Speak EV - Electric Car Forums banner

301 - 314 of 314 Posts

·
Registered
Joined
·
973 Posts
I'm a bit confused actually.

I am not really up on all the 'standard' driver assistance things that are normal for cars to come with, so I am currently looking at what is on the Seat Arona I have just ordered.

So looking at the NCAP video of this at 2'44" in the following video;-

this cheap car I have ordered (£133pm lease) comes with "AEB pedestrian/city" (NCAP speak) on all models including the most basic model seems to do easily what this Uber car struggled with.

At slightly faster speeds, as in the video at 2'52", it shows the difference between driving at 35kph and 45kph, in the former it stopped for the pedestrian and in the latter reduced speed to 27kph.

If this cheap basic car can do the job of reacting to this sort of hazard, I am stumped as to how this super-fandango AI system failed.
The Uber car was doing 70 kph though. Maybe at that speed the AEB is programmed not to engage as that might pose different threats?

At least my understanding of AEB is that when going too fast, it stops trying to brake.
 

·
Registered
Joined
·
777 Posts
The autobraking was shown in the vid to work at 70kph but that was a slow moving vehicle.
Impressive that at 35kph with a car blocking the path, the car managed to stop for the pedestrian
 

·
Registered
Joined
·
1,220 Posts
Don't know if I'm a bit late with this, but I've just seen a video on YouTube that says the victim was an Apple engineer, and that he'd complained that the model X AP swerved towards the barrier every 7 times out of 10 . Now its not nice to talk ill of the dead, but WTF !!!!!
 

·
Registered
Joined
·
11,791 Posts
That crash was at the Habsheim airshow and some of the safetys were disengaged to do a high alpha go-around at 100ft... it was found to be pilot error... not autoland failure. Captain Asseline, First Officer Mazière, two Air France officials and the president of the flying club sponsoring the air show were all charged with involuntary manslaughter. All five were found guilty.

There is a theory that the automation was the primary reason for the crash but that has not be proven.
How is this different to inappropriate use of Autopilot in a Tesla?

Ultimately most aviation accidents can be traced to inappropriate control of the aircraft systems (where it is not a mechanical failure). We had a similar issue with the Canadian EH101 where they were flying as per mechanical controls, not full fly-by-wire and didn't understand the adjustments the system was making in the background to keep the thing flying despite the pilot's best efforts!
 

·
Registered
Joined
·
11,791 Posts
Don't know if I'm a bit late with this, but I've just seen a video on YouTube that says the victim was an Apple engineer, and that he'd complained that the model X AP swerved towards the barrier every 7 times out of 10 . Now its not nice to talk ill of the dead, but WTF !!!!!
Well yes - it doesn't take long to understand there are limits to the system and realise that it might get you killed if you don't keep an eye on it.

Bit early to say, but it looks like he may have taken his eyes off the road at a critical point in the highway. I think US roads are worse for this as everything looks the same.
 

·
Registered
Joined
·
7,948 Posts
How is this different to inappropriate use of Autopilot in a Tesla?
My point was that it was not a system failure. As far as anyone can tell the automation performed as expected.

I am not going to try to defend this any more. Clearly you have made up your mind over this and so I respect your opinions.

I see autopilots on aircraft as completely different to advanced cruise controls on cars - sorry :)
 

·
Madam Legurtz
Joined
·
407 Posts
Colour me surprised.

"The Tempe police report said the crash was "entirely avoidable" if the Uber operator, Rafaela Vasquez, had been watching the road while the car was operating autonomously."
 

·
Registered
Joined
·
2,517 Posts
Two points here.

First, Uber initially put out reports that the employee was watching logs being produced by the system. Someone at the company clearly lied. I guess this should be no surprise.

Second, it's clear that Voice or no Voice, it's literally impossible for a human to be fully engaged and alert for anything more than 15 or 20 minutes at a time when monitoring a system like this. It's not uncommon for them to even fall asleep. Any system that relies on longer periods of continuous attention is bound to fail. Uber had one just employee per car because it was cheap. Despite the fact that supposedly there are billions of dollars to be madethey still skimped and literally risked bystanders lives to save a few $.

And in the spirit of Monty Python a third point, Uber's autonomy is clearly nowhere near ready for prime time when a human is meant to realise there's an issue, analyse the situation, and take corrective action, within less than 5 seconds. The autonomous system didn't even attempt to slow down. If any autonomous system needs that kind of overview, the only roads it should be allowed on are virtual ones inside a computer.
 

·
I'm not crazy, the attack has begun.
Joined
·
23,685 Posts
Two points here.
...

And in the spirit of Monty Python a third point, Uber's autonomy is clearly nowhere near ready for prime time when a human is meant to realise there's an issue, analyse the situation, and take corrective action, within less than 5 seconds. The autonomous system didn't even attempt to slow down. If any autonomous system needs that kind of overview, the only roads it should be allowed on are virtual ones inside a computer.
Hurrah! The paradox I keep mentioning i finally getting through to a few people!

An 'autonomous' system that requires continuous human monitoring is useless because instead of the human 'merely' making the one decision about what they should be doing, they now have to make to decisions; 1) do they have to intervene and [still the same thing] 2) what they should be doing if they intervene!

An autonomous system that requires continuous supervision doubles the work load on the 'driver/monitor' if used correctly.

The fact is that these systems are NOT intended to be used properly, it is merely a get-out to say 'oh, you must pay attention all the time' knowing people won't.
 

·
I'm not crazy, the attack has begun.
Joined
·
23,685 Posts

Worst bit of it, really ....
"Tragically, Volvo’s own City Safety automatic braking system had been disabled because its radars could have interfered with Uber’s self-driving sensors."

But the whole thing is just an utter tragedy and, if I am reading it right, doesn't sound like it went anywhere near an automotive-industry type safety review. Sounds like some software developers bummed it together.
 

·
Registered
Joined
·
2,709 Posts
Like isn't really the correct sentiment but you hopefully know what I mean.

Whilst mistakes happen and all that, there is clearly something very wrong in the ability to disable a series of safety systems without an appreciation of the consequences. Hindsight is wonderful but leaving the Safety braking system in place might have been a fail to safe mode rather than the tragic outcome here.

Reading elements of the report however there seems to be a lot of talk of pushing, but also about doing the right thing in terms of reporting dangerous behaviours, at some point that all went out the window with what appears to be little or no oversight from then. Very difficult to report your colleague who you are riding with for dangerous behaviours when they move to single driver/monitor model, wonder if he was streaming on a personal device or on a company one.
 
301 - 314 of 314 Posts
Top