autonomous vehicle lane

AnotherJoe

forgotmyname wrote: »

It detected the object 6 seconds before the impact. Approx 360feet away.

Not seen the timings for how quickly it went from object to bicycle though, but i imagine it would have been fairly quick. The software knew the object was moving and in a collision path.


I doubt that it did or it would have taken evasive action. From what I've read it was busy trying to work out what it was and calculating an intercept doesn't seem to be part of its ability or it was dumbly trying to work out what it was rather than avoid it first worry what it was second.


Would the display the driver was looking at not have shown the object detected and then the path it was taking?

I dont think it was a "display" in the sense of a picture showing objects projected paths etc.

AIUI it was a scrolling log of events thats going on.



If you look at the video you can see the absolute look of surprise on the face of the "driver" just before impact, so it seems unlikely the computer predicted it would intersect in 6 seconds and displayed it on a screen, since it would have surely slowed down or took avoiding action in that case, and also the "driver" would surely have at least looked up earlier than milliseconds before.

forgotmyname

Yes the driver was in shock but from what i gather they looked up during the impact. So what was on the screen to monitor to closely without the need to look up?

The computer did spot the person 6 seconds before, the investigation had the evidence from the cars data recorded. It also proved the lighting was not as dark as the dashcam.

Volvo's own system detected the need to brake but Uber had deactivated the emergency braking, the accident may still have happened with the Volvo system. But not at 43mph. EEVBlog worked it out and estimated the accident would have been around 18mph?

But Ubers system detected it in plenty of time to stop.

AnotherJoe

forgotmyname wrote: »

Yes the driver was in shock but from what i gather they looked up during the impact. So what was on the screen to monitor to closely without the need to look up?


Dont know. Apparently they used to have two people in the cars, one watching the road the other the logs. But guess what its half as expensive to only have one person


The computer did spot the person 6 seconds before, the investigation had the evidence from the cars data recorded. It also proved the lighting was not as dark as the dashcam.

Volvo's own system detected the need to brake but Uber had deactivated the emergency braking, the accident may still have happened with the Volvo system. But not at 43mph. EEVBlog worked it out and estimated the accident would have been around 18mph?

But Ubers system detected it in plenty of time to stop.

It spotted something 6 seconds before the impact but kept changing its mind about what it was, and crucially while it was taking its own good time to categorise the object, it made no attempt whatsoever to slow down or to steer to avoid it. What kind of AV system is that?



Making a mockery of this one ten thousandth error rate of human drivers.



Reminds me of the Challenger enquiry, when it became clear that NASA simply made up the error rate they expected from the shuttle program. As in, literally just made it up. I expect no less from anything Uber would do and say it goes with their whole corporate history and it seems now the same is true of AV proponents as well.

Car_54

AnotherJoe wrote: »

Reminds me of the Challenger enquiry, when it became clear that NASA simply made up the error rate they expected from the shuttle program. As in, literally just made it up. I expect no less from anything Uber would do and say it goes with their whole corporate history and it seems now the same is true of AV proponents as well.

To be fair to Uber, this particular claim did not come from them, but from the UK's Institution of Engineering and Technology.


They announced this 1 in 10,000 figure in 2014, before they'd even started trials.

almillar

If you want a guide on how not to do it, I think that incident covers most of it.

This was Uber's trial, and they've cancelled it.

AnotherJoe

[quote=[Deleted User];74393964]To be fair to Uber, this particular claim did not come from them, but from the UK's Institution of Engineering and Technology.
They announced this 1 in 10,000 figure in 2014, before they'd even started trials.[/QUOTE]

Sorry I realised that but my post wasn't clear. What I meant was, it makes a mockeryof that sort of claim made by AV proponents, and numbers like that get a life of of their own and get repeated by other people as facts when, they are just plucked from thin air with no basis behind them other than wishful thinking.