Let’s Talk Seriously About Driverless Trains

With BART losing ridership and revenue off-peak, take a cue from Uber, Waymo, etc.

The Copenhagen Metro (seen here in  Vanløse station) is one of many systems with no drivers on board the train. Photo: Wikimedia Commons
The Copenhagen Metro (seen here in Vanløse station) is one of many systems with no drivers on board the train. Photo: Wikimedia Commons

BART ridership is down. That may seem hard to believe, given rush-hour overcrowding, but fewer people are riding BART off-peak, with a drop of nine percent on weekends. A recent KQED story looked at the numbers:

The district reported bringing in $480 million in fare revenue for fiscal 2015-16 and projected an increase of about 6 percent, to $510 million, for this fiscal year. BART financial staff now projects the agency will collect about $492 million in fares — $18 million less than forecast.

It’s not hard to figure out what’s going on with BART’s off-peak ridership. Suppose you want to go from 24th and Mission to San Francisco International on a Sunday afternoon. The trip only takes 24 minutes by BART–about as long as it takes by car. The problem is, if you time it wrong, you can end up waiting another 20 minutes for a train. As Streetsblog explored in a previous post, to maximize transit ridership, service has to be utterly reliable and frequent–especially in the age of Uber and Lyft, one can’t expect to be competitive with 20 minute headways on an urban metro system.

What if instead of BART running six-to-ten-car trains every 20 minutes at night and on weekends, it ran two-car trains every five minutes? That kind of frequency would make transit more competitive. The problem, of course, is labor costs–BART would lose money on the proposition, thanks to the additional drivers.

Everybody knows how to solve this problem.

Uber, Waymo and others are in the process of developing driverless car technology, while transit systems in the Bay Area continue to rely on one operator in the front of every single train (even though computers already drive BART trains–the operator is just opening and closing doors and acting as a backup).

Meanwhile, there are some 20 fully driverless, “GoA4”  train systems (meaning there is no requirement to have an operator on board) throughout the world. Many have been in continuous operation for decades.

Citilab did an excellent breakdown of all the systems a few years ago. The advantages can’t be overstated: better frequency, punctuality, reliability, capacity, and better fare-box recovery.

So why isn’t BART, the transit system of the supposed tech capital of the world, actively pursuing the obvious technical solution to its ridership woes? “A higher level of automation at BART would benefit passengers by allowing more frequent trains,” said BART director Robert Raburn. “I asked for a pilot project between Millbrae and SFO.”

That will be a great start, if it actually happens. In Streetsblog’s conversations with other transit officials, two reasons for resisting GoA4 are usually given.

One is safety. In an earlier interview with Streetsblog, Director Lateefah Simon said “When BART trains go under the Bay with hundreds of millions of pounds of water pressure above, I want to know there’s staff on those trains. We live on earthquake faults. We need trained people who understand disaster preparedness.”

First of all, if the Transbay Tube rips open in an earthquake, it’s hard to imagine what an on-board driver is going to do besides become another victim. Furthermore, if that’s really the issue, backup safety “operators” with emergency training could simply ride back and forth between Embarcadero and West Oakland. The rest of the system could be operator free.

The other issue that’s frequently brought up, and it’s probably the bigger obstacle, is organized labor. “I don’t believe the contract would even allow for this. An operator is required to run a train,” wrote Alicia Trost, a BART spokeswoman, in an email to Streetsblog.

But union resistance is based on a false assumption: that automation means the drivers get canned. It doesn’t.

Laurent Fortune is an engineer who managed the automation systems for Line 1 and Line 14 of the Paris Metro. Drivers weren’t fired–in fact, they were promoted to “supervisor” or transferred to other areas.  “There are still lots of humans around, they’re just not at the front of the trains,” said Fortune. He added that they are switched to places where “they’re more useful.” Additionally, these “supervisors” still have to be trained to drive the trains in an emergency, or to bring a train back to the shop if the automation on a particular train has failed.

A stop on one of Paris’s automated lines. There are still “humans” in the system, they’re just not at the front of trains. Photo: Wikimedia Commons

That said, with full automation “you get around 30 percent more service, you have a better service with more trains all day long, because train operation is less costly,” he said.

In other words, you don’t fire operators, but you also don’t have to hire more of them or pay overtime to run more trains, improve service, and increase revenue.

A GoA4 system is also far more dynamic. Fortune recalled a time there was an unexpected political demonstration around 2:30 in the morning. At the request of the police, the RATP (the Paris transit system) was able to keep the subway open and get the demonstrators home. As he explained, BART could never respond to a last-minute demonstration because you wouldn’t get the operators in place “until the next morning,” when it would no longer do any good. With a fully automated system, staffing is consistent in the control centers, regardless of how many trains are running.

In fact, Streetsblog confirmed that the Clipper system already maintains a count of how many people are entering and leaving BART at any given moment at any station. In a GoA4 system, that data could be fed into the operational computers, which could lengthen trains and/or send them more frequently as needed. And each train in an automated system carries more people, since there’s no need for dedicated control cabs (usually there’s a  pop-up control counsel for emergencies in a GoA4 system).

There would be much work to do on BART to automate it fully and get rid of the human door operator. Many of the world’s systems don’t have platform doors throughout, but it would obviously be prudent to add platform doors/a barrier along the platform edge in BART’s busiest stations. The trains would need object detection: computers that can stop the train in case someone or something fell onto the tracks, but, unlike driverless car tech, this is off-the-shelf, reliable and proven technology for trains.

BART’s labor contract expires in 2021. It’s not too soon for BART management and union leadership to figure out a way to get the one-train-one-operator clause out of the contract, perhaps by building in guarantees that drivers will not be fired. Then BART can start working on full automation. Because with Waymo and Uber’s automated cars sure to come online soon, alternatives to BART are only going to get cheaper and more available, and BART’s revenue issue is only going to get worse. BART has a responsibility to start solving this problem now–especially since the tools for doing it are already available.

Skytrain, Vancouver's driverless Metro, has been running safely for over 30 years. Image: Skytrain
Skytrain, Vancouver’s driverless Metro, has been running safely for over 30 years. Image: Skytrain
  • Jeffrey Baker

    BART would need a new fleet to do this, of course. Automation won’t be able to deal with a dragging brake, a flat spot on a wheel, or a stuck door, all daily occurrences on BART.

  • Andy Chow

    BART already has driverless trains which is the Oakland Airport Connector.

  • mx

    I do the calculation you describe for most of my airport trips. I live a few blocks from a BART station, so I would seem like a target customer for BART to SFO, at least for most trips where I don’t have a ton of luggage to carry. But I have to factor in the ~10 minutes to walk to the station and get down to the platform, 5+ minute buffer time because I’m stuck there for 20 minutes if I miss the train, ~30 minute ride to SFO, and the walk at the airport because BART drops me at the far end of the international terminal. And I’ll pay $8.95 for it. Without traffic, it’s about a 30 minute drive and around $22 in a Lyft.

    Yes, it’s more expensive (which isn’t so much a factor for business trips, most companies will be fine with either), but the total trip time is close to half. And there’s no danger of a long wait at the BART station if I timed it wrong. It’s hard to treat BART as a good option here unless there’s heavy rush hour traffic.

  • xplosneer

    That’s the point? Drivers rehired as supervisors would be able to fix these problems more effectively if they weren’t driving… They could be on the train and have a fault alarm system, but wouldn’t need to be up front.

  • Patrick Devine

    Aren’t BART trains already Type 2 or Type 3? The technology has been around for decades to fully automate the entire system. Vancouver’s system has more annual riders and more stations (and around the same number of daily riders) in an area with far fewer people. Oh, and the trains come during rush hour about once ever 90-120 seconds. What’s crazy is that there still is overcrowding and they’re having to lengthen the platforms.

  • vcs

    Could have happened back in the 1970s, but probably will never happen. They would hire cheaper janitors and station agents first, and that will probably never happen. The founding ideal of BART was it is as much of a jobs program as a transit system.

  • thielges

    I’m afraid organizational inertia is the biggest impediment to driverless BART. The technical problem isn’t that hard and as this article demonstrates has been successfully deployed decades ago.

    The irony is that even though solving the technical problems of driverless cars is at least a hundred times more difficult, we’re likely to see automated taxi cabs roaming our streets before a fully automated BART.

  • calwatch

    You’d probably have ambassadors/conductors on board during the rush hour to help manage crowds. Rather than one driver at the front there might be two or three making sure doors close properly. I would not expect peak train throughput to increase much, but the off peak headways could be reduced greatly.

  • Alex

    Bart tube’s under 60psi, not “millions of pounds”… just sayin. I guess it would still hurt.

  • david vartanoff

    If you dig up pre operation BART agit-prop, they were planning driverless trains. Unfortunately the first generation train control system was grossly dysfunctional, so it was easy to insist on human backup. As to frequencies, mx is correct. Twenty minute headways (evenings, all day on weekends) renders BART close to useless if one has any other options. The other way to increase non rush hour ridership is to offer passes (as Muni riders can buy for intra SF rides only). New York city showed that unlimitd use passes greatly increased off hour usage. Time for BART to behave like a real urban subway system.

  • Isnt BART already capable of being driveerless with existing technology? The conductors open and close doors not because they have to, but because they need a distraction so they dont fall asleep

  • DragonflyBeach

    Skytrain doesn’t have 420,000 weekday riders like BART does. However, it is a larger percentage of the population.

  • Patrick Devine

    It’s less than 10% difference. Skytrain has 390k daily riders vs. 420k weekdays riders, which is why I said “and around the same number of daily riders”. Skytrain also has a lot more riders on the weekends.

  • Ray

    BART needs to focus on where they are cost-competitive, during peak commute times. Implement congestion-pricing on the roads, and BART will become even more cost-competitive. Leave the times of the day when there is plenty of empty road capacity to the road-based vehicles. They provide the best service. Just provide a subsidy to the poor to use Uber-like services during these times. It will be cheaper than operating BART.

  • Rachel H

    I’m struck by these sentences: “What if instead of BART running six-to-ten-car trains every 20 minutes at night and on weekends, it ran two-car trains every five minutes?” and “with full automation ‘you get around 30 percent more service…'”

    I am outraged that labor relations are blocking a transition to this remarkable improvement to our transit system’s backbone. GoA4 systems have been proven safe, sensible and effective. We could have this without needing to approve, buy or construct any new track, tunnels or rolling stock.

    The Mercury News reported that as of 2013, BART employed 3,430 people [1]. If about 3/4 of those are drivers, then about 2,500 people are blocking transit improvements that affect our region’s population of 7 million people.

    I applaud BART director Robert Raburn for requesting a pilot project between Millbrae and SFO. It should start as soon as possible. People are stuck in traffic, waiting for it. Please do what it takes to promote drivers to “supervisor” or transfer them to other areas.

    It’s bad enough that the new federal administration is blocking funds for CalTrain electrification, and that state republicans are blocking high speed rail, another technology that the rest of the developed world takes for granted. I didn’t know until tonight that job paralysis was also blocking this obvious step to keep “the innovation economy” moving forward. [2]


    [1] http://www.mercurynews.com/2013/07/26/bart-workers-paychecks-already-outpace-their-peers/

    [2] https://www.nytimes.com/2017/03/06/us/trump-and-republicans-block-caltrain-grant.html

  • Vooch

    if you get rid of the fireman & driver who will shovel the coal into the locomotive ?

  • Lee Haber

    Actually it’s a bit higher than BART now. With the latest Evergreen Extension, Skytrain now carries 450,000 riders per weekday. http://www.tricitynews.com/news/updated-evergreen-extension-ridership-hits-30-000-per-day-1.10392055

  • bdawe

    to be fair, the platforms are fairly short to begin with

  • Tran Carl

    BART is still better than Caltrain. Caltrain’s weekend service runs every 60 min and soon reduce frequency to every 90 min. Thanks for strict labor regulation during electrification construction.

  • DominicL

    There didn’t appear to be any labor hangups when implementing the driver-less Oakland Airport connector. Or were there?


BART Staff Opposes Rush Hour Bike Access on Rear Cars

BART staff has shot down a proposal from Director Robert Raburn to allow commuters to bring bicycles aboard the back cars of rush hour trains. In a memo [pdf] to the BART Board, Executive Planning and Budget Manager Carter Mau recommended maintaining the current policy, which prohibits bicycles aboard its most crowded trains during peak […]

BART Transit Blogger Roundtable, Part II

Flickr Photo: pbo31 Last week regional transit bloggers sat down with BART management and asked questions of General Manager Dorothy Dugger about her transit operation.  This is part two of that roundtable: Impact of economy on operating budget The biggest challenge that I’m focused on right now is certainly the effects of the economic reality […]