Transit by Wire – Automating New York’s Aging Subways

When I left New York in January, the city was in high spirits about its extensive Subway System.  After almost 50 years of construction, and almost 100 years of planning, the shiny, new Second Avenue subway line had finally been completed, bringing direct subway access to one of the few remaining underserved areas in Manhattan.  The city rallied around the achievement.  I myself stood with fellow elated riders as the first Q train pulled out of the 96th Street station for the first time; Governor Andrew Cuomo’s voice crackling over the train’a PA system assuring riders that he was not driving the train.

In a rather ironic twist of fate, the brand-new line was plagued, on its first ever trip, with an issue that has been effecting the entire subway system since its inception: the ever present subway delay.

A small group of transit workers gathered in the tunnel in front of the stalled train to investigate a stubborn signal.  The signal was seeing its first ever train, yet its red light seemed as though it had been petrified by 100 years of 24-hour operation, just like the rest of them.

Track workers examine malfunctioning signal on Second Avenue Line

When I returned to New York to participate in a summer internship at an engineering firm near Wall Street, the subway seemed to be falling apart.  Having lived in the city for almost 20 years and having dealt with the frequent subway delays on my daily commute to high school, I had no reason to believe my commute to work would be any better… or any worse.  However, I started to see things that I had never seen: stations at rush hour with no arriving trains queued on the station’s countdown clock, trains so packed in every car that not a single person was able to board, and new conductors whose sole purpose was to signal to the train engineers when it was safe to close the train doors since platforms had become too consistently crowded to reliably see down.

At first, I was convinced I was imagining all of this.  I had been living in the wide-open and sparsely populated suburbs of Massachusetts and maybe I had simply forgotten the hustle and bustle of the city.  After all, the daily ridership on the New York subway is roughly double the entire population of Massachusetts.  However, I soon learned that the New York Times had been cataloging the recent and rapid decline of the city’s subway.  In February, the Times reported a massive jump in the number of train delays per month, from 28,000 per month in 2012 up to 70,000 at the time of publication.

What on earth had happened?  Some New Yorkers have been quick to blame Mayor Bill De’Blasio  However, the Metropolitan Transportation Authority, the entity which owns and operates the city subway, is controlled by the state and thus falls under the jurisdiction of Governor Andrew Cuomo.  However, it’s not really Mr. Cuomo’s fault either.  In fact, it’s no one person’s fault at all!  The subway has been dealt a dangerous cocktail of severe overcrowding and rapidly aging infrastructure.


Thinking Gears that Run the Trains

Anyone with an interest in early computer technology is undoubtedly familiar with the mechanical computer.  Before Claude Shannon invented electronic circuitry that could process information in binary, all we had to process information were large arrays of gears, springs, and some primitive analog circuits which were finely tuned to complete very specific tasks.  Some smaller mechanical computers could be found aboard fighter jets to help pilots compute projectile trajectories.  If you saw The Imitation Game last year, you may recall the large computer Alan Turing built to decode encrypted radio transmissions during the Second World War.

Interlocking machine similar to that used in the NYC subway

New York’s subway had one of these big, mechanical monsters after the turn of the century; In fact, New York still has it.  Its name is the interlocking machine and it’s job is simple: make sure two subway trains never end up in the same place at the same time.  Yes, this big, bombastic hunk of metal is all that stands between the train dispatchers and utter chaos.  Its worn metal handles are connected directly to signals, track switches, and little levers designed to trip the emergency breaks of trains that roll past red lights.

The logic followed by the interlocking machine is about as complex as engineers could make it in 1904:

  • Sections of track are divided into blocks, each with a signal and emergency break-trip at their entrance.
  • When a train enters a block, a mechanical switch is triggered and the interlocking machine switches the signal at the entrance of the block to red and activates the break-trip.
  • After the train leaves the block, the interlocking machine switches the track signal back to green and deactivates the break-trip.

Essentially a very large finite-state machine, this interlocking machine was revolutionary back at the turn of the century.  At the turn of the century, however, some things were also acting in the machine’s favor; for instance, there were only three and a half million people living in New York at the time, they were all only five feet tall, and the machine was brand new.

As time moved on, the machine aged and so did too did the society around it.  After the Second World War, we replaced the bumbling network of railroads with an even more extensive network of interstate highways.  The train signal block, occupied by only one train at a time, was replaced by a simpler mechanism: the speed limit.

However, the MTA and the New York subways have lagged behind.  The speed and frequency of train service remains limited by how many train blocks were physically built into the interlocking machines (yes, in full disclosure, there is more than one interlocking machine but they all share the same principles of operation).  This has made it extraordinarily difficult for the MTA to improve train service; all the MTA can do is maintain the again infrastructure.  The closest thing the MTA has to a system-wide software update is a lot of WD40.


Full-Steam Ahead

There is an exception to the constant swath of delays…two actually.  In the 1990s and then again recently, the MTA did yank the old signals and interlocking machines from two subway lines and replace them with a fully automated fleet of trains, controlled remotely by a digital computer.  In a odd twist of fate, the subway evolved straight from its Nineteenth Century roots straight to Elon Musk’s age of self-driving vehicles.

The two lines selected were easy targets, both serve large swaths of suburb in Brooklyn and Queens and both are two-track lines, meaning they have no express service.  This made the switch to automated trains easy and very effective for moving large numbers of New Yorkers.  And the switch was effective!  Of all the lines in New York, the two automated lines have seen the least reduction in on-time train service.  The big switch also had some more proactive benefits, like the addition of accurate countdown clocks in stations, a smoother train ride (especially when stopping and taking off), and the ability for train engineers to play Angry Birds during their shifts (yes, I have seen this).

The first to receive the update was the city’s, then obscure, L line.  The L is one of the only two trains to traverse the width of the Manhattan Island and is the transportation backbone for many popular neighborhoods in Brooklyn.  In recent years, these neighborhoods have seen a spike in population due, in part, to frequent and reliable train service.

L train at its terminal station in Canarsie, Brooklyn

The contrast between the automated lines and the gear-box-controlled lines is astounding.  A patron of the subway can stand on a train platform waiting for an A or C train for half an hour… or they could stand on another platform and see two L trains at once on the same stretch of track.

The C line runs the oldest trains in the system, most of them over 50 years old.

The city also elected to upgrade the 7 line; the only other line in the city to traverse the width of Manhattan and one of only two main lines to run through the center of Queens.  Work on the 7 is set to finish soon and the results looks to be promising.

Unfortunately for the rest of the city’s system, the switch to automatic train control for those two lines was not cheap and it was not quick.  In 2005, it was estimated that a system-wide transition to computer controlled trains would not be completed until 2045.  Some other cities, most notably London, made the switch to automated trains years ago.  It is though to say why New York has lagged behind, but it most likely has to do with the immense ridership of the New York system.

New York is the largest American city by population and by land area.  This makes other forms of transportation far less viable when traveling though the city.  After a the public opinion of highways in the city was ruined in the 1960s following the destruction of large swaths of the South Bronx, many of the city’s neighborhoods have been left nearly inaccessible via car.  Although New York is a very walkable city, its massive size makes commuting by foot from the suburbs to Manhattan impractical as well.  Thus the subways must run every day and for every hour of the day.  If the city wants to shut down a line to do repairs, they often cant.  Often times, line are only closed for repairs on weekends and nights for a few hours.


Worth the Wait?

Even though it may take years for the subway to upgrade its signals, the city has no other option.  As discussed earlier, the interlocking machine can only support so many trains on a given length of track.  On the automated lines, transponders are placed every 500 feet, supporting many more trains on the same length of track.  Trains can also be stopped instantly instead of having to travel to the next red-signaled block.  With the number of derailments and stalled trains climbing, this unique ability of the remote-controlled trains is invaluable.  Additionally, automated trains running on four-track lines with express service could re-route instantly to adjacent tracks in order to completely bypass stalled trains.  Optimization algorithms could be implemented to have a constant and dynamic flow of trains.  Trains could be controlled more precisely during acceleration and breaking to conserve power and prolong the life of the train.

For the average New Yorker, these changes would mean shorter wait times, less frequent train delays, and a smoother and more pleasant ride.  In the long term, the MTA would most likely save millions of dollars in repair costs without the clunky interlocking machine.  New Yorkers would also save entire lifetimes worth of time on their commutes.  The cost may be high, but unless the antiquated interlocking machines are put to rest, New York will be paying for it every day.