Adam is an international treasure

You are viewing a single thread.
View all comments
42 points

I’m really tickled by the fact that we can’t fully automate trains yet. I never thought about it, but put into perspective how asinine self-driving cars are if we can’t achieve the same thing with a train, something that is vastly more tractable and less chaotic than road traffic.

permalink
report
reply
7 points

a couple years back I recall seeing some interesting stuff re the nyc subway (link1, link2)

leaving aside how hilariously far they’ve overrun constructions and estimations (wow nyc is bad at this), it was kinda wild to learn that they didn’t even know where trains were during operation. made perfect sense in context with the info, but still wild

permalink
report
parent
reply
13 points

The video actually mentions why it’s not quite possible yet, but it’s slowly moving towards that direction. It’s just that the EU train system is very FUBAR’d due to all the 20th century conflicts

permalink
report
parent
reply
1 point

@db0 u wot m8

permalink
report
parent
reply
5 points

permalink
report
parent
reply
39 points

If what you took away from that is “EU trains are FUBAR because something something conflicts” then you weren’t paying attention. Automating trains isn’t that hard to do (London’s Victoria Line has been more or less self-driving since 1967) with some kind of transponders-and-in-cab-control arrangement, but they still to this day have to have a train operator (i.e., a driver) in the cab. This is not because the automation can’t make the train go and make it stop again at the right place, it’s because actually pushing a lever back and forth is only a tiny part of the job of driving a train. The rest is about knowing rules many of which are extremely safety-critical, evaluating rules, and applying knowledge and experience to make sure those rules are correctly applied. For instance, you can put a passenger train on a fenced-off track with no intermittent route changes and it can drive itself from A to B using existing technology. The problem, however, is what happens if something goes wrong? A wire connecting a trackside transponder fails – the train will stop because it doesn’t know what to do. A foreign body is detected on the track in front? The train will have to stop until someone moves it. And not only that train will be stopped, but all the trains behind it will be stopped until someone can get there in person… and let’s hope they don’t have to use the strech of track that’s blocked to get there.

So you still need a human on the train to resolve these problems - a signal failure means a two-way conversation with the signalbox to confirm what’s going on and get given manual permission to proceed, usually at a reduced speed. A foreign object can be examined on the spot, moved if the driver is able to do so, and the track checked over to make sure it hasn’t been damaged by the impact. And this is a very simple example. Driving a train is one of those jobs (a lot like being a pilot, and few people seem to be talking about getting rid of airline pilots) which is 99% routine but 1% exceptions, and the possible number of exceptions is nearly infinite. Automation in the cab is certainly a useful thing just as automation in your car is a useful thing and for the same reason - it frees up expensive human eyeballs and brains to worry less about the repetitive mechanics of the 99% routine so they can pay more attention to any potential 1% exceptions coming down the line. Automation simply can’t meet the safety requirements – there’s no “acceptable number” of accidents or fatalities in railway operations unless that number is zero.

There are, to be fair, some extremely niche operations where full automation can and does work – mostly on isolated metro systems where the infrastructure is expansive enough, there are no level crossings, and the line operates effectively in a vacuum. Even in that case, the Victoria Line can’t meet the safety requirements as the tunnels have no side walkways and passenger evacuation means walking people off through the middle of the cab onto what have to be assumed to be live electrical rails without going through complex safety procedures to be sure they’re safe.

Railway safety in Europe is nothing like what a lot of people think it is (i.e., akin to highway safety). It’s taken very, very seriously and no compromises are ever acceptable. Even many rules which seem hard to explain today exist because a massively improbable series of failures at some point in the past caused disaster or near-disaster and could still repeat themselves today if not for this rule. It’s complex, sure, but for a system that’s undergone 200 years of continuous evolution and development and still remains extremely safe it’s anything but FUBAR.

permalink
report
parent
reply
18 points

mostly on isolated metro systems where the infrastructure is expansive enough, there are no level crossings, and the line operates effectively in a vacuum.

Also, critically, metro systems are relatively “dense”, so you can have a human respond to an exception quickly - you’re always in a populated area with not much space between individual stations.

permalink
report
parent
reply
10 points

I’m also wondering what the ROI is on automation costs and maintenance thereafter versus paying a conductor to just drive the thing.

In Copenhagen, the metro is fully automated and there are discussions of doing the same with the commuter rail (all grade separated). I think the more interesting benefit to the commuter rail is they will run closer together in time and run more frequently at night (every 10 vs 20 or 30 minutes)… But I don’t think the financials would ever really pay for the upgrades needed…

permalink
report
parent
reply
23 points

Other thing to add to this is that there’s just one or two people in the train providing service for hundreds of other people or millions of dollars worth of goods. Automating those people away is simply not economical, not even in terms of the headcount replaced vs headcount that has to be hired to maintain the automation software and hardware.

Unless you’re a techbro, who deeply resents labor, someone who would rather hire 10 software engineers than 1 train driver.

permalink
report
parent
reply
3 points
*

The DLR is driverless, but there’s continuous remote monitoring. (There’s continuous remote monitoring on all the lines, of course. With chat back and forth with the driver. My wife used to work on the Tube doing this - Line Information Assistant, occasional Line Controller - so I have heard all the after-shift rants.)

A Tube driver’s real job is largely safety inspection, with a side order of driving a train.

permalink
report
parent
reply
3 points

Yeah, the light-rail system in Vancouver, BC is basically fully-automated, but they do still have to have a person hanging out in the train in case something goes badly wrong.

permalink
report
parent
reply
26 points
*

It’s not so much that self-driving is easier than automatic trains, but that it’s easier than passenger rail in the US at all. Everyone hates rail here it’s wild, and the safety expectations in the US are (weirdly) much lower for cars than anything else.

permalink
report
parent
reply

SneerClub

!sneerclub@awful.systems

Create post

Hurling ordure at the TREACLES, especially those closely related to LessWrong.

AI-Industrial-Complex grift is fine as long as it sufficiently relates to the AI doom from the TREACLES. (Though TechTakes may be more suitable.)

This is sneer club, not debate club. Unless it’s amusing debate.

[Especially don’t debate the race scientists, if any sneak in - we ban and delete them as unsuitable for the server.]

Community stats

  • 381

    Monthly active users

  • 201

    Posts

  • 2.4K

    Comments