Visually impaired Paralympian hit by self-driving bus | Hacker News Reader

Visually impaired Paralympian hit by self-driving bus

france24.com

32 pointsphreeza37 comments

I’m all for criticism of reckless self driving implementations but it seems this can’t be blamed on software:

“ Two bus operators -- on board to supervise the autonomous driving -- said they noticed the athlete, but thought he would stop walking as the bus was approaching, according to Japan's Asahi daily.

The Mainichi daily said the bus automatically stopped but the operators pressed the start button as they did not think he would walk out on to the road.”

>The Mainichi daily said the bus automatically stopped but the operators pressed the start button as they did not think he would walk out on to the road.”

Yeah, if that's the case then this story is literally the 100% exact opposite of what the headline implies: it was the human drivers who were the weak link and caused an accident just is normally the case. Self-driving systems can be patient and just wait when in doubt, rather then feel any emotional temptation to make assumptions and get a move on.

There was a thread on this earlier this (last?) week with a video of a Waymo car hesitating as a person walked near a cross walk.

I mean, yes, headline might be misleading but it's not "100% exact opposite of what the headline implies". The headline ("Visually impaired Paralympian hit by self-driving bus") implies that someone was hit by something, that something was a self-driving bus. All of that is true, and is what happened. What's left out though, is whoose fault it was, it's not part of the headline at all. One might think it was because of the self-driving system instead of a human, but it wasn't.

So yes, misleading/clickbait it was, but not "100% exact opposite".

It should just say "hit by bus" given the self-driving nature of the bus had nothing to do with the incident and in fact would have prevented it.

But that wouldn't be news. It matters because most people will just read the headline and go away thinking self-driving technology is dangerous.

Sort of disagree... It's a misleading title but the fact that it's self-driving is what makes this story interesting. Title should have added 'after human intervention' for more clarity rather than less.

Why is it that the human drivers are the weak link. While driving through the city there are numerous times where pedestrians are waiting near cars and if they suddenly acted irrationally and stepped in front of the vehicle there would be a collision. Humans have to navigate that, a self driving system that refuses to attempt to read the intentions of pedestrians would be stalled forever. Not to mention in this story it could be the pedestrians fault and not on the human drivers.

Who designed an autonomous bus with a “run over people anyway” button?

You mean the gas pedal? Feature of ever vehicle since it's invention?

The use of a gas pedal implies a human operator is in control. This seems different than a “just go” button on a bus that is otherwise autonomous.

I know that it is frowned on here at HN to ask 'did you read the article' however -

The Mainichi daily said the bus automatically stopped but the operators pressed the start button as they did not think he would walk out on to the road. (fourth paragraph from the end).

Perhaps you missed that part?

The bus should have stopped even when the operators told it to continue. Basic safety design.

But that's exactly what this question refers to...

Missed what part? What you quoted is to me exactly the same as a "run people over anyway" button. Sure, you and the article author can call it a "start" button if you like, but in this case it served perfectly well as a "run people over regardless of what the bus computer says" button.

Everybody who designs their self-driving bus to err on the side of caution.

Such buses will stop too often. If you don’t allow a human to tell the bus that it can continue (not a “run over people” button, but a “trust me, it’s safe to continue” button), it may never continue.

Advantage over a bus that drives more aggressively is that this doesn’t require the human operator to continuously pay attention.

Were the operators truly in the wrong or was it one of those classic cases of "self driving vehicle gets super skittish because it can't read body language", the operators overrode it based on body language and the other traffic participant did something unexpected and so it turned out the skittish AI was right?

Without video any attempt at allocating fault is mostly speculation.

I wrote software for that bus! I’m glad it didn’t hit the pedestrian because of a bug in one of the rtos drivers. Instead it crashed due to a bug in the human drivers.

It is a bug in the safety case. Single point of failure. Is not even ASIL B ( error - pedestrian hit - but no diagnosis - no big warning "overrighting controls could result in human injury"). If I was the Safety expert for such a project i would have never given a green light for this.

My reaction: Why did the bus not have a couple buttons in between 'Stop' and 'Go' - something like 'Creep Forward' and 'Go Cautious and Very Slow' - for crowded and uncertain situations? (I have no background whatever in self-driving vehicles, nor their software.)

Might you be able to address that aspect?

People are saying it's purely human driver error.. perhaps. But the situation often occurs where you are driving past crosswalks with people on the side waiting. If one were to suddenly step out just before you passed you would hit them. That wouldn't be on the driver it would be on the pedestrian.

Also, say an observant driver should have been able to tell that the Paralympian intended to walk. So it's human driver fault right? But maybe even then not entirely.. say the self driving system acts skittish way too often. Then it conditions the human drivers to be fed up with its automatic skittishness and just punch the proceed button routinely when it gets to intersections with pedestrians on the side.

Just putting this out there because I see no one suggesting the pedestrian could be to blame or recognizing the subtleties of the shared responsibility of human and self driving system.

Yep, and this is exacerbated in cities, and in areas where, for some reason, they make the bad choice of putting crosswalks near bus stops. Much of the time, people standing near crosswalks are just waiting for a bus, and usually successfully communicating that to drivers does not happen. And as we know, when there IS a bus using the stop, using the crosswalk is a lot more dangerous.

You live somewhere where people on the crosswalk don’t have right of way? That’s interesting. Where I live you have right of way on the crosswalk if you demonstrate intent to cross. There’s pretty much zero chance for the driver to not be liable in case of an accident.

Same here. If a pedestrian is waiting at a crosswalk, vehicle traffic must stop by law.

There are places where pedestrians IN the crosswalk have the right of way, but not AT the crosswalk.

Pedestrians can enter the crosswalk without showing (much of) intent to cross.

As an extreme example, a pedestrian could be standing on the sidewalk and suddenly take a step backwards onto the crosswalk.

In some countries if you’re waiting at the crosswalk, it’s assumed you will cross and cars will wait for you. I was yelled at in Berlin once for (unintentionally) standing at a crosswalk and not crossing (I wasn’t paying attention and didn’t realise, I was chatting with someone) while the car waited for me to cross.

The number of times I've been almost hit by a Tesla because I was walking in a parking lot and didn't hear them (because they didn't have any warning sounds) is too damn high. If they're going slow, they should play a sound or something. I check behind me, no cars, I cross the parking lot road, only to be almost hit by a Tesla because they just pulled out, and I didn't hear them. A simple sound, whooshing, beeping, whatever, would have solved that problem. Why Tesla didn't implement such a thing is anyone's guess.

Product page:

https://global.toyota/en/newsroom/corporate/29933371.html

"Low-speed automated driving with a focus on safety"

- "The e-Palette is equipped with a specially-designed automated driving system that includes control hardware, software, and advanced sensors such as cameras and LiDAR. Combined with high-accuracy 3D mapping and an operation management system, e-Palette will realize low-speed automated driving at SAE level 4."

- "To support safe operation, the e-Palette features an external human-machine interface designed to assist communication with those around the vehicle, including pedestrians, during automated driving. Both front and rear lamps on vehicle mimic eye contact to inform pedestrians of vehicle actions."

- "The automated driving system constantly monitors for obstacles across a full 360-degree field of vision around the vehicle and operates at a speed appropriate and in accordance with the surrounding environment. As a preventative measure, an on-board safety operator will also monitor the vehicle's movement and be prepared to take control if necessary."

So, no sound warnings to pedestrians is what I'm reading. Blind be damned.

I guess "Bus-driver orders self-driving bus forward at stop, striking visually impaired Paralympian disguised as able-bodied person", just doesn't have the same ring.

It's difficult to see how this could at all be Toyota's fault. The system stopped at the crosswalk correctly. The pedestrian was apparently not carrying any visually recognizable means to identify them as sight-impaired. There was even a crossing guard who also did not stop the pedestrian from aimlessly wandering into the intersection.

Is the problem actually that the electric busses are too quiet and the pedestrian didn't realize it was there? I know some cities have had problems with this, but I thought installing beeper-devices on electric busses was pretty standard by now. Seems like the sort of mistake that a podunk town in bumble-fuck nowhere would make, not something that would happen in Tokyo.

I'm assuming this person is used to walking around without a cane and has thus far managed without incident. Or maybe he's getting hit by cars all the time and this is the first we're hearing about it because it was an automated vehicle.

Regardless, there are a dozen scenarios that are more likely than anything being done by anyone involved with the bus, driving it, supervising it, or manufacturing it.

Please not more beepers. Some low level white noise is much less intrusive and easier to localize.

Some vehicles use a white noise that sounds almost like the vehicle fans. I think this works pretty well without filling the area with useless beeps.

My BMW beeps rather angrily if you dare to walk behind it whilst I am reversing. It's neat.

>It's difficult to see how this could at all be Toyota's fault

You're not trying very hard.

"Skittish AI stops for flagrantly false positives so frequently it trains the operators to override it at the drop of a hat" is a pretty common failure mode for automated systems.

I think you should append “due to human error” to the title.

This is why self-driving vehicles will never be accepted in the west. OK, I'm including Japan in the west for this.

- Self driving software correctly stops.

- Human decision making overrides, and says go.

- Pedestrian gets hit.

- Self-driving vehicles are taken out of service.

Does anyone else see the irony of this situation? The software was right, and the humans were wrong, so shelve the software, and have all humans drive buses for the rest of the paralympics.

In the west, it's the story that counts.

But in China, the math would be calculated differently. The self-driving buses would keep rolling, without human supervisors.

Much worse for the visually impaired are not cars but people on electric scooters going fast on the sidewalk.

Reply on news.ycombinator.com