Spotted on page mill road. Had MFG plates and some type of velodyne spinning camera on top
What is tesla testing?
What is tesla testing?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
larmor
Active Member
If you don't blink your memory will be erased and will have no recollection of the events...
Ulmo
Active Member
My wild guesses (and I'm probably way off):
Learning which camera positions give the best information.
Data collection for camera processing development. (Software development sample data.)
Camera testing.
Roadbed baseline data collection for fleet collection data quality and meaning analysis.
Proof of concept for new visual features using known good input, for later fitting into production sensors. (Visual features such as AP, suspension damping, whatever.)
Backup to in-development pieces, for analysis and automated quality checks.
Understanding the competition.
Learning which camera positions give the best information.
Data collection for camera processing development. (Software development sample data.)
Camera testing.
Roadbed baseline data collection for fleet collection data quality and meaning analysis.
Proof of concept for new visual features using known good input, for later fitting into production sensors. (Visual features such as AP, suspension damping, whatever.)
Backup to in-development pieces, for analysis and automated quality checks.
Understanding the competition.
stopcrazypp
Well-Known Member
From what I can tell it does look like the velodyne logo written in the middle of the unit, so I would assume it is a lidar unit.
Specifically, the HDL32E:
HDL-32E
Ford's self-driving Fusion fleet uses the same model:
Ford's self-driving Fusion Hybrid to hit the streets of California next year
Canuck
Well-Known Member
stopcrazypp
Well-Known Member
A lot of people seem to think Tesla is "against" LIDAR (the spinning kind specifically). Elon only commented that he felt it was unnecessary, but the gist I get is that if it can be made cheaply Tesla would happily use it. The biggest problem right now is it's very expensive right now, so it makes sense right now to search for other alternatives in the mean time.
Canuck
Well-Known Member
Yes, but to expand a bit... Elon Musk:
"I don’t think you need LIDAR. I think you can do this all with passive optical and then with maybe one forward RADAR… if you are driving fast into rain or snow or dust. I think that completely solves it without the use of LIDAR. I’m not a big fan of LIDAR, I don’t think it makes sense in this context."
I'm pretty deeply familiar with both technologies, and I also want to put on the record I'm not saying LIDAR is worse than RADAR full stop. I do agree with Elon on two points:
(1) the current 77GHz phased array patch antenna based radar system from Bosch seems like there's more room for advanced signal processing. And that seems like what they delivered.
(2) radar is just LIDAR at a different wavelength. There's no good reason why future directional radar can be made to be more similar to the beam forming based solid state LIDAR tech recently shown off, delivering both x-ray vision and also highly directional point clouds
Sir Guacamolaf
The good kind of fat
Tech gets cheaper with time. Today you can get LIDAR cameras from $8K to $15K (or higher).
They look like that contraption. In 2-5 years, they will be seamlessly integrated into the car.
So it's reasonable to assume that Tesla is developing software for it that takes advantage of this tech - before making it mass market.
Come to think of it, once Model 3 comes out, and if it has the same AP capabilities as S .. the S naturally has to jump up a notch to justify it's premium price.
They look like that contraption. In 2-5 years, they will be seamlessly integrated into the car.
So it's reasonable to assume that Tesla is developing software for it that takes advantage of this tech - before making it mass market.
Come to think of it, once Model 3 comes out, and if it has the same AP capabilities as S .. the S naturally has to jump up a notch to justify it's premium price.
thegruf
Active Member
Is lidar monochrome or full visible spectrum?
ultimately this is all to do with how "visible" objects are at different frequencies combined with the ability to see through clutter (rain/snow/dust) etc.
Monochrome, it emits one wavelength of light, typically 904nm as patterned pulses of light. It is almost strictly time of flight, while modern radar uses a combination of FM, pulsing, and Doppler shift. One radar return tells you distance, relative speed, direction, and size!
It's also interesting because radar correlates better with "don't hit that".
Giant sheet of white paper? Looks just like a wall or a building to LIDAR. Looks invisible to radar.
Darkly painted black truck? Very poor LIDAR reflectivity, but looks bright as day from a mile away to radar.
thegruf
Active Member
and, as a follow-up thought could a lidar be integrated in the center of the roof with light guides/fibre optics to "lenses" blended into the bodywork maybe at the top of the a, b and c pillars?
I wonder what kind of interference there will be once there are 100's of thousands or millions of these on the road?
Yes, there are actually a lot of sleek looking automotive LIDAR setups. Quite honestly, a lot of the examples are police speed-measuring LIDAR jammers (which is one of my hobbies), but Mazda has a LIDAR setup that goes where our AP MobilEye sensor goes and is about the same footprint. Others hide LIDAR behind the grill, or with a faux logo cutout, etc. There's newer solid-state LIDAR technology coming on the market that use phased optical arrays which are totally solid-state.
In general though, fiber optics make transmitters extremely easy to hide. However, receivers tend to take up more space because they need lenses and optical filters to pick up reflected pulses.
There will be an inherent limit to the density of these systems due to the fact that a car takes up space and a finite number of dorks buy Smart Cars (oops did I say that out loud?)
But at any rate, the pulses are patterned, both in terms of pulse width and the spacing between pulses, so that a car can statistically assert that "yes, I'm 99% sure this LIDAR return is for the pulses I sent out 200ns ago, and it means there's something 30 feet away".
However, that's not really the end of the story. Other sources like the sun and oncoming headlights can cause a lot of LIDAR glare and interference, much like shining a flashlight into a camera. And in conditions such as fog and rain, you really do have to accept marginal LIDAR returns to have even a shred of visibility, which increases the room for error. That's not even to point out that overhead road signs, sides of trucks, reflective chrome surfaces, guardrails, etc can reflect LIDAR and create ghosted readings and false obstacles being sensed. However, mathematically speaking, such pulses always come back at a longer distance. So your point cloud will be noisy, but it won't necessarily cause you to believe in a non-existent imminent obstacle. You just might think there's perpetually a giant wall dangled 500 feet in front of you moving at your speed.
All of this is not mentioning any intentional source of interference. LIDAR is relatively easy to intentionally interfere with, at least almost every form of LIDAR on the market. That includes police laser guns, which all advertise some level of "anti jam", and even $10,000+ automated traffic enforcement systems like the Poliscan that can ticket 8 lanes' worth of traffic at once. It's a cat and mouse game where the mice are, quite frankly, winning because the cats refuse to believe mice exist. And from a regulatory standpoint, in the USA there are tough laws (federal prison sentences and fines) for interfering with RF transmissions like radar, but basically nothing for interfering with LIDAR. LIDAR is just regulated by the FDA to ensure eye safety. Nothing says you can't transmit a specially encoded but weak sequence of pulsed lights that make a Google self-driving car think that there's a semi truck approaching it.
EDIT: To be clear, I'm not endorsing or condoning sabotaging autonomous vehicle LIDAR. That's insane. All I'm saying is that there are already regulatory and inherent security benefits to radar vs LIDAR. Radar interference is harder to do and much easier to detect for a variety of reasons.
Similar threads
