PDA

View Full Version : Self-driving car kills pedestrian.


G-CPTN
19th Mar 2018, 23:24
An autonomous (self-driving) car travelling in autonomous mode - but with a human monitor behind the steering wheel - failed to register a pedestrian crossing the road at 10pm (not at a designating crossing place) in Tempe, Phoenix, Arizona.
The pedestrian was struck at 40mph and died from her injuries.

What was the human monitor (not) doing?

jack11111
20th Mar 2018, 00:00
The whole commercial world and tech community are so jazzed about the coming of self driving cars that nothing will stop their introduction.

This is just a slight speed bump (sorry), in that arrival.

And they will hang the victim for not being in a crosswalk.

aox
20th Mar 2018, 00:02
Oh. The news programme I saw didn't even say a human driver/monitor. It was only a brief mention in the programne though, not a full report.

Lantern10
20th Mar 2018, 01:50
Well it was only ever going to be a "matter of time"

I suppose it's a given that the monitor will be after some compensation for the trauma!

llondel
20th Mar 2018, 02:44
Just like automation in the cockpit - when you decide to trust the computer it will go wrong and bad things can happen.

Lascaille
20th Mar 2018, 03:02
Or, on the other hand they could have been dressed 100% in black complete with a raised hood, and the human monitor didn't take any action because they didn't see them either...

garyscott
20th Mar 2018, 05:19
These things are regularly jumping red lights too . . :eek:

sitigeltfel
20th Mar 2018, 05:41
The victim was walking her bicycle across the street when she was struck. Press photos show the mangled bike at the side of the road.

Blues&twos
20th Mar 2018, 05:58
Is it now known that the automation didn't detect the pedestrian? The reports I've seen so far have only mentioned the collision and the (irrelevant) 'not on a crosswalk', but not the full circumstances.

ORAC
20th Mar 2018, 06:26
Jaywalker killed by car. Happens every day. Let’s wait for the report on the circumstances.

Automation does not change the laws of physics or stopping distances. If someone steps out in the path of a vehicle both still apply - which is why speed limits are applied in built up areas even with manned vehicles.

Blues&twos
20th Mar 2018, 07:18
This was what I was wondering. Clearly if the pedestrian stepped immediately in front of the car giving it no time to stop, the collision was inevitable whether manned or automatic.
However, if the automation failed to detect an obstacle in the road, that's more concerning.

ExSp33db1rd
20th Mar 2018, 07:20
.......QED......

Pontius Navigator
20th Mar 2018, 07:23
The victim was walking her bicycle across the street when she was struck. Press photos show the mangled bike at the side of the road.
Bicycle, yes, struck, yes, at a crossing, no. Was she walking? Again, a missing piece of evidence.

jolihokistix
20th Mar 2018, 08:48
Orac, "...which is why speed limits are applied in built up areas even with manned vehicles."


Are you suggesting that the computer was breaking the speed limit? If not, then the simple fact of having a speed limit there was meaningless in this context.

CargoMatatu
20th Mar 2018, 08:50
Authorities are reported as saying that she stepped out from between two parked cars, not on a pedestrian crossing, into the immediate path of the vehicle.

Thus, the outcome was probably inevitable - in automated or manual mode.

Anilv
20th Mar 2018, 09:20
A meaningful comparison will be to compare number of incidents from a period where no automatic driving cars were in existence with a period where only automatic driving cars exists.

Anilv

vapilot2004
20th Mar 2018, 09:37
Technology is only as perfect as the programming code. We've yet to invent the infallible human.

PDR1
20th Mar 2018, 09:52
Presumably the AAA spokesman will give a statement to the effect that this is very sad, but it's a price worth paying for the right to have cars...

PDR

ORAC
20th Mar 2018, 09:53
Are you suggesting that the computer was breaking the speed limit? If not, then the simple fact of having a speed limit there was meaningless in this context.

No, I am saying that, in an area with a 40mph limit, if someone steps in front of a vehicle (see CargoMatatu) then they will be killed or badly injured whether it has a human behind the wheel or a computer. The laws of physics, not robotics, apply.

Gertrude the Wombat
20th Mar 2018, 09:55
Jaywalker killed by car. Happens every day. Let’s wait for the report on the circumstances.
What is a "jaywalker"?

TWT
20th Mar 2018, 09:59
Gertrude :

https://dictionary.cambridge.org/dictionary/english/jaywalk

ATNotts
20th Mar 2018, 10:05
Looks as though in this instance the vehicle / monitor are "guilty until proven innocent".

I doubt the pedestrian, whatever she was doing is going to have any blame apportioned to her, certainly not by the media.

jolihokistix
20th Mar 2018, 10:05
Thanks Orac.


By the way the original argument for these infernal contraptions still stands, i.e. people will still get killed on the roads, but the numbers will fall (not disappear completely) as computers take over.


Using this logic, the argument for automated cars in the US over 5 years is 100,000 deaths vs 1 so far, or two if you include the Tesla crash in Florida. Expect the numbers of road deaths by automation to increase but hopefully very slowly, even as the numbers of deaths from manual machines fall drastically. Well, that is the theory.

Gertrude the Wombat
20th Mar 2018, 10:16
Looks as though in this instance the vehicle / monitor are "guilty until proven innocent".
Well, people describing the victim as a "jaywalker" appear (according to that definition) to have decided that the victim is guilty until proven innocent.

VP959
20th Mar 2018, 10:18
I'm not convinced that you can apply the same behavioural trends to the introduction of autonomous cars as you can to other improvements in vehicle technology. Often these seem counter intuitive.

When seat belt legislation was introduced here it significantly reduced the incidence of death and injury to car drivers and passengers, but gradually increased the incidence of death and injury to pedestrians, cyclists and motorcyclists from car accidents. The theory is that of risk compensation, where people will tend to adopt a pattern of behaviour that is influenced to some degree by their appetite for taking risks. Drivers wearing seat belts started to feel safer, so they took more risks, which then led to more vulnerable road users being injured or killed.

Autonomous cars are a great idea, but the idea of mixing autonomous cars with non-autonomous cars and other road users is a pretty lousy one, in my view. It's highly probable that an autonomous car may be intrinsically safer than one with a human driver (it can't get mad and suffer from road rage, for a start), but it's also highly probable that other road users will adapt to autonomous cars by taking advantage of their enhanced safety features, so putting themselves at greater risk.

Once people know that when they see an autonomous car coming it will act in any way it can to avoid an accident I predict that they will quickly just take advantage of this by doing daft things, like walking across the road in front of them, or pulling out of junctions when one is spotted, trusting the technology in the autonomous car to keep them safe, or at least within their own risk comfort zone.

longer ron
20th Mar 2018, 10:26
Perhaps if they build an autonomous car with ultra quiet electric drive - that will be ultra safe for pedestrians :cool:

Katamarino
20th Mar 2018, 10:45
Once people know that when they see an autonomous car coming it will act in any way it can to avoid an accident I predict that they will quickly just take advantage of this by doing daft things, like walking across the road in front of them, or pulling out of junctions when one is spotted, trusting the technology in the autonomous car to keep them safe, or at least within their own risk comfort zone.

And if they get hurt or killed then it'll be their own stupid fault. Hopefully natural selection will take its course, but I fear that the autonomous cars will be good enough that we'll avoid any meaningful clean-up of the gene pool.

TWT
20th Mar 2018, 10:49
Very good point longer ron.

Do any of these 'autonomous' cars being tested, or indeed any EV's, have a synthetic sound to warn pedestrians of their approach ?

With EV's, one of the senses which people have relied on for their whole lives, is no longer useful to assist with 'separation'.

cattletruck
20th Mar 2018, 10:50
When you enter your autonomous vehicle to be driven somewhere, how do you know all the sensors, software library functions and reactive mechanical tasks are in full working order? Or do you find this out along the way....

"Fixed in version two-point-oh" is just not good enough.

jolihokistix
20th Mar 2018, 11:03
And how do all the cameras work in rain or snow, or simply when dusty?

hiflymk3
20th Mar 2018, 11:09
Next it will be. "Driver terminates jaywalking robot..."

treadigraph
20th Mar 2018, 11:11
Next it will be. "Driver terminates jaywalking robot..."

Well, most humans with a smart phone in their hand appear to be automatons... Add a pair of headphones and they are completely remote from the world about them.

IFIX
20th Mar 2018, 11:30
I have often wondered if and to which extent these vehicles are capabe of judging what the safe speed is for the given conditions.
I guess that when the posted speed limit is 40 mph, then 40 mph is what it will do.

Do these vehicles take into account the situations a human driver should?
Kids playing with a ball on the sidewalk for instance make me expect a ball from between parked cars and a kid chasing it.
I will adapt my vehicle speed as required to the situation.

G-CPTN
20th Mar 2018, 11:40
With EV's, one of the senses which people have relied on for their whole lives, is no longer useful to assist with 'separation'.

How come that there were pedestrian fatalities involving early trains?
They (the trains) were hardly silent . . .

The earliest recorded fatality caused by a steam locomotive was an unnamed woman, described as "a blind American beggar", fatally injured by a train on the Stockton and Darlington Railway on 5 March 1827.
At least two people were killed on the Liverpool and Manchester Railway prior to its opening to the public.
William Huskisson was mortally wounded at the opening of the Liverpool and Manchester Railway when he was run down by Stephenson's Rocket.

IFMU
20th Mar 2018, 11:51
Well, most humans with a smart phone in their hand appear to be automatons... Add a pair of headphones and they are completely remote from the world about them.
Perhaps in the near future we will have an app to warn human automatons about car danger. Or more likely, the autonomous cars will track the humans by their phones.

I say we go back to manual transmissions!

cattletruck
20th Mar 2018, 12:03
Perhaps in the near future we will have an app to warn human automatons...

Funny you should say that as this morning I had an idea. I'm going to fight fire with fire by writing an app that plays an automobile horn out aloud. This should scatter them zombie automatons out of my way next time I'm walking along the footpath (probably into the path of real cars :}).

ORAC
20th Mar 2018, 12:03
Gertrude.

Jaywalking is crossing the road except in a marked crosswalk. It is an offence in many states of the USA.

It would appear Arizona is one of those states.

https://www.azcentral.com/story/news/local/glendale/2014/09/30/what-is-jaywalking/16496643/

longer ron
20th Mar 2018, 12:04
How come that there were pedestrian fatalities involving early trains?
They (the trains) were hardly silent . . .

In those days most people probably were not used to anything travelling faster than walking pace or a gently trotting horse pace.
I have seen pedestrians almost struck by modern quiet cars - it does not necessarily have to be an EV - but EV's are deathly quiet and if there is any background noise.... well !!??
Even adults can get 'target fixation' on something across the road (or simply make a mistake with direction of traffic/speed etc).

As IFIX posted - if there are kids playing on the pavement or between parked cars - I always slow down because of the risk of running out after a ball etc - I doubt an autonomous car would register them as a risk.

VP959
20th Mar 2018, 12:08
There may be a difference in outlook between those here living in countries where "jaywalking" may be an offence, and those living in countries where crossing a road on foot anywhere is legal.

Here in the UK it's legal to cross any road, except a motorway, on foot, and doing so is commonplace pretty much anywhere. Pedestrians are warned in the Highway Code to avoid crossing roads from behind stopped or parked cars, but drivers here know full well that it happens all the time, especially with children.

As a consequence, most drivers here will half-expect a pedestrian to appear from behind, or between, parked cars and try to cross the road, especially in some crowded urban areas. That may not be the case where pedestrians are taught to only cross roads at specific points.

I remember being physically grabbed by the arm and hauled back on to the pavement (sidewalk) by a colleague on my first visit to the USA, for example. I'd just assumed that the opposite sense of the UK Highway Code rule applied, so "Look Left, Look Right, Look Left again, and if safe cross" to account for the reversed traffic directions in the US. I had no idea that jaywalking was an offence.

Wyler
20th Mar 2018, 12:08
This was a deliberate attack on a riderless bike by a driverless car. The human was irrelevant and expendable.

Welcome to the future.

longer ron
20th Mar 2018, 12:11
How come that there were pedestrian fatalities involving early trains?
They (the trains) were hardly silent . . .

Just checked - Steam Trumpet and later Steam Whistles not fitted to Locos until after 1833,prior to that the most a driver had was a Mouth - blown Horn,if anything.

longer ron
20th Mar 2018, 12:15
I remember being physically grabbed by the arm and hauled back on to the pavement (sidewalk) by a colleague on my first visit to the USA, for example. I'd just assumed that the opposite sense of the UK Highway Code rule applied, so "Look Left, Look Right, Look Left again, and if safe cross" to account for the reversed traffic directions in the US. I had no idea that jaywalking was an offence.

My other half giggles at me when we are in the States because I become much better behaved at crossing roads,prior to our first trip there together she did not know about Jaywalking.

meadowrun
20th Mar 2018, 12:28
While I don't run across highways, I tend to cross normal streets wherever I need to, when traffic allows.
If you can't use your senses to cross a road safely, if you are not situationally aware, perhaps you shouldn't be out at all.
Always amuses me, when out really late or early, to see pedestrians waiting for the traffic lights to change when there is not one headlight to be seen in any direction.

Golf-Sierra
20th Mar 2018, 12:30
I have often wondered if and to which extent these vehicles are capabe of judging what the safe speed is for the given conditions.
I guess that when the posted speed limit is 40 mph, then 40 mph is what it will do.

Do these vehicles take into account the situations a human driver should?
Kids playing with a ball on the sidewalk for instance make me expect a ball from between parked cars and a kid chasing it.
I will adapt my vehicle speed as required to the situation.

What I wonder about is will there be an independent inquiry into this accident of the same scrutiny as would be the case for say an airplane crash, severe adverse effect of medication or some incident in the nuclear industry? Will someone look at this, ask the same sort of questions you have asked and make recommendations? Will such a report become available to the public?

Golf-Sierra

Gertrude the Wombat
20th Mar 2018, 12:48
Jaywalking is crossing the road except in a marked crosswalk. It is an offence in many states of the USA.
How does this work then? Does every suburban cul-de-sac have a marked crossing every few yards, or does this rule only apply to major roads (in which case how are roads marked so that you can tell the difference)?

Carbon Bootprint
20th Mar 2018, 12:54
What I wonder about is will there be an independent inquiry into this accident of the same scrutiny as would be the case for say an airplane crash, severe adverse effect of medication or some incident in the nuclear industry? Will someone look at this, ask the same sort of questions you have asked and make recommendations? Will such a report become available to the public? Apparently, both the US National Highway Traffic Safety Administration and the National Transportation Safety Board have said they were sending teams to Tempe. Rather unusual for a motor incident, but given the "world's first" status they likely felt compelled to look into it. I believe anything those agencies publish becomes a matter of public record.

The NTSB did publish a preliminary report (https://www.ntsb.gov/investigations/AccidentReports/Pages/HWY16FH018-preliminary.aspx) on the guy whose car drove itself under a truck while he was reportedly watching a Harry Potter DVD.

RAT 5
20th Mar 2018, 13:16
It could be an interesting 'who's at fault' case. If the car didn't stop for technological reasons then is it the designers/engineers/construction manufactures at fault. If the human didn't intervene, but could have and prevented the death, are they at fault? Was the monitor even insured for such an event? I hope the manufacturer has their back if deep pockets are needed. No doubt they are an employee, but even then it depends on how closely they were monitoring?

Golf-Sierra
20th Mar 2018, 13:59
I just had a look at my local Uber homepage. Not a word about the accident.

er340790
20th Mar 2018, 14:29
New technology often does have unintended consequences.

When ABS equipped cars first appeared in numbers in the 80s, insurers were shocked that their accident rates were HIGHER than the average...

Why? Because non-ABS equipped cars, unable to stop as promptly, kept running into the back of them!!! The accident rates only fell as ABS became standard equipment.

I predict similar unintended consequences for self driving cars, at least until they become mandatory for all.

Pontius Navigator
20th Mar 2018, 14:43
There is a difference in behaviour between human and robot in a speed control zone.

The robot will be aware of the speed limit, it will assess traffic and miss distances, and drive at the highest authorised and safe speed - based purely on logic.

A human will frequently ignore the speed limit, assess that the road is clear and drive, generally safely, at a speed in excess of the limit. Equally they may note the time, observe children playing with a ball near the road, and assess that there may be a sudden incursion into the drive line. A slower speed will be driven. Largely this is based on experience and instinct.

Only when humans recognise the limits of robotic cars they take appropriate precautions. You would probably be safer crossing a motorway where robot cars were travelling at 70 than stepping out from between parked vehicles,

ORAC
20th Mar 2018, 15:35
Gertrude, see the link below. The rules vary widely but generally apply to where there is a marked pedestrian crossing within 30-100m. Interestingly the page has been updated to state that the fine in Tempe, Arizona, where the accident occurred, is up to $118.

https://en.wikipedia.org/wiki/Jaywalking

How the rules are applied also vary wildly.....

https://www.telegraph.co.uk/news/uknews/1539148/Historian-pinned-to-ground-by-US-police-and-beaten-for-jaywalking.html

jolihokistix
20th Mar 2018, 15:44
Thinking about being run over by an automatic vehicle makes my blood boil. I want to smash something. Being run over by an unobservant driver is somehow more forgivable.

Will this incident turn out to have been the first example of a robot killing a human, I wonder?

ORAC
20th Mar 2018, 15:54
Lots of robots have killed people - mainly through having been fed bad data or orders. Aircraft navigation systems for example.

People has also been killed by robots in factories (http://www.independent.co.uk/news/world/americas/robot-killed-woman-wanda-holbrook-car-parts-factory-michigan-ventra-ionia-mains-federal-lawsuit-100-a7630591.html).

n5296s
20th Mar 2018, 16:13
Not sure if jaywalking is an offence (or indeed offense) everywhere in the US. The only place I've ever heard of serious enforcement is the Seattle area, where I know of people getting tickets for it. Round here in the suburbs of San Francisco I've never heard of any enforcement. In downtown Mountain View people mostly use marked crosswalks, but not always.

mickjoebill
20th Mar 2018, 17:59
And how do all the cameras work in rain or snow, or simply when dusty?
Some use radar and IR.

The police said there was no evidence that the car had braked before the impact.
Well you would not expect any skid marks from an auto braking robot with ABS so it was a stupid thing to say that suggests the system did not recognise the victim!

In fact it probably did recognise the victim but the laws of physics still apply to a heavy Volvo doing 40mph.

Mjb

Gertrude the Wombat
20th Mar 2018, 18:06
Interestingly the page has been updated to state that the fine in Tempe, Arizona, where the accident occurred, is up to $118.
Yup. A fine of $118, not a death sentence.

ORAC
20th Mar 2018, 18:53
Mayhap, but on pedestrian crossings the pedestrian has right if way; when someone crosses elsewhere the laws of physics and Darwin apply.

Pontius Navigator
20th Mar 2018, 19:18
ORAC, they may be in the right, but the laws of physics still apply. How many drivers slow down when they see a pedestrian approaching a crossing on the off chance they will suddenly decide to cross?

ORAC
20th Mar 2018, 19:43
Well I always do, on the basis that if they put a foot on the crossing I am legally obliged to stop before I reach it.

Gertrude the Wombat
20th Mar 2018, 19:52
Well I always do, on the basis that if they put a foot on the crossing I am legally obliged to stop before I reach it.
What pisses me off is pedestrians who walk up to a crossing, then stop, so I have to stop, then only when I have stopped do they start crossing. Whereas if they'd never stopped in the first place I wouldn't even have had to slow down.

treadigraph
20th Mar 2018, 20:04
What pisses me off is pedestrians who walk up to a crossing, then stop and stare at their mobile bloody phone, so I have to stop, then only when I have stopped do they start crossing. Whereas if they'd never stopped in the first place I wouldn't even have had to slow down.

You missed out a few words! ;)

vapilot2004
20th Mar 2018, 20:11
On January 25, 1979, 25-year-old Robert Williams climbed into a storage rack to retrieve parts from a malfunctioning robot at Ford’s Flat Rock plant in Michigan. The robot, not able to sense Williams' presence, swung round and struck him on the head, killing him instantly. The robot kept working for 30 minutes as Williams lay dead on the floor. His death, nearly forty years ago, makes Williams the first person to be killed as a result of actions by a robot.

Robotic machinery, cars and trucks, and even modern aircraft have all caused deaths due to programming errors. While far from complete, and over a decade old, Wired offers their short list of some whoppers - History's Worst Software Bugs - Wired 2005 (https://www.wired.com/2005/11/historys-worst-software-bugs/)

Kiltrash
20th Mar 2018, 21:53
and wait for the driver-less trucks doing 56mph on the motorway nose to tail and the boy racer comes down the slip road at 90 knowing the trucks will all give way to him
F55K the multi car pile up behind the stopping trucks

Condolences to the family of the victim but until all vehicles and humans are automated ......

Gertrude the Wombat
20th Mar 2018, 22:32
F55K the multi car pile up behind the stopping trucks
Any car driver must at all times drive in such a way that if the truck in front of them does an emergency stop for no visible reason they won't go into the back of it, taking account of the fact that the truck has better brakes than they do. But you knew this.

ZeBedie
20th Mar 2018, 23:09
If the car was indeed driving itself, I'll be interested to hear the reason for it exceeding the speed limit - seems unlikely that it would, unless badly programmed or overridden by the human.

M.Mouse
20th Mar 2018, 23:44
Any car driver must at all times drive in such a way that if the truck in front of them does an emergency stop for no visible reason they won't go into the back of it, taking account of the fact that the truck has better brakes than they do.

A 44 tonne truck will most definitely not stop quicker than most cars!

Some numbers. (http://www.ukmotorists.com/hgv%20braking%20distances.asp)

jimjim1
20th Mar 2018, 23:58
If the car was indeed driving itself, I'll be interested to hear the reason for it exceeding the speed limit

The limit is 40. According to the sign visible on Google Street View.

jolihokistix
21st Mar 2018, 00:51
ZeBedie, there is no mention that the car was speeding.

b1lanc
21st Mar 2018, 01:42
How does this work then? Does every suburban cul-de-sac have a marked crossing every few yards, or does this rule only apply to major roads (in which case how are roads marked so that you can tell the difference)?
There are marked crossing lanes in which you are supposed to cross in most metro areas. Some folks cut corners, i.e. cross at a 45 instead of waiting for a traffic light to cross at 90 and waiting for a second traffic light to cross 90 again. That can range up to 5 minutes total to continue journey. But, there are not only traffic signals, but pedestrian signals that count down from say 15 seconds. For example, it is legal for me to make a vehiclular right turn on red - except if a pedestrian is crossing and the countdown time has not reached "0". Technically, I'm supposed to wait until the display reads 0 even if no person is in sight. Most people go if pedestrian is visible but I've seen people ticketed for doing so. In urban areas, there could be miles between traffic lights and crossings so anyone can cross anywhere in between. That said, if an autonomous vehicle hit the brakes, I'd guess it would be rear-ended by any number of following vehicles. No quarter!

If ever in NYC, good luck on avoiding jaywalkers. I'm just curious how these autonomous cars are going to deal with the spontaneous windshield washers that descend on all vehicles that stop for every traffic light and then demand 'tips'.

You never know what you find. In my rural area, I have a cul-de-sac with two houses on it that has a yield sign (from both directions since they joined two roads at the cul-de-sac) and a basketball hoop...Hmmmm. I'm not sure if I'm supposed to yield to basketball players or people coming out of their driveways....

The better half and I have also had a 'near' head-on with a google 'autonomous' vehicle taking street views. This took place on a rural two lane road (well marked with center lines) with a 48kmh limit. At the time, the autonomous vehicle swerved into our lane and supposedly there was a driver monitoring. We never saw any human form in the car. Of course, we were in the ditch on the other side:ugh:

b1lanc
21st Mar 2018, 02:06
Any car driver must at all times drive in such a way that if the truck in front of them does an emergency stop for no visible reason they won't go into the back of it, taking account of the fact that the truck has better brakes than they do. But you knew this.
That's called self preseveration. However, you can't fix 'stupid'!

Traffic_Is_Er_Was
21st Mar 2018, 02:50
truck has better brakes than they do.
Hydraulically operated discs all around vs air operated drums? The amount of dual wheel skidmarks on my local motorway indicates to me that lockup (on the trailers mostly) is common, and once the tyres are sliding...

crippen
21st Mar 2018, 02:55
Have they programmed the cars to do a runner after an accident,the driver often runs off in Thailand/:uhoh:

Ascend Charlie
21st Mar 2018, 07:57
The biggest challenge to an autonomous car has to be driving in Jakarta. 4 lanes of traffic on a 3-lane marked road, changing lanes at random without signal, dodging potholes, making u-turns whenever they feel like it, plus add in the scooters and tuk-tuks (bajaij) and you have something that only a human could interpret.

Amazingly, you can do almost ANYTHING up there, as long as you do it at a slow and predictable pace, and in 20 years of going there I have only ever seen 1 scooter knocked over.

radeng
21st Mar 2018, 08:19
There is a saying with some truth in it: 'To err is human: to really foul things up, you need a computer".

ORAC
21st Mar 2018, 09:06
Regardless, the family is bound to suit Uber, Volvo and whoever else has money for being responsible.....

Driverless Uber car ‘not to blame’ for woman’s death (https://www.thetimes.co.uk/edition/news/driverless-uber-car-not-to-blame-for-woman-s-death-klkbt7vf0)

A self-driving Uber vehicle that struck and killed a pedestrian in Arizona was probably not to blame for the accident, according to a police chief who reviewed footage of the crash......

Sylvia Moir, the chief of police in Tempe, a city to the east of Phoenix, watched footage from the cameras in the Uber car. It was in self-driving mode at the time of the crash. The accident happened at 10pm on Sunday when Ms Herzberg was crossing a six-lane road near a junction. She pushed her bike across two lanes, over the central reservation and across another four lanes to a grass bank. She appears to have almost reached the far side when she was struck by the Volvo, which was said to have been travelling at 38mph. The car did not brake, according to police.

From the videos “it’s very clear it would have been difficult to avoid this collision in any kind of mode [autonomous or human-driven] based on how she came from the shadows right into the roadway,” Ms Moir told the San Francisco Chronicle. “The driver said it was like a flash, the person walked out in front of them. His first alert to the collision was the sound”.......

ZeBedie
21st Mar 2018, 10:33
ZeBedie, there is no mention that the car was speeding.

https://www.telegraph.co.uk/technology/2018/03/20/ubers-self-driving-car-speeding-killed-pedestrian-police-reveal/

sitigeltfel
21st Mar 2018, 10:38
Scenario..

Airliner using Autoland technology strikes and kills airport worker who strays onto the runway in the dark.

Who would be to blame?

Yamagata ken
21st Mar 2018, 10:48
Any car driver must at all times drive in such a way that if the truck in front of them does an emergency stop for no visible reason they won't go into the back of it, taking account of the fact that the truck has better brakes than they do. But you knew this.

Utter tosh from someone who's never driven a lorry or a car with front drums. Lorries have drum brakes, cars have disc brakes. Disc brakes out-perform drums at every level except for cheapness and convenience to arrange a hand brake. :=

bingofuel
21st Mar 2018, 11:18
Utter tosh indeed!

It is not the brakes that stop a vehicle, brakes stop or slow the wheels from turning, it is the coefficient of friction between the tyres and road surface that dictate how quickly the vehicle will stop.

Golf-Sierra
21st Mar 2018, 11:59
Utter tosh from someone who's never driven a lorry or a car with front drums. Lorries have drum brakes, cars have disc brakes. Disc brakes out-perform drums at every level except for cheapness and convenience to arrange a hand brake. :=

Modern lorries have disc brakes.

cattletruck
21st Mar 2018, 12:18
These developers of autonomous vehicles will also need to program their computers to recognise children playing cricket on the street.

Carry0nLuggage
21st Mar 2018, 13:06
Autonomous vehicles need to recognise that a hazard is not just something on a constant bearing with decreasing range but locations where such hazards may appear from. Hidden driveways, gaps between parked cars, occupied parked cars, buses, delivery vans, dog on a lead attracting the attention of another dog not on a lead, the deer following the one which crossed in front of you. All those things we pick up from our own behaviour as road users.

As told to me by someone in the haulage industry, trucks running empty can lock up their wheels very easily. They are designed to stop 44 tonnes, take off 28 tonnes of payload and you've got a lot of spare braking power. Also, those thick black tyre marks swerving onto the hard shoulder come from trucks with a brake system leak. The driver wants it off the road quickly while it can still move, before all the wheels lock.

Blues&twos
21st Mar 2018, 13:47
When I first started driving lorries I discovered to my alarm that braking on approach to a roundabout at the end of a dual carriageway with an empty flatbed truck can cause the back wheels to lock, which in turn causes the back wheels to bounce up and down. This seriously reduces your braking power and scares the life out of you...

MG23
21st Mar 2018, 16:18
Airliner using Autoland technology strikes and kills airport worker who strays onto the runway in the dark.

Who would be to blame?

I'm guessing people aren't expected to walk across runways unless the runway is closed. They are expected to walk across roads.

Similarly, no-one would blame a driverless train if it hit someone who'd decided to take a walk along a railway line.

In ten or twenty years, we'll probably have the tech to let cars safely drive themselves. The cheerleaders pushing them onto the roads today are just asking for trouble, because it's clearly nowhere near safe enough.

Ancient Observer
21st Mar 2018, 16:24
[QUOTE=Carry0nLuggage;10091563]Autonomous vehicles need to recognise that a hazard is not just something on a constant bearing with decreasing range but locations where such hazards may appear from. Hidden driveways, gaps between parked cars, occupied parked cars, buses, delivery vans, dog on a lead attracting the attention of another dog not on a lead, the deer following the one which crossed in front of you. All those things we pick up from our own behaviour as road users.

Quite right.

RAT 5
21st Mar 2018, 20:21
The discussion follows the lines of it'll be OK when all cars are self-drive. There is a huge difference between little shopping town cars and the motorway km eaters. The manufactures are satisfying the macho delights of the rich by making absurdly powerful & fast cars, and their owners like to release their heads on the German autobahns.
Is anyone suggesting that mixing self drive 80-100kph & 200kph cars would be a safe idea, especially as there are too many who just pull out last minute to pass the 90kph lorry. To satisfy safety ideals, cars will all need to be limited to a max speed on the open road, and autobahns likewise. That will require a massive cultural change for the petrol heads. The politics of that are mind-numbing.

G-CPTN
21st Mar 2018, 22:45
Does anyone know whether the speed of autonomous cars is governed by the operators or by the technology?
Are there (currently) any autonomous cars operating at 200kph?

Blues&twos
21st Mar 2018, 22:54
The point raised earlier in the thread about a driver's experience and the ability to spot potential hazards not immediately related to driving is interesting.
Car 'body language' can often give away a dodgy driver.

G-CPTN
21st Mar 2018, 23:15
On October 19th 2014 an Audi RS 7 piloted driving concept raced the Hockenheimring driverless at the physical limit:-

https://www.youtube.com/watch?v=DeWriarFlsU

jack11111
21st Mar 2018, 23:20
If you think gun owners raise a stink about loosing their guns, wait until you tell Americans they can no longer drive a car because it is incompatible with the safety of self-driving cars.

Ha, ha, ha.

The steering wheel from "my cold dead hands".

jolihokistix
22nd Mar 2018, 01:12
Better get a 28th Amendment in there right away. "Our inalienable right to hold a vehicular handle on the public highways and byways."

WingNut60
22nd Mar 2018, 02:34
With dash cam footage from the vehicle now released, there are obvious questions arising from that vision.

Such as:-

What made the woman think that it was safe or OK to cross that road with her back to oncoming traffic?
But also, why did the vehicle systems not see her as an obstacle on the road. There was no obvious indication of braking.

jolihokistix
22nd Mar 2018, 03:13
Does Uber have in place a set of rules defining percentage of liability in any accident where a human is retained to watch over the automatic driving?

WingNut60
22nd Mar 2018, 04:24
Does Uber have in place a set of rules defining percentage of liability in any accident where a human is retained to watch over the automatic driving?

No, but apparently the law does.
The safety driver is in the spotlight apparently, for not being in control of the vehicle.

sitigeltfel
22nd Mar 2018, 05:49
Dashcam video of the incident

https://youtu.be/5mDxiYguNPI

Is the "driver/observer" reading from a phone or some literature?

MungoP
22nd Mar 2018, 08:39
'Driver' unfocused.. as way too many are these days.. Stupid woman with bike wearing dark clothing fails to spot vehicle 14ft long, 6ft wide with lights on all corners.

treadigraph
22nd Mar 2018, 08:41
Neither of them seem to paying much attention; she looks as though she might be looking down at something too.

ORAC
22nd Mar 2018, 08:50
Want to research the number of people crossing poorly lit highways/freeways/motorways who are hit and killed by vehicles and the drivers had no idea what they had hit?

Even if the driver had bee; staring intently through the windscreen I doubt that the outcome would have been any different, or any automated system would have made any difference.

Pontius Navigator
22nd Mar 2018, 09:14
Friend of mine, driving down A1 at night, little traffic and no speed limit in those days hit and killed a tramp crossing the road.

A police car, not far ahead of him, only just missed him. They testified my friend was not to blame. In this day and age it would of course be different.

ShyTorque
22nd Mar 2018, 10:14
It appears that the driver in the video (did she consider herself more of a passenger?) was looking down for much of the time in the seconds leading up to the collision, certainly not alert as a driver ought to be.

*It also appears that the sensors on the car did not pick up the crossing pedestrian at all. I thought they were meant to be "smarter" than humans?

Having said that, it certainly isn't sensible to be wearing dark clothing and not showing lights whilst crossing the road in the complete darkness under those circumstances (especially when pushing a bicycle, which considerably limits one's personal manoeuverability).

Swiss cheese effect.

*I wonder how these sensors work, in view of the adverse weather we've had this winter. Do they cope with snow or do they become "blocked" by it? How does a "robotic" car cope with lying snow and icy surfaces? I think I already know the answers. Driving in icing conditions prohibited....?

ORAC
22nd Mar 2018, 10:32
I think the jury is out on whether the vehicle brake does or not. A said, with ABS it would not have left any skid marks, and I doubt if the driver would have noticed in the moments before the collision. It will have to wait till the vehicle black box has been examined.

Certainly the photo of the car shows less damage than I would expect. The data I have been able to find shows that at speeds around 40mph both the bike and pedestrian should have lifted over thee bonnet and into the windscreen.

https://pbs.twimg.com/media/DYw2h07WsAEo50W?format=jpg

Hydromet
22nd Mar 2018, 11:13
From the video, it seems that in the visual spectrum the woman was only visible for a short time, probably too little time for the car to stop. But, I thought these cars had other sensors, such as IR and radar, which I would have thought would detect her earlier. Comments from those who know more than I would be appreciated.

IFMU
22nd Mar 2018, 11:24
The pedestrian crossed an entire lane before ending up in front of the car. If the driver was driving, perhaps they would have had high beams selected as it seems there were no oncoming traffic. Then they could have seen the pedestrian.

VP959
22nd Mar 2018, 11:33
There seem to be a host of issues here that combined to cause this accident, but how do we know which were the most critical, as far away observers, only reading news reports?

I wouldn't read too much into the external dash cam footage that's been released, solely because I've had a forward facing dashcam fitted to my cars for years, and they really aren't great at night - my eyes can see a great deal more, especially peripheral movement, at distance, than ever shows on the dash cam recording.

The internal camera is more interesting, as it shows that the "driver" was barely paying any attention to what was going on outside the car. That raises the question as to whether the "driver" of one of these test cars has the same, or similar, responsibilities to a normal driver or not. Not something I know the answer to as I don't know the laws that apply in that state for this trial. My view is that his behaviour was careless, as I believe he was supposed to be the "safety driver", who's primary role was to keep a good look out and act if the autonomous systems failed to detect a hazard. He doesn't seem to have been doing that at all.

Clearly the pedestrian seems to have not seen the oncoming car at all, from what's been released, so seems to have behaved as carelessly as the car "driver".

The unanswered questions seem to revolve around the ability of the sensors to detect the woman pushing the bicycle at distance. Can the sensors sound the car horn as a warning, for example? If so, doing this may have stopped the woman before she reached the path of the car (assuming she had reasonable hearing).

I thought that most of these autonomous cars had a mix of visual and lidar sensors, plus, perhaps, doppler radar for closing speed measurement. They seem able to look ahead, day and night, to detect the edge of the road or road lane, and detect stationary and moving obstacles day and night, so the big question here is what the sensors actually saw. My guess is that during this trial the cars will be fitted with some form of telemetry or data recording system, so there should be recordings from the car sensors for the period leading up to the accident.

The question the lawyers will probably fight over is "Who was the most careless, the pedestrian pushing the bike across the road or the driver who wasn't looking where he was going much of the time?". Whether the autonomous systems are blamed may not be here nor there, as I suspect that ultimately the car has to be "under the control" of the "driver", as far as the law may be concerned.

G-CPTN
22nd Mar 2018, 11:51
If, as the driver, you believe that the system is infallible, why would you remain poised to respond 100% of the time with your foot over the brake pedal?

We have seen videos of Tesla drivers 'relaxing' and reading whilst car's autonomous system controlled the vehicle.

Hotel Tango
22nd Mar 2018, 12:10
I'm not pro self-driving cars in the least, but having watched the video I very much doubt that the outcome would have been any different with a human behind the wheel. What the heck was going on in the mind of the woman crossing? Was she under the influence of drink or drugs? How could she not see the car's headlights? Tragic whatever, but I really doubt that in the prevailing conditions a human driver would have managed to avoid her.

dogsridewith
22nd Mar 2018, 13:15
Were both of the Volvo's headlights on? In the in-car forward-facing camera, it seems pretty to dark to the driver's left where the bike was coming from.

Headlights not on would be a problem for the pedestrian crossing the road.

Particularly in urban settings w/ a lot of ambient lighting, one encounters a fair amount of cars driving w/ headlights off.

PDR1
22nd Mar 2018, 13:15
Except that had a human been at the wheel the lights would probably have been set to main beam rather than dipped, making the pedestrian visible from much further away. But as VP says, the video isn't a reliable guide to how visible the pedestrian actually was.

Having said that, I'm surprised how little reaction to the oncoming car can be seen in the pedestrian.

PDR

ORAC
22nd Mar 2018, 13:30
Not directly related to this accident, but on the safety of such systems. Echoes the concerns on the increased automation of cockpit systems and decay of pilot skills. But they will keep on being used because the overall accident rate continues to go down - because humans are the cause of most accidents.....

....”Joshua Brown was killed in May when his Tesla Model S collided with a truck while Autopilot was activated. NHTSA closed the investigation and determined Autopilot was not at fault because Brown had seven seconds to hit the brakes before the car collided with the truck.

The accident raised concerns that people were beginning to over-rely on Level 2 autonomous systems, believing them to be more capable than they actually are.

After that accident, Consumer Reports called on Tesla to rename Autopilot and to disable its hands-free operation to make it clear the system wasn't fully self-driving. A warning will now sound if a Tesla driver takes his or her hands off the wheel while Autopilot is activated.

Both Ford and Waymo, Alphabet's self-driving-car company, have said they are developing fully self-driving cars because they fear people become too complacent in Level 2 systems.

Although Level 2 driving systems come with their risks, they are also much safer than cars on the road today. Crash rates for Tesla vehicles have fallen 40% since Autopilot was first installed in 2015.”.........

ShyTorque
22nd Mar 2018, 13:57
Surely as the law stands, the captain of the ship/aircraft/vehicle remains responsible for safety. I don't see how I could ever relax in a "self driving" vehicle.

Highway1
22nd Mar 2018, 13:58
Surely as the law stands, the captain of the ship/aircraft/vehicle remains responsible for safety. I don't see how I could ever relax in a "self driving" vehicle.

Doesnt that kinda kill off the whole point of it?

ShyTorque
22nd Mar 2018, 14:24
Doesnt that kinda kill off the whole point of it?

Correct. That's why I will never buy one.

VP959
22nd Mar 2018, 14:36
Doesnt that kinda kill off the whole point of it?

Remember this isn't legally a fully autonomous car, it's at best a semi-autonomous one that is on trial and requires a virtually hands-on "safety driver" in order to operate legally under the trial conditions.

The best analogy I can think of is a student driver, driving a car with dual controls under the supervision of an instructor (same scenario as a student pilot, for those countries that don't have dual control driving school cars).

The student driver is the autonomous system, which is still learning how to deal with a wide range of different driving conditions and the skills needed to handle them (which I believe is the case with the self-learning part of these autonomous systems).

In the event of an accident, it isn't the student who is legally "in control" of the car, it's the instructor in almost all cases. In this case the "safety driver" is equivalent to an instructor, in that he/she is able to take over control at any time if the student gets into a situation that is beyond their skill level.

If there isn't a condition in the agreement for this trial on public roads that says that the "safety driver" has a duty to keep a good lookout and take over control in the event of a situation arising that is outwith the capability of the car's systems, then I will be surprised, especially as Tesla have already made it clear that drivers have this duty of care when using their cars in autonomous mode.

It may be that the "safety driver" wasn't briefed on this, or if he was he ignored it, but, notwithstanding the carelessness of the cyclist I wouldn't have expected a safety driver to have his head down inside the car most of the time.

Ancient Mariner
22nd Mar 2018, 15:21
Doesn't that kinda kill off the whole point of it? ;)
Per

MG23
22nd Mar 2018, 15:57
Doesn't that kinda kill off the whole point of it?

Pretty much, yes. I'd like a system that can at least drive on the highway during the day, where I can see potential hazards well ahead of me and be ready if the car screws up. But if I have to keep my hands on the steering wheel the whole time, what's the point?

If it's true that the driver was texting at the time of the crash, rather than looking out the windows, I'm guessing they're screwed.

tdracer
22nd Mar 2018, 20:06
Except that had a human been at the wheel the lights would probably have been set to main beam rather than dipped, making the pedestrian visible from much further away. But as VP says, the video isn't a reliable guide to how visible the pedestrian actually was.

At least on this side of the pond, using the high beams in urban, populated areas is frowned on even if there is no other vehicular traffic (blinding and annoying to others, especially the newer, much brighter high beam systems). I've literally gone months without using the high beams because I've not ventured out into any rural areas.
That being said, missing a distracted pedestrian on a poorly lit street is an area where I'd expect an autonomous vehicle to actually be better than human since the sensors are not dependent on the visible light spectrum. I'd expect the investigation to focus on why that wasn't the case - at least after they get through crucifying the human 'monitor' driver...

MungoP
22nd Mar 2018, 20:41
missing a distracted pedestrian

The Darwin effect.. Dark clothes at night and not paying attention.. Stepping into a road at all is something that should demand total focus.. it's where the cars live. I see no reason to discard advances in technology over the stupidity of both the woman stepping into the road while not paying attention or the stupidity of the woman behind the wheel not paying attention. What IS it with these people that can't keep their minds on the job ?

ORAC
22nd Mar 2018, 20:43
MungoP,

If you want to prove the technology you want the lowest common denominator behind the wheel, not the highest....

Ogre
22nd Mar 2018, 21:04
I haven't read everything that seems to be published about this incident or the car setup, so perhaps someone can enlighten me on the following:

What was the "driver" looking at on the internal video? Does the car have some form of instruments or displays that need to be monitored which would distract the driver?

Exactly how much control is the driver supposed to have?

I assume the car was fully serviceable at the time, as any mechanical fault or failing would affect the autonomous performance (doesn't matter how good the autonomous system is, if there are bad brakes the car won't stop as fast as the system expects).

When I heard the original report it said that the woman had been pushing her bike along the aide of the road, in which case the issue would have been one of lane keeping, but the video shows her pushing the bike across traffic, which is dangerous enough when the cars are being driven by humans!

I can trust a machine to do what it's told more often that I trust a human!

ZeBedie
22nd Mar 2018, 21:08
I think a human driver would have been able to take at least some action, whether successful or not, but for a system using 'see in the dark' technology to do nothing is shocking. I suspect the people developing driverless cars in other companies are furious about the damage done to their emerging industry by an event which should not have happened.

Blues&twos
22nd Mar 2018, 21:50
The obstruction in the road not being 'seen' by the car's sensors is the only issue here (if, it turns out, the car really didn't see it).

The obstruction being human, and the monitoring driver not taking action are not relevant in the context of ultimately allowing totally self-driving vehicles to share our roads.

vapilot2004
22nd Mar 2018, 21:58
It would appear that the pedestrian took three steps while (at least partially) under the illumination of the car's main beams. Viewing the video the beam cutoff is clearly visible, strongly suggesting that high beam lights certainly would have helped, if they were selected and the driver (or the self-driving computer) was paying attention.

Of course, radar and IR sensors, if so equipped, do not require visible light to detect obstacles and should have picked up the pedestrian and sounded the alarm before she crossed into the car's lane. Judging by the car's speed, braking alone would not have served to prevent contact with the girl and her bicycle, as it appears there is insufficient stopping distance available, setting aside reaction times, although a quick jog into the other lane would have done so.

Had this been a standard automobile, the driver would be guilty (technically if not legally) of over-driving their headlights. Main beams have between 1/5 and 1/3 the effective distance of high beam headlights. It is generally known that at speeds greater than 35mph, main beams are insufficient in range for preventing a collision at night, particularly on an unlit or poorly lit road, as the beam's illumination range is less than the stopping distance for most cars, again, leaving aside the reaction time.

bbpathfinder
23rd Mar 2018, 04:11
I live in the area where this happened, so took a little detour on the way home just now. Drove the exact route and speed the uber did, its dark now [just like the time of the accident] and my 54 yr old eyes cant be as good as they used to be. The exact location of the accident is not pitch dark, but has street lights, as well as backround light from an office building up the road a few hundred feet. The video appears MUCH darker than the area actually is...
With all that said, if any driver was paying attention I'm almost positive they would see at least an outline of someone crossing that road. Plus, if the bike the lady was pushing had anything reflective on it that would've been shown brightly in the lights. The lady crossing was in a bad spot, but I feel I would've seen her, the uber driver should have, and I cant believe the uber car didn't sense her in the road.

As an avid cyclist I worry about getting run over by texting teens, now I gotta worry about self driving cars too?

Ancient Mariner
23rd Mar 2018, 08:17
Don't know about the fab Swede, but my humble Skoda has front lights that automatically adjust to it's surroundings. It will increase/decrease light intensity and widen/narrow the beam as required.
It needs,however, to be turned on.
Per

MungoP
23rd Mar 2018, 08:42
I believe that the current state of this technology as stated by the manufacturer is proved. They are not claiming that these cars can be trusted with only your children aboard to find their way to their grandmother, only that they relieve the driver of much of the demands of driving. The very fact that the car had a 'driver' in place as a back-up to the systems is evidence of the current level of automation. The fact that the driver was not paying attention is not the fault of the technology. We simply have to accept that two stupid people can overcome the sophisticated technology at its present level of development.

Sallyann1234
23rd Mar 2018, 08:51
We know of two fatality accidents involving semi-autonomous cars.
The question I'd like answered is how many fatalities might have been saved over the same period of time if every manually driven car had been automated?

ORAC
23rd Mar 2018, 09:05
Sallyann1234, I refer you to my post #110.

“Both Ford and Waymo, Alphabet's self-driving-car company, have said they are developing fully self-driving cars because they fear people become too complacent in Level 2 systems.

Although Level 2 driving systems come with their risks, they are also much safer than cars on the road today. Crash rates for Tesla vehicles have fallen 40% since Autopilot was first installed in 2015.”......

cattletruck
23rd Mar 2018, 11:33
I can't bring myself to watch the fatality from the dashcam footage, but was the victim walking across the road with her bike positioned on the impact side? Could the rigid bike frame hitting her body rather than the crumple zoned designed vehicle have been a factor?

Many years ago I made several attempts to trigger a speed camera by riding my bicycle up and down at speed past the camera but with no success. I spoke to the law enforcement officer later and he seemed to think that the radar signal was getting confused by reflection from the spokes of the wheel.

G-CPTN
23rd Mar 2018, 11:43
When I first watched the video I was 'struck' by the lack of illumination of the cyclist as they approached the path of the car.
UK headlamps have (or had) a beam that illuminates the nearside kerb, but, of course, the cyclist approached from the offside.

MungoP
23rd Mar 2018, 12:19
officer later and he seemed to think that the radar signal was getting confused by reflection from the spokes of the wheel.

I'm off to get wire wheels fitted to my car..

VP959
23rd Mar 2018, 12:20
When I first watched the video I was 'struck' by the lack of illumination of the cyclist as they approached the path of the car.
UK headlamps have (or had) a beam that illuminates the nearside kerb, but, of course, the cyclist approached from the offside.

My experience is that dash cam footage often doesn't show as much headlight illumination as you see with your eyes. I suspect the dynamic range of these cameras may be lower than our eyes.

Cunliffe
23rd Mar 2018, 12:39
With regard to ABS I watched a police documentary some years ago which showed that ABS brakes do in fact leave skid marks. Instead of a continuous line they left a series of dashes. Estimating the speed of a vehicle from the marks was said to require a different technique from standard marks. Has ABS changed so that no skid marks are left?

G-CPTN
23rd Mar 2018, 12:50
he seemed to think that the radar signal was getting confused by reflection from the spokes of the wheel.

I'm sure that, when hand-held radar guns were introduced, there was a ruse involving aluminium foil (though I cannot recall the details).

The logic was that the centre of the wheel travelled at vehicle speed, but the bottom (in contact with the road) was always stationary and the top twice vehicle speed.

Blues&twos
23rd Mar 2018, 13:03
It's an absolute certainty that if a self-driving car operates safely and faultlessly pretty much all of the time, requiring no input at all from the 'driver', then the driver will very quickly cease concentrating.
This happens even when people are driving normal cars on long or familiar journeys!

Checkboard
23rd Mar 2018, 13:13
It just shows that the philosophy of these things is the wrong way around.

Humans are rubbish monitors - easily distracted - but clever manipulators.
Computers are excellent monitors - but difficult to get to manipulate.

So these cars - if they are not fully autonomous - should have the human driving, and the computer system monitoring with over-ride authority to slow/stop/swerve the vehicle.

G-CPTN
23rd Mar 2018, 13:20
the computer system monitoring with over-ride authority to slow/stop/swerve the vehicle.

but it didn't . . .

sitigeltfel
28th Apr 2018, 15:18
New technology creates a new breed of idiot.

A driver who moved into the passenger seat after putting his electric car into autopilot while at 40mph on a motorway has been banned from driving.

Bhavesh Patel, 39, of Alfreton Road, Nottingham, pleaded guilty to dangerous driving at St Albans Crown Court.

A witness in another car filmed him sitting in the passenger seat of his Tesla S 60 on the M1 between junctions 8 and 9 near Hemel Hempstead.

Patel said he was the "unlucky one who got caught", the court was told.


Hemel Hempstead autopilot Tesla seat switch driver banned - BBC News (http://www.bbc.com/news/uk-england-beds-bucks-herts-43934504)

glad rag
28th Apr 2018, 15:57
[QUOTE=ORAC;10092568]I think the jury is out on whether the vehicle brake does or not. A said, with ABS it would not have left any skid marks,

Point of order!

There WILL be marks on the road surface, just not skid marks.

these are visible under very close examination and are the result of the wheel/ABS system reaching the ABS trigger points. These are often highlighted by police accident investigators as a trail of dots with highlighting spray paint...

glad rag
28th Apr 2018, 16:00
New technology creates a new breed of idiot.

A driver who moved into the passenger seat after putting his electric car into autopilot while at 40mph on a motorway has been banned from driving.

Bhavesh Patel, 39, of Alfreton Road, Nottingham, pleaded guilty to dangerous driving at St Albans Crown Court.

A witness in another car filmed him sitting in the passenger seat of his Tesla S 60 on the M1 between junctions 8 and 9 near Hemel Hempstead.

Patel said he was the "unlucky one who got caught", the court was told.


Hemel Hempstead autopilot Tesla seat switch driver banned - BBC News (http://www.bbc.com/news/uk-england-beds-bucks-herts-43934504)


How does an idiot like that be in a position to buy a Tesla in the first place :confused:

treadigraph
28th Apr 2018, 17:18
How does an idiot like that be in a position to buy a Tesla in the first place :confused:

To be fair, plenty of other expensive cars such as BMWs and Mercedes are often driven by idiots too!

VP959
28th Apr 2018, 17:30
I think there is a fair bit of evidence (globally) that you can be an apparent idiot and still make a great deal of money!

Solar
28th Apr 2018, 20:40
As an aside to robotic cars I'm surprised that it is not now compulsorily for car manufacturer's to install decent all round vehicle camera coverage which in theory should reduce insurance premiums, not that that will happen but it may reduce the myriad of scams as well as reducing court costs. Mind you anything reducing the legal fees will be a hard sell but at least it would be a practical plus selling point for the vehicle makers.

Ogre
28th Apr 2018, 23:48
As an aside to robotic cars I'm surprised that it is not now compulsorily for car manufacturer's to install decent all round vehicle camera coverage .
I'm just to decide whether you are installing the cameras for evidence of an incident or to assist the drivers? If it's the latter then the probelm would be the driver spending all their time looking at the picture from the camera which would distract them from the road ahead. For the former, well I recently installed a dashcam which gives me a great view out the windscreen and even indicates g forces on the car. I recently had an incident where the bloke behind decided I wasn't going fast enough so he tried to speed me up by getting really close to my back end. This went on for quite a while before he had the chance to overtake at a junction, so I decided that I had justification to report it to the police.

I described the incident to the police, told them I had a dashcam, but when I went back and reviewed the footage all I could prove was that the incident started at this point because that's when I audibly swore at the other driver! My dashcam only shoots forwards, so in this case a 360 degree camera view would have proved something or other.

The only other things my dashcam shows are that I have a dubious taste in music and that I shout at the radio because it has an audio record function. I now have to be careful what I say in the car as I know it will be several hours before it gets overwritten....

Dan Gerous
29th Apr 2018, 10:54
Aside from seatbelts, I think we should be removing safety features from cars. To many people are becoming reliant on some of these, and not paying attention to the actual driving. The only thing I would add as a safety feature, is the oft mentioned 6" spike protruding from the centre of the steering wheel.

Blues&twos
29th Apr 2018, 11:12
The problem with the theory behind removing safety features from cars to make people drive more carefully, is that the 'non-fault' occupants also need protecting, as well as the non-fault occupants of other cars.
Drive as carefully as you like, if some idiot collides with you through no fault of your own, you and your passengers will be needing as many airbags, seatbelt tensioners and crumple zones as you can lay your hands on, so to speak.

​​​

ORAC
25th May 2018, 05:22
The question being who turned the braking system off - and how. (I can understand how revertin* to manual control would disable the entire system, but what would be the logic in allowing the system to be able to remain in control but be able to disable the braking subsystem?). I would imagine a manslaughter charge, or charges, will follow.

https://www.thetimes.co.uk/edition/news/self-driving-uber-in-fatal-crash-had-emergency-braking-off-wm52zhtqb

The emergency braking system of a self-driving Uber car that killed a woman (https://www.thetimes.co.uk/article/driverless-uber-car-chose-to-hit-woman-who-died-wthkf5vt6)in the US had been disabled “to reduce the potential for erratic vehicle behaviour”, according to a report published yesterday.

The car’s radar system had spotted Elaine Herzberg (https://www.thetimes.co.uk/article/uber-footage-shows-victim-of-self-driving-car-crash-emerging-from-shadows-in-tempe-arizona-gc6shvfss)about six seconds before she was struck in Arizona in March, the report by the US National Transportation Safety Board found.

A video of the crash showed the safety driver looking down just before the vehicle struck and killed Ms Herzberg, 49, whose death was the first of a pedestrian involving a self-driving vehicle. Sensors on the Volvo XC-90 SUV spotted her while the car was travelling at 43mph and determined that braking was needed 1.3 seconds before impact, according to the report.

A diagram showed that the system had determined that the car needed to brake when it was at least 20 metres from Ms Herzberg.

Blues&twos
25th May 2018, 06:16
Emergency braking system disabled “to reduce the potential for erratic vehicle behaviour”?!
It will intetesting to find out how erratic vehicle behaviour is defined in the context of an emergency stop.
Actually the apparent causal factor put me in mind of the Southall rail crash (UK)... safety system disabled, distracted driver.

​​​​​​

Pontius Navigator
25th May 2018, 06:19
We might guess that the erratic braking was the result of false positives. One or two might be tolerated, frequent braking not so. Why only the sub-system being switched off could be an unintended design feature. A feature where the car reverts to cruise control but with adaptive features off (as many cars have) and lane-assist (which some have). Couple that with the type of person whose manual reversion features are also off . . . .

MurphyWasRight
25th May 2018, 22:40
From the ntsb prelim report:https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf

At 1.3 seconds before impact, the self-driving system determined that an emergency braking maneuver was needed to mitigate a collision (see figure 2). 2 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

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.

They turn off the cars built in auto brake system, have the safety driver "responsible" for monitoring a screen that requires taking eyes off road and finally don't even raise an alarm if emergency braking is required. Glad I am not on Ubers legal team.

Gertrude the Wombat
26th May 2018, 09:24
Well, I wouldn't ride a bicycle that didn't have wheel reflectors, but that wouldn't have helped in this case if

the car knew it was a bicycle (accepted)
the car knew it needed to perform an emergency stop (accepted)
the emergency stop feature was switched off (accepted)
the human wasn't paying attention (disputed).

Now this last point seems to be disputed - the operator seems to be claiming that the "driver" was paying attention, contrary to earlier news reports. So a field day for the lawyers. Particularly if the human hadn't been briefed that the automatic safety systems had been switched off.

(The reason I wouldn't ride a bike without wheel reflectors is that I have driven a motorbike, and at night in the rain there are lights of all colours flashing at you from all directions, distorted by the raindrops on your visor, and the wheel reflectors are the only strong clue you get that there's a bicycle already on the roundabout you wish to enter. But perhaps they don't have roundabouts in this part of the USA, and/or perhaps the cyclist didn't have my motorcycling experience - it's not compulsory training for a cyclist, unlike cycling for bus drivers which is compulsory training round here.

In town I may or may not be wearing stealth clothing on the bike, as I rely on all the lights and reflectors for visibility, and motorists are expecting to see cyclists all over the place anyway and are looking out for them. In the unlit countryside where motorists might not come across a bicycle at night from one month to the next I'm much more likely to wear hi-ris.)

VP959
22nd Jun 2018, 10:17
Interesting to read the report in the media now that the NTSB have released more info: https://www.bbc.co.uk/news/technology-44574290

Seems like the safety driver was head-down, watching "The Voice" streaming on a device just before the fatal accident, The car systems seem to have detected the bicycle OK, but classified it as a bicycle rather than a pedestrian pushing a bike. It may be that this is relevant, as there may be an assumption in the self-driving algorithm that expects bikes to turn to go in the direction of traffic, rather than across it like a pedestrian might. Either way, it looks like the safety driver may face charges.

M.Mouse
22nd Jun 2018, 17:10
That is interesting.

Automation is fantastic but humans are very poor at monitoring for long periods. Until cars are fully autonomous, and in my view reaching that level is some years away, then I think a car driving automatically but needing constant monitoring by the driver is a recipe for disaster.

Blues&twos
22nd Jun 2018, 17:44
M.Mouse is absolutely right about expecting people to be monitoring for long periods, especially when for 99% of the time, they need to make no input.
You only have to look at the number of accidents which happen to drivers of ordinary cars where they very definitely have to pay attention and make inputs continuously to see this.

Pontius Navigator
22nd Jun 2018, 18:59
You only have to look at the number of accidents which happen to drivers of ordinary cars where they very definitely have to pay attention and make inputs continuously to see this.
A motorway is getting close to creating the same low arousal state as an autonomous car. Relative speeds in one lane might be only 1-2 feet per second, self-centreing steering keeping the car in lane with very light touch on steering ​wheel and no change from this steady state for minutes at a time.

.Scott
2nd Jul 2018, 12:01
From the ntsb prelim report:https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf

They turn off the cars built in auto brake system, have the safety driver "responsible" for monitoring a screen that requires taking eyes off road and finally don't even raise an alarm if emergency braking is required. Glad I am not on Ubers legal team.
from the NTSB report: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 am not comfortable with having the driver "monitor" and "tag" - even though they apparently made this a fairly simple task. The diagnostic messages are on the "center stack", but the driver would still have to look down and focus several times a minute. Tagging was probably just pressing a button - no worse or better than dialing a cell phone.

Also, it appears that that "monitor" and "tag" requirement was a factor in this incident. from the report:The inward-facing video shows the vehicle operator glancing down toward the center of the vehicle several times before the crash. In a postcrash interview with NTSB investigators, the vehicle operator stated that she had been monitoring the self-driving system interface.

That setup surprises me. When monitoring the data collection or taking notes is required, the norm would be to either do it while not driving or to have an engineer on board.

ORAC
7th Mar 2019, 06:25
https://www.thetimes.co.uk/article/uber-cleared-of-criminal-charges-after-self-driving-car-death-zrlzqcb7m

Uber cleared of criminal charges after self-driving car death

Uber has been cleared of any criminal charges after one of its self-driving cars hit and killed a woman in what is believed to be the first fatality involving a pedestrian and an autonomous vehicle. Prosecutors said it had not found any evidence to charge Uber for the fatal crash in Tempe, Arizona. Uber, best known for its taxi-hailing app, halted all its self driving tests after the incident.

In March last year, a semi-autonomous Volvo SUV modified by Uber was travelling at 40mph when it drove into Elaine Herzberg, 49, a homeless woman who had been wheeling her bicycle across a four-lane road near a junction. However, Sheila Polk, from the local attorney’s office, said that after a “very thorough review of all the evidence presented”, the office has determined there was “no basis for criminal liability for the Uber corporation”.

It is still to be decided whether Rafaela Vasquez, the safety driver in the car, will be charged. Local police said Ms Vasquez, 45, was watching an episode of The Voice on her phone during most of her shift, looking up just half a second before hitting Herzberg. The review suggests more evidence is needed before Ms Vasquez can be charged, adding a widely shared video of the crash “likely does not accurately depict the events that occurred”...........

meadowrun
7th Mar 2019, 10:09
http://www.youtube.com/watch?v=BjHpBFPz2xI

ve3id
7th Mar 2019, 10:32
Well, most humans with a smart phone in their hand appear to be automatons... Add a pair of headphones and they are completely remote from the world about them.
I saw traffic stopped at a light, police car speeding towards the intersection on the outside of traffic, in the opposite direction lane to pass the stopped traffic. Siren blaring and all lights flashing. As the police car came up to the intersection, woman with headphones on and playing with her cellphone stepped right out in front of the police car from behind the stopped cars. The police car stopped. But you cannot expect every driver to have advanced driver training as police do, nor can you expect that of a computer!

VP959
7th Mar 2019, 11:50
It's about time that greater action was taken against those who are so absorbed with music being played through headphones, or who choose to walk around with their heads down using their phone, and by doing so put their own life, and that of others, at risk. The Highway Code already has rules covering pedestrians crossing roads, and yet these rules are broken with impunity all the time, usually by people who are engrossed in listening or watching something like a phone.

meadowrun
7th Mar 2019, 12:50
who choose to walk around with their heads down using their phone

That bit I kind of enjoy.
I rarely change course for them when walking.

Blues&twos
7th Mar 2019, 12:57
Of course, as a pedestrian there is no obligation to have read the highway code or be aware of its advice to pedestrians.
There are no actual enforceable laws regarding wandering about using a phone or wearing headphones.
It's clearly not sensible to stroll into a busy road without being fully aware of what's going on around you, but some people are just like that - clueless. Probably the same people who would turn round whilst driving to chat to their children in the back seat.

VP959
7th Mar 2019, 15:15
Of course, as a pedestrian there is no obligation to have read the highway code or be aware of its advice to pedestrians.
There are no actual enforceable laws regarding wandering about using a phone or wearing headphones.
It's clearly not sensible to stroll into a busy road without being fully aware of what's going on around you, but some people are just like that - clueless. Probably the same people who would turn round whilst driving to chat to their children in the back seat.

I think we should have a jaywalking law here. First time I went to the US I was grabbed by the arm by a colleague as I just went to cross the road, who pointed out that I had to use the crosswalk. First time I'd ever encountered a place that had laws controlling how pedestrians behave.

Kiltrash
7th Mar 2019, 16:01
Unfortunately several years ago I was involved in a fatal RTC, The elderly female walking the same way I was driving just stepped off the pavement in front of me and despite hard breaking from 30mph and steering away the pedestrian head contacted my near side windscreen.
She died instanly
Went through the whole Police breath test, car check for roadworthy (company car)
Devastated and off work for 2 months,
At the coroners court it transpired she had stepped off the pavement before at the same place to cross the road to the local shop several times before and had been injured twice before needing hospital attention. Her doctor at the coroners court testified that both she and her daughter had been advised not to let her out on her own and the Police confirmed they knew she was a local concern
I was absolved of any blame, I was the unlucky sod...wrong place wrong time
Of course it is rightfully the drivers fault and I cannot see how the UBER safety driver will not be held to blame. Even though it was the pedestrian that collided with his car.
RIP

VP959
7th Mar 2019, 17:39
A bit over 10 years ago a chap out running (listening to music via earphones) ran across a dual carriageway just as I was was accelerating to overtake a slower car in the inside lane. He didn't seem to look, and hit the nearside front of my car, was thrown up on to the top of the windscreen and the force from his body caved the roof above the centre of the windscreen down so that it touched my head.

The chap was seriously injured, but thankfully there were a number of eye witnesses who all said that there wasn't a thing I could have done to prevent the accident. I was breath and drug tested, cautioned by the police and interviewed, but no charges were brought. A few months later, once the chap had partially recovered from his injuries, I received a claim from a no-win, no-fee claims vulture, with the instruction to pass it on to my insurer. I did this, but told my insurer that I'd be pursuing a counter claim against the pedestrian, as there was overwhelming evidence that he had acted negligently in causing the accident, and I'd suffered uninsured losses plus ongoing stress, sleepless nights, etc as a consequence of the accident (the sound of a human body hitting a car at speed is not something you ever forget).

I felt very sorry for the chap, but didn't see why my insurer should have to pay out even more money just because he'd been a negligent idiot. He'd already cost them around £8k in repairs to my car, and by rights I felt that my insurer should really pursue him for that money, not him try to get money out of them. He did drop the claim a while later, as I think the no-win, no-fee vultures probably thought they were unlikely to be able to profit from him/my insurer.

One consequence is that I feel pretty strongly that pedestrians should be made responsible, in law, for their own reckless acts.

M.Mouse
7th Mar 2019, 19:16
I have a Tesla Model S. It has an 'Autopilot'. Essentially if the road markings are good it will stay in lane, change lanes if I engage a turn signal, stop/start and slow down/accelerate at a uniform distance from the vehicle in front and so on.

I think it is dangerous. First of all a driver has the ultimate responsibility for the behaviour of the car. If it does something unexpected or erratic the driver is supposed to intervene. Having spent decades monitoring airliner autopilots, and as any pilot will tell you, monitoring in a state of constant alertness is impossible. In an aircraft the aeroplane systems will warn you if the autopilot disconnects or is degraded and any erratic behaviour is unlikely to lead to running into another adjacent aeroplane or crashing into a bridge! In my Tesla it will give warnings of various sorts but the time between warning and possible catastrophe will be small.

Secondly I find the Tesla cruise control side of the autopilot extremely random. It will, for instance, on accasion suddenly brake when it 'sees' a bridge over the motorway. Only momentarily but enought to make you jump.

I use the Autopilot in dense stop/start traffic on the M25, etc.It is extremely good at relieving the tedium of stop/start low speed driving. It also isn't going fast enough for any erratic behaviour to lead to a serious accident.

Tesla and others have done a pretty amazing job with self-driving cars but in my opinion we have a log way to go before we see fully autonomous and virtually foolproof driving ability.

ShotOne
7th Mar 2019, 20:02
+1 to that mmouse, not because the autopilot is inherently faulty but because of the unrealistic expectations of (some) operators. Creating a grey area as to who’s driving is asking for trouble.