3 ms·
From my DARPA Grand Challenge days, a few comments: You want a high-resolution LIDAR. The trouble is, they cost too much. That's due to the low volume, not the
by Animats 3y ago
From my DARPA Grand Challenge days, a few comments:
You want a high-resolution LIDAR. The trouble is, they cost too much. That's due to the low volume, not the technology. A few years back, there were several LIDAR startups using various approaches. There were the rotating machinery guys, the flash LIDAR guys, the MEMS mirror guys, and some exotics. I was expecting the flash LIDAR guys to win out, because that approach has no moving parts. Continental, the big auto parts company, acquired Advanced Scientific Concepts, a LIDAR startup, and demoed a nice little unit. But nobody wanted enough of them to justify setting up full scale manufacturing. ASC sells some space-qualified units, and the Dragon spacecraft uses them for docking. Works fine, costs too much.
Another of the startups, Luminar, made its founder a billionaire without the company having shipped much product. "Luminar has not generated positive cash flows from operating activities and has an accumulated deficit of $1.3 billion as of December 31, 2022." - Wikipedia. They've announced many partnerships and have been at this since 2012, but you still can't order a LIDAR from their web site.
Velodyne makes those spinning things Waymo uses. They had the first automotive LIDAR, a big car-top spinning wheel, at the DARPA Grand Challenge. It fell off the vehicle. The technology has improved since, but it's still expensive. Velodyne just merged with Ouster, which sold similar spinners.
Ouster's web site has a "Order Now" button, but it leads to an email onboarding form, not a price list.
Many of these things need custom indium gallium arsenide detector chips, which would be far cheaper if the market was 100,000 a month instead of 100 a month. This technology hasn't scaled up in volume yet. That's one thing holding this back.
If you want coverage over a full circle, you either need one of those silly looking domes on top of the car roof, or multiple units, each with a narrower field of vision. Waymo has both. Works fine, costs too much.
On the radar side, there's been improvement in both processing and resolution. Not as much as expected over 20 years. Automotive radar has moved up from 24 GhZ to 77 GhZ, so resolution can now potentially be sub-centimeter. True millimeter radar (300GhZ and up) has been demoed [1] but not deployed beyond lab systems. Once you get up there, it's almost as good as LIDAR. Humans show up in detail. That's going to be a useful technology once it gets into commercial products.
So the sensor situation is pretty good, but expensive because it can't mooch off some other high volume technology such as cell phones.
[1] https://www.fhr.fraunhofer.de/en/the-institute/core-competencies/High-frequency-systems/300-GHz-SAR-measurements-of-streets-and-facades.html https://www.fhr.fraunhofer.de/en/the-institute/core-competen...