Since the early fall, my commute has involved only IRT trains, and as I’ve written in the past, I find the countdown clocks to make waiting for the next subway a far less stressful experience. As all of my stations now have these clocks, I’ve also started focusing on what the wait times can tell us about subway headways, and I’ve often taken to Twitter to post my observations about train bunching. It seems to happen far more often than we would like.
Take, for instance, off-peak patterns on the 2 and 3. Oftentimes, I’ll see a long wait for a downtown 2 train and a much shorter wait for the 3 train right behind it. The wait times may average out to acceptable headways, but the reality of the situation is far more subtle. I witnessed another strange happening this past Saturday when three downtown 1 trains past within 5 minutes of each other at around 5:15 p.m. Accordingly to the MTA’s schedule, those trains should have been operating with eight minutes of headway rather than 100 seconds.
Some of this bunching may be unavoidable. Perhaps an earlier delay on the 1 line caused Saturday’s bunching, but the 2/3 bunching happens often enough that I wonder if it’s a scheduling issue. The 3 makes just one stop after leaving its terminal before joining up with the downtown 2 that the dispatcher should be able to avoid such bunching. I’ve noticed, however, that the late-night schedule features Manhattan-bound 2 and 4 trains running with similarly mistimed gaps in Brooklyn.
One day, I hope the MTA releases the wealth of data currently being collected by the underlying technology supporting the PA/CIS system. It would provide a glimpse of how trains are running with what spacing as they make their journeys to their ultimate destinations. Perhaps we’d see bunching, and perhaps too we would be able to isolate why bunching happens by discovering which trains are experience high dwell times. For now, though, we just suffer through mystifying and frustrated bunching as the countdown clocks tick down the minutes until the next two trains arrive.
28 comments
Has anyone noticed trains arriving to the station even when the clock is showing 2 mins? I take the 4/5 from Bowling Green most days of the week, and I find that the “train is entering the station” message is announced even when the clock has just switched to showing 2 mins.
Now, when headways are longer – say a late weeknight – I haven’t noticed this issue, which leads me to believe the rush hour service levels are somehow to blame.
Anyone have any insight?
Thanks.
My possibly outdated understanding is the times are based on printed schedules, not real-time data.
No. That’s incorrect. The countdown clock times are based upon the real-time location of the train based upon location data from the block signalling system.
To answer Adrian’s question, the two-minute issue has been addressed somewhat by Transit, but since the system relies on the block system, the train may be closer to the station than a full 120 seconds away. That’s why it sometimes jumps from 2 minutes to “now arriving.”
Have you asked Transit for an explanation of the regular bunching? You seem to have journalist status and could probably get a response.
My understanding is that bunching is to some extent stochastically inevitable. Over the course of a long run, delays are unavoidable, and once they happen there is no way to make the time back.
That doesn’t necessarily explain every case, but it is one reason for it.
No, what is inevitable are delay-causing actions by passengers. Some loads might be unusually high for some reason, or maybe a narcissist decided his time is more valuable than ~1000 other people’s time and held a door. These can be scheduled for.
Can’t speak to NYCTA procedure, but if it is “stochastically inevitable” every day that a train simply can’t make its scheduled stops,* then they’re probably doing it wrong. If the system is run properly, you can make up the time: you go a little faster. Not so much faster that you are traveling at an unsafe speed, of course, but there should be some padding in the schedule to go faster.
* Not every station needs to be scheduled either. If you have a string of stations A, B, C, D, E, F, G, H, I, J, K, L, M, N, and O perhaps there is a scheduled time to leave A, and then arrive at every third station at a scheduled time. It’s fine for there to be a small variance in when it reaches the other stations, at least on a frequent transit system.
Schedules are written based on typical running times for the time of day, including typical delays. They don’t (and can’t) account for the worst possible case.
Trains already go as fast as they can, within the limits imposed by the (safety-critical) signal system and the skittishness of the train operator. (By the way, this is one of the arguments in favor of ATO: a computer isn’t skittish.) Once the delay has already happened, it’s hard to undo it – the train is probably overcrowded now, with people holding the doors at each stop.
Conceivably, the schedule could include little bits of padding up and down the line, but then most trains would be way ahead of schedule by the time they reached the third or fourth timepoint (unless they’re held at each timepoint, but that delays through riders). Recovery time is traditionally included only at the terminals, because terminals have no through riders and terminals are where crews usually have access to such amenities as restrooms.
Not every station is a timepoint. Most timepoints have holding lights, so that, if a train is running substantially early or a delay has developed behind it, a dispatcher can hold it back, but that’s assuming that a dispatcher is paying attention. Whether due to understaffing or lack of interest or engagement in a more critical activity, the dispatcher usually isn’t paying attention.
So is the IRT using a different system then the L? What I said seems quite accurate with regard to the L, or at least it was back in 2008:
They installed the real-time location monitors in some stations because they gave you an idea of train location, which was supposed to give you a clue about whether the time indications were close to right! :-O
Is that still true? Those computer screens change when there is a problem with a train or if the train is stalled. They don’t use a predetermined system at all. In fact they are pretty darn accurate. I use that station and those screens almost everyday.
Yes, the computer screens do not use scheduled times. They provide real time location data.
It’s the countdown clocks that are based on schedules, at least circa 2008 on the L Train. I was asking Ben if the system installed on the IRT is significantly different from what was described in 2008 on the L. And I’d be curious, if they are different, if the L has been or will be upgraded.
I think you’re misunderstanding the technology.
The system uses real time train location data to determine where the trains are. It then uses predetermined scheduled running times to estimate how long it will take the first train to travel from where it is now to where the sign is.
That’s how it’s been from day one.
Ah, yeah, that makes sense, thanks for the clarification.
And the graphical train locations are based entirely on real time data, right? (Presumably the same as is used in the estimates.)
Yes, as far as I know.
The L line and the IRT use two different systems from two different vendors. They both start with a schedule (yes, there actually IS a schedule) and then the predictions for arrival times are computed based on real-time train movement. A tremendous amount of effort has been expended trying to make the predictions as accurate as possible, but they are systems created by humans and dependent upon the actions of humans riding the subways and operating the trains. That adds lots of variables which no system can be created to handle. Oh, and the TLC monitors on the L are based off the same data as the signs.
They are two different systems, but aren’t they both Siemens products?
The other problem with bunching is that it seems to require the trains to slow down. During my morning commute on the 7th Avenue Express I rarely if ever hear any of the dreaded “We are delayed due to train traffic ahead of us” or similar messages. In the evening, however, when when there’s often a long wait followed by a two-minute wait between trains, it seems like we’re far more likely to stop or significantly slow between stations. This is just another example of the inefficiencies created by bunching.
Yes, traditional block signals require large spacings between trains. And when trains switch tracks, it takes time for the interlocking “points” to be set. This all means that bunched up trains will cause delays.
You can see the same thing on the BMT Broadway Local (and even express sometimes). All the trains in 1 direction and especially the N & R come within a minute of each other and then there are no more trains for 9 minutes until the 2 show up together. It was like this with the W too. They sent the uptown W out right after the R! So what’s pointof the W then? Same on the 60th St tunnel where all 3 come 1 after another and then there is a huge gap before the next 3.
While I agree that it’s frustrating, scheduling is a pretty complicated task, especially on the BMT. The N, Q, and R share trackage between Queens and 42nd, and if the N and R are scheduled to have even headways, then the Astoria line has to have uneven headways. The N then diverges from the R and merges back with the Q at Prince – but as soon as the N and Q get into Brooklyn, they diverge from each other, with the N merging with the D and the Q merging with the B.
On the IRT, the top constraints are the flat junction where the 2/3 meet in Harlem and the infamous Nostrand/Rogers interlocking. Getting that right is more critical than having perfectly even headways. At night, while having 2 and 4 trains spaced evenly is good for Atlantic-to-GAP riders who can use either train, it’s bad for West Side-to-New Lots and East Side-to-Flatbush riders who need to transfer. For transfers, the optimal service would have both trains meeting at Nevins, with one then going ahead of the other. (Northbound, the first train into Nevins would be held for the connection with the second to allow transfers in either direction.)
And that’s all assuming that trains are running normally and that dispatchers are paying attention to the road.
I know this is way off the point. I wish that the MTA would post pictures of the 63rd street rehabiliation progress “F” line. I noticed that when they reroute the F for the 2nd ave subway what they are doing there when there is no trains. It like a secret mission
Anyone know why there are some countdown clocks at B division stations such as 28th st on n/r or at 96th street? they have just said the time and date for years….anyone know?
I believe those were part of the L train’s CBTC contract with Siemens. The countdown clocks were going to be set up if the ones on the L worked, but due to various delays and issues, the MTA cancelled the contract after the L was finished.
The MTA does seem to have something with weird scheduling, and it’s not limited to the trains. For instance, the Q43’s printed schedule shows an alternating gap of ten and four minutes between buses during the PM rush hour. To compound this, all Q43s during this time are limited buses, so it’s not like there’s any specific reason to schedule bus bunching (and at the terminal at Jamaica Station, no less).
In Seoul, the government hires mathematicians to coordinate buses, regional rail, and a comparably-sized subway. While that’s never going to happen here because Wall St. will pay top dollar for mathematicians as opposed to the MTA’s comparatively low managerial pay, one can dream, right?
http://www.mta.info/nyct/bus/s.....043cur.pdf
Where do you see that?
Four of the Q43’s start at 179 St, leaving a long gap at the west end of the route. But aside from that, the headway seems pretty steady.
I based my comment off of the schedule at the physical bus stop, not the schedule online. The one online seems a lot more rational, although that brings up the problem of where all the extra buses have been…
Are you looking west of 179 St? Perhaps you’re running into the gaps due to the four buses that only run the partial route.
Is the schedule posted at the stop up-to-date?
Yes, I’m looking at the Jamaica Station stop. If I remember correctly, the stop is updated, and I probably am running into the short-turned service gaps west of 179th.
From my memory, bus bunching is also bad at 179th St, but that’s probably because of the horrible traffic and lack of bus lane enforcement in the area.
Speaking of bus bunching, weren’t the buses supposed to be also subject to some form of GPS positioning? I would love to know if anyone sees all those M60’s sitting at the Marine air terminal between 4 and 5pm (coffee clutch anyone?) … especially when ONE or TWO M60 buses pass us by because they are TOO FULL from the airport to pick up passengers on Astoria Blvd. and then no one else shows up until just before 5pm (because, of course, the bus drivers want to go home and need to get to 125th Street to switch out drivers). Eight buses are supposed to run between 4 and 5 and we never, and I mean never see 8 buses running at that time (unless you count the “Next Bus Please” buses). I have considered bringing my camera with me to make a record of bus numbers …
Speaking of countdown clocks, I saw some on the Eighth Avenue (A)(C)(E) lines and some on the Queens Boulevard (E)(F)(M)(R) lines recently. I also recently saw some on the Broadway (N)(R) line at 28th, 23th and 14th Street stations.