Self-Driving Uber Car Kills Arizona Pedestrian (Tempe: 2013, felon, stats)

[Cloudy Dayz]

Originally Posted by craig11152

Only if the car could reasonable be expected to stop. Even in a crosswalk a pedestrian can't step out in front of a car going 35 mph and just 15 feet away. If that happens the driver will NOT be held responsible.

Except that their story makes no sense. If she had stepped off the median while the car was just 15 feet away, she couldn't even have run across four lanes fast enough to get hit by the Uber car, in the far right lane. We are just not getting the complete story.

[Cloudy Dayz]

Originally Posted by hikernut

I think it's crucial to know exactly where the collision occurred. Most of the articles I've seen imply that the woman stepped immediately in front of the Uber vehicle and gave it no time to react. For this to be true the Uber car would have been in the far left lane, which is quite different than the NYT diagram shows.

And also quite different than the pictures of the accident scene.

[lvmensch]

Originally Posted by Cloudy Dayz

It doesn't make any sense to me either. I want to see the dashcam video, before I'll believe it. As described by the NY Times, I don't see how a human driver could not have avoided hitting her. The closest situation I can think of was some years ago in California, I was driving a delivery truck on an expressway, in the right lane, and had a green light ahead of me. I spotted a teenage girl talking on her cellphone start walking across from the medium against the light. I could see that she would be in my lane by the time I got to the crosswalk. So I started to brake and then started blasting my horn for about 10 or 12 seconds until my truck came to a complete stop about 15 feet from her. I wouldn't even say it was a hard stop. I could have stopped even quicker if I had too. Why this car couldn't stop I have no idea. I was going the same speed or faster then this car, and I had no trouble stopped for the oblivious girl on her cell phone. If the reports are correct, I think this was a technology failure and a driver failure. I don't see how else this could have happened.

On this one we agree. I do not have a problem with the human missing her. Lots of visual clutter and a dark night and the human would have a problem.

But there was almost ideal conditions for the AV. And she apparently passed over 30 or 40 feet of pavement moving at walking speeds. Hard to see how the AV missed her.

Then again the police officer who has seen the front end videos says it was unavoidable.

I don't know that anyone has seen the LIDAR results which are the critical thing. In fact off hand I would think there was a LIDAR failure or the NY Times diagram is wrong.

[hikernut]

Originally Posted by lvmensch

Yes indeed. But the photos show the bicycle on the right side and the stopped AV in the right lane. Hopefully this will all come out relatively quickly. UBER is not going to want their testing shut down for an extended period. And I would think AZ will want some answer quickly.

Well I think that is the catch. Uber testing should be shut down if the accident happened on the right hand side of the street. There's simply no excuse for an autonomous car to cause a accident in that situation. There should have been plenty of time to notice and react to a woman pushing a bicycle across the (fairly wide) street. The only possible exception I can imagine is if there were other cars on the street that were blocking the pedestrian from view.

[hikernut]

Originally Posted by lvmensch

Then again the police officer who has seen the front end videos says it was unavoidable.

I expect his opinion is taken from a human point of view. A human driver cannot see in the dark very well, but the autonomous car should have "seen" everything just fine as there were no weather conditions that would have compromised the vehicle's capability. The officer probably doesn't understand that(?).

[lvmensch]

Originally Posted by hikernut

Well I think that is the catch. Uber testing should be shut down if the accident happened on the right hand side of the street. There's simply no excuse for an autonomous car to cause a accident in that situation. There should have been plenty of time to notice and react to a woman pushing a bicycle across the (fairly wide) street. The only possible exception I can imagine is if there were other cars on the street that were blocking the pedestrian from view.

Agreed. That is why the LIDAR record is of great importance. No way this could have happened unless another car was involved. I would think though another car would also be on the front end video. Still hard to think of a dynamic situation where the AV did not get a reasonable view of the lady.

[Cloudy Dayz]

Originally Posted by hikernut

Well I think that is the catch. Uber testing should be shut down if the accident happened on the right hand side of the street. There's simply no excuse for an autonomous car to cause a accident in that situation. There should have been plenty of time to notice and react to a woman pushing a bicycle across the (fairly wide) street. The only possible exception I can imagine is if there were other cars on the street that were blocking the pedestrian from view.

That's a good point about other cars possibly blocking the view. But you would expect, they would mention that if that was the case. Even if other cars were blocking the view, the next question would be, where they stopped for the woman to cross? If the traffic in the other lanes stopped for her, then the Uber car should have at very least been slowing down. That would be reckless to shoot by stopped traffic at 45 mph.

[craig11152]

Originally Posted by Tiffer E38

Over 10 pedestrians are killed every single day by human driven cars and no threads are started on that. Per million miles traveled, semi autonomous and autonomous cars are so far vastly safer than humans at driving, for all those reasons mentioned.

That is simply not true at all. Human driven cars cause about one pedestrian fatality per 688 million miles. Those numbers come from the NTSB. The sum total of all driver-less miles probably isn't 200 million although those numbers are not well kept. Most of that is in controlled conditions or limited access roads where hitting a pedestrian is nearly impossible. So even if a third of those miles, say 67 million was on public roads with pedestrians "available" then autonomous vehicles are 10 times more likely to kill a pedestrian than a human driven car.

The last part is conjecture on my part because there is no clear data yet on how many driver-less miles in real time traffic has been logged. But it isn't remotely close to 688 million miles which is how often a human driven car kills a pedestrian.

[craig11152]

I just read a Forbes story that suggests Waymo cars (Google) are north of 5 million miles on public roads. Uber is at 2 million miles according to Forbes.
American drivers go about 3.1 trillion miles a year. When you break that down we drive 7 million miles in a little over a minute.

As I said above human driven cars kill a pedestrian every 688 million miles. It appears a computer driven car kills a pedestrian every 7 million miles.

[hikernut]

Originally Posted by craig11152

I just read a Forbes story that suggests Waymo cars (Google) are north of 5 million miles on public roads. Uber is at 2 million miles according to Forbes.
American drivers go about 3.1 trillion miles a year. When you break that down we drive 7 million miles in a little over a minute.

As I said above human driven cars kill a pedestrian every 688 million miles. It appears a computer driven car kills a pedestrian every 7 million miles.

Bottom line is the number of miles driven to date in purely autonomous mode is way too small to make any sort of significant statistical conclusion. We have one fatality to study, and IMO it's the details of this accident that are important today.