Uber Suspends Driverless Car Program After Pedestrian Is Struck and Killed

[ImpulsE69]

There's a piece here no one is talking about. The backup driver. You guys really think that some person is just going to sit there hands on wheel while driving everyday but not doing anything? Think again. It will become routine and they will be back to not paying attention, so this whole idea that a human behind the wheel not doing anything is going to stop screw ups isn't as foolproof as you think it is. This is not the same thing as a pilot doing his daily job. This is joe schmoe going too and fro drinking his cup-a-joe and watching his favorite show paying absolutely no attention to where he go.

It always baffles me why everyone is SO concerned about human drivers though. It's just one more thing that can kill you, and I'm guessing damn near everyone of you doe plenty of other things that can and will kill you. It's to the point it's just another "omg we gotta fix this issue because it's a threat to me!" moment. I'm not against computers driving cars, but this whole movement to replace humans completely is a bit much.

 

[Red Storm]

After seeing the video... I have to think something was up with the car's sensors. I thought the lady came from the other side, like off the median right into traffic and then she was hit a moment later. From the video it looks like she already crossed one lane. Due to the dark spots I don't expect a human driver to have been able to avoid the person in that scenario, but I thought these self driving cars had the tech to "see in the dark" and avoid crashes in these specific scenarios? Maybe the bike screwed with the sensors? I dunno, but to me it now looks like something must have been up with the car's sensors.

 

[DietDrThunder]

Never mind, I hate photobucket now.


So I captured a image of where the woman first became visible to their camera system. From this image you can see that there are no parked vehicles on the left lane, she crossed all of one empty lane and is partially through the second lane when she became visible to the camera. She SHOULD have been picked up by the LIDAR. Even if they were using a crude RADAR, she should have been detected by the system.

 

[rh71]

Interesting that the autonomous vehicle was also speeding - 40 in a 35.

 

[PeterScott]

Never mind, I hate photobucket now.


So I captured a image of where the woman first became visible to their camera system. From this image you can see that there are no parked vehicles on the left lane, she crossed all of one empty lane and is partially through the second lane when she became visible to the camera. She SHOULD have been picked up by the LIDAR. Even if they were using a crude RADAR, she should have been detected by the system.

I agree, Uber's systems seems to have really messed up big time. Object recognition not refined enough? Tolerance turned too low?

Corners cut by Uber. Who'd could have guessed that?

 

[GagHalfrunt]

So you're saying whatever device you are using to view these forums has never locked up, reset, crashed, rebooted, resized, or behaved in any other unplanned or illogical manner?

No zippy, that's not what I'm saying. What a sad, pathetic attempt at deflection that was. You can almost smell the desperation in that.

What I'm saying is that the problems I might have when using the forums are like 10 billion times more likely to be human error. My computer doesn't crash, I hit the red x instead of hitting the minus to minimize the browser window. My keyboard doesn't malfunction, I typo. The browser doesn't hang, I get distracted. My machine doesn't spontaneously reboot, I'm far more likely to smack the reset button while reaching for a pen.

Dumb ass.

 

[slayer202]

After seeing the video... I have to think something was up with the car's sensors. I thought the lady came from the other side, like off the median right into traffic and then she was hit a moment later. From the video it looks like she already crossed one lane. Due to the dark spots I don't expect a human driver to have been able to avoid the person in that scenario, but I thought these self driving cars had the tech to "see in the dark" and avoid crashes in these specific scenarios? Maybe the bike screwed with the sensors? I dunno, but to me it now looks like something must have been up with the car's sensors.

She comes out of nowhere on the video but you're right, you'd think the tech would have picked her up (she shares some blame too, walked right into the path of a car). "Driver" wasn't paying attention and may or may not have been able to see better than the camera/video that was posted. Seems like a screw up all around. So I guess uber takes the brunt of it, followed by the pedestrian. I think for this whole thing to work, pedestrians are going to have to adapt and not assume right of way 100% of the time. This woman was not too bright, pun partially intended

 

[PeterScott]

Interesting that the autonomous vehicle was also speeding - 40 in a 35.

Nope, it's doing 38 MPH in a 45 MPH zone.

That zone there is being incorrectly reported as 35 MPH.

It is in fact a 45 MPH zone. Unless this is some kind weird trap. Because the sign immediately before and after the collision area says 45 mph.

Here is the sign in streetview.
https://www.google.com/maps/@33.435...4!1sx-K4_17J8MVthFRapvIa2A!2e0!7i13312!8i6656

And no, it hasn't been changed, someone drove through after the accident with their camera recording video, and it still says 45 MPH.

 

[zinfamous]

Retired. And yes, I do see all those idiots not paying attention to the road. Too many gadgets in cars now. But I will still trust myself and not some computer to do the driving for me.

But we aren't talking about you. We're talking about those other people that you are also complaining about. You agree with everyone else, then turn the argument on yourself, again. Again: not about you, it's about distracted assholes running into you.

It's a long-accepted truth among city planners and such that the only reason we have traffic, and traffic accidents, are because of humans and their stupid brains. Yes, it sounds icky to be on the verge of the new paradigm and as humans we tend to be slow to adapt and accept these truths, but robot cars are the answer, and they are the only answer. It's happening, anyway, and I guess many will stubbornly accept it, if they ever do. Fortunately, stupid humans (meaning: all of us) won't really have a choice in the end. This is the future, and it is happening.

 

[DietDrThunder]

In my opinion this form of testing being conducted on the general public without strict guidelines by the NTSB, without controls in place, and without our consent is no better than the Tuskegee syphilis experiments that were conducted between 1932 and 1972. The only major difference is that, whether we agree with this testing or not, we are aware that it is going on.

This is government and industry run a muck to provide a "safe" system without the full consideration of safety during testing.

By the way, other than the person behind the wheel being a driver, what qualifications does this person have for being a "fail safe" mechanism?

 

[Red Storm]

In my opinion this form of testing being conducted on the general public without strict guidelines by the NTSB, without controls in place, and without our consent is no better than the Tuskegee syphilis experiments that were conducted between 1932 and 1972. The only major difference is that, whether we agree with this testing or not, we are aware that it is going on.

This is government and industry run a muck to provide a "safe" system without the full consideration of safety during testing.

By the way, other than the person behind the wheel being a driver, what qualifications does this person have for being a "fail safe" mechanism?

There is no way any human driver was going to be able to react in time to avoid that person, let alone someone who is there as a backup and has even less time to react.

We'll have to wait and see for info on the car's sensors, and whether or not Uber had a part in the car not detecting the person.

 

[DietDrThunder]

There is no way any human driver was going to be able to react in time to avoid that person, let alone someone who is there as a backup and has even less time to react.

We'll have to wait and see for info on the car's sensors, and whether or not Uber had a part in the car not detecting the person.

I never said a human driver would have been able to react in time. I'm saying that adequate safety measures are not in place and testing at this point should not be where the general population could be at risk.

More to come:

Fox
http://www.foxnews.com/auto/2018/03...uldve-spotted-pedestrian-in-deadly-crash.html

AP
https://www.apnews.com/6877b1912111...ises-questions-about-Uber-self-driving-system

Smith said the video may not show the complete picture, but "this is strongly suggestive of multiple failures of Uber and its system, its automated system, and its safety driver."

 

[Red Storm]

Not gonna lie, when I first heard it was an Uber car my immediate thought was "Of course it would be Uber." Any other company I might suspect no more than a simple equipment failure, but with Uber I actively wonder if it could be something more such as cost cutting measures leading to this accident or straight up conspiracy level stuff (Uber is trailing other companies in this industry race - hence them stealing tech from competitors - so they figure the easiest way to slow everybody down is to get someone killed on the road).

I don't actually believe that conspiracy theory, but I wouldn't put it past Uber to pull some shady shit that indirectly led to this lady's death.

 

[slayer202]

I never said a human driver would have been able to react in time. I'm saying that adequate safety measures are not in place and testing at this point should not be where the general population could be at risk.

More to come:

http://www.foxnews.com/auto/2018/03...uldve-spotted-pedestrian-in-deadly-crash.html

Smith said the video may not show the complete picture, but "this is strongly suggestive of multiple failures of Uber and its system, its automated system, and its safety driver."

what would you suggest? at some point the testing needs to be moved to the real world, and considering there will never be perfection, you certainly can't perfect things before testing in the real world. so many of these accidents could be prevented if people simply followed "do not walk in front of a moving vehicle, even if you have the right of way". You'd think that would be common sense...Are people too lazy to pause for a moment to make sure an approaching car is going to stop?

I'm not sure what the driver should/would have done if she had seen the pedestrian. Do you assume, rightfully so, that the vehicle should and will stop? How long do you give the car to do its job?

 

[rh71]

Well the real-world accidents will drive home the fact that this is a serious problem. If Uber hits a crash-test-dummy, do you think they would go to the lengths they will have to now in order to address the problem?

 

[ImpulsE69]

Another thing is that driving on a multilane highway or interstate is not at all the same as driving on residential streets in a city, but then again, how many people or accidents have google cars caused since they've been on the road?

 

[DietDrThunder]

what would you suggest? at some point the testing needs to be moved to the real world, and considering there will never be perfection, you certainly can't perfect things before testing in the real world. so many of these accidents could be prevented if people simply followed "do not walk in front of a moving vehicle, even if you have the right of way". You'd think that would be common sense...Are people too lazy to pause for a moment to make sure an approaching car is going to stop?

I'm not sure what the driver should/would have done if she had seen the pedestrian. Do you assume, rightfully so, that the vehicle should and will stop? How long do you give the car to do its job?

Just as I had stated before. First the NTSB must come up with testing guidelines similar to those of the DOD and the FAA. Just like the DOD and FAA, there has to be acceptance testing by the NTSB prior to releasing these vehicles on the public. On tactical aircraft systems, just on the flight control and avionic software alone, we go through years of white box testing, SEAL1 certification testing (assembly level branch and decision coverage analysis to ensure there is no dead code and to cover every possible corner case imaginable), black box testing, and integration testing long before first flight. All approvals from the DOD and FAA have to be met before first flight is conducted. First flight and subsequent flights are conducted with two support aircraft. All data is analyzed and has to be approved by the DOD and FAA. The aircraft is then escorted to the test range at Eglin, Edwards , and Pax River for formal flight test. Testing is conducted in these remote areas as to not endanger the lives of the general public. Flight testing at the sites goes on for years before any branch of the military signs off on and declares it fit for initial operational capability.

Yes, we find errors in hardware, flight controls, and avionic software. When this happens changes are made and the testing essentially starts from scratch, but we have new data and corner cases which now have to be tested for prior to the next release to flight test.

X-35 was initially designed prior to 1994.
Selected to compete in 1996.
2001 contract award.
December of 2006 F-35 first flight.
First production aircraft 2011.
Verification and developmental test flight ongoing
Marine Corps F-35B Initial Operational Capability (IOC) December 2015.
The Air Force F-35A IOC December 2016.
The Navy version, the F-35C, will reach the IOC milestone by February 2019.

So basically 21 years before the first branch of the military declared the F-35 operational.

 

[quikah]

Another thing is that driving on a multilane highway or interstate is not at all the same as driving on residential streets in a city, but then again, how many people or accidents have google cars caused since they've been on the road?

Last one I remember was sideswiping a bus a year or 2 ago. Waymo stop publishing accident data on their website. This is interesting data for California:

https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/autonomousveh_ol316

BTW: just saw a Waymo semi-truck at the gas station a day ago. Something they have yet to program was pulling to the pump as the driver was doing all the maneuvering.

 

[slayer202]

So basically 21 years before the first branch of the military declared the F-35 operational.

I mean, aircraft, especially fighter jets, have a different standard than cars. And we're not talking about building a new vehicle from scratch, these are existing cars we are using. It's not really a fair comparison, not too far off from your Tuskegee syphilis experiment reference

 

[Ns1]

Not gonna blame the vehicle for hitting someone jaywalking across a darkened street when humans routinely run down others in broad daylight then drive away.

 

[DietDrThunder]

I mean, aircraft, especially fighter jets, have a different standard than cars. And we're not talking about building a new vehicle from scratch, these are existing cars we are using. It's not really a fair comparison, not too far off from your Tuskegee syphilis experiment reference

But it is a fair comparison between Tactical Aircraft Systems and Autonomous vehicles because these companies are building autonomous driving systems from scratch. I would say the testing on autonomous vehicles should be more stringent because they are more likely to cause a fatality. I'm not saying it should take 21 years, but there have to be NTSB guidelines similar FAA and DOD guidelines, to do acceptance testing before unleashing these systems on the general public.

But I guess if you guys don't mind code developed in the hack it, jack it, and pack it methodology, then feel free to put them to the test when you see them on the road.

 

[rh71]

Check out these stats:
https://gizmodo.com/how-likely-you-are-to-get-killed-by-a-car-depending-on-1778993900

Hitting at 35mph is a huge difference from 45mph.

 

[slayer202]

It's still not the same. If a plane fails, it will fall out of the sky. If a car fails, there's less of a chance of catastrophic disaster. Human drivers are not very good. What was the stat I read recently, there are 100 people killed a day in the USA by car accidents? One a day for autonomous at this point would be atrocious considering how few are on the road, but have there been other fatal accidents? How many vehicles are on the road and how long have they been driving?

 

[preslove]

You guys know that uber is totally unprofitable, right? As is Tesla

I'm bearish on autonomous cars bc they're being pushed by companies that are living off the promise of future profitability. And why would established car makers spend the enormous sums it would take to get them on the road, with the added liability?

 

[DietDrThunder]

It's still not the same. If a plane fails, it will fall out of the sky. If a car fails, there's less of a chance of catastrophic disaster. Human drivers are not very good. What was the stat I read recently, there are 100 people killed a day in the USA by car accidents? One a day for autonomous at this point would be atrocious considering how few are on the road, but have there been other fatal accidents? How many vehicles are on the road and how long have they been driving?

Here is where your statement is wrong. It is not the car that would fail, it is the AI autonomous system that would fail. If an auto pilot on an aircraft fails, it is still unlikely that the aircraft will collide with another aircraft. If the autonomous system in an automobile fails, it is HIGHLY likely that someone, either an occupant in the vehicle, an occupant of another vehicle, or a pedestrian nearby, will be injured or killed.

FAIL: