Hacker News Reader

Comment by guywithahat | Hacker News Reader

The more I've looked into the topic the less I think the removal of lidar was a cost issue. I think there are a lot of benefits to simplifying your sensor tech stack, and while I won't pretend to know the best solution removing things like lidar and ultrasonic sensors seem to have been a decision about improving performance. By doubling down on cameras your technical team can remain focused on certain sensor technology, and you don't have to deal with data priority and trust in the same way you do when you have a variety of sensors.

The only real test will be who creates the best product, and while waymo seems to have the lead it's arguably too soon to tell.

View on HN

Having multiple sources of data is a benefit, not a problem. Entire signal processing and engineering domains exist to take advantage of this. Even the humble Kalyan filter lets you combine multiple noisy sources to get a more accurate result than would be possible if using any one source.

What I've heard out of Elon and engineers on the team is that some of these variations of sensors create ambiguity, especially around faults. So if you have a camera and a radar sensor and they're providing conflicting information, it's much harder to tell which is correct compared to just having redundant camera sensors.

I will also add in my personal experience, while some filters work best together (like imu/gnss), we usually either used lidar or camera, not both. Part of the reason was combining them started requiring a lot more overhead and cross-sensor experts, and it took away from the actual problems we were trying to solve. While I suppose one could argue this is a cost issue (just hire more engineers!) I do think there's value in simplifying your tech stack whenever possible. The fewer independent parts you have the faster you can move and the more people can become an expert on one thing

Again Waymo's lead suggests this logic might be wrong but I think there is a solid engineering defense for moving towards just computer vision. Cameras are by far the best sensor, and there are tangible benefits other than just cost.

Counterpoint: Waymo

> solid engineering defense for moving towards just computer vision

COUNTERPOINT: WAYMO

We don't know enough about the internals of either one of them to make a judgement. The only one that is, is Comma.ai.

Waymo just reached 20 million public unsupervised rides. When will it be validated enough for Tesla fanboys? (Answer: never)

Fanboy or not, we don't know how much Waymo's model relies on an army of contractors labeling every stop light, sign, and trash can that, sure, they're using LIDAR to detect them and not cameras. We also don't know much about Tesla's Robotaxi initiative and how much human help they're relying on either.

first of all, their approach does not rely on map level labeling for runtime. They do that for training, but so does every other player. High precision maps are used as a form of GPS in their solution. They also use it to determine delta if road conditions change and alert ops.

Second of all, they’re using cameras to detect things like signs and trash cans! I don’t know where this misconception came from that Waymo only uses lidar, but it demonstrates a lack of familiarity with the problem space and the solutions

i cant edit this but it's bugging me "does not rely on map level labeling for runtime" isn't quite true. they do use tagging in their centralized mapping that the waymo driver uses as one source of truth. but ultimately, the onboard sensors and the onboard classifier and pathing engine are used.

this allows them to tag things that their cars are not recognizing, but then are fed back into the training set and validated. it allows them to "live patch" troublesome sections of road but it's not intended for long term use.

they also use this to close sections of road, such as for car accidents, flooding, etc.

From my previous comment, in case you didn't see it

> Again Waymo's lead suggests this logic might be wrong but I think there is a solid engineering defense for moving towards just computer vision. Cameras are by far the best sensor, and there are tangible benefits other than just cost.

Waymo could be piloting cars remotely with an operator per-vehicle for all we know.

Your logic is correct, however these challenges can be solved and then you get synergy effects from using different sensors.

I don’t understand how running into difficulties when trying to solve a problem can be interpreted as “[taking] away from the actual problem”.

In our case if we're spending a lot of time on something that doesn't improve the product, it just takes away from the product. Like if we put 800 engineering hours into sensor fusion and lidar when the end product doesn't become materially better, we could have placed those 800 hours towards something else which makes the end product better.

It's not that we ran into problems, it's that the tech didn't deliver what we hoped when we could have used the time to build something better.

Kalman filters and more advanced aggregators add non-trivial latency. So even if one does not care about cost, there can be a drawback from having an extra sensor.

Yes, there are tradeoffs to be made, but having to reconcile multiple sensors is not intrinsically a negative.

But also, if you didn’t get the right result, I don’t care how quickly you didn’t get it.

The latency from video capture and recognition is going to be so significant that it makes all other latency sources not even worth mentioning.

Cars and roads are built for human reaction times. That's why you have low speed limits in urban areas. You can have a pile of latencies attributable to processing a scene and still have superhuman reaction time that contributes to outperforming humans.

It's analogous to communications latency. High latencies are very annoying to humans, but below a threshold they stop mattering.

To tell what? Waymo is easily 5 years ahead of the tech alone, let alone the roll out of autonomous service. They may eventually catch up but they are definitely behind.

This is a solved problem. Many people I know including myself use Waymo’s on a weekly basis. They are rock solid. Waymo has pretty unequivocally solved the problem. There is no wait and see.

Nevermind the Waymos rolling by stopped school busses.

https://www.npr.org/2025/12/06/nx-s1-5635614/waymo-school-bu...

This seems solvable, no? Not saying it isn’t really damn important, but those have stop signs and flashing lights. It seems like they can fix that.

Solvable yes, but it's a perfect example of it not being solved yet despite this person's anecdotes.

Engineering problems aren't limited to a single solution anyhow. Anyone ruling out a camera based solution has very little imagination.

Unless it’s a mathematical proof, solved usually means works in the overwhelming majority of cases. It’s solved for all intents and purposes.

Waymo has barely seen the US. They just got on the highway and don't operate anywhere with snow.

If that's your definition of solved, be my guest, but it's a pretty silly one.

Highway? They’re in San Francisco my guy.

Detroit, Denver, Philadelphia and DC this year.

Yeah, it's been a month.

Indeed. The problem was already fixed.

You can cavil about this but it’s weak.

I mean if waymo had unequivocally solved the problem the country would be covered in them, and the only limit to their expansion would be how many cars they can produce. Currently they're limited by how quickly they can train on new areas, which is likely slowed by the fact they're using over 20 sensors across four different types. On the other hand, Tesla could spread across the country tomorrow if they were reliable enough. I would think solving autonomous driving would imply you could go nation wide with your product

Tesla literally has a guy sitting in there.

Right, nobody has solved it. If either company had solved self driving, it would be in basically every US city. While it is my opinion that Waymo is further ahead, no company has solved the problem yet and because of that it's still not clear what the best solution will be.

Sure, but Tesla is already losing the race. They were ahead a few years ago, but not anymore. They bet in getting autonomous driving done with cameras only that are cheap and have a simple and will understood tech stack and ecosystem.

It didn't work out though and now multi sensor systems are eating their lunch.

Honestly I think it's more that he was backed into a corner. The Teslas from ~9 years ago when they first started selling "full self driving" as an option, had some OK cameras and, by modern standards, a very crappy radar.

The radar they had really couldn't detect stationary objects. It relied on the doppler effect to look for moving objects. That would work most of the time, but sometimes there would be a stationary object in the road, and then the computer vision system would have to make a decision, and unfortunately in unusual situations like a firetruck parked at an angle to block off a crash site, the Tesla would plow into the firetruck.

Given that the radar couldn't really ever be reliable enough to create a self driving vehicle, after he hired Karpathy, Elon became convinced that the only way to meet the promise was to just ignore the radar and get the computer vision up to enough reliability to do FSD. By Tesla's own admission now, the hardware on those 2016+ vehicles is not adequate to do the job.

All of that is to say that IMO Elon's primary reason for his opinions about Lidar are simply because those older cars didn't have one, and he had promised to deliver FSD on that hardware, and therefore it couldn't be necessary, or he'd go broke paying out lawsuits. We will see what happens with the lawsuits.

> he was backed into a corner

He "painted himself into a corner" is the more accurate expression when one is the cause of their own problem

Usually you would go in with the max amount of sensors and data, make it work and then see what can be left out. It seems dumb to limit yourself from beginning if you don’t know yet what really works. But then I am not a multi billionaire so what do i know?

Well we know that vision works based on human experience. So few years ago it was a reasonable bet that cameras alone could solve this. The problem with Tesla is that they still continue to insist on that after it became apparent that vision alone with the current tech and machine learning does not work. They even do not want to use a radar again even if the radar does not cost much and is very beneficial for safety.

Human performance won't be sufficient. Self-driving vehicles have to be noticably (order of magnitude) better than humans to be accepted.

> Well we know that vision works based on human experience.

Actually, we know that vision alone doesn't work.

Sun glare. Fog. Whiteouts. Intense downpours. All of them cause humans to get into accidents, and electronic cameras aren't even as good as human eyes due to dynamic range limitations.

Dead reckoning with GPS and maps are a huge advantage that autonomous cars have over humans. No matter what the conditions are, autonomous cars know where the car is and where the road is. No sliding off the road because you missed a turn.

Being able to control and sense the electric motors at each wheel is a big advantage over "driving feel" from the steering wheel and your inbuilt sense of acceleration.

Radar/lidar is just all upside above and beyond what humans can do.

Human vision is terrible in conditions like fog, rain, snow, darkness and many others. Other sensor types would do much better there. They should have known that a long time ago.

To be fair, lidar is arguably worse in rain and snow. I don't think we know of a sensor yet which works well in these conditions

>The only real test will be who creates the best product, and while waymo seems to have the lead it's arguably too soon to tell.

Price is a factor. I’ve been using the free self driving promo month on my model Y (hardware version 4), and it’s pretty nice 99% of the time.

I wouldn’t pay for it, but I can see a person with more limited faculties, perhaps due to age, finding it worthwhile. And it is available now in a $40k vehicle.

It’s not full self driving, and Waymo is obviously technologically better, but I don’t think anyone is beating Tesla’s utility to price ratio right now.

Seems to me rather that Teslas are self driving cars with a handicap; they are missing some easily obtainable data because they lack the sensors. Because their CEO is so hard headed.

Simplifying things doesn't always make things easier.

Reply on news.ycombinator.com