Arizona Shuts Down Uber Self-Driving Cars

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,817
Uber and its self driving cars have been in the news a lot recently, and not for good reasons. After one of Uber's self-driving cars killed a woman crossing a dark street at night, and the footage of the accident came to light, it did not paint a very good picture of Uber's technology. Even Waymo came out and publicly stated that its technology would not have allowed the fatality to happen, and others are questioning it as well. It seems as though Arizona Governor Doug Ducey might be leaning that way as well, as he has now pulled Uber's ability to test those cars in his state. The accident video may be NSFW. Thanks cageymaru.

Check out the video.
Gov. Doug Ducey on Monday sent a letter to the CEO of Uber saying he was suspending the ride-sharing company's tests of self-driving cars on Arizona roads after a fatal accident March 18 in Tempe.
 
Was this person in a crosswalk or did they just expect the oncoming car to see them crossing and stop...AT NIGHT. Idk about you guys, but at night I tend to err on the side of caution as to whether or not a vehicle has seen me yet. That being said....this shouldn't have happened with all of that tech on board.


As a side note, what will the world look like when people know cars will just stop if they walk in front of them with impunity.....chaos.
 
Was this person in a crosswalk or did they just expect the oncoming car to see them crossing and stop...AT NIGHT. Idk about you guys, but at night I tend to err on the side of caution as to whether or not a vehicle has seen me yet. That being said....this shouldn't have happened with all of that tech on board.


As a side note, what will the world look like when people know cars will just stop if they walk in front of them with impunity.....chaos.

hence why drivers should be doubly cautious when driving at night , from drunk drivers driving badly, to reckless daredevils on bikes in the dark
 
Couple points:

1) The video is way darker than the actual road is at that time of night. People have posted videos of that stretch of road from their own dashcams and it appears a LOT better lit than the video from the Uber car. That video is potentially just a really bad quality video.

2) I see lots of people crossing the road without looking out for cars. It is disturbing how people don't seem to give a shit about their own life when crossing a road. A guy crossed an intersection in front of me a couple weeks ago with his head down, hood up, and headphones on. He couldn't see cars coming from either direction very well but he crossed the road anyway. People had to brake hard to avoid him. I honestly thought he might have had developmental problems given how weird an approach he took.

3) The manufacturer of the sensors that Uber uses say that the sensors should have seen the woman and other self-driving developers say that their tech would have seen her. I can only imagine that this points to some flaws in Uber's software development.

4) The "driver" or "tester" wasn't paying attention to the road. Looking at your lap (either a readout from the car or your smartphone) while you're supposed to be watching for dangers is not exactly normal for this sort of operation. As many have noted, Uber switched to one person in the car while Waymo and others have stuck with two drivers until the software and hardware is extremely mature.

All things considered .... no, she shouldn't have been crossing the way she did but the technology should have seen her anyway and the woman in the driver's seat should have been paying attention. Even if they're not legally at fault, Uber deserves a black mark against them on this one.
 
I posted this link in another thread but that released footage seemed heavily doctored or the camera system on the vehicle is equivilant to a flip phone's camera from the early 2000s:

https://arstechnica.com/cars/2018/0...victim-came-from-the-shadows-dont-believe-it/

In the Uber car video you can clearly see the street lights are on yet somehow the illumination on the actual road is nowhere to be seen. In videos that normal civilians have taken of the same area, the place is very well lit, and there's no way that an attentive driver would have not seen the person crossing much less a self-driving car worth a damn.
 
I'm hoping the court/government makes Uber release all the information from all the cameras and sensors on the car leading up to the crash (assuming the car records this information).

At least then we can have experts weigh in on what went wrong and everybody can learn something.
 
3) The manufacturer of the sensors that Uber uses say that the sensors should have seen the woman and other self-driving developers say that their tech would have seen her. I can only imagine that this points to some flaws in Uber's software development.
Let's say a perfectly clean sensor has 100% efficiency. As the car is driven road dust and dirt is going collect on the car and sensor degrading the sensors efficiency to an increasing degree. How much degradation is acceptable? How often will sensor cleaning/maintenance be required? Changing weather conditions can decrease time between required maintenance. Is the sensor technology at a level such that we can trust self-driving cars?

In the push to develop self-driving cars NOW I feel as though many important questions are not being asked and therefore are not answered. How much money is being thrown at self-driving car development? Hundreds of millions of dollars? Billions? Hundreds of billions? In order to maintain the flow of money into somebodies pockets is the public being feed an overly optimistic narrative? Is this old ladies death the actual reality of self-driving cars in 2018 (nowhere near prime-time)?
 
Uber will probably skate on criminal liability since the woman was not in a crosswalk, but you better believe that a wrongful death lawsuit is being drafted as I type this. Named in the suit will be Uber, the state of Arizona, the city of Tempe, and maybe the manufacturer of the sensors. And it will be successful because the public roads are no place to beta test this kind of system.
 
In the push to develop self-driving cars NOW I feel as though many important questions are not being asked and therefore are not answered. How much money is being thrown at self-driving car development? Hundreds of millions of dollars? Billions? Hundreds of billions? In order to maintain the flow of money into somebodies pockets is the public being feed an overly optimistic narrative? Is this old ladies death the actual reality of self-driving cars in 2018 (nowhere near prime-time)?

WT everlovin F does this mean!? Yes, they're testing autonomous vehicles. Yes, this one fecked up and killed a woman. Overly optimistic narrative? Christ man. Every company on the damn planet feeds us this every day. It's all about spin now. Can you out-spin the true negatives and shed light on the positive? Uber's response- sure, while our technology failed this time and tragically killed an innocent pedestrian, we learned a great deal from the experience. We've settled with the family and another person won't be contributing to the carbon dioxide global warming problem. Smiles for everyone!
 
Let's say a perfectly clean sensor has 100% efficiency. As the car is driven road dust and dirt is going collect on the car and sensor degrading the sensors efficiency to an increasing degree. How much degradation is acceptable? How often will sensor cleaning/maintenance be required? Changing weather conditions can decrease time between required maintenance. Is the sensor technology at a level such that we can trust self-driving cars?

In the push to develop self-driving cars NOW I feel as though many important questions are not being asked and therefore are not answered. How much money is being thrown at self-driving car development? Hundreds of millions of dollars? Billions? Hundreds of billions? In order to maintain the flow of money into somebodies pockets is the public being feed an overly optimistic narrative? Is this old ladies death the actual reality of self-driving cars in 2018 (nowhere near prime-time)?

The software shouldn't drive the car if any of the sensors are blocked. Should be easy to make software that can tell if the sensors are not working.
Can't wait till they find out why the software didn't see the person crossing. This should of been a really easy to see object by the self driving software.
 
The software shouldn't drive the car if any of the sensors are blocked. Should be easy to make software that can tell if the sensors are not working.
Can't wait till they find out why the software didn't see the person crossing. This should of been a really easy to see object by the self driving software.
My point is that sensors are not just "working" or "not working". There are degrees of degradation. Was this actually a software failure? What happens if a sensor is working and suddenly gets splashed with muddy water. Does the car stop? Does it pull over? Does it keep going? I am by the way assuming there are redundant sensors. If one only plans for the best without adequately planning for the worse than old ladies die. If enough old ladies die new technology dies as well until the public forgets and newer technologies are developed.
WT everlovin F does this mean!? Yes, they're testing autonomous vehicles. Yes, this one fecked up and killed a woman. Overly optimistic narrative? Christ man. Every company on the damn planet feeds us this every day. It's all about spin now. Can you out-spin the true negatives and shed light on the positive? Uber's response- sure, while our technology failed this time and tragically killed an innocent pedestrian, we learned a great deal from the experience. We've settled with the family and another person won't be contributing to the carbon dioxide global warming problem. Smiles for everyone!
What does this mean? You said: "Can you out-spin the true negatives and shed light on the positive?" That is exactly what I mean by "over-optimistic narrative". It seems you are in support of continuing to advance (at this time) a primitive and immature technology because you want tomorrow's technology today. A couple more dead old ladies and the government will shut down self-driving car testing. Is that what you want? A couple months ago the editors for Car and Driver Magazine predicted affordable self-driving cars are 30 years away. I think that is about right...
 
What does this mean? You said: "Can you out-spin the true negatives and shed light on the positive?" That is exactly what I mean by "over-optimistic narrative". It seems you are in support of continuing to advance (at this time) a primitive and immature technology because you want tomorrow's technology today. A couple more dead old ladies and the government will shut down self-driving car testing. Is that what you want? A couple months ago the editors for Car and Driver Magazine predicted affordable self-driving cars are 30 years away. I think that is about right...

I'm in full support of the tech and its advancement. Is it ready for prime time? No. It's not. But you make it sound as if the companies are lying to you. Of course they'll tell you its perfect. Welcome to Earth.
 
looks like natural selection to me, crossing a road with no reflectors on bike, wearing a black jacket and not even looking at to whats coming, even if the driver had seen her prolly wouldnt have stopped in time anyway
 
Couple points:

1) The video is way darker than the actual road is at that time of night. People have posted videos of that stretch of road from their own dashcams and it appears a LOT better lit than the video from the Uber car. That video is potentially just a really bad quality video.

2) I see lots of people crossing the road without looking out for cars. It is disturbing how people don't seem to give a shit about their own life when crossing a road. A guy crossed an intersection in front of me a couple weeks ago with his head down, hood up, and headphones on. He couldn't see cars coming from either direction very well but he crossed the road anyway. People had to brake hard to avoid him. I honestly thought he might have had developmental problems given how weird an approach he took.

3) The manufacturer of the sensors that Uber uses say that the sensors should have seen the woman and other self-driving developers say that their tech would have seen her. I can only imagine that this points to some flaws in Uber's software development.

4) The "driver" or "tester" wasn't paying attention to the road. Looking at your lap (either a readout from the car or your smartphone) while you're supposed to be watching for dangers is not exactly normal for this sort of operation. As many have noted, Uber switched to one person in the car while Waymo and others have stuck with two drivers until the software and hardware is extremely mature.

All things considered .... no, she shouldn't have been crossing the way she did but the technology should have seen her anyway and the woman in the driver's seat should have been paying attention. Even if they're not legally at fault, Uber deserves a black mark against them on this one.


Uber has said that they usually employ two people, one to monitor the systems, and a second to record data into a laptop. They have stated that lately they have been short people and frequently only a single employee is in the car and must perform both duties. There has been no requirement in Arizona to have a "safety driver". Many people believe that they are there to take over in a situation like this, which is not true. They are there to do their thing for Uber, and to move the vehicle out of the road and deal wit it if it fails.

It would seem from these statements that the employee was entering data into the laptop per instruction and was doing their job, not distracted on a cell phone or whatever others have been assuming.
 
They should make the car accelerate when it detects jaywalkers. ;)
That might alert the driver attendant that he/she should quit looking at the phone and watch the road, averting the disaster entirely.
The vehicle should just maintain its current speed and if anything it could swerve into the jaywalker at the last second.
 
That might alert the driver attendant that he/she should quit looking at the phone and watch the road, averting the disaster entirely.
The vehicle should just maintain its current speed and if anything it could swerve into the jaywalker at the last second.


You need the car to communicate to the pedestrian that you have no intention of running them over, but that you are driving in this direction and you are not responsible should they get in the way.

That's the Uber way (y)
 
I posted this link in another thread but that released footage seemed heavily doctored or the camera system on the vehicle is equivilant to a flip phone's camera from the early 2000s:

https://arstechnica.com/cars/2018/0...victim-came-from-the-shadows-dont-believe-it/

In the Uber car video you can clearly see the street lights are on yet somehow the illumination on the actual road is nowhere to be seen. In videos that normal civilians have taken of the same area, the place is very well lit, and there's no way that an attentive driver would have not seen the person crossing much less a self-driving car worth a damn.

The uber camera has an anti-glare filter on it. Second, the video was not released by UBER, it was released by the police, which they got the footage from the car that is in THEIR possesion, not from uber. It was not doctored, unless you believe the police doctored it.
 
Self driving cars are not ready. Really feel everyone pushed this so quickly.
 
Something that occurred to me. The saying often heard is 'how many have to die before ________" and you can pretty much fill in the blank. Well with these self driving cars that number seems pretty low and I think that's great. On the other hand, how many till a proper street light, stop light, speed bump/turn circle etc. Those numbers seem to be much higher until action is taken. Kind of makes you wonder.
 
woman in the driver's seat should have been paying attention.

That was a woman?

Yes the video is dark and she was crossing the road without reflectors, but that seems like a strange argument to give. I thought these cars were super safe, more safe than real people, and used radar and infa red? Last I heard neither of those things needed daylight to function. If they don't have them then I'm sending my patent in fast!
 
looks like natural selection to me, crossing a road with no reflectors on bike, wearing a black jacket and not even looking at to whats coming, even if the driver had seen her prolly wouldnt have stopped in time anyway

She for sure darwined out.
 
Uber, the shit just hit the fan.

Arizona halts Uber self-driving tests; supplier says Uber disabled Volvo safety system before fatality
Also on Monday, the auto-parts maker that supplied the radar and camera on the Volvo SUV that struck and killed the woman last week said Uber had disabled the standard collision-avoidance technology in the vehicle
A conclusion they made from working backwards from it wasn't their fault. "Our system wouldn't let this happen, therefore it must have been turned off." Those were not conclusions after examining any technical data or a hardware post-mortems. Made by companies that would be instantly ruined faster than Uber if fault did land on them. So yeah must be conclusive.
 
Self-Driving cars will never be "ready" as long as people keep expecting miracles from them. There will always be accidents that are NOT the fault of the self-driving cars, and these accidents do NOT represent problems with the cars, the software, the supervising drivers, or anything else.

I look forward to the day when dipshits can blow through red-lights because everyone in their self-driving cars will automatically stop on a green for them :rolleyes:

Also, what does Arizona think that they are doing by shutting down Self-Driving cars in their state? Do they think that this is going to change things, as opposed to just making self-driving development go somewhere else and help some other economy instead? Arizona isn't exactly a populous state... There are more people just in the San Francisco Bay Area than there are in all of Arizona.
 
Last edited:
Let's say a perfectly clean sensor has 100% efficiency. As the car is driven road dust and dirt is going collect on the car and sensor degrading the sensors efficiency to an increasing degree. How much degradation is acceptable? How often will sensor cleaning/maintenance be required? Changing weather conditions can decrease time between required maintenance. Is the sensor technology at a level such that we can trust self-driving cars?

In the push to develop self-driving cars NOW I feel as though many important questions are not being asked and therefore are not answered. How much money is being thrown at self-driving car development? Hundreds of millions of dollars? Billions? Hundreds of billions? In order to maintain the flow of money into somebodies pockets is the public being feed an overly optimistic narrative? Is this old ladies death the actual reality of self-driving cars in 2018 (nowhere near prime-time)?

A good system should be able to detect when its sensors are getting dirty. Typically, in any transmit-receive type sensor you can gauge when all the signal returns are getting weak. The software should detect that the returns are being obstructed and either very weak or totally obscured.
 
A good system should be able to detect when its sensors are getting dirty. Typically, in any transmit-receive type sensor you can gauge when all the signal returns are getting weak. The software should detect that the returns are being obstructed and either very weak or totally obscured.
Or iced over, or covered with condensation. crash at 1:45
 
Back
Top