Link operator Kevin Gumke isn’t worried a robot will take his job anytime soon.

At least 55 metro lines in 37 cities around the world are fully automated, according to the International Association of Public Transport, an advocacy group that promotes public transit.

Many of these fully automated lines are closed systems, in contrast to Sound Transit which, at times, mixes with other vehicular traffic. The agency says it will eventually study automating or semi-automating parts of the Link system, but today, ST uses a hybrid system with operators and computers working together to operate the train. Here’s how it works: 

20 Replies to “Operating Link Light Rail”

  1. Thanks, Lizz. And congratulations. Congressional Medal of Honor is easier to get than a cab ride. But refresh my memory: Do drivers run the DSTP in ATP or manual mode? And what about between Westlake and UW Station?

    I’d caution everybody about treating complete automation of our particular railroad as a given. Let alone a good idea. In our current standing-up stage- we looked like African mongooses-us people have had several million years of practice staying alive amid real quick life and death changes.

    And also the sole ability to “program” – meaning keyboard in the answer to every single decision our machine has to make – into a machine. Old adage: Want anything done right…..?

    You’ve got five seconds to Google the pertinent speed, and location,of the double-bottomed gasoline tanker westbound on Holgate, and link to the right answer. And hope whoever was on programming duty morning after game-night caught the last Malabyte caution.

    So look at it this way. Even after we undercut every major arterial and eliminate every single grade crossing….we might finally be able to run MLK with hands-recently fully- opposed thumbs and all-taken off the controls.

    But we’ll be losing a critical learning and training facility. We’ll have sections that can permit personnel to be out helping and informing passengers. Until somebody needs to take the lid off those controls and run the fully-informed passengers to next station before trip to The Shop. Like Sky Train.

    MLK line would make a terrific museum train ride through Old Columbia City. But just remember: If “Driverless” isn’t an old-time railroader’s joke about an inept human operator, it should be.

    Mark Dublin

  2. This is a really cool video, Lizz! Thanks for putting it together!

    Something else I’ve always wondered about light rail operations: how are switches set on a light rail line? Are the switches set by the control center based on the train’s assigned route, or is there hardware on the train that requests switches to flip to a particular direction? And are humans controlling the switches, or does the human just pick a destination and the computer changes all the switches necessary on the way?

    1. On operating trackage there is the equivalent of “CTC” (“Centralized Traffic Control”). Computers with available human override control the turnouts which are rarely used except at terminal stations and at route divisions. Trains operate under signal control.

      I believe, though, that in the Maintenance Facility when a train approaches a facing-point turnout the switch uses power-on, power off like Electric Trolley Bus overhead.

      Trailing-point turnouts may be “spring switches” in which the points are by default set for the through track and bend to allow a train from the entering track to pass, or they may be powered and line for the direction from which a train approaches.

      Speeds are so low and movements so relatively rare that line of sight operation is possible

      1. Switches operated by power on or power off may also be on the mainline. Next time you are at Gateway look north and notice the “reverser in neutral” signs.

        First Hill Streeetcar would be ideal for spring switches. You always want the car to follow a particular track when the car reverses direction.

  3. I continue to wonder why operations along MLK can’t be improved. Seems like you should be able to achieve very close to 100% signal pre-emption. Sometimes it works well, but anecdotally i see (or am riding) trains that have to stop and wait at signals…sound transit always blames it on emergency vehicles screwing up the signals…

    1. Good chance that SDOT has some needs, or wishes, that conflict with ours. Is there a new building- and its utilities- being built anywhere along the line?


    2. I notice that the signal detectors seem to be failing all over Seattle, and signals change without any cars or pedestrians there. I’m not sure if there is an effective system in place to report them. When I’ve reported them, I’ve gotten messages from SDOT staff that the problem was addressed — when it either wasn’t fixed or only was working right for just a few days.

      Lately, I’ve noticed that train approaching lights go on then seem to time out, just in time for the train to arrive and get delayed!

      Signal priority is really cool and useful. It still seems to require good human monitoring or it gets dysfunctional pretty quickly. It only takes one or two malfunctioning signals to create major delays.

    3. MLK was a lot worse the first year or two. Since then I have occasionally experienced a train waiting for signals, but not often. Has it gotten worse recently?

  4. Al, City Council member Rob Johnson ,4th district around UW used to be CEO of Transportation Choices Coalition.

    Currently on the Sound Transit Board, chairman of the Audit and Reporting Committee.

    So pretty sure he knows the answer and cares about it. So at least he’ll give you a straight story.


  5. Lizz;

    Congrats on getting the best seat in the house and on a great video! You did awesome!

    That said, a few notes on the video:

    a) Happy to see sexy light rail break 50 miles an hour. I would hope it could break 75 MPH like an Amtrak Cascades going between Tacoma & Seattle. Not too optimistic about my North by Northwest… ;-)

    b) I sure wish you could have a Light Rail Simulator but sadly kids – for some time – aren’t into driving your local trains on the computer or playing SimTransit where you’re a boardroom star. I’ll leave it at that, this is a transit blog after all.

    c) We simply cannot have a ST4 that does not start with elevating the Rainier Valley (even a cut and cover tunnel like some of TransLink’s Canada Line into hard basalt would be impossible) part of the Ballard-Tacoma line…

    Great work Lizz!


    1. Joe, where is this “hard basalt” you mention? Honestly curious; the USGS’ geological map for the city only shows three types of bedrock in the entire city, and of those three only one is basaltic andesite. That formation is actually some distance south of the Link line where it crosses I-5 along the Boeing Access Road. There is a decent amount of bedrock in SE Seattle, but it is predominantly sandstone (and even at that the Link line/MLK misses most of it). If we were to undercut through intersections rather than tunneling the entire line, we’d only need to do so at seven locations (Alaska, Orcas, Graham, Othello, Kenyon, Cloverdale, and Henderson – the last two are close enough that one perhaps could be eliminated). Of those, only Orcas (possibly) and Othello are over the sandstone bedrock strata.

      Even if it were basalt, that’s not impossible to cut through – drive on almost any highway in Central Washington and you’ll see that. Fortunately, though, there is little bedrock to deal with in our area. That doesn’t mean elevating the line is necessarily a bad idea, just that it isn’t a requirement for grade separation.

  6. Nice to see that the train can only speed for two seconds before the computer automatically takes over (and the operator presumably gets a talking to), so the Amtrak situation can’t realistically happen on Link.

    1. Alex, it’ll be a very brief one-way conversation, and I don’t think you’d like the conclusion if you’re the subject. But main reason it’s not going to happen on LINK is that a lot of conditions for the accident are missing.

      None of LINK was recently upgraded from a small-scale freight railroad with a top speed closer to thirty than eighty. Or featured a dangerous curve left uncorrected for budgetary reasons.

      Or a railroad that would permit a man at the train controls who honestly did not know exactly his train’s location and speed, by both training and instinct.

      Computer train controls? Of course use them. One more fallback could still save many lives. But don’t blame the lack of them for the loss of Train 501. Or that their presence will save a lot on wages, training, and retention.

      When anything digital fails, it’ll go with no warning at all. And the technology really hates heat, dirt, shock and rough handling. Meaning ground transit itself.
      So good habit of mind is that in our hearts and minds, we never see a computer as our final Train Protection.

      Real last safety measure? Two very old mechanisms. A spring-loaded lever with a linkage that if the driver’s hold ever lets go of, the train’s every brake locks the wheels to the tracks.

      In the grip of a supremely skilled, trained, and respected worker. Who, if the worst happens, will have the train at best speed, location and control humanly possible when he or she dies.


  7. Great video. I’ve been taking Light Rail almost daily and still a little stumped with how inconsistent the acceleration and deceleration can be. Now I’m wondering if it has something to do with the throttle the driver is using or if certain drivers are to blame. Some starts and stops can be really smooth while others really jerk the train around with more pulsed deceleration and torque-y non-linear acceleration. I notice this because I am usually loading/unloading my bike from the hook during the accel/decel periods and find it surprisingly difficult to have stability even with a wide stance. Can’t wait for this operation to eventually be fully automated since the computer will probably have much more consistent actions with gyroscopic feedback and calibration. Still would be glad to have a driver up front to deal with any abnormalities in the future.

  8. Andrew, your last sentence is the more reliable answer to smoothing out LINK operations than the one before it.

    Exactly like with trolleybuses, smooth operation requires some pretty extensive training in vehicle-handling. There’s even a specifically correct way to handle the control lever- which does not come unlearned.

    Which costs money, but less than recompense for damaged bikes and body-parts. And also public relations- rendered more expensive by Lyfft, Uber, and bright green bikes. Helps clear the ones thrown down station stairs, too.

    Meantime, encourage your fellow passengers, especially the bike-securing community that you’ve got enough experience with good train-handling that you won’t accept any excuse for the other kind.

    Buy reporter Mike Lindblom an ORCA card if his Seattle Times editor won’t. And KIRO has Go-To man named Jess for complaints about public aggravations. Pretend you’ve already bought gift cards for Steve O’ban and Bob Hasegawa but don’t really do it.

    Because Steve will take you up on it in front of cameras if he was the one who ordered the training budget cut in the first place.

    But can recall recent subway rides in Stockholm where subway ought know better, where automatic mode was definitely rougher than average driver. Remember, automatic control is only as smooth as whoever programmed it can run their own train,

    Incidentally, sentence last comment about the role of the “Dead Man” control wasn’t about honorable suicide for shame over an imminent disaster.

    Meant that only really reliable last defense is a driver whose train operation is so deeply drilled and indoctrinated that he or she is constantly in such good control of the train that if the lever lets go and triggers the train is in its best possible position and condition to minimize death and destruction.

    Starting from the time the driver first grips the “Dead Man” same motion as putting control lever to leave the platform.


    1. Mark, all I’m saying is that a driver is sitting in a comfortable seat facing forward in the first seat on the train. They are not getting as much feedback from the train as say a passenger standing towards the back, facing sideways. A slight acceleration/deceleration from the perspective of a driver comfortably seated in the cockpit could feel like nothing while its tossing around everyone in the rear of the train. I suspect a programmer would take the job seriously and utilize feedback sensors throughout the train as the eyes and ears of the computer. We all like to pretend that drivers arms are well calibrated machines but I suspect we would be sorely disappointed in a head to head competition with a well calibrated computer.

  9. What isn’t a clear glass wall between the main cabin and the operator so riders get this view?

  10. … and precisely what is this video teaching us? Nice view, but is the driver doing and why?

Comments are closed.