That's a tough problem - distinguishing wet pavement from deep water.
Humans make that mistake frequently.
Autonomous vehicles should probably be equipped with a water sensor. (We did that in our DARPA Grand Challenge vehicle back in 2005). Then they can enter water very cautiously and see if it's too deep. This may make them too cautious about shallow puddles on roads, though.
It’s a particularly hard problem in Texas. We get torrential rains and the landscape is relatively flat. Couple that with shallow soil over lots of limestone and it means flooding is really common. We also have roads that have a “low water crossing,” where a road crosses a creekbed that is normally dry but which will flood. There are often water depth signs there (basically a vertical ruler with feet marks so you can see where the water is up to). We lose people to this scenario (driving into flood waters) every year. It’s particularly problematic when it’s dark and you miss a warning sign. Before you know it, you’re in deep water and the flow can sweep the whole car downstream until it gets pinned against a tree, possibly with water forcing its way into the car.
Yeah, people are bad at guessing and the usual "Plan continuation bias" kicks in.
I was travelling in a group to eat lunch with friends once, after heavy rains. We reached a site where the road needs to fit under a bridge and is known to flood, there's standing water, and the driver figured it's probably not too high, he drives in and nope, water over the air intake, bye bye engine and we walked the rest of the way to lunch
I absolutely should have said "No, don't" but the plan says we have to drive under that bridge, there is no plan B. Of course plan A being "Wreck car" is a stupid plan, but the bias meant I didn't say "No" and I should have.
You wouldn't die there, just trash the car, the flooding is localised - but there are definitely other sites around here where in flood conditions you could die if you drove into water that's deeper than you realised.
This case makes me think of my brother's place in rural Tennessee. To get to his house, you drive through a small creek, year round. For a hundred years in their community, they've managed without a bridge. I'm not sure driverless cars are ready for edge cases like this. Also, no one tell Enterprise I drove their rental through a creek.
Heck, that was the way I took into the city for work for a few years, shaved a good 30 mins off the commute.
You'd have to hold off for a few weeks every season change while the ice hardened up/melted or get stuck in it (thankfully I tended to get there after someone else found out).
I've seen several places in England (and at least one in the western United States) where they have fords.
For those not familiar, water runs over the road full-time, and people are expected to just drive through it like it's no big deal. Except for right after a storm, when it is a big deal. It's essentially the intersection of a road and a stream where a bridge should be, but nobody ever built one.
Ha, the stream is not that bad in 2021 when the Google car drove it in your link. But if you go forward two steps, the date changes to 2024, and if you pan back to the river there is a much stronger/higher flow. Maybe they drove through it in 2021 but said no in 2024.
Also, there's a quizzical cow up the road a bit and now I want to live in this place. Thanks.
Make a mental note of the level of the air intake for the engine in your car, if the water doesn't make it that high you should be fine as long as you don't get stuck (no, i don't where it is on my car).
The engine gets damaged when water gets into the piston chamber i think. Water compresses less than air so important things bend or crack if the engine is running too long with water in it.
I fear that advice might make some people overconfident when the water isn't stationary. Flooding doesn't have to be that intake-high to sweep a car sideways off the road.
Such detailed database of fine grained road geometry gets stale very quickly, due to road maintenance and road construction. In US highway lanes are shifted sideways frequently.
So the first waymo to get to this less used road to john’s will not have the data rather than every waymo that travels down a new highway, that then becomes a problem if it rains.
One car with an issue of first coincides with rain on a less used road?
Well, it's closer to: any car with stale data and sufficient water depth is a financial and PR disaster. These cars are not cheap, and a tiktok of someone being driven into the water is even more expensive!
I’m still amazed at the people who claim that Apple Maps is as good as Google Maps nowadays. If you live anywhere where there’s lots of development, it’s definitely not. It’s also terrible when businesses or places of interest move. My wife’s business moved a half mile down the road and a single message to google maps got it moved in a couple days. Apple Maps took about a year with multiple requests and even multiple messages to their special “escalation” email address.
I don't know how they don't notice thousands of users driving through these "cornfields" at 60mph every day, though. You'd have thought that'd raise some alert?
You underestimate how frequently details like this change in the real world and how difficult it is to reliably integrate them into the mapping models with very low error rates.
Aggregating this data in something close to real-time, verifying and corroborating that the change to the road model is real and correct, and then pushing those model updates to every vehicle that may need it almost immediately is not really a solved problem.
Waymo does have a database of every today they drive, but for this they don't need one.
If the car comes to a road covered with water, and that road is in the database, and the water level appears low compared to the historical level of the road in the DB, then the car could cross. if the road is not in the DB, then a different decision might be made.
This is similar to humans: you might make different decisions depending on whether you know the road well or not.
Isn’t that the Waymo data model, though? They extensively pre-drive every new market, building dense volumetric maps of the entire service area before they begin service, so they essentially do have that database of every road (that they drive on).
Granted, I am not sure exactly how Waymo operates, but I thought that the extensive testing was mostly for legal reasons+just handling edge cases.
I am saying this, because I noticed that they typically start with a low-tier restrictive permit to operate (with a rather small number of cars in the fleet). Then they run it for a year or two, iron out edge cases particular to a given city (e.g., climate particularities, crazy spaghetti junctions in ATL, etc.), and log a lot of data. Then they take that data, go to the city/state, say "we have all this data that demonstrates we were very above the board while running the test pilot program, we are safe, and now we want to expand out of a very limited test pilot program."
And then it either goes well (Bay Area, LA, etc.) or goes off the rails for other reasons (often failing earlier for entirely unexpected reasons, like the pushback against it from taxi driver unions in NYC).
My point being, I could be entirely wrong, but I don't think that they literally map every single inch of the road before being allowed to operate. I just don't see it as being possible in any large populated city, given how often things change there. Just in 3 years living at one apartment in Seattle, I had a road directly adjacent to me changed from 2-way to 1-way, and then had 3-4 lanes that were basically highway entrances/exits (a block away from me) created and the whole area being rerouted entirely.
Tesla is less "HD", they have standard maps like we all think of, and a lane level "see-ahead" system where they basically just grab a satellite image tile, and align it with what the car sees for "FSD".
They actually do significant mapping. Where it operates currently it is not unusual to see this. It will be a waymo with a human driver operating someplace not currently in the waymo zone and clearly not en route to any maintenance facility either. Stuff like windy canyon roads with no thru access anywhere that are currently gated away, you might see a waymo with a human today.
Waymo is not the only company making lidar maps right now either. I've seen UPS deliver trucks with retrofitted lidar scanners on the roof now. I've even seen this on a police car already, looked like a black rooftop industrial ventilator on a 2ft mast installed directly on the crown victoria roof.
This is also why they recommend not to use polarized while cycling, it can obscure slicks or water in certain sections. I still use mine but I know it's not as ideal as photo chromatic lenses.
It would have to know the height of that car above ground, how high the air intake is, etc. A lifter offroader could make it through a much deeper body of water than eg. a prius.
It doesn't take much of a rainstorm to see localized flooding. Some debris over the storm drain is enough to flood a street. Hard to anticipate that happening.
If they've mapped the surface of the water relative to themselves... couldn't they slowly wade in and just calculate the depth based on that 3d model without extra sensors.
Assumes there's no abrupt cliff to fall off... but short of the ability to make a 3d map underwater that seems inevitable.
By a water sensor do you mean a sensor to detect the water level relative to the chassis? It seems like a very inexpensive downward-facing ultrasound sensor could work.
Is ultrasound less expensive than a moisture sensor?
The problem with both is they effectively require the vehicle to be in the water already. They need something that can tell depth before the vehicle has to slow down.
They have been known to make that mistake. To use the word "frequently" demonstrates a misunderstanding between number of incidents and total miles driven. It also ignores that humans often drink and most of these types of accidents happen after 2am and most often in the state of Florida.
> equipped with a water sensor
Car washes will be fun.
> DARPA Grand Challenge
The problems the grand challenge ignores are more important than the ones it solved.
I've never made that mistake; I'm not aware of anyone I know doing it. I very rarely see it myself, except on news footage. Of course it happens some time somewhere but that says nothing about frequency.
> That's a tough problem
Not really. Don't drive where you don't know it's safe. Definitely don't drive into moving water - puddles only, and only if not too deep: I can usually figure it out based on the rest of the road - unless it's a sinkhole, the geometry is somewhat consistent - and especially by looking at objects in the water such as other cars driving through it. Sorry your friend isn't competent to figure it out.
People here are always quick to defend the autonomous cars, like a close friend. How often will we fall in love with a technology or company? It always distorts the truth.
It’s definitely a thing humans do a lot in certain places. Perhaps where you live, it isn’t as much of an issue, so naturally you and nobody you know has encountered it.
I suppose we can redefine 'a lot' to mean many things, but 99.9..% don't do it.
It's the exhausted talking point of the autonomous vehicle industry that humans are awful drivers and we are better. What is sad is seeing HN users doggedly repeat it like PR reps - it was the first line of the original comment, even though irrelevant to it - as they've done with other passions like Uber, Musk's entities, etc.
I rode my motorcycle into a hole that almost swallowed the front tyre entirely in rural Australia. That hole had just been a slight depression that collected water the last time I rode through it, and there was no visual indication that it was now deeper.
This is the naive "if you can't stop you're following too close" circular definition based take. Makes for good rightthink points on reddit and communities of similar quality membership but you're not actually gonna build anything useful thinking like that.
In order to drive reasonably humans need to drive through water that is 6-12in deep on occasion. That's just how it is. Near me it's whenever the storm drain at the bottom of the hill clogs.
Any human can't necessarily tell the difference between an inch of water, which is perfectly safe to drive on (if slow enough that you don't hydroplane), and a flooded street. They can tell the difference between an inch of water and wet pavement, though.
Article's current (possibly original), less ambiguous title: "Waymo recalls 3,800 robotaxis after glitch allowed some vehicles to ‘drive into standing water’"
IOW 3,800 Waymo vehicles aren't currently sat spinning their wheels in water.
Every time an issue is found, no matter how minor, it's fixed and updated everywhere. From now on, every car of that model (and future models, and related models) will no longer have that problem. Several passes of that improvement cycle, and self-driving cars become safer (and more efficient/comfortable/etc) than human drivers. At least, that's how it's supposed to work.
If we're lucky, this'll help get us better roads. I'm never exactly happy slowing down to 10mph on the freeway because it isn't obvious whether the 100ft long puddle is an inch deep or 2ft, and I can sidestep some of that danger sometimes by proactively choosing safer lanes, but our roads are dangerous and don't handle water correctly.
It's an interesting case of whether it's possible to infer the condition of wading and avoid having to install a sensor specific to a one in a million trips circumstance.
The inference would come from standing water slowing down the vehicle and likely require steering correction, in combination with some machine vision for identifying standing water.
Then there's the advantage of being Google and having hundreds of thousands of people in the same area using Google maps and navigation. Accelerometers in phones can detect crashes pretty reliably. There's a good chance they can reliably detect deceleration from standing water and report the location of the hazard.
How is this only being solved now? Isn’t it a very common thing that happens on roads or what am I missing? At least few times a year I have to go around major puddles, Waymo just speed boats through them? I doubt to believe.
Around 40 years ago, I was cycling around Andover (Hants, UK). Me and a mate were whizzing around near a small artificial lake. I decided to run into what I thought was shallow water and it wasn't. With hindsight that was a really daft failure of perception but you live and learn.
Forty years later touch wood I have not yet broken myself or a car ...
This seems tautological, but in practice, you might expect to see different results.
Engineering hours are finite, so if they're spread across interpreting signals from two different sources, they might not go deep enough to make either one as good as it could be.
Having your engineering resources more focused on a particular approach might actually yield better results.
I say this as someone who's dealing with LiDAR + vision vs pure vision in a different domain, and at this point, I actually think our pure vision systems are better.
More often than not, constraints refine and focus a project, rather than restricting it. It’s best to start work with as few variables as possible, and only add new ones when absolutely necessary; You make a lot more progress that way.
For very complex things like AVs, it is critically important to keep the number of such variables down, since each acts on complexity & workload not as an addition but more like a quadratic, or worse—combinatorial explosion.
Kind of unrelated. That issue was due to a misguided effort to be cautious by having vehicles requesting human-review when they didn't really need it. Waymo fixed the issue by allowing the vehicles to operate in their normal, independent, mode.
part of the problem is that SFs traffic lights just turn off in a power outage, rather than flashing red battery power as I have seen in many other jurisdictions
This is one of the reasons why I'm suspicious of camera-only systems, here in Finland. Half the year there's a lot of snow and ice around. Which I imagine means most of the view is "white" and "shiny". Coupled with the dark winters it's gotta be a nightmare to deal with.
Could you use a different spectrum of EM radiation to detect water? There are parts of the microwave band that attenuate the signal by absorption and I wonder if you could use that. The only clue a human driver has in that situation is in the visible spectrum. The lines of the road disappear from view, which can be challenging to see at night.
Not necessarily. Depending on angle and water depth, multi-return LIDAR can give you returns from both water surface and the road surface beneath, in the same way multi-return LIDAR can produce returns from vegetation and the ground beneath.
They suspended service areas they deem high risk until the software update can be applied. So while, yes, it's just a software update, it's a recall in the sense that they've temporarily pulled all the cars off the road in certain areas
"Recall" is a technical term meaning: "public dangerous defect notice".
A "recall" is stating that the defective version of the product in the field must be "removed/recalled" and replaced/updated with a non-defective version at the manufacturer's expense. It just so happens that the removal and replacement of defective software from the field can occur remotely.
The important part is that the manufacturer delivered a defective product that risks your safety, that fixing that safety defect is the responsibility of the manufacturer, and the system is unsafe until that occurs.
Yes, this is a common terminology issue. "Recall" is legally defined in terms of the kind of problems that require one, not the solution to those problems, because the relevant regulations were written when there was no way to fix consumer products other than physically delivering them to the manufacturer or an authorized repair person.
How about a Mastodon, Lamb of God take with Floods of Triton:
Heap data upon this modern age
All human drivers now phased away
A lidar's glow, the soft wheel's echo
Autonomous force of code remains
We are last of the before rides
Now hear the robot cars rise
Hum into eternity
Remember this, all roadways lead to the fleet
That gives me a horrifying idea for a short story about a Waymo being hacked to carry out an assassination where it purposely drives off a bridge and into a lake.
Since recall on cars no longer means doing anything to the car's physical location I think the regulator NHSTA should update this term
It just creates alarmist headlines for what's really an over the air update, although "recall" is still currently a regulatory accurate term in the vehicle space
Cars, especially EVs, have many similarities to being gigantic phones. Imagine if a routine software update from Apple was called a "recall", that functionally describes what's happening here
NHTSA should at least distinguish between "omg we have to get these cars off the road and bring them to the shop immediately!" versus "over the air software update"
Exactly what I was thinking, the CNBC article feels very clickbity because they don't say that in the opening lead that it's just a software update. They make it sound like they need to be taken back to some factory somewhere and get their systems updated. Which is not true because they just get a software update.
They are rolling these out in New Orleans soon. Standing water is everywhere, and sometimes you have big hidden potholes. You just need to know the roads. Should be fun.
On a normal day it should suffice to train the model to use its judgment and maybe monitor how other cars are reacting to water covering the road, but when it starts flooding everywhere maybe they should pause the service until it dries out.
Maybe you drive into flood waters, but I don't. That's not a difficult skill to pull off.
We're still in the early days of self driving cars, and as much simulation and miles as they have, they're still constantly getting exposed to real world conditions that are new to them. The world is dynamic, so this will always remain true.
It remains to be seen where we'll converge on capability, incident rate, and acceptance.
> It remains to be seen where we'll converge on capability, incident rate, and acceptance.
I think we're already there with Waymo as the example. We may later choose to diverge from this now-accepted path, but for the moment we have a blueprint, and fixing edge cases with a software update is apparently acceptable, if you just look at all the Waymos operating legally right now.
Maybe you don't drive into flood waters, but your Uber driver might, and that's what Waymo is trying to replace, not your personal driving.
In that context I think comparing it to the average human driver makes a lot of sense, because even if you personally are an even better driver, or even if human drivers are better at some specific things, we have more than enough data to show that Waymo reduces accident rates overall in their current rollout.
The world is dynamic, so sure, it will always be true in some technical sense. But I am confident that eventually we’ll have trained them on enough scenarios that novelty will have a smaller and smaller effect on their ability to safely navigate through the world.
>A product recall is a request from a manufacturer to return a product after the discovery of safety issues...
I think using the term for a software update is abusing the language a bit. And may confuse people who have a real recall where the thing has to go to the dealer.
Waymos are fleet vehicles. Recalls go to the owners, just like with other fleet vehicles such as rental cars, taxis, limos, delivery services, utilities, and city/state/federal government. It doesn't really matter who is whose customer.
I really want car companies to just automate publishing “recalls” for every commit pushed to any car ever. Flood this broken term and force a distinction between “the airbags will literally explode and destroy your face” and “the radio volume is too quiet sometimes”
A "recall" is a specific regulated action. It is announced as a recall because that is what is legally required according to the NHTSA. There is no wiggle room here.
Yes, we need to change the rules to create a distinction. The meaning of “recall” in common understanding vs. industry has diverged, and it’s almost certainly causing car manufacturers to do suboptimal things to avoid having “recall” tied to their name in the press.
There is no issue in understanding unless you are talking about only reading the headline that a media outlet decides to use. How about we all just use our brains and understand that things can be fixed in different ways, but it is important that they get fixed.
Suboptimal behavior from companies is what leads to recalls. I cant even understand an example of what you are talking about there. And now you want to carry water for the industry by creating some diluted term. Does the car have a safety issue that is should not? Then its a recall. The manufacture can now decide how to resolve it. Sometimes that can be done via an OTA update.
I think its is in the interest of consumers to know ALL the ways these corps are putting your life at risk through their engineering efforts or lack there of. If your car manufacture is doing weekly OTA bug fixes on the vehicle that you drive you kids in everyday, you (the apparent beta tester) should sure as well know. Then you can make an informed decision.
We really need a better term for when an urgent software update for a vehicle is issued. The extreme majority of the population completely misunderstands it when a "recall" is done when it's actually just an OTA software update.
That you need world models to sensibly deploy "thinking" machines in the real world. Else they do stupid shit like drive straight into water. You can bruteforce some semblance of thinking by training on literally all knowledge that can be digitized but even that is proving to not be quite enough.
This seems like an odd take. Don't existing self-driving cars already have rather a lot of world-model? It's not like they're just hooking the driving apparatus to the output of an LLM or something.
(Of course there is also scope for debate about how much world model today's LLMs have; it seems like it's more than none even though it has to be built out of token-shuffling parts. But that's not relevant here.)
Just this morning I was almost killed twice on my bike ride to work by two separate drivers, one of whom looked to be 80 and could barely see over the dashboard, and one who was on their phone. I didn’t even bother trying to remember the plate numbers, knowing that the odds of any kind of consequences are absolute zero.
No, we can’t go back to driving our own vehicles. Waymo everywhere and human driving outlawed, ASAP.
Agree. Multiple people I know have bought Teslas because they don’t trust themselves or their spouses to drive safely, and want them to use FSD. There should be incentives to get people onto self driving.
"Failure to follow all warnings and instructions can result in property damage, serious injury or death."
"Driver intervention may be required in certain situations, such as on narrow roads with oncoming cars, in construction zones, or while going through complex intersections."
"Always remember that Full Self-Driving (Supervised) (also known as Autosteer on City Streets) does not make Model S autonomous and requires a fully attentive driver who is ready to take immediate action at all times."
Different states have different rules about what sort of things insurers are allowed to charge different rates for. In the states that allow it, Tesla does offer insurance discounts for FSD usage.[1] Lemonade also offers discounts for FSD usage.[2]
That's a tough problem - distinguishing wet pavement from deep water. Humans make that mistake frequently. Autonomous vehicles should probably be equipped with a water sensor. (We did that in our DARPA Grand Challenge vehicle back in 2005). Then they can enter water very cautiously and see if it's too deep. This may make them too cautious about shallow puddles on roads, though.
I thought the same thing. A very small float switch would work here. Somewhere between the radiator and the bumper.
It’s a particularly hard problem in Texas. We get torrential rains and the landscape is relatively flat. Couple that with shallow soil over lots of limestone and it means flooding is really common. We also have roads that have a “low water crossing,” where a road crosses a creekbed that is normally dry but which will flood. There are often water depth signs there (basically a vertical ruler with feet marks so you can see where the water is up to). We lose people to this scenario (driving into flood waters) every year. It’s particularly problematic when it’s dark and you miss a warning sign. Before you know it, you’re in deep water and the flow can sweep the whole car downstream until it gets pinned against a tree, possibly with water forcing its way into the car.
Yeah, people are bad at guessing and the usual "Plan continuation bias" kicks in.
I was travelling in a group to eat lunch with friends once, after heavy rains. We reached a site where the road needs to fit under a bridge and is known to flood, there's standing water, and the driver figured it's probably not too high, he drives in and nope, water over the air intake, bye bye engine and we walked the rest of the way to lunch
I absolutely should have said "No, don't" but the plan says we have to drive under that bridge, there is no plan B. Of course plan A being "Wreck car" is a stupid plan, but the bias meant I didn't say "No" and I should have.
You wouldn't die there, just trash the car, the flooding is localised - but there are definitely other sites around here where in flood conditions you could die if you drove into water that's deeper than you realised.
Uncle Rodney has left the crankcase.
This case makes me think of my brother's place in rural Tennessee. To get to his house, you drive through a small creek, year round. For a hundred years in their community, they've managed without a bridge. I'm not sure driverless cars are ready for edge cases like this. Also, no one tell Enterprise I drove their rental through a creek.
Heck, that was the way I took into the city for work for a few years, shaved a good 30 mins off the commute.
You'd have to hold off for a few weeks every season change while the ice hardened up/melted or get stuck in it (thankfully I tended to get there after someone else found out).
I've lived here over a decade. Lived through multiple floods. Never once have I driven into water without being unaware of it.
Texas has it easy.
I've seen several places in England (and at least one in the western United States) where they have fords.
For those not familiar, water runs over the road full-time, and people are expected to just drive through it like it's no big deal. Except for right after a storm, when it is a big deal. It's essentially the intersection of a road and a stream where a bridge should be, but nobody ever built one.
For example: https://maps.app.goo.gl/eqjHKqDxGxFNz47D7 complete with the google streetview car (presumably) driving through it
And a collection of videos https://www.youtube.com/@jawalton2001/videos - it goes without saying, these aren't major thoroughfares.
Ha, the stream is not that bad in 2021 when the Google car drove it in your link. But if you go forward two steps, the date changes to 2024, and if you pan back to the river there is a much stronger/higher flow. Maybe they drove through it in 2021 but said no in 2024.
Also, there's a quizzical cow up the road a bit and now I want to live in this place. Thanks.
It seems to me you could do something interesting with floating warning devices that only appear when the water's too deep.
Sounds like overkill.
Stick in the ground will do it.
https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2F...
As long as the ground doesn't dip halfway across...
My dad went through one of those in England where it was a bit deep. The car had to have a new engine.
Make a mental note of the level of the air intake for the engine in your car, if the water doesn't make it that high you should be fine as long as you don't get stuck (no, i don't where it is on my car).
The engine gets damaged when water gets into the piston chamber i think. Water compresses less than air so important things bend or crack if the engine is running too long with water in it.
I wonder how electric cars fare with deep water.
I fear that advice might make some people overconfident when the water isn't stationary. Flooding doesn't have to be that intake-high to sweep a car sideways off the road.
I wonder how electric cars fare with deep water.
The ground clearance for my EV is 7 inches. The manual specifies it can handle 18 inches of water.
I don't know if that's the point where water messes with the electronics, or a swift current would start to move it sideways.
If they have a laser measurement of the road from before, couldn't they see that the level of water vs the expected road surface?
Such detailed database of fine grained road geometry gets stale very quickly, due to road maintenance and road construction. In US highway lanes are shifted sideways frequently.
But are they not continuously updating the road database with their fleet?
For common routes, yes. For getting to John's house, where the path there sometimes floods, no.
So the first waymo to get to this less used road to john’s will not have the data rather than every waymo that travels down a new highway, that then becomes a problem if it rains.
One car with an issue of first coincides with rain on a less used road?
Well, it's closer to: any car with stale data and sufficient water depth is a financial and PR disaster. These cars are not cheap, and a tiktok of someone being driven into the water is even more expensive!
As soon as the car descends below what was mapped it should be able to know there is a discrepency.
Satellite monitoring is also available for detecting extensive road work which they could use to invalidate and send out something to remap.
Earthquake or sinkhole?
I traveled to Austin 3 weeks ago and there were entire highways not on Google Maps.
Apparently they were built in just a few months.
There's some places where Apple still thinks I'm driving through a cornfield even though the development is a few years old, now.
I suppose I could inform them somehow, but it's not worth the bother.
I’m still amazed at the people who claim that Apple Maps is as good as Google Maps nowadays. If you live anywhere where there’s lots of development, it’s definitely not. It’s also terrible when businesses or places of interest move. My wife’s business moved a half mile down the road and a single message to google maps got it moved in a couple days. Apple Maps took about a year with multiple requests and even multiple messages to their special “escalation” email address.
I don't know how they don't notice thousands of users driving through these "cornfields" at 60mph every day, though. You'd have thought that'd raise some alert?
I’ve literally watched my car (only car around) trigger the yellow heavy traffic warning (because I was driving slow to look for something).
Yet thousands of cars doing 60 mph through a cornfield and over a river doesn’t trigger a “maybe a freeway was built here?”
Pretty sure they already rely on such a database for positioning, so they already have that problem.
But yes, this wouldn't work for other self-driving systems that don't rely on HD maps.
That seems a very risky assumption for any car (self driving or human driver) during flash floods. "Turn around don't drown":
You think you know how deep it is under because you've taken that road many times before (or in your case you have historical laser measurement)
But you don't know:
- Maybe the road under fully collapsed
- Maybe the flow of water is extremely strong, so you need to accurately estimate that too.
Flow should be able to be done with vision, radar can as well: some bridges use surface flow monitoring radar.
You underestimate how frequently details like this change in the real world and how difficult it is to reliably integrate them into the mapping models with very low error rates.
Aggregating this data in something close to real-time, verifying and corroborating that the change to the road model is real and correct, and then pushing those model updates to every vehicle that may need it almost immediately is not really a solved problem.
That's so much extra complexity
If they have a pre-existing database of every road, sure. And if it's kept up-to-date at all times in all vehicles.
Waymo does have a database of every today they drive, but for this they don't need one.
If the car comes to a road covered with water, and that road is in the database, and the water level appears low compared to the historical level of the road in the DB, then the car could cross. if the road is not in the DB, then a different decision might be made.
This is similar to humans: you might make different decisions depending on whether you know the road well or not.
Isn’t that the Waymo data model, though? They extensively pre-drive every new market, building dense volumetric maps of the entire service area before they begin service, so they essentially do have that database of every road (that they drive on).
Granted, I am not sure exactly how Waymo operates, but I thought that the extensive testing was mostly for legal reasons+just handling edge cases.
I am saying this, because I noticed that they typically start with a low-tier restrictive permit to operate (with a rather small number of cars in the fleet). Then they run it for a year or two, iron out edge cases particular to a given city (e.g., climate particularities, crazy spaghetti junctions in ATL, etc.), and log a lot of data. Then they take that data, go to the city/state, say "we have all this data that demonstrates we were very above the board while running the test pilot program, we are safe, and now we want to expand out of a very limited test pilot program."
And then it either goes well (Bay Area, LA, etc.) or goes off the rails for other reasons (often failing earlier for entirely unexpected reasons, like the pushback against it from taxi driver unions in NYC).
My point being, I could be entirely wrong, but I don't think that they literally map every single inch of the road before being allowed to operate. I just don't see it as being possible in any large populated city, given how often things change there. Just in 3 years living at one apartment in Seattle, I had a road directly adjacent to me changed from 2-way to 1-way, and then had 3-4 lanes that were basically highway entrances/exits (a block away from me) created and the whole area being rerouted entirely.
Waymo explicitly lidar scans and "HD maps" the area:
https://waymo.com/blog/2020/09/the-waymo-driver-handbook-map...
Tesla is less "HD", they have standard maps like we all think of, and a lane level "see-ahead" system where they basically just grab a satellite image tile, and align it with what the car sees for "FSD".
Some of the Tesla robotaxi deployments have relied on HD maps too, or at least they were spotted extensively lidar mapping Austin.
They actually do significant mapping. Where it operates currently it is not unusual to see this. It will be a waymo with a human driver operating someplace not currently in the waymo zone and clearly not en route to any maintenance facility either. Stuff like windy canyon roads with no thru access anywhere that are currently gated away, you might see a waymo with a human today.
Waymo is not the only company making lidar maps right now either. I've seen UPS deliver trucks with retrofitted lidar scanners on the roof now. I've even seen this on a police car already, looked like a black rooftop industrial ventilator on a 2ft mast installed directly on the crown victoria roof.
I live minutes from Ford's new HQ in Dearborn and I see multiple lidar equipped vehicles daily. Or at least vehicles loaded up the wazoo.
This is also why they recommend not to use polarized while cycling, it can obscure slicks or water in certain sections. I still use mine but I know it's not as ideal as photo chromatic lenses.
This is how I got my shoes wet climbing around rock pools last weekend.
They should really just park themselves on the side of the road (or observe in real time), wait for another car to go first, then follow that path
It would have to know the height of that car above ground, how high the air intake is, etc. A lifter offroader could make it through a much deeper body of water than eg. a prius.
Pretty sure the right answer mainly involves the car knowing about the weather and other emergency events.
It doesn't take much of a rainstorm to see localized flooding. Some debris over the storm drain is enough to flood a street. Hard to anticipate that happening.
Dangerous localized flooding has also occurred for other reasons unrelated to weather, like broken dams or embankments.
It can also occur downstream many many miles from the rain - see flash floods in the desert from mountain rains far away - https://www.youtube.com/watch?v=ORJtxkuD62E
All that it can take is a broken fire hydrant in the wrong place.
Doesn't Land Rover historically have like a wading sensor?
If they've mapped the surface of the water relative to themselves... couldn't they slowly wade in and just calculate the depth based on that 3d model without extra sensors.
Assumes there's no abrupt cliff to fall off... but short of the ability to make a 3d map underwater that seems inevitable.
By a water sensor do you mean a sensor to detect the water level relative to the chassis? It seems like a very inexpensive downward-facing ultrasound sensor could work.
When you're going 35 mph and suddenly hit a 2 ft deep puddle (I've done this), that sensor isn't going to help at all.
Is ultrasound less expensive than a moisture sensor?
The problem with both is they effectively require the vehicle to be in the water already. They need something that can tell depth before the vehicle has to slow down.
What kind of moisture sensor are you thinking of?
Probably an exposed set of electrical contacts and a resistance meter.
Unfortunately, those get screwed up in climates where there is salt used on the roads.
I've used an ultrasonic sensor to detect the water level in a tank before, I don't think it would work as you describe.
Also, the sensor didn't work in that context either as condensation kept forming on it.
Human drivers look for Waymos up ahead, water up to their windows.
<jk>
> Humans make that mistake frequently.
They have been known to make that mistake. To use the word "frequently" demonstrates a misunderstanding between number of incidents and total miles driven. It also ignores that humans often drink and most of these types of accidents happen after 2am and most often in the state of Florida.
> equipped with a water sensor
Car washes will be fun.
> DARPA Grand Challenge
The problems the grand challenge ignores are more important than the ones it solved.
Do they make this mistake frequently? How frequently? I've seen people overestimate things, but I don't think this is as hard as one might think
My commaai can do this. I'm pretty sure tesla can as well
> frequently
I've never made that mistake; I'm not aware of anyone I know doing it. I very rarely see it myself, except on news footage. Of course it happens some time somewhere but that says nothing about frequency.
> That's a tough problem
Not really. Don't drive where you don't know it's safe. Definitely don't drive into moving water - puddles only, and only if not too deep: I can usually figure it out based on the rest of the road - unless it's a sinkhole, the geometry is somewhat consistent - and especially by looking at objects in the water such as other cars driving through it. Sorry your friend isn't competent to figure it out.
People here are always quick to defend the autonomous cars, like a close friend. How often will we fall in love with a technology or company? It always distorts the truth.
It’s definitely a thing humans do a lot in certain places. Perhaps where you live, it isn’t as much of an issue, so naturally you and nobody you know has encountered it.
> humans do a lot
I suppose we can redefine 'a lot' to mean many things, but 99.9..% don't do it.
It's the exhausted talking point of the autonomous vehicle industry that humans are awful drivers and we are better. What is sad is seeing HN users doggedly repeat it like PR reps - it was the first line of the original comment, even though irrelevant to it - as they've done with other passions like Uber, Musk's entities, etc.
Any human can distinguish wet pavement from a flooded street. Some voluntarily drive into the flooded street.
And that is the difference. In a Waymo you are a prisoner, in your own car you can turn around.
I rode my motorcycle into a hole that almost swallowed the front tyre entirely in rural Australia. That hole had just been a slight depression that collected water the last time I rode through it, and there was no visual indication that it was now deeper.
This is the naive "if you can't stop you're following too close" circular definition based take. Makes for good rightthink points on reddit and communities of similar quality membership but you're not actually gonna build anything useful thinking like that.
In order to drive reasonably humans need to drive through water that is 6-12in deep on occasion. That's just how it is. Near me it's whenever the storm drain at the bottom of the hill clogs.
Why do you ignore the context (humans cannot distinguish?). If what you say is true, Waymos are useless in your areas. Thank you for confirming this!
Humans can distinguish well enough to know that it's not a "normal puddle amount" of water on the road and slow down.
Any human can't necessarily tell the difference between an inch of water, which is perfectly safe to drive on (if slow enough that you don't hydroplane), and a flooded street. They can tell the difference between an inch of water and wet pavement, though.
Article's current (possibly original), less ambiguous title: "Waymo recalls 3,800 robotaxis after glitch allowed some vehicles to ‘drive into standing water’"
IOW 3,800 Waymo vehicles aren't currently sat spinning their wheels in water.
This is important as flooded vehicles are a common sight on the salvage-title market.
Though the idea of a single rider calling for a Waymo and slowly one-by-one 3,800 Waymos drove into a flood and were washed away ...
That's the promise of self-driving cars.
Every time an issue is found, no matter how minor, it's fixed and updated everywhere. From now on, every car of that model (and future models, and related models) will no longer have that problem. Several passes of that improvement cycle, and self-driving cars become safer (and more efficient/comfortable/etc) than human drivers. At least, that's how it's supposed to work.
If we're lucky, this'll help get us better roads. I'm never exactly happy slowing down to 10mph on the freeway because it isn't obvious whether the 100ft long puddle is an inch deep or 2ft, and I can sidestep some of that danger sometimes by proactively choosing safer lanes, but our roads are dangerous and don't handle water correctly.
It's an interesting case of whether it's possible to infer the condition of wading and avoid having to install a sensor specific to a one in a million trips circumstance.
The inference would come from standing water slowing down the vehicle and likely require steering correction, in combination with some machine vision for identifying standing water.
Then there's the advantage of being Google and having hundreds of thousands of people in the same area using Google maps and navigation. Accelerometers in phones can detect crashes pretty reliably. There's a good chance they can reliably detect deceleration from standing water and report the location of the hazard.
How is this only being solved now? Isn’t it a very common thing that happens on roads or what am I missing? At least few times a year I have to go around major puddles, Waymo just speed boats through them? I doubt to believe.
Around 40 years ago, I was cycling around Andover (Hants, UK). Me and a mate were whizzing around near a small artificial lake. I decided to run into what I thought was shallow water and it wasn't. With hindsight that was a really daft failure of perception but you live and learn.
Forty years later touch wood I have not yet broken myself or a car ...
Does anyone with a better understanding about LIDAR vs camera approach to autonomous drivng explain how would Tesla handle such situation ?
This is a HW4 Tesla on FSD 14.3.2 trying to drive into a lake five days ago (a la The Office): https://www.reddit.com/r/TeslaFSD/comments/1t9rl2u/fsd_tried..., so I would not say Tesla has solved standing water yet.
That said, FSD seems quite capable of routing around standing water in many cases (e.g. https://xcancel.com/planoken/status/2030754820462633031, https://www.reddit.com/r/TeslaFSD/comments/1pw9f2m/fsd_navig..., https://xcancel.com/BLKMDL3/status/1991862465328779317, https://xcancel.com/JVTacoma/status/2046313902749921638), so handling the remaining cases seems more like a model intelligence / data issue rather than a sensor limitation. Lidar beams generally bounce off mirrorlike surfaces without returning to the sensor, so I think all lidar would tell you about standing water is "there's something shiny/reflective within this region of the image", which you already know from cameras+headlights.
Waymo has LIDAR and cameras, so it is better equipped for every situation.
This seems tautological, but in practice, you might expect to see different results.
Engineering hours are finite, so if they're spread across interpreting signals from two different sources, they might not go deep enough to make either one as good as it could be.
Having your engineering resources more focused on a particular approach might actually yield better results.
I say this as someone who's dealing with LiDAR + vision vs pure vision in a different domain, and at this point, I actually think our pure vision systems are better.
More often than not, constraints refine and focus a project, rather than restricting it. It’s best to start work with as few variables as possible, and only add new ones when absolutely necessary; You make a lot more progress that way.
For very complex things like AVs, it is critically important to keep the number of such variables down, since each acts on complexity & workload not as an addition but more like a quadratic, or worse—combinatorial explosion.
What if they oppose each other? Which one do you trust?
Google ”Sensor fusion”.
Unless the power is out
https://abc7news.com/post/san-francisco-leaders-press-waymo-...
Kind of unrelated. That issue was due to a misguided effort to be cautious by having vehicles requesting human-review when they didn't really need it. Waymo fixed the issue by allowing the vehicles to operate in their normal, independent, mode.
part of the problem is that SFs traffic lights just turn off in a power outage, rather than flashing red battery power as I have seen in many other jurisdictions
LIDAR isn't helpful for water. Standing water behaves like a mirror on LIDAR.
This is one of the reasons why I'm suspicious of camera-only systems, here in Finland. Half the year there's a lot of snow and ice around. Which I imagine means most of the view is "white" and "shiny". Coupled with the dark winters it's gotta be a nightmare to deal with.
Plenty of snow and ice in parts of the US. Hopefully you'll get a chance to try the self-driving for yourself.
do humans drive in it?
Could you use a different spectrum of EM radiation to detect water? There are parts of the microwave band that attenuate the signal by absorption and I wonder if you could use that. The only clue a human driver has in that situation is in the visible spectrum. The lines of the road disappear from view, which can be challenging to see at night.
Not necessarily. Depending on angle and water depth, multi-return LIDAR can give you returns from both water surface and the road surface beneath, in the same way multi-return LIDAR can produce returns from vegetation and the ground beneath.
If the LIDAR can sense the road close enough to the front of the car, then it could estimate how far underwater the car is.
There are better ways to measure how far underwater the car is: https://www.landrover.com/ownership/off-road/wading-depths.h..., this assumes you driving over something more or less flat underwater and that car is built for this.
Here the goal is avoiding driving into the water in the first place.
Maybe they're secretly developing Waymo submarines..
What is a recall in this case? Is them getting a software update a recall now?
They suspended service areas they deem high risk until the software update can be applied. So while, yes, it's just a software update, it's a recall in the sense that they've temporarily pulled all the cars off the road in certain areas
"Recall" is a technical term meaning: "public dangerous defect notice".
A "recall" is stating that the defective version of the product in the field must be "removed/recalled" and replaced/updated with a non-defective version at the manufacturer's expense. It just so happens that the removal and replacement of defective software from the field can occur remotely.
The important part is that the manufacturer delivered a defective product that risks your safety, that fixing that safety defect is the responsibility of the manufacturer, and the system is unsafe until that occurs.
I think so. For some kind of legalese reasons that's generally what a Tesla "recall" amounts to these days.
Yes, this is a common terminology issue. "Recall" is legally defined in terms of the kind of problems that require one, not the solution to those problems, because the relevant regulations were written when there was no way to fix consumer products other than physically delivering them to the manufacturer or an authorized repair person.
Waymo: *locks doors, chorus to Floods by Pantera starts playing, guns it into the water*
“Wash away maaaaan, take him with the floooood”
How about a Mastodon, Lamb of God take with Floods of Triton:
That gives me a horrifying idea for a short story about a Waymo being hacked to carry out an assassination where it purposely drives off a bridge and into a lake.
Since recall on cars no longer means doing anything to the car's physical location I think the regulator NHSTA should update this term
It just creates alarmist headlines for what's really an over the air update, although "recall" is still currently a regulatory accurate term in the vehicle space
Cars, especially EVs, have many similarities to being gigantic phones. Imagine if a routine software update from Apple was called a "recall", that functionally describes what's happening here
NHTSA should at least distinguish between "omg we have to get these cars off the road and bring them to the shop immediately!" versus "over the air software update"
Exactly what I was thinking, the CNBC article feels very clickbity because they don't say that in the opening lead that it's just a software update. They make it sound like they need to be taken back to some factory somewhere and get their systems updated. Which is not true because they just get a software update.
They are rolling these out in New Orleans soon. Standing water is everywhere, and sometimes you have big hidden potholes. You just need to know the roads. Should be fun.
On a normal day it should suffice to train the model to use its judgment and maybe monitor how other cars are reacting to water covering the road, but when it starts flooding everywhere maybe they should pause the service until it dries out.
Go fish
This is ok though because humans drive into flood waters too.
Look, you can't make progress without getting your feet wet and then diving straight into the deep end.
Maybe you drive into flood waters, but I don't. That's not a difficult skill to pull off.
We're still in the early days of self driving cars, and as much simulation and miles as they have, they're still constantly getting exposed to real world conditions that are new to them. The world is dynamic, so this will always remain true.
It remains to be seen where we'll converge on capability, incident rate, and acceptance.
> It remains to be seen where we'll converge on capability, incident rate, and acceptance.
I think we're already there with Waymo as the example. We may later choose to diverge from this now-accepted path, but for the moment we have a blueprint, and fixing edge cases with a software update is apparently acceptable, if you just look at all the Waymos operating legally right now.
Maybe you don't drive into flood waters, but your Uber driver might, and that's what Waymo is trying to replace, not your personal driving.
In that context I think comparing it to the average human driver makes a lot of sense, because even if you personally are an even better driver, or even if human drivers are better at some specific things, we have more than enough data to show that Waymo reduces accident rates overall in their current rollout.
The world is dynamic, so sure, it will always be true in some technical sense. But I am confident that eventually we’ll have trained them on enough scenarios that novelty will have a smaller and smaller effect on their ability to safely navigate through the world.
"recall" = applies software update
Wikipedia has
>A product recall is a request from a manufacturer to return a product after the discovery of safety issues...
I think using the term for a software update is abusing the language a bit. And may confuse people who have a real recall where the thing has to go to the dealer.
Yes, "recall" brings to mind serious issues like the gas tank exploding on the Ford Pinto.
I had one for my car because the airbag inflators were starting to go off more like a hand grenade than as intended.
Also I think it's wrong to call something a recall if it's not owned by customers. Waymo is a service.
Waymos are fleet vehicles. Recalls go to the owners, just like with other fleet vehicles such as rental cars, taxis, limos, delivery services, utilities, and city/state/federal government. It doesn't really matter who is whose customer.
aw, I was having fun imagining 3,800 Johnny cabs just immediately changing route to go to headquarters.
The difference between that and usual software updates I'm guessing is the cars are pulled from service until the update takes place.
Recall makes for better headlines.
I really want car companies to just automate publishing “recalls” for every commit pushed to any car ever. Flood this broken term and force a distinction between “the airbags will literally explode and destroy your face” and “the radio volume is too quiet sometimes”
A "recall" is a specific regulated action. It is announced as a recall because that is what is legally required according to the NHTSA. There is no wiggle room here.
Yes, we need to change the rules to create a distinction. The meaning of “recall” in common understanding vs. industry has diverged, and it’s almost certainly causing car manufacturers to do suboptimal things to avoid having “recall” tied to their name in the press.
There is no issue in understanding unless you are talking about only reading the headline that a media outlet decides to use. How about we all just use our brains and understand that things can be fixed in different ways, but it is important that they get fixed.
Suboptimal behavior from companies is what leads to recalls. I cant even understand an example of what you are talking about there. And now you want to carry water for the industry by creating some diluted term. Does the car have a safety issue that is should not? Then its a recall. The manufacture can now decide how to resolve it. Sometimes that can be done via an OTA update.
I think its is in the interest of consumers to know ALL the ways these corps are putting your life at risk through their engineering efforts or lack there of. If your car manufacture is doing weekly OTA bug fixes on the vehicle that you drive you kids in everyday, you (the apparent beta tester) should sure as well know. Then you can make an informed decision.
Gah, thanks for this. Thought I was used to that slight-of-hand but this one got me
Legally and technically true, and I hate it.
We really need a better term for when an urgent software update for a vehicle is issued. The extreme majority of the population completely misunderstands it when a "recall" is done when it's actually just an OTA software update.
We've updated the title above. Thanks!
LeCun is right.
About what
That you need world models to sensibly deploy "thinking" machines in the real world. Else they do stupid shit like drive straight into water. You can bruteforce some semblance of thinking by training on literally all knowledge that can be digitized but even that is proving to not be quite enough.
Like this?
https://waymo.com/blog/2026/02/the-waymo-world-model-a-new-f...
Or did you mean strictly in operation?
This seems like an odd take. Don't existing self-driving cars already have rather a lot of world-model? It's not like they're just hooking the driving apparatus to the output of an LLM or something.
(Of course there is also scope for debate about how much world model today's LLMs have; it seems like it's more than none even though it has to be built out of token-shuffling parts. But that's not relevant here.)
FFS, can we just go back to talking to each other in person and driving our own vehicles? Where'd the 90s go?
If the car drives itself we will have more time to talk to each other in person.
Or invest in public transport instead
Just this morning I was almost killed twice on my bike ride to work by two separate drivers, one of whom looked to be 80 and could barely see over the dashboard, and one who was on their phone. I didn’t even bother trying to remember the plate numbers, knowing that the odds of any kind of consequences are absolute zero. No, we can’t go back to driving our own vehicles. Waymo everywhere and human driving outlawed, ASAP.
Agree. Multiple people I know have bought Teslas because they don’t trust themselves or their spouses to drive safely, and want them to use FSD. There should be incentives to get people onto self driving.
Tesla cars are not capable of driving autonomously according to the company and regulators.
My dad doesn't care what the regulatory definition is, he just presses "Start Self-Driving" and off he goes.
https://www.tesla.com/ownersmanual/models/en_us/GUID-59736DF...
"Failure to follow all warnings and instructions can result in property damage, serious injury or death."
"Driver intervention may be required in certain situations, such as on narrow roads with oncoming cars, in construction zones, or while going through complex intersections."
"Always remember that Full Self-Driving (Supervised) (also known as Autosteer on City Streets) does not make Model S autonomous and requires a fully attentive driver who is ready to take immediate action at all times."
If self-driving is better, then presumably cheaper insurance costs would be an incentive.
Different states have different rules about what sort of things insurers are allowed to charge different rates for. In the states that allow it, Tesla does offer insurance discounts for FSD usage.[1] Lemonade also offers discounts for FSD usage.[2]
1. https://www.tesla.com/support/insurance/fsd-discount
2. https://www.lemonade.com/fsd
California is one of those states that does not allow FSD usage to impact insurance rates
Actually, we’ve just returned to 2007.
https://youtu.be/DOW_kPzY_JY
> can we just go back to talking to each other in person
He posts on an internet message board
If they are recalled, do the cars drive themselves back to the factory?