1500nm band LIDAR unit — large today but sure to shrink in the future.
Blackmore
Aurora is young, but with a standout and experienced team, and it has quickly generated a double unicorn valuation for itself.
Aurora is making pretty much the opposite bet from Tesla. They are saying that not only is LIDAR necessary, but that super-LIDAR is the right and fastest path to a car that is safe enough. In addition, this moves them from a pure software company to a mixed software and hardware company, which will alter the nature of the deals they do with automakers.
ADVERTISEMENT
Aurora’s position is to be envied. While Waymo is (correctly) widely considered to be the leader in the robocar space, because Alphabet is the world’s #1 or #2 company, they have an ego and balance sheet to match. This allows Waymo to be picky about who it deals with and demanding as a partner. Faced with being unable to get the deal they want from Waymo, automakers see attraction in doing a deal with the man who led Waymo’s tech team for almost all of its existence, along with leaders from other major players. This gets them over the hump of working with a brand-new company, something they usually are afraid to do.
Three other players in Doppler LIDAR remain. Aeva was founded by two former Apple employees. SiLC is producing a single photonics chip. Insight LIDAR claims to have an FMCW device steered with no moving parts but a range of only 200m in their parts. Both Aeva and SiLC claim they have single-chip solutions that will be smaller, cheaper and lower power once in production. No doubt other players are sitting in stealth right now. There are also many companies making 1500nm LIDARs without Doppler. On top of all this, there are companies hoping to make high resolution radar, which would be superior to LIDAR if it got such fine resolution, since it sees through fog. Waymo builds their own LIDAR of this long-range class, and has been making it for several years and gotten the price down to reasonable levels.
You can leave a comment on this site.
Disclaimer: The author holds shares in GOOG, TSLA, and Quanergy (a LIDAR company) among others.