Law Commissions open consultation into new rules for UK’s self-driving future

[kwackers]

Of course if these things are so superior perhaps we shouldn't have monkey brains riding bicycles either.

Why? How many people do cyclists kill?
What makes self driving superior is that it handles cyclists and pedestrians with ease and makes the correct allowances all the time without getting narked or knocking you off because "you're in the middle of the road".

I ride and drive a lot and rarely come across the problems you claim to suffer. Do I sense a certain amount paranoia in your use of the phrase " actively try and kill us" and a certain lack of logic too. If they were you'd already be dead.

I'm sure you don't.
How many miles do you do a year? How many in rush hour on the roads joining two 'city' centres together (and including said city centres)?
I do in excess of 10k and I've no shortage of video footage that shows event where had I not taken evasive action there'd have been a collision.

Lets take a look at what's happened on the roads - ignoring the 3 adults and a 1 year old killed by some monkeys in a car in Sheffield, locally this morning we have a 17 year old cyclist killed in a hit and run by a suspected drunk driver who then crashed a second time not long after, we also have an 8 year old girl also killed this morning with the driver arrested on suspicion of death by dangerous driving, last night we have a woman killed and a man seriously injured by a man arrested on suspicion of dangerous driving (and drink driving).
And that's just local!

There are 28,000 hit and runs reported each year.
Let that sink in. Twenty eight thousand...

So am I supposed to accept that level of carnage just because it's done by my fellow monkeys and ignore any possibility of reducing it?

[fastpedaller]

There are 28,000 hit and runs reported each year.
Let that sink in. Twenty eight thousand...

So am I supposed to accept that level of carnage just because it's done by my fellow monkeys and ignore any possibility of reducing it?

I absolutely agree - but will automation reduce that figure or increase it? No real evidence either way yet.

[kwackers]

I absolutely agree - but will automation reduce that figure or increase it? No real evidence either way yet.

Except that's not true.

There are a lot of self driving test cars running around at the moment, collectively they've covered many 10's of millions of miles.
We can also include all the Tesla's running around - although technically they're driver assist since the driver is meant to be in charge (but usually isn't), Tesla's alone have covered getting on for 2 billion miles in autopilot mode and they report back to base with the data they collect which is used to test and train new software.

Which brings me to another point about self driving cars, you can test them without actually having them on the road.
They have no way of knowing the data they're getting is simply recorded rather than real which makes it possible to see how software changes work in the real world without actually being in the real world.
That means that at any point in time tens of thousands of self driving cars in labs are driving around in simulations using real world data, testing 'difficult' scenarios as well as those actually driving around.

All of this is enough to know that so far these things aren't just safer but way safer and don't forget this is still early technology.
IMO mass roll of genuine self driving (no driver assistance required) is 5 - 10 years away by which time pretty much all the issues will have been sorted.

Finally there's another slant on this.
All over the world manually driven cars are testing the sensors and software that's being used for self driving cars. Anti collision, lane keeping, parking etc etc
That runs to millions of current cars all performing a subset of the technology required.

There's a lot of convergent technology coming into play which will ultimately give us self driving - and in the short term give us anti collision systems that should make even manual controlled cars a lot safer which given events locally to me over just one day it can't come quick enough.

[thirdcrank]

If we accept that human drivers are often unsafe, why is there the need to even consider modifying the rules to allow driverless vehicles to emulate human behaviour?

[kwackers]

If we accept that human drivers are often unsafe, why is there the need to even consider modifying the rules to allow driverless vehicles to emulate human behaviour?

I think you read far more into the OP than actually exists.

Nothing there says that will happen, it's a consultation which means you need a reason if that is the behaviour you want to emulate.

For example on a thread on this very forum about self driving cars there were a number of people who were of the opinion that a self driving car would need to be able to nudge pedestrians out of the way.

Now imo that's nonsense but the problem with consultations is they need to pay lip service to folk who think that to be true. But it doesn't mean we will get cars that nudge pedestrians, far from it. But at least to those who think that they'll feel that they're being listened to.

OTOH, back to the motorway thing. Why not allow one self driving car to tailgate another? That makes perfect sense but would require changes in the law.

At some point the law commission will publish their report and we can come back and discuss facts rather than speculate on imagined worst case scenarios.
And if you're worried the worst case scenarios might come true then I believe there's a consultation that you can contribute to.

[thirdcrank]

Why is there any need to consider these changes?

[fastpedaller]

I absolutely agree - but will automation reduce that figure or increase it? No real evidence either way yet.

Except that's not true.

There are a lot of self driving test cars running around at the moment,

But still the numbers must be a very small percentage of the traffic - any a lot are being driven by 'test drivers' which may or may not be representative of real-life. I can agree there will be some instances where the automation may be beneficial, but there must be many where human any=ticipation cannot be matched by a machine.

[pete75]

Of course if these things are so superior perhaps we shouldn't have monkey brains riding bicycles either.

Why? How many people do cyclists kill?
What makes self driving superior is that it handles cyclists and pedestrians with ease and makes the correct allowances all the time without getting narked or knocking you off because "you're in the middle of the road".

I ride and drive a lot and rarely come across the problems you claim to suffer. Do I sense a certain amount paranoia in your use of the phrase " actively try and kill us" and a certain lack of logic too. If they were you'd already be dead.

I'm sure you don't.
How many miles do you do a year? How many in rush hour on the roads joining two 'city' centres together (and including said city centres)?
I do in excess of 10k and I've no shortage of video footage that shows event where had I not taken evasive action there'd have been a collision.

I drive over 20,000 miles in an average year on all sorts of roads and at most times of day. PLenty of miles on motorbikes and bicycles too. Never felt the sort of paranoia that'd lead to installing a so called dashcam though.

[rfryer]

Why is there any need to consider these changes?

There will be driving scenarios where taking an action that isn't strictly legal is reasonable. For example, getting out of the way of emergency vehicles, escaping from a threatening situation, or avoiding a serious accident.

For autonomous vehicles, those kinds of behaviours need to be thought through and programmed in advance. I'd prefer that autonomous vehicles were programmed to behave within the law, which means that the law needs to cover situations where such "infractions" might be allowed, or even encouraged.

Put another way, an individual human driver can make a decision to serve the greater good by breaking the law, with some confidence that the action could be justified if necessary.

However, I'd hope that a car manufacturer would be less enthusiastic to release a car that was deliberately programmed to behave illegally; they would be opening themselves up to some extremely complex and expensive litigation should that deliberate decision lead to a death or serious injury.

Hence the need to agree in advance what behaviours are acceptable, even those that are not currently enshrined in law.

Thinking about this, I do find the "threatening situation" side of things troubling. If cars were programmed to always be "good", then it would be easy to terrorise their occupants simply by standing in the way of the car. This might range from gangs of yoofs "tagging" cars coming through their patch, to robbery, carjacking and worse. Is the solution to program the car to drive over the aggressor (which, frankly, is what I might choose to do in extremis), or do we just accept this as a limitation and maybe program the cars to drive around in groups when transiting dodgy areas at quiet times?

[kwackers]

Why is there any need to consider these changes?

rfryer above answers this well.

The problem with self driving cars is they obey the rules - always.
Doesn't matter what the emergency vehicle trying to make it's way through traffic does the car will sit obediently at the lights until they change and it has a clear path through.
Ditto the pavements thing.

There are hundreds of these edge cases and unless we program the car to make some allowances then we have the problem of people whining about how "stupid" the cars are and that a "human driver wouldn't have done that".
In that respect the cars can't win.

Simply saying "why is there any need" is imo a very odd statement that suggests an emotional rather than logical approach to how these vehicles operate.
Even if nothing changes and the cars stupidly stick to the rules no matter what then it would be remiss of us not to consider the big picture. (And some of that includes things like deciding culpability for the cars actions).

[thirdcrank]

To take the example of getting out of the way of emergency vehicles, there will have to be some way of the emergency vehicle identifying itself and then indicating it needs priority. Presumably that would be some sort of electronic signal, rather than flashing lights and going woo-woo. A driverless vehicle could presumably be programmed to respond to that signal only.

If general exceptions are created to cater for occasional circumstances then the exception will become the general rule.

What I'm failing to express is a fear that the law is going to be amended to cater for expediency rather than safety. The results of any consultation will be shaped by the form of the consultation. This consultation seems to be leading to expediency rather than safety.

=========================================================================

PS The consultation appears to me as though the manufacturers have presented the government with a shopping list of changes they want if they are going to develop this technology in the UK.

[kwackers]

I drive over 20,000 miles in an average year on all sorts of roads and at most times of day. PLenty of miles on motorbikes and bicycles too. Never felt the sort of paranoia that'd lead to installing a so called dashcam though.

And yet time and time again on these very forums we discuss examples of miscarriages of justice because the dead cyclist can't bear witness to how they died.

I consider my cameras as an insurance policy which should I be an unfortunate victim then at least whilst I can feel my life ebbing away in the gutter I can hope justice will be served and my family will be recompensed by the guilty.

I refer to an earlier post:
28,000. That's twenty eight thousand cases of hit and run per year in this country alone.

[Cyril Haearn]

But the camera is tiny, you need a big notice:
CAMERA IN USE!

28 000 is a gross underestimate of course

[thirdcrank]

I still have difficulty with the idea that human drivers are terrible so when we introduce something that has the potential to be better we try to ensure it can emulate human drivers.

[pete75]

To take the example of getting out of the way of emergency vehicles, there will have to be some way of the emergency vehicle identifying itself and then indicating it needs priority. Presumably that would be some sort of electronic signal, rather than flashing lights and going woo-woo. A driverless vehicle could presumably be programmed to respond to that signal only.

If general exceptions are created to cater for occasional circumstances then the exception will become the general rule.

What I'm failing to express is a fear that the law is going to be amended to cater for expediency rather than safety. The results of any consultation will be shaped by the form of the consultation. This consultation seems to be leading to expediency rather than safety.

=========================================================================

PS The consultation appears to me as though the manufacturers have presented the government with a shopping list of changes they want if they are going to develop this technology in the UK.

Yes - in order that these things don't commit a "crime" you alter the law to make what they do legal. A novel approach to crime reduction indeed.

Who is going to develop them here anyway? All teh car manufacturing plants here are owned by German, Japanese, French and Indian car companies. Most of these will do or, more likely, have done their development in their own countries. The vehicle systems would just need some fine tuning to make them fit the road systems of the UK. The one that does do development work here is Jaguar Land Rover but it looks like they're going down the pan anyway with financial losses, plant shutdowns and three day weeks etc. Sounds like the latter days of BL.