The myth of the driverless tube train | Hacker News Reader

The myth of the driverless tube train

londonreconnections.com

66 pointsrwmj143 comments

What they are actually seeking is

"...progress towards the conversion of at least one Underground line to Grade-of-Automation 3 (driverless, but with an on-board attendant, as on the Docklands Light Railway), subject to a viable business case and its statutory responsibilities."

They are only aiming to do one line, which just happens to meet a complex set of constraints outlined in the article. Doing the entire network would require a massive set of changes as outlined in the article (e.g. to rolling stock, platforms and signalling) to meet the safety requirements.

The requirement for an on-board attendant means that the driverless trains will still be at risk of strike action, in contrast to the line taken by some commenters here.

Note that the actual title is 'political myth'. The article suggests that the main reason for the change is to demonstrate political vision rather than, say, cost savings.

There are many times when the DLR trains have no employees onboard

There aren't, although I guess in a crowded DLR train you might not realise.

A DLR train won't go anywhere until the operator closes the doors, which they can do from any set of doors on the train.

Because the DLR is GoA 3 there doesn't need to be a plan for how central operations can co-ordinate rescue of a failed train without somebody at the scene, there will be an attendant, who is trained to drive the train, onboard somewhere.

Interesting. A very strange job then, most of the time you just close doors, and once a year you drive a train

The DLR closes most nights. When it re-opens, the first train of each day on each section is driven manually to confirm that maintenance crews haven't managed to leave anything on the track. Yes there are procedures to ensure they don't do that, but no procedure is foolproof.

Attendants may also choose to drive a train at other times, although doing so will generally be slower even if the computer is overseeing the journey (and the train won't accelerate above a crawl if the computer isn't overseeing) so it would be avoided in normal service.

The main job of these personnel are to assist passengers. So it's more a people job than train driver. Wondering if the Stratford destination is "the Shakespeare one"? They can help. Are the lifts working at the stop you need? They can check. Is it far to walk from this next stop to the museum? Does this ticket work on a ferry or only trains?

Does the automatic system limit where/when the manual trains can go to avoid crashes? What about syncing times multiple trains use the same track?

> The requirement for an on-board attendant means that the driverless trains will still be at risk of strike action, in contrast to the line taken by some commenters here.

That's true, although it might be a lot better than the current setup, if they're able to more easily replace that person (say, if they don't have to have specialised training).

I'm not familiar with TfL and their strike habits, but over here in strike-happy Paris, where conductors decide to stop driving on a whim, the automatic lines are practically never concerned [0]. They are, however, "level 4", as in there's no human present on board. They do require people at the control centre, though.

---

[0] I say "practically" because while I don't remember there ever being any reduction of traffic, even during major national strikes, I don't take the metro that often and I may have missed some. Though I'm pretty confident that it would have made the headlines.

Striking has historically been a significant problem on the Tube as well. That is partly because there is a limited supply of train drivers with the appropriate qualifications and if they strike then you can't just parachute someone untrained in to replace them. As you say, the risk of disruption to the travelling public from strikes might be reduced, though not eliminated, if the roles that still required human action had more limited responsibilities and required less skill/training. The DLR is a reasonable example of this in London, not immune to strike action, but historically less disrupted by it.

Note the weasel words. "Progress towards". Not conversion. Just progress.

Why would it matter?

You need so few subway drivers/conductors that does it really matter if you need one per train?

Taxis are probably average a driver per 1.5 ish customers. Subway/train is hundreds.

This reminds me of optimisation of non critical code. Sure its nice, but the cost/benefit is just not there.

Not just any 1 person, a driver, who must have years of training, if they don't turn up to work 1000's other people don't get to work. It's not so much one person as a single point of failure. If there were enough drivers with redundancy then it would not matter.

Won't be popular opinion but, one benefit of driverless subway is: decreased risk of industrial action disruptions.

In the city I live underground drivers hold ridiculous privileged (even compared to other underground workers) that allows positions to be effectively hereditary: children of drivers are more likely to get into the job than others.

> Won't be popular opinion but, one benefit of driverless subway is: decreased risk of industrial action disruptions.

Very true, though equally the increased risk of automated safety system stopping the train in the middle of a long tube tunnel because one of it's sensors got dirty or a rat triggered a sensor as it went splat on the tracks are things drivers can work around easier.

So if a perfectly safe and working train stops due to a sensor getting dirty quickly (which for the London tube is more likely than most places), with a driver there they could see and know that issue and safely carry on. Automation will go into health and safety mode and stop that train in the middle of nowhere needing human attention. So delays become longer.

I'd rather deal with planned outages, than longer drawn out automated random ones.

> Automation will go into health and safety mode and stop that train in the middle of nowhere needing human attention. So delays become longer.

In Germany, trains have a system called "Notbremsüberbrückung" / NBÜ. This system overrides the passenger emergency brake in tunnels... basically (i'm shortening it a bit here), as soon as the last carriage leaves the platform (in an underground) or the train can't fully stop before entering a tunnel, the e-brake is disabled, and only re-enabled when the train enters a platform / exits the tunnel.

There's nothing preventing automated trains from using the same principle. Additionally, tunnels can be retrofit with cameras or motion detectors, to further prevent something larger than a rat from getting in a dangerous position.

A number of tube lines in London also has this. Central Line, (parts of) Jubilee Line, both do if I recall. Those lines are "semi-automatic" where the driver is providing oversight and operates the doors whilst the train will automatically drive. Driver can override easily.

And a driver can when some one is blocking the doors not proceed and get the blockage removed quickly.

Pre covid in the rush hour this was very common.

That's more a job for the platform staff isn't it? Apart from making (several) announcements I've never seen a driver fix this.

This. London's tube system is essentially stopped a day or so a year due to strikes, which is pretty dramatic for a city of 10 million people who rely on it.

Most strikes affecting the underground have not been just drivers, so making the system driverless would be unlikely to change that.

True, and been over safty concerns/reduction in safty staff.

Though: https://www.bbc.co.uk/news/uk-england-london-19741282

Nicely sums up how the whole driveless tube/train aspect has been rumbling on for a while now.

London's transport strikes have been shown to ultimately benefit the wider economy, as people have broken out of their local maximums and found more efficient ways to travel around.

That you can identify a benefit doesn't mean it's outweighing the costs.

Perhaps the drivers should be paid a sum of money which closer reflects how much marginal value they add to the city, and this would convince them not to strike.

That sounds like a corruption problem that could be fixed via regulation, investigation, and prosecution.

This point is well addressed in the article. There are more staff members than just drivers.

Some tube lines in London have been closed lately due to unavailability of staff. Not sure if it’s drivers specifically but could be a reason to consider this.

Good point and down to cost, which is ironic as you have a system built and used mostly from people going to and from work and they are focusing upon reducing the wrong job in that equation is a very fair perspective.

Does make you wonder, if years/decades later they fully automate it, turns out cost way more than just keeping human drivers. Equally even if on paper for TFL it breaks even cost wise or cheaper, those people will still need jobs/income and may well be case of shifting that problem elsewhere in the governmental chain of things.

Germany for example has about 35.000 train conductors. Subways are just an obvious first step. Add renewable energy to the equation and you can build a fantastic system running 24/7 without any waiting for trains.

>Add renewable energy to the equation and you can build a fantastic system running 24/7 without any waiting for trains.

You'll need some storage or natural gas power plant as most renewable energy are intermittent.

Also, unless you have alternative routes, you cannot have a subway running 24/7 because maintenance needs to be done somehow.

You're missing the obvious solution of nuclear trains. I think it's self-evident that these would have no drawbacks. In the unlikely event of a disaster all you'd need to do is seal the tunnel and dig a new one.

rocket trains sounds better, and put them in vacuum tubes!

That would be revolutionary \s

Mainline trains is a non-starter for many decades, though, at least in the UK which still has a large number of level crossings, as well as stations with "fun" issues such as being too short for the longest trains. There is tech solutions in place to prevent the wrong doors from opening. Except when it doesn't work, where you really need someone on the train.

And yes, it does fail. Been on trains when the doors refused to open because the system failed to recognise the station. For ages, the trains also waited for ~30 seconds to open doors at London Victoria, supposedly because the system that recognised the stations used GPS, and the station concourse at Victoria is too deep for the trains to get a signal. Don't know if that is true or not, but in any case automating even door control was a long, complex project that still has issues years later.

Steps will keep being taken towards automating more of it, but there is decades old rolling stock, so any upgrades will take accordingly long to spread through the system as everything gets upgraded.

Isle of Wight for example, ran on London Underground 1938 stock trains until last year, when they started upgrading to refurbished British Rail Class 484 rolling stock built in the 1970's and 1980's...

With respect to renewables, not all of the UK network is even electrified yet...

Sounds overly negative to me. With those interest rates, we should really invest into any useful infrastructure project, especially when there are wins for or at least no cost to the environment. Let's generate some sustainable economic growth. We can solve the problems step by step. Starting at least brings us closer to a solution.

I'm all for investment (though not in London - I'd rather we invest in drawing business out of London, to create a more sustainable and scalable distribution of the UK economy), but that does not change the fact that these upgrades are incredibly resource intensive, not just in terms of money but also skilled staff.

There are many tens of billions of investment in prerequisites such as getting rid of level crossings and properly secure rail and upgrading stations and trains that'd be required before you can even think of upgrading mainline rail to driverless. For the London Underground, it means a rolling program of platform upgrades, signalling upgrades and rolling stock upgrades that'd similarly be in the many billions.

And yes, now would be the right time to finance it, but even if you start now this is just not happening in the next couple of decades, that was my point.

So by all means, it's worth starting to make the pre-requisite improvements, but it won't bring driverless trains on most of these lines anytime soon.

It sounds like this could be all done remotely, with cameras, and one worker physically present at each station, plus a roving maintenance team.

If we can fly drones through the deserts of Arabia why can't we drive trains remotely through the suburbs of Croydon?

Losing contact with your UAV is a lot less risky than losing contact with a runaway train.

Depends on where and how the UAV goes down.

I don't know if you suggested Croydon because you've seen me mention living there, or if you picked it at random or because it's one of the main locations with trains heading for Victoria, but Croydon is one of the highest trafficked parts of the UK rail network, with an interconnected set of some of the most complex junctions in the country, and a large rail depot that is hard to secure because of its sheer size...

It's if anything one of the worst places in the entire UK to start such a project.

The problem with what you suggest is that a ten minutes delay at East Croydon during rush hour, if you're lucky and a problem happens near a station, can easily lead to rush hour traffic from large parts of the South Coast to both London Victoria and London Bridge piling up with effects lasting for an hour or more. It's a horrible chokepoint. There are plans to fix that [1], but those plans also involve multi-year projects. In fact the planning has been going on for years at this point.

A similar issue is a problem for the London Underground - while there are fewer complex junctions, the margins are far tighter. A train that's stuck on the station for a couple of minutes can have knock on effects over a larger area.

All of this is solvable, but the costs involved in various improvements to mitigate them means it'd take a very long time before those investments would start paying off.

[1] https://www.networkrail.co.uk/running-the-railway/our-routes...

And we will call it 'The Snow Piercer'

The stated reason in the article is largely: to avoid strikes. Though the article also argues that this just pushes the power to strike on to other places.

They're not just trying to de-skill driving the trains though, they're trying to de-skill all aspects of running the network.

Driverless trains can run 24/7 without anyone needing to stay up all night

The trains discussed in the article will still require an onboard attendant.

Also, track maintenance still requires service downtime, which currently happens at night.

It's rail maintenance that stops trains running 24/7 in London. The rails are too old, often not fit for purpose and no one wants to pay to replace them. So every night they need polishing (literally), grinding, replacing, checking, etc.

>no one wants to pay to replace them. So every night they need... replacing

Not sure how that works. More seriously, I am unaware of the Tube tracks being unusually old, or requiring unusual amounts of maintenance because of their age compared to other comparable systems. Have you got a source on that?

There's also the fact that the deep Tube lines only have one track per direction, in contrast to the New York Subway where there are express and local tracks so trains can be rerouted via one set of tracks while maintenance is done on the other set.

Plus, of course, the equipment needed to do maintenance on the lines usually needs to be carried by hand through stations and into the tunnels.

You're quite right to ask for a citation. Sadly I can't give you one as the TFL rail condition assessments aren't public. Instead I'll point out that Metronet and Tubelines spend billions trying to dig out of the backlog and failed

https://en.wikipedia.org/wiki/Tube_Lines

>no one wants to pay to replace them. So every night they need... replacing

Yeah, stupid right?

Imagine a signal is broken (and TFL still used vaccume tubes for its signal systems as recently as 2010). So you have to reduce speed on the line. So now the centra-pedal forces on the corner after the speed restriction aren't enough to counter the camber of that curve. So the wheels on ever train grind down the railhead. So now you have the replace the rail (or polish it etc) much sooner than expected.

Failing to replace one part forces you to replace another part. Replacing that part uses up the time/money you have so you can't now replace the first part.

Take the central line anywhere between Bank and Stratford some time. The screaming of the wheels isn't just deafening noise. It's the actual rail being turned into dust because no one has invested a weeks closure and money to fix the rest of the infrastructure. So instead every night someone has to go check it and make sure it hasn't cracked yet...

4000 drivers at an average of around £70K per year is not chump change (£280M) + employer costs for staff. Out of a budget of maybe £5B per year, it is not insignificant but I would expect there are other areas where savings could more easily be gained.

Maintenance is crazy due to the short maintenance hours and therefore everyone getting paid enhanced rates to work 0:00 to 04:00. Significantly modernizing signalling and control would be expensive but would probably back back over time.

A few years back I was in a control room in a tube station and saw some steampunk looking machine that we assumed was related to the old destination boards on the station, which were long gone but the machine was still operational and left alone. Why? Because who would risk breaking something just to remove some random old piece of kit!

A few years ago my father did some consulting work for LU and he motioned that a lot of the pre war equipment had far less faults that the modern 1980's stuff.

It also had much less requirements and probably required more maintanence. You can think this type of claims everywhere; and pretty nowhere those are true.

For what it’s worth you ought to be careful arguing your position considering that GP has a second-hand account and you only have a general principle.

I see that you hedged your point, and I'm not even disagreeing with you at all. Just an interesting thing to consider.

They'd want it even if it cost the same or more. It's the ability to break strikes they're after.

Frankly, it is chump change.

You put your finger on the real issue with the London Underground - the maintenance needs which are onerous and prevent the trains from running overnight on most nights.

That kind of kit has reliability properties that are hard to imagine. There is a number of things that can go wrong and that gear basically guarantees that if you pass a green sign the track is free. 100% bug free and the only glitches there are is that if a bulb burns out, the system doesn’t run until it is replaced.

You can rebuild that kind of system but it’ll be exactly the same hardware behind a more modern looking panel.

It's political.

In the uk, train drivers unions are one of the last meaningful unions left and the right wing government hates that.

Also, London has a left wing mayor and a very good public transport system. The right wing government hate both the mayor (he's heir apparent to lead the party nationally) and the concept of public transport.

That's also the real reason private train operators were bailed out but public TfL was been hung out to dry...

> he's heir apparent to lead the party nationally

He’s not even an MP! He’s not allowed to be the leader.

I don't think he has to be? But he could easily be dropped into a safe seat. That's what they did for Boris in basically the same circumstances.

Maybe Heir apparent is a bit strong? He's not defonate. But I can't think of anyone else with the name recognition...

> I don't think he has to be?

Of course he does. He has to be a commoner MP.

> The leader and deputy leader of the Party shall be elected or re-elected from among Commons members of the PLP

https://labour.org.uk/wp-content/uploads/2019/04/Rule-Book-2...

> But he could easily be dropped into a safe seat.

Hmmm not sure how many of those they have available right now!

Thanks!

Agreed on the safe seats. Corbyn/Starmer has been pretty dire for Labour.

The rules are made by the party, and the party can change those rules. It has before and it will again.

In this case however I think it's much more likely that if Khan is the choice, room will be found for Khan to win a byelection somewhere. Labour has plenty of London MPs still, they've struggled mostly with the North, progressive policies sell very well among socially aware people inside the M25 but aren't popular among people for whom the whole point was that the Unions ought to run the country.

In lots of Labour seats in London it really wouldn't matter who you run. That's why Vauxhall's MP used to be Kate Hoey. Nice part of London, very gay friendly, many of the people there have never eaten meat, much less gone fox hunting and of course they opposed Brexit - and yet of course Kate was pro-hunting, pro-guns, anti-gay and fiercely favoured Brexit. Why did they elect her? Because it said "Labour" next to her name and they don't do research.

Sadiq Khan wouldn't even make most Labour members top 5. If you're going to look at mayors, Burnham is far more popular among Labour members.

(but of course same issue: would need to become an MP again first)

The Party can have whatever rules it wants. If it should even prefer for its leader to be a Hampster, it can do that.

As you might expect the political parties in the UK are organised differently according to their principles, so the Labour Party is an unincorporated association which only owns companies for purposes for which an informal group would be illegal (e.g. "People who come to our meetings" cannot legally own a plot of land, but "Labour Party Meetings Limited" could) while whatever the current Nigel Farage vehicle is (Reform? Brexit Party? Whatever he's calling it now) is a private company entirely owned and controlled by Nigel and his friends.

> The Party can have whatever rules it wants.

Not sure what point you're trying to make. Yes they can have whatever rules they want... and the rule they want is that the leader must be a commoner MP.

But Labours rules prevents him from being leader without being an MP.

Labour could change that, but if they were to be changed, it would not be for Sadiq Khan. Maybe for Andy Burnham. Khan doesn't have enough support among Labour members. Realistically it won't be changed for either, as it would cause too many difficulties.

Nor was the last Tory mayor. What ever happened to him?

> The right wing government

I'd love to know by what scale you consider the British government "right-wing"?

The normal EU /ROW one :-) of course US right wingers consider Maggie Thatcher a hard leftie.

I suppose Corbyn was the right wing candidate in the last general election then!?

Eh, what are you on about? Who says there has to be a left wing and right wing candidate?

Realistically, they are all pretty centrist - at their most radical, they lean slightly one way or the other.

> hung out to dry

£1.8 billion in loans/grants

Compared to 3.5bn in free money to private operators? Who were allowed to cut their services? When TFL lost 600m in the first 8 weeks?

Yes, London got hung out to dry as usual.

> Compared to 3.5bn in free money to private operators?

Sadly London is not yet more important than the rest of the UK combined.

I didn't say it was, but giving us nothing, taking 3.5bn for yourself and then implying we're the ones with a chip on our shoulders?

Maybe r/uk should kick us out as we're clearly holding you all back.

Interestingly, this is a situation created by previous Conservative governments. When the national rail system was privatized, lots of smaller individual train operating companies were created, each with a franchise to run different lines. They are incentivized not to cooperate with each other, because taking over another franchise, if another operator fails, doubles the size of your business.

So the employers were divided and the highly unionized employees were united, which meant that they did extremely well in negotiations. The membership overall are extremely loyal to their union leadership, because of how successful they've been in maintaining conditions and improving wages.

Yeah, that was a weird time in the uk. The tories sort of ran out of basic competence, even to rig the system to fit them. But then they went on "governing" for a few more years. See also "Black Wednesday" and other similar faff.

https://en.wikipedia.org/wiki/Black_Wednesday

Point of reference.

Barcelona has two fully automated lines (AFAIK, there might be some remote control) L9 and L10. They were built already with the idea of having no driver.

Nontheless when a strike is in effect their frequency go down too. AFAIK it was negotiated this way between the unions and the subway company.

Here are a couple of articles in Catalan about different strikes and the reduction of trains in all lines.

https://beteve.cat/mobilitat/vaga-metro-10-maig-amiant/

https://www.elpuntavui.cat/societat/article/5-societat/19353...

I think the Docklands Light Railway (DLR) model is fine. The author seems to be as ignorant of "machine learning" as he claims the general public to be of the concept of "driverless trains". For those unaware of what the DLR is, it's a 35 year old driverless train system in London with a human "train captain" who is basically like any other passenger but that can also operate the doors and override things if needed. You don't need machine learning and AI to drive a train on a one dimensional track, just a half decent team of Engineers to design something properly.

Yes the "train captains" can strike but they are far more replaceable than highly trained and very highly paid human train drivers. The evidence is clear, the DLR is far less affected by industrial action than regular tubes. Industrial action is so rare on the DLR that it becomes delicious news when it does happen: https://www.independent.co.uk/voices/comment/strike-proof-do...

You don't need machine learning and AI to drive a train on a one dimensional track, just a half decent team of Engineers to design something properly.

Most of the article was about the reasons "designing something properly" is an almost impossible task in this situation, once you take into account the reality of what is there right now, the goal(s) you might actually want to achieve, and the likely costs of doing so.

They speak of "myth", but there are many cities with driverless subway trains (e.g., in neighboring Paris, but definitely in more than the four "exceptions" that they mention). Instead of such grandiose philosophical statements, which are ultimately empty, they could make an argument based on facts and observations from the reality.

Tube is in the title, not subway. I think the author is talking about London’s system.

Yes, the author is totally talking about the London underground railway system. Most Brits - at least people in London and the surrounding area - call this 'the tube' and use the term 'subway' to mean underground pedestrian tunnels.

[Edit] We also never use the phrase 'the metro' to mean 'the tube' (i.e. the London underground') although some UK regions call have named their transport systems 'metro', e.g. https://www.nexus.org.uk/themes/custom/nexus/images/metro-ma...

"Metro" (short for "metropolitan railway") is the clearest generic, English term.

I always use it when I'm not in the UK, since subway is ambiguous, and underground / tube are more-or-less specific to London.

It's a long article, and I didn't read everything, but I guess this is one of the main points:

> Yet unless you are removing humans from the entire operation of the network, all you are actually doing is shuffling pieces, not removing them from the board. Station staff can strike, and do. DLR ‘Train Captains’ can strike, and do. Control centre staff can strike, and do. Signallers can strike, and do.

> As plenty of mainline railway franchisees have discovered, changing the relative responsibility of the roles doesn’t shift the balance of power between employer and employee. It simply alters which Union you need to negotiate with most

If something goes wrong in the middle of a 120 year old tunnel, 30 metres beneath the city, I want there to be a trained, responsible member of staff in place to manage the situation. It's that simple.

Lol, what are they going to do? Keep you calm? The situation would most likely be outside of their control.

Actually, yes, British people do rather respond well to calm people with authority during a crisis.

Yes, one of the reasons would be to keep the 1200 people on the train calm. It's important in an emergency for someone to be in charge of the situation. What's wrong with that?

Using a rough number for a yearly travel card, 1200 * £2000 = £2.4m yearly revenue from those passengers aboard that train. I think we can spare £60k for the driver.

Doesn't always work though. It might help but there are plenty of times where passengers take things into their own hands and follow survival instinct against the advice of a driver, both above and below ground.

During the July 7 bombings the train drivers led passengers out of trains and out exits. I would be glad to have someone familiar with that process on hand in the middle of that situation.

In train accidents, usually the first compositions are the most badly hit. The train driver is the most susceptible to die. You're better assisted by a remote operator...

That's in the very narrow situation of a head-on collision, something I don't think has happened on the tube in my lifetime. I'm talking about any kind of general emergency. I remember 7/7/05 like it was yesterday.

The 1975 Moorgate accident [0] was a single train accident that killed 43 people including the driver, who failed to brake his train as he approached a dead end.

[0] https://en.wikipedia.org/wiki/Moorgate_tube_crash

Anything from a more relevant era, perhaps one in which passengers aboard an underground train weren't allowed to smoke a cigar?

There's a list.

https://en.wikipedia.org/wiki/List_of_London_Underground_acc...

Assuming you've read the list, you will know that the answer is no. No head-on tube collisions in my lifetime or in any other relevant modern era of any definition.

> Anything from a more relevant era

Here you go: https://en.wikipedia.org/wiki/List_of_London_Underground_acc...

Incidentally, I narrowly avoided one of these accidents: The Kings Cross fire of 1987 [1] which killed 31 people. This was caused by a fire under a (wooden) escalator that hadn't been cleaned. I passed through about an hour before the fire started. I was returning from a conference in London. A friend and I had planned to have a quick pint before travelling back but he felt ill so we didn't. Otherwise we would have been right in it. This was before mobile phones so my wife saw the event on the news and was of course desperately concerned that I had been it.

[0] https://en.wikipedia.org/wiki/King%27s_Cross_fire

This subthread you are commenting on is about head-on collisions. There have been none since 1975 and I think you will agree that given that passengers on that train were wearing hats and smoking pipes, it's not particularly relevant to any kind of modern railway.

>> I'm talking about any kind of general emergency.

Hence me posting the list of general accidents

> The train driver is the most susceptible to die. You're better assisted by a remote operator

The driver didn't die in any of the recent incidents, so they would still be in place to manage the situation.

Interestingly, 7/7/05 isn't on the list, presumably because it wasn't an accident. That's exactly the kind of incident that is relevant in the context of this parent thread.

It would be interesting to where on the train is statistically the most survivable.

> Interestingly, 7/7/05 isn't on the list, presumably because it wasn't an accident

Yes, would also be interesting to know how many crime / related incidents are 'managed' by crew or station staff.

Even on a derailment, the driver seems also likely to die. E.g. https://en.wikipedia.org/wiki/Valencia_Metro_derailment

> I want there to be a trained, responsible member of staff in place to manage the situation

There will be. The article states that the trains will still have an onboard attendant.

As the article also states, there will be no point at all in moving to onboard attendants because the onboard attendants will strike just the same as drivers do.

Here's a list of places for you to avoid (granted, not all of them run in a tunnel): https://en.wikipedia.org/wiki/List_of_automated_train_system...

And I want one man for each 20 people "trapped" underground. Scratch that. One man for each 5 users. All of them paid minimum wage and doing absolutely nothing for 99.9% of the time. Otherwise it's unsafe. It's that simple. (/s).

This isn't really the argument though. Sure, we want a member of staff to help but they could be a £30K train manager instead of a £70K driver.

As the article explains, the train managers will strike just the same as drivers do, leaving us in exactly the same situation as before.

Not exactly the same situation. Less wages to pay, less pension debts to fund.

If there are still strikes then it is the same situation. Your plan solves nothing, for reasons both the article and I have outlined already. If we put you in charge tomorrow to execute your plan of moving from £70k drivers to £30k hi-vis vest-wearers, you would cause weeks of strikes, massive disruption and solve nothing.

One of cost of a few weeks of strikes to reduce cost of drivers by over half? Sign me up. Also the unions strikes are selfish in nature, they all team up to support one another to keep remuneration artificially high by holding the transport system to ransom anytime sometime tries correct it. The Tube Mafia would be a more apt name.

How badly paid do you want the people responsible for your safety to be?

A 30K train manager is still more than a 25K bus driver.

And there are a lot more road accidents than train accidents.

The last metro crash in Paris, quite a few years back, was due to the driver overspeeding in a curve for some reason.

Then you shouldn't use the tube, should you ? The argument that they are trained, responsible and competent is only used to justify their benefits. If you believe that, I have a bridge to sell you.

> There are over 100 metro systems in the world built before 2000. Of those, why have only four been converted to full automation?

Because the retrofits that the article details, no matter if installing platform doors, re-wiring infrastructure or replacing old carriages to actually enable consistent door positions, costs a lot of money, and public transport isn't at the top of public spending priority - car infrastructure is sucking up all the funds. Everywhere.

You want automated trains? Lobby your politicians to redistribute funding.

And FWIW the platform doors are really only required for scare-mongering reasons.

I don’t know if that changed by now but Lyon’s Line D went GoA4 in 1992 with no platform doors (for the longest time there weren’t even turnstiles, you could just walk walk / run from the entrance to the train with no stops).

The only accident I’ve heard of is someone falling on the train from a mezzanine above the carriage (they were drunk and apparently tried to jump to the quay, but managed to land on the train itself).

nuremberg underground also has not platform doors

Interesting that they are doing the "Waterloo and City line" first and for background, that is a isolated line/train that just goes from Waterloo station to the City central. I used that many times in the past and out of all the lines, was more civilised - would have commuters line up along the platform were the doors would appear once the train stopped and really a contrast to other tube/train stations.

So for a first trial/go of converting existing stock/track - that I will agree is a perfect choice.

Now when they get to the circle line - that IMHO will be the hardest and as a commuter.

I just hope they build sensors that can handle that environment. Many memories of commuting and having my ears and nose covered in a layer of black soot. But then that's best read up about here: https://www.wired.co.uk/article/london-tube-dust

The entire section makes it clear that this was put in so that DfT, and by extension politicians, get to claim they've pushed for driverless trains, and TfL get to claim they've engaged constructively about driverless trains, yet everyone knows nothing much will happen, and if anything does happen with will be Waterloo and City, which is in effect a token upgrade.

But they get to point out they've also pushed for Piccadilly, where the business case will point out lots of cash will need to be provided for signalling upgrades first, and it will be quietly put aside.

I think it depends on the stops. People naturally queue up at Canary Wharf for example.

So to keep TfL afloat the government has to give them money and as part of that they have to spend a large amount of money investigating something most people don’t think is viable. Seems like a good use of public funds.

Best of all, the money is a loan that London taxpayers will be expected to repay. Meanwhile private TOCs get free cash...

Aside from the strikes, another aspect of automated subways is worth considering.

Most subway lines in Paris have a few minutes interval between one train leaving and the next arriving. On the automated ones, it sometimes gets down to ~30 seconds.

Seems automating the trains enables much higher frequencies, probably by eliminating human reaction times and reducing margins of error.

Now, other countries have that without automation… but I’m not sure I’ll ever live to see the day when people over here run their trains as efficiently as it is done in, say, Japan.

One key factor in Paris is that automated trains are more "aggressive" in closing doors right on schedule where human drivers would tend to wait one or 2 seconds because you have a lot of people getting onboard.One or 2 seconds doesn't seem much but it can ruin the whole schedule because if you do it on 2 stations then it can add to 4 seconds, and then the next train needs to brake or even stop and lose even more time, and it cascades etc. until you can't keep a super tight schedule.

It has even happened to me once that the station and trains were so packed that the train came, people barely had time to get out and the doors closed already, with not a single person able to go onboard before the doors closed. Obviously feels bad but helps train keep schedule and it is a net positive in the end, but I suppose it is psychologically tough for drivers who would have to have that kind of constraints, they would really feel like a machine.

The Victoria line in London is human-driven, and has a frequency of 60-90 seconds at rush hour.

The Victoria Line was the first automatic (computer controlled) railway in the world.

The driver presses a button to close the doors, the rest is done by computer.

it seems down for me, archive link:

https://web.archive.org/web/20210609102038/https://www.londo...

I don't get why the train attendant has to be able to move the train. That's not required in the driverless systems I know. Either the train is unable to move and has to stop immediately, or it tries to drive to the next station, right? The only job then for the attendant is to guide the passengers to safety and watch out for unruly passengers. I know Nurembergs underground quite well and it just works. No platform doors and no attendants.

For the exceptional case that the train is unable to drive itself without an accident occurs a special team should be enough. I don't remember that it happened for the nuremberg underground.

Nuremberg is a small town where emergency personal, which can drive the train can be deployed quickly to all locations, without having lots of on call staff ready. Deploying this in London is a different challenge.

However I agree that with remote control the need should be mostly avoidable.

Well, that's true but then the emergency personal could at least just be spread around in the city. I don't think the response time should be that critical, if the train is burning nobody is needed to drive it to the next station, assuming staff is on the train that knows the emergency procedures. They should only need it for technical problems.

For some scale:

The shortest line in London (ignoring 2.5km Waterloo &City) is Victoria line with 21km.

The longest line, U1, in Nuremberg is 18.5km and still doesn't run automated. They started with U3, which is running 9.2km. Even with only two ready rooms maximum distance is 2.5km, which can be travelled in a few minutes in case of emergency (using a car and emergency signals it's less than 5 minutes, but of course getting underground and from nearest entry point to train will take longer)

https://en.wikipedia.org/wiki/List_of_automated_train_system...

https://fr.wikipedia.org/wiki/Liste_des_syst%C3%A8mes_de_m%C...

I don't understand why you can't have PEDs for multiple kinds of rolling stock. That is stated but never qualified. It seems like a tractable design problem considering there are a limited set of types of train you are serving.

It's not impossible, but it's difficult if the doors are different sizes and have different spacing. There are a lot of design compromises.

Here's an article about it (why Vancouver can't have PEDs): https://dailyhive.com/vancouver/translink-skytrain-station-p...

This is not really a big issue for London, though. The vast majority of London Underground platforms only have one type of train serving them. The few exceptions (eg: parts of the Piccadilly and Bakerloo lines) are mostly above ground and not the busiest parts of the network, so you could just skip the PEDs there and use manual door closing.

Food for thought: The Skytrain in Vancouver was "GoA 4" grade in 1986.

London's DLR is the worst offender. The conductor relegated to manually closing the doors through an override switch.

So, this article is about that GoA 4 is not reachable and thus you should not call it "driverless tube"?

Is that right or I am missing something?

When he writes that they looked at the system that Siemens provides, then I think I know the system, which is considered. The very same trains system is actually running here in my town for more than 10 years now. I think it should be deployed originally by 2006. But as any project in Germany, this was late to and as such was deployed in 2008.

Yes, it is not GoA4 but GoA3. Yes, it was a technically challenging. I personally know people who where involved in the project and specifically in the functional safety aspects. The first sensors were installed somewhere around the end of the last century on the lines that are run by the drivers to get experiences and to improve the safety system. At that time the system was somehow unique, because it was specifically design for retrofitting existing tubes. It is also designed from the beginning that driverless trains and trains with drivers can run in parallel on the very same rail, if needed, because some lines may share some parts of the rail system.

Yes, it is (only) GoA3, that means there is a central control room. There are actual people sitting monitoring and "supporting" the automated decisions. They are supporting in a sense, that if some situation occurs the system does not have an automated decision to, the operator can view a sequence and decide what to do. If the operator decides, that he is unable to decide remotely, he can sent an onsite operator. That means there are no (operating) people on the trains itself, they are only required at the platform.

This gives the benefit, that the number of trains currently running can be decreased or increased depending on demand without having the need for having drivers on standby. That makes personnel planing much more easier and can save costs. This is paid for with increased costs to run the technical system.

This is actually not cheap as well. Here in Nuremberg there are currently 3 lines. Line 1 is the oldest and still run by a driver. Line 2, the second oldest, was originally run by a driver but converted to be driverless. Line 3, the newest, started completely driverless. Line 2 and 3 largely share the same rails. Line 1 runs on their own rails. Only when maintenance is happening, then all lines share in some parts the very same rails. Because of Line 2, the system is mostly retrofit. The city decided a few years ago not to go driverless for line 1, because of the costs involved for retrofitting the complete line.

Besides the costs, there is also a benefit for the drivers to be driverless. I have some friends who used to be drivers. They say, this job is very very challenging when you are the whole time underground and basically most of the stations looking almost the same. After the second round you loose your mind and concentration, so that sometimes they announced the wrong way they where going. For the safety aspects that is even worse.

So going driverless or not is not a myth as such, only if you consider GoA4 as the only possible driverless. But the decision for or against is very complex and even can increase safety for the passengers, because the automation does no get tired.

The unions opposing it actually indicates that is a progressive idea.

I'll note that the actual title is (now?) "The Political Myth of the Driverless Tube Train" not merely "The Myth of the Driverless Tube Train"

And I'm sure that as a claim about politics "myth" may feel about right, but it clearly doesn't work as a technical claim and the article seems to hedge about whether that's what it wants to say.

It pokes Boris, former mayor (a role for which I felt he was well suited) and present Prime Minister (a role in which he's incapable) which is politics, but then it also talks about Grades of Automation, rolling stock upgrades, signalling, and Platform Edge Doors, which are technology not politics.

The sort of contradictions shown are effective for making Boris look foolish (compare Thatcher's pun "You Turn if you want to. The Lady's Not For Turning" to Boris having difficulty remembering what he answered on the same question yesterday...) but they have the reverse effect on matters of fact.

For example, we're reminded repeatedly that the London Underground is very old. But then we're also told, for a different reason, that the Docklands Light Railway is old too. A GoA 3 system, right there in London, operating very safely for decades. Complexity, we're told, is the enemy of automation, but the DLR has only become more complex, adding more branches, longer trains, newer service patterns.

Of course automation of the London Underground is also very old, it had one of the first GoA 2 automated railways in the world. When I was a child the Tube still had guards. If you're a New Yorker this probably seems unremarkable, but that guard is completely unnecessary in ordinary service there's barely enough work for a driver, let alone also needing a guard, and this role was abolished in the London Underground last century.

Safety considerations will make GoA 4 operation of the Deep Tube very difficult to deliver cost effectively in my lifetime. But in practice the objections to automation aren't about some hypothetical unsafe GoA 4 proposal, they're mostly about trying to ensure the RMT retains power in the capital by delaying real, practical projects that would mean London doesn't need so many of their extremely well compensated members.

Turnham Green is a particular red herring. If you've used Turnham Green you may not even be aware that Picadilly trains stop there, because in practice for most purposes they don't. They pass through but most don't stop. A future upgrade might make it practical to have all Picadilly trains once again stop at Turnham Green some day. Or not. Hardly the sort of show-stopper it's presented as in this piece.

In other places where there could be a clash, Selective Door Operation might resolve it. SDO is already used extensively on the London Underground for S-stock.

The point of bringing up the technical issues is not that they are unsurmountable, but that they require political will to fund upgrades.

The RMT would be easy enough to deal with if there was political will to offer them a suitably lucrative agreement for their present drivers to accept that their numbers will steadily decline.

It's a reasonable sounding argument but I don't find it totally convincing.

Firstly, a big deal is made about the need for PEDs. The justification given is vague - essentially that the regulator would require them, because it's now "known" that PEDs are required, and the DLR doesn't have them because, well, really it should but we don't care enough to actually upgrade it. But new stuff must.

This looks like a case of creeping safetyism. If PEDs are really required for the acceptably safe operation of a railway then clearly both the DLR and LU must be considered unsafe, because the number of passenger-on-track or passenger-in-door situations that only a driver can solve seems very small indeed. In fact it's surely the case that platform attendants are better placed to observe these problems as they can easily look along the entire length of even a busy platform, whereas a driver must rely on a mirror. Saying automation "requires" PEDs is a bold statement given the successful and apparently safe operation of the DLR for decades, yet no evidence is provided beyond the opinion of the ORR, which as a regulator is hardly unbiased - they have zero incentive to make automation happen and every incentive not to.

The arguments about strikes also looks superficially reasonable, but I'm not sure it holds on closer inspection. Drivers have a lot of strike power because there are a lot of them and they require a lot of training, so they are impossible to replace as a bloc. Presumably a well automated railway would require only a much smaller number of people in the control room, and those people would require less training, as the software will be carefully checking their inputs and doing the bulk of the work. So if they did decide to strike then you could probably have teams in reserve or call up the manufacturers of the equipment and ask them to send over some of the engineers who built it, to temporarily operate the equipment and train up a new team of control room staff. The striking power would surely be much lower.

Even for physical maintenance staff, yes they can strike, but they can't immediately paralyze a live railway like drivers can. There's a limited amount of time for which a railway can operate without regular maintenance, and if the equipment is already in good condition, it seems likely that the railway can outlast the strikers. Also a lot of physical maintenance is not especially complex to train people on, so alternate workers could be found more easily.

The main reasons the Tube isn't being automated are clearly not that it's some sort of insurmountable challenge, at least not according to this article, but rather than automation has been conflated with general wide-reaching safety and simplification upgrades, based on a dubious argument that drivers are much safer for passengers than any automated system can be. I'm not at all sure that's true.

I'm sure an automated system can be built that is safe enough - the Elizabeth Line demonstrates that. But the reality is that upgrades cost a lot of money, and nobody wants to spend that money. That is what it boils down to.

With respect to attendants, they're mostly there on the busiest times at the busiest stations. Most of the time most platforms are largely unattended, and the driver is the only person present to take action unless you want to wait for the train to alert station staff to a situation that requires intervention.

If the solution to automate is to put in place attendants permanently on every platform I'm not convinced it'd serve much purpose.

With respect to the DLR, it's less than a year since the last time someone was killed by falling onto the tracks. While the DLR overall has a great safety record, it's far less crowded than several of the other lines. I don't think you'll get anyone willing to sign off on overriding the regulator there, given people have called for safety upgrades of the DLR too despite the few fatal accidents.

I think the idea is that attendants are much easier to train and thus replace quickly than drivers are. All they have to do normally is push a button saying "it's safe", more or less. Perhaps very rarely they'd need to go into the tunnels to reach a stuck train, but that's exceptionally rare. In a squeeze people could be trained to do the normal tasks in a day or less.

DLR safety - aren't you agreeing with me? The DLR is not perfectly safe but in reality people accept it. Nobody is clamouring in London for the DLR to be shut down on the grounds that it's a death trap. London society has made its peace with the level of safety the DLR provides. Would more safety be nice, of course. Is it actually a hard requirement that blocks other upgrade projects? That's not at all clear and the argument you're making here is the same circular one the article makes - automation upgrades require safety upgrades because the regulator (+other undefined people) say they require safety upgrades, and the regulator says it requires safety upgrades because automation requires safety upgrades. But the link between these things seems rather brittle.

On your first point about PEDs, I think you've missed that there's an element of safety they offer that prevents non-accidents (often referred to as a "person under the train" or other euphemisms). These definitely happen on the DLR [0] , along with many other lines. Among other obvious repurcussions of these incidents, these often result in a line closure of at least a full commuting slot.

[0] - multiple incidents on this FOI request from 2012, and I'm sure there will be more recent data available https://www.whatdotheyknow.com/request/persons_under_trains

Reply on news.ycombinator.com