Hacker News new | past | comments | ask | show | jobs | submit login
TfL abandons plans for driverless tube trains (ianvisits.co.uk)
29 points by edward 12 days ago | hide | past | favorite | 74 comments





I’ve never really understood the argument for driverless tube trains.

Take the Victoria line doing approximately a full lap of the circuit in an hour, north to south and back again.

That one train with one driver has a capacity of 260 seated, 1000 in total. People get on and off all of the time so the actual usage can be far higher, most people do a few stops or half the route max. Based on riding the tube I would guess that the average train has at least 1000 users an hour e.g. 500 northbound, 500 southbound using this example.

If the train driver gets paid £100 an hour (feels high) then it’s 10p per person per trip or approx 30 seconds of work at the minimum wage.

The entire thing feels like optimisation for the sake of it, I’m happy for my tube driver to get 10 bloody pence from my fare. Even if my assumptions are wrong then say it’s 20p or 50p or whatever that feels totally reasonable.

A taxi is completely different because in that case the hourly rate of the driver is basically the entire fare, you are pretty much paying for the time of one person 1:1.


Your assumption is missing a few important points. It's a weird, stressful and depressing job, while also being inconvenient. People need transit to function early in the morning, late at night, during holidays. Having enough train drivers to serve all of those comes at a cost, and is sometimes quite hard because people aren't lining up to stay stationary and press buttons in a tunnel from 4am or during Christmas/New Years (even if they don't personally celebrate it).

Automated driverless trains allow for higher frequency, reduce the need for a shitty job, and allow operations in inconvenient for humans hours and times.


>people aren't lining up to stay stationary and press buttons

I've always gathered there's quite a waiting list to get in. This from a tube driver:

>You need to find a role within TfL. It doesn't have to be customer facing, but it has to be something within TfL. Once you're employed for 6 months with TfL, you can apply for Train Operator, when the job is available. If you pass all the interviews and assessments, you'll be entered into a queue before you start training. This could be a couple of years. In short, yes it is pretty hard to get the job of tube driver. https://www.reddit.com/r/LondonUnderground/comments/ti7skk/h...

Mostly because there aren't that many lowish skilled jobs paying £70k for a 35 hour week in London.


And the Victoria line is already driverless in the sense that all the "driver" does during normal operation is operate the doors.

More or less the same as the DLR, except the Victoria line has a permanent driver's seat/cabin where as the DLR operator can move throughout the train.

(Personally I'd rather be a DLR "operator", at least you're on your feet much of the time and getting some exercise. But I bet the Victoria line "drivers" get paid more!)


> "I would guess that the average train has at least 1000 users an hour e.g. 500 northbound, 500 southbound using this example."

Yes, seems not far off. There's something close to 500 scheduled Victoria line services per day in each direction (less on Sundays, and not all services will operate if there's disruption/delays). Annual Victoria line ridership in 2019 was 302 million, or 827k per day on average. So, roughly, 827 passengers per train service on average (but not all at once!).

Each service takes around 35 minutes to do the full line from Brixton to Walthamstow, plus turnaround time at each end. Some services turn back early at Finsbury Park or Seven Sisters. So each train will be doing a bit more than 1 service per hour.


The big thing is not the cost of the driver it is the capacity. As the article says:

"The main benefit of driverless trains is capacity, as the computerised control systems can allow trains to run closer together, which increases passenger capacity on the railway."

However as the article says this isn't as big a gain on the Underground since it is pretty optimized already and gains would be minimal for the cost.

For most metro-type rail systems driverless is the way to go though, it can offer increased capacity and lower running cost (especially off-peak).


> That one train with one driver has a capacity of 260 seated, 1000 in total. People get on and off all of the time so the actual usage can be far higher

You're just using the wrong units. You'd normally measure transportation in passenger-miles. (Or, as the case may be, passenger-km.)


The relevant figure in my mind is per entry and exit to the system, in fairness the same person usually will take two trains so that’s two drivers to pay.

Basically, per passenger served.

Per mile basis is weird in a metro system because of distortions like how a mile in the centre is both slower and far more in demand than a mile at the edges.


> Per mile basis is weird in a metro system because of distortions like how a mile in the centre is both slower and far more in demand than a mile at the edges.

What's the problem supposed to be? Passenger-miles automatically take that effect into account. They go down when speed goes down, and up when demand goes up.

> both slower and far more in demand

By the way, it sounds like you think this is something strange about a metro system, but higher demand means lower speed for every form of transportation.

> Basically, per passenger served.

This is a mistake; the guy who crosses the whole city has gotten more value out of the metro system than the guy who goes one stop.


I think that you're trying to force metro travel patterns into a normalized format for comparison with other modes that doesn't match actual usage patterns.

If we have to do that, then for my original post, the relevant factor is passenger minutes/hours travelled because the driver's cost scales with time not distance.

My cost is not proportional to either, though. If I go to my nearest station and go the next tube stop or if I go five tube stops the fare is the same. If anything it's inversely proportional e.g. the longer I spend on the train, the less I contribute to the driver's hourly wage.

The amount of time I spend and the amount of value I derive is entirely irrelevant to any of this.

I'm not especially interested in any of this beyond my original point that the cost of the driver is trivial on a per passenger basis, my back of the envelope calculation gives the same result (doesn't matter) even if you 3x or 0.33x the outcome.


> "the guy who crosses the whole city has gotten more value out of the metro system than the guy who goes one stop."

Not necessarily. The value you receive from a transportation service is mostly a function how badly you want/need to be at the destination, as well as things like how comfortable and fast the ride is. If anything, the guy travelling a shorter distance got more value because his ride took less time!

Or to put it another way, there is little value to me in travelling a long, laborious route all the way across the city if I don't have a pressing need to be at that destination. But if I need to be at work in 30 minutes I get huge value from taking the metro to the nearby station, even if it's only 2 stops away.


Your analysis doesn't work. In your terms, the guy who crosses the city has paid more (by spending more time) for what he received, and so you still assess him as having valued the trip more than the one-stop guy.

> But if I need to be at work in 30 minutes I get huge value from taking the metro to the nearby station, even if it's only 2 stops away.

Not really; as you just noted, the subway is a very slow method of covering distance. If you're in a hurry, it's not going to beat the alternatives.


> "the guy who crosses the city has paid more (by spending more time) for what he received, and so you still assess him as having valued the trip more than the one-stop guy."

No, I think you misread or misunderstood what I wrote. My point is that the "value" derived from a trip is _not_ proportional to the distance travelled.

> "Not really; as you just noted, the subway is a very slow method of covering distance. If you're in a hurry, it's not going to beat the alternatives."

In London, the tube or train is often the fastest way to get somewhere, particularly at rush hour. But if I place a high value on comfort, perhaps I'd consider that I get more value from a taxi ride even if it takes the same amount of time: the tube can be very hot and crowded! I'd then have to consider whether the extra value gained is enough to justify the higher cost of the taxi ride.


Salary is just a part of the cost like other people hinted at. Benefits, recruitment, capacity management, health and safety, performance management, training, all other sorts of HR processes.

No doubt the RMT will take the blame for this but as the article points out:

> It’s now been confirmed that the study reached the same conclusion that every other study into the issue has already reported — it’ll cost an awful lot of money for very little benefit.


I wonder if they compared with Paris which has similarly old infrastructure, but managed to automate the two most heavily used lines (1 and 4).

I wish they'd kept conductors on buses but that ship left 3 decades ago. Some of my fondest memories are the conductors on Melbourne's ageing W class wooden trams.

Trains feel like something where drivers and conductors and platform staff are a social good. We're beyond cost at this point, it's about public utility.


I think the combination of driverless trains and automated platform screen doors is the gold-standard for new metro-like systems. The new REM in Montréal is able to have very high frequencies with much lower cost, so transit users should be able to get more transit for the same cost and the municipality isn't worried about the burden of high operating costs for the next century. The platform screen doors are great too, anyone who's taken the Metro enough has had to wait for hours because someone has fallen onto or committed suicide on the track, which feels like such a sad problem to have when the solution exists.

I do like Asian metro systems I use with screened entry to trains. But I also like seeing staff. Porque no los dos?

Vancouver has driverless trains, but you pretty frequently see staff at the stations.

Oddly enough, you almost never see them on the trains unless someone has pushed the silent alarm.


Skytrain is quietly amazing, such a joy to use compared to even other train based metro systems in my opinion.

Fair enough though one doesn't interact much with a subway driver.

On the flip side, I've seen "funny" videos from Japan where there is staff at the stations who push passengers in to fit everyone as much as possible?

I used to think that was crazy until I started using the metro every day. There will be a hundred people packed by the doors and the few people in the middle of the car don't think to compress a little. Suddenly I wanted those staff to come in and make the ride more equitable!

Also the ratio of driver to passenger especially compared to say a taxi, is negligible so the amortized cost is basically zero. That said, they’re not exactly useful on trains, you don’t chit chat with the man in the armored cab.

You would think so but e.g. airlines pushed quite heavily to reduce pilots from 3 to 2 and occasionally lobby for 1.

> Airlines are trying to squeeze every penny out in a depressing race to the bottom because they couldn't really differentiate themselves and travelers overwhelmingly just pick the cheapest flight

They also removed the roles of radio operator, navigator, flight engineer and third officer without compromising safety or capability [1]. I'd want to see studies. But single pilot + relief crew should be safe enough to start with for long-haul flights.

[1] https://en.wikipedia.org/wiki/Aircrew


Single pilot is fine as long as nothing goes wrong. When a problem occurs then workload can get really severe really fast. You need one pilot flying and the other running checklists and troubleshooting. Even if there is a relief pilot onboard they might be sleeping and unable to make it to the cockpit fast enough to help. The higher risk level might be judged acceptable for cargo flights but I think it will be a long time until automation makes it viable for scheduled passenger flights.

But then there are so many crashes where having 2 pilots was kinda the cause of crash...

How many crashes?

> When a problem occurs then workload can get really severe really fast. You need one pilot flying and the other running checklists and troubleshooting

This argument was made for each of those now-obsolete roles. I'd want to see data before concluding there is a benefit to a second person on checklists. In reality, that advantage would only manifest if the captain is inexperienced, self-check systems have failed and the radio/remote ops have failed.

> Even if there is a relief pilot onboard they might be sleeping and unable to make it to the cockpit fast enough to help

Can you name a crash which occured because the FO was in the bathroom?


> Can you name a crash which occured because the FO was in the bathroom?

Germanwings 9525. Can’t discount human factors, especially in an industry where human factors are very frequently the root cause of incidents.

More importantly I can name any number of flights where two pilots became task saturated and flew a plane into the ground. Or flights where a single pilot used bad CRM to force the other pilot to do something stupid, even though they both knew it was stupid. Both of those would be more common without a second pilot, general aviation is proof of that.


> Germanwings 9525. Can’t discount human factors, especially in an industry where human factors are very frequently the root cause of incidents.

Not a great argument for a second pilot unless the second pilot never goes to the bathroom. The human factors involved in Germanwings 9525 didn't randomly occur during the flight, they specifically occured while the co-pilot was in the bathroom.

> Both of those would be more common without a second pilot, general aviation is proof of that

Between the reliability of piston engines versus jets to the experience of someone with a PPL versus ATP, this comparison includes far more variables than number of pilots.

> I can name any number of flights where two pilots became task saturated and flew a plane into the ground

I can do that for any value of N. The question is what the ratio is between N = 1 and N = 2, and how that has changed with evolving automation.

> where a single pilot used bad CRM to force the other pilot to do something stupid, even though they both knew it was stupid

Not really an argument for a second pilot either...

I'm not saying we have the evidence to say a single pilot is sufficient. I'm arguing that nothing you've brought up shows it's critical. That said, I'm sure we'll wait until another aviation authority pioneers the way. American aviation ceded the first-mover position decades ago.


> Not a great argument for a second pilot unless the second pilot never goes to the bathroom. The human factors involved in Germanwings 9525 didn't randomly occur during the flight, they specifically occured while the co-pilot was in the bathroom.

One of the changes from that incident is that there is always a second person in the cockpit. If a pilot needs to leave the cockpit for some reason, a second crew member must come in and ensure that the pilot flying doesn’t lock the other pilot out.

As to the rest of the argument: the whole point is that two pilots is safer than one by a very large margin. We eliminated engineers, radio operators and navigators because we know have systems that do that better than a human. We do not yet have systems that do the job of pilot as well as a human. Commercial autolanding has existed since the 1960s but there are still situations where the pilot has to land the plane. We aren’t there yet.


> Not a great argument for a second pilot unless the second pilot never goes to the bathroom. The human factors involved in Germanwings 9525 didn't randomly occur during the flight, they specifically occured while the []pilot was in the bathroom.

While that's true, it is a good argument for a third pilot.


> it is a good argument for a third pilot

To stop one crash in decades amidst hundreds of thousands of flights per day [1]? If you place that high of a value on a human life, there are better optimisations to do first at much lower cost.

[1] https://financesonline.com/number-of-flights-worldwide/


That's a different argument about the third pilot.

I'm just pointing out that while my parent was correct that the argument failed as a justification for two pilots, it doesn't fail for any other number. It's similar to the proof that all horses are the same color.


> while my parent was correct that the argument failed as a justification for two pilots, it doesn't fail for any other number

It doesn't fail as much as not work. One can always come up with hypotheticals justifying more redundancy. Whether that's worth the tradeoff requires actually looking at the facts.


One of the changes from germanwings is that a third crew member must remain in the cockpit in the event that one of the pilots leaves for any reason.

> Can you name a crash which occured because the FO was in the bathroom?

I don't think that's a strong argument. The only time a pilot goes to the bathroom is during the most boring portion of the flight when everything is going exactly as expected. Planes so rarely have excitement occur at that phase of flight that the odds are really low of it happening at the exact moment one of the pilots is using the head.


> only time a pilot goes to the bathroom is during the most boring portion of the flight when everything is going exactly as expected

Planes don't have something go wrong to the point of running checklists in an expected way.


You’d think they were already investigating the use of automation in emergencies. Like going through checklists, debugging situations in real time, etc…it seems like an area where AI could actually be useful beyond current auto pilot/landing tech.

It's just a bunch of if conditions, no need for AI. The problem is that sometimes something not planned in those if conditions happens, and then you need deduction and high knowledge of physics and the aircraft systems.

> something not planned in those if conditions happens, and then you need deduction and high knowledge of physics and the aircraft systems

This is an argument for offloading checklists to an automated system. Not having more pilots to run more checklists.


On modern planes there is a very high degree of automation already. But it's a fine balance, because too complex automations can be confusing or unpredictable to the pilots. There are a few crashes due to pilots not understanding/fighting with automation.

You just killed a Copilot

Airlines are trying to squeeze every penny out in a depressing race to the bottom because they couldn't really differentiate themselves and travelers overwhelmingly just pick the cheapest flight.

Public transit is quite different. It's not as opex-intensive as flying, it's usually subsidized by the state, and it's often pretty much a monopoly.


The numbers are pretty different, though. A subway driver might make $50K/yr and the train has a thousand people, with short runs at high frequency. A long haul flight will have 300-400 passengers for hours and 3-4 pilots each making $250K/yr.

On the flip side, a subway/bus ticket is usually around $1-$10, while an airplane ticket is $100-$1000.

That's way more than the end-of-career salary for a pilot at a major UK airline, around £140k. Newly-qualified pilots at Ryanair make around £23k.

It's a bit more lopsided in the US. A regional carrier pilot makes peanuts (due to how they are paid, only when wheels are up, there have been pilots effectively paid lower than minimum wage). At regular carrier like American/United/Alaska, pilots make waaaaay more. A captain may make half a mil. If you can suffer through the regionals and make your way up the ranks, it eventually pays off.

The conductor wages over the system as a whole are a large part of operating costs. Another issue you can have is a lack of conductors, which is a big issue in the Netherlands apparently.

More to the point, staffing isn't setting the price of transport, it's subsidised almost everywhere. Sure. It's a cost. The cost:benefit here is more than just apparent budget impact because public transport is a utility function.

I don't disagree. One thing that RMTransit on Youtube talks a lot about is that transit is a lot more expensive to build in the Anglosphere of the US, Canada, Britain, and Australia, and so despite these countries investing a lot of money into it recently, there's an order of magnitude less return on investment than similar projects happening in the rest of Europe and Asia.

Return on investment not being profit for the transit system, but moreso a more extensive, faster network for riders.


I used scotrail for the first time in years as a casual return visitor to the UK from Oz. It was ok and on time and clean, and the QR code on/off experience was fine but the confusion of choices trying to book.. complete nightmare. Privatisation has only partially been unwound and multi app multi provider timetable and ticket exposure to casual users is away with the bees.

if you look at the accounts/statistics: staff is the majority of the cost

the official figures here are however misleaading as their "staff" numbers only include those directly employed by TOCs and Network Rail

under "other" there's the vast army of contractors, including cleaning staff, track maintenance contractors, train maintenance contractors, private security, and so on

what's left is fuel, interest payments, depreciation, premium payments to government and some capital expenditure (not much...)


Platform edge doors could be done half-height style without massive rebuilds of platforms. I’ve seen the half-heights in Japan and Malaysia. That also then avoids the need for super precise alignment which is an issue for platforms supporting multiple lines and train types.

None of the above would enable driverless trains. It’d just go towards accident reduction —maybe suicide too, depends on the individual’s determination.


In London there are platforms which are shared between lines that use trains with completely different proportions, car lengths, and door spacing. Even with half-height doors it would be difficult to come up with a system that could handle both the small deep-tube trains of the Piccadilly line and the large, sub-surface trains of the District line.

Other metros around the world that have retrofitted PEDs had the advantage of having fleets of trains with standardised dimensions.

Of course, this wouldn't stop them selectively installing platform edge doors at stations that do not have this issue. The most modern lines (central section of the Elizabeth line, and the eastern section of the Jubilee line) do have PEDs.


Step one, standardize the doors. Step two, wait 30 years until the non standard door train is replaced.

Or just put 2 doors on the barrier

that works if there are only two different door layouts. as you get more you run into the door support for on train being in the wayof a door for another. Even for two layouts sometimes you need platform doors very wide doors two cover two.

Yes I acknowledge the differences.

We could still install PEDs with imprecise doors. Ie more doors than required. Larger doors than required. The main thing is for the door to only open while a train is in front of it, and all the train doors be unobstructed.

Ie won’t hurt much to have a PED open to the train wall or opens double the size of the current trains’ door.


> "Ie won’t hurt much to have a PED open to the train wall or opens double the size of the current trains’ door."

I imagine there would be safety issues with that approach since it might create an opportunity for people to get trapped between the platform doors and the train. No sense solving one safety issue just to introduce another one. You'd also need to be sure the platform doors didn't take any longer than the train doors to open. Anything that increases boarding & alighting time, even by a second or two, would likely be rejected in London.


There’s already plenty of instances of the JLE doors opening out of sequence, misaligned etc. hasn’t stopped TfL from using them.

> Platform edge doors could be done half-height style without massive rebuilds of platforms.

Make them chest height. There are no words for how much I despise guardrails that reach to below my center of gravity. I want guardrails that will stop me from tumbling over a cliff, not guardrails that will make sure that if I do go, I go headfirst.


> Despite the claims that it would prevent strikes on the tube, the reality is that it wouldn’t, as driverless trains would still have staff on board

Why would they? Paris lines 1,4, 14 and soon 15 have zero staff onboard


Same in Nuremberg. And they don't even have automated gates. It seems to work well.

You obviously want a human competent enough on board to manage the train, even if everything is automated.

they could start with the waterloo and city

I can't believe that would cost billions given it has two stations and is fully underground

good leverage against the unions too...


The first two paragraphs of the article specifically talk about a study done on Waterloo & City coming to a negative conclusion.

How is it good leverage against the unions? Staff is required, even on driverless trains, and in the stations, and on maintenance. The driver is just a tiny part of the labor involved.

> Staff is required, even on driverless trains

Why? Is that some union requirement? Staff is absolutely not required on driverless trains. There can be at stations and of course for maintenance, but there's no reason to have staff in each train.


The DLR is only partly staffed. Often the trains and or stations don't have anyone.

compare dlr "train captain" pay with that of tube drivers



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: