Driverless trains - why limited progress on the national rail network?

najaB said:

I know that, and I fully expect we will see more and more systems like Tesla's Autopilot over the next few years. However, at some point even the easy task will be more than the automation can handle - it could be something as simple as water getting into a connector causing the sensors to drop out.

The point I was making is that, if that were to happen, a remote operator in a contact centre somewhere in the world isn't going to be much use to me when I'm heading towards a bridge abutment at 70mph (or, for that matter, if the failure is in the HGV that's behind me and my car has to brake rapidly).

Click to expand...

I think, in the early iterations, this will be the failure mode - stop the vehicle, driver takes control. Over time, as the failures get rarer, the need for a driver will be eliminated - the remaining occupant, be they passenger, goods handler or whatever - will most likely have a guided process to reset the system, and if the failure persists, follow prompts to enable a degraded, 'limp home' mode. If all else fails, you stay put and call for an engineer - not really so different from today's vehicles.

NotATrainspott said:

Because if it doesn't, the cost of traditional signalling systems will cripple the railways financially. It might not be called ETCS Level 3 and it may not have the moving-block part but you can be guaranteed that you'll be using purely wireless control for movement authority eventually.

Click to expand...

That's complete rubbish. The cables for TCB are already in place. And unless it can be proven to be completely safe and reliable then it will not happen.

Dave1987 said:

That's complete rubbish. The cables for TCB are already in place. And unless it can be proven to be completely safe and reliable then it will not happen.

Click to expand...

How can you say Notatrainspotter is talking rubbish? Your evidence please.
I,& others,have found his postings exciting,thought provoking & really positive compared to your persistent negativity. The idea of low cost cableless sensors that draw power from the 25kv AC varying field is very very interesting & really useful for a safe automated railway. I only did A level physics, & some years ago now,i.e. not a degree. But my worry with this interesting idea is that if there is no train in the 25kv section, or if the trains in the section are coasting with very little auxiliary power being drawn,perhaps empty stock with heating,air con off,then there will be no current in the overhead wire & so no field. Also neutral sections of wire will have the same zero field problem. Am I right that no amps in overhead means no field & no field means no power for sensors set in rails,or cableless sensors on passing trains? Is this a serious barrier to these low cost sensors?

najaB said:

I think the truth is somewhere in the middle. It's likely more than ten but less than thirty years before the majority of road vehicles are capable of fully autonomous driving.

Click to expand...

Normally the cost of learning to drive is split over about 50 years (from 20 to 70) of driving. So now it's down to 30 years, probably less than that. I suspect I'll be the last generation of my family to learn to drive, by the time my eldest is old enough your timescale will be down to about 15 years, and that's for the majority. Add the cost of insurance, fuel, and most crucially the availability of things like uber (perhaps automated) means it wouldn't be required.

The march of automation will be driven by more effective utilisation of assets. There's a large cost up front to develop the technology, so how fast it happens depends on how much money will be saved. In the case of HGVs, automated driving would not only allow the removal of the driver (saving £10 an hour), but also allow all those pesky tacograph stops (allowing a HGV to be utilised for 20 hours a day rather than 8).

There are fewer financial benefits of automating a plane completely -- the cost of the pilots is rather small compared with the cost of running a plane (you'd still need cabin crew), and while there are occasional issues down-route, especially on long haul, which lead to delays or cancellations they are fairly small. I suspect that automation will remove the need for a co-pilot at some stage though (there still will be co-pilots on some flights, if just for the training purposes)

Trains are somewhere between. The 1 car 'train' I got today ran for 4 hours, had 2 members of staff aboard, and few passengers. Ripe for automation or replacement with something from the 21st (or even 20th) century. On the assumption that the majority of road vehicles will be capable of fully autonomous driving in 30 years, why will this train continue? Either passengers will be using more efficient cheaper automated taxi/busses, or the train itself will be automated.

A mainline train travelling at 180mph with 500 passengers on board has less to save from removing the driver though, and less competition from automated road vehicles.

The saving from automating driving is like the saving Southern are hoping from DOO. Not to remove the guard (at least initially), but to ensure that if there's no guard for some reason the asset is not sat idle because of a lack of staff.

XDM said:

How can you say Notatrainspotter is talking rubbish? Your evidence please.

Click to expand...

Evidence that there are cables in the ground? Surely that would be the fact that there are cables in the ground.

paulweaver said:

Normally the cost of learning to drive is split over about 50 years (from 20 to 70) of driving. So now it's down to 30 years, probably less than that. I suspect I'll be the last generation of my family to learn to drive, by the time my eldest is old enough your timescale will be down to about 15 years, and that's for the majority. Add the cost of insurance, fuel, and most crucially the availability of things like uber (perhaps automated) means it wouldn't be required.

Click to expand...

This is what I find most interesting - peoples' decision to become motorists being deferred by a technology that will likely get deferred by any number of legal, societal and possibly technical reasons.

When cars came along the railways responded with internal combustion powered "railcars", but of course this did not protect them from losing a lot of their business. Now that today's Digital natives get mobility as required from the cloud, the car people are responding to this by attempting to make their cars mistakable for smartphones.

I personally think that the wider society has had enough of untrammelled computerisation for now - the current US president articulates these views. I think that society is becoming less willing to embrace driver-less cars with each passing month.

Now with regard to NotATrainspott and his visions, I reckon the downfall of everything being networked relates to cyber security. Aircraft and train systems - at least those that have the authority to drive - should in my view be hardwired (or at least on a read-only memory) and offline.

Deleted

paulweaver said:

Normally the cost of learning to drive is split over about 50 years (from 20 to 70) of driving. So now it's down to 30 years, probably less than that. I suspect I'll be the last generation of my family to learn to drive, by the time my eldest is old enough your timescale will be down to about 15 years, and that's for the majority. Add the cost of insurance, fuel, and most crucially the availability of things like uber (perhaps automated) means it wouldn't be required.

The march of automation will be driven by more effective utilisation of assets. There's a large cost up front to develop the technology, so how fast it happens depends on how much money will be saved. In the case of HGVs, automated driving would not only allow the removal of the driver (saving £10 an hour), but also allow all those pesky tacograph stops (allowing a HGV to be utilised for 20 hours a day rather than 8).

There are fewer financial benefits of automating a plane completely -- the cost of the pilots is rather small compared with the cost of running a plane (you'd still need cabin crew), and while there are occasional issues down-route, especially on long haul, which lead to delays or cancellations they are fairly small. I suspect that automation will remove the need for a co-pilot at some stage though (there still will be co-pilots on some flights, if just for the training purposes)

Trains are somewhere between. The 1 car 'train' I got today ran for 4 hours, had 2 members of staff aboard, and few passengers. Ripe for automation or replacement with something from the 21st (or even 20th) century. On the assumption that the majority of road vehicles will be capable of fully autonomous driving in 30 years, why will this train continue? Either passengers will be using more efficient cheaper automated taxi/busses, or the train itself will be automated.

A mainline train travelling at 180mph with 500 passengers on board has less to save from removing the driver though, and less competition from automated road vehicles.

The saving from automating driving is like the saving Southern are hoping from DOO. Not to remove the guard (at least initially), but to ensure that if there's no guard for some reason the asset is not sat idle because of a lack of staff.

Click to expand...

The HGV stuff is spot on, and why automation will be more beneficial than just the reduction in labour cost. Hauliers will be able to move just as much stuff with fewer trucks, and in a free market that inevitably means lower costs for the rest of the economy.

Cargo aircraft are more ripe for automation than passenger ones for those reasons, and the ones that I have previously described. Also worth remembering when you automate a vehicle completely is that you no longer need to waste money on keeping it habitable for the drivers or pilots. A pilot-less cargo aircraft wouldn't need to be pressurised, resulting in huge cost savings for airframe design, maintenance and a much longer lifespan for the asset.

The main advantage of automating a train is that it allows far more optimisation of capacity than is economically feasible otherwise. One notable example of this is that Elizabeth line trains terminating at Paddington from the east will be permitted to drive themselves unattended into the Westbourne Park sidings and then back into the eastbound platform. This isn't intended for use when things are running to time but if there is disruption it will allow the driver to walk from one cab to another while the train is reversing, minimising turnaround time. It seems quite plausible therefore that the first application of full autonomy-based UTO on the heavy rail network will be in ECS operations.

najaB said:

Evidence that there are cables in the ground? Surely that would be the fact that there are cables in the ground.

Click to expand...

Cables that will need replaced at the end of their lifespan.

squizzler said:

This is what I find most interesting - peoples' decision to become motorists being deferred by a technology that will likely get deferred by any number of legal, societal and possibly technical reasons.

When cars came along the railways responded with internal combustion powered "railcars", but of course this did not protect them from losing a lot of their business. Now that today's Digital natives get mobility as required from the cloud, the car people are responding to this by attempting to make their cars mistakable for smartphones.

I personally think that the wider society has had enough of untrammelled computerisation for now - the current US president articulates these views. I think that society is becoming less willing to embrace driver-less cars with each passing month.

Now with regard to NotATrainspott and his visions, I reckon the downfall of everything being networked relates to cyber security. Aircraft and train systems - at least those that have the authority to drive - should in my view be hardwired (or at least on a read-only memory) and offline.

Click to expand...

People will not have a choice in the matter, because the bulk of these decisions are driven purely by market forces and not personal choice. People want cheap groceries and if Tesco start using autonomous trucks to move things from their distribution centres to their supermarkets, then they'll be able to offer lower prices on essentials than their traditional competition. Soon enough all of the companies will have no choice but to follow. Also worth remembering is that much of the labour upheaval caused by automation is happening behind closed doors. If Tesco opens an entirely automated distribution centre then no one outside of the company would ever see or understand that it exists, just as people don't really know or understand how their Amazon or Ocado orders are processed. All people are exposed and really care about in the end is the price, and people want more things for less money. As people find it harder to gain meaningful employment they'll have less money to spend on essentials, making autonomy-heavy companies even more competitive due to their intrinsic lower prices. The unemployed taxi driver who might find work somewhere may well have no option but to use the autonomous taxi that made him unemployed to get there.

NotATrainspott said:

Cables that will need replaced at the end of their lifespan.

Click to expand...

Cables in the ground have a near-infinite lifespan if properly insulated and not disturbed.

NotATrainspott said:

People will not have a choice in the matter, because the bulk of these decisions are driven purely by market forces and not personal choice. People want cheap groceries and if Tesco start using autonomous trucks to move things from their distribution centres to their supermarkets, then they'll be able to offer lower prices on essentials than their traditional competition. Soon enough all of the companies will have no choice but to follow. Also worth remembering is that much of the labour upheaval caused by automation is happening behind closed doors. If Tesco opens an entirely automated distribution centre then no one outside of the company would ever see or understand that it exists, just as people don't really know or understand how their Amazon or Ocado orders are processed. All people are exposed and really care about in the end is the price, and people want more things for less money. As people find it harder to gain meaningful employment they'll have less money to spend on essentials, making autonomy-heavy companies even more competitive due to their intrinsic lower prices. The unemployed taxi driver who might find work somewhere may well have no option but to use the autonomous taxi that made him unemployed to get there.

Click to expand...

Off topic but people absolutely do have a choice in the matter. The economy (and therefore market incentives, subsidies etc) is in the hand of politicians. And yes, most complex technology does require subsidy in one way or another - the allowing of firms to externalise their costs and internalise the profits - of course they will never call it a subsidy as such. And if the political system fails to heed the population for long enough, well that is how revolutions start.

Some day I'm sure you will look back fondly at how you once evangelised the silicon valley hype. All part of growing wise through experience. Of course some things will stick, but life will go on, and there will probably still be semaphore signals somewhere.

Oh and I did notice you ducked my suggestion that aircraft autopilots and trains must not be connected to the wider internet (unless this was already addressed elsewhere). That seems the main barrier to the more whimsical ideas - and the cyber security threats seem only to be growing still. We don't know how compromised our existing critical systems really are, not everybody will show their hand.

najaB said:

Cables in the ground have a near-infinite lifespan if properly insulated and not disturbed.

Click to expand...

Only for as long as the cables are fit for purpose. Resignalling works will eventually be needed and that's when you often need to replace the cables. At that point in time, why would you arbitrarily decide to use a more expensive way of delivering the same, or less, capacity?

squizzler said:

Off topic but people absolutely do have a choice in the matter. The economy (and therefore market incentives, subsidies etc) is in the hand of politicians. And yes, most complex technology does require subsidy in one way or another - the allowing of firms to externalise their costs and internalise the profits - of course they will never call it a subsidy as such. And if the political system fails to heed the population for long enough, well that is how revolutions start.

Some day I'm sure you will look back fondly at how you once evangelised the silicon valley hype. All part of growing wise through experience. Of course some things will stick, but life will go on, and there will probably still be semaphore signals somewhere.

Oh and I did notice you ducked my suggestion that aircraft autopilots and trains must not be connected to the wider internet (unless this was already addressed elsewhere). That seems the main barrier to the more whimsical ideas - and the cyber security threats seem only to be growing still. We don't know how compromised our existing critical systems really are, not everybody will show their hand.

Click to expand...

Automation doesn't require subsidy though. An autonomous taxi will be cheaper than a traditional taxi in every respect. There is no negative externality of automation in the traditional sense because the act of automation isn't in itself a bad thing. The only reason why people need to be afraid of it is if the government doesn't sort out the welfare system to cope with this new post-work reality. These changes should have happened decades ago, because we no longer have a contributory welfare system based upon the idea of full employment for all of society. In many respects automation is no different to de-industrialisation writ large.

The only feasible welfare system in this world is the Universal Basic Income. That policy will mean that people have absolutely nothing to fear from automation, as their basic income is guaranteed. If you know that you don't need to work to feed yourself, then you will not accept degrading work at low pay as people currently have to in order to survive. People will find themselves living in a perpetual weekend where they're free to do what they want. If you really like playing with semaphore signals then you'll be able to do so every day. However, unlike the times when people were employed to do it, when you get bored you can just walk away and do something else, since ultimately no one else relies upon you doing your job to be able to get around their lives.

Systems which could cause death are already connected to the internet. There is no inherent reason why they must be less secure than a closed system. When Airbus or Boeing do a digital system they have to work much harder to secure it than some random toy manufacturer. It is worth remembering that even if humans were preserved in important roles, they would still rely upon technology to such an extent that you could make the same criticisms about security. If you can hack an Airbus airliner today then it really doesn't matter whether there's a human at the controls, as they're really just telling the computer what to do.

Noticed this news story on the BBC website:

Uber suspends self-driving cars after Arizona crash

Uber has pulled its self-driving cars from the roads after an accident which left one of the vehicles on its side.

Pictures posted online showed the car on its right side on an Arizona street, next to another badly damaged vehicle.

The car - a Volvo SUV - was in self-driving mode at the time of the crash, on Friday, Uber said. No one was hurt.

A spokeswoman for the police in Tempe, Arizona said the accident occurred when another vehicle "failed to yield" to the Uber car at a left turn.

Click to expand...

Interesting that, according to the police, the accident was caused by a human driver not following the rules of the road.

najaB said:

Noticed this news story on the BBC website:

Uber suspends self-driving cars after Arizona crashInteresting that, according to the police, the accident was caused by a human driver not following the rules of the road.

Click to expand...

It's actually something they've discovered repeatedly with the self driving cars. The computer follows the rules of the road to the letter which very few human drivers do. IIRC the Google ones have had several nose to tails all of which were by tailgaing human drivers.

TwistedMentat said:

It's actually something they've discovered repeatedly with the self driving cars. The computer follows the rules of the road to the letter which very few human drivers do. IIRC the Google ones have had several nose to tails all of which were by tailgaing human drivers.

Click to expand...

Which makes me wonder if one of the biggest roadblocks to widespread driverless cars (and trains) is the fact that there's still lots of human-controlled cars (and trains) out there being illogical and dangerous.

najaB said:

Which makes me wonder if one of the biggest roadblocks to widespread driverless cars (and trains) is the fact that there's still lots of human-controlled cars (and trains) out there being illogical and dangerous.

Click to expand...

How would a human driver cope in these circumstances? Obviously 70% of drivers are above average, and 29% are average{1} and will read the mind of other drivers on the road via some sort of gut instinct, which is why there are so few accidents.

If someone pulls out infront of you, isn't it normally counted as that driver's fault?


{1} http://www.theaa.ie/blog/70-of-drivers-rate-themselves-as-above-average/ - "69% of the 15,000 motorists surveyed during the latest AA Motor Insurance said they would rate their own driving as above average. Inversely only 0.8% described their driving as below par"

najaB said:

Which makes me wonder if one of the biggest roadblocks to widespread driverless cars (and trains) is the fact that there's still lots of human-controlled cars (and trains) out there being illogical and dangerous.

Click to expand...

That argument works for cars, not for trains. Train drivers are highly trained (as we know) so their behaviour should be highly predictable and within the letter of the rules.

najaB said:

is the fact that there's still lots of human-controlled cars (and trains) out there being illogical and dangerous.

Click to expand...

So train drivers routinely drive illogically and dangerously do we?

What a winker (misprint)! :roll:

455driver said:

So train drivers routinely drive illogically and dangerously do we?

Click to expand...

From the perspective of a computer (if were capable of forming an opinion), I'm sure you do. And it's not just drivers, passengers and bystanders all do things that don't strictly conform to the rules.

najaB said:

And it's not just drivers, passengers and bystanders all do things that don't strictly conform to the rules.

Click to expand...

I'm perplexed now. Given that these are surely the areas where some of the most illogical and dangerous behaviour originates from, is this an argument in favour of passenger-less and bystander-less trains?

najaB said:

From the perspective of a computer (if were capable of forming an opinion), I'm sure you do. And it's not just drivers, passengers and bystanders all do things that don't strictly conform to the rules.

Click to expand...

Huh? The utter stupidity from car drivers who believe they can get a slight advantage over others is just not comparible with train drivers in any way, shape or form. Please explain what you believe train drivers do that don't strictly confirm with the rules?

najaB said:

Which makes me wonder if one of the biggest roadblocks to widespread driverless trains is the fact that there's still lots of human-controlled cars out there being illogical and dangerous.

Click to expand...

(snipped a bit )

EXACTLY - level crossings, bridge strikes you name it. Idiot road users et al not trains. After 1 incident the RAIB gets involved and produces a huge report and imposes huge cost - never happens after a road accident.

455driver said:

So train drivers routinely drive illogically and dangerously do we?

Click to expand...

Maybe not illogically and dangerously, but I'm sure that some drivers are tired, thinking about personal issues, need the toilet, hungry, thirsty etc. which could potentially distract them, whereas a machine wouldn't have this problem.

But I don't see how that means driverless trains couldn't be introduced - there is no need for any direct interaction with another train driver, unlike in a car when you flash someone to let them go first.

Dave1987 said:

Please explain what you believe train drivers do that don't strictly confirm with the rules?

Click to expand...

Because even the most conscientious train driver is human, and will make a mistake at some point. Most of the time those mistakes are minor - to the point that they have no impact on safety - but very occasionally they are not.

absolutelymilk said:

unlike in a car when you flash someone to let them go first.

Click to expand...

I suggest you read the highway code if you think that is what flashed headlights mean! :roll:

driverless trains would be easy to introduce if all they had to worry about was other trains, it is all the other issues which cause the problems, people running across tracks, level crossings etc. In the same way as driverless cars would cause no issues if all vehicles were driverless and there were no pedestrians cyclists or anything else to interfere with their operation.

philthetube said:

driverless trains would be easy to introduce if all they had to worry about was other trains, it is all the other issues which cause the problems, people running across tracks, level crossings etc. In the same way as driverless cars would cause no issues if all vehicles were driverless and there were no pedestrians cyclists or anything else to interfere with their operation.

Click to expand...

Which is the point I was making - the biggest impediment to autonomous road/rail vehicles is the non-autonomous environment they have to operate in.

Uber have withdrawn their cars not because there's a propensity for them to crash themselves but because human-driven cars crash into them.

philthetube said:

driverless trains would be easy to introduce if all they had to worry about was other trains, it is all the other issues which cause the problems, people running across tracks, level crossings etc. In the same way as driverless cars would cause no issues if all vehicles were driverless and there were no pedestrians cyclists or anything else to interfere with their operation.

Click to expand...

If a person runs across the track, or a level crossing in blocked, how does a train driver running at 90mph deal with the issue?

On the assumption that cameras can detect these objects (as they can with cars), what specific skills does a driver have that a computer can't do?

1) Detecting the object
2) Activating the brakes
3) Activating the horn

najaB said:

Uber have withdrawn their cars not because there's a propensity for them to crash themselves but because human-driven cars crash into them.

Click to expand...

This is a PR problem, not a technical problem. If another driver drives into you, what can you do that a computer can't?

If a human driver could anticipate it, what is the basis of that anticipation, and can that anticipation be implemented in a computer. If so, will that have knock-ons that cause more crashes?

455driver said:

I suggest you read the highway code if you think that is what flashed headlights mean! :roll:

Click to expand...

Drivers seem to be able to determine who is flashing to yield, and who is flashing to let you know they are there.

delete - what happened to dual post merging?

And of course, an automated train is just part of the whole system which is an automated railway. Fitting AI-based CCTV to every single level crossing, from the most basic foot ungated foot crossing up to a full barrier multi-lane road crossing, is going to happen and will give the railway more information about its state than any practical human-based system ever could. Cameras cost nothing. AI systems that can detect humans and other obstacles from video footage cost billions to develop and then cost nothing to produce. Indeed, if are Uber or Google or Tesla or whatever, the best way of making back those billions you've sunk into your AI system is to go around implementing it in as many areas as you possibly can.

paulweaver said:

If a person runs across the track, or a level crossing in blocked, how does a train driver running at 90mph deal with the issue?

On the assumption that cameras can detect these objects (as they can with cars), what specific skills does a driver have that a computer can't do?

1) Detecting the object
2) Activating the brakes
3) Activating the horn

Click to expand...

There are a large number of problems with your argument that have already been detailed earlier in the thread.