Author Topic: EEVblog #1068 - Autonomous Uber Incident Update  (Read 177101 times)

0 Members and 1 Guest are viewing this topic.

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
EEVblog #1068 - Autonomous Uber Incident Update
« on: March 27, 2018, 11:32:03 pm »
An update on the autonomous self driving Uber Volvo XC-90 involved in the pedestrian fatality.
It is being reported that Uber disabled the Intel Mobileye collision avoidance sensor that is factory fitted in Volvo XC90.
Intel have ran the dashcam footage of the accident through the Mobileye system and said that even with the dark footage it would have detected the pedestrian a second before the incident.

 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #1 on: March 28, 2018, 12:14:38 am »
Those systems are supposed to be toys in comparison. If one of those can detect and prevent only 5 % of accidents that would be great. However the standard for a robotic car must be virtually 100%, certainly much better than human. Even 99 % wouldn't be good enough since hitting 1% of all jaywalking pedestrians is obviously not acceptable. Something was seriously wrong with the Uber car.

Unsurprisingly Uber have had their licence to operate selfdriving cars in Arizona suspended: http://ideas.4brad.com/arizona-bans-uber
 

Offline orion242

  • Supporter
  • ****
  • Posts: 746
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #2 on: March 28, 2018, 02:40:01 am »
Silly to see they wouldn't integrate that sensor since it was already there.

Better angle, lets see all the safety driver video.  Did Uber know their safety drivers where distracted most of the time and choose to ignore it?

Google wrote off the safety driver concept because of the obvious problem with keeping them engaged.  See Waymo safety report.  Seems common sense, why did Uber not figure this out?

Crash and burn, goodbye for Uber's AV program I suspect.
 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
 

Offline Cnoob

  • Regular Contributor
  • *
  • Posts: 132
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #5 on: March 28, 2018, 03:35:41 am »
Unlike what the Sheriff said, it's looking more like a preventable accident. From Ubers video I thought the incident happened in a less built up area.
But those mobile phone videos show otherwise. Even though  digital cameras over expose at night when set to automatic, the mobile videos show many light sources and Ubers video shouldn't been so dark.     
 

Offline Stray Electron

  • Super Contributor
  • ***
  • Posts: 2039
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #6 on: March 28, 2018, 04:31:08 am »
  IMO the Sheriff in Arizona was much too quick to say that it was NOT Uber's fault. 
 

Offline Brumby

  • Supporter
  • ****
  • Posts: 12288
  • Country: au
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #7 on: March 28, 2018, 05:27:54 am »
Not really.

The cycle-pushing pedestrian contributed the most to this accident.  They caused the hazard in the first place.
 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #8 on: March 28, 2018, 05:48:13 am »
Not really.
The cycle-pushing pedestrian contributed the most to this accident.  They caused the hazard in the first place.

The issue is here is not the actual accident and who's at fault, sadly fatal car accidents happen dozens of times a days. This is being talked about because of the unique nature of the technology involved, that is was a technology trial of essentially a prototype system, and the huge ramifications it has in the short and medium term on the implementation and adoption of autonomous cars, a subject that technology that will potentially change our way of life.
 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #9 on: March 28, 2018, 05:49:19 am »
Unlike what the Sheriff said, it's looking more like a preventable accident. From Ubers video I thought the incident happened in a less built up area.
But those mobile phone videos show otherwise. Even though  digital cameras over expose at night when set to automatic, the mobile videos show many light sources and Ubers video shouldn't been so dark.   

And it probably doesn't matter either way, because it's the LIDAR that should have picked it up and it didn't, under almost ideal practical circumstances.
 

Offline Brumby

  • Supporter
  • ****
  • Posts: 12288
  • Country: au
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #10 on: March 28, 2018, 06:01:11 am »
The issue is here is not the actual accident and who's at fault,

Oh, I quite agree.  I was simply responding to this:
  IMO the Sheriff in Arizona was much too quick to say that it was NOT Uber's fault. 

What you and I are interested in is the background as to why the tech in play did not (as it appears) act at all.

This is, by far, the most important aspect of this whole incident and is key to the refinement of AV technology.


There's going to be some name-calling and finger-pointing coming out of all this, but that's just butt-covering crap that nobody really wants to get caught up in.  In 5 years' time, this incident will be a memory - but the tech will have evolved.
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 16560
  • Country: 00
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #11 on: March 28, 2018, 08:21:02 am »
It totally failed to detect the cyclist at 6:17 in the video.



I'm sure that's just a coincidence. :popcorn:
« Last Edit: March 28, 2018, 01:46:16 pm by Fungus »
 

Offline ChrisLX200

  • Supporter
  • ****
  • Posts: 458
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #12 on: March 28, 2018, 08:21:42 am »
The facts are that the Uber system failed spectacularly when it clearly should not have. All the fluff about 'dark video/she came out of nowhere' is irrelevant.

The system was not Fit for Purpose and should not have been in use on public roads in live traffic. Someone signed off that it WAS fit for purpose and will shortly get their ass handed to them in a basket. I can think of no excuse they can offer other than hardware failure and that possibility should have been screened out before these trials started.
 

Offline sakujo7

  • Contributor
  • Posts: 39
  • Country: au
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #13 on: March 28, 2018, 09:01:25 am »
The system was not Fit for Purpose and should not have been in use on public roads in live traffic.

By this logic, human drivers are not fit for purpose and should also be banned from roads, because this sort of accident happens all the time.

Of course, Uber still needs to get their shit together because a LIDAR system should easily outperform humans in cases like this.

I really want to see the data. Unfortunately they will do their best to make sure it never gets out (other than shitty dashcam footage), as it will only demonstrate their incompetence.
 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #14 on: March 28, 2018, 09:34:35 am »
By this logic, human drivers are not fit for purpose and should also be banned from roads, because this sort of accident happens all the time.

The problem is we humans except that other humans fail sometimes, and whilst we try to make that batter through regulation and training, we essentially live with that.
But technology on which our lives depend is held to a different standard, we expect it to work and work reliably when our lives depend on it.

Quote
I really want to see the data. Unfortunately they will do their best to make sure it never gets out (other than shitty dashcam footage), as it will only demonstrate their incompetence.

They will pay any price to make this go away. Both the driver and family of the victim will be made offers they can't refuse.
 

Offline orion242

  • Supporter
  • ****
  • Posts: 746
  • Country: us
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #15 on: March 28, 2018, 11:42:14 am »
Don't think I have heard anyone talk about the lidar blind spot near the car caused by relying on a single roof mounted unit.

https://www.reuters.com/article/us-uber-selfdriving-sensors-insight/ubers-use-of-fewer-safety-sensors-prompts-questions-after-arizona-crash-idUSKBN1H337Q


CA has also announced it will not renew Uber's testing permit, they are done at the end of this month.  Another nail in their coffin.
« Last Edit: March 28, 2018, 11:50:24 am by orion242 »
 

Offline ChrisLX200

  • Supporter
  • ****
  • Posts: 458
  • Country: gb
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #16 on: March 28, 2018, 12:46:51 pm »
The system was not Fit for Purpose and should not have been in use on public roads in live traffic.

By this logic, human drivers are not fit for purpose and should also be banned from roads, because this sort of accident happens all the time.

Of course, Uber still needs to get their shit together because a LIDAR system should easily outperform humans in cases like this.

I really want to see the data. Unfortunately they will do their best to make sure it never gets out (other than shitty dashcam footage), as it will only demonstrate their incompetence.

These trials are not about testing critical safety mechanisms, they are for fine tuning a host of other things (predicting traffic flow, optimising road positioning and so on). Do you really think letting 2 tons of metal travelling at 50mph on public roads weaving through traffic and pedestrians is a good way of testing basic safety systems?
 

Offline omooswald

  • Newbie
  • Posts: 1
  • Country: de
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #17 on: March 28, 2018, 01:00:26 pm »


this video is interesting. anybody want to translate some german? it seems mobileye is tested


indeed it is, they tested different integrated systems, and Volvo and Lexus were the ones that could prevent an accident with a kid appearing from behind another car.
mobileye gets mentioned, but only for upgrade kits ... which are useless.
another interesting thing is: Mercedes had the most cameras and sensors in the car, but didn't get the best results... so more isn't always better
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 6693
  • Country: nl
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #18 on: March 28, 2018, 01:10:29 pm »
Don't think I have heard anyone talk about the lidar blind spot near the car caused by relying on a single roof mounted unit.
It's not really relevant here. Also at close range stereo works a lot better. So LIDAR at range with stereo for the blind spots makes a fair amount of sense.
 

Offline CoffsHackerSpace

  • Contributor
  • Posts: 10
  • Country: au
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #19 on: March 28, 2018, 02:03:34 pm »
i found this very fascinating
know i want to understand more about the volvo camera
thankyou for sharing
 

Offline sakujo7

  • Contributor
  • Posts: 39
  • Country: au
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #20 on: March 28, 2018, 02:07:19 pm »
These trials are not about testing critical safety mechanisms, they are for fine tuning a host of other things (predicting traffic flow, optimising road positioning and so on). Do you really think letting 2 tons of metal travelling at 50mph on public roads weaving through traffic and pedestrians is a good way of testing basic safety systems?

No. I never said anything like that.
 

Offline G7PSK

  • Super Contributor
  • ***
  • Posts: 3859
  • Country: gb
  • It is hot until proved not.
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #21 on: March 28, 2018, 02:13:05 pm »
The other videos of that bit of road show a well lit building, it is fairly certain that the building was there before and during the indecent so why is it not in the uber dash cam video. Was the camera really that bad or was the video hurriedly tampered with in order to try and shed blame. 

If the equipment was faulty why did the car not stop, surely the prerequisite of any moving machinery is that in the event of any failure it safely stops.
« Last Edit: March 28, 2018, 02:17:04 pm by G7PSK »
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #22 on: March 28, 2018, 03:04:51 pm »
It totally failed to detect the cyclist at 6:17 in the video.



I'm sure that's just a coincidence. :popcorn:
In this case it wasn't important though, since that cyclist wasn't anywhere near that car. Modern image analysis software is very impressive, but not yet good enough that you could rely on them 100% as far as I can tell (might get there soon though). But that is why companies like Waymo still rely on lidar since then you don't have to try to interpret 2d images.

For collision avoidance it doesn't matter if it's not perfect though, it's there as a backup to stop the car when the driver makes a mistake, so even if it only detects and prevents some accidents its still a win.
 

Offline Stray Electron

  • Super Contributor
  • ***
  • Posts: 2039
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #23 on: March 28, 2018, 04:51:14 pm »
In this case it wasn't important though, since that cyclist wasn't anywhere near that car.

  I completely disagree. The cyclist is much closer than some of the other pedestrians that the system did detect and highlight.  By the color of the boxes it appears that the system priorities each pedestrian but it appears to have completely overlooked the cyclist.  The only reason that he was "unimportant" was that he didn't pull in front of the automated car and get run over.  I don't know who's car this was but if someone had looked at this video sooner and realized that there was a problem with the system not recognizing cyclists then perhaps the accident in Arizona would have been prevented.  (And hind sight is always 20/20 as they say!)
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 16560
  • Country: 00
Re: EEVblog #1068 - Autonomous Uber Incident Update
« Reply #24 on: March 28, 2018, 05:18:08 pm »
It totally failed to detect the cyclist at 6:17 in the video.
In this case it wasn't important

Right, and the cyclist hit by the Uber car wasn't important either, not until the very last second...

I really can't understand your answer. How is the cyclist not worthy of a blue rectangle?

Even worse: The cyclist "blocks" the car's view of the van as he passes in front of it.


« Last Edit: March 28, 2018, 05:43:10 pm by Fungus »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf