First Driverless Car Death

Forums › Transport (cars, bikes and boats) › First Driverless Car Death

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12

28622 posts

Uber Geek

Trusted

Lifetime subscriber

#1984582 27-Mar-2018 22:39

bmt:

I'm sure we will find out soon enough what the true cause of the accident was. Air crash investigations don't take two weeks!

If even a small % of cars on the road these days were self driving, but lets say a majority, you would assume the death rate would go down on average. Is it possible for the overall rate to go down while your own personal risk of death (being in a self-driving car) goes up because of the programming unknowns?

It's not just about deaths though. Self driving cars are better drivers than humans - a study found adding just a couple of self-driving cars to traffic will even out traffic flows as they are not stop-start like many human drivers are.

IMO why they are better is they drive properly, obey the rules, are far more attentive, and take the correct action, and save 3/4 second reaction time. Little of which applied to the Uber car. Oh, and the driver is alert too, another fail

GoodSync

GoodSync. Easily back up and sync your files with GoodSync. Simple and secure file backup and synchronisation software will ensure that your files are never lost (affiliate link).

FireEngine

1217 posts

Uber Geek

Trusted

#1984650 28-Mar-2018 08:15

itxtme:

Debo, I believe the yeild to bikes sign indicates dont pull into the right turning lane (20m up the road)without giving way to bikes that are in the inner "cycle" lane, it is not related to someone walking across the road with their bike.

I think he was just being ironic and humorous :-)

That said, this must not be allowed to become any debate about the legality of the pedestrian/bike on the roadway, driverless cars must be able to detect and avoid this kind of incident, this cars system patently could not.

Regards FireEngine

nunz

1421 posts

Uber Geek
Inactive user

#1984655 28-Mar-2018 08:37

FireEngine:

tdgeek:

FireEngine:

I'm not talking a "silly" mistake but it is easy to envisage real-world scenarios that the developers didn't cater far...that is largely the nature of software defects per se.

Ok, if I sat down and noted as many situations that my car can face in the wild, I pick the number 100. I even pick that I will struggle to get to that number. Thats a very low number to code for. At the lowest level its item A is approaching me and will hit me. Item A may be stationary or moving. There are many other different scenarios but not that many. So we list these 100 . Expert drivers say what should be done to mitigate the accident, taking into account what the car, the tyres the conditions allow for. Its not really rocket science, unless you are a disbeliever

Detect item A

Determine what item A is, its currently movement path if any and assess its possible and likely movement paths and speeds...

Detect item B

Determine what item B is, its currently movement path if any and assess its possible and likely movement paths and speeds...

Then determine priority of A over B if required, determine if A and B will interact, does this create an increased likelihood of A or B changing speed and or path to result in correction needed...

No its not rocket science, it is however complicated. Dealing with 1 scenario at a time is straightforward but that is not driving. Driving is potentially tracking the behaviour and possible/likely future behaviour of up to 20 or so vehicles in traffic, at speed. And we did just see a car using the current state of the art technology mow down a pedestrian without hesitation despite on the face of it, it not being a very unlikely scenario...so its clearly not <easy> to code for.

One scenario that would be really tough to code for would be merging lanes ,either two into one like a zipper or on ramps when you are travelling down a highway. in the highway situation you have a car moving at speed headed in your direction from the side - either going to end up in the lane beside you, crash into you or merge into the lane in front / behind of you.

As humans we predicatively give way (assuming we are decent driver), recognise the chances of the car being out of control and side swiping us are low or slow down / speed up to get the gap right for the merge. That's a tough call to make coding wise because an out of control car heading into you from the side is almost the same as cars merging. Imagine the random breaking at every highway on ramp trying to get that right.

FireEngine

1217 posts

Uber Geek

Trusted

#1984662 28-Mar-2018 08:46

nunz:

FireEngine:

tdgeek:

FireEngine:

I'm not talking a "silly" mistake but it is easy to envisage real-world scenarios that the developers didn't cater far...that is largely the nature of software defects per se.

Ok, if I sat down and noted as many situations that my car can face in the wild, I pick the number 100. I even pick that I will struggle to get to that number. Thats a very low number to code for. At the lowest level its item A is approaching me and will hit me. Item A may be stationary or moving. There are many other different scenarios but not that many. So we list these 100 . Expert drivers say what should be done to mitigate the accident, taking into account what the car, the tyres the conditions allow for. Its not really rocket science, unless you are a disbeliever

Detect item A

Determine what item A is, its currently movement path if any and assess its possible and likely movement paths and speeds...

Detect item B

Determine what item B is, its currently movement path if any and assess its possible and likely movement paths and speeds...

Then determine priority of A over B if required, determine if A and B will interact, does this create an increased likelihood of A or B changing speed and or path to result in correction needed...

No its not rocket science, it is however complicated. Dealing with 1 scenario at a time is straightforward but that is not driving. Driving is potentially tracking the behaviour and possible/likely future behaviour of up to 20 or so vehicles in traffic, at speed. And we did just see a car using the current state of the art technology mow down a pedestrian without hesitation despite on the face of it, it not being a very unlikely scenario...so its clearly not <easy> to code for.

One scenario that would be really tough to code for would be merging lanes ,either two into one like a zipper or on ramps when you are travelling down a highway. in the highway situation you have a car moving at speed headed in your direction from the side - either going to end up in the lane beside you, crash into you or merge into the lane in front / behind of you.

As humans we predicatively give way (assuming we are decent driver), recognise the chances of the car being out of control and side swiping us are low or slow down / speed up to get the gap right for the merge. That's a tough call to make coding wise because an out of control car heading into you from the side is almost the same as cars merging. Imagine the random breaking at every highway on ramp trying to get that right.

Yep - I suspect tdgeek's "100" scenario's quickly become near-infinite

Interesting that collision-avoidance (TCAS) is quite common in airliners now but it focuses on detection (which it is real good at given a very limited number of threats at any one time), possibly some advice coordinated with the TCAS system in the other plane (so one gets advice to climb, one to dive) - then hands over decisions and action to the pilots...

Regards FireEngine

FireEngine

1217 posts

Uber Geek

Trusted

#1984702 28-Mar-2018 09:11

In fact that sign begs a new question - are these Driverless cars going to need to interpret every roadside sign directly (ie read them), speed limits are already tagged in many GPS navigation systems from the location directly, not the signage - but what about local/temporary/minor signage like that "Yield to Bikes"???

Regards FireEngine

tdgeek

28622 posts

Uber Geek

Trusted

Lifetime subscriber

#1984703 28-Mar-2018 09:16

FireEngine:

Yep - I suspect tdgeek's "100" scenario's quickly become near-infinite

Interesting that collision-avoidance (TCAS) is quite common in airliners now but it focuses on detection (which it is real good at given a very limited number of threats at any one time), possibly some advice coordinated with the TCAS system in the other plane (so one gets advice to climb, one to dive) - then hands over decisions and action to the pilots...

Yes. merging is another, well inside my 100 limit. How could the scenarios be infinite?? Th3e car knows where the road is, where the merge is, the approach to the merge, detecting the competing cars, collision avoidance

tdgeek

28622 posts

Uber Geek

Trusted

Lifetime subscriber

#1984713 28-Mar-2018 09:19

FireEngine:

In fact that sign begs a new question - are these Driverless cars going to need to interpret every roadside sign directly (ie read them), speed limits are already tagged in many GPS navigation systems from the location directly, not the signage - but what about local/temporary/minor signage like that "Yield to Bikes"???

Down the track maybe the signs will be chipped or barcoded, so they can be read. Or GPS is forced to include all signs, with a quick update process

kryptonjohn

2523 posts

Uber Geek

Lifetime subscriber

#1984716 28-Mar-2018 09:24

I expect they are better at merging than most humans who are absolutely hopeless at it.

FireEngine

1217 posts

Uber Geek

Trusted

#1984721 28-Mar-2018 09:30

tdgeek:

How could the scenarios be infinite??

Because I think your "merge" is actually just a huge category of scenarios that would need to be allowed for:

Different priorities at the merge

Different amounts of time to merge

Different traffic levels/speeds/competing car behaviour

Easy to envisage driverless car gridlock due to unforeseen scenarios causing everything to grind to a halt as the cars don't know what to do - especially in Google vs Uber car behaviour LOL

Regards FireEngine

tdgeek

28622 posts

Uber Geek

Trusted

Lifetime subscriber

#1984726 28-Mar-2018 09:32

FireEngine:

tdgeek:

How could the scenarios be infinite??

Because I think your "merge" is actually just a huge category of scenarios that would need to be allowed for:

Different priorities at the merge

Different amounts of time to merge

Different traffic levels/speeds/competing car behaviour

Easy to envisage driverless car gridlock due to unforeseen scenarios causing everything to grind to a halt as the cars don't know what to do - especially in Google vs Uber car behaviour LOL

Disagree. Yes, not every merge is identical, there can be a few variants, but the principle is the same. If the scenarios are infinite then then driverless is a permanent no go, simple as that.

FireEngine

1217 posts

Uber Geek

Trusted

#1984774 28-Mar-2018 10:21

tdgeek:

Disagree. Yes, not every merge is identical, there can be a few variants, but the principle is the same. If the scenarios are infinite then then driverless is a permanent no go, simple as that.

NASA, Boeing and Airbus all still have humans because "you can't code for everything" - sure the variants come right down if ALL cars are driverless and operating to the same set of rules (not just the road laws).

I think a mixed driverless/driven scenario may well be too hard to properly code for with a high degree of confidence, we should view driverless cars as we would any other new technology introduced by commercial companies, it will not be perfect on day 1 by any means. Regulation may help but that too needs to ideally be given a chance to evolve over time rather than introduced with its own assumption that it can be perfect. New rules were introduced in Arizona on March 1st but on a 30-day compliance model so did not affect this incident.

Regards FireEngine

Linuxluver

5824 posts

Uber Geek

Trusted

Subscriber

#1984788 28-Mar-2018 10:39

nunz:

One scenario that would be really tough to code for would be merging lanes ,either two into one like a zipper or on ramps when you are travelling down a highway. in the highway situation you have a car moving at speed headed in your direction from the side - either going to end up in the lane beside you, crash into you or merge into the lane in front / behind of you.

As humans we predicatively give way (assuming we are decent driver), recognise the chances of the car being out of control and side swiping us are low or slow down / speed up to get the gap right for the merge. That's a tough call to make coding wise because an out of control car heading into you from the side is almost the same as cars merging. Imagine the random breaking at every highway on ramp trying to get that right.

When most cars are autonomous it will be easy for them as they will know and expect the behaviour or other cars.

What they can't predict is the timid driver going down an on-ramp at 40kph.....and slowing down as they approach the merge before suddenly doing a "Hail Mary" into the 100kph lane.

It's the incompetent (including suddenly ill / incapacitated) humans that will cause enormous problems for any AI.

Elderly people having a TIA and driving through the front entrance of an office tower (as I saw one day in Featherson St, Wellington), for example.....

_____________________________________________________________________

I've been on Geekzone over 16 years..... Time flies....

tdgeek

28622 posts

Uber Geek

Trusted

Lifetime subscriber

#1984791 28-Mar-2018 10:44

Linuxluver:

nunz:

One scenario that would be really tough to code for would be merging lanes ,either two into one like a zipper or on ramps when you are travelling down a highway. in the highway situation you have a car moving at speed headed in your direction from the side - either going to end up in the lane beside you, crash into you or merge into the lane in front / behind of you.

As humans we predicatively give way (assuming we are decent driver), recognise the chances of the car being out of control and side swiping us are low or slow down / speed up to get the gap right for the merge. That's a tough call to make coding wise because an out of control car heading into you from the side is almost the same as cars merging. Imagine the random breaking at every highway on ramp trying to get that right.

When most cars are autonomous it will be easy for them as they will know and expect the behaviour or other cars.

What they can't predict is the timid driver going down an on-ramp at 40kph.....and slowing down as they approach the merge before suddenly doing a "Hail Mary" into the 100kph lane.

It's the incompetent (including suddenly ill / incapacitated) humans that will cause enormous problems for any AI.

Elderly people having a TIA and driving through the front entrance of an office tower (as I saw one day in Featherson St, Wellington), for example.....

That will be easier. But even so, the humans that dont pay by the rules will still be caught by the driverless (its not AI) following its nose

The driverless will navigate, follow road rules, keep away from other obstacles or vehicles, be aware of any unexpected obstacles or vehicles and avoid those.

Linuxluver

5824 posts

Uber Geek

Trusted

Subscriber

#1984801 28-Mar-2018 10:59

networkn:

http://www.bbc.com/news/business-43459156

It was inevitable.

Of course, there will be plenty of commentaries that humans kill people in cars as well.

The difference is, people can be held accountable.

Uber has suspended it's driverless operations whilst it investigates, a logical and smart thing to do.

Interestingly, there was a human monitor in the car. The person killed was not on a ped crossing.

It was night.

I've seen the video now.....

The person was crossing the road. It's hard to say how fast they would have been moving prior to the collision.....but they appeared to running across the headlights. Possibly 2-3 seconds before they were also hurrying across opposing lanes of traffic (LHD). Ideally, the car would have detected an anomalous moving object and slowed. I did note it was doing 38mph in 45 zone (police said 35mph, but Google Maps shows a 45mph sign.....perhaps it recently changed.....and maybe because of incidents like this one - speculating).

Maybe the car did slow.....don't know.

Jaywalking at night isn't a good idea on a wide road with multiple lanes......but that's what defensive driving is about: be alert and ready for the unexpected. The car failed there...and so did the human driver.

Most of us are just one glance at the speedo away from the same situation.

_____________________________________________________________________

I've been on Geekzone over 16 years..... Time flies....

networkn

Networkn
30223 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1984802 28-Mar-2018 11:02

Linuxluver:

Jaywalking at night isn't a good idea on a wide road with multiple lanes......but that's what defensive driving is about: be alert and ready for the unexpected. The car failed there...and so did the human driver.

Most of us are just one glance at the speedo away from the same situation.

Sure, would a human in full control of the car have stopped, it would depend, could the automated car of stopped allowing for physics, it would seem on the surface it should have.

Who gets held responsible for the death? In some regards it's easy to be flippant and say, to make a good omelette you need to break a few eggs, however, if that was my relative, I'd want someone held accountable.

I am still moderately unsure on driverless completely automated driving technology.

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12