Pedestrian killed by self driving car.

Articles of interest to the AD community, currently in the news.
User avatar
Strangely Brown
Posts: 1018
Joined: Sun Sep 27, 2015 8:06 pm
Location: Sussex

Re: Pedestrian killed by self driving car.

Postby Strangely Brown » Thu Mar 22, 2018 1:25 pm

Horse wrote:Interesting that the pedestrian decided to cross 'then' (ie in front of an approaching vehicle) - I wonder whether she was somehow 'impaired'?


Interesting, but irrelevant. The car should have seen her even if the driver could not. Maybe the car /did/ see her and just failed to react.

TheInsanity1234
Posts: 544
Joined: Fri Oct 02, 2015 8:03 pm

Re: Pedestrian killed by self driving car.

Postby TheInsanity1234 » Thu Mar 22, 2018 1:41 pm

Here's an interesting article: bloomberg article

So, the car was travelling at 38 mph, but I would have hoped most drivers would have been travelling at 30 or less, owing to the relatively short throw of the headlights. But this is interesting, as it means the autonomous vehicle was speeding, as it was doing 38 in a 35 zone...

The other thing that's interesting is, the technology supposedly detected the pedestrian, but failed to react as if it was an actual pedestrian, which implies it had detected the bicycle with the pedestrian behind, but assumed it wasn't a human about to jump out in front of it.

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Pedestrian killed by self driving car.

Postby Horse » Thu Mar 22, 2018 1:54 pm

Strangely Brown wrote:
Horse wrote:Interesting that the pedestrian decided to cross 'then' (ie in front of an approaching vehicle) - I wonder whether she was somehow 'impaired'?


Interesting, but irrelevant.


Irrelevant in the same way that some in the recent thread about mobile phone use were blaming the driver of the minibus.
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
Strangely Brown
Posts: 1018
Joined: Sun Sep 27, 2015 8:06 pm
Location: Sussex

Re: Pedestrian killed by self driving car.

Postby Strangely Brown » Thu Mar 22, 2018 2:07 pm

TheInsanity1234 wrote:The other thing that's interesting is, the technology supposedly detected the pedestrian, but failed to react as if it was an actual pedestrian, which implies it had detected the bicycle with the pedestrian behind, but assumed it wasn't a human about to jump out in front of it.


So it's probably a classification failure. I am sort of relieved that the car was not using visible spectrum only but dismayed that the classification failed, especially when the "identified object" started moving in front of the car. You would have hoped that the car would have, at least, slowed for "an object" moving directly across its path. How can it not have known that it was going to hit whatever it had seen?

As for the "safety driver"... I accept that they probably wouldn't have seen the pedestrian soon enough to avoid an impact but had they been paying attention rather than "looking at their lap" they might at least have hit the brake.

So many questions...

User avatar
akirk
Posts: 1659
Joined: Sun Sep 27, 2015 6:58 pm
Location: Bristol

Re: Pedestrian killed by self driving car.

Postby akirk » Thu Mar 22, 2018 2:29 pm

dvenman wrote:
Strangely Brown wrote:The thing that is most obvious to me is the appalling use of the headlights


I wouldn't have had full beam on because of the vehicles in front. And I only have visible light-capable vision systems, so LIDAR / radar / sonar wouldn't have been much use to me either.

And she really does just pop out in front of the vehicle...


from comments circulating the internet, there was c. 60 foot between car and pedestrian at the point when she became visible...
US laws on lighting requires dipped beams which provide coverage over 60m ahead... over 3 times that distance - and easily time in which to stop from that speed...

correct lighting in use, or running on DRLs?

not that it makes any difference what lighting was in place:
- car shouldn't need it with the other sensors
- driver was clearly paying no attention anyway

Alasdair

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Pedestrian killed by self driving car.

Postby Horse » Thu Mar 22, 2018 2:37 pm

Strangely Brown wrote: ETA: Oh, and it /looks like/ the "safety driver" was more interested in his phone than supervising the car.


Or he could have been looking at a display showing what they car is doing / seeing? Whatever, he looks down for 3 -4 seconds - just too long to be looking away from the road ahead - as mentioned in another thread recently . . . viewtopic.php?f=8&t=445&p=10668&hilit=ambulance#p10668
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
Strangely Brown
Posts: 1018
Joined: Sun Sep 27, 2015 8:06 pm
Location: Sussex

Re: Pedestrian killed by self driving car.

Postby Strangely Brown » Thu Mar 22, 2018 5:09 pm

Horse wrote:
Strangely Brown wrote: ETA: Oh, and it /looks like/ the "safety driver" was more interested in his phone than supervising the car.


Or he could have been looking at a display showing what they car is doing / seeing?


It's possible I suppose, but I suspect unlikely. I would expect screens of that nature to be mounted properly within the vehicle, not something that is held on the "safety driver"'s lap. How could they be expected to "take control" if they first had to put the laptop away.

I suspect it was a phone and it is a good example of what normal drivers will when their car is in a self-driving mode. IMO, there is no way that semi-autonomous cars will ever be "safe". If the driver is required to be alert and ready to take control then they might as well drive it themselves. In practice they will likely be so disengaged from the driving task that taking control will prove impossible.

User avatar
jont-
Posts: 1522
Joined: Mon Sep 28, 2015 7:12 am
Location: Herefordshire

Re: Pedestrian killed by self driving car.

Postby jont- » Thu Mar 22, 2018 5:28 pm

Strangely Brown wrote:If the driver is required to be alert and ready to take control then they might as well drive it themselves. In practice they will likely be so disengaged from the driving task that taking control will prove impossible.

Indeed. And some manufacturers already agree and have said they will skip L3 for this reason (remember L4 is fully autonomous but geofenced).

Pyrolol
Posts: 73
Joined: Sun Oct 04, 2015 1:52 am
Location: San Francisco

Re: Pedestrian killed by self driving car.

Postby Pyrolol » Thu Mar 22, 2018 6:19 pm

Please don't read too much into the video. A human in that situation would have been able to see far better, and so should the car.

The car had lidar, which shouldn't have been affected by the darkness at all, and likely much higher quality cameras that didn't get ran through such aggressive compression.

The video's purpose seems to be make it look like it wasn't their fault. We'll see if the NTSB's investigation turns up anything.

More interesting is the safety drivers apparent complete lack of attention (they may well be made a scapegoat), and absence of braking even when the victim could be seen in the video (later than the car or safety driver should have seen them).

fungus
Posts: 439
Joined: Sun Dec 13, 2015 5:26 pm
Location: Dorset

Re: Pedestrian killed by self driving car.

Postby fungus » Thu Mar 22, 2018 7:25 pm

Strangely Brown wrote:
Horse wrote:
Strangely Brown wrote: ETA: Oh, and it /looks like/ the "safety driver" was more interested in his phone than supervising the car.


Or he could have been looking at a display showing what they car is doing / seeing?


It's possible I suppose, but I suspect unlikely. I would expect screens of that nature to be mounted properly within the vehicle, not something that is held on the "safety driver"'s lap. How could they be expected to "take control" if they first had to put the laptop away.

I suspect it was a phone and it is a good example of what normal drivers will when their car is in a self-driving mode. IMO, there is no way that semi-autonomous cars will ever be "safe". If the driver is required to be alert and ready to take control then they might as well drive it themselves. In practice they will likely be so disengaged from the driving task that taking control will prove impossible.


+1

So will you be able to use a fully autonomous car without a driving licence? :shock:

Nigel.


Return to “In The News”

Who is online

Users browsing this forum: No registered users and 18 guests