First self driving car fatality

Dr. Courtney
Education Advisor
Insights Author
Gold Member
2018 Award
3,003
1,947

Answers and Replies

Borek
Mentor
28,127
2,634
Considering the small experimental installed base
Do you know how large the base is and how it does compare to the deaths/mile for cars driven by humans?
 
Spinnor
Gold Member
2,094
307
"A self-driving vehicle made by Uber has struck and killed a pedestrian. It’s the first such incident and will certainly be scrutinized like no other autonomous vehicle interaction in the past. But on the face of it it’s hard to understand how, short of a total system failure, this could happen, when the entire car has essentially been designed around preventing exactly this situation from occurring."

From, https://techcrunch.com/2018/03/19/heres-how-ubers-self-driving-cars-are-supposed-to-detect-pedestrians/

From, https://news.google.com/news/story/dWZXEd51GqIuoLMy_mmpiTM-SCwsM?ned=us&hl=en&gl=US

That sucks.
 
Choppy
Science Advisor
Education Advisor
Insights Author
4,510
1,603
One of the potential bright sides to this tragedy, as opposed to most human-caused motor vehicle fatalities, is that you can have a team of engineers review the incident and data collected about it in detail with a much higher likelihood of determining the principle causes and then get to work on strategies for avoiding similar situations in the future. That will make not just this individual autonomous vehicle better, but potentially *ALL* autonomous vehicles better. In that way something useful can come out of the tragedy.

Something else to remember is that not all motor vehicle accidents are completely unavoidable. If a car is traveling along at the speed limit and someone jumps out in front of it from a concealed position, its options are limited by physics and in some cases ethical decisions.
 
Vanadium 50
Staff Emeritus
Science Advisor
Education Advisor
23,434
5,863
There are 93 vehicles (as of October). Assuming 8 hours a day at 30 mph for the whole year, that's about 8 million miles, or 12.5 deaths per 100 million miles. The rate of pedestrians and cyclists killed by living drivers is 1.25 per 100 million. So there's a factor of 10 different (at least).
 
HAYAO
Gold Member
306
138
My condolences to the family who lost one of their important members.

I am very concerned about how it happened. Was it a bug? Bad programming? Fundamental flaw in the technology? Or simply an unavoidable accident (for example the pedestrian tripped over something and fell on the road)? If it is the first two, then we have a simple fix. However, if it is some fundamental flaw, then we need to use different kind of technology for this. If the last one, then it just cannot be helped unfortunately. I will not be quick to judge that a technology is good or bad simply based on this incident because no information as to how it happened was provided.

I am also concerned about the legal aspect of this. How do laws view this incident?
 
BillTre
Science Advisor
Gold Member
2018 Award
1,199
2,156
I could have been some kind of pedestrian error.
Today I was driving down a residential road, turned the corner and was faced the a guy riding no-hands on a bike, holding his phone in two hands and doing something with it while he was looking down and not where he was going.
Sometimes people do stupid things.
 
Bystander
Science Advisor
Homework Helper
Gold Member
5,121
1,138
russ_watters
Mentor
19,016
5,168
One of the potential bright sides to this tragedy, as opposed to most human-caused motor vehicle fatalities, is that you can have a team of engineers review the incident and data collected about it in detail with a much higher likelihood of determining the principle causes and then get to work on strategies for avoiding similar situations in the future. That will make not just this individual autonomous vehicle better, but potentially *ALL* autonomous vehicles better. In that way something useful can come out of the tragedy.
This is why American airliners don't crash anymore: every crash makes the next one less likely.
Something else to remember is that not all motor vehicle accidents are completely unavoidable. If a car is traveling along at the speed limit and someone jumps out in front of it from a concealed position, its options are limited by physics and in some cases ethical decisions.
Yes - at the outset we know that there is fault to be found in the victim here; she was not in a crosswalk. That doesn't necessarily make the accident unavoidable, but it at least makes it less avoidable.
 
russ_watters
Mentor
19,016
5,168
Considering the small experimental installed base, this gives me serious concerns with the technology. Not ready for prime time.
We'll see what caused this, but either way I have a fundamental moral objection to beta testing technology that can kill you on the public - including people who didn't choose to be part of the test. This would never be allowed with airplanes.
 
3,377
942
According to news, "jaywalking."
Yes apparently the killed person crossed the road away from a designated crossing. (though not very far away).
I am sure a human driver would recognize that kind of situation easily.
It might annoy them considerably. but they definitely would stop
 
StoneTemplePython
Science Advisor
Gold Member
1,128
535
This is why American airliners don't crash anymore: every crash makes the next one less likely.
Yes, though in terms of incentives, it's worth pointing out that if there's a crash, pilots go down with the plane. The issue of people 'in charge' externalizing losses completely on the general public, is an issue, which I think ties into...

We'll see what caused this, but either way I have a fundamental moral objection to beta testing technology that can kill you on the public - including people who didn't choose to be part of the test.
- - - -
In general I'm a fan of self-driving cars, but it's perilous, delicate road ahead. There are a lot of concerns to be had, including these.
 
156
63
In particular, if the accident is found to be avoidable, who goes to jail for involuntary manslaughter? This leads back to the whole "corporations are the same as people" garbage.
 
gleem
Science Advisor
Education Advisor
1,504
889
There are 93 vehicles (as of October). Assuming 8 hours a day at 30 mph for the whole year, that's about 8 million miles, or 12.5 deaths per 100 million miles. The rate of pedestrians and cyclists killed by living drivers is 1.25 per 100 million. So there's a factor of 10 different (at least).
I don't think this data can say much about the probable rate of accidents from autonomous cars vis-a-vis driven cars since the projected uncertainly in the autonomous car accident rate does not exclude the driven car rate at this point.

As far as determining the cause one would think that considering these cars are experimental there should be equipment on board like dashcams and black box recorders monitoring everything from speed to accelerations to computer functions. Maybe they do.
 
TeethWhitener
Science Advisor
Gold Member
1,527
834
berkeman
Mentor
56,063
6,121
From Bloomberg:
https://www.bloomberg.com/news/articles/2018-03-20/video-shows-woman-stepped-suddenly-in-front-of-self-driving-uber
The victim was walking her bike by the center median at around 10pm at night and stepped suddenly out into the roadway. The car had a backup driver but was in autonomous mode. The backup driver didn't see the victim either. The car was going about 3 mph above the speed limit.
Great find, thanks @TeethWhitener -- finally it starts to make sense. Wonder if it was a suicide...
 
Borek
Mentor
28,127
2,634
Seeing comments in my internet bubble I am starting to think future for autonomous cars looks bleak. Each casualty will be reported by media and blown out of proportion for the sake of hype/viewership/ratings/clickability/whatever and in time it will build a skewed opinion about how these cars are inherently unsafe, no matter what the numbers will say. Similar social mechanism that fuels anti-vaccine movements because of rare vaccine injuries.
 
russ_watters
Mentor
19,016
5,168
Seeing comments in my internet bubble I am starting to think future for autonomous cars looks bleak. Each casualty will be reported by media and blown out of proportion for the sake of hype/viewership/ratings/clickability/whatever and in time it will build a skewed opinion about how these cars are inherently unsafe, no matter what the numbers will say. Similar social mechanism that fuels anti-vaccine movements because of rare vaccine injuries.
I believe you are correct, and again, planes provide the analogue: for perception reasons, right or wrong, planes need to be MUCH safer than cars or people will be afraid to fly them.

And from a legal perspective there may actually be more culpability in a self driving car than a human driven car due to the fact that humans are known/allowed to be flawed, but machines are required to be perfect.
 
russ_watters
Mentor
19,016
5,168
In particular, if the accident is found to be avoidable, who goes to jail for involuntary manslaughter? This leads back to the whole "corporations are the same as people" garbage.
Along that vein, the AP is reporting that the driver ("operator") was a convicted felon. It is against the law for convicted felons to be rideshare drivers.

https://www.google.com/amp/www.foxnews.com/us/2018/03/20/operator-self-driving-uber-vehicle-that-killed-arizona-pedestrian-was-felon-report-says.amp.html

Since reporters don't have thinking caps, despite using the right terminology, the report does not pick up on the fact that he wasn't a 3rd party "driver", but an employed "operator" and may not be covered by that law.

In either case, it is too rare but corporate heads do sometimes go to jail for the crrimes of the company. What makes it difficult is pinpointing which person casud the crime. In the Enron scandal, many were pinpointed and went to jail and in addition the accounting firm was given a corporate "death sentence" and killed-off by government mandate (Enron died on its own).
 
1,345
332
Approx 15 pedestrian fatalities per day, short of inventing mind reading to let the car know when someone decides to step into traffic ( I had to drive down a strip mall at walking speed the other day, as the pedestrian kept turning to step out, see me step back, but would not look at me or waive, I was just guessing - 10 steps, turn, turn back, 10 steps, turn.... arg) . This situation is practically unavoidable, and not all auto-accidents are the fault of the driver. I am still convinced Self-driving is far safer, and will become more safe over time as weakness can be addressed and rolled out to the whole "population" as an update. You can not re-program human drivers.

And the whole "stand-by" emergency driver piece is just a warm fuzzy, there is very little chance that a human, who is not engaged in the act of driving, can instantly step in and make a proper timely decision, they will not have complete situational awareness, IMO potentially making the problem worse.
 
Borg
Science Advisor
Gold Member
1,845
2,176
People can be really careless around cars. They frequently think that they can be seen when they can't, don't look before entering the road, or just expect that the driver will have to avoid them like it's some kind of game. I've had to avoid each of these more than I can count.

I came very close to a situation like the Uber accident once but fortunately it was daylight and I saw that the person was clueless. I was driving southbound on this stretch of road where the speed limit is 55 mph (traffic is usually going faster than that). A girl was walking on the edge of the road with her back to traffic where the red arrow is. I saw that as a very dangerous situation for her and kept focused on her as I approached. Just before I got to her position, she abruptly turned right and stepped into my lane without looking. She barely stepped back as I swerved away from her. If I had been paying as little attention as she had, she would have been killed for sure.

NearAccident.jpg
 

Attachments

TeethWhitener
Science Advisor
Gold Member
1,527
834
People can be really careless around cars. They frequently think that they can be seen when they can't, don't look before entering the road, or just expect that the driver will have to avoid them like it's some kind of game. I've had to avoid each of these more than I can count.

I came very close to a situation like the Uber accident once but fortunately it was daylight and I saw that the person was clueless. I was driving southbound on this stretch of road where the speed limit is 55 mph (traffic is usually going faster than that). A girl was walking on the edge of the road with her back to traffic where the red arrow is. I saw that as a very dangerous situation for her and kept focused on her as I approached. Just before I got to her position, she abruptly turned right and stepped into my lane without looking. She barely stepped back as I swerved away from her. If I had been paying as little attention as she had, she would have been killed for sure.

View attachment 222387
Hey, I know that road! NOVA in the house!
 
715
158
From Bloomberg:
https://www.bloomberg.com/news/articles/2018-03-20/video-shows-woman-stepped-suddenly-in-front-of-self-driving-uber
The victim was walking her bike by the center median at around 10pm at night and stepped suddenly out into the roadway. The car had a backup driver but was in autonomous mode. The backup driver didn't see the victim either. The car was going about 3 mph above the speed limit.
How can any backup driver help in these instances? A driver has to be fully engaged with the environment, I find it difficult to believe that a backup driver can remain concentrated and respond to emergency situations in time. We do research on micro-sleep here and even driving drivers fall asleep for short spells.

Cheers
 
gleem
Science Advisor
Education Advisor
1,504
889
And of course the ubiquitous smart phone will be assisting in the distraction.
 
1,345
332
"of course the ubiquitous smart phone will be assisting" While I am confident this is statistically valid in pedestrian fatalities - no report of this in this case. Personally I will fight tooth and nail to call out this speculation, as I have said, AV is (IMO) far safer than human drivers.

This is NOT to say I want to take away driving as an activity, if you want to drive - go ahead ( a fully engaged driver may be as good as an AV. But I think an AV oversight of a human driver, is much different then an human driver oversight of an AV one.

Humans - next to a hazard ( like traffic) - will always find a way to get hurt.
 

Related Threads for: First self driving car fatality

Replies
121
Views
7K
  • Last Post
Replies
3
Views
2K
  • Last Post
Replies
10
Views
2K
  • Last Post
Replies
13
Views
2K
  • Last Post
Replies
6
Views
967
  • Last Post
Replies
21
Views
6K
  • Last Post
Replies
8
Views
2K
Replies
19
Views
742
Top