Hacker News new | past | comments | ask | show | jobs | submit login
Does Automating the Metro Save Money? (2017) (cat-bus.com)
41 points by luu 8 days ago | hide | past | web | favorite | 75 comments

> The problem with automation is that in order to create a completely unattended system, the industry best practice is to install platform screen doors to ensure that nobody can ever enter the tracks.

What...? The problem with non-automated metros is the same! Either way, best practice is to have doors for safety. Several train stations in Tokyo have doors/gates, despite those lines having drivers. Meanwhile, Vancouver's automated metro has no gates protecting the tracks, but a generally good safety record nonetheless. The attendants are mainly there to deal with issues that gates wouldn't solve anyway: medical emergencies, confused passengers needing directions, drunk/violent passengers, etc.

Vancouver's automated system without platform doors has been in place since 1985, and as you mentioned, has worked quite well.

The key benefit to not having doors is that you can use different train types and consists because the train doors don't have to line up with the platform doors. Platform doors also make future procurement of new trains potentially extremely expensive, because you're pretty much locked into buying a specific type of train.

Or...you could just replace the doors/gates, which is far cheaper, easier, and generally years faster.

The cost of replacing all of the doors in a single transit system would be equal to the cost of one or two trains worth of new cars. (They're very expensive.)

You can't mix/match then though. If you have a training running every 10 minutes and want to go to every 5 you may need to double the amount of trains you have. If there is nothing wrong with the existing trains (they are not wore out) it doesn't make sense to replace them all.

Also it means you need to shut down the entire line for a month or so while you replace the doors in all the stations. Shutting down one station is easier to pull off without losing riders. If tell everyone you are remodeling for a month and here are the buses, your customers will adjust for that month (getting to the station a little early...), but if you shut down the entire line the system doesn't work for most riders and they buy a car and quit riding.

Your comment doesn't make any sense. Doubling the amount of trains don't change the trains themselves, unless you meant doubling the cars per train, which is a much bigger change that would usually require the reconstruction of the stations to handle the doubled length of the trains. That's a multi-year process per station.

In contrast, if you just want to double the frequency of trains, you just need to double the number of running trains. Doors/gates have nothing to do with that.

If you want to double the number of running trains you need to have double the number of trains. The new trains need to have the doors in the same place as the doors on the tracks, you might not be able to buy the same door configuration on new trains anymore.

There are a couple of exceptions, but they are absurd and expensive so I doubt they apply. Maybe you bought too many trains in the first place and they are just sitting unused somewhere and you can use them. Or maybe you run the train, and then let it sit for a long time before turning it around and running again.

I see a middle ground here...

Stream CCTV footage from every platform to a control room. Have operators observe the arrival of every train, and have a big red emergency stop button for when people fall on the tracks.

In a big network, with say 300 stations with a train every 4 minutes in each direction on average, that works out to ~40 operators required to watch those cameras for 10 seconds during each train arrival.

Not-so-advanced image processing could be used to show only the train arrivals at the busiest platforms or where there was most recent movement on the track, bringing you down to 1 or 2 operators.

Thats a lot cheaper than both platform edge doors, and drivers in every train.

That's almost exactly what Vancouver does. Biggest difference is that instead of image processing on the video feed, there's sensors on the tracks at every platform.

Actually Vancouver had no gates at all. The ticketing was automatic and honor based.

This was about gates on the platform, not gates to get into the station. Those new station gates have been a complete boondoggle however. Translink spent $200 million [1] on a system which was supposed to save $2 million a year in fare evasion but people just jump through anyway. They try to put a bright spin on this saying revenue went up $30 million, but they don't talk about ridership level increases at the same time. Even if they saved $10 million a year, it would take 20 years to pay for the system, far longer than the useful life of the fare gates.

[1] https://vancouversun.com/news/local-news/translink-still-los...

And the gates are very badly designed! The single-door design can't handle the peak crowds that go through them, and their 'beep' that confirms fare payment is a long monotone whine which makes it impossible to tell if you've paid successfully or the person next to you has, since all the gates are just constantly emitting a high pitched squeak.


Sort of honor based. You could be asked to show proof of payment and would get fined if you couldn't show you had paid.

Nitpick: Automation does not imply instllating platform doors.

Nuremberg U-Bahn uses radar instead to monitor the tracks.

I don't know the cost differences between those approaches.


The following paragraph:

"On 30 October 2008, the first accident resulting in a death occurred at Rathenauplatz station. A passenger fell onto the track as a driverless train approached the platform. The train was unable to halt in time because of the short distance to the fallen passenger and limitations of its braking system. Even if the same incident had occurred with a manually operated train, the driver would not have been able to halt in time either."

from that page is completely without a source.

The argument is bogus: either way they should be installing doors in each station preventing track access. This is basic safety: keep people off the tracks so they can't be hit by a train. This is also air quality: keep brake dust out of the air in the station.

Maybe you decide automation isn't worth it, but you still want the doors, even if it means hiring someone to open them.

The above was also stated by one of the comments, but it is worth bringing up.

Hmmm. London Underground carries about 1.5 billion people reach year. In the last year for data I can find, around 80 ended up on the tracks. Only a few stations have any pbarrier between track and platform.

All/most of the new stations on the Elizabeth line are going to have barriers afaik. It seems to me like all the new/refitted underground stations are getting the barriers these days too - I think it's a great improvement personally.

Alternatively they cost a lot and reduce flexibility (want to change your train fleet / door configuration?)

London has extremely cramped, extremely crowded stations. Platform screen doors become necessary as a way to ensure safety of _crowds_ not individuals. In most metro stations around the world, this is not true.

How about making a study, counting all metro stations noting whether they have platform screen doors, automation, and the year built. You'll very likely see there's a strong connection between automation and platform screen doors, but less so for the other variables.

Most of the stations on the older bits of the London Underground have curved platforms which means each place would need custom doors (+ some larger gaps between train and doors).

Newer lines, like the new section of the Jubilee line have straight track and doors.

The article doesn't mention one more thing that automation might bring: flexibility in working hours. Automated trains could work an extra hour for routes and districts that would benefit from it, or it could also be not working, e.g. extra trains that only work in peak hours.

This is especially true in the context of self-driving cars, but might be true for Metro as well.

The first comment on the article mentioned that the automated sky train in Vancouver comes every 2 minutes even at 1AM. Sounds like an awesome benefit.

That is the biggest benefit. When you already have an automated train and it isn't in the shop for maintenance the cost to run it empty is low enough that you may as well run it instead of store it. This means that people feel more confident in the system overall and are more likely to use it.

That is automated trains eliminate the "what if I have to work really late - how will I get home" problem.

For good reasons or bad, but... machines don't strike. That on its own is a very good reason to fully automate the metro.

Automated metro lines aren't unsupervised though. The automated lines in Paris and Lille also stop when there is a strike.

No during the last strike (September) in Paris the line 1 and 14 (the only 2 lines with automated metro) were both operational.

That's because the conductors were striking but not the supervisors, though, not because an automated line is inherently immune to striking.

I thought line 14 was stopped because of a strike some time ago but it was apparently just because of an incident, it might never have stopped because of a strike indeed. But that just means the employees that manage it don't strike.

> That's because the conductors were striking but not the supervisors, though

That's the whole point.

Is there a technological reason that it needs supervisors or is that an administrative requirement?

Of course, because however controlled is the environment it runs in, all sorts of unplanned things can happen. A tyre blowing out. A passenger pressing the emergency stop button. A person or animal entering the tracks after defeating the barriers. Flooding in a tunnel.

All of these events need a fast response regardless of whether the hardware detects the problem and stops by itself or not, even if just to evacuate the people that are inside the stopped cars.

Living in Paris, where the people that run the trains have threatened to strike from December 5 "indefinitely" (we'll see what that means), that was also my first thought when reading the article.

Perhaps not strike, but they do sometimes malfunction, go out of order, sometimes due to vandalism, or stop working because of electricity issues.

Automated trains are still run by humans who can go on strike.

So is the biscuit factory, but you may notice that striking biscuit factory workers don't find themselves in such a great position.

Train drivers (and aircraft pilots) have a bunch of special knowledge that isn't transferable and is safety critical. So when the person who is supposed to be doing the job isn't available you can't replace them in a timely fashion.

Automation often allows us to get rid of the specialisation, which makes those striking drivers replaceable, which makes striking less effective.

In the UK for example, my train from Southampton (on the South Coast) to London, has a crew of two people, a driver and a guard, both of whom need to be trained specifically in the exact model of train (let's say it's a 444 Electric Multiple Unit) and the exact route (in this case let's say the Wessex Mainline more or less direct from Southampton to London Waterloo)

If _either_ of those people aren't available, maybe they're on strike but maybe they have flu, that person must be replaced by someone with those two properties. A driver with 20 years experience driving an InterCity 125 on the ECML can't drive that train, but even a driver who has driven exactly the same train, but on the Portsmouth route, can't drive this train because they lack route knowledge. The guard, even though they don't drive the train, needs route knowledge just the same, because of the regulatory changes made in response to Ding-Ding And Away years ago.

Now, if instead these trains used ECTS level 2 automation, any driver, from anywhere in the ECTS system could safely operate that train, because it works the same everywhere. Doubtless local drivers might know a few little non-safety critical edge cases, and so a replacement driver from say, Germany, driving a train in Southampton would be a little slower, but the train doesn't need to be cancelled.

The (poor) economics of removing drivers from existing metros just goes to show what an extremely efficient way of moving people they already are.

The tech crowd is really big on cutting out people when the reality is that people are 1.) Pretty versatile, even in fairly routine jobs, handling unexpected events and 2.) Often aren't that expensive in the scheme of things. Even personal transportation (taxis, etc.) drivers are only about half the cost. That's certainly not nothing. But it's also not the difference between expensive and almost too cheap to meter.

Would you want to be a metro driver?

It's a dream job for many people, and is an extremely skilled job when dealing with technical problems and emergencies is taken into account. There's a lot of worse jobs we should get rid of first.

A garbage truck man is also a dream job for many people. Then the reality hits. I don't think we should simply not improve something when there's an opportunity because "there are other jobs to automate". Well yeah, there are, but the opportunity is there, not with the other jobs. There are many people in the world, we're capable of parallel innovation.

Garbage men report real high levels of job satisfaction. When asked why, they’re answer is simple: everyday the job gives their life meaning - they clean up for others and for everyone. Their role is fundamentally important.

Contrast this with my job. Energy researcher. Let’s assume energy research is necessary. Odds are that in my career I won’t make a ground breaking discovery. So I get up drag myself to work knowing that most likely than not nothing I will do that day will mean anything to anyone.

Not to say that sanitation work shouldn’t be automated. But the lack of respect for ppl’s livelihoods commonly shown in HN, or in general neoliberal circles, is astonishing.

I really don't think you should call me a neoliberal (and I am not sure what it means, but I am definitely not a part of the group that is commonly described as that). I am not sure what lack of respect are you seeing there, I have massive respect for the people doing the jobs I wouldn't mentally be able to do. That does not mean we should not innovate just because (still haven't heard a single reason other than "these are prestigious jobs" which is not really a reason), especially in a case where the technology is rather simple and available for several decades, well proven. A computer (as in human profession) used to be a prestigious job as well; so was a writer.

Do you know any garbage men? Also the proper term I sanitation worker. It’s a critical job that literally keeps you from dying of highly contagious diseases.

Most of the automated trash collection services haven’t automated much. They simply moved the responsibility of rolling the bins to the curb from the workers to the residents. Now it’s my job to wake up at 5am to move the bin to the street so the grappling arm can catch it. And then move it back to the house before 8am or get fined.

Where do you live that you have to take it out at 5 and bring it back by 8? That sucks!

I live in the city now. Previously I lived in an upper-middle class florida suburb. A long time ago, the trash collector was a 3-man team and one would roll the bin from the side of your house to the truck, then back. Now that they city has "automated trucks" the driver never leaves the cab. You roll the bin to the curb and the robot arm takes care of the last 5 feet. Lots of cities have ordinances about when and how long you can leave the trash bin out. And you should not leave them out overnight less you have wild animals digging through your trash.

I'm not going to weigh in on what should or shouldn't be automated but the trash/recycling industry is generally a pretty fulfilling place to work because you basically never have adversarial interactions with people and the impact of your work is immediately obvious. This is why many people prefer to run scrap and drive garbage trucks instead of flip burgers and drive bus.

Of course the compensation and working hours aren't as good as the six figure office jobs most people here have but dealing with refuse is pretty decent compared to other jobs of similar pay.

Answer to that question going to prove what? In my city metro drivers are unionized and paid on pair with software developers. Similar for bus. There is absolutely no shortage of the bus or metro drivers.

I am not saying it is not well paid. But when you try the work (I did for a day), you realize that you spend whole days in tunnels, mindlessly pushing a button, waiting for a computer to finally and rightfully replace you. Even worse job than being a bus driver (which is IMHO way worse than being a taxi driver).

So all jobs you personally dislike should be abolished?

I am sure the people who enjoy driving metro can do so in VR. Why would we forcefully hold onto a job that wastes human potential?

Maybe the issue is the difference in my understanding of human potential. There is simply so much more to do for these people that I can't see why someone would want to forcefully hold onto old ways. Why would you force a person to spend all days in tunnels?

Technology for automated metro is there for many years. Not using it is foolish. If you don't like people losing jobs, simply take the savings and employ the people elsewhere in the city, e.g. as bus drivers.

If want to help humans achieve their full potential, you need to have a system in place to enable that before you automate away their jobs. If you replaced all train conductors with robots today, most of them will wind up desperately scrambling for another (probably worse) unfulfilling job to keep food on the table.

Are you sure there is absolutely no shortage of workers and thus few dozens of people will be completely unable to live their lives anymore after you automate metro? I don't think that is true.

I didn't say anything close to that.

You said that there was "so much more to do for these people" that doesn't "waste human potential" if only we didn't "force" them to do dreary metro jobs. I'm saying that if you take away their dreary metro jobs, most of them will just wind up in dreary non-metro jobs unless we have a better situation standing by ready to receive them, which we don't.

If you want to free people from a life of drudgery, you need to offer an alternative, not just lay off all the drudges.

Edit: Or are you specifically opposed to underground jobs, not bad jobs in general?

The point is elsewhere, it doesn't matter whether the job is bad in my opinion or not. A metro rides on tracks in empty tunnels, and automated metros are technology from 1990's, not using it definitely wastes human potential. On the other hand, tram, bus or car driving is still a hard problem because of the unpredictability, so obviously, even though you're maybe not using your human potential to the fullest as a driver on the ground, it's at least something that a computer can't do, making it a valuable addition to the society, and it also saves these people from being locked to tunnels. I'm not sure about NYC in particular, but there are not enough bus drivers in the whole EU.

Maybe human potential isn’t that great. At least not for most people.

At a local grocery store they employ mentally handicap people to man the check out counters. They aren’t even aware that the automated counters make their jobs totally useless.

They are extremely slow at checkout.

And yet, I am very pleased with this grocery store because they’ve given a menial task to people who would otherwise have nothing. A chance to be in society and interact with us. Maybe, given the self checkout, it’s theater; but I don’t care. What’s the alternative for these people? Euthanasia? Do you see value in their lives?

You are missing my point. What you are describing is charity, and sure, let's do a lot of charity, way more than we do. I am also shopping at the local store where disabled people are the cashiers and I really like the social aspects. That whole concept is orthogonal to this discussion.

Let's not do something like non-automatic metro because ... What? I still don't know a single reason other than "these are jobs and they are considered prestigious and well paid" which is not a reason to me, especially since people here would say it about nearly every job I could talk about. There is not that much jobs today we can automate as well as riding trains, so why not start there?

I would hate myself if I decided that it's a good idea to force a person, even if they're mentally incapable (that would make it worse, actually), to spend whole days in tunnels, having to mindlessly push a button, when there is a proven way to easily automate. I just don't know why would someone do such a crazy thing from my point of view?

Except, not really, right? If you mean at existing service levels, then yes, of course. It's not really worth it.

Where automation really helps though is with _frequency_ of trains. If you have headways which are half as long (like in Vancouver on the Expo line the headways are every 90 seconds during peak periods) it makes a huge difference compared to something like SF's BART which has a weekday frequency of 15 minutes.

It also makes a massive difference when you're building a new line. If you can half the headway, you can half the size of the platform which makes building the line considerably cheaper. Less land to appropriate, less concrete to pour, less land to dig, etc.

There is one problem though. If you build a super convenient metro system with trains every minute or two, people will start to use that system. A lot. Vancouver's system is running into problems (particularly on the Canada Line) with massive overcrowding because even with the high frequency of trains, there's just not enough capacity.

As the article says, many metros (including the one in the article) combine automation with drivers. You only need to pay for the bit of automation that actually increases frequency and use the driver for all the fiddly bits.

The economics goes behind the limited analysis of this article. The main problem with train and metro drivers is that they need to be hired and trained to a certain level, and you also need other position like traffic regulator, etc. Here in Switzerland there is a lack of train driver creating problems as some services cannot be maintained to the expected level. In my city a suburban train line had to go from one train every 15min back to one every 30min, because of a lack of trained resources regarding traffic regulation (seems that 2 peoples were sick and another one quit), and cannot be as easily replaced.

They're efficient in terms of drivers.

Trains require large headway (gaps between trains, like following distance in a car) for safety reasons. This limits the number of trains you can run on a line. Which places a natural limit on the number of drivers you need.

Instead, you scale up passenger throughput by adding cars to trains. When you do this, you scale up capacity independent of number of drivers. Sort of a marginal cost of zero.

Another efficiency question, though, is bang for the buck. Trains are expensive to build. Headway limits the number of trains, and you can't make trains longer than the platform length. So you hit kind of a brick wall of scalability.

So it's worth asking whether automation could let you squeeze more return out of your investment in building the train line. Maybe it would unlock capacity increases through reduced headway. Or maybe some other way. I think it's an open question how much this can help, though.

Just one thing: It failed to mention that those extra pane w/ doors on the metro platform also help safety.

People falling on the metro line (intentionally or not) is a real issue, and, aside from the obvious immediate consequence to the body integrity of the person that fell, it also yield a significant cost.

It is become increasingly common to put those type of protective pane even on non automated metro lines. I saw them in Paris and Barcelona for example.

Sigh. The Lyon metro is only fully automated on line D. There is an ongoing project to automate line A and B. Lyon metro line D is relatively famous in mass transit signalling because it is one of the very first driverless metro and it features no platform doors. This is but one of the inaccuracies I found in this uninformed article.

You should automate because it eliminates future budget. When the next recession comes (which could be transit only: when political winds change and voters give transit less money, or won't allow fares to raise), the automated system is the last one cut because there is so little money to save. Since the trains are your busy routes and the backbone it means when things get better latter your most important users are still users and there are enough riders. If your system isn't automated you are tempted to cut service on the main routes to the point where people who are most likely to use transit buy a car anyway and then they forget that transit works because the car is their habit and they forget the downsides.

Might be too late for anyone to see this...

Cost estimates are far too low, and fail to include substantial back-loaded pension benefits as well as training and such.

I used the Montreal Metro a few years ago and they were automated and didn't use the doors. Are they coming eventually or are they being rolled out in phases?

I think you're mistaken. The Montreal metro is not automated and has never been. The train controls are mostly automated, but they are still operated by a driver.

automated (ATO), but not unattended (UTO)

I find it curious that we are attempting to create an automated car, but we can't seem to make an automated train that is on a defined rail?

No, I think it's pretty common in this case. [1]

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

Of course there exist automated trains. The question is whether automating an existing system makes sense, in terms of cost. The simple model chosen here is to assume that automation requires platform screen doors -- and showing that the cost of installing those is on par with the cost of the drivers. Meaning the drivers represent a small amount of the total cost of the system.

The paris metro has been automated for decades. Get out of here with your middlebrow dismissal

My ambiguous usage of "we" to refer to my particular country of the United States was misleading. I'm glad these things exist around the world and I hope we in the US can implement one someday.

Ever been on one of those wee trains at an airport?

Personal attacks are against site guidelines. A simple factual rebuttal suffices here.

My completely uninformed assumption has been that there are more unions preventing the automation of trains than automobiles.

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