Will driverless cars save lives? (more than 130 collisions)
Discussion
bigothunter said:
Dave Finney said:
But safety is the primary criteria.
Authority frequently sets-up smokescreens to hide their genuine objectives. Safety is a favourite. Are you convinced safety really is the primary objective? I'm not.I meant that safety is the primary criteria for me to get a driverless car.
Dave Finney said:
Well no, not for the authorities, I agree.
I meant that safety is the primary criteria for me to get a driverless car.
Some way to go before driverless cars qualify as safer than the average UK human driver.I meant that safety is the primary criteria for me to get a driverless car.
Dave Finney said:
We are responsible, sober drivers, who always wear seat belts. So we need to remove from the human road deaths those that involved: Drink drive, no seatbelt, reckless, stolen etc.
And motorcycle deaths need to be removed too.
You mentioned motorcycle deaths previously (death rate 40 times higher than cars). How do you propose solving that problem? Should motorcycles be banned?And motorcycle deaths need to be removed too.
bigothunter said:
You mentioned motorcycle deaths previously (death rate 40 times higher than cars). How do you propose solving that problem? Should motorcycles be banned?
By finding out when, where and why they occur,eg it may be that the increase in fatal collisions at speed camera sites primarily involves motorcyclists.
and No.
I feel less safe knowing that driverless cars are likely to be out and about. Someone may not have written the algorithm for a previously unforeseen set of circumstances.
My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
ATG said:
FMOB said:
ATG said:
The idea that you can't write stable, predictable software is bilge. There's tons of it around keeping machines running. Aircraft are a pretty good example of what can be achieved. They are properly complicated systems running huge amounts of software and they are highly fault tolerant.
.
The key here is fault tolerant, the fact there are extra bits to make it fault tolerant is accepting the software isn't reliable and faults have to be mitigated with backup systems, etc..
Reliable software is a fallacy.
Either it works as intended or it doesn't.
People who say Software is not reliable are actually confusing, the fact that it has intrinsic failure modes, where a certain set of inputs, at an unplanned level have not been catered for. Every eventuality not identified, programmed and or tested.
If you like. Its A design fault.
The problem is that the inputs involved now include inputs from radar, sonar, and camera, and there must be an infinite variability in what they see.
So how do you test every possible outcome ?
Miles. And more miles, and user feedback. Looks like a reliability issue, because eventually it will fall into an error state. But it's not.
rdjohn said:
I feel less safe knowing that driverless cars are likely to be out and about. Someone may not have written the algorithm for a previously unforeseen set of circumstances.
My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
What the eff...My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
NFT said:
rdjohn said:
I feel less safe knowing that driverless cars are likely to be out and about. Someone may not have written the algorithm for a previously unforeseen set of circumstances.
My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
What the eff...My nearest close encounter to date was while riding my bike. Being slow, I was hogging the edge of carriageway marking when an Audi 8 passed me without considering my needs. It kept absolutely to the centre of its narrow lane. Simple lane guidance without hazard detection. The Highway Code now suggest you should ride in the centre of the lane, in order to dominate the space
If I ruled the world. No driverless vehicle would be authorised for the Highway until they can successfully demonstrate that they can use and reuse every approach at roundabout on the Arc de Triumph in Paris, all day, without the need to request a human takes over.
I guess that will keep them off the road for this century.
Because there are no lines and no lanes everything is very carefully moving so as to ensure they don't crash, it seems to.work.okay.because clearly if you crash its your own fault. Not fast but safe.
Self driving cars will talk to each other and agree protocol on which goes first. Probably that's one method of raising revenue. Pay extra for a priority car.
Pit Pony said:
Looks like the crossing about 100 yards up the road, is regulating the flow of traffic coming on to the roundabout.
Because there are no lines and no lanes everything is very carefully moving so as to ensure they don't crash, it seems to.work.okay.because clearly if you crash its your own fault. Not fast but safe.
Self driving cars will talk to each other and agree protocol on which goes first. Probably that's one method of raising revenue. Pay extra for a priority car.
That theory may work fine when 100% of vehicles are autonomous, and protocols have been agreed at an international level. My guess is that take-up will be even slower than for EVs and so designers are obliged to deal with the world as it is, rather than some utopian future. Each car manufacturer tends to be going their own way with software design and procurement. The Internet of everything also runs the risk of being “everything out of date”. Roadworks suddenly appearing etc.Because there are no lines and no lanes everything is very carefully moving so as to ensure they don't crash, it seems to.work.okay.because clearly if you crash its your own fault. Not fast but safe.
Self driving cars will talk to each other and agree protocol on which goes first. Probably that's one method of raising revenue. Pay extra for a priority car.
Most of the 130 crashes referred to by the OP have taken place in the US, where roads tend to be easy-peasy straight with either signal control or graded interchanges. Roundabouts in Europe where gap acceptance is all important, especially in dense traffic, is where many collisions are likely to occur. This is a massive issue that needs to be resolved. Writing protocols that mimic human behaviour will be tricky. Humans are pretty handy at assessing collision points within moving traffic and calculating how to give-way at the entry point to be in the right place at the right moment having a good sense of acceleration and brake performance of the vehicle..
In the Arc de Triumph situation, the protocol will likely be, cannot compute; driver must take over.
Not sure if correct place for this but there is an autonomous cars Bill going through Parliament ...
https://www.pinsentmasons.com/en-gb/out-law/news/a... - english&utm_content=your weekly round-up - rest of the world(17)
"The Automated Vehicles Bill is designed as framework legislation, aimed at establishing high level principles and powers for the UK government to make regulations in future to set out the details of regulatory requirements for self-driving cars before they can be used. among other things, it provides for a new authorisation regime under which self-driving vehicles would have to pass a “self-driving test” and satisfy a range of safety standards before they could be used on UK roads."
"Entirely new concepts that the Law Commission recommended be established are provided for in the Bill. This includes the concepts of an ‘Authorised Self-Driving Entity’, which is an entity that is responsible for vehicles when in an autonomous mode; a ‘user-in-charge’, which is an individual in charge of a vehicle in autonomous mode who is in position to control it but is not controlling it; and ‘transition demands’ – a timed demand, communicated by equipment of a vehicle, requiring the user-in-charge to assume control of the vehicle."
TX.
https://www.pinsentmasons.com/en-gb/out-law/news/a... - english&utm_content=your weekly round-up - rest of the world(17)
"The Automated Vehicles Bill is designed as framework legislation, aimed at establishing high level principles and powers for the UK government to make regulations in future to set out the details of regulatory requirements for self-driving cars before they can be used. among other things, it provides for a new authorisation regime under which self-driving vehicles would have to pass a “self-driving test” and satisfy a range of safety standards before they could be used on UK roads."
"Entirely new concepts that the Law Commission recommended be established are provided for in the Bill. This includes the concepts of an ‘Authorised Self-Driving Entity’, which is an entity that is responsible for vehicles when in an autonomous mode; a ‘user-in-charge’, which is an individual in charge of a vehicle in autonomous mode who is in position to control it but is not controlling it; and ‘transition demands’ – a timed demand, communicated by equipment of a vehicle, requiring the user-in-charge to assume control of the vehicle."
TX.
Tiglon said:
Driverless cars have been in a few collisions? They must surely be banned for all eternity so that vehicles can continue to be safely and flawlessly operated by humans who, famously, never make errors.
Driverless cars are bound to cause a few deaths until the bugs are ironed out. But the prize justifies that sacrifice. Who's going next? 
rdjohn said:
Someone may not have written the algorithm for a previously unforeseen set of circumstances.
That isn't generally how these sorts of systems work or are built. E.g. you don't have someone write a specification saying something like "when a motorbike is to your left, keep to the right hand side of your lane to give them more room" or "when you head into Paris start praying for an EMP to put yourself out of your misery". If you're trying to control a simple, closed system, e.g. controlling a washing machine or a machine tool, then that kind of "if A do B" programming can work fine, but in complex systems like driving through a town, it's going to be hopeless. What you do instead is teach the system with examples, very much mimicking how people learn stuff. You feed its sensors lots of example data and then tell it that "that's the set of signals that means "road" and that's the set of signals that means "vehicle", etc, etc". The system learns what a road looks like in a general sense, or what a vehicle looks like in a general sense. It can then start recognising vehicles and sections of road that it has never seen before as being vehicles, or being roads. And you take a similar approach with teaching it how to negotiate a road, e.g. loads of examples plus feedback about whether the outcome was good or bad. The idea is that the system ends up knowing/understanding things like "how to negotiate a roundabout" and "not cutting up other road users".
rdjohn said:
Someone may not have written the algorithm for a previously unforeseen set of circumstances.
This a common issue I've found with 'driver aids', in my case particularly because I drive on a lot of gravel roads and driver aids often don't seem to be programmed for those conditions.For example I've experienced a number of cars (worse with older ones to be fair) where the ABS got horrible confused by gravel (especially if the surface was corrugated) and refused to allow enough brake force to slow the car no matter how hard I braked, leaving me resorting to the handbrake. You know the ABS programming is bad when there is enough grip to stop the car with the handbrake but ABS won't let you slow down with all 4 wheels braking...
My current 10 year old work Hilux does ok for ABS on varied surfaces, but it has horrifically bad traction/stability control - clumsy and oversensitive, reacting to the slightest non-issue amount of slip by cutting power, grabbing brakes and generally lurching the thing around. I've had it suddenly cut power at 90km/h on a straight road because one rear wheel (poxy open diff) obviously had a tiny amount of slip as it went through a puddle or muddy patch.
Its safer and far more predictable for a competent driver to turn it off for gravel roads, but you have to be stopped and hold the button down for about 8 seconds to turn it off!
The previous one with an LSD and no traction or stability control was a much better and more predictable drive.I know driver aid systems have generally got better over the years, but the ones I have experience of are often still more of an annoyance than a benefit, which doesn't inspire me to want even more interfering with my control of a vehicle.
FMOB said:
PF62 said:
FMOB said:
Maybe acknowledge to yourself you are using dodgy statistics rather than quoting as gospel something that has more potholes in it than your average bit of tarmac.
And yet you have absolutely no statistics whatsoever, dodgy or not, to back up your views, so perhaps you might acknowledge that turning off LKA every time you get in the car isn't a sensible thing to do.I now have even more motivation to keep turning the sodding thing off because it seems to annoy you!
The day in 1985 that I spun my dad's 1.6 Sierra estate backwards over a central reservation, after pressing the accelerator too hard as I left a wet roundabout. Lucky there was no damage.
No traction control (or even rear seat belts or head rests it was so basic)
And the day in 2017, when I came off a wet roundabout too fast and pressed the accelerator too much in a 2001 MV6 omega. It's very crude traction control, went, "No you don't" and did something at the rear, which stopped me going sideways at 60 mph. I felt a little "twinge" a kind of split second deceleration from one rear wheel, and that was that. No drama.
GravelBen said:
Its safer and far more predictable for a competent driver to turn it off for gravel roads, but you have to be stopped and hold the button down for about 8 seconds to turn it off! The previous one with an LSD and no traction or stability control was a much better and more predictable drive.
I know driver aid systems have generally got better over the years, but the ones I have experience of are often still more of an annoyance than a benefit, which doesn't inspire me to want even more interfering with my control of a vehicle.
Driver aids need to cover the whole spectrum of drivers not just the competent ones. Consistent with the dumbing down policy across motoring. The previous one with an LSD and no traction or stability control was a much better and more predictable drive.I know driver aid systems have generally got better over the years, but the ones I have experience of are often still more of an annoyance than a benefit, which doesn't inspire me to want even more interfering with my control of a vehicle.
Surprised we are still allowed to turn these 'safety systems' off. Can't imagine that freedom continuing indefinitely.
Gassing Station | Speed, Plod & the Law | Top of Page | What's New | My Stuff