3 ms·

Obviously the scale is huge, but conventional trains have inefficiencies. Suppose the average rider's trip is 10 stops. At every stop where the rider does not

Obviously the scale is huge, but conventional trains have inefficiencies. Suppose the average rider's trip is 10 stops. At every stop where the rider does not get off, there is inefficiency because they sitting there, taking up a seat on a train that is stopped. You're not only wasting their time, you're also wasting train capacity. The naive solution is to make the train longer. Which works, but it's costly because you also have to make the platform longer. In a subway system, that's expensive. And the worst part is, you have to do it for every station the train will stop at, even the ones that aren't as busy. This is, of course, why express trains exist. Eliminate some stops and you reduce all these problems. But the more stops you eliminate, the fewer people can use the express (because it doesn't stop where they're going), so there's an unavoidable compromise. In the days before automation, this was probably the best that could be done. But now that automation exists, new approaches are possible. Since you no longer need a driver per train, you now have the option of making trains smaller. And it's also possible to group people together by starting point and destination. So in theory you could make every train an "express train", one with zero stops other than starting point and destination. No seats wasted on making people wait in a non-moving train while others get on/off. So you can eliminate an inefficiency that you never could before. Of course, every train needs headway for safety reasons. More smaller trains instead of fewer larger trains means that's harder. So it's not without its own, different limitations. Maybe automation has fast enough reaction time (or other tricks) that can cut down on the physical space needed for headway. Anyway, the point is that much smaller vehicles might actually increase efficiency, and this is what the Boring Company seems to be planning to do.

jcranmer 8y ago
You're completely wrong. On a fixed guideway system--that includes vanilla railways and things like monorails--a train that is stopped completely blocks the entire track, and cannot be routed around except via switches. The minimum headway between trains is dominated by switch fouling: how long it will take a train to pass over the switch, and then for the switch to switch to a different track, before the signal can go green again. You'll also want some extra padding on top of that for timetabled trains, to take into account unexpected long station dwells and schedule recovery. The throughput of most subways is about 26 trains per hour. With good signalling and track management, you can push it to about 36 trains per hour on a branched line; an unbranched line (which simplifies management a lot!) can get up to 45 trains per hour, but those are definitely not typical numbers. Train length doesn't impact these numbers much, definitely not at the scale of variation being discussed. Express trains require dedicated tracks to be able to bypass parked trains, or insanely good timetable-keeping to not get stuck behind a local train before a bypass. NYC built a lot of its trunks to 4 tracks, which allows it to make local/express trains, but most systems have found it's easier to just build another 2-track trunk a few blocks away than to try to add 4-tracks to the 2-track trunk. The idea of trying to use tiny cars with much narrower safety margins for headways is the idea behind personal rapid transit, but PRT systems have been shown to have utterly lousy throughputs: pulling 10,000 people/hour on a trunk PRT line would be a significant milestone, whereas the standard 8-car train on a 26TPH subway easily pulls 25,000 people/hour without needing much effort.
- avn2109 8y ago
  This is a really great comment and should be much higher on the page, and I appreciate that you took the time to write it all out.
- steve_gh 8y ago
  26 trains an hour is pushing it hard. The London underground generally runs at about 24 TPH at rush hour on a single line (in each direction) If you go to a moving block signalling system you can push the system a bit harder, but the problem is that as the trains get more crowded, their braking distance increases (which increases the block size - so fewer TPH). Also, at rush hour the dwell time at each station increases because more people are getting on and off - again impacting the headway. So pushing the TPH on a metro line is hard...
  - iggldiggl 8y ago
    > The London underground generally runs at about 24 TPH at rush hour on a single line (in each direction) Not quite. All of the resignalled lines manage more than that. The Central line reaches something between 27 and 30 tph, the Jubilee Line 30 tph (and I think there are still plans to go even higher) and the Victoria Line 36 tph. The Northern Line is a bit special because of its branching, but the single section south of Kennington towards Morden manages 30 tph as well. And even the Subsurface Lines (which for the largest part are still conventionally signalled, although not for much longer) manage around 27½ tph on the north (Baker Street - Aldgate) and south (Tower Hill - Gloucester Road) sides of the circle. > the problem is that as the trains get more crowded, their braking distance increases A properly designed brake system will a) be designed to achieve the required deceleration even under maximum loading b) include load compensation, i.e. the braking force for each car is adjusted depending on its loading to achieve the same deceleration so no, that shouldn't really matter.
- adrianmonk 8y ago
  Basically, I see three ways of looking at what Boring Company is trying to do: 1. It is being built by people who don't know what they're doing and don't understand the field, and it isn't going to work. (This seems to be the way you look at it.) 2. They expect to be able to reduce the costs of tunneling so drastically that the lower throughput of PRT won't even be a problem. 3. They plan to use an unprecedented level of automation to advance the state of the art in how PRT can be done. (This is my guess.) Obviously, nobody really knows which of the above is the case. They haven't shared a lot of detail. But in support of the idea that it _could_ be #3, there's a reason I think there's room for innovation in PRT. It's so expensive to build any transit system that relatively few attempts have been made at PRT. And some things have changed since many of those tries, mainly advances in automation. So for example, if switching is a bottleneck for headway, maybe there is a way to build a much faster switch but nobody has put the R&D money into it yet.
mncharity 8y ago
> Maybe automation [...] can cut down on [...] headway. Indeed. One can think of a BC tunnel as a narrow rail tunnel containing a single highway lane. An express lane restricted to networked self-driving non-truck vehicles. Headway buys fewer vehicles sharing the damage and acceleration of a lead accident vehicle's decel - because with networking, they needn't ever exceed it. One might literally do full-speed bumper-on-bumper. It's a really pretty point in design space that the Boring Company is targeting. But the commentary I've seen has pervasively been confused. Which may be for the best - industry disrupting startups are often better off remaining disregarded for as long as feasible.
- jcranmer 8y ago
  > Indeed. One can think of a BC tunnel as a narrow rail tunnel containing a single highway lane. An express lane restricted to networked self-driving non-truck vehicles. Headway buys fewer vehicles sharing the damage and acceleration of a lead accident vehicle's decel - because with networking, they needn't ever exceed it. One might literally do full-speed bumper-on-bumper. How do you do stations? If you have the station on the mainline, your entire train of vehicles has to stop at the station just like, well, a train, so what you have is just a less safe version of a train. If you have turnouts for stations, you have the issue of switch fouling preventing you from bunching vehicles too close together, and high-speed turnouts have to be long to deal with turn radii. But the real killer with turnouts is trying to get a train to merge back onto the trunk at speed--with little-to-no space between cars, a slight mismatch in speed is going to equal a high-speed collision.
  - mncharity 8y ago
    > How do you do stations? [...] high-speed turnouts have to be long to deal with turn radii [...] to get a train to merge back onto the trunk at speed It's a highway. Paying with tunneling to avoid surface real-estate constraints. And needing fewer lanes for similar throughput due to density and speed. With minibuses. That are self driving. And don't have to worry about human drivers. And can be street legal. So how would you do X? Well, how would you do it now, with a highway and minibuses? What if you could restrict it to networked automated minibuses? What if they were tightly integrated with infrastructure? So for stations? Now: highway ramp to/from a lot, platform curb, or street. Automated: similar, but with tighter non-human margins. Integrated: so many possibilities... including elevators? Consider a subway station. A short station. For a single minibus. Minibus and mini platform and elevator. Everyone gets off the minibus, crosses the platform, and gets on the elevator. Platforms are expensive. Why have a platform? Ok, so minibus and elevator. Everyone steps off the minibus, directly into the elevator. But the minibus isn't all that heavy, so why craft an extra human space? Ok, so you could have the minibus get on the elevator. At the "top", of ramp or elevator, in a mall, or lobby, or platform, or garage, or street, the minibus could play elevator cab, or PRT, or kiss-and-ride Lyft, or lot rideshare, or neighborhood minibus trundling off down the street. And continuous elevators for tightly-integrated automated vehicles could have surprising throughput. The artist sketch[1] of a point-to-point Chicago-O'Hare station has them playing PRT. [1] https://www.boringcompany.com/gallery/2018/6/14/io1dfpyvs1y6dj79tsbrjgvo8p76h9 https://www.boringcompany.com/gallery/2018/6/14/io1dfpyvs1y6...

jcranmer 8y ago

You're completely wrong. On a fixed guideway system--that includes vanilla railways and things like monorails--a train that is stopped completely blocks the entire track, and cannot be routed around except via switches. The minimum headway between trains is dominated by switch fouling: how long it will take a train to pass over the switch, and then for the switch to switch to a different track, before the signal can go green again. You'll also want some extra padding on top of that for timetabled trains, to take into account unexpected long station dwells and schedule recovery. The throughput of most subways is about 26 trains per hour. With good signalling and track management, you can push it to about 36 trains per hour on a branched line; an unbranched line (which simplifies management a lot!) can get up to 45 trains per hour, but those are definitely not typical numbers. Train length doesn't impact these numbers much, definitely not at the scale of variation being discussed. Express trains require dedicated tracks to be able to bypass parked trains, or insanely good timetable-keeping to not get stuck behind a local train before a bypass. NYC built a lot of its trunks to 4 tracks, which allows it to make local/express trains, but most systems have found it's easier to just build another 2-track trunk a few blocks away than to try to add 4-tracks to the 2-track trunk. The idea of trying to use tiny cars with much narrower safety margins for headways is the idea behind personal rapid transit, but PRT systems have been shown to have utterly lousy throughputs: pulling 10,000 people/hour on a trunk PRT line would be a significant milestone, whereas the standard 8-car train on a 26TPH subway easily pulls 25,000 people/hour without needing much effort.

avn2109 8y ago
This is a really great comment and should be much higher on the page, and I appreciate that you took the time to write it all out.
steve_gh 8y ago
26 trains an hour is pushing it hard. The London underground generally runs at about 24 TPH at rush hour on a single line (in each direction) If you go to a moving block signalling system you can push the system a bit harder, but the problem is that as the trains get more crowded, their braking distance increases (which increases the block size - so fewer TPH). Also, at rush hour the dwell time at each station increases because more people are getting on and off - again impacting the headway. So pushing the TPH on a metro line is hard...
- iggldiggl 8y ago
  > The London underground generally runs at about 24 TPH at rush hour on a single line (in each direction) Not quite. All of the resignalled lines manage more than that. The Central line reaches something between 27 and 30 tph, the Jubilee Line 30 tph (and I think there are still plans to go even higher) and the Victoria Line 36 tph. The Northern Line is a bit special because of its branching, but the single section south of Kennington towards Morden manages 30 tph as well. And even the Subsurface Lines (which for the largest part are still conventionally signalled, although not for much longer) manage around 27½ tph on the north (Baker Street - Aldgate) and south (Tower Hill - Gloucester Road) sides of the circle. > the problem is that as the trains get more crowded, their braking distance increases A properly designed brake system will a) be designed to achieve the required deceleration even under maximum loading b) include load compensation, i.e. the braking force for each car is adjusted depending on its loading to achieve the same deceleration so no, that shouldn't really matter.
adrianmonk 8y ago
Basically, I see three ways of looking at what Boring Company is trying to do: 1. It is being built by people who don't know what they're doing and don't understand the field, and it isn't going to work. (This seems to be the way you look at it.) 2. They expect to be able to reduce the costs of tunneling so drastically that the lower throughput of PRT won't even be a problem. 3. They plan to use an unprecedented level of automation to advance the state of the art in how PRT can be done. (This is my guess.) Obviously, nobody really knows which of the above is the case. They haven't shared a lot of detail. But in support of the idea that it _could_ be #3, there's a reason I think there's room for innovation in PRT. It's so expensive to build any transit system that relatively few attempts have been made at PRT. And some things have changed since many of those tries, mainly advances in automation. So for example, if switching is a bottleneck for headway, maybe there is a way to build a much faster switch but nobody has put the R&D money into it yet.

mncharity 8y ago

> Maybe automation [...] can cut down on [...] headway. Indeed. One can think of a BC tunnel as a narrow rail tunnel containing a single highway lane. An express lane restricted to networked self-driving non-truck vehicles. Headway buys fewer vehicles sharing the damage and acceleration of a lead accident vehicle's decel - because with networking, they needn't ever exceed it. One might literally do full-speed bumper-on-bumper. It's a really pretty point in design space that the Boring Company is targeting. But the commentary I've seen has pervasively been confused. Which may be for the best - industry disrupting startups are often better off remaining disregarded for as long as feasible.

jcranmer 8y ago
> Indeed. One can think of a BC tunnel as a narrow rail tunnel containing a single highway lane. An express lane restricted to networked self-driving non-truck vehicles. Headway buys fewer vehicles sharing the damage and acceleration of a lead accident vehicle's decel - because with networking, they needn't ever exceed it. One might literally do full-speed bumper-on-bumper. How do you do stations? If you have the station on the mainline, your entire train of vehicles has to stop at the station just like, well, a train, so what you have is just a less safe version of a train. If you have turnouts for stations, you have the issue of switch fouling preventing you from bunching vehicles too close together, and high-speed turnouts have to be long to deal with turn radii. But the real killer with turnouts is trying to get a train to merge back onto the trunk at speed--with little-to-no space between cars, a slight mismatch in speed is going to equal a high-speed collision.
- mncharity 8y ago
  > How do you do stations? [...] high-speed turnouts have to be long to deal with turn radii [...] to get a train to merge back onto the trunk at speed It's a highway. Paying with tunneling to avoid surface real-estate constraints. And needing fewer lanes for similar throughput due to density and speed. With minibuses. That are self driving. And don't have to worry about human drivers. And can be street legal. So how would you do X? Well, how would you do it now, with a highway and minibuses? What if you could restrict it to networked automated minibuses? What if they were tightly integrated with infrastructure? So for stations? Now: highway ramp to/from a lot, platform curb, or street. Automated: similar, but with tighter non-human margins. Integrated: so many possibilities... including elevators? Consider a subway station. A short station. For a single minibus. Minibus and mini platform and elevator. Everyone gets off the minibus, crosses the platform, and gets on the elevator. Platforms are expensive. Why have a platform? Ok, so minibus and elevator. Everyone steps off the minibus, directly into the elevator. But the minibus isn't all that heavy, so why craft an extra human space? Ok, so you could have the minibus get on the elevator. At the "top", of ramp or elevator, in a mall, or lobby, or platform, or garage, or street, the minibus could play elevator cab, or PRT, or kiss-and-ride Lyft, or lot rideshare, or neighborhood minibus trundling off down the street. And continuous elevators for tightly-integrated automated vehicles could have surprising throughput. The artist sketch[1] of a point-to-point Chicago-O'Hare station has them playing PRT. [1] https://www.boringcompany.com/gallery/2018/6/14/io1dfpyvs1y6dj79tsbrjgvo8p76h9 https://www.boringcompany.com/gallery/2018/6/14/io1dfpyvs1y6...