Meeting the Challenge of Serving Large Complex Systems 
                     with Group Rapid Transit
                                   by   Richard Arthur
                                    February 3, 2004

Group Rapid Transit (GRT) systems have been criticized for not being able to
handle the complex passenger sorting process inherent in grouping people by
destination. It is argued that as system size and complexity increases, the
challenge of grouping passengers by destination quickly becomes
insurmountable and consequently only Personal Rapid Transit (PRT) can truly
provide a sufficiently attractive level of service to become commercially
viable. I do not believe that this is an either/or issue. I think that both
PRT and GRT will become commercially viable technologies with appropriate
applications. I think that in some cases, they will actually share the same
track as some such as Kirston Henderson have proposed with his Mega-Rail /
Micro-Rail design.
Automated Rapid Transit technologies including GRT, PRT, and Dual Mode
systems are being proposed as a response to the breakdown of the
transportation system resulting from the spatial inefficiency of the
automobile. This spatial inefficiency has two costs: traffic congestion and
the dominant wasteful role that meeting parking requirements has on the
urban form. But regardless of what any of us thinks of the automobile, its
role in society is so closely intertwined with the perceived attractiveness
of suburban living that it will doubtless continue to play a large and
probably dominant role in our transportation system. The ability to
cost-effectively serve remote locations coupled with the truly personalized
service stemming from owning the vehicle has attractions that no mass
transit device can match.
Consider also that all transportation projects are expensive and difficult.
Consequently any initiative to build new transportation capacity will
necessarily be the result of an overwhelming desire to relieve a major
problem with congestion in the existing network. The goal of the
transportation planners will be to get the biggest bang for the buck. Since
the enthusiasm grand schemes is somewhat quiescent these days, my guess is
that the new systems will be relatively small, relatively simple, and
designed to siphon off as much volume as possible from known choke points.
(This may not be the most cost-effective long term solution and may upset
the purists and planners, but until there is a wholesale change in the tenor
of the political discussion and the political players, I fear that
incrementalism and caution are the watch-words of the day.)
This argues that initial systems will be small and simple and characterized
by having relatively sharp peaks of passenger volume. Such a system may be
more cost-effectively handled by a slightly larger vehicle. As I have said
many times before, as systems grow in size and complexity, the average
volume of demand between any given origin and destination at any point in
time will tend to drop. However, there will still be some high volume
origins and destinations (primary or central activity centers) which will
still maintain relatively high demands even as average demand over the
entire system drops. Finally there is the consideration that since it only
takes a small reduction in traffic volumes to produce a significant
improvement in highway traffic flow, local complacency will quickly
re-emerge as the dominant innovative strategy.
In this world, the first systems may very well turn out to be GRT systems
and if this industry is going to succeed, these initial GRT systems must be
successful. Here is an examination of some approaches to making GRT
successful.
As mentioned above, as system size and complexity grows, the problem of
organizing and grouping passengers by destination quickly begins to boggle
the mind.
However, just as PRT has the analogue of the personal automobile, GRT has,
as its analogue, the elevator. In high rise office buildings, individual
banks of elevators are dedicated to selected destinations and passengers
quickly learn to follow signage to the correct elevator bank. Then they
travel in express mode to a small number of floors where they travel in
local mode. However, unlike traditional transit, the elevators only stop at
those floors where there is a request for a passenger stop. In all but the
highest of the high rises, this method works very quickly and efficiently,
transporting large numbers of passengers quickly to their specific
destinations.
The only significant difference between the elevator operations and GRT is
that each elevator has its own shaft whereas the individual GRT vehicles
each share common main-line tracks. This makes the GRT operation
considerably more efficient, if a bit more complex than the elevator
operation.
In very tall buildings with 80 plus stories, the need to restrict elevator
shaft space results in a series of compromises. One technique, employed
effectively in the old World Trade Center, was to double up on the
elevators. Each elevator opened on two floors simultaneously and passengers
traveling to odd numbered floors would depart from the first floor while
passengers traveling to even numbered floors would board at the second
floor. This approach has no relevance for group rapid transit. Another
technique, also employed in the World Trade Center buildings, the Empire
State Building and many other very tall buildings involves the use of sky
lobbies. A sky lobby is a floor about 50 stories up which would serve as a
staging area for floors above. Passengers would take the express to the sky
lobby and then would transfer to locals for their particular destination.
This approach would double the usage of the elevator shafts since a single
shaft could serve local passengers in the 1-49 range as well as in the 50-99
range.
The difficulty of using the elevator analogy for GRT is that it presupposes
a situation where the vast majority of travelers start or end their
journey at a single destination i.e. the ground floor. This may be a viable
model for a GRT system functioning as a commuter rail system serving a
central downtown, however, in a true regional GRT application, while there
may be a primary flow of travel during the two rush hours, the actual
numbers of origins and destinations is large and the pairings of origins and
destinations is complex.
Nevertheless, there is a precedent for this pattern as well. Here the
precedent is the national airline industry. In the airline industry the
problem of grouping passengers by destination efficiently is solved through
the use of a small number of regional or national hubs. This hub and spoke
system design is similar to the sky lobby model but with the express cars
reaching the sky lobby from a large number of remote destinations.
So once again we find ourselves visiting the sky lobby.
Transit traditionalists are quick to reject this model citing statistical
models showing a decline in ridership for each "seat change". In this
research, the quickest way to lose a passenger is to ask him or her to
change vehicles in the middle of a trip. There are weaknesses to this
reasoning, however. Much of the work documenting ridership reductions to
required seat changes has been done in reference to changes in mode i.e.
between train and bus without allowing for factors such as exposure to
weather and uncertainty regarding the arrival of the new vehicle. Ridership
reductions stemming from poorly timed modal interfaces which leave
passengers shivering the night seem very reasonable.
On the other hand, any traveler on the New York City subway system or several other
major international transit systems will witness a fluid exchange of
passengers between subway lines. I have even observed a significant amount
of activity between express and local trains in search of mere moments of
time savings. (Riding the northbound IRT #6 train - the Lexington Avenue
line, I have seen the same passenger boarding at 59th street who left the
train for the express at 42nd street.) The key to this fluidity is the fast,
frequent service provided by the Transit Authority. This fluidity declines
markedly during off-peak hours when the perceived delay of waiting for an
express which has not reached the station exceeds the potential time
savings. Conversely, during peak periods, one can change from a local to an
express in the hopes of catching the previous local at the next express
stop. This is not the case during the off-peak hours, so passengers tend to
grab a seat on the local and sit.
Since the large number of small vehicles in the GRT system design maximizes
the likelihood of fast, frequent service, the applicability of the peak hour
NYCTA passenger fluidity model seems realistic.
A possible manifestation of this model could be a corridor oriented GRT
serving a series of PRT's acting as local feeder/distributors as well as
circulators within regional activity centers. In the northeast, where there
is a well established commuter rail network utilizing very high capacity
heavy rail cars, the GRT could then act as a feeder/distributor to the
commuter rail line. Thus a hierarchy of vehicle sizes could be matched with
a corresponding hierarchy of corridor capacity requirements.
Another factor to consider is the ready availability of intelligent software
which will allow the vehicle control system to "learn" the travel patterns
of large groups of its regular passengers. As travel patterns emerge over
time, specific trips could be formally or informally built into the vehicle
routing software. For example, if it appears that a group of passengers
tends to travel to between a small group of origins and a particular
destination every day during the morning rush hour, a vehicle could be
formally scheduled for that route. A schedule could be published showing
that a vehicle will leave station N at T time, making scheduled stops at
stations n1...nn and arriving at station D at time T2 or before. The
advantage of the schedule is that it helps the passengers minimize their
wait time at the originating stations, they can time their arrival at the
station to allow time to make the T train. In the absence of a schedule,
vehicle departing and loading times could be adjusted to ensure the optimal
load for passengers traveling to station D.
The difficulties of developing and operating this software in a way such
that vehicle occupancy rates are high during peak periods and overall
passenger wait times are minimized with no passenger encountering undue
delay should not be minimized but they are well within the scope of existing
scheduling algorithms.
The sum and substance of this argument is that in an imperfect world where
automobile drivers face the hassle of traffic and inconvenient parking
spaces and even PRT passengers will have to face the compromises inherent in
getting to and from stations, GRT systems may play a vital role in meeting
regional mobility requirements.

                                                                                 home2.gif (1492 bytes)


                                                                               Last modified: February 19, 2004