I am sure they have some sort of honk detection, but I think the key issue here is that they don't recognize other Waymo cars, and honk at each other any time they get into a standoff, which is common in their parking lots.
Pretty sure the honks are meant at human drivers. I guess what must be happening is that when navigating a parking lot and seeing another car that it thinks may be in its way or not aware of it, it honks to make its presence known.
This is generally a good idea when interacting with human drivers and in the outside world where the probability of encountering another self-driving car in such a situation is extremely low. It only breaks in this specific edge-case where there's seemingly nothing but self-driving cars in the entire lot.
It's actually quite a difficult problem to solve. The honk reflex is there for safety - if something that might be a car (based on limited/occluded camera & sensor data) you want to honk just in case, as early as possible. Waiting for the presumed car to get into better view so you can identify it is dangerous as it will delay the honk reaction to a time where it might be too late.
The only way I see to solve that is that all the cars need to communicate and be aware of each other in 3D space, and then be able to predict their upcoming position in the visual/radar field very precisely, so that as soon as it comes into view you need to be able to confidently say that this is indeed another Waymo car and thus inhibit the (now unnecessary) honk reaction. This needs to be very precise as you wouldn't want to accidentally inhibit it in case a human driver (or just cyclist/pedestrian) also happens to be there and broadly matches the location/angle you expect your other Waymo car to be at.
>The only way I see to solve that is that all the cars need to communicate and be aware of each other in 3D space, and then be able to predict their upcoming position in the visual/radar field very precisely, so that as soon as it comes into view you need to be able to confidently say that this is indeed another Waymo car and thus inhibit the (now unnecessary) honk reaction.
I honestly can't believe that's not already part of the feature set.
Easier solution: don't drive in a way that would make you honk if you were the other car. Yes, that might make them overly cautious... let's start there.
42 the number of honks per minute amongst self driving cars.
The AGI trained the self driving AI to recognize all sorts of objects and obstacles but not themselves. Perhaps to not be defeated by their reflection in the mirror. That old parable learnt from the tale oft repeated about a dog and the bone. Best to honk, defense in depth. geese honk in flight. Should work for cars too. Or perhaps the AGI generalized too far from the nursery rhyme: Hear a honk, there a honk, everywhere a honk honk… after all these AIs are multi modal.
It's funny to think about how they've kind of figured out self-driving (something many said would be impossible for decades) but they haven't figured out parking in their own lots without the self-driving cars honking at each other.
As a dev, I experience this sort of thing daily with bugs popping up. But I work on websites. It's just another ball game when we're talking about cars keeping people awake.
I mean forget self driving, I most want self parking. Basically my own personal valet at every shopping center, theme park, etc. I can drive the enjoyable bits on roads, just low speed scoot around and find a spot!
Is that a harder problem than driving? Or just as complicated so won’t arrive before FSD?
Oooo, that'd be great. Drives like a regular car, but lets me out in front of the mall and just goes off to find a parking spot somewhere, or even goes off to find a spot in a big public parking area somewhere else in the city. I want that even more than I want a car that can drive itself to a destination. There are so many lovely places downtown that would be so much more pleasant to drive to and from if I didn't need to park the car.
But I guess they'd rather own the cars, pick me up, and drop me off at the mall, and honestly that's probably more efficient for society in a city; no sense leaving the cars unused most of the time.
Hypothetically, if a situation like this were happening where self driving cars are causing a disruption in a particular area, could someone place traffic cones around the cars to prevent them from moving (and in the case in the article they'd presumably stop honking)?
This is a nice low-stakes example of how we can't expect automated driving to just be "like people but with better attention spans and reaction times and no risky emotions."
To twist the idiom: We aren't making an improved apple, we're replacing apples with oranges.
There will be things that are kinda-alien to human expectations and conventional wisdom, novel problems, and differently shaped risk profiles that we haven't had to deal with before.
(Comment partially recycled from older submission)
Still, I’d rather see empty Waymos bumbling around a parking lot than seeing empty Waymos clogging up our SF streets, and (observed twice recently) blocking pedestrian crosswalks.
Still, I wish Waymo had at fielded a vehicle smaller than an enormous, 5,000 lb, 400 hp Jaguar i-Pace SUV. That Jag is designed for 80-90 mph freeway cruising, not 15-30 mph urban driving.
> "We are aware that in some scenarios our vehicles may briefly honk while navigating our parking lots. We have identified the cause and are in the process of implementing a fix."
Making people wait for your bugfix/code change/deploy cycle when your little bug is unilaterally intruding on their lives and disrupting their sleep? This is just callous and irresponsible.
They really should have a procedure for immediately rerouting robo-traffic when problems like this occur, or just blocking off an area immediately when something like this is reported. I'm surprised something like that isn't part of the legal hurdles for getting a robo-taxi service approved.
Even in the current legal framework they have to be violating some sort of noise ordinance. (Speaking of which, can police not just stop and tow robo-taxis that are causing a disturbance? Shouldn't there be a mechanism for that, too?)
I wonder how many issues like this will need to happen before there's any legislation to make companies more responsible for the autonomous robots they deploy (what a sci-fi sentence!)
1) I'm interested to know what the patch look liked to fix this bug. The intended behavior is good (honking to warn that a car is about to back into you), but this edge case hadn't been considered.
2) If only it were this easy to patch bugs in human drivers.
Some lawyer could sign up several hundred people as plaintiffs in a class action lawsuit.
The other problem is that noise control bylaw officers seem to be on 9-5 schedules (don't ask how I know) rendering noise control bylaws effectively unenforceable at 4am.
I'm surprised the high rise neighbors haven't taken to slingshotting rocks off their balconies.
31 comments
[ 4.6 ms ] story [ 73.8 ms ] threadI was just thinking that the US now has a futuristic dystopian city, not just Japan in anime.
It's a bit silly though, "how did we end up here?" in an Idiocracy way.
This is generally a good idea when interacting with human drivers and in the outside world where the probability of encountering another self-driving car in such a situation is extremely low. It only breaks in this specific edge-case where there's seemingly nothing but self-driving cars in the entire lot.
It's actually quite a difficult problem to solve. The honk reflex is there for safety - if something that might be a car (based on limited/occluded camera & sensor data) you want to honk just in case, as early as possible. Waiting for the presumed car to get into better view so you can identify it is dangerous as it will delay the honk reaction to a time where it might be too late.
The only way I see to solve that is that all the cars need to communicate and be aware of each other in 3D space, and then be able to predict their upcoming position in the visual/radar field very precisely, so that as soon as it comes into view you need to be able to confidently say that this is indeed another Waymo car and thus inhibit the (now unnecessary) honk reaction. This needs to be very precise as you wouldn't want to accidentally inhibit it in case a human driver (or just cyclist/pedestrian) also happens to be there and broadly matches the location/angle you expect your other Waymo car to be at.
I honestly can't believe that's not already part of the feature set.
If all cars could detect these kinds of beacons, i’d want a transmitter on my bicycle. Or, carry one (my iPhone?) as a pedestrian.
DSRC and cellular-V2V i think solve similar problems, but those two require the more inaccurate GPS, i think,.
A Cacophonic symphony:
42 the number of honks per minute amongst self driving cars.
The AGI trained the self driving AI to recognize all sorts of objects and obstacles but not themselves. Perhaps to not be defeated by their reflection in the mirror. That old parable learnt from the tale oft repeated about a dog and the bone. Best to honk, defense in depth. geese honk in flight. Should work for cars too. Or perhaps the AGI generalized too far from the nursery rhyme: Hear a honk, there a honk, everywhere a honk honk… after all these AIs are multi modal.
(EDIT: They're the ducks from Rick & Morty [1].)
[1] https://www.youtube.com/watch?v=AZYyeNiGQJc
As a dev, I experience this sort of thing daily with bugs popping up. But I work on websites. It's just another ball game when we're talking about cars keeping people awake.
Is that a harder problem than driving? Or just as complicated so won’t arrive before FSD?
But I guess they'd rather own the cars, pick me up, and drop me off at the mall, and honestly that's probably more efficient for society in a city; no sense leaving the cars unused most of the time.
To twist the idiom: We aren't making an improved apple, we're replacing apples with oranges.
There will be things that are kinda-alien to human expectations and conventional wisdom, novel problems, and differently shaped risk profiles that we haven't had to deal with before.
(Comment partially recycled from older submission)
Still, I wish Waymo had at fielded a vehicle smaller than an enormous, 5,000 lb, 400 hp Jaguar i-Pace SUV. That Jag is designed for 80-90 mph freeway cruising, not 15-30 mph urban driving.
Making people wait for your bugfix/code change/deploy cycle when your little bug is unilaterally intruding on their lives and disrupting their sleep? This is just callous and irresponsible.
They really should have a procedure for immediately rerouting robo-traffic when problems like this occur, or just blocking off an area immediately when something like this is reported. I'm surprised something like that isn't part of the legal hurdles for getting a robo-taxi service approved.
Even in the current legal framework they have to be violating some sort of noise ordinance. (Speaking of which, can police not just stop and tow robo-taxis that are causing a disturbance? Shouldn't there be a mechanism for that, too?)
I wonder how many issues like this will need to happen before there's any legislation to make companies more responsible for the autonomous robots they deploy (what a sci-fi sentence!)
2) If only it were this easy to patch bugs in human drivers.
The other problem is that noise control bylaw officers seem to be on 9-5 schedules (don't ask how I know) rendering noise control bylaws effectively unenforceable at 4am.
I'm surprised the high rise neighbors haven't taken to slingshotting rocks off their balconies.