UKC

Brit transport pundit Christian Wolmar on why the driverless car is on a 'road to nowhere'  https://www.theregister.co.uk/2018/01/11/christian_wolmar_interview_selfdri...

"If self-driving cars are to be considered at all safe, then they will be programmed to come to a halt as pedestrians enter the road space. In many places, however, such as London's Holborn Underground station at rush hour, pedestrians are typically spilling over into the road en masse."

But but but...

To me, the logical thing to do is:

1. Ban all cars in London (zone 1, at least). Dedicate the less-busy roads to busses and bicycles.
2. Use self-driving cars outside of London where the "Holborn problem" doesn't exist and collaborating, emotionless, self-driving vehicles can achieve much better efficiency than nasty human drivers by reducing queues and wotnot.

So, in no way, do I believe that this problem presents a sufficient reason to discredit the self-driving car.

What do you think?

I think it’s a stupid argument.

If a person can drive a car in XYZ circumstances, it’s only a matter of time before a machine is sufficiently advanced to also drive a car in XYZ circumstances.

Pedestrians spilling out of a Kings Cross en mass doesn’t seem to stop thousands of drivers and explode their minds on a daily basis.

The only real points of debate in my mind are (1) should we permitt fully driverless in mixed-mode conditions and (2) when will it be possible?  For what it’s worth I think (1) emphatically yes and (2) it’s already happening in one part of the USA for real.  I think within a decade it’ll be old hat, and a lot of pundits will be left looking as stupid and blinkered as ever.

Banning cars in London will come long before driverless cars are ready.

Some level of driverless cars will be commonplace soon, just imagine how smoothly motorways could run if all vehicles on them were automated and communicated with each other.

One of the biggest issues in a mixed user environment (cars, pedestrians, cyclists, etc) are the 'moral' decisions that the driverless car will have to make, ie if an accident is unavoidable what is the most acceptable accident to have - is it ok to kill the driver if it will save a child, does a horse come above a dog, etc etc. We don't consciously think about these moral decisions when driving but driverless cars have to be given rules to follow so some decisions have to made on what the moral rules should be. Added complexity is caused by different cultures having different views on some things, e.g. is a cow worth more or less than a horse.

I think Christian underestimates the depth of human endeavour to solve problems.

And his understanding of machine learning, artificial intelligence and the rate of progress in computing power is probably not his strong point. No doubt they will find a way.

Self driving cars don't need to be perfect.  They just need to have less accidents than human drivers.

I think they will need to learn to drive like humans. Otherwise there will be nationwide gridlock.

It will be interesting to see how the legal issues around such questions as who is responsible in the event of an accident are resolved. 

Edit: and just remembered, there's the software control and security issue.

https://www.theguardian.com/technology/2015/dec/23/the-problem-with-self-dr...

Agreed. But, define a matter of time.

Using the current technology – No chance. Yes, currently it’s a proof of concept, but there is still a long way to go before the technology is sufficiently advanced before it becomes a common-place reality.

AI hasn’t really advanced since the late 80’s, despite the hype (OK it’s faster and has more memory but is still basically stupid). Given a quantum leap in machine design, then it’s a possibility, but using the current technology, not a hope in hell.

You can make a nice list of cities where they are already being tested. 

AI may not have advanced significantly, if you're looking for a silver bullet thinking machine, but cpu power and code development has made the implementation of machine learning increase like heck

I sort of agree and sort of disagree.  Artificial neural networks are just a faster/more evolution of those of the 1980s.  Recurrency is really just a temporal folding up of a large multi layer network with staggered inputs (and by extension the equivalent 3-layer network) and convolutional layers are just a particularly specific hidden layer.  But, the combination of recurrence, convolutional layers and bigger hardware really has moved ANNs from “toy” to “super capable”.  

AI doesn’t have to be that smart to enable a self driven car - it’s mainly used for categorising what is in images/videos etc, and for that it needs to be highly capable but not “smart”.  The smarts are less AI, more codified rules on what to do based on what the AI processed image sensors and other sensors see and what the physics models are predicting etc.

The key point is even with just a linear progression of “more” (memory, GHz etc) the technology would recently have gone from “not good enough” to “good enough,”, but because it’s done that, people are now making custom silicon optimised to do more of it, meaning there’s a significant boost in price/performance and performance/watt looming in our very near future.  All this in turn unlocks the third gain - improved understanding towards the optimal ANN architecture and training scheme for a given problem as the new chips significantly speed up R&D.

I think his entire argument can be summed up in the first part of this quote.

“So billions are being spent developing technology that nobody has asked for, that will not be practical, and that will have many damaging effects.”

It's the "nobody has asked for bit". It seems that for most tech nerds ( the sort of people who read The Register) their car is part of their personality, and they can't imagine life without this. They don't believe that there is any demand, because driving is an important part of their life. For them, driverless cars are unimaginable and unwanted, and they can't grasp the fact that this feeling is not universal.

I'm a techy nerd myself BTW, but I'm saving up already for the first true driverless car, which I believe will be here within 10 years. I'd gladly never drive again.

The people who can’t imagine driverless is for them perhaps haven’t had to care for an aging parent who becomes medically unable to drive.

Autonomous cars could change my future almost completely - we live in a rural location with no public transport within 4 miles.  If we both become unable to drive  30-60 years from now (I know people driving/not able to drive across that age range), we’d realistically have to move to a town and leave our orchard and memories behind.  Autonomous cars - the problem evaporates.  

Lol, that’s exactly my point. Self driving – no problem. Unanticipated/Unrecognised conditions – WTF???
Did you see Guy Martin and the AI race car? Nowhere near what he could do (after learning the setup). When it tried it spun off. Now how the hell did that happen? Surely it was fitted with sensors to detect traction and accelerometers, but it still ended up spinning off the track.
AI – lots of hype, little intelligence.
Neural networks – great at recognising conditions it have been explicitly trained for, but as for extrapolating, self learning and error back propagation – No Chance.
Genetic Algorithm’s – Maybe, but still lots of work to do and greater processing power required for real time problems.
Another solution – TBD. But there has to be quantum leap in what we have now before we have really ‘intelligent’ machines.

Problem already solved. Call a Taxi

I was going to suggest you were joking but then I remembered your contribution to a past thread on vehicles.

Do you seriously not understand how different it is between relying on taxis when living in the middle of nowhere and having your own autonomous car?

I don’t think you’ve really thought about it.  At all.

Each unanticipated solution only has to be solved once, and then it’s known what to do.  The best solution I’ve seen mooted is that in the rare case (once per 4000 miles and rising in Google’s testing) such a situation occours, is that the car connects with a remote human who figures it out.

Sure, AI isn’t a very technically correct name for most of what’s out there, but that doesn’t change what it can do.

You say that but published research disagrees - recurrent networks are now teaching themselves to play games and win games with no help - precisely self learning.  As in - the network is not trained to win a game, the network is trained so that when live, you can give it one of many different (unseen during training) games, and it will self learn how to play and win that game.

I don’t know why you even mention GAs; they’re one of a vast class of optimisation algorithms, and are really nothing to do with self driving.  They’re got a more biomimetic name name the simplex or stochastic gradient descent but they’re a lot like those.

I totally agree about “intelligent” and have said as much on previous threads, but I don’t think that kind of intelligence is needed for widespread autonomous vehicles.

What's your take on the Guy Martin comparison?

Are there plans for self driving motorcycles?

What about buses? More of an issue of passenger security from other bus passengers than the bus actually driving itself.

I don't understand why pedestrians spilling onto the road en-masse is viewed as an unsolvable problem for a self driving car.   Firstly the car's sensors will easily detect pedestrians and it will respond by reducing speed or stopping.  If the condition persists and progress along the planned route is impossible then after a predetermined period the vehicle will issue a warning to disperse and engage its anti-personnel weapons systems.

I saw Guy Martin.

His first foray into the self driving car involved him converting his van's manual gearbox so a robot with arms could pull the lever and depress the clutch.  Later on in the program when assessing impact avoidance he chose to save a dog over a human family.

OK if you like Bear Grylls type TV but not for me.

And, to agree with you, a boy racer could buy a tyre temperature kit for a couple of hundred quid so I'm puzzled why the state of the art car spun off because its tyres had cooled down, unknown to its handlers.

Strongly disagree with this. 

Most grid lock I see every day is caused by humans.  The yellow grid / keep clear sign doesn't apply to me, so what if I block all the traffic as long as it gets me where I am going a a few seconds faster. 

The utopia (unless you enjoy driving) will be 100% driverless and everything will be smooth and fast. 

No need for traffic lights, the cars will talk to each other and decide the priority. No more stuck at a junction waiting to get out, again the cars will talk the the ones on the main road will slow down enough for the joining car to slip in to the flow.

Most traffic disruption is caused by a combination of human selfishness and human error. 

Even in an urban setting it could be a life saver.

My grandpa should never have been in control of a car in his later years, but after his stroke he was not open to reason and logic. He wouldn't give up driving and he fell out with the whole family when we tried to convince him to. A self driving car would have been excellent for him, as he wouldn't have felt he was giving up his car.

He could easily have paid for taxis, in fact he could easily have employed a driver(!!) but that's not something his generation did and he was very stuck in his ways.

If it it was a typical Guy Martin show then Bear Grylls is an exact analogy and about as relevant to the real world as Desparate Dan 

Retaining personal choice is vital.

For me one of the biggest questions is will your "driverless" car have the ability to read and react to all the subtle nuances of human behaviour that human drivers are constantly using as they drive. 

Surely family and friends already fulfill that role whilst also providing some vital human interaction?

Potholes will be less of an issue when we get over our silly obsession with alloy wheels and low profile tyres.

Couldn't agree more about the silly tyres, but a self driving car should be better at spotting potholes than a human. It should be possible to have a sensor to look through the puddle and see how deep it is for example.

Well I would rather have spent my time in the garden or house with the relative paying attention to them than sat in a car paying to the road and driving home just how far their independence has eroded, being well aware how much this upsets them.

As with the suggestion of taxis above, the f&f route also has a lot of problems autonomous cars won’t.  I’ve given one example above common to both suggestions, there are plenty more.

I'm not sure it is that helpful.

Various surveys have shown that something like 80% of drivers believe they are of above average ability. That's a lot of people who seriously overestimate their decision making ability. Given personal choice in a driving situation, a large number of people will make a worse decision than an AI.

I haven’t had a chance to watch it yet...  

Self driving cars will never have the judgement needed to pass the risk assessment to make them safe on roads. It's not what they're for though, it's a trial of AI for the railways and trams that doesn't upset the unions.

I think that's right- 100% driverless is the way to go. Real issues arise when mixing human driven cars with autonomous ones precisely because of the human element. Remove that, have all cars linked to every other car and controlled centrally by supercomputers and we would have the dream. Car ownership as a concept would be consigned to history- you want to go somehwere then call up the next vehicle on your phone, there should be one along in 30 seconds.

Not very good for getaway drivers admittedly.

That being the dream, when are we going to have it?

They can also consult and update a live database of potholes.  Get it right and at worst, one car should drive through a particular pothole at speed ever.  Most of the time sensors will catalogue them in advance.

I suspect that it is quite possible that 80% above average, some of those at the bottom end drag standards down an awfully long way

We aren't incorporating that sort of telemetry into today's cars why should we assume that is an advantage of autonomous cars?

What other choices are you happy to lose?4

I agree with all comments above saying that self driving cars will be on our roads in the near future, but I don't believe it will lead to a utopian world with no accidents and fewer traffic jams.

The pedestrian issue is an interesting one, but actually comes down to human behaviour as much as it does the intelligence of the vehicle.  At present people don't step off the kerb in front of a moving car for 2 main reasons 

1.  The driver may not be paying attention, resulting in injury to the pedestrian
2. There is a social element, it's unreasonable to force a driver to slam on the brakes so you can cross the road, and people recognise that the smooth progress of everyone relies on following these conventions.

A move to driverless cars completely changes these reckonings, the car will always be paying attention, and you're upsetting a machine not a human.  And as machines can't actually get upset, the controlling emotions in the scenario come entirely from the pedestrian.

Was also reading an article saying that current commonly used sensors (Lidar) don't work in snow, and are hampered by rain, autonomous vehicle researchers in Russia are trying to develop AI systems that run purely from video images captured by cameras around the car, but this is a much more involved AI project than the one currently being tackled by most Western researchers.

I'm sure they're coming, and in the long run I'm also sure they'll be safer and reduce traffic hold ups.  Personally I don't think I'll ever buy one, I suspect the common model will be leasing or shared ownership with people calling them like taxis as and when needed.  Current cars spend the vast majority of their life stationary, not a very effective way to use an expensive asset.

An upshot of this is that there will be far fewer cars parked on our streets, and houses with parking will command less of a premium.  Of course garages will still be useful for storing bikes and outdoor kit...

1) Because “living” databases of high resolution mapping of areas are already being built in advance as part of the r&d towards self driving cars.  Maps down to features a few cm in size.

2) Because self driving cars contain all the necessary sensors for this sort of thing, current cars don’t.

3) Because the conatant updating of “living maps” by autonomous vehicles is an active area of research.  It’s not just potholes, it’s roadworks, broken down vehicles, XYZ other problems.

4) Because it’s a relatively trivial addition to future autonomous cars compared to cost of pothole damage to fleet operators, and I think there will be a lot of fleet owned autonomous stuff.

I'm happy to lose choices where I believe that a machine or another human will make better decisions than me. 

The point is that I'm realistic about my abilities, unlike a lot of keen car drivers.

The best autonomous vehicles can now drive a few thousand of miles on real roads without getting confused* compared to the first DARPA challenge (2004) when the best car did 7 miles. 

We've not improved by a factor of 500 since 2004 and if the autonomous cars can improve by a further factor of ten or a hundred the failure rate gets down to once every few years or once in a lifetime of human driving. Then it starts to sound very realistic.

*they come to a stop or a human takes over, they don't spin off the bend destroying an orphanage like we do. 

The weakness is not death and destruction, the major weakness is what to do when it fails safely. Maybe the AA will come out to rescue, reboot or tow your confused vehicle every few years.

For commercial usage, paying 1 person to do rescues for a fleet of 100 vehicles eliminates the wages of 99 drivers so there is a huge incentive there.

You think people will own autonomous vehicles? Paris has already talked of banning it.

Edit: for fake news

Can you imagine the improvement for Scottish Winter Weekends! Instead of arriving at the north face car park at 2am, frazzled  after van 8hr drive you could have snacked and snoozed ready for a big day on the Ben. Then, on Sunday night, shattered from a mixed battle and a harrowing walk off in a blizzard, you collapse in the back of the motor, tap "home" into the nav system and curl up on the back seat.

Technically possible I'm sure, but that's not the same thing as being desirable. Both politicians and manufacturers appear to be so entranced with the possibilities of the technology they haven't paused to think about whether any of us would actually want to buy one. Of course if we don't, Big Business will be lobbying Big Government to ban driving so we have no alternative, and what Big Business wants Big Government usually provides.

Considering the number of upgrades that computers require, the fact that aircraft, which are much more expensive than cars require 3 computer systems (for backup in case of problems) and are also maintained at far greater frequency (by a full team), and the problems caused by hackers - both those doing it for fun and those with malicious intent - the likelihood that we could both keep traffic moving and not have some major incidents is a little unlikely.

I would also like to point out that many people and most businesses do not have off-road personal parking - so going "home" or to a business can mean parking anywhere nearby, which means some form of intelligence to make decisions, particularly with regard to when goods are transported.

We will hopefully see 100% driverless but there will still be a place for ownership rather than summoning a pool car. Lots of people carry enough equipment to make summoning a random car impossible. As an example some muppet ran in to my van in the snow and it's going for repair next week. 

A hire van is being provided but it will take me around 2 hours to transfer my tools and spares and then rather than having everything racked and in place and easy to find I will spend a few days tearing my hair out digging through boxes on the floor. It's always a nightmare. 

I see a mix of pool vehicles and both private and business ownership sharing the roads but all driverless and intigrated in to a common computer standard. 

I also think it will be at least 2040/2050 before we get there fully but I think we will see driverless deliveries and taxis by the mid 2020s

It's all a matter of choice, if you would rather  use autonomous cars that is finem but it shouldn't compromise other peoples freedom of choice.

Therefore autonomous cars are going to need to be capable of sharing the road with other people who choose to drive for themselves.

You appear to have sidestepped the question!

We can do all of those things on good old fashioned cars with human drivers, we don;t have to go down the road to autonomy to achieve such benefits.

Surrendering your own freedoms is your choice.

We need to be more careful about removing other  people freedoms and foisting our choices onto the,.

I think that self driving cars will probably become widespread in one form or another. Personally, I won't like it and millions of control freaks just like me will not either.

Kindly show me where I have advocated a curtailing of other people’s freedom of choice.  I believe I have said I am emphatically for mixed usage of the roads.

It’s just as well that this is what the people spending billions of $,£ and € on are doing.

Mind you, whilst I am all for choice, a million deaths globally a year from bad human driving are only tolerated because it’s the least worst option.  When people start choosing to drive manually and then kill people, it becomes a very different matter.  I think the social changes around autonomy are going to lead to a much higher required standard of manual driving and tougher penalties for messing up.  The autonomous fleet will have much more evidence quality data on most human mistakes as well, so liability will be clearer than is often the case now.  

In reply to wintertree:

A self driving campervan would be amazing. set off just before bedtime and wake up at your destination!

Indeed!  Totally transformative of big continental trips.

I’m a great fan of doing “linear” walks - it’d be great to be dropped of in upper Weardale and collected many hours later in Hexham for example.  Just not viable self driven and with public transport.

Odd I thought I answered it. 

I’ll say it more explicitly.

Autonomous cars are doing this now to some degree in and its in the plan.  They have a truck load of additional ££££ kit to give the capability to do it (and needed for autonomy).  Manual cars do not have most of this kit.  

Would you really spend £1k on an additional sensor suite and cloud data service to have your car warn you about potholes?  Really?  But many people will spend it for autonomy and will get all the other benefits.

Roads are shared resources.  They only work if people play by agreed rules.  If autonomous cars become common and then dominant, it will almost certainly result in  manual cars being removed from most roads, otherwise the benefits won't be available to the majority.  For similar reasons we don't allow pedestrians on motorways, or lorries on some roads.  

You need someone in charge of the vehicle and ready to take over at a moments notice so going to sleep would not be an option.

I assume you use taxis? So you're happy to surrender your freedom and allow another human to drive, but not a machine? Even if the machine has been proven to be far safer?

I didn't suggest removing anyone's freedoms.

>"Therefore autonomous cars are going to need to be capable of sharing the road with other people who choose to drive for themselves."

Or insist on pedalling around on two wheels. In the Guy Martin show the Czech boffin claimed that satellite accuracy meant that movements could be judged down to a couple of centimetres. Pretty handy for those users who consider a 15 cm clearance enough for passing a road cyclist.

Should we be concerned that in the event of an accident between an autonomous vehicle and a manually driven car the "evidence" will be held by a global corporation with a huge vested interest in proving that they weren't at fault?

Realistically I would be far more likely to be able to afford retrofit sensors than a whole new autonomous vehicle.

Therefore for the right suite of telemetry gear and data the answer is an unequivocal yes.

Surely at the very least it is time that live smart motorway data was incorporated into sat navs?

Sadly I suspect that any such move will exclude a great number of motorists who have never managed to afford a brand new car from car ownership.

You seem to have forgotten why this sub-discussion about pot holes is going on.  It wasn’t about a purported benefit of self driving cars, but two posters suggesting that potholes won’t be the problem some consider them to be and some suggestions as to why being given.

As I said there’s probably no market for a cloud based pothole warning retrofit, but I think it’ll become normaI on future vehicles (not as a feature itself but as a natural part of “live maps”).  

Of course I use taxis on the very rare occasions when they are the best option. That amounts to 3 taxi journeys in the last 2 years.

However that doesn't mean that being forced to use autonomous vehicles is either financially or logistically realistic for a great many drivers.

Autonomous vehicles must be able to safely co-exist with manually driven vehicles.

I look forward to the day when we can all benefit from such data regardless of our choice of vehicle

Who suggested forcing people? It wasn't me.

Of course. They wouldn't be allowed on the road otherwise. 

What's your point?

Manual cars will be around for a good few years. As will pedestrians, cyclists, horse riders. Then there will be flooded fords, fallen branches, wild life.  There will be mist and fog and snow and ice to detect. An autonomous car will need to deal with all safely.

There might be owls!  What about owls?  What if an owl swoops down?  An owl swooping down during during an earthquake?  A parliament of owls divbombing during an earthquake.  Don’t know about you but I’d not drive my manual car if being divebombed put a parliament or owls during an earthquake whilst aliens attack.  Perhaps I’d call for a taxi, huh?

Mind you many drivers are appallingly dangerous in fog, snow and ice.  The bar isn’t exactly set high for autonomy here.  Knowing when not to use fog lights and leaving a 4-second separation in adverse conditions and limiting speed on ice are all trivial to automate and would be a damned sight safer than many squishy meat bags...

Timjones going on about freedom has got me thinking, and I've come to the conclusion that the autonomous car will inevitably mean the end of manually driven cars on the road. It simply won't be justifiable any more.

Driving a car is not a fundamental human right, any more than owning a gun is. Cars and guns are both machines capable of killing people when misused. When we allow people to operate a dangerous machine, we have to weigh up the advantage to people against the damage caused. We quite rightly IMHO banned handguns in this country, despite the fact that the number of deaths was tiny, because the advantage of owning a gun was tiny - a few people's entertainment. Humans driving cars cause a huge amount of damage, but we tolerate this because we don't have a realistic alternative at the moment, and IMHO freedom of mobility is a human right. Once there is a practical alternative to human drivers, I can't see how we can continue to justify the death toll that they cause - since the only reason to drive yourself will be for your own entertainment.

But will an autonomous car work on a treadmill?

And what has an autonomous car ever done on grit? That's what I want to know.

You do realise that with radar and traction control a self driving car is far better equipped to deal with rain, snow fog and ice than any human, unless you know someone who can see through these.

Every time it snows there is carnage on the road, people don't slow down enough or leave a big enough gap. I'm currently looking for a new van as a result of the last snow. Every time it's foggy or raining heavy with poor visibility you get people flying past you on the motorway and see multiple accidents.

Autonomous cars will be far safer in the fog because whatever radar system they settle on will see through it, also if the car in front has to hit the brakes for any reason it will broadcast this to the cars around it who will also break with far far faster reactions than a human can. 

Autonomous cars will still have accidents because occasionally things will go wrong but it will be far far less than with humans in control and I can certainly see the day when self drive is banned from public roads and restricted to track days and the like. 

You make it sound like an autonomous car would not slow down in rain, fog, ice when it most definitely should.  Results so far where the surface is slippier than an auto car equations supposed have not ended well. The auto car still will not see round corners or know if the way is clear so will need to be programmed slow right down when approaching them. 

Surely you get the best of both worlds if you incorporate traction control and radar into a manually driven car?

Carnage?

Are you sure that you aren't over-dramatising things a little

Maybe a little over egging but the one day I was called out over Christmas was the one day, well morning that it snowed in Leeds. 

I was out for 2 hours and drove 24 miles to site and back. During that drive I saw the aftermath of 6 accidents, someone lost control and slid in to me, I lost count of how many cars I saw skid (I slid myself once) and had multiple vehicles overtake me at speed going down hill on an un-gritted road. Taking people out of the equation can only be a good thing.

I can't wait for a self driving van, that is once we get to the point where the driver doesn't have to sit ready to take over. I can lose 3 to 4 hours of a working day to traveling between sites when I could be catching up on paperwork and planning. It will be hard to give up the motorbike once self driving is banned though. 

Given that the driver was given clear warning is this reckless, what can he be charged with and what's the insurance situation

Radar cannot not see black ice

To a degree, but we have some epic ones and I have broken a couple of springs on a Nissan XTrail over the past few years hitting big potholes.

I have to admit that if I was still allowed to ride motorbikes I wouldn't be so enthusiastic about giving up driving.

That's the problem with humans in control. It's snowy and icey so most of us slow down but after a while when our wheels have not slipped we speed up a bit, still no slippage so speed up a bit more and a bit more suddenly they do slip but your going too fast to do anything about it. 

we will see , they can't even manage a slightly damp corner at the moment

Nonsense.  

I’m not sure what you are basing this claim on?  Please do share....  Unless it’s something like the daft video someone linked above which is more like light entertainment than any of the serious autonomous road program.

I meant to say earlier, you are totally wrong about radar and ice.  There are lots of different ways RF waves can be used to detect ice.  Current automotive radars may not detect ice, but that’s not what you said.

This one drives in the rain

https://www.theverge.com/2017/2/14/14610614/drive-ai-self-driving-car-rain-...

The idea rain is a showstopper is absurd, given the by-in from manufacturers.

I do wonder about how fine detail will work though off road where things aren’t mapped. There will I assume need to be some human way of controlling things here.

it it cannot rely on mapping as we all know from our sat navs mapping often falls way behind what it looks like on the ground.

Please provide a link to ice detecting radars do how it works round bends.

I’m sure you’re smart enough to google “RF backscatter ice”.

You didn’t mention seeing round bends before.

An autonomous car doesn’t need to see round a bend, especially in black ice forming conditions.  It’ll do what people should and drive so that it’s stopping distance is less than it’s forwards visibility.

So no links then?

If you can’t be bothered to google “RF backscatter from ice” or “radar ice detection” I doubt you could be bothered to read any link that I post.  

If I did, you’d probably point out that whatever I linked to isn’t an automotive radar, ignoring my entire post.  I’ll repeat it here to head that off.  There are lots of different ways RF waves can be used to detect ice.  Current automotive radars may not detect ice, but that’s not what you said.

I did , nothing came up in first four pages of links I clicked through about back scatter from ice and its use in road going radar.  I was looking for evidence this exists or has even been thought about. You present no evidence that it could be used and is in use for autonomous vehicles at all for black ice on tarmac roads.  Please present evidence rather than belief.  

Did you read my post, or my follow on post?   I’ll repeat myself again.  There are lots of different ways RF waves can be used to detect ice.  Current automotive radars may not detect ice, but that’s not what you said.

Well it’s clear I just thought about it.  Perhaps I’m the first person to do so.  Golly, that seems unliklely.

I never claimed I had any.  I was very explicit from my very first mention that I was simply correcting your brief and wrong statement that “radar cannot see black ice”.  You were wrong.

What it appears you meant to say is that “No currently publicised automotive radar can currently see black ice on tarmac around a bend”.  If you read the first post of mine that you replied to, you will see I made it clear that I wasn’t talking about current automotive radar.  So I’m not sure why you’ve been arguing rather than stating your full position.

Plenty of evidence that radar can detect ice.  That was my point, you seem to be wilfully trying to turn it into a different argument.  Can’t help you there.

The corners thing is nonsense.  You really think they know where to go by direct measurement from satnav?  How do you think lane detection works? The car sees the road

Only in America ... "0:30 Narrow street with parked cars and glare from oncoming traffic."

Looks like your average relatively major UK A road.

And WHAT pray tell did you do, so that it means you are no longer allowed to ride a motorbike ? (and also means that you wish others to no longer enjoy the privilege of same).

Said ban did not make things any safer - all it did was annoy some people who were perfectly law-abiding, whilst allowing some others to put on a smug face. The problem with hand guns, just like any weapon, was with those held by the criminal element - who are now using acid to throw in peoples faces. And before you bring up Dunblane, if the Assistant Chief Constable had done his job then the criminal (paedophile) in question would have been locked up and his guns (including illegals !) would have been destroyed long before it happened - the official report has been buried, never to be seen.

Humans driving cars cause a huge amount of damage, but we tolerate this because we don't have a realistic alternative at the moment, and IMHO freedom of mobility is a human right. Once there is a practical alternative to human drivers, I can't see how we can continue to justify the death toll that they cause - since the only reason to drive yourself will be for your own entertainment.

What about any driving NOT on the public highway - farmworking, estates, maintenance. And passing-places roads - how will the computer-controlled vehicles decide which one reverses ?

There are many scenarios which I could list which do not fit into the urban driving situation. Unfortunately it is the urban population of our major metropolises who will try and impose a system designed to "fix" their problems on the rest of the country - no matter how flawed and unworkable it is - after all, that is what normally happens.

This is new-ish technology and the example is optical rather than radar, but I would imagine something like this:

https://www.technologyreview.com/s/603314/new-camera-can-see-around-corners...

We had a similar experience with my Grandad.  My Mum and her brother argued with him many times about whether he should be driving and confiscated his keys more than once.  In the end a health issue forced him to give up his licence temporarily and he never passed the assessment he needed to get it back, not because the health issue persisted but because he just wasn't competent any more.  He loved his car and refused to sell it so it rotted on his drive for a decade.  My Mum and her brother opened an account for him at a local taxi company and paid the tab monthly, but he only used it a handful of times - he felt like relying on taxis wasn't independence.  He would have loved to be able to use a self-driving car of his own as and when he wanted to.

My Mum his going the same way now - she still drives but her eyesight is deteriorating rapidly and I doubt she will pass her next assessment.  She can and does use buses and taxis (mostly because she remembers arguing with her Dad and doesn't want to argue with me the same way) but she hates it and would probably also love a self-driving car if the technology was ready.

I've wondered about this, given statements about cars explicitly not having a human controlled aspect.  My guess is there will continue to be hybrid type cars with which allow human control and autonomous control when suitable.

I'd imagine they will "talk" to each other in advance and adjust speeds so there is no need for reversing, or even stopping.

Why? No previous advance in road safety technology however significant has lead to legacy vehicles being excluded from the road. It's still perfectly possible with minimal modification (lighting and numbers) to register and use 100yo vehicles that are by modern standards barely controllable and unrecognisable from a modern user or pedestrian safety perspective. I see no reason for that to change, there is far too much monetary and emotional value tied up in classic vehicle collections, many held by the rich and the influential for them to be barred from the road. I do suspect the use of polluting legacy vehicles will be taxed as it hasn't been before to encourage limited use.

jk

How do you see around bends? Actually, simpler than that: how do you detect black ice?

jk

I repeat for your benefit. Please provide links to these black ice detection systems you say exist.  

You're right.  The lack of a link to meet you ever changing demands means that autonomous vehicles are impossible and will never happen.

I don’t see manual vehicles being banned, but I see changes with similar effects.

• Large volumes of evidence quality telemetry from autonomous vehicles precisely implicating a bad driver in both normal driving and accidents.
• A toughening societal attitude to those who cause injury or death by bad driving once it becomes a choice made for purely self driven reasons and not utility.

My hope is towards a much higher level of excepted competence and much more robust enforcement of that level.  I also think that soon enough most new manual and optionally manual cars will have the same safety features as fully autonomous cars, using the same sorts of sensors.  Take my current car - bloody thing overrides the TCS disable switch in bends unless you pull 3 fuses... 

I might be wrong but I'm not sure the car buying public will go for turning each and every car into a mobile speed camera ready to spy on them and others for the state. Then again technology appears to have pretty much tamed teenagers in <20 years so maybe we will be next. Maybe we already are.

As far as I can tell the traction control over-ride switch in mine does nothing but illuminate another warning light on the dash. No fun at all, give me a pair of big tyres and an LSD over the always-on electronic trickery any day

jk

I don’t think they will either, but the manufactures and operators of shared ownership / on demand fleets will.  They have a sound business reason (encouraging adoption) to vigorously defend the public perception of the safety of their vehicles, not to mention reducing insurance payouts.  The regulation of this sort of thing is one area that doesn’t seem to be getting much government attention either side of the Atlantic.

Those were the days....  Per-wheel electric motors should be close to the perfect solution for safety (if not wonton hooliganism) but it’s not really catching on.

This is an interesting area.  There are all sorts of driving behaviours mediated by human-human interaction (nods, waves etc.), and also behaviours that aren't really legal but widely accepted.  How are these going to work with a mix of autonomous  and manual vehicles?  I liked the thought someone had of manual drivers "bullying" autonomous vehicles but playing on their propensity to stop if uncertain about things.

I have to run so can’t dig out the link, but Arstechnica had an interesting article on this sort of thing recently - autonomous testing cars are involved in a higher than normal number of low speed accidents where the other (manual) car is at fault.  The conclusion is that human drivers make all sorts of assumptions (that they don’t validate) about the behaviour of other cars deviating from the rules.  It’s going to be a complex adjustment period...

This is a really good idea.  I bet that even without computer controlled vehicles if councils were willing to spend some money on roadside equipment you could build an intelligent system for single track roads to sense where vehicles were and using computer displays tell them to either slow down, speed up or wait in next layby and minimise the overall amount of waiting and likelihood of anyone having to reverse.  It would be a cool project for an undergraduate EE/CS student to figure out the algorithms.

Great idea but people will ignore them. A message saying slow down to make things faster for everyone translates in to why should I slow down to make things faster for everyone else, it's going to slow me down and I'm not having that. It's the same with temporary speed limits on motorways, everyone ignores the flashing 50 limit until they can actually see the delay then they join the choke point aggravating the problem. If everyone slowed down to 50 there wouldn't be a choke point and we could all progress but people are selfish. 

Although you often go for mile upon mile of emptyish road with the 50 limit flashing to find no obstruction - and that's basic technology stuff. I fully agree with Mr Wolmar.

I think with the single track road if the scheduling was working well it would be really obvious.  You'd see you were arriving at laybys when needed with reference to oncoming traffic.  If you ignored the system and ended up blocking somebody that was playing by the rules they wouldn't be happy.

I was just speculating on the morality of driving. It wouldn't happen quickly, but I think that manual driving will get less and less common until eventually there are so few people doing it that it will be banned on most roads.

From our current perspective this seems unbelievable, but in some ways it's our current view on driving that is strange. We just seem to accept deaths caused by driving as a fact of life, but I struggle to think of any other activity where this would be acceptable. If climbers killed innocent bystanders on a regular basis then climbing would be a criminal activity and we'd be social pariahs.

Think back to the smoked filled pubs of your youth and imagine telling people that in a few years time smoking in public would be illegal and unacceptable. I'm sure most of them would have laughed at you.