Forums

LIDAR is coming.....

LIDAR is coming.....

At CES this year there are many commercial manufacturers of LIDAR units that are very compact. They are for use in autonomous cars and other uses, but it seems that the LIDAR industry is predicted to be $1.6 billion market now that these LIDAR sensors are the size of a deck of playing cards. In the article stated that Velodyne's leader Gopalan says that Musk is out of date regarding LIDAR and is "out of touch" with his approach to self driving.

I have always wondered if we (Tesla car owners) have enough sensors and associated AI software to accomplish FSD as advertised by Tesla. It is worrisome that an large industry is springing up around LIDAR for self driving, and other uses. Will the Tesla software ever get smart enough to use existing sensors for FSD? What if it never happens with what he currently have but have been promised? I have bought into FSD and want it to work. But it seems this newly minted market that is huge is flying directly in the face of the Tesla approach.

What do all of you think?

coselectric | 07. Januar 2020

TSLA crosses $470, brace for scared shorts joining the message board...

TexasBob | 07. Januar 2020

Musk's response to Lidar question on Autonomy day: https://youtu.be/Ucp0TTmvqOE?t=9200

Joshan | 07. Januar 2020

Elon has been pretty clear that LIDAR will not be on Tesla's thus far. He called it a fools errand.

JAD | 07. Januar 2020

My basic understanding is the lidar works very well 95% of the time. The problem is the hard 5% of the time lidar doesn't work you need sensors that do. once you get those sensors working for the hard 5% of the time they have no trouble with the easy 95% of the time that the lidar works. So why bother with the lidar?

TabascoGuy | 07. Januar 2020

Gopalan is badmouthing Musk because, if Musk is right, that $1.6 billion market will shrink faster than Tesla's stock price is rising.

TexasBob | 07. Januar 2020

Trying again. That is not at all what he said. Musk's response to Lidar question on Autonomy day: https://youtu.be/Ucp0TTmvqOE?t=9200

Magic 8 Ball | 07. Januar 2020

Hillary is running again, you don't say...

WhiteWind | 07. Januar 2020

Poor Fish forgot that master plan is transition from ICE to EV. FSD is just sweet bonus for people who wants it.

TabascoGuy | 07. Januar 2020

@Fish "...it was extreme fog so Tesla's camera system is not as robust as Subaru's..."

If it's too foggy for the cameras to see, it's too foggy for you to see. Neither you nor your car should be driving in those conditions.

Keithdorschner | 07. Januar 2020

I don’t read FISHEV any more since he’s been proven dishonest and fraudulent. If I can’t trust someone why even bother with what they say? Impossible to separate the wheat from the chaff. So let me guess— FISH loves LIDAR, right?

Effopec | 07. Januar 2020

I think I recall other OEMs recently pulling back from Lidar, but I can't recall which ones. Nissan maybe??

My personal experience: My company makes components that can be used in Lidar systems. A couple of years ago we were getting a lot of interest and RFQs from all over the world. Now most projects have gone cold. Take from that what you wish.

TabascoGuy | 07. Januar 2020

That's not the point. If a human can't see under certain conditions, the car should not be on the road. Same thing for cameras.

There will not be "VFR" like rules for AP as there is for aircraft so cars don't need to see through bad conditions.

TabascoGuy | 07. Januar 2020

@Keithdorschner, Fish loves whatever Tesla has dismissed as inefficient or unnecessary.

jallred | 07. Januar 2020

Oh no. Here we go again.

The wavelength of the laser used in LIDAR is very close to the spectrum that we see. It suffers from fog, rain and snow for the exact same reasons that a camera does.

In can work better in the dark because it provides its own light. But it can be worse in fog, rain and snow because the light has to go through the fog, rain and snow twice. If the path from the laser or back to the sensor gets blocked then LIDAR doesn't work.

A lot of people are working on making LIDAR work better in fog, rain and snow, and clouds. But it is not magic pill for low visibility scenarios.

RedPillSucks | 07. Januar 2020

FYI, Tesla uses RADAR as well as cameras, so there IS an orthogonal sensor for at least forward looking data.
According to Musk it's not JUST that LIDAR was bulky and expensive, it's also that it requires the route to be PREMAPPED and GEOFENSED. All the current users of LIDAR have to premap the environment that the cars will drive on, while Tesla is working towards FSD that will work even on roads that's not been mapped. So, for instance, when a road gets reconstructed/redesigned so the route is different, FSD will still work whereas the LIDAR implementation will not unless the road gets remapped.

Tronguy | 07. Januar 2020

@Keithdorschner: Fish likes whatever causes the most turmoil and FUD in the forum. Hijacking threads, inserting false information, leading newbies down the garden path and making them disappointed enough so they don't reappear, making random visitors who might be potential buyers think that Teslas fail all the time, etc., etc., etc.
He,she,it cycles from sounding reasonable to stating factually incorrect stuff about the engineering of Teslas. At one point, it became clear that he she it's talking point involved clear errors in basic energy vs. power.
A long time ago in a galaxy far away, my older brother came in the house with a spoof record, the vinyl kind, where some pretty superb musicians played music, purposely, badly. Like being off by a fifth of an octave and playing a duet like that: It was like somebody skritching chalk on a blackboard. It wasn't just like an occasional bad note which creeps into any clarinetist's practice: They could play off key like this for minutes at a time.
Obviously, just to be able to do that, they had to _really_ know their stuff.
FISH shows definite signs of being in this class. He's so bad, and switches back and forth from being bad to good and bad again so often, that it looks like he's not just dumb, he's really doing it on purpose.
M8B calls him evil. I'll go with that. But I've gone a little farther. I personally suspect that FISH is run by a bunch of beltway bandits that are being paid to disrupt one of Tesla's strengths, Social Media.
There also appears to be a select group of sock puppets who often trail around with FISH. And, perhaps, some others who are attempting to emulate troll behavior.
As to who might be paying these beltway bandits, that's another question. But there are other people on the web, much better known than I am, who have repeatedly hypothesized that Big Oil is behind many of the shorts on Wall Street. Being able to deny Tesla access to the funds available from the stock market may be a cost effective move on Big Oil's part; every day they delay Tesla is another billion or so for them.
If that's true, then a paltry million or two to a beltway bandit to disrupt these forums is not even pocket change, it's pocket fluff.
Thing is, Bil Oil definitely has played FUD with respect to climate change for nigh onto 50 years now, funding so-called scientists to deny that climate change even exists. I've even read reports that, of all people, cigarette companies were getting hints directly from the Big Oil guys on how to go about the business of denying facts.
All of the above sounds precisely like a conspiracy theory. To which I say: It's not paranoia if the bad guys really are out to get you. Think Koch brothers. And the current US and Australian administrations.

jallred | 07. Januar 2020

It's quite a stretch to say that LIDAR and video are orthogonal. They both use light which puts them neatly in the same Venn Diagram circle. Or in the same Venn Diagram circle of "those that can be disrupted with a piece of tape over the sensor".

bjrosen | 07. Januar 2020

On a related topic, are the sensors in a Tesla sensitive to a wider spectrum than the sensors in a photography camera? For photography only the visible spectrum is useful but for self driving the infrared and possibly the UV portions of the spectrum can give additional information.

majassow | 07. Januar 2020

For those that think that LIDAR is required, here is a look at a video shared by Intel at CES 2020 which shows Mobileye navigating heavy city traffic for ~25 minutes using ONLY cameras.
https://youtu.be/hCWL0XF_f8Y
No radar. No LIDAR. No rear cross traffic alerts. OK, don't know about that last one, as they never reversed :-).

Yes, it is daylight, and it is good weather. But this video shows a heck of a lot of things that many claim won't be possible for many years. And as others have pointed out, LIDAR isn't meant to help in bad weather (it can be worse than cameras alone).

However, I wouldn't want to give up the forward radar: it saved my bacon at least once when my car braked prior to the truck in front of me nearly plowing into stopped traffic and swerving out of the way at the last second.

pnagy | 07. Januar 2020

I am an amateur astronomer and I take pictures of the heavenly sky with dedicated astro camera. Astro cameras use CCD or CMOS sensors. CMOS sensors are very sensitive that can capture both UV and IR (or near IF). So it's possible for Tesla cameras to capture UV and IR assuming it does not use filters to block UV and/or IR.

Peter

rdavis | 07. Januar 2020

Shorts must really be feeling it now... I mean, throw it all at the wall and see what sticks.

@Majassow... great link to the Mobile Eye demo. That was REALLY cool.
@TexasBob... great link to Elon talking about LiDAR. That whole video is one of my favorites they've done. It was truly eye opening and help me to understand just how Autonomy works.

Whdame | 07. Januar 2020

MobileEYE is using as (12) camera based system where as we only have (8) in our Model 3's. Wonder if it is still possible with (8) cameras and sensors compared to just (12) cameras and no sensors.

andy | 07. Januar 2020

Following this with interest. Have experienced the limitations of cameras (e.g. glare, weather and contrast), but the point about LIDAR using similar wavelength was new info.

Effopec | 07. Januar 2020

@Majassow - great video, shows that cameras can do the job, at least in this instance. They have something there that is generally not available around where I live - lane markings.

majassow | 07. Januar 2020

@Effopec: yes, good lane markings certainly help. But there were several times where the vehicle had no issues with ignoring those markings to navigate around blockages. Just like shown in the Tesla Autonomy day, it appears like the car is recognizing "navigable space", as well as employing some rules for yielding to vehicle conflicts within that space.

Regardless, neither LIDAR or radar could take advantage of lane markings...

howard | 07. Januar 2020

Level 5 will not be achieved with the current Tesla sensor suite. LIDAR may just be one more necessary component if Elon is to ever achieve his infamous Robo Taxi Fleet.

Joshan | 07. Januar 2020

howard has spoken ignore the scientists and engineers who have spent a decade on this, we have an internet expert!

WhiteWind | 07. Januar 2020

I was about to say. Who do I listen to? Elon or Howard hahah

Joshan | 07. Januar 2020

In September, Tesla CEO Elon Musk had to say about the use of LiDAR sensors:

We do not anticipate using lidar. Just to make it clear, lidar essentially is active photon generator in the visible spectrum – radar is active photon generation in essentially the radio spectrum. But lidar doesn’t penetrate intrusions so it does not penetrate rain, fog, dust and snow, whereas a radar does. Radar also bounces and lidar doesn’t bounce very well. You can’t do the “look in front of the car in front of you” thing. So I think the obvious thing is to use radar and not use lidar.

jimglas | 07. Januar 2020

Hide your heart girls

howard | 07. Januar 2020

Ah, the other experts have spoken. Guess time will tell. Who cares really. It is not worth the risk, at least until I am legally not able to drive then It won’t matter.

Joshan | 07. Januar 2020

Fish wants to buy a 200k car it seems. Every sensor, system and useless item in the world. The more the better, amirite?

andy.connor.e | 07. Januar 2020

what he really wants is the turbo

howard | 07. Januar 2020

Tesla states AP may not perform in heavy rain, snow, fog and bright driving conditions. What do the Tesla experts propose to over come the stated limitations to reach level 5 as promised?

howard | 07. Januar 2020

Clue software is not the answer.

majassow | 07. Januar 2020

Fish's Subaru has a cane for sensing potholes and curbs, and can read the Braille on the drive-thru ATMs.

Gee: Mobileye, Qualcomm, Tesla all developing LIDAR-less autonomous vehicles? Must not be anything there.

jallred | 07. Januar 2020

None claimed it was if you read the self driving technical experts comments -fish

I never said anyone claimed anything. Will you quit arguing for the sake of arguing?

Adding more sensors and sensors of different types helps. You can visit San Francisco area and see all kinds of test cars with stacks of sensors on the roof. Not practical for real life. It’s great that lidar sensors are getting more affordable. That helps maybe, but doesn’t change that lidar is a very similar sensor to a camera. And it is yet to be seen what the performance is of these new lidar units. Most lidar spins around, these are unidirectional.

jallred | 07. Januar 2020

a massive 20% accuracy increase, -fish

What do you mean by accuracy?

WhiteWind | 07. Januar 2020

"Taycan comes standard with wireless charging which is a $100 third party option for Tesla."
Did you just compared 250k car to Model 3 starting for 39 k?

WhiteWind | 07. Januar 2020

Compare*

TeslaTap.com | 07. Januar 2020

Ok, LIDAR is a great way to get to autonomous driving tech quicker without as much software development when cost, styling, and maintenance are unimportant and it only has to work in dry climates.

Of course LIDAR systems today along with the massive computing power to process all the data costs between $50K-100K per car. Then there is the complexity of keeping these things clean. The main solution is rotating the array, which seems very impractical on a production vehicle. Solid-state LIDAR, is sampling now, seems promising, but it requires some kind of wiper assembly. The best spots for LIDAR are not in the front windshield. Headlight assemblies with LIDAR and small washer/wipers are one solution proposed, but it is a bit ugly and non-aerodynamic.

We're years away from LIDAR being practical in any reasonably priced vehicle. My guess is vision/radar systems as Tesla uses today will be the only practical solution.

apodbdrs | 07. Januar 2020

I don't think Musk is speculating regarding LIDAR as he has direct experience in designing LIDAR sensors. Paste the link below to your browser and at 11:06 into the You tube video he discusses LIDAR.

https://www.youtube.com/watch?v=uj1YAnzUYCo

TexasBob | 07. Januar 2020

ANCAP (Australia's equivalent of IIHS) 2019 evaluation of driving assist technology:
Subaru: 77% out of 100.
Tesla Model 3 and X: Best of anything tested. 94% out of 100.
Real data from actual engineers at an independent testing agency. “Both the Model 3 and Model X achieved the highest Safety Assist scores recorded to date at 94% - well ahead of all others rated this year.”

https://www.ancap.com.au

TexasBob | 07. Januar 2020

^^^ Responding to fishy comments about cameras - not directly relevant to the Lidar OP.

apodbdrs | 07. Januar 2020

I don't think Musk is speculating regarding LIDAR as he has direct experience in designing LIDAR sensors. Paste the link below to your browser and at 11:06 into the You tube video he discusses LIDAR.

https://www.youtube.com/watch?v=uj1YAnzUYCo

howard | 07. Januar 2020

Waymo uses a suite of vision, radar and three different LIDAR systems. They are the only company in public trials of Level 5. They report the costs of their in-house self driving systems have come down 90% and they are now selling complete self driving packages to others expecting costs to come down even more. The technology is rapidly expanding with expectant future costs to be substantially lower. The ergonomics of the sensor packages will also continue to improve until hardly noticeable. This is the exciting truth of it all.

You can postulate all you want but Tesla is way behind compared to others. Maybe they will get there relying on relatively inexpensive cameras and sensors but it is potentially a very long road while others are racing on ahead.

coselectric | 07. Januar 2020

Why y'all arguing with the Fish? Tesla stock just hit $470, he's obviously having a bad day so just let him be.

WhiteWind | 07. Januar 2020

Hmm who is correct. Guy that sends rocket to space and gets it back or Mr Fish and his sock firend H. and their great analytical skills .

WhiteWind | 07. Januar 2020

Friend*

jallred | 07. Januar 2020

Check out waymo.com to see what your car would look like.

Pages