Sunday, June 27, 2010

Driverless metros: All they are cracked up to be?

For many transit activists, the "gold standard" of rapid transit is the driverless metro.  While expensive to build, driverless systems are held to have a significant advantage over human-operated systems--the absence of a driver, and his/her salary and benefits.  In the developed world, it is typically the case that a) labor is the predominant cost in transit operations, and b) most transit systems have a farebox recovery ratio of less than 100%--these two factors combined produce the result that in many systems, the limit on how much service can be provided isn't the amount of rolling stock or the capacity of the infrastructure, but the amount of operating subsidies which are available.   By greatly reducing the operating subsidy (or in some cases, creating lines which turn a profit), a transit agency can offer much more frequent or comprehensive--and therefore more attractive--service.

The problem, of course, that driverless metro is expensive to build, and politically difficult to boot.  Current automatic train operations systems are designed for a protected right-of-way--which generally means grade-separation.   Allowing the public to cross the tracks seriously complicates things.  And the driverless control systems themselves are expensive to design and install.  An additional issue in many large cities, where the expense can be justified, is that transit unions are often sufficiently powerful to prevent driverless metros from being installed (labor is generally opposed to such systems, for obvious reasons)  A third issue is public skepticism of the control technology--after the 2009 accident on the Washington Metro red line, blamed on a failed train-detection sensor (since repaired), the automatic train operation system has remained turned off.  (The DC metro was never driverless, as operators were present on the train even when the ATO systems were enabled; but the switch to full-time manual operation has lowered system throughput).    Despite the stated advantages of the technology, there are at present less than two dozen fully driverless rail systems in operation worldwide.

Say what?

However, the benefits of driverless metro, once installed and operational, are generally considered to be axiomatic by transit activists.  Which is why I was astonished to find British Columbia transit activist "zweisystem", one of the contributors to the Rail for the Valley blog, make the following claim on a thread on the Transport Politic.

Jay, SkyTrain, being driverless doesn’t mean there is no one present on the metro system. Vancouver’s SkyTrain system has a large cadre of attendants who are on the train and at stations to maintain the integrity of the metro system. Added to this, we also have the SkyTrain police to counter increased crime on the metro system.
All driverless transit systems also maintain a Small army of electrical specialists, technicians and operators to keep the metro in operation. Unlike LRT, which has drivers, which can deal with small problems, on a driverless or automatic transit system, when a problem arises the metro stops until someone tell the computer it is safe to proceed. Sometimes this means an attendant must walk along the guide way to ensure there is no problem with the metro and is very time consuming.
SkyTrain has more employees (and higher operating costs) than LRT systems and what was once the flavour of the month in the 1980′s, is now reserved for the most heavily used metro lines in the world, where there are cost benefits with automatic operation.
Those who keep on proposing automatic metros for minor transit lines, really don’t know what they are talking about.

The thread was about Seattle's Link system, a light-rail system which is fully grade-separated (but not driverless).   Zwei seems to be making two separate claims here:
  1. The overhead of maintaining a driverless system cancels out any cost reduction from not having to pay drivers, except at very large scales (which the SkyTrain system allegedly doesn't meet)
  2. The lack of drivers makes such systems (or at least SkyTrain in particular) less reliable, as many contingencies that could be handled by an on-board operator instead require a more difficult manual intervention, resulting in greater disruption of service.
Zweisystem is a bit of a controversial figure in many transit forums; generally due to his dogged opposition to SkyTrain expansion in the greater Vancouver.  His opposition to SkyTrain stems from a variety of factors, but a significant complaint seems to be that BC authorities are spending a disproportionate amount of transit dollars in the metro Vancouver area, and neglecting the Fraser River valley east of Surrey.  Furthermore, zwei claims that the expenses involved with SkyTrain leave little money for improving transit elsewhere in the province.  Zwei frequently calls official TransLink statistics into question, and there have been other accusations of the system being the beneficiary of pork-barrel politics.  I agree with zwei that commuter rail would make sense in the Fraser valley, should funding be available.  Not being a British Columbian (or even a Canadian), I am neutral on the question of what provincial funding priorities for the province ought to be.

However, zwei's anti-SkyTrain advocacy frequently morphs into a general anti-metro (and pro-surface rail) position, independent of any specific application.  Zwei has published numerous articles on the railforthevalley blog which underlie his claims:  Several may be found here, here, here, and here.  I generally consider such blanket praise (or condemnation) of a particular technology to be questionable--so with that in mind, let us consider each argument in turn. 
    Does driverless cost more?

    Zwei's first claim is undoubtedly true if rephrased:  Driverless systems incur a significant overhead--the staff needed to maintain the control systems and infrastructure (computer systems, track sensors, on-board control equipment).  Some of these duties are essentially fixed, and other duties scale with the amount of trackage and rolling stock.  There is some level of service--I don't know what, and it depends on many local factors, in particular the state of the labor market--below which driverless operation would be more expensive than manual operation, even if the capital costs could be ignored.  This is true for many technology choices--there's a service level below which manual grade-separated metro doesn't make sense; there's a level below which surface rail doesn't make sense--and there's a level below which even plain local bus service is inefficient (and paratransit or dial-a-ride service would be more economical).  This analysis focuses entirely on operational efficiency, it should be noted, and not on the quality of the service to passengers.  (Were transit agencies to focus entirely on financial metrics, the logical outcome for most of them would be to cease operations altogether--obviously, not a desirable outcome for many reasons).

    Some of the other roles zwei mentions--police, security, station attendants--perform functions which generally aren't performed by drivers on manually-operated rail systems.  MAX operators, for instance, are not charged with duties such as customer service (beyond dealing with on-board incidents), fare collection/enforcement, giving directions, etc--they are focused on operating the trains.  In addition, many of these roles, even if assumed to be required for a driverless system, scale with the amount of trackage or stations, not with the number of trains in service.

    An important question to ask in any analysis:  What is the marginal cost of train service?  In other words, assuming that the rolling stock and line capacity exists, how much does it cost to add an additional train?   

    Zwei uses, for the foundation of his claims, an 1989 paper written by Gerald Fox (a fellow who was involved with the design of MAX) entitled A Comparison of Some New Light Rail and Automated Guideway Systems (abstract here, full paper appears to be unavailable in electronic form).  The abstract, unfortunately, doesn't shed any light on the situation--it claims that LRT "may" offer lower operating costs than driverless systems, but not stating under what circumstances this is true.  Given the advances in computer and control technology which have occurred in the two decades since the paper was written, and the changes in the labor market (public sector wages are generally higher compared to private-sector wages than was the case in 1989), what was true back in the 1980s may no longer be true in the 2010s. 

    What about SkyTrain?

    Zwei is fond of comparing SkyTrain to C-Train, Calgary's excellent light-rail service, and one which has similar levels of ridership despite operating in a city half the size of metro Vancouver.

    Currently, SkyTrain's Millennium and Expo lines operates peak service at 108 second intervals over their combined segment--a figure which is likely the capacity limit of the existing infrastructure.  That's 33 trains an hour.  C-Train's peak service frequency over its combined segment through downtown is less--presently 5 minutes (12tph), with plans to reduce peak headways to 3 minutes (20tph); the minimum possible headway is 2 minutes.  Both systems peak headways are excellent; at these frequencies, adding trains becomes more of an issue of increasing capacity rather than improving service.

    However, there is a significant difference in off-peak service:  C-Train service frequencies reduce to 3-4 trains per hour during the off-peak hours; whereas SkyTrain's service frequency, on the Expo and Millenium Lines, never drops below 7.5 tph.  Operations in off-peak hours aren't constrained by capacity; they're generally instead constrained by cost and staffing issues. 

    SkyTrain's Canada Line is another matter, unfortunately.  Despite also being driverless, this line's peak headway (8 minutes) is no better than the off-peak headway of the other SkyTrain lines; and headway along the branches increases to 20 minutes during late-night hours.  South of the Bridgeport Station, the line splits into two branches serving Richmond any Vancouver International Airport, respectively; these branches include significant single-track sections.  OTOH, the decision not to run trains more frequently during off-peak hours is curious.  Unlike the Expo/Millennium line, which runs diagonally across the street grid for a considerable distance, the Canada Line is much shorter and runs parallel to numerous city streets with excellent bus service.  

    Is driverless less reliable?

    Zwei's second claim--that driverless metro is less reliable compared to other types of transit--also flies in the face of conventional wisdom.  Zwei argues that driverless metro systems experience a longer recovery time from minor incidents (such as a passenger maliciously or accidentally pressing an alert button without a genuine emergency onboard), which could be handled more expeditiously with operators on board--turning them into major events affecting the system as a whole.  (In particular, he suggests that many incidents require manual intervention from inspectors who must travel down the trackways on foot to reach a stopped train).  Unfortunately, no reliability or availability statistics are provided by Zwei to support this claim.

    Assuming zwei's observations are true (which, if they are, might be an issue with the particular implementation of SkyTrain--minor incidents which could be handled by an onboard operator can generally be handled remotely, including closed-circuit communication with passengers; and true failures of the trains, signals, or tracks are generally beyond the capability of a driver to resolve), one must balance these against the issue of human error causing a service disruption. 

    This paper reports on Paris' experience with driverless systems, and claims that the driverless portions of the Paris Metro have availabilities of 99.2-99.5%, with a significant part of the disruptions which do occur being caused by passenger misconduct.  Paris opened its first driverless line (a rubber-tired metro) in the 1990s, and has been converting many of its existing rail subway lines to driverless operation over the years; and the experiences of both the operator and the public has generally been positive.

    The preceding observations don't take into account the issues with surface-running rail; where service disruptions caused by collisions, track blockages, etc. are more likely to occur.  (Surface rail cannot be made driverless with the current state-of-the-art in control technology).

    In many ways, this line of argument is a common one in the transit industry--more complicated (and expensive) systems which are capable of higher performance under normal circumstances, are frequently held to be more subject to severe service disruptions than are systems which are lower-tech or more distributed.  Similar arguments are frequently made in support of busses and against rail--busses can simply drive around any incidents; whereas major incidents on rail lines often take the entire line out of service (and typically resort in the deployment of busses to ferry passengers around the affected section of track).


    There are many valid criticisms to be made of TransLink, and of BC provincial government.  There are also many valid criticisms to be made of the SkyTrain system.  However, the more general conclusion that metros are bad (driverless or otherwise), except in cities of New York or London scale, doesn't seem to hold water. Is Portland "ready" for a system such as SkyTrain?  Probably not--out transit patterns are too downtown-focused, and the Portland metro area is too auto-centric outside the inner city.  If the population continues to grow, and the newcomers are accommodated by increasing density rather than by sprawling outwards, then maybe--but today, the present system is probably adequate.  (Conversion to a more metro-like system will probably occur in phases, such as construction of a downtown tunnel to bypass the slow slog through downtown streets that MAX presently makes).  But for other cities which more transit-friendly densities, fully grade-separated metro is an option that makes sense; and driverless operation then becomes an option worth considering.


    1. Just a note regarding SkyTrain.

      The light-metro has been on the market now for over 30 years and during that time has undergone at least four name changes (ICTS > ALRT > ALM > ART) and only seven such systems have been built.

      During the same period, over 200 new LRT systems have been built or are under construction, or are in the final stages of planning.

      The problem with SkyTrain is that there is no bang for your buck, so to speak.

      It also should be known that the new Canada Line metro system is incompatible with the present SkyTrain system, which should give many pause to wonder: "If SkyTrain is so good, why didn't they build the Canada Line with SkyTrain?"

      Also the newly opened Canada Line's construction costs escalated so much that the scope of the product was greatly reduced and the now unofficial $2.8 billion price tag (Susan Heyes court case)now sees a subway system so designed can only achieve a fraction of the capacity of a $1.8 billion cheaper light rail line!

      I have never been anti metro and when ridership demands (300,000 to 400,000 passengers per day), then an automatic metro in a subway would be a correct solution.

      We do not have this type of ridership in Vancouver as SkyTrain is being force fed bus passengers (Translink claims 80% of SkyTrain's passengers first take the bus to the metro) which is a great deterrent in attracting new ridership.

      In fact, despite a now $8 billion+ investment in SkyTrain and RAV metro, the percentage of ridership using public transit has not noticeably changed in the past decade and a half and has stagnated around 14%.

      Added to this, car use and gas consumption is up, despite the fact that Vancouver has the highest gas prices in the country and there is no money to start building the Evergreen SkyTrain Line, East of Vancouver.

      Also please don't use the density gimmick, until one states what density is needed to sustain what type of transit, to date I know of no density/transit matrix, showing what transit works in what densities.

      The metro lobby speak bravely, but as in Europe, building metro for the sake of building metro has lead to economic woes for the operating authorities.

      In Vancouver, the talk of increased density translates into windfall profits for friends of the government, who happen to buy land at key station sites, by up-zoning density on their newly acquired properties.

    2. @zwei:

      When I speak of "SkyTrain", I speak of the entire system; not of just the linear-induction-powered Expo and Millenium Lines. When I speak of "driverless metro", I refer to all such systems in existence.

      Certainly, there are more LRTs in existence than driverless metros; like there are more bus systems in existence than LRTs. Many applications are better met with the cheaper solution, and as noted in the article, driverless is expensive to build.

      To address the rest of your points:

      * The cost issue between the Canada Line (using traditional trains) and the rest of the system, with linear induction, is the cost of the propulsion systems; not the cost of the control systems. LI, while having the advantage of enabling really light trains, is expensive and proprietary, certainly, but this has nothing to do with who's driving the train.

      The Canada Line designers made a significant mistake by single-tracking the ends of the line--limiting capacity to a fraction of what one observers on the other parts of the system. This, too, has little if anything to do with driverless.

      Where do you come up with the 300000-400000 figure for metro (driverless or not) being more efficient than LRT? I'm not questioning it--I'm simply curious. (And is that figure for a given LINE--few transit lines get that ridership, and only in the largest of cities--or for a system?)

      Taking the bus to the train doesn't strike me as a problem. One of the point of major trunk lines is that they are generally served by more local services on less infrastructure-intense technologies, particularly local bus. Were a city to build a network of surface rail in a grid pattern, as Prof Condon seems to suggest for Vancouver, one still would likely need to change trains for many trips.

      If you're arguing that "no new commuters" were attracted because formerly all-bus trips were changed into bus+rail trips, I don't see the problem there, either--improving transit times for existing passengers is also important. (The same criticism gets made here in Portland concerning MAX--that a lot of MAX users are former or transferring bus users. In either case--so what?)

      There isn't likely a "density/transit" matrix available; such things depend on many local factors (culture, socioeconomics, etc).

      Here in Portland, we wish we had 14% transit use. The facts on the ground are quite a bit different than in Vancouver--we've got lots of freeways, and lots of human-powered commuters, but our transit usage is far lower than Vancouver despite the two cities being of similar size (Portland is slightly smaller).

      At any rate, metro (driverless or otherwise) is a tool, just as LRT, BRT, and local bus service are tools. I think we agree on that--we may disagree somewhat on where the cutoff points are. I have no comment on the motivations of BC authorities in their transit decisions.

    3. I have built and operated both driverless and manual systems all over the world. The reason that they are in the minority is the initial capital cost that most smaller cities cannot afford. The problem with North America in particular is that most cities do not forecast their needs properly and go on the cheap. Indeed, most politicians vote for extensions, but do not consider existing infrastructure or that you might need a few more trains if you are going to make the system longer. They also usually opt for light rail when Medium or Heavy would be the better long term choice for the taxpayer.
      However, the best advantage for a driverless system is the reduced administration and coordination needed to match trains with drivers. A manual system requires on average 1-2 hours to mobilise a driver to an empty EMU if there is a sudden demand to increase service. A driverless system allows immediate insertion of a train into service at a moment's notice. Throughput is another advantage that manual systems cannot match.
      I would also suggest that if you check the Reliability, Availability,Maintainability and Safety stats (RAMS), you will find that driverless systems lead manual systems by a significant margin. If they are designed correctly (the old SEL system designed in Canada and now sold by Thales for example), they are incredibly safe. The on-board supporting systems on Bombardier trains also have automated fault response to ensure correct actions are taken on board in the case of a detected fault. Drivers often miss some of these alarms or respond to them incorrectly causing significant damage often to the point of jeopardising safety.


    Keep it clean, please