Hacker News Reader

Comment by potato3732842 | Hacker News Reader

The real problem is that it didn't recognize and stop for the stop signs on the school bus. The child is basically an afterthought designed to appeal to the emotion of those whom logic fails. Even if no kids materialized the way a bus stop works (bus stops, then kids cross) means that detecting the kid really shouldn't be the primary trigger for stopping in this situation, the stop sign needs to be recognized and acted upon. Any ability to detect and stop for pedestrians is secondary to that.

View on HN

I don't agree with this. Not hitting pedestrians should not just be an afterthought. Of course the car should recognise the stop sign, but there are cases in which stop signs are obstructed or missing, and in those cases pedestrians should still not be hit by a car.

>I don't agree with this. Not hitting pedestrians should not just be an afterthought.

You're disagreeing with something I didn't say. There's a difference between afterthought and the primary initiator of the stop in a situation like this.

>Of course the car should recognize the stop sign, but there are cases in which stop signs are obstructed or missing, and in those cases pedestrians should still not be hit by a car.

The surprise pedestrian test is one that any vehicle can be made to fail by sheer physics. Avoiding errant pedestrians like in the video will likely only come as a byproduct of better situational behavior by self driving vehicles. The overwhelming majority of drivers know to ignore the speed limit if the situation is rife with pedestrians or otherwise sus and are generally fine with missing/obstructed stop signs. I don't know what route self driving software will take to approximate such behavior but it likely will need to.

> The surprise pedestrian test is one that any vehicle can be made to fail by sheer physics.

There's different degrees of failure as well, did the Tesla try to brake beforehand or apply brakes after hitting the doll?

> The overwhelming majority of drivers know to ignore the speed limit if the situation is rife with pedestrians or otherwise sus and are generally fine with missing/obstructed stop signs.

This sentence makes me want to cry. Sure, I may know drivers treat speed limits as minimum acceptable speed at best, but seeing it unironically spelled out like this hurts.

Please, it's just two simple words: 1) speed 2) limit. L I M I T. It's the maximum allowable speed. No, it's not the minimum speed, no it's not the target speed, no, it's not "more of a guideline than anything". You are not allowed to go faster.

It says nothing about a minimum speed. It says nothing about what speed you should drive at. All it does is limit the maximum speed. Repeat after me, it's a speed L I M I T.

Just kidding though, I know even speaking like I would to a 5 year old won't do it, this mentality runs far too deep. There's no hope.

Yes, but recognizing a pedestrian when he jumps in front of your car is useless -- you don't have time to stop anyway.

What you WANT to recognize is conditions when such an event is possible (obstructed vision) and to slow down in advance even if you don't see/detect any pedestrians at the moment.

This obviously includes the case with the school bus and the stop sign but, as you correctly point out, is not limited to that. There are more cases when a pedestrian, especially a child, can jump under your car from behind a big vehicle or an obstacle.

Recognizing these situation and slowing down in advance is a characteristic trait of a good-intentioned experienced driver. Though I think that most of the time it's not a skill you have straight out of driving courses, it takes time and a few close calls to burn it into your subconsciousness.

At 25 mph, which I would hope would be the speed limit on roads next to schools, slamming on the brakes even seconds before colliding with children can make an enormous difference in how fast the car is going when it hits the kid.

Speed is the factor in collisions (other than weight), and modern brakes are incredibly good.

Not to mention that the car, with it's 360 degree sensors, could safely and efficiently swerve around the children even faster than it can brake, as long as there's not a car right next to you in another lane -- and even if there is, hitting another car is far less dangerous to their life than hitting the children is to yours.

These things should be so much better than we are, since they're not limited by unidirectional binocular vision, but somehow they're largely just worse. Waymo is, at best, a bit better. On average.

25mph is 36 feet per second, about the length of a school bus. The stopping distance at 25mph is 30 feet, assuming perfect reaction time and dry pavement. Human reaction time is about 750ms, so stopping distance is about 2 school bus lengths. You don't have seconds.

25mph is too fast for any street where kids may jump out behind parked cars. Not just school zones, but all residential streets. There's a knee at about 20mph in the traffic fatality stats where below that speed pedestrian collisions are unlikely to be fatal. Above 20mph fatalities become more common quite quickly.

This is why school buses flash yellow warning lights before deploying the stop sign and opening the doors.

It should never be the case that someone is surprised by an instantaneous bus stop. The are plenty of context clues in advance. Including the fact that the bus is around at all, which should already heighten attention.

For the knee to be useful you have to go below it -- thus 15 mph is a much better neighborhood and school speed limit.

To add a data point. In Belgium, the max speed around schools is enforced to be 30 km/h. That should be a bit less than 20mph.

Even if you won't prevent the collision, reducing the velocity at which it happens is still very desirable, And, given that kinetic energy of the car is proportional to velocity squared, even a little bit of reduction means a lot less energy dumped into the pedestrian.

Sure, if you've hit the brakes by then. The reaction time delay means you travel significant distance before hitting the brakes.

> The stopping distance at 25mph is 30 feet,

Stopping is nice, but not the entire point of braking. The lower the collision speed, the better.

> Human reaction time is about 750ms

No. Human reaction time is around 250ms on average. What you point to is the time it takes to react to an unexpected stimuli. The number I've seen quoted is about 1s. But that assumes a completely unexpected event that you're not prepared for at all.

So if you're mindlessly passing a school bus at 25mph, then a 1s delay is expected. But if you're doing so with your foot covering the brake while hyper focused on reacting to any sudden movement in front of you, you can do much, much better than 1s. Of course, at that point you might as well drive correctly by slowing down.

I wasn't talking about human reaction time.

Kids don't get dropped off only at school. They get dropped of at their homes and can be on 55 mph roads. You are quibbling with the main point anyway. The whole reason for the stop signs on busses is because kids will be kids and really its too late when they run out.

> At 25 mph, which I would hope would be the speed limit on roads next to schools

I regularly drive on a two lane 55mph highway that school buses stop on and let kids out.

It runs through a reservation and has no sidewalks at all.

> modern brakes are incredibly good.

They're probably not worth that much of a superlative, and they're fundamentally limited by the tires.

This is just a pet peeve of mine, since it is used by people to argue that modern vehicles are so much vastly better than cars in the 1980s that we should be able to drive at 90mph like it is nothing.

But reaction times and kinetic energy are a bitch, and once traction control / stability assist hits its limits and can't bail you out, you might find out the hard way that you're not as good of a driver as you think you are, and your brakes won't stop you in time.

> Speed is the factor in collisions

This I will definitely agree with. Say it louder for everyone in the back.

There are a lot of videos Waymo has posted of split second swerves they’ve had to do in SF and Austin. It looked to me like a combination of hard braking and swerving could have avoided the collision. Now to be fair to Tesla, the dummies in this test didn’t look very realistic, but not even slowing down for the school bus shows that FSD is not close to being ready for unsupervised use.

Bah. The car should slow down if the view is restricted or when it's passing a bus, especially a school bus, regardless if there's a stop sign or not.

Merely slowing down is not enough. If the car doesn't come to a complete stop and remain stopped while the bus has its stop sign extended, it's driving illegally.

I want them to drive legally, but I also want them to be able to react to objects even if they don't have signs on them.

Signs are often obstructed by trees or are simply inadequate for safe driving, even when combined with the rules of the road. Any even "partially" automated driving should be able to trivially recognize when its view is compromised and proceed accordingly.

Totally agree. "Stop for a stopped school bus" and "Drive slow enough to stop in an emergency, in areas where pedestrians are obscured" are two separate, barely related problems that car companies need to reliably solve.

They're saying if there's a school bus inside of a narrow corridor, that a prudent driver would slow down and use caution IN ANY CIRCUMSTANCE.

They're obviously not arguing that the car shouldn't stop with a sign deployed.

Arguing from a point of bad faith doesn't advance the discussion.

I'm saying slowing down is better, but not enough. If the manufacturer can program the system to slow down when it sees a school bus inside of a narrow corridor, then it can also program the system to (correctly) stop in that situation.

Reply on news.ycombinator.com