Will Tesla ever do LIDAR?

[Ulmo]

The current thinking by Elon & co is no.

 

[diplomat33]

The current thinking by Elon & co is no.

I started this thread because I thought that even if LIDAR is not strictly needed, it might still be good to have for redundancy. But after watching the presentation, I think Elon and Karpathy made a compelling case for not using LIDAR.

 

[electronblue]

I started this thread because I thought that even if LIDAR is not strictly needed, it might still be good to have for redundancy. But after watching the presentation, I think Elon and Karpathy made a compelling case for not using LIDAR.

I thought that was the least compelling part after Elon’s jokes. Why even discuss Lidar at all? Just show your greatness. Dissing on Lidar didn’t go down well in the media either. It was not believable. They should have just focused on what makes their solution great instead.

 

[diplomat33]

I thought that was the least compelling part after Elon’s jokes. Why even discuss Lidar at all? Just show your greatness. Dissing on Lidar didn’t go down well in the media either. It was not believable. They should have just focused on what makes their solution great instead.

I kinda agree with you. And I think they did make a good case for why their solution is great. I particularly liked the video of how Tesla created a 3D reconstruction using just the cameras. But there are experts who believe that LIDAR is a non-negotiable requirement for FSD. If they believe that, then it won't matter how great your vision, they will still dismiss your approach as never being good enough for FSD. So you have to make the case against LIDAR not just for your system. Honestly, Elon is probably just tired of getting the LIDAR questions so he went off a bit.

 

[mhan00]

Sure Mobileye and Waymo are ahead in terms of tech but Tesla is catching up fast. Yesterday, Tesla proved that they have excellent hardware, very good camera vision, they don't need LIDAR, and they have the fleet learning. And Tesla can deploy their software to more cars faster.

By the way, your signature is super annoying and obnoxious.

Dunno if they “proved” they don’t need Lidar. Nothing is proved until they deliver FSD. But they certainly made a compelling argument for why it isn’t necessary and why it might even be a hindrance to achieving FSD by using Lidar as a crutch. I bought FSD, so hopefully i’ll Be able to evaluate for myself by the end of next year or so. I’m also probably in the minority where I’d be satisfied with merely EAP on city streets and highways requiring human supervision as NoAP works now by then, with a hopefully clear path using data gained from that to true level 4 and 5 self driving.

 

[iTech]

It troubles me that you feel you need to keep your location confidential.

It's troubling that you think location matters. Isn't Tesla trying to accomplish autonomy without regards to specific locations?

 

[StealthP3D]

I thought that was the least compelling part after Elon’s jokes. Why even discuss Lidar at all? Just show your greatness. Dissing on Lidar didn’t go down well in the media either. It was not believable. They should have just focused on what makes their solution great instead.

Wow! Did you even watch the presentation? The opinions presented on LIDAR were in direct response to analysts questions about LIDAR. They would have looked really foolish to say "We're not going to discuss LIDAR today."

 

[StealthP3D]

It's troubling that you think location matters. Isn't Tesla trying to accomplish autonomy without regards to specific locations?

Of course they would like full autonomy to be released continent-wide all at once but they have let it be known that, due to various local jurisdictions, it probably won't be approved continent-wide or nationwide all at once. It will get started in limited areas and spread out.

But I'm not asking for your birth certificate or your address, just the city (or county) you live in.

 

[malcolm]

By the way, your signature is super annoying and obnoxious.

I think that's his USP.

You can turn off signatures in the preferences anyway, but my take is that it's a bit like Hydrogen:-

When will we see Hydrogen fuel-cell cars go mainstream? In about ten years time - Hydrogen is (always) the future.

How many times does Blader have to be right about tesla? At least one more time

 

[johnm]

Level 5 (no steering wheel required) robotaxi.

Can you get to "no steering wheel required" without being level 5? For example with remote control ability to help the car get out of (hopefully quite rare) situations it can't handle?

To actually be level 5 does the car have to handle things like a police officer coming to the window and instructing the car to turn around and drive the wrong way on the highway because the road is blocked ahead? Does anyone think the car is likely to do this sort of thing in the near future? (I am also pretty convinced that LIDAR will not help with this scenario.)

 

[StealthP3D]

To actually be level 5 does the car have to handle things like a police officer coming to the window and instructing the car to turn around and drive the wrong way on the highway because the road is blocked ahead?

In my 40 years of being a motorist, I've never had to drive the wrong way on a freeway. But I've seen a couple of drunk humans do it accidentally. Pretty scary stuff.

 

[heltok]

Gonna cross post this here as I think it is relevant:

My perspective on this. When I did my master thesis we could choose between doing Lidar SLAM or Camera SLAM. We choose Lidar because we felt it suited us better compared to the other team who were more suited for the camera project.

The camera team had it somewhat easier because they could pretty much just download ORBSLAM and have a fancy demo running without too much work. We as the Lidar team had struggle with many of the steps ourselves, such as feature point extraction, which back then was far from trivial. But we had an easier time with particle filters etc for positioning.

It seems that we have two fields converging:
- Probabilistic Robotics, Sebastian Thrun et al. Particle filter, graphSLAM and classical hand made tools trying a little bit of Machine Learning
- Computer Vision, Andrej Karpathy et al. CNN and other computer science tools trying a little bit of Robotics

The probabilistic robotics guys love their Lidars, it works in the same bird’s eye framework as they see the world. The Computer Vision guys love their cameras, the input comes in a nice structured matrix, the same way as they see the world.

We are now seeing deep learning making great depth maps out of camera images and we are seeing classical point clouds from camera images making great object detections. The first runs great on GPUs/TPUs, the latter will complicate how to pipe the code a lot... But the main takeaway is that the two fields are starting to overlap. A very interesting fusion of domains that will confuse a lot of people in both domains.

We are at a time where we have a lot more computer scientists coding than we have roboticists coding, but we have more roboticists building vehicles than we have computer scientists building vehicles. Cameras are cheaper, they are passive sensors. Lidars are getting cheaper fast but there will likely always be a difference of some magnitudes. Lidars rely less on intelligence, if you don’t get a reading in front of you, you can be pretty certain that there is free space in front of you. But with some clever software and a gigantic amount of data the camera is catching up. Thus the price and power benefits starts to favor the camera.

Imo at this point, cameras are easier to work with but hard to do well. Lidars are hard to work with, but easier to do well. I think Teslas approach will turn out to be the right one and I am very impressed by Elon’s ability of coming to this conclusion much earlier than most other experts. I was wrong on this.

 

[electronblue]

Wow! Did you even watch the presentation? The opinions presented on LIDAR were in direct response to analysts questions about LIDAR. They would have looked really foolish to say "We're not going to discuss LIDAR today."

Yes, including the part (on which I commented on the Autonomy Investor Day thread) where Karpathy went off-character and added a clearly agreed-upon diss on Lidar to his otherwise stellar presentation. It was a strawman and not believable — and neither were Elon’s answers and negative comments on Lidar.

But we will have Level 5 no geofence feature complete systems for our cars at the end of 2019, Tesla should focus on delivering that. That in itself is all they need to do to convince the market, dissing on Lidar is completely unnecessary. Mind you, I’m not talking about showing their camera SLAM, that is of course showing your own features and perfectly nice.

It is like those traditional automaker ads going after Tesla. No, that is stupid. Deliver greatness, that is all that is needed. Attacking your opponent is a sign of desperation.

This my opinion, you have yours and I respect that.

 

[StealthP3D]

Yes, including the part (on which I commented on the Autonomy Investor Day thread) where Karpathy went off-character and added a clearly agreed-upon diss on Lidar to his otherwise stellar presentation. It was a strawman and not believable — and neither were Elon’s answers and negative comments on Lidar.

I don't know how Musk could answer the question about why they don't use LIDAR without explaining the (negative) reasons why they won't be using LIDAR!

There is nothing wrong with speaking negatively about the lack of abilities of a technology, it's not like LIDAR has feelings that Musk needs to protect! He was simply explaining why they don't use it and why they have no plans to use it in the future (at least not for autonomous driving).

This my opinion, you have yours and I respect that.

It's not clear to me why you refuse to disclose your location. What could you possibly have to hide?

 

[electronblue]

I don't know how Musk could answer the question about why they don't use LIDAR without explaining the (negative) reasons why they won't be using LIDAR!

In my opinion there would have been plenty of valid reasons to give for that without saying, multiple times, Lidars are unnecessary appendices and Lidar users are doomed. In my view they would have been more believable if they would have just focused on the benefits of their approach and left the choices of others without too much comment. Certainly stuff like ”doomed” and smirking was completely unnecessary. That was not an engineering answer either, that was just sheer arrogance — and it was not believable to me, and judging by the market reaction, not quite to the market either.

There is nothing wrong with speaking negatively about the lack of abilities of a technology, it's not like LIDAR has feelings that Musk needs to protect! He was simply explaining why they don't use it and why they have no plans to use it in the future (at least not for autonomous driving).

This is our disagreement. I think differently. I think they attacked technology choices of other companies unnecessarily and it diminished the believability of their argument. I respect your opposing opinion and I have mine.

It's not clear to me why you refuse to disclose your location. What could you possibly have to hide?

I think everyone chooses themselves what they disclose online and it seems normal netiquette to respect that. Secondly even when people disclose a location, there is no way for regular readers to verify that anyway. You could be in the UK for all I know driving a Mazda.

 

[StealthP3D]

I think everyone chooses themselves what they disclose online and it seems normal netiquette to respect that. Secondly even when people disclose a location, there is no way for regular readers to verify that anyway.

OK. Based on you using the most minuscule and unimportant things to attack Tesla and Musk at every possible opportunity, I'm going to assume you are either in Detroit or Manhatten.

 

[electronblue]

That’s the great thing about the written word, it can exist separate of its writer and its merits can be analyzed independently. Indeed often it benefits to do so because it is the great equalizer. Everyone’s word counts the same independent of who they are.

There’s something beautiful about that idea — at least as long as the dialogue remains civil. Everything has its flipside of course.

 

[MikeS1]

This ignorance is real for a lot of folks here.

You don't need lidar when you get almost the same results of a point cloud system using imaging.

1 You NEED imaging for autonomy
2. A point cloud system can be obtained from vision but then ask why do you need lidar...

 

[StealthP3D]

This ignorance is real for a lot of folks here.

You don't need lidar when you get almost the same results of a point cloud system using imaging.

1 You NEED imaging for autonomy
2. A point cloud system can be obtained from vision but then ask why do you need lidar...

Exactly! It's like Musk said, LIDAR is a development crutch that will ultimately slow down the successful development of full autonomy.

Some people see Musk's statements as an unfair "attack" on his competitors who are using LIDAR. I see his statements simply as the plain and accurate description of the subject by a person who actually understands the complexities of full autonomy and how to achieve it. Much like the Wright brothers understood the problem of heavier than air flight and used the pragmatic approach to develop it.

The argument that autonomy requires LIDAR (because all the other major FSD competitors use it) is about as strong a popular belief that was common before we had heavier than air flight:

"An aeroplane requires flapping wings because all creatures that fly use flapping wings."

And it was true that most people working to develop heavier than air flight wasted a lot of time on craft that had flapping wings. The ones who succeeded didn't fall for such unsupported beliefs.

 

[GM EV1]

We all know that Musk has been quite adamant that he is against LIDAR. And a few years back, I think it was somewhat understandable. Back then, LIDAR was clumsy and expensive. There was just no way that Tesla could afford to put LIDAR in every car they sell, not to mention the issue of ruining the aerodynamics and aesthetics of the cars with a LIDAR tower on the roof. So I think back then, it made sense for Tesla to go the camera vision only route. After all, if they could manage to achieve the same result with camera vision only for a fraction of the cost of LIDAR, why not?

But today, these problems with LIDAR are pretty much solved. LIDAR is cheaper and smaller. Tesla could integrate LIDAR in the car in a way that does not ruin the aerodynamics or the aesthetics of the car at a much more affordable cost. Plus, there is no question that even if Tesla does manage to achieve great things with camera vision alone, LIDAR would offer more redundancy and make true Full Self-Driving much more robust. In other words, even if camera vision works, why not have that extra redundancy of camera vision, radar and LIDAR to give the car an even fuller picture of the environment to make FSD even better? There is no downside. So I am thinking that Tesla will eventually cave in a few years and add LIDAR to the FSD hardware.

Thoughts?

Here is article in Cornell Chronicle also from last Monday where they are finding the benefits in camera vision vs LIDAR, too, and confirm the massive cost difference.
New way to ‘see’ objects accelerates future of self-driving cars | Cornell Chronicle
They will present their paper in Long Beach in June