Comment by Animats | Hacker News Reader

Comment by Animats | Hacker News Reader

Parent Full thread

I could see Page saying that. Urmson seems to have been the one who took the Waymo technology from "sort of works" to "works reliably". There's a long, slow grind phase of testing, logging, and dealing with more unusual situations before the technology is ready for deployment. Levandowski tends to skip that part, as with the Otto demo. One can see why he got along with Kalanick.

Levandowski comes out of this fine. Google can't sue him. Google has an employee contract which forces employee disputes into arbitration. That arbitration has already happened and can't be reopened. The settlement terms are not public.

The LIDAR design is a side issue. There will be lots of good, low-cost LIDAR units as soon as someone is able to order a few hundred thousand of them. Continental's flash LIDAR will probably be the first successful mass-market flash LIDAR. Quanergy made a lot of noise, but never shipped. Now they seem to be pivoting from automotive to border security. (They want to make Trump's border wall a virtual one.) Tetraview has funding but no products. Velodyne has been making flash LIDAR noises but hasn't demoed yet. Innoviz has demoed an experimental unit and claims to be selling it. Meanwhile, Velodyne is trying to get their rotating devices down to $500 in quantity.

Knowing the Tetravue founders personally and having seen the tech years before any the funding came in I can say that their product is at least as real as their competitors'.

Who?

Currently studying arbitration. Do you have a source on Levandowski’d arbitration already having been completed? Interested to know more there...

https://www.google.com/search?q=Levandowski+arbitration

Details: https://techcrunch.com/2017/03/29/waymo-pursued-arbitration-...

I'm curious what LIDAR solves for border security, that infrared cameras can't already do. I guess they could more easily identify slow moving objects and camoflage?

PMD/IFM has been trying to drive their costs down for a while as well.

That's a modulated carrier type system. Those aren't useful in sunlight, because they have to outshine ambient lighting continuously. Pulse-type devices only have to outshine the sun in a narrow frequency band for a nanosecond, which is quite possible.

That's the idea behind the modulation, though. It's relatively easy for the receiver to lock to the modulated AC frequency in the presence of sunlight ("DC" at baseband).

Same concept as TV remotes that frequently use a 36-38 kHz carrier. If they just gated the IR LED on and off at the baseband data rate, interference from room light would be a problem.

Those things are fun. The idea is that you modulate the outgoing beam with a RF carrier, and measure the phase shift of the incoming beam relative to the outgoing beam. Today you can do this directly, because electronics is fast enough to count in the gigahertz range. It used to have to be done by superheterodyning, like a radio receiver. You down-convert both the incoming and outgoing signals using a local oscillator, and get two new signals at a much lower frequency. This transformation has the useful property that the phase difference between the lower-frequency signals is the same as the phase difference at the higher frequency. But at the lower frequency, it's easier to measure the phase difference. Also, you get the selectivity of a superheterodyne receiver.

That's fine for a scanner, where you only need one copy of the electronics, but doing it for every pixel means something like a radio receiver behind each pixel. Still, you could probably do that now.

The SwissRanger [1], a short-range time of flight imaging camera, worked like that. It's been discontinued.

(I fooled around with this in the early 1990s, trying to build a laser scanner. Too many noise problems.)

[1] http://hptg.com/industrial/

Yeah, that's an intriguing piece of hardware. A camera that outputs a Z buffer at conventional digital-photo resolutions and data rates would have lots of uses, but it sounds like it fell short of that goal.

Luminar inked Toyota deal…

Reply on news.ycombinator.com