Hacker News Reader

Camera vs Lidar+Camera is not a narrow question. Cameras lack sufficient dynamic range to work in many situations, and therefore cannot alone be used for a real self-driving solution where the driver naps.

Your example of needing to read a stop sign isn't a great example. At least in North America, a hexagonal sign is always a stop sign. A better example of your point would be a speed limit sign.

Yeah, world of difference between having Lidar or not. Tesla demonstrably has issues detecting objects with their cameras.

>At least in North America, a hexagonal sign is always a stop sign.

the back of a stop sign is often a slow sign, or a do not enter sign. The same shape of sign, but different meaning from different directions. The slow variant is often held by construction workers, hence the GPs example.

Please re-read this part of my post:

> Say you're driving down the road and there's a bit of construction, there's a guy holding SLOW/STOP sign directing traffic.

Here's a picture of a guy holding a hexagonal SLOW sign. They are very common. https://nj1015.com/how-slow-should-you-go-through-constructi...

LIDAR cannot tell you what's on that sign. It cannot read any sign, nor any road markings, nor streetlights, blinkers, stop lights, etc. If you believe that cameras are not capable of being used as input due to dynamic range or any other reason, then that's fine, you just believe that self-driving is impossible (lidar or not). But to believe that one can safely drive using nothing but a completely blank and unlabeled 3D scene (ie, LIDAR-only)? That's pretty crazy.

> Say you're driving down the road and there's a bit of construction, there's a guy holding SLOW/STOP sign directing traffic. LIDAR will tell you it's a hexagonal sign, but it can't tell you what it says, you need a camera to read the sign and tell you what it says.

But Waymo never said you don't need cameras. Hell, they have 29 cameras in each vehicle compared to Tesla's 8.

Your point about their approaches being more alike than different is somewhat true, but you wrongly attribute the LiDAR vs camera debate to Waymo marketing. It's Elon and Tesla fans who started it and incessantly repeat it even to this day. Most rational folks say use whatever you can to get it working (which Waymo did) and optimize later.

Indeed, cameras are absolutely necessary for self-driving. They're the only ones that can read lane markings, signs, etc. LIDAR alone is not sufficient - you cannot navigate the roads using nothing but an unlabled 3D scene. It simply does not have the necessary information for you to drive ie, is that hexagonal sign a STOP or a GO, pretty important bit of info.

So the question is: is LIDAR also necessary or are cameras sufficient? IE, can cameras+motion give you an accurate-enough 3D scene the way LIDAR can. And that's a narrow technical question, and it isn't even the most question when you consider self-driving as a whole.

"LIDAR is necessary" is not exclusively a Waymo talking point - it is shared by all companies using LIDAR, suppliers of LIDAR etc. But it is just a talking point, there's no reason to think it's actually true.

The reasons to think it’s actually true is because:

1. Only companies who have LiDAR in their stack have shown fully autonomous driving i.e. driverless.

2. Better 3D scene construction, which LiDAR, unquestionably provides permeates as an advantage throughout the stack.

I do think “LiDAR or not” is a narrow argument. But the advantages are massive and undeniable, so it becomes necessary especially in light of rapidly falling costs.

It's not necessarily a case of "LIDAR is necessary": it's more a case of how much more difficult it is to extract the necessary information with the necessary reliability from cameras alone (one important aspect I think is that LIDAR + HDmaps is very reliable at telling you 'something is there', even if it's not able to distinguish details. From the point of view of the car operating safely, that's a big deal). As it stands Tesla is behind waymo and making things harder for themselves (fundamentally because they've been trying to make it a consumer product, which constrains costs much more than aiming for robotaxis). It might mean they eventually get a much more cost competitive product, but that won't matter if they're beaten to market by 5-10 years (at the moment you can argue tesla is ahead in the areas their driver assist can operate and number of vehicles, but I think it's going to be much easier for waymo to scale up and to different areas than it is for tesla to progress their tech).

The whole thing is largely probabilistic in many ways/parts, and it seems like more sensors, especially more sensors that operate in different modalities, is better, assuming your sensor fusion is working properly so that each additional sensor adds certainty to your predictions.

There are atmospheric conditions and obstructions that lidar can see through that cameras can't.

Cameras also seem prone to being blocked by a small splash of mud/dirt. Is anyone on this thread knowledgeable enough in the domain to know if that's an issue? I thought of it while moving my head sideways to see around a temporary sight obstruction on my windshield. Luckily the windshield is big, and I can move my head. Cameras are small. I guess you just put several so you have an effectively large camera array? It does mean more redundancy is necessary than I would have initially thought.

Camera obstructions are big problem, yes. A combination of hardware fixes (making it less likely for obstructions to happen in the first place), redundancy, and detection of the issue coupled with an escape plan, is the general approach to fixing it. (the escape plan is a big one: a lot of effort goes into making sure the car can stop/pull over safely if something goes wrong)

Thanks! That all makes perfect sense…

Tesla and Wayno are so unalike that Tesla could have to start from scratch if a detailed 3D map and the sensors to sense one's position in that map prove to be required. Tesla hasn't got, and can't retrofit that hardware. Tesla hasn't got the mapping data. Tesla hasn't got the many years of steady progress toward thousands of uneventful trips per day with no human touching the wheel.