Author Topic: EEVblog #1066 - Uber Autonomous Car Fatality - How?  (Read 10811 times)

Fredderic and 1 Guest are viewing this topic.

Offline Decoman

  • Regular Contributor
  • *
  • Posts: 152
  • Country: no
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #275 on: March 29, 2018, 03:39:01 am »
Somehow, I suspect that autonomous vehicles with mediocre hardware/software (whatever is current really), might be as "impractical" (as in being directly hazardous, because of poor operational security) and as dangerous as having people in boxes go ballistic through pipes just to go as fast as one can that way.
« Last Edit: March 30, 2018, 05:56:06 pm by Decoman »
 

Offline TheDane

  • Regular Contributor
  • *
  • Posts: 152
  • Country: dk
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #276 on: March 29, 2018, 04:39:42 am »
If it was hacked that would be relatively easy to tell afterwards. Doubt it's the case here but if autonomous cars are networked they will likely be susceptible to hacking, which is a big problem imho. You can build unhackable systems in theory but it seems very hard to do in practice.

You could say it was 'hacked' - literally - if removing almost all of the LIDAR's is equal to 'hacking a system into defective pieces'  :-//
https://www.reuters.com/article/us-uber-selfdriving-sensors-insight/ubers-use-of-fewer-safety-sensors-prompts-questions-after-arizona-crash-idUSKBN1H337Q
 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #277 on: March 29, 2018, 12:02:51 pm »
Maybe this is a case of someone "Muntzing" themselves to the lowest cost solution.  And going at least one step too far.
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 3457
  • Country: nl
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #278 on: March 29, 2018, 08:29:28 pm »
Everything is build down to a price. I don't necessarily see the problem with using LIDAR + stereo.

Stereo cameras in IR should be able to get a good depth picture near the car.
 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #279 on: March 30, 2018, 12:45:01 am »
Everything is build down to a price. I don't necessarily see the problem with using LIDAR + stereo.

Stereo cameras in IR should be able to get a good depth picture near the car.

The car had an adequate set of sensors installed.  But various reports indicate one or more of the sensor systems was disabled.  No comments have been made on the processors, but there is always the possibility that some processing capacity was also disabled.

I agree, everything is built to a price.  But not everyone agrees on the minimum required performance, and there are certainly mistakes made on the route to whatever performance level a particular group is trying to achieve.
 

Online Brumby

  • Supporter
  • ****
  • Posts: 6979
  • Country: au
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #280 on: March 30, 2018, 12:09:34 pm »
Everything is build down to a price.

I will challenge this statement (but not too harshly).

Certainly, when it comes to mass production, this statement is true, but we are looking at research and development at this stage - and cost per unit considerations are not as important as establishing reliable solutions.

Once those reliable solutions are found, then the cost pressures come to the fore.
 
The following users thanked this post: Tom45

Offline Stray Electron

  • Frequent Contributor
  • **
  • Posts: 353
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #281 on: March 30, 2018, 01:11:00 pm »
  That's a great point. If this can happen on what is arguably the best system available, regardless of the costs, then what happens when the bean counters are placed in control and the system is designed to be mass produced at the absolute minimum costs?   Every mass production car on the road is filled with countless examples of systems that were designed for minimum production cost at the expense of reliability, maintainability, repairability and performance. The radios, the carpet material, the seats, the ultra thin wiring, the poor quality connectors, plastic head lamp covers, cheap plastic body parts, the list goes on and on.
 

Offline SilverSolder

  • Regular Contributor
  • *
  • Posts: 65
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #282 on: March 31, 2018, 02:14:57 am »
  Every mass production car on the road is filled with countless examples of systems that were designed for minimum production cost at the expense of reliability, maintainability, repairability and performance. The radios, the carpet material, the seats, the ultra thin wiring, the poor quality connectors, plastic head lamp covers, cheap plastic body parts, the list goes on and on.

Cars (and every component in it) are built to a spec:  100,000 miles or 10 years.  Most cars beat that spec, if well taken care of you can double or even triple it. 
 

Offline emptech

  • Regular Contributor
  • *
  • Posts: 55
  • Country: us
    • Empirical Technology
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #283 on: April 01, 2018, 09:14:10 am »
There is no law against stupidity, there will always be stupid people.  Unfortunately this person paid the ultimate penalty.  If it was a normal vehicle, I'd say the driver was over-driving his/her headlights, according to what I saw in the video.  I know that when police use LIDR, shiny objects give the best signal, I suppose there was nothing reflective on the bicyclist.  I didn't say victim, not sure if that is the automobile or the bicycle.

Jim
"What this country needs is less people telling us what this country needs"
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 7963
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #284 on: April 01, 2018, 10:01:55 pm »
I wonder how much a company would have to pay somebody to take a hit from an autonomous car?

It seems a good way to put a competitor out of business.  :popcorn:


 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 657
  • Country: se
  • Hobbyist
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #285 on: May 25, 2018, 08:52:25 am »
National Transportation Safety Board (NTSB) preliminary report:
https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf

Apparently the car saw the woman and realised it had to brake. But apparently it wasn't programmed to neither brake on its own nor warn the safety driver (because the system is so bad it generates too many false positives). The safety driver wasn't looking at her phone, instead she was looking at a monitor with diagnostics information (as instructed):

Quote
According to Uber, the developmental self-driving system relies on an attentive operator to intervene if the system fails to perform appropriately during testing. In addition, the operator is responsible for monitoring diagnostic messages that appear on an interface in the center stack of the vehicle dash and tagging events of interest for subsequent review

I don't understand how Uber expect their safety drivers to stare on a diagnostics monitor and at the same time keep track of what happens on the road.
 

Offline orion242

  • Supporter
  • ****
  • Posts: 496
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #286 on: May 25, 2018, 09:09:02 am »
For facks sake, the NTSB has an invalid SSL cert on their page?
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3052
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #287 on: May 25, 2018, 09:10:09 am »
National Transportation Safety Board (NTSB) preliminary report:
https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf


Another quote which I find shocking,  the car was designed NOT to stop !  Surprised the regulators allowed this  :--
Quote
According to Uber , emergency braking maneuvers are not enabled while the vehicle is under computer control, to reduce the potential for  erratic vehicle behavior.  The vehicle operator is relied on to intervene and take action.  The system is not designed to alert the operator.
YouTube and Website Electronic Resources ------>  http://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline orion242

  • Supporter
  • ****
  • Posts: 496
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #288 on: May 25, 2018, 09:27:49 am »
More surprised if the locals don't grab their torches and pitchforks and camp out in front of the politicians that welcomed that bull shit in.  Alpha test on public roadways here!
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 4837
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #289 on: May 25, 2018, 10:37:09 am »
Most of the locals are probably excited about the concept, having very little technical understanding they have bought into the hype. I suspect most people have absolutely no idea just how challenging it is to make a sufficiently reliable fully autonomous car. I also worry that partially autonomous cars will breed complacency by making it possible to get away with not paying attention to what's happening, most of the time.
 

Online Brumby

  • Supporter
  • ****
  • Posts: 6979
  • Country: au
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #290 on: May 25, 2018, 11:53:28 am »
National Transportation Safety Board (NTSB) preliminary report:
https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf

Apparently the car saw the woman and realised it had to brake. But apparently it wasn't programmed to neither brake on its own nor warn the safety driver (because the system is so bad it generates too many false positives). The safety driver wasn't looking at her phone, instead she was looking at a monitor with diagnostics information (as instructed):

Quote
According to Uber, the developmental self-driving system relies on an attentive operator to intervene if the system fails to perform appropriately during testing. In addition, the operator is responsible for monitoring diagnostic messages that appear on an interface in the center stack of the vehicle dash and tagging events of interest for subsequent review

I don't understand how Uber expect their safety drivers to stare on a diagnostics monitor and at the same time keep track of what happens on the road.
The poor operator was set up to fail - even doing the "right" thing.

"The system is not designed to alert the operator."  MAJOR fail, IMO.  The system started detecting something and didn't notify the operator straight away, which one can understand - to a point - but it would have been nice to have some sort of "imminent danger" warning.  Even a couple of seconds might have provided enough warning to lessen the severity of the impact.
 
The following users thanked this post: apis

Offline r3bers

  • Contributor
  • Posts: 10
  • Country: ru
    • solovjov.net
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #291 on: June 23, 2018, 02:12:07 am »
 

Offline chris_leyson

  • Frequent Contributor
  • **
  • Posts: 827
  • Country: wales
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #292 on: June 23, 2018, 02:28:43 am »
Posted this under another topic, oops.
A police report released Thursday would seem to imply that the driver was looking at her mobile phone and only looked up 0.5s before the crash. Wouldn't be surprised if the driver faces charges of vehicle manslaughter. It's illegal in the UK to drive whilst using a mobile phone and it's probably illegal in a lot of other countries. Probably illegal in the state of Arizona as well, but people still do it.
https://eu.azcentral.com/story/news/local/tempe-breaking/2018/06/21/uber-self-driving-car-crash-tempe-police-elaine-herzberg/724344002/
Looking back at the dash cam footage, I initialy thought the driver must have been looking at some Uber intrumentation screen, obviously not and your cellphone provider knows exactly what you're up or down loading at any given time. Definitely vehicle manslaughter you can't blame AI.
« Last Edit: June 23, 2018, 02:49:07 am by chris_leyson »
 

Offline Rasz

  • Super Contributor
  • ***
  • Posts: 1994
  • Country: 00
    • My random blog.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #293 on: Today at 07:44:44 am »
They found main reason!
It's Hulu show... "The Voice"
https://www.cnbc.com/2018/06/22/uber-driver-streamed-the-voice-before-self-driving-car-crash.html

I will totally trust US Police over NTSB investigation findings!!1 Its not like Police in US lie all the time, and Tempe Chief of Police didnt run to the papers proclaiming driver blame 30 minutes after the accident, or released misleading $10 Dash Cam recording.

Yes, this was main reason, not the software coded to plow thru obstacles, even when it recognizes live person with a bicycle on the road.
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Online Brumby

  • Supporter
  • ****
  • Posts: 6979
  • Country: au
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #294 on: Today at 11:33:14 am »
I agree.  Focussing on the female occupant for blame is a distraction from the real issues.

A fundamental question about which I haven't seen any information is: How many hours had she been doing this kind of work?  This is critically important in understanding the human factor.

Put a someone who is a real driver into such a vehicle for the first time and for the first hour, they are going to be hovering over the controls like a nervous nellie.  After 3 hours of the vehicle driving itself like it was on a licence test and they are going to relax.  After 5 hours they are going to start to relax and after 10 hours, they are going to become rather complacent and their mind will start wandering, if it hasn't already.  Some people may take longer to get to that point ... and some may get there much quicker.

IMO, the female occupant simply cannot be held culpable because the whole vehicular environment was set up to compromise her ability to act in the same way as a driver of an ordinary vehicle ... and I would expect any half-decent lawyer to get her off if charges were brought.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf