IMG_1736

After three days of riding ULink, my experience has been one of nearly unqualified satisfaction, frequently bordering on jubilation. Sound Transit expertly managed the project’s construction and put together a fantastic opening weekend, and it’s not a stretch to say that the service will revolutionize my life. But one thing has already consistently disappointed, and that’s the supposed real-time arrival signage that Sound Transit is “piloting” at Capitol Hill and UW.

I’ve been watching the signs closely, and it has been immediately clear that the signs are not estimating train arrivals in real-time. There’s a bit of awkwardness inherent in trying to depict real-time arrivals, especially when travel times are dynamic due to bus/rail interaction and because ST has chosen to publish a static 8 minute schedule from Westlake to UW even though the most common duration is 6-7 minutes. Northbound trains are consistently arriving at Capitol Hill 2 minutes early due to the padding in the schedule, yet the signs always say “2 minutes” when the train is in fact already sitting on the platform. In addition, the signs nearly always display even headways of 6 or 10 minutes, betraying that they are either reflecting scheduled information or using a cruder method of real-time arrival estimation.

One possibility is that the data is being fed directly from the Supervisory Control And Data Acquisition (SCADA) system to the information signs, producing a number that estimates an arrival based upon which line segment that scheduled train currently occupies. For instance, if a train is departing Westlake, a Capitol Hill platform sign would say “4 minutes” even though the train will usually only take 2 to get there. But in cases where a train scheduled to arrive is not yet active in the system – for instance just pulling out of the OMF facility – the signs would show a gap in the schedule because the train is not yet traveling the line (see photo above).

I have an email in to Sound Transit about this issue, and will update the post with comment. But let’s hope that Sound Transit will be able to display data that riders can trust, especially now when first impressions are so crucial. Having no information (or having a 2-minute warning like the rest of the line) would likely be better than having visual data that’s often wrong.

75 Replies to “ULink’s Real-Time Arrival Signs Are Anything But”

  1. Totally agree with you on that. For all the wonders of ULink the lack of real time info at both the platforms and the bus stops is disappointing.

  2. I recall SoundTransit posting a position for a real-time arrival project. That was perhaps 1-2 years ago. Where is our real time arrival info? And not just at the new stations, but as a google transit feed so that OneBusAway, Google Maps etc can be used for planning purposes.

    I know ST is reticent to promise availability of anything, but at least tell us that you’re working on it and what the plans are. Be conservative with the go-live date if needed.

  3. Why on earth would anyone care the 3rd or 4th train will leave in 40 minutes? It’s not like they say “gee, I have time to go back to the surface for a quick lunch or something”

    1. Four is a bit much, but it does give you an idea of the train frequency which might not be useful right now, but could inform how you plan future trips.

      1. It makes a difference late at night when you’re not sure if there is actually going to be a 3rd or 4th train.

    2. Imagine the sign when we have East Link running on the same tracks at peak hours. “Red Line in two minutes, Blue Line in five, Red Line in eight, Blue Line in eleven”, and suddenly the sign makes a bit more sense.

    3. Great if you need to run a errand on Capitol Hill… You can glance to see when you need to be back at the station

  4. Don’t hesitate to drop a line that nudges them to change the verbiage of “Departs to Seattle and SeaTac/Airport” ; )

    1. Get rid of that word “Departs”, or at least put it in smaller type in a row above.

      Also, I noticed today that the overhead signs at Westlake say “Link: Univ. of Washington”. Notice what’s not there? The word “Station”. It’s also not on the signs on the side walls identifying the station. So why is it in the electronic signs on the trains? A station by definition is where a train stops to let people on/off, so the word is redundant. The only time it’s sometimes useful is on the outside of a station, if you’re concerned that “Tukwila Station” without the “Station” might be mistaken for “Tukwila Something Else”.

      1. Be glad there are no locales in the area with ‘station’ as part of their name. I could just see Sound transit calling a stop “College Station Station”

  5. The “The next train … northbound … is arriving in …. two minutes” soothing robotic voice has been off since opening too. Both times I’ve used it to the U Dist from U St, the warning has sounded about 15 seconds before the train shows up. Maybe because of this same SCADA issue?

  6. On Saturday, I noticed the same thing. Trains routinely departed while the sign said “4 minutes” or “8 minutes”. It seems accurate enough until you consider that the train is running 10 minute headways. They could have simply programmed the sign to expect departures every 10 minutes, starting at “rand() % 10” minutes after the hour (e.g. no real-time connection to the trains whatsoever), and been just as accurate.

    Accuracy of the data notwithstanding, I also still have issues with the layout of the sign. The large-print text says “Departs to Seattle and SeaTac airport” and the actual numbers are much smaller. I would rather have the directional messages shrunk to simply say “south” and “north”, so that the actual time interval numbers can be larger. Larger numbers means numbers that are readable from further away, which means more time to pick up the pace when necessary to make a train.

  7. It’s not surprising that running trains in the bus tunnel has been problematic if ST can’t predict when the next train is going to arrive. FWIW it seems as if the real time arrival info for RR-B on 148th is close to reality it’s simply because of random variation.

  8. One of the few things that the mbta does well is provide real time arrival information. They have managed to get the system set up so that the arriving announcement is made at the point where the train comes into sight of the platform. That being said they have a similar issue at several underground stations on the green line with multiple branches on the same tracks. In those cases the reader boards display how many stops away the next train is. Eg “e line 2 stops away”. They also have some amazing apps they have let developers create. I especially like the ones that show where every train and bus is on the line. For routes that I use often I won’t even look at the expected arrival and just look at where the nearest bus or train on the route is since bus and trains will move a bit fast or slow sometimes.

  9. I was in Portland a couple of years ago, waiting on the streetcar in The Pearl District. The arrival time said “2 minutes”. I couldn’t see any streetcars coming up the street. I watched the sign count down to “Now”, and still no streetcar. When the streetcar did arrive, the sign said it was due in 12 minutes.

    The problem isn’t limited to Sound Transit.

    1. This has been an issue on the Portland Streetcar for a long time, and I don’t understand why.

      MAX seems to work better.

      I’m guessing that there is a problem with the vehicle location system inside the streetcars, if they even have it installed. In my limited communications with Portland Streetcar, I am extremely unimpressed with how they handle things, and seem to have no interest in installing anything not from the manufacturer into the cars.

    2. This morning on Seattle Streetcar’s First Hill line there was no arrival info, either on the reader board or at nextbus.com.

  10. How do other systems in the world handle this? I think I remember reading that about 80 years ago, on single track stretches, one interurban line had light bulb on every power pole along the line.

    If the lights were all dark, a motorman knew the track was clear through the whole section. When he entered the section, he reached out and threw a switch that turned on a whole section of lights.

    When he left the section, he turned off the lights. With present electronics, couldn’t the train just roll by an antenna?

    Because, and ironclad reason there’s no such thing as “artificial intelligence”, is that all a computer knows is the latest thing the worst-informed idiot told it.

    Mark Dublin

    1. They don’t need any antennas. There are track occupancy detectors (1920s technology really) that sense the current in the track and determine if there is a train there or not. This information goes back to the dispatcher and is indicated on the train location and signal system that the dispatching center has.

      Turning that into a semi-accurate arrival time indication, even when there isn’t any GPS indication, shouldn’t be rocket science.

  11. I disagree with this post totally. The reason we don’t have real time arrival in the rest of the tunnel is obsessive attention to accuracy. It’s extremely useful to have a sense of whether the train is a couple of minutes away or not — that’s what OBA precision is, and it’d be a huge step up over the two minute warning.

    1. What exactly do you disagree with? Ride ULink a few times and you’ll see that the screens don’t give you a sense of whether a train is coming soon or not. Off-peak they’re pretty good, but that’s only because schedule adherence is easier then.

      1. “no more than a couple of minutes off” – so how do you disagree with the post? Consistently being a couple minutes off is NOT real-time arrival information.

      2. If the errors are +/- 2 minutes, that’s useful information, particularly when headways drop to 15 minutes. OBA is generally within this precision and is INCREDIBLY useful.

        ST’s excuse for not rolling out real-time arrival system-wide is the inability to be extremely precise; they’re not willing to do the extrapolation that OBA does in the DSTT. I’ve thoroughly disagreed with that attitude. But this post shows they’ll be reamed if they roll out useful real time information if it’s not perfect.

      3. +/- two minutes isn’t too bad.

        When you are at the platform, the “how long?” is somewhat less of an issue than the “has it gone past yet?” question.

        The problem that Zach notes is that it still shows a train two minutes away when it is at the platform. If that is the case, then someone could be expecting a train that has already arrived and figure they will just take the next one – when the next one is actually closer to 10 minutes away.

        The newest of the MAX station signs seems to automatically take trains off the signs if they have already past, even if the two minute estimate is off. I can’t say if it really works that way, but I have not seen anything on MAX quite like what Zach is reporting. Portland Streetcar is a different story unfortunately – which perhaps means the vehicle location system comes off the shelf with this issue and perhaps TriMet had to make some adjustments to their signs so they work better.

      4. Sure, plus or minus two minutes is better than nothing, but it’s frustrating because there’s no need for it to have such a large margin of error. OBA has large uncertainty because it is dealing with an uncontrolled environment, so a pedestrian, a car waiting to make a left turn, a slight change in traffic can either put you ahead or behind by one light cycle.

        But in the tunnel, you should be able to predict to within 30 seconds just from knowing when the train left the last platform. And we’re living in the golden age of sensors and data processing – it should be easy to set up a vastly more accurate system.

        (You could get location data from cameras, microphones, a data feed from the train itself, the electrical usage of the train, etc. The sensors are all really cheap, so you can deploy a ton of them and just use “how much light/sound is there, where” or you could rely on processing power being cheap, and analyze the pitch, volume, and timbre from just a few microphones, and figure out where the train is by how the tunnel changes the train’s sound (especially powerful if you combine it with the electrical draw of the train, from which you can figure out acceleration, deceleration, hills, etc. And all of this is now doable by machine learning, where you just feed it a bunch of data and the computer figures out how to get the data to predict the train’s arrival accurately).

        What I’m saying is, yes, real time info isn’t that important for high frequency transit, and yes, what we have is helpful for when transit frequency is low. But there is zero technological reason why we can’t have as accurate of arrival info for the next train as we want.

      5. Sure, plus or minus two minutes is better than nothing

        Wrong information is worse than no information. Really!

      6. >When you are at the platform, the “how long?” is somewhat less of an issue than the “has it gone past yet?” question.

        this. Having a sign that says “2 minutes” when actually the train just rolled by and is 8-12 minutes off, is damnably frustrating.

        OBA is useful, but is still not that accurate. On a dedicated right of way where they own everything, don’t have to contend with traffic, and have 90 year old technology to tell them where the trains are, shouldn’t they be able to have better accuracy than OBA? If it’s going to be wrong on future events, at least it could tell you about past events, as in, the train that’s shown as 2 minutes away has actually already gone by?

      7. Moscoiw and St Petersburg have count-up clocks that reset to zero when a train departs. Low-tech and foolproof as long as it can sense the train passing.

    2. I don’t see any excuse when the operations center has location information based on the signal system that should be more than enough to give fairly accurate real time arrival information.

      There really is no excuse.

  12. It’s an annoyance that real-time train info isn’t there, but in a practical sense, it’s much more important for there to be accurate bus info. Are the bus signs installed? Say I’m going to Lake City, I’d need to know if I should wait for the 65 outside the station, or head to Stevens Way for the 372 or 75. Which means solving the “1st stop” problem, where the 372 is listed as 25 minutes late because it’s sitting on Campus Parkway waiting to start its route on time but the system assumes that means stuck in horrible traffic.

  13. On the photo of the arrival sign it shows a 20 minute gap between arrivals. I noticed that yesterday around noon at UW station and I made sure not to dilly-dally and miss the upcoming train. Is there an actual 20 minute gap in the schedule at certain times of day?

    1. No, there isn’t a 20-minute gap. It seems to be a quirk of the way the data is managed.

  14. The arrival signs are simply showing the scheduled arrivals. This became super clear yesterday evening when there was a disabled train at Beacon Hill and trains all over the system were held in place for many minutes. The signs at Capitol Hill kept counting down the arrivals of three consecutive trains at Capitol Hill, even though none of them arrived. But the signs kept counting down.

    More evidence … when a counter reaches zero it never says “Now”. It simply resets to the next scheduled headway.

      1. Mine too. I just missed a northbound train at CH and the sign dutifully counted down from 10 minutes…. The platform filled up, the sign flipped over from 1 min to 10 min again and several people around me were confused to the point of considering going up and getting a bus. I told them the train was probably delayed because of the heavy loads and it would be here soon, which it was.

        Not everybody is used to just knowing a train (or short headway bus route) will be arriving soon regardless of when you get there–and on those rare occasions when something goes wrong it’s just sad the “real-time” count does not reflect that.

  15. Our wretched communications with passengers are more than a nuisance. In any life and death situation on our system, present level of communications is going to get a lot of people killed.

    Real security is more than uniforms and guns. No army in the world goes into a fight without a reliable supply line, medics, and above all, every bit of communications it can get.

    For passengers? What’s happened, where, how to evacuate, which way to go. Real-time to the minute. Security guards? Nobody in uniform on a platform who doesn’t know every escape route, and the training a combat medic would have.

    And communications in the hands of people who all know the system by heart, starting with the location of every emergency exit.

    Budget? Call it Homeland Security. Since it really is, add a lot of fake crap so we can get the money.

    Mark

  16. In Istanbul, some of the train stations have screens which count down to the second, and in my experience, they were pretty accurate

    1. I remember seeing similar things in Hamburg 20 years ago, and agitated German businessmen tapping their toes when we got down to the 30-second mark :) Clearly there are places that know how to do this and do it well. The fact we can not is kinda sad.

  17. I went to the Capitol Hill Station yesterday. The sign said the next train would arrive in 20 minutes. When it arrived seven minutes later the sign didn’t notice and said the next train was arriving in 13 minutes.

  18. My read is simply that there is a clock mismatch somewhere in the system.

    If it takes weeks or months of data to get the signs right, ST should roll out a message that the time that trains depart is either approximate or the system is still in testing. The lack of that makes ST look like they don’t care about passenger information and by association its passengers.

    1. That’s a good point – after all, those screens have a nice place for general and emergency info that currently just tells you what station you’re at, in case the signs on all the walls didn’t tell you. A disclaimer like that would fit quite nicely there.

  19. I consider myself to be one of the nation’s leading arrival sign experts. Why is there a slash between SeaTac and Airport? “Because, Sam, they are making a distinction between the city of SeaTac and the airport!” I understand that. But what do other systems do? MARTA goes to Atlanta’s airport, which isn’t in Atlanta, but College Park, but the station name is not College Park/Airport, it’s Airport Station.

    Sam, arrival sign expert

    1. “Westlake/Seattle” is the same way. It’s just the goofy thing ST decided. “SeaTac/Airport” has the advantage that it’s pronounced like people usually say the airport name, but the slash also hints that it serves the city of SeaTac. There’s the discrepency of SeaTac city not having a hyphen but Sea-Tac Airport does, but the hyphenless style can be seen as stylish and modern like the slash.

      Also, the LA subway has station names like “Hollywood/Highland” which refer to the street intersection, but are spoken as “Hollywood and Highland”.

      And yes, I do believe you consider yourself an expert on arrival signs.

      1. So why isn’t the station name SeaTac Airport, without the slash? I honestly don’t understand why they put the slash in there.

      2. Because the City of SeaTac would be protest that it’s not mentioned. The station is both the gateway to the airport and to the city.

  20. I gotta say those signs should be redesigned by Oran & SounderBruce…. the times could be bigger… a lot bigger.

    Also a 30-seconds-to-board announcement could be made how long to stop taking pictures and dash aboard the chrome train…

    But enough lamenting. I’m not the Washington Policy Center!

  21. The signs seem almost redundant. Even the next train in 2 minutes at Rainier Beach where I used to live was way WRONG.

    The Sounder station signs no better, after 15 years? they don’t tell you anything useful, not even which track the next NB or SB train will be on, or is due. Just the station name (people know where they are already since they found the station). If the ETA of trains was displayed, bus operators could also use this to decide its worth hanging around a couple minutes to collect some additional riders, or if a regular train they collect riders from is a few minutes late. Simple things can add a lot of value.

  22. Funny that you mention SCADA.

    While doing the little light reading of the ST Agency Progress Reports, I noticed that up until December or so SCADA was on a “critical path” for ULink, which either means high priority to install or issue with the installation. Had the contractor resolved the issue completely? Or fixed it just enough to open (I mean, cripes 12 minutes door to door between Husky Stadium and Pioneer Square) with resolution continuing?

  23. What is the root cause of this problem? Not sure why so many other agencies can provide real-time train arrival information and ST can’t.

    1. Other agencies have had it running for a while and fixed any problems earlier. ST just started four days ago. The RapidRide B signs also have a hit-or-miss relationship with reality.

    2. 12th & Jackson has just one display for both directions. It doesn’t even try to say when the next streetcar will come. But I took it outbound yesterday around 6:30pm and just missed a train, and the next one came just six minutes later, which was less than the 15 minutes I expected.

  24. Yesterday, I took a Pronto to Capitol Hill and the sign for northbound said 4 minutes while the southbound said three minutes. When the southbound train arrived, as predicted it said one minute for the northbound train. Then it suddenly switched to saying “6 minutes”, which proved to be correct. No explanation. Obviously, the train wasn’t sitting in the tunnel for five minutes.

  25. Another benefit of automatic operation. Skytrain are accurate because they are running off the same computer that is running the whole train.

    1. Not to mention mostly no obstacles that make scheduled time and actual time extremely different. Witness TriMet’s twitter feed at most any given moment.

  26. And nobody makes any mention of how worthless platform signs are for Seattle Street Car. Signs say 20 or 40 minutes and a car shows up two minutes later.

  27. Yes, the times are bogus. I roude southbound through Capitol Hill Station this afternoon, and the display was visible from the train; it said 3 minutes. As I sat in my seat the train collected people and departed, while the sign switched from 3 minutes to 2 minutes.

    1. Regarding excess dwell times at Capitol Hill to avoid waiting in the tunnel for buses to clear at Westlake: it did wait in the tunnel on that trip. To make sure you get the entire DSTT experience, I guess. This was around 4:20pm.

Comments are closed.