As I explained a few years ago, automating new lines and even existing lines would have many advantages. Even Sound Transit has recently considered automating the Ballard extension. Some people still consider automated train control (ATC) as unproven even though many airports (SeaTac, JFK AirTrain…) and subways use this technology such as Vancouver Skytrain and new lines on Hawaii and in Montreal already use the technology. Let’s look at the history.
Automating trains is much easier than having fully autonomous cars. As long as the trains run on tracks in a separate right of way, no complex video or even lidar systems are required but track sensors and a more powerful signaling and train control system. Platform doors are often added for additional safety.
Boston already started to experiment with automating trains in 1901. Berlin in 1928. Paris operated a short line from 1952 to 1957 but didn’t have funds for upgrades until the first main line was upgraded in 1966. More lines have been upgraded since. Tokyo started limited operation in 1962, more lines have been added since. Hitachi automated the Disney and Linimo monorail. London started tests in 1968. The Victoria Underground Line finally started fully automated operations in 1967.
While initially the trains still had a driver, fully driverless operation started in Kobe (Japan) and later in Lille (France), Barcelona and airports such as JFK.
In Nürnberg (Germany) they even allowed mixed operation between traditional and driverless trains in 2008.
More modern light metros such as Vancouver’s Skytrain (in 1985), London’s Dockland Light Railway (in 1987) and Copenhagen were designed as automated systems from the beginning allowing for shorter trains without a cab, smaller stations, and higher frequency. Recent new lines in Montreal (REM) and Hawaii also leveraged these advantages. Now many systems around the world are driverless. While Italian Ansaldo/Hitachi Rail STS developed their train systems around ATC, established rail provides (Alstom Urbalis Fluence and Siemens Trainguard MT) added ATC support to some of their existing rail offerings. Siemens AutomatedTrain effort is even automating regular heavy rail operation including the video detection of obstacles along the tracks.
As I explained in my prior post, Sound Transit had evaluated various modes in 2014. This included automated light metro systems such as Skytrain. At that time they decided on light rail technology for the spine to allow for both DSTT as well as at-grade operation. However, for some extensions Sound Transit is now planning, such as Ballard, it might be better to keep them separate rather than interlining them to leverage ATC advantages such as higher frequency and smaller stations.
As the northern Link line is fully grade separated, Link trains could even run fully automated from Everett to downtown. This would allow more precise acceleration and breaking, reduced headways (as close as 60sec) and faster turnaround at the end of the line, reducing the need for pocket tracks. Alstrom claims it reduces trackside equipment usage by up to 20% and energy consumption by up to 30%.
I hope Sound Transit would seriously consider automation where feasible.


I love SkyTrain! Their worst scheduled headways are about 12 minutes late evenings and between two and five minutes during peak hours. I understand that they can theoretically operate with as low as 75 to 80 second headways.
The latest generation sets (Mark V) are over 278 feet long and feature open gangways so riders can walk from one end of the train to the other.
With the exception of Canada Line, the trains use linear induction motors. The first generation cars are not fully retired yet. Though from a passenger comfort aspect, they are abysmal compared to the newer generation trains, the sound they make as they accelerate is so cool.
Marta in Atlanta, WMATA in DC, and BART all have state of art automation at their time. They are not super impressive to these days in a way that many people didn’t know they are also automated to some extent.
I see a pattern here that most of the mass transit automation success story don’t happen in the US, which makes citing those overseas success stories less relatable.
When you have a transit system that is super underfunded, a one time capital investment of state-of-art infrastructure doesn’t solve everything.
Of course it doesn’t “solve everything”, because nothing can do that. In fact, no combination of “somethings” can do that.
But that doesn’t mean that the investments were not worth making. BART has operators on its trains who basically just initiate the departures from station stops. They do that to ensure that the train doesn’t start with someone caught halfway through a door. Nowadays platform doors do that.
Why is this even a question? CBTC is safer than human operators, platform doors make schedule-keeping much easier by eliminating door-holding, and fully allocated costs per hour for operators is nearly $50 per hour. It was foolish for ST to dig the bores from the Convention Center to Northgate without including the electronics for CBTC, but apparently they didn’t. It makes more sense to delay both WSLE and BLE in favor of modifying the tunnel and cars to automate Lynnwood to CIDS at the earliest possible date in order to get the best performance possible from the most important and only truly “critical” portion of the system.
Some people certainly thinks it solves everything. That’s the vibe I pick up here. There are many terribly operated automate system in the US. For starter, many the escalators at Link station.
And what exactly are you asking for here, driveless metro or just CBTC? What Martin described are train with GoA4 or equivalent level of automation that doesn’t require drivers, which can save some operating cost. CBTC is a signal protocol for modern high capacity transit, but it doesn’t give you automated train just already. Many of the New York Subway lines are now on CBTC. It is just one step to improvement efficiency and necessary infrastructure for automation. It can improve reliability for extreme operating headway, but it probably won’t save ST any money at the end of the day unless Sound Transit is all in for driverless technology. It is an investment on better transit quality not really a cost-cutting measurement. Making sound transit train driveless needs another layer of serious investment beyond upgrading signal to CBTC.
A new line such as BLE should go with GoA4 to allow for more frequent and potentially 24h operation. The northern Link line may start with CBTC to allow for more trains, operational efficiency etc. I’m not sure that driverless operation is useful if it only works for a portion of a line. Over time driverless operation may even be possible on the at-grade portions, then of course you could go GoA4 all the way.
I disagree with Martin on few things, but this is one of them. Yes, the north half of Link should be fully automated, with platform doors and CBTC to provide precise train locations and constant speed data. This should happen now that “The Spine” is complete in the high-density sections, even if West Seattle and Ballard have to wait a couple of years for their stubs.
Drivers can board and alight at Westlake with plenty of time to tie down or activate the manual controls before they need to take over or depart.
Thr improvements to the rider experience and capacity — especially important with DSTT1’s narrow platforms — should be the first priority.
It’s not that expensive!
You have to have “CBTC” or something very like it so that the position and speed of trains can be exactly known in tunnels. You could use cell phones on the surface or elevated sections and maybe modern automation systems do that to avoid having another set of wires every where. In any case, the first step is determining exact train location and speed in the tunnels.
Is “GoA4” a Siemens product? If so, I guess that would be the functional technology that you’d retrofit. Can it work with the Kinkisharyo cars?
And I support the idea of upgrading signal to CBTC. It is a less controversial upgrade for an existing train line. It opens the possibility for running more trains per hour. I agree that it was unfortunate they didn’t build the tunnel in a way that can make CBTC upgrade easier. I think ST is pretty much still seeing Link as a light rail while it is actually a light metro.
Bingo! Exactly this. It became a light metro when the extension south of Rainier Beach Station made the step up to structure at BAR and never stepped down.
Yes they made a couple of screw-ups in Bellevue and only saved a few tens of millions of dollars by doing so. But elevating all the way to Lynnwood after they said “No” to Aurora because of the necessity of elevation (and, yes, “it would be four minutes slower”) showed their cards.
I’m glad they have made the transition. But please, let’s not let “Light Rail constraints” be a limitation in the future.
Thanks, Martin. Given the narrow platforms and propensity for crowding delays in DSTT1, it’s sort of amazing that ST hasn’t automated Lynnwood to CIDS already. Yes, it’s a big capital expense to retrofit all the cars, but removing labor costs from 40% of the trackage in the system has to make long-term economic sense. It would also allow the frequently over-loaded North Spine to host short turns between Northgate and CIDS by using the pockets just west of Judkins Park Station and just north of Northgate for very little cost.
Running the West Seattle Link stub with one-car automated trains every two minutes would transform the rider experience for people along California north of the Junction, Alki, or 35th SW from a pretty useless average wait of ten minutes in two transfers inbound to six minutes and thirteen minutes to eight outbound. And of course, riders from all over the region would benefit from having two minute headways on BLE through South Lake Union instead of ten minutes.
That ST plans to build a West Seattle line which must by engineering requirements be entirely grade-separated, at least in the stub-section, but refuses to automate it and plans a half-mile of at-grade running along the BNSF tracks in Interbay in BLE, subverting the prospect of automating that line, are the epitome of “penny-wise, pound-foolish” in the extreme. Just “mini-elevating” the trackway six feet along the playfield would allow automating BLE, avoiding labor costs forever and two-minute headways with smaller stations throughout the line.
The consulting companies should be sued into bankruptcy for such malpractice. They should be screaming for The Bored [sic] to change course. But, ironically, when “cknsulting” on a system that boasts of having “The Spine”, they have none.
Tom, I think you can automate an at-grade section as long as there are no crossings. You may want to fence it off though.
In fact, I’m surprised that ST paved the tracks and didn’t provide more physical barrier through RV which would block cars from crossing or even driving on the tracks (accidentally or intentionally). In Europe such tracks would either run in gravel/ballast or even grass which provides a visual deterrent.
The thousands of Waymo vehicles on the road right now as I write this show that it isn’t necessary to automate only the grade separated sections north of CIDS. That thinking is reflective of the 40-60 year old technology that was used to originally automate the Victoria line and Skytrain. You can absolutely automate an at grade section that still has grade crossings and does not have a fence. If Waymo can have autonomous vehicles safely driving varied routes in mixed traffic, pulling over then back out into traffic, following traffic signals and signs, etc. then automating a vehicle on a fixed route on tracks is completely feasible.
Dude, you miss the point: government can NOT kill people, except through a judicial process or war.
People will be killed by an automated — sometimes as suicides — and the anti-government Maniacs will go into Hyperdrivr.
Link kills people right now, but it can almost always be argued down to negligence by the victim or the operator. An automated train is going to take the blame every time.
No, it’s not “fair”, but it very much IS.
“As the northern Link line is fully grade separated, Link trains could even run fully automated from Everett to downtown.”
Martin, what do you think the chances are Sound Transit will do that? I know you’ll probably want to say you have no idea, but make a wild guess. A 50% chance? 10%? 1%? Zero?
ST has already considered a separate BLE starter line using automated trains. I would say that’s a 30% chance. They also talked about upgrading the DSTT signaling system. They’re also working on their next generation trains. I think eventually that will lead to automation, the question is just when it will happen.
I think it’s worth noting that SeaTac has had automated trains for several decades.
I think the basic points are these: Given the tens of billions being spent, automation avoidance is a political decision. And any thing that won’t open for another 9 to 50 years should be designed with automation in mind — either from day 1 or for conversion later.
It would be like planning horse drawn trains in the 1930’s.
Ooos. I missed the SeaTac reference in the article! Sorry!
I think it bears emphasizing though. There are those who still insist automation is a gadgetbahn — even though it predates the selectric typewriter of 50 years ago!
I’d be shocked if ST used automated trains here, as there are numerous stakeholders who financially and politically benefit from the existing system. Unfortunately, ST is a parochially-run system, where most of its board members or hand-picked based on their political symmetry and are not regular transit riders. As such, they are dependent on transit planners whose decisions reflect the able-bodied riders while excluding the non (long journeys to LRT are common). If board members were ever elected in districts throughout the region and were open to views that many in the well-attuned STB community express, we would see a far superior system.
The fact that you even have to write this article is a testament to the stupidity and structural problem of the ST Transit Board. They clearly don’t know what being a transit leader means. They are still in the toy train mode. And they show no interest in doing their own personal investigating. And then they stick in a fellow member as CEO that has never learned from working on other rail systems anywhere.
It’s time the Board has people who actually get transit as part of their jobs. It should not be a “hobby” board that rubber stamps everything in two hour feel good meetings once a month.
The Board has lots of arrogance and our tax money but no clothes.
The earliest non-experimental automated line I know of was the London Post Office Railway, designed in 1913 and completed in 1927. It was built due to street level congestion delaying mail sorting. Sadly, in 2007, the line was abandoned in favor of surface congestion.
An automated coal hauling railroad was built in Ohio in 1968. It was entirely surface running but separated by fences. I don’t know that they had any trespassers get run over.
The iron ore railroads in northwest Australia have been running their mine to port trains without crews since 2019.
So, automatic trains are already here, and the push for more shareholder dividends will increase the pressure for it’s acceptance on mainline railroads as time goes on.
Transit may well be one of the last places crewless operation persists.
Don’t you mean “crewed operation”, Glenn?