PDA

View Full Version : Traffic engineering



Jens
2008-Jun-09, 01:27 AM
I've been wondering about something, and think that since there are tons of scientifically minded people here, somebody might be able to steer me in the right direction.

I take the train to work every morning, and partly because the schedule is very right (there are local and express trains, and they come every few minutes), they often run late. And the conductor often makes an announcement, like "this morning an umbrella got stuck in a door, so the trains are running late." But the problem is, sometimes the schedule just gets more and more messed up as time goes by. The umbrella must have only delayed the train by a few minutes, but at some point the trains are running 15 minutes late. So I'm wondering, is there a simple reason for this? I know that with car traffic, there is the problem of "rubbernecking," but with trains, which are controlled by computerized schedules, I don't see why it would happen.

I guess what I'm asking is, is there a simple way to look this problem up?

01101001
2008-Jun-09, 06:09 AM
But the problem is, sometimes the schedule just gets more and more messed up as time goes by. The umbrella must have only delayed the train by a few minutes, but at some point the trains are running 15 minutes late. So I'm wondering, is there a simple reason for this? I know that with car traffic, there is the problem of "rubbernecking," but with trains, which are controlled by computerized schedules, I don't see why it would happen.

See the poorly titled topic Wave-particle duality... of traffic (http://www.bautforum.com/general-science/74152-wave-particle-duality-traffic.html).

It's not just rubbernecking with traffic, and I'd expect that with trains it is related: simply that the performance of a system that is near capacity, when there is a slight upset, suffers disproportionally during recovery.

Roads don't have a centralized intelligence, but I doubt that is critical to the matter; emergent behavior of a collective of individual decision makers still adds up to an overall behavior.

I think it's just a matter of that behavior, the policies, whether centralized or distributed, that defines how effects are related to causes. But they are probably not easily changed. It wouldn't surprise me that when systems are near capacity then effects cannot vary directly with the size of causes, without sacrificing safety or some other measure of economy.

Stuart van Onselen
2008-Jun-09, 06:00 PM
For that exact reason, all sorts of systems, from computer networks to electricity supply, are designed to have a reasonable degree of over-capacity, often from 10% to 20%. Run without that "little bit extra", and small problems rapidly become huge.

For example, a major power failure occurred in the USA, because the grid was running very close to full capacity, and then some overhanging branches shorted out just a couple of wires. Cue a chain-reaction that blacked out half the Eastern Seaboard.

Larry Jacks
2008-Jun-09, 07:33 PM
Not knowing much about trains, I wonder if it'd be better to add an additional passenger car or two to each train to maintain passenger capacity and allow a few more minutes between each train. Would that allow a better margin to prevent cascading disruptions for minor delays?

jokergirl
2008-Jun-09, 07:45 PM
Not really. People are very efficient in squishing in, but they want a lot of trains to go.

The problem with train delays is that they affect the whole grid very quickly. Many old tracks only allow for trains passing each other in certain places, so if one train is late, the other has to wait for it, creating a chain reaction of delays.

Here they cancel trains if the delays get too large. Needless to say, I'm pretty annoyed when they cancel mine. :P

;)

Ilya
2008-Jun-09, 07:46 PM
Semi-off topic: one thing which utterly ****es me off about American medical service is that with VERY few exceptions, all doctors schedule patients as if every appointment ran smoothly and without delays. Naturally, as delays occur appointments down the line get delayed more and more. With (theoretically) 15 minute visits it is not so bad; if every visit is a (theoretically) hour-long procedure, you may show up at your scheduled time and end up waiting for hours.

If procedure takes 45 minutes when all goes well, but with complications can run for two hours, why can't doctors give 1 hour 15 minutes to each patient, hence allow delays to die down instead of propagating?

I noticed this is much less of a problem with dentists, perhaps because dental work is more predictable.

Delvo
2008-Jun-09, 09:00 PM
Many old tracks only allow for trains passing each other in certain places, so if one train is late, the other has to wait for it, creating a chain reaction of delays.And that's where a lot of the extra minutes come from. If one train is delayed three minutes and another stops for it and then waits for it to pass, the second one gets delayed three minutes, plus the time for the first one to pass, plus the time for the second one to start up again, plus the time to reach wherever its next stop was supposed to be. Cars, at least, can often go past each other.

HenrikOlsen
2008-Jun-09, 09:07 PM
Here they cancel trains if the delays get too large. Needless to say, I'm pretty annoyed when they cancel mine. :P

;)
And even that creates new problems, since that means the canceled train isn't at the other end where it's expected in the plan so they have to reschedule another or cancel the return ride as well.

Jim
2008-Jun-09, 09:47 PM
Not really. People are very efficient in squishing in, but they want a lot of trains to go.

Most people miss a train because it leaves before they get to the station, not because there's no room for more passengers.

With that in mind, it's better to shorten the wait for the next train.

Larry Jacks
2008-Jun-09, 09:51 PM
However, doesn't shortening the wait for the next train increase the potential for disruption that the OP mentioned? If the time between trains is already so tight that any problem causes widespread delays, it seems that reducing the time would make the problem even worse. That's why I asked if you increased the time between trains by a few minutes but increased each train's capacity, it seems you'd have fewer disruptions.

A similar thing happens at airports that are too tightly scheduled - any disruption causes a cascading series of delays.

Jens
2008-Jun-10, 06:04 AM
And that's where a lot of the extra minutes come from. If one train is delayed three minutes and another stops for it and then waits for it to pass, the second one gets delayed three minutes, plus the time for the first one to pass, plus the time for the second one to start up again, plus the time to reach wherever its next stop was supposed to be. Cars, at least, can often go past each other.

Maybe that's part of it. Actually, the train line I live on has four "lanes" around where I live, so the expresses can almost always pass the locals without trouble. The problem is that they try to "connect" the trains. So the local will wait for the next express, to allow people to switch between the trains.

Somebody asked about the length of the trains. I suspect that one limitation is the acceleration. The distance between stations is pretty small (I can actually see the train parked at the station before mine and see when it leaves, because the headlights go on). Sometimes the train basically accelerates to full speed and then immediately starts decelerating.

Jeff Root
2008-Jun-10, 07:24 AM
Can you post a link to a route map? Especially if it has distances between
stations marked. That might be the only part of it I'd be able to read.

-- Jeff, in Minneapolis

Jens
2008-Jun-10, 07:49 AM
Here (http://en.wikipedia.org/wiki/Odaky%C5%AB_Odawara_Line)'s the best I can do, a diagram from Wikipedia. The station I get on at is Komae, and you can see that the distance to the next station out (Izumi Tamagawa) is just 600 meters. I only ride that train to Yoyogi-Uehara, and then change to the subway. I don't think it would help much with the problem. :)

SeanF
2008-Jun-10, 01:42 PM
And that's where a lot of the extra minutes come from. If one train is delayed three minutes and another stops for it and then waits for it to pass, the second one gets delayed three minutes, plus the time for the first one to pass, plus the time for the second one to start up again, plus the time to reach wherever its next stop was supposed to be. Cars, at least, can often go past each other.
Can't add "the time for the first one to pass" in there, nor "the time to reach wherever its next stop was supposed to be." That time would already be included in the normal travel time for the second train. The only additional delay to the second train should be the re-acceleration time, and I don't even think we'd need to count all of that... :think:

EDIT: Actually, if we consider we've got communication between the trains - the second train could be told the first train is running three minutes late, and do the deceleration and acceleration ahead of time such that it reaches the passing point three minutes later than it normally would have. That way it can just continue on through with no additional delay beyond the three minutes at all. :)

Veeger
2008-Jun-10, 06:05 PM
Answering the question of OP is one thing. Solving the problem quite another. And the one who can design networks that solve the problem can probably name their price at a company like AT&T or IBM.

;)

Tobin Dax
2008-Jun-10, 10:16 PM
EDIT: Actually, if we consider we've got communication between the trains - the second train could be told the first train is running three minutes late, and do the deceleration and acceleration ahead of time such that it reaches the passing point three minutes later than it normally would have. That way it can just continue on through with no additional delay beyond the three minutes at all. :)
Unfortunately, I suspect that this may not be a two-body problem. ;) Put a third train into the equation somewhere (so to speak), and it becomes a bit more complicated.