Uber Suspends Driverless Car Program After Pedestrian Is Struck and Killed

[formulav8]

Uber Suspends Driverless Car Program After Pedestrian Is Struck and Killed

A self-driving car from Uber Technologies Inc. struck a woman who died Monday in Tempe, Ariz., local police say, in what is believed to be the first known fatality of a pedestrian from a driverless vehicle.

Following the accident overnight, Uber is temporarily pulling its self-driving cars off the roads in Tempe, San Francisco, Pittsburgh and Toronto, where it is testing them, a spokeswoman said. She said Uber is investigating the incident and cooperating with authorities.

The Uber vehicle, which included a human driver to assist, struck a woman who later died from her injuries, according to Tempe police. More details are expected after an investigation is concluded.

“Some incredibly sad news out of Arizona,” wrote Uber CEO Dara Khosrowshahi on Twitter Monday. “We’re thinking of the victim’s family as we work with local law enforcement to understand what happened.”

Edit: A liitle more info:

Tempe police Sgt. Ronald Elcock at a Monday afternoon news briefing identified the Uber driver behind the wheel as Rafael Vasquez, 44. He said the driver was not impaired when the car, a Volvo, hit the woman, who had stepped out of the crosswalk mid-block. The car was going about 40 mph in autonomous mode, police said.

Earlier, Tempe police spokesperson Detective Lily Duran said in an email the crash was reported around 10 p.m. Sunday night. She did not have details on when the woman, identified as 49-year-old Elaine Herzberg, died after she was taken to a hospital.

 

[rh71]

Makes you wonder how many people have been struck without fatalities.

And maybe I'm missing something, but why is a rideshare company doing the driverless testing when it should be a car/tech company who created/programmed them?

 

[Blackjack200]

Anybody else getting the feeling that progress on driverless cars may have been a wee bit oversold the last few years?

 

[PottedMeat]

And maybe I'm missing something, but why is a rideshare company doing the driverless testing when it should be a car/tech company who created/programmed them?

they did 'develop' and implement it themselves. then theres that whole google->waymo->otto->uber tech clusterfuck


even their old ceo knows 'laser is the sauce' to driverless tech
https://arstechnica.com/tech-policy...o-pound-of-flesh-kalanick-has-the-best-words/

 

[Muse]

I don't have stats but obviously one fatality isn't going to be the end of driver-less cars when cars with drivers kill thousands. I still feel safer in a plane than on a train (or in a car).

 

[ImpulsE69]

I don't have stats but obviously one fatality isn't going to be the end of driver-less cars when cars with drivers kill thousands. I still feel safer in a plane than on a train (or in a car).

You're talking 100's (are we in 1000's yet?) of driverless cars versus millions of cars with drivers. That ratio is going to need to come down just a wee bit.

 

[momeNt]

Anybody else getting the feeling that progress on driverless cars may have been a wee bit oversold the last few years?

I have heard from insiders, that people at Mobileye were thoroughly amazed these companies were moving forward so quickly with the autonomous driving, and they are the developers of the crash avoidance tech....

 

[GagHalfrunt]

You're talking 100's (are we in 1000's yet?) of driverless cars versus millions of cars with drivers. That ratio is going to need to come down just a wee bit.

It seems likely that the ratio will come down. Yes, there are going to be some teething problems and it will take some time to get things straightened out. That's the way of new tech. Whatever the math says now, it's going to change because the tech will keep getting better and there's nothing to suggest that humans won't keep getting worse. As long as the computers are not drunk, texting or putting on make-up they're going to win eventually because humans are too stupid to ever learn.

 

[Blackjack200]

You're talking 100's (are we in 1000's yet?) of driverless cars versus millions of cars with drivers. That ratio is going to need to come down just a wee bit.

Also, let's not forget that these "driverless" cars have human operators that are supposed override the system when it tries to mow someone down. So its certainly possible that the operators have already saved a few lives with the brake pedal.

 

[formulav8]

I think Uber is using Nvidia tech? That could partially explain why their stock is being pounded today. Down almost 5% right now.

 

[SKORPI0]

Maybe the operator of that driverless car stepped on the gas instead of the breaks?

 

[Muse]

Maybe the operator of that driverless car stepped on the gas instead of the breaks?

Is there no news concerning this beyond the fact that a pedestrian was killed by a driverless car that had a backup human at the wheel?

Personally, I haven't kept abreast of the technology but it has always seemed to me that there are serious challenges in getting it to work reliably. Even so, I have the impression it will take over and welcome the day. Seeing as drivers aren't going to get more reliable, as long as driverless gets a lot more reliable than relying on humans to make the decisions, that should determine where we go concerning "motorized" vehicles. And I imagine that some of the issues that make it difficult to get driverless reliable can be addressed, things like how we design roads, intersections, laws stipulating road/highway design, the entire vehicle environment design spectrum can be rethought to make it work better by and by.

I personally feel/think that most accidents happen at least in part because someone screwed up, i.e. the driver. But with a pedestrian death, you suspect that the victim had a hand in it (excuse the pun!), IOW maybe they "didn't look both ways," something that I see all the time where I live. For example, a child can dart out from between parked cars chasing a ball. What driverless (or drivered) vehicle could avoid that?!

 

[Kaido]

I have heard from insiders, that people at Mobileye were thoroughly amazed these companies were moving forward so quickly with the autonomous driving, and they are the developers of the crash avoidance tech....

That's one of the things that bugged me about Tesla. Their car was in self-driving mode on the highway when it ran into a semi-truck & killed the driver. Now, that sounds insane, because it shouldn't have hit the semi-truck at all (obviously), BUT - the MobileEye system in the car at the time did not have lateral detection or whatever. i.e. that's literally not a feature that was built into that version of the car (the explanation is a bit more complicated - the hardware version the car had apparently didn't detect stuff above the hood, considering it a sign or a bridge or whatever, to prevent false braking & possible accidents from that, plus the semi-truck was white against a bright sky). But Tesla is selling it as "Autopilot". It's not Autopilot. It's dumb to call that because it's confusing to the consumer. It's semi-automated driving. Like, with Honda's Sensing system (similar but not quite as advanced - it has adaptive cruise, lane-keep assist, etc.), they are VERY clear that it is only a convenience feature & not a safety feature. Tesla is (imo) pretty reckless in how they currently market Autopilot. Eventually it will be near-perfect, but it's going to take years to get there, and calling it Autopilot now is (again imo) a very bad idea because it sends the wrong message to the consumer.

And places like Uber are pushing the technology faster than it should be going because they really want to make it out to the market as quickly as possible, and now people are paying the price. But it's also, unfortunately, something that is going to happen, and when you look at it statistically, this technology really DOES need to happen because if we can trade 30,000+ automotive deaths in America annually for just a few thousand, then it's more than worth it to continue to develop it. But what a rotten way to go...killed by a robot. Just an awful situation

 

[rh71]

But Tesla is selling it as "Autopilot". It's not Autopilot. It's dumb to call that because it's confusing to the consumer. It's semi-automated driving. Like, with Honda's Sensing system (similar but not quite as advanced - it has adaptive cruise, lane-keep assist, etc.), they are VERY clear that it is only a convenience feature & not a safety feature. Tesla is (imo) pretty reckless in how they currently market Autopilot. Eventually it will be near-perfect, but it's going to take years to get there, and calling it Autopilot now is (again imo) a very bad idea because it sends the wrong message to the consumer.

They went into autopilot discussion in this video for a bit. They say it's disabled but the hardware is in place for when it becomes legal. I don't think they're advertising autopilot to the consumer outright.

 

[darkswordsman17]

Also, let's not forget that these "driverless" cars have human operators that are supposed override the system when it tries to mow someone down. So its certainly possible that the operators have already saved a few lives with the brake pedal.

In Arizona no driver is necessary now (they passed a law fairly recently that allows them to operate driverless cars without human operators at all on Arizona roads now). But then that's not the case here (did have a driver), but they're already moving towards it.

Added: Waymo released video of some of the fully driverless ones taxiing people.
https://www.carscoops.com/2018/03/d...s-first-reactions-video-autonomous-pacificas/

I wonder how good of awareness those vehicles have (personally for me, I'd have some area where I'd be throwing dummies, even shooting them out of like cannons at high speed and see how quickly it reacts). Like would it see someone running in its direction (or general area where its going to be) from how far away and how would it react. Basically give it worst case scenario (what happens if an object, or person falls off/out of a car ahead at highway/freeway speeds for instance). But at the same time, not freak out and cause an accident by reacting quickly when it wasn't even going to be an issue.

They went into autopilot discussion in this video for a bit. They say it's disabled but the hardware is in place for when it becomes legal. I don't think they're advertising autopilot to the consumer outright.

I think that's just the Model 3, I am pretty sure they were marketing it on the Model S though (there was a spat between the company that used to provide the hardware and Tesla when that accident happened, and I think there were other people - by that I mean like car industry people - saying that Tesla should not have been allowed to use the term Autopilot to describe their system as it was giving people a false belief of what it was actually capable of; I believe some government agency even investigated over that specific issue).

I'd say it applies to the Model S and X based on this:
https://www.tesla.com/autopilot

https://electrek.co/2018/03/15/tesla-autopilot-crash-cyclist-scooter-denied/

Yeah, I think they should be clamping down on its use. At minimum people need to be going through some sort of training with regards to it before they let them operate it.

 

[Kaido]

I'd say it applies to the Model S and X based on this:
https://www.tesla.com/autopilot

https://electrek.co/2018/03/15/tesla-autopilot-crash-cyclist-scooter-denied/

Yeah, I think they should be clamping down on its use. At minimum people need to be going through some sort of training with regards to it before they let them operate it.

Agreed. If the system didn't have Autopilot, Tesla is always quick to say so (or say if it was not enabled at the time of the crash), but if it IS installed & enabled, then they still blame the driver because hey, just because it says Autopilot doesn't mean it really has "Autopilot", doesn't that make sense?? lol

 

[Red Squirrel]

Wait I did not realize driverless cars were ACTUALLY live yet. I thought the tech was still highly in prototype stage and that there was always a driver.

Interestingly enough driverless cars and true "auto pilot" is probably more complicated than autopilot on a plane. A plane is in the open sky with nothing around it, and has lot of time and distance to react. Using various sensors like altitude and compas, and ADS-B and radar, they can more or less avoid each other. Cars on the other hand are always operating on split second decisions and moving within feet of obstacles and have to account for road conditions and stop and go a lot. Planes don't stop, they just move around anything and have lot of time to do so.

Then again, ships are probably even simpler than planes in terms of traffic avoidance because of being on the open sea with relatively little traffic compared to a road, yet the US military has managed to crash at least 2 within a year. lol

 

[rh71]

^ there's a TED talk on youtube that showed how predictive the systems have become even on surface streets. It's only a matter of running them through many many instances now to account for all scenarios it seems.

Skip to 7:49 if you want:

 

[Red Squirrel]

Yeah guess I'm really underestimating the sensory tech. As long as they don't get covered in snow and ice... Nothing will save you there. I imagine they do a self test before they start though, so it would tell you that it's blocked. Maybe they also have deicers to keep them from getting full of snow while driving in storms. To me that would be the biggest issue. On TV shows they always show them in a perfect scenario, nice clean paved road with lines (what's lines? My city does not know what that is) and even sidewalks and no snow etc but in the real world it's rarely like that.

 

[darkswordsman17]

Is there no news concerning this beyond the fact that a pedestrian was killed by a driverless car that had a backup human at the wheel?

Personally, I haven't kept abreast of the technology but it has always seemed to me that there are serious challenges in getting it to work reliably. Even so, I have the impression it will take over and welcome the day. Seeing as drivers aren't going to get more reliable, as long as driverless gets a lot more reliable than relying on humans to make the decisions, that should determine where we go concerning "motorized" vehicles. And I imagine that some of the issues that make it difficult to get driverless reliable can be addressed, things like how we design roads, intersections, laws stipulating road/highway design, the entire vehicle environment design spectrum can be rethought to make it work better by and by.

I personally feel/think that most accidents happen at least in part because someone screwed up, i.e. the driver. But with a pedestrian death, you suspect that the victim had a hand in it (excuse the pun!), IOW maybe they "didn't look both ways," something that I see all the time where I live. For example, a child can dart out from between parked cars chasing a ball. What driverless (or drivered) vehicle could avoid that?!

Driverless cars can "see" more than humans can in some regards. And generally their computer systems are quicker reacting than humans. There were people lauding the Tesla Autopilot after in some video it started slowing down and then an accident happened. With development, we'll hopefully create better sensors (and add new ones) that can detect more. Plus they want cars to communicate, so that if there's a crash miles ahead it can forewarn other cars to help prevent pileups, or help it see things it couldn't directly itself. Eventually they want that to be able to offer improved efficiency (like when you're at a stoplight and there's like a chain of delayed reactions as each driver waits for the one ahead to get moving, they could instead get all the cars synchronized to start accelerating at a reasonable but consistent rate, but it would reduce the queuing issues at lights). You'd also have the lights and cars communicating, so say when there's one car on a side road, it could give it a quick opening but quickly let the more trafficked road not get bogged down as much.

Agreed. If the system didn't have Autopilot, Tesla is always quick to say so (or say if it was not enabled at the time of the crash), but if it IS installed & enabled, then they still blame the driver because hey, just because it says Autopilot doesn't mean it really has "Autopilot", doesn't that make sense?? lol

Tesla actually has data that shows what the driver does though, so they have a good idea of if the driver actually took control, and if they did, what they did (so if they jerked the wheel, or hit the gas instead of the brake they'd know). Its why Toyota was stuck in a bad position, their data revealed it was almost always the drivers that were hitting the gas instead of the brake or didn't do certain other things. But they couldn't come out and go "Toyota owners are idiots that keep pressing the gas instead of the brake", because they didn't really want to insult people. Unfortunately people are...people, and often double down on stupid instead of going "oh, wow, yeah, I screwed up, not sure why I did that" they go "no, their data is wrong, I've been driving for 30 years, I know the difference between the gas and brake pedals, rabble, rabble, rabble". That happened one time and I think police showed the woman video where her brake lights did not come on at all, and she still insisted she hit the brakes and not the gas. We have problems admitting that we're wrong and accepting our limitations. I know I've done things at times where I was like, whoa, what am I doing (almost merged into a car as didn't see it in my mirror initially). I haven't had any serious accidents (only accident was hitting a deer, where I was lucky and my reflex wasn't to hit the brakes as I hit just the lead deer, there were several others and I would've ended up having them smash the windshield and side of the car), but I'm not going to pretend I'm infallible. Hell I was even an idiot kid once too.

I do worry that in the interim where driverless tech is available but not quite robust enough, we're gonna have lots of problems where people will then get hugely distracted (just a matter of time before we have instances of people having sex while the car drives, sleeping, doing anything and everything else; I mean, people already do that shit with just cruise control or even before that was common, its gonna get worse).

 

[darkswordsman17]

^ there's a TED talk on youtube that showed how predictive the systems have become even on surface streets. It's only a matter of running them through many many instances now to account for all scenarios it seems.

Skip to 7:49 if you want:

Thanks will have to check that out. I saw an article that showed what the car "sees" from Waymo but hadn't gotten around to actually checking it out.

Yeah guess I'm really underestimating the sensory tech. As long as they don't get covered in snow and ice... Nothing will save you there. I imagine they do a self test before they start though, so it would tell you that it's blocked. Maybe they also have deicers to keep them from getting full of snow while driving in storms. To me that would be the biggest issue. On TV shows they always show them in a perfect scenario, nice clean paved road with lines (what's lines? My city does not know what that is) and even sidewalks and no snow etc but in the real world it's rarely like that.

Funnily enough, that is a problem currently. They have to have people go and manually clean off the sensors. Here in Arizona there's lots of dust (and there's haboobs during certain times of the year) so its not just ice and snow that cause problems. Yeah I'm sure it gives some feedback, but the problem is, it probably isn't all or nothing. It could have something reducing its capability, but hopefully it'll have some self check test where reduced capability is noted and it alerts you so you can check it out. But I'd guess eventually they'll end up with sensors integrated into normal parts of the car (like the lights), and there are cars that have headlight cleaning systems (where it extends a little arm that has a water jet like the windshield washer ones). But I'd guess the plan is to make sensors as small and cheap as possible, and then have them all over the car, so that if some are limited or go bad it has some redundancy, but it'd give it a full view, with probably some more powerful sensors in certain key spots.

That brought up a funny discussion. Can't recall where I heard it, but someone was saying how people could screw with autonomous vehicles by changing/screwing with lights, signs, lines. The thing is, that would work on humans too. I think over time we'll adapt our roads to them more, there will be "hidden/invisible" sensors/markings that the car reads with certain sensors. It would actually help make our roads more clean too (they've done studies that show that more signs actually don't help driving safety as it provides too much info and people start just dismissing most of it anyway; there was somewhere where they removed all of the street signs and people drove more safely because they had to actively be more aware since they had nothing guiding them; would be interesting if that held up over time or if people eventually start pushing boundaries again).

Oh and on the sensors, the Waymo vans have these parts that stick out and have silver colored spinning cylinders on them. I can imagine they probably have issues with people screwing with them (either out of curiosity or just to be assholes).

 

[VirtualLarry]

this technology really DOES need to happen because if we can trade 30,000+ automotive deaths in America annually for just a few thousand, then it's more than worth it to continue to develop it. But what a rotten way to go...killed by a robot. Just an awful situation

Disagree. Freedom and autonomy and self-determination are FAR more important, than a few thousand deaths each year.

 

[ElFenix]

Disagree. Freedom and autonomy and self-determination are FAR more important, than a few thousand deaths each year.

my dad can't see well enough to drive himself. a car that can drive for him is freedom and autonomy.

 

[VirtualLarry]

my dad can't see well enough to drive himself. a car that can drive for him is freedom and autonomy.

Yes, until he has an unpopular political opinion or two, and becomes "unwanted", then he might be signing his death warrant, just by getting into an autonomous vehicle.

See China's "Social Credit System" for what they are implementing. Hopefully, that never happens here. And if they try it here, there's going to be a lot of 2A defenders of freedom and self-determination.

 

[Muse]

Driverless cars can "see" more than humans can in some regards. And generally their computer systems are quicker reacting than humans. There were people lauding the Tesla Autopilot after in some video it started slowing down and then an accident happened. With development, we'll hopefully create better sensors (and add new ones) that can detect more. Plus they want cars to communicate, so that if there's a crash miles ahead it can forewarn other cars to help prevent pileups, or help it see things it couldn't directly itself. Eventually they want that to be able to offer improved efficiency (like when you're at a stoplight and there's like a chain of delayed reactions as each driver waits for the one ahead to get moving, they could instead get all the cars synchronized to start accelerating at a reasonable but consistent rate, but it would reduce the queuing issues at lights). You'd also have the lights and cars communicating, so say when there's one car on a side road, it could give it a quick opening but quickly let the more trafficked road not get bogged down as much.

When they really get these technologies worked out it is bound to help with highway congestion tremendously. This especially when all cars have the technology running and communicate with each other. With computers and laser sensors and the A.I. that's going to coordinate it all, we are willy nilly moving into science fiction territory here.