Hacker News Reader

Hacker News

Comment by fngjdflmdflg | Hacker News Reader

Full thread

fngjdflmdflg

This is important if nothing else because Miami sees much more rain than SF and Phoenix:

Miami: 57 in. (https://www.ncei.noaa.gov/access/us-climate-normals/#dataset...

SF: 25 in. (https://www.ncei.noaa.gov/access/us-climate-normals/#dataset...)

Phoenix: 7 in. (https://www.ncei.noaa.gov/access/us-climate-normals/#dataset...)

View on HN

Also relevant, when it rains in Miami during the summer, it pours.

As in, zero visibility for 15-30 minutes, then it's past.

So if it can handle Miami tropical rain, it should be okay with all sorts of normal rain.

Out of curiosity, what's Waymo's current production sensor suite mix? I'd assume lidar and radar would also be very unhappy with the surrounding space suddenly being ~10%(?) liquid water droplets.

I would expect service to be canceled while it is pouring down. Do we have reasons to believe that they have the ability to ride safely during heavy rain? I haven't been keeping up.

If it's too dangerous for humans to drive (regardless of whether they do anyway, humans do all manner of things which are unacceptably dangerous) then I don't expect Waymo to offer service even if they believe they technically could have the Waymo driver [their software] continue to deliver service.

The weather becomes too dangerous to drive because of mechanical (lost of adherence) problems way sooner than humans have sensorial problems anyway.

So if it's too dangerous for people, it's also too dangerous for computers.

You moved the goalpost by introducing the "too dangerous for humans to arrive" qualifier. The person you're replying to never said that. They asked if it would refuse to drive in pouring rain. They never asked if it would refuse to drive in scenarios where it's too dangerous for humans to drive.

The problem is that "pouring rain" is vague. So that's why I drew a more specific line.

They can handle heavy rain, see time 6:00: https://youtube.com/watch?v=Bm1A3aaQnh0

From their August 2023 blog post:

> During this past winter season in California with its record rain, high winds, and thunderstorms, we were able to maintain 99.4% fleet uptime

Thank you, that was fascinating.

How does it make out lane markings? Or is it all just GPS-based?

Localization is primarily based on visual registration, i.e. matching the current surroundings to the closest data in its map. Lane markings are based on map data and what it's able to see in real-time.

Maybe similarly how humans do it.

Humans do it super badly in a heavy rain/snow though. We're basically blind and just toodle along following the guy in front and hoping for the best.

A machine driver should not accept these conditions.

On the contrary, sticking strictly to the lane markers when everyone else is blindly straddling lanes seems the worst of all worlds.

Right, but I’m arguing that the correct behaviour for everyone is to pull over and wait for conditions to improve — robots and humans alike.

Where I come from that might be called a "shower". Heavy rain here is like fog. It's so thick you can only see a few meters, and the windscreen wipers can only give you a brief glimpse of what's ahead.

It happens rarely. When it does, more cautious drivers give up and pull over, even if they are on the freeway. That makes travelling at high speed down on freeway at high speed in those conditions near suicidal.

It only lasts a few minutes. I expect Waymo would handle like any human. Stop, or just creep forward.

Example of moderate Miami tropical rain: https://m.youtube.com/shorts/HAvKGwtQ2Uo

Far less than 10%. During a heavy downpour, by volume, about one part in a million is liquid. In a cubic meter of heavily rain, there are only a few tens of raindrops.

Imagine if it were 10%, though. During the time it took for a droplet to fall 1 metre, you’d have 10cm of water on the floor.

I reckon it’d feel quite heavy.

Rain falls at 9m/s, so in a second you'd almost completely fill that space with 900 liters. For the immediate area around you, imagine an olympic pool of water falling roughly every second.

According to NASA, terminal velocity for the largest droplets is ~10 m/s.

Which seems oddly close (in magnitude) to Earth gravitational acceleration (9.8 m/s^2).

Weird!

Yep, but what matters for radar/lidar is the projection. I mean what percentage of 1 _square_ meter (not cubic meter) is occupied by droplet projections. Or, in radial terms, what percentage of "Solid angle" is occupied by rain droplets.

LiDAR sensors, like for example from SICK can have multiple 'layers' of sensors, which combined with various algorithms can handle rain pretty good.

> So if it can handle Miami tropical rain, it should be okay with all sorts of normal rain.

I feel like a lot of "How well does it handle rain?" comes down to how the roads are built and maintained (Huge puddles, proper drainage, etc) rather than about the car itself, as the car you could test by blasting it with water from different directions and amounts.

There’s also the question of how other drivers handle the rain. And I have to imagine it’s nontrivial to, on a test range alone, permute the full range of different surfaces’ handling characteristics under different precipitation conditions.

I wonder whether, like many human drivers, Waymos might be wont to pull over and wait out the short-but-extreme Miami squalls.

Hopefully it does, because the drivers that don't stop are just driving blind and won't see the waymo.

Not just visibility in the rain, but diminished or fully obstructed visibility due to ponding and full flooding. Then there are the physical navigational problems associated with that. They probably shouldn't be driving though a foot of sea water.

Lidars perform very well in the rain [1].

[1] https://ouster.com/insights/blog/lidar-vs-camera-comparison-...

Thanks! Super informative link, and what I was hoping to get.

Interestingly is there a potential moral issue in the making here? What happens if/when self driving dependency is so prevalent that the majority of inhabitants in a city don't know how to drive. In addition add the fact that self driving cars don't have a steering wheel so even people who know how to take over driving can't actually take over.

What happens if there's an event that requires a mass evacuation such as a Category 5 hurricane and the major self driving car companies deem it too risky to drive in the conditions that precede the storm?

Interesting random fact: when it rains, waymo turns on the windshield wipers

Is this relevant? Are any of its cameras/sensors behind the windshield? Or are there wipers directly on the external cameras?

I think it's just for passengers' experience

Probably also a legal requirement to run them during rain, even if it's not actually needed for the self-driving cars to work.

Plenty of south florida rain is not helped by windshield wipers. Anyway I wonder if waymo sensors actually have better visibility in such conditions than people do

I think the comment you are replying to is implying that it's a bit funny/weird that the wipers turn on, because there aren't any sensors that are looking out the window to see. (As others pointed out, it could just be default auto-wiper functionality, and of course passengers still like being able to see out the window, even if they aren't controlling the vehicle).

> Interesting random fact: when it rains, waymo turns on the windshield wipers

The jaguar i-pace does this independent of the waymo use case.

My 2011 Mazda also did this.

Cars dating back to 2008 if not earlier. It can also be annoying/doesn't work very efficiently

The interesting part here is... the Waymo has no reason to. There is no driver. All cameras and sensors are outside. It's just for not freaking out the passenger :)

> The interesting part here is... the Waymo has no reason to. There is no driver. All cameras and sensors are outside. It's just for not freaking out the passenger :)

We have very different thresholds for what's interesting.

The platform provides this feature out of the box, why would waymo go out of their way to disable it. Obviously potential occupants would appreciate seeing out the windshield if it's raining, why that is interesting escapes me.

Waymo has wipers on it's LIDAR dome

https://www.reddit.com/r/oddlysatisfying/comments/6biyr8/way...

some places have laws that require wipers to be on while it's raining. Seems like a smart thing to have until laws are updated just to prevent your cars from getting pulled over by police.

They would probably have to go out of their way to disable the auto-wipers, no?

They definitely go out of their way to make significant modifications to their vehicles.

Rain + lidar = challenges

I'm guessing you've never tried driving in a tropical rainstorm. It's as bad a driving in heavy fog. Sometimes you only really have visibility of a few feet.

The.... driverless car... doesn't look out the window....

Who said anything about windows? I would imagine LIDAR looses some accuracy when its refracted by raindrops.

When I took Waymo in Phoenix, I booked a ride from a suburban hotel to a restaurant in a strip mall. One of the things I noticed was that I was picked up far away from the entrance of the hotel (eg, not under the overhang that protects from sun and rain, where every Uber has picked me up and dropped me off). I recall thinking that it was good there was no weather in Phoenix b/c I had to walk far enough I'd have gotten soaked in a decent rainstorm.

Have they changed this?

One thing I've noticed about the SF deployment is that it's slowly gotten better at this. At first it was very cautious about where it would pick you up/drop you off, but now it offers much closer options (from a menu -- a bit like Uber at airports).

I suspect this might be something that is human-added from data collected in past trips.

Also Waymo will not pick you up on private streets. I live in a small community with a private street and I have to walk to the nearest public one (2 mins).

Waymo can handle heavy rain, see time 6:00 here: https://youtube.com/watch?v=Bm1A3aaQnh0

And yet I regularly get stuck behind a Waymo in SF when there's just a little bit of fog.

Good point! Though also worth mentioning that they're already in Atlanta, which gets ~50in (and 59in so far this year, despite the mind-bending "first October without rain in recorded history")

I think they've announced they're headed to Atlanta in early 2025. So they may be testing there, but I don't believe they are at GA in GA :)

They are frequently spotted but not yet available in ATL.

Good point, I guess I missed when that happened. Looking at some news sites and Waymo's blog, it seems that they are testing in Atlanta and will start accepting customers in 2025.[0]

>It currently operates fleets of driverless cars in San Francisco, Los Angeles, Austin, and Phoenix. It also plans to launch a robotaxi service in Atlanta in an exclusive partnership with Uber.[1]

[0] https://waymo.com/blog/2024/09/waymo-and-uber-expand-partner...

[1] https://www.theverge.com/2024/12/5/24313346/waymo-miami-robo...

The rainfall can pose serious visibility risks that will be as much of a challenge as picking up and dropping off customers on a rainy day. Extreme high tides do still flood some roads on Miami Beach with brackish water, which isn't something you want to drive through in an electric car.

On the less challenging side, the city has zero snow, no road ice to worry about.

Reply on news.ycombinator.com