It's getting harder than ever to see out of modern "safer" cars, too. I regularly entirely lose cars hidden by the B-pillar on my Model 3. I have to rock my head back and forth before each lane change to make sure something's not hidden there. It's compounded by the fact that the driver's side mirror doesn't physically move far enough to the left to cover the blind spot like it would in most cars, and our terrible flat mirrors in the US don't provide the wide angle that european driver's side mirrors have. (still working on sourcing a european mirror glass; currently Tesla only sells the entire mirror body+motors+glass as a single part, for $400ish)<p>It's VERY easy to lose a pedestrian in the A-pillar (windshield pillar) of modern cars, particularly due to how large these units have gotten to accommodate side curtain airbags.<p>Note that the fixtures on the body don't have to 100% optically obscure things for them to be 'invisible' to the driver.<p><a href="https://www.portsmouthctc.org.uk/a-fighter-pilots-guide-to-surviving-on-the-roads/" rel="nofollow">https://www.portsmouthctc.org.uk/a-fighter-pilots-guide-to-s...</a><p>>It gets even worse. Not only can we not see though solid objects; research has shown that we tend not to look near to the edges of a framed scene. In plain language, we tend not to look at the edges of a windscreen. So, not only do the door pillars of a car represent a physical blind spot, but our eyes tend not to fixate near to it, leading to an even bigger jump, or saccade, past a door pillar. This is called windscreen zoning.
I love cars, driving, and freedom in general. Just wanted to put that up front, because I'm going to suggest this isn't an issue with the manufacturer of motor vehicles, but with the massive government system(s) built around regulating them.<p>There is nothing stopping this terrific government thing everyone is raving about from passing some regulations that require people who operate motor vehicles on the roadways, actually be capable of operating motor vehicles on the roadways. Instead, we get short HS drivers education classes, that really only teach you what the 2 pedals in American cars does what, and perhaps a quick refresher on how "wheels" work.Then you get put on the highway, and taught to speed, but not too much, and generally ignore every other major rule there is.<p>I ride motorcycles, have taken several high performance driving classes, as well as drive in an amateur (wheel to wheel) auto racing series, and I still catch myself making mistakes on the roads. Its not an easy task, and the current system is woefully under performing when it comes to equipping people to handle it. This needs to be fixed by changing the culture around driving, if for no other reason than because sometimes machines fail, and the person behind the wheel needs to be responsible.
Lately my (foot) commute has come to involve a whole lot of people blasting down residential streets at 40mph while staring at their phones and only barely slowing down in acknowledgement of stop signs.<p>My current belief is that the single most effective pedestrian safety feature that car manufacturers could possibly implement is some mechanism for preventing Waze from working.
> Nearly 6,000 pedestrians were killed in U.S. traffic accidents in 2017, the latest year data were available, according to the National Highway Traffic Safety Administration. That was up 35% from 2008<p>I wonder how much of this is due to distracted driving because of smart phones. I would bet it's more than we care to admit.<p>This is where the material from the college engineering ethics course I was required to take kicks in. It's stuff like this we should pay attention to.<p>> At 20 miles an hour, the cars struggled with each test, AAA found. The child was struck 89% of the time, and all of the cars hit the pedestrian dummy after making a right turn. The systems were generally ineffective if the car was going 30 mph. The systems were also completely ineffective at night<p>When companies advertise these systems, I wonder how many people feel they can pay less attention while driving.<p>I also wonder what's up with their QA testing. This stuff could use some chaos engineering thrown at it.
Car companies seem complicit in promoting distracted driving in their commercials, too, by promoting automatic braking for drivers who aren't paying attention:<p>1. <a href="https://youtu.be/bS19g7Va6jg?t=30" rel="nofollow">https://youtu.be/bS19g7Va6jg?t=30</a><p>2. <a href="https://www.youtube.com/watch?v=T9d8PPrCB38" rel="nofollow">https://www.youtube.com/watch?v=T9d8PPrCB38</a>
1. Even if they fail to stop 90% of pedestrian accidents, that an improvement from the 100% of previous car generations.<p>2. These systems are primarily for avoiding car-on-car collisions. They're <i>very</i> good at that. Which is maybe why it's not mentioned in the article...
An interesting study on US pedestrian fatalities: <a href="https://www.ghsa.org/sites/default/files/2019-02/FINAL_Pedestrians19.pdf" rel="nofollow">https://www.ghsa.org/sites/default/files/2019-02/FINAL_Pedes...</a><p>Some notes: Pedestrian fatalities are up in the last decade and there isn't a great smoking gun on why. The "every driver is starring at a smartphone" explanation doesn't have much evidence. The vast majority of pedestrian deaths happen at night, and not at an intersection. Almost all the additional pedestrian deaths are at night; daytime fatalities have been flat since 2008.
You can also look this from a completely different mentality.<p>> The child was struck 89% of the time, and all of the cars hit the pedestrian dummy after making a right turn.<p>That means that in that simple no-turn circumstance, the car DOESN’T hit the child 11% of the time.<p>It’s better than nothing, and it’s a start. You still have to drive your car. You still need to stay off your phone. But these systems only have to work <i>sometimes</i> for them to be worth installing.
I think a huge part of the problem is that car companies aren't properly incentivized to protect pedestrians. It's easier and cheaper to prevent my car from hitting something that can damage it than to prevent my car from hitting something that it can damage (easier to detect buildings and other cars than to detect children). If you were the CEO of a car company and had a finite amount of money to spend on research on driver assist features would you rather spend it on the easy problem of protecting your customers or the expensive problem of protecting other people? Furthermore, as a consumer, would you rather buy a car that costs x dollars that protects you or costs x+y that protects other people?
It seems to me this encouragement of partial attention while driving just doesn't work for human reasons. The more you give people an excuse not to participate the more they will withdraw and blame technology when something goes wrong. You're either driving or you're not - there is no middle ground.
Pedestrian safety is of almost not value to us manufacturers. The could improve it significantly but there were actually a couple cases where manufacturers would strip off pedestrian safety features off cars originally designed to European standards to make them more aggressive looking.
It took a bunch of old NIMBYs to have scooter companies install speed limits within certain locations. We have the technology to have cars automatically limit speeds to the maximum allowed on the given road. We choose not to. Instead people keep getting killed.<p>This is not pie in the sky science fiction to drool about. We could do it now. We could do it 5 years ago.
I am 100% blind. This article confirms my fears. In the future, it will be more dangerous to walk the streets then it already is. I was born with an deep fear of automatic doors. And I will likely die due to the autonomous car revolution.
Relatedly, a <i>1939</i> vision of the future (discussed recently at <a href="https://news.ycombinator.com/item?id=21104762" rel="nofollow">https://news.ycombinator.com/item?id=21104762</a> ) shows a simple --- but probably quite expensive --- solution; people and cars can't collide if they're separated:<p><a href="https://upload.wikimedia.org/wikipedia/commons/5/59/Street_intersection_Futurama.jpg" rel="nofollow">https://upload.wikimedia.org/wikipedia/commons/5/59/Street_i...</a>
Sadly, this is well known inside the industry. The functionality of these systems significantly degraded when dark.<p>I've been told that there is similar degradation in the performance of autonomous emergency braking and forward-collision warning not only when dealing with pedestrians, but with cars that have their lights on and are perfectly visible to human drivers.
If any of those systems use deep learning, they _by design_ will not necessarily recognize a _dummy_ as a _person_, particularly in difficult lighting conditions or if the dummy does not show up on IR (if the system is multi-spectral).
That headline seems exceedingly tautological to me. If a safety feature fails, then the situation isn't going to be the safest, or if it was, who would notice?
> the Chevrolet Malibu, Honda Accord, Toyota Camry and Tesla Model 3 ... When testers drove the cars directly at a dummy crossing the road in the dark, however, the system failed not only to stop or slow the car but also to provide any alert of a pedestrian’s presence before a collision.<p>Oh good.