So, only when you're stuck in traffic then?

Yes. That's where some level of automated driving might actually be useful and low-risk. Both for the driver of the car, and the reduction of unnecessary speeding up and braking that causes unstable "waves" in the traffic flow.

As soon as you include higher speeds and non-motorway driving, the ethical problems (and the political question of where to set the risk/performance trade-off point) become HUGE.

Yes, it's exactly this kind of scenario that automated driving technologies will be most useful.

However to fully function, all the vehicles need it, and ideally you need some form of central control to monitor and make decisions based on all the vehicles.

Absolutely not. Just need vehicle to vehicle comms within a speed-dependent radius, and better sensors than just relying on cameras i.e. radar, lidar.

There's simply no need for some kind of overlord watchdog system to know where all vehicles are at all times. There's enough of that with ANPR as is.

> ... reduction of unnecessary speeding up and braking

Dream on. They'll just do it automatically.

It's certainly a risk if the control loops aren't optimally damped (including how they interact with each other) - could end up with some "fun" oscillations if a resonant frequency can be found ;-)

I've always felt that the standard of driver awareness would be considerably enhanced if the driver side airbag was replaced with a spring loaded spike which shot out of the middle of the steering wheel in the event of a crash.

spring loaded spike

If you do that then the average Tesla driver will be sitting in the back seat with a beer.

shot out of the middle of the steering wheel in the event of a crash.

Save the NHS!

" How can we all be satisfied that drivers of vehicles fitted with ALKS know how and when to take back control of their vehicle? "

For me, the answer is simple : we can't.

As long as there are bloody Tesla idiots who put themselves in the back seat of their car, I am not going to put an ounce of trust in any technology that purports allowing the driver to not pay attention to the road.

Give me a true autonomous vehicle, or get lost.

It seems that last point is actually harder than was advertised.

My last three cars claimed to be able to keep me in lane (ish) and avoid accidents but they all failed to detect or wrongly detected where lanes were at one time or the other. They all also falsely alerted me to accidents which weren't about to happen (no it's not me driving like and idiot and it was also reported by other drivers of the cars concerned).

I'd love to say the errors were very uncommon but they weren't and all three cars were from different manufacturers.

Some new technology has been useful, like blind spot monitoring (as long as you still look), radar cruise control (as long as you sill pay attention) and reversing cameras/radar (as long as you still look) but I think autonomous driving is either too difficult to achieve (except in very tightly controlled environments) or will require constant supervision which rather spoils the point.

I have the occasional misfortune of having to drive through the roadworks where the M3 and M27 join. New lane markings have been painted over partially scrubbed out old markings and the car fights me over which set of markings we should follow. I figure we both have about 50% accuracy except when the road is wet and the sun is glinting on the puddles, at which point we are both bloody confused.

I very carefully selected the poverty spec version of a recent car purchase precisely because it doesn't have all the technotoys.

Though I do feel that leaving the keyfob button to unlatch the boot but not including the function itself was a bit mean.

I have a vehicle that has ALKS which has never been engaged . If I am not able to maintain the correct lane, then I should stop and get some rest (unfortunately, not everyone thinks that way).

Having been in the technology business for over half a century, I am naturally highly sceptical of 'breakthrough' technology as it usually isn't. I am also highly sceptical of autonomous anything.

There are some autonomous vehicles in appropriate settings where the traffic density is low enough to permit such operation and / or where the hazards to a human operator are very high. That is one of the reasons that autonomous (sometimes semi-autonomous) minehunters have been developed among other things. There are autonomous (very large) trucks used at mining operations but that is hardly a high traffic density zone.

New technology though?

ABS? They may be electronic now, but they were fitted to aircraft designed in the 50s which although mechanical had the same underlying mode of operation.

Regenerative braking? That was in use in South Africa (railways) at least 50 years ago. Trains were timed so as to have one going down a steep incline while another was going up . The train going down fed electricity back into the local power section so the train going up had enough energy available. My power electrical lecturer waxed lyrical about it in the early 70s.

Lane keeping at it's simplest is not rocket science; there are lots of problems to contend with, though, particularly in reduced visibility.

There are some useful pieces of technology in the vehicle, though; it is an automatic and when I am going down an incline and I brake moderately hard, the system will automatically downshift me to the lowest appropriate gear which is very useful in the rather hilly terrain around here.

Then there are things that have uses in some countries but are a waste of space and resources in the UK; cruise control comes to mind. For the most part, trying to use cruise control in the UK is more effort than it is worth. I found it very useful in the USA, though.

I have yet to see any real breakthroughs for road vehicles.

> ... not rocket science

Commentards have previously pointed that it's rocket engineering which is difficult. The underlying science is relatively simple.

The difference between simple and easy. The rocket equation is simple. Turning it into a working reliable rocket is not easy.

For the most part, trying to use cruise control in the UK is more effort than it is worth

That depends on driving style. If you're the kind of driver that is forever overtaking other vehicles just because they are driving at 5mph less than you or because you think shaving twenty seconds of a half hour journey is worthwhile then, yes, CC will seem pointless.

But I use it almost constantly on motorways, especially now that I have a car with radar control. I just plop myself in lane one, set it to 60mph and for the most part ignore it. A couple of times an hour if there's a long line of HGVs ahead I might pop out into lane two and get my foot down because that probably means there's a dirth of HGVs after that. Also if a hill is severely slowing them down, But otherwise I just bimble along having a nice relaxed drive and probably adding no more than fifteen minutes to the duration.

I also sometimes use it on single carriageways if I'm familiar with them. I nearly always use it on the A422 Middleton Cheney bypass though it's only a mile or two long. I'd use it along the whole stretch to Brackley (have done in rush hour on occasion) but I'm marginally better at managing fuel consumption on the relatively hilly bits so I tend to only use it where it's flat.

I'd never want to own a car that didn't have it ever again.

Sounds to me like using ALKS as it exists today is going to be more difficult for a non-suicidal driver than driving without it is. You have to do everything you'd do while driving except the trivial job of manipulating the steering wheel plus watching out for early warning that your car is going to try to kill you. What possible benefit does the driver gain from this technology until it is truly ready for prime time?

I have a lane system on my car....I "play" with it once in a while. Would I trust it or depend on it? Absolutely not!

Mine, AFAIK, depends on "seeing" the road markings to determine if in lane. How many times do the lane markings disappear, are worn out or the system not "see" them? A lot in my experience. I also have a system that picks up on road speed signs and displays the value on the dash. I would say it is only accurate 60-70% of the time. That for me makes it bordering on the useless.

bordering on the useless

No, that's just dangerous.

You're blaming the self-drive cars for making the same mistakes that the average driver makes all the time?

"Sorry officer, we're on I-95 and I thought that was the speed limit."

These car manufacturers don't seem to state where their control units have been made and which country has jurisdiction on workers being able to access remotely the car and modify it's behaviour.

There is a possibility that in time of conflict, a foreign power can take control over the cars and use them as weapons.

For me a self driving car is a no no and if they are let on the road without any oversight...

Uuuuuum... Yeah... Can we have some of what you've been smoking? It's been a long week.

"restricted to motorways and to speeds of 37mph (+- 60km/h)"

Er, what?

60km = ~37miles

So '37mph (+- 60km/h)' = 0 to ~74 mph

Still makes no sense to me why they would mix km and miles though!

Why not just state zero to 120km/h, or up to 74mph?

The +- 60 km/h meant "37 mph equates to around 60 kph", not that it works in a range of 0 to 120 kph!

Though most likely it works up to EXACTLY 60 kph, and the 37 mph should have had the +- treatment.

It's a traffic jam management system.

Plenty of traffic jams end up with additional collisions during the jam because people aren't paying attention.

The lane keeper system in my car makes it hard to avoid potholes, bicycles and broken down vehicles. So it is firmly turned off.

Really - the lane keeper in mine is easily overridden, feels like driving up a small gutter slope.

The biggest stumbling block to adopting self-drivng cars is the necessity for a driver in the seat.

Get rid of the driver, and the controls, use ML and you have the world's most experienced driver taking you everywhere.

No, shut up! There are no 'edge cases' - do you thinkg the world's most experienced driver doesn't know what to do?

I feel like maybe you're overestimating the quality of the training sets, and the abilities of ML.

A very experienced cyclist in my local area was killed a few years ago by a Tesla driver. While not knowing the inquest / legal outcome of this case, given the location it happened there is practically zero chance that this wasn't down to the driver not paying attention to the road, and with it being a Tesla a good chance that autopilot was engaged as well. The last thing we currently want to allow is hands off lane assist.

I was on site at a DXC office a few years back, during a break one of the managers mentioned to us that they had fully autonomous driving solved and it was only the insurance companies stopping them from being used. I somehow managed to maintain professionalism and avoided shouting bullshit while laughing my arse off at him.

The article says, " industry and regulators have solved all the issues relating to how drivers interact with advanced driving assistance systems. " but they are not mentioning all the issues of how advanced driving assistance systems interact with the designated driver.

Back in the 1950's, as a 5 year old kid I could just catch the bus every day to go to school and then back home - if we get this new technology to work (I doubt it) does this mean that every five year old will have their very own voice controlled vehicle?

When self-driving vehicles were first described I thought this might be a nice technology advance but now, after watching it fail too often I wonder if we would not all be better off returning to the days of decent bus and rail services.

Somewhat worryingly, DXC have a contract to provide support to BMW, for their autonomous driving development, including up to level 3 fully autonomous driving due in 2025!

It can also [1]mis-identify diverging lane edge markers at an off ramp as a lane, enter it, accelerate towards a bloody great black and yellow chevroned plate and smash into it at over 70 mph .

For some inexplicable reason, the "artificial intelligence" could identify a moving car in front as an obstacle, but not the stationary big yellow and black rectangular plate.

And actually it's not an autopilot in the strict sense. But even if it were, the big difference between aviation and road traffic is the separation between vehicles and obstacles and the time available for decision making. There's a fallacious assumption that road vehicle avoidance automation should make decisions about the nature of obstacles, which makes the process much more complicated (and complexity is hazardous).. Aviation collision avoidance systems very sensibly aren't expected to discriminate between different categories of obstacle - just to warn that there's one (of any old kind) ahead.

[1] https://trid.trb.org/view/1707883

In aviation there's never a stationary aircraft, so anything that isn't moving is a mountain.

My Corolla has lane assist and I avoid it most of the time. It likes to sit at the right side of a lane rather than the centre which suggests it's not been correctly updated for LH drive vehicles. It also sways a bit as it goes and is constantly twitching the wheel. If I relax and let it do its own thing it moans that I'm not holding the wheel. But if I'm holding the wheel there's a constant low grade struggle going on which is more tiring than me just holding onto the wheel unassisted.

"Legal terms presented to the user potentially operate to their detriment."

Legal terms presented to the user *always* operate to their detriment.

Those who make the rules are falling for the idea that humans make mistakes (they do) and that driver "assistance" things do not (provably false, time and time again.) A big difference is that when a human makes a mistake, they take immediate corrective action. The assistance things just blunder on.