Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 35
35
stop data and produces adjusted boarding and alighting data additional APC units to enhance the ridership database,
at the stop level. but this argument is difficult to make if staff is not com-
pletely satisfied with the underlying quality of the data
The first problem is errors within the APC technology. produced by the existing APC units. On the other hand,
Boardings and alightings do not balance on any given trip, staff also struggles with how much effort to dedicate to
resulting in incorrect loads. A second problem that contrib- generating other outputs from the APC data, given the lim-
utes to the first is that the system is set up to parse data that ited fleet of APC buses available for data collection, and
may have occurred "off-route." Every trip is programmed the complications described earlier regarding the complex
as a series of stop intervals. As a trip progresses, the sys- scheduling requirements.
tem calculates distance traveled and constantly confirms
adherence to the preprogrammed compass headings and Overall, Metro is moderately dissatisfied with the perfor-
proximity to the coordinates for the next stop. Distances, mance of its APC system in terms of its ability to deliver
stop coordinates, and compass headings were populated usable data on passenger counts. The primary benefit to date
by driving a staff car. Data errors, odometer problems, or is a general picture of boarding and alighting trends at the
excessive lane changes can result in mismatches from the bus stop level. Problems include inaccuracies in boarding
preprogrammed adherence guidelines, which can result in and alighting counts across an entire trip, which lead to inac-
the bus falling back into an "off-route" status. The sys- curate loads, and lost data owing to the inability to maintain
tem will disregard the raw APC data collected during any route adherence correctly. The latter problem often results
period of time the bus was either correctly (owing to actual from detours or other unusual operating circumstances, such
detour) or mistakenly (as a result of data mismatch) in this as special events, but can also stem from insufficient or inac-
"off-route" status. curate route data programming.
One solution to load issues is to "zero out" the load at the If Metro could go back and change only one thing, the
end of each trip. Metro schedules a fair amount of interlines agency would go into the process with a better understand-
and loop routes, and in many cases passengers remain on ing of potential pitfalls. The request for proposals for the ITS
the bus at the "end" of a trip. Another issue with interlin- procurement included a requirement for 95% accuracy in the
ing is that it becomes more challenging to collect a single APC units, and the prime contractor has subsequently been
day's data on a given run. Some core routes may have up to on-site addressing various issues. More staff involvement
20 blocks providing trips. This issue is exacerbated by the and attention in the testing and acceptance phases would
geography of Madison. There is a narrow isthmus between have been useful, but staff limitations played a role here as
two lakes, with three primary corridors that buses use to well. The primary lesson learned is to set better equipment
access the downtown and university campus. This results acceptance standards and testing.
in a network structure where multiple routes provide service
along these three trunk corridors. Metro is not able to get a
single-day 100% count at a given stop on one of these three
trunk corridors.
Metro has only been able to find effective use for its APC
data at the stop level. Each trip serving a given bus stop is
assumed to have an APC bus assigned on at least a few days TRIMET (TRI-COUNTY METROPOLITAN
within a pick (typically 90 days). An external reporting pro- TRANSPORTATION DISTRICT), PORTLAND, OREGON
cess is used to calculate the average number of boardings and
alightings at each stop for each trip, and then to sum these TriMet was a case study for TCRP Synthesis 29 and is a long-
averages for all the trips serving this stop over the course of time user of APCs. TriMet began operation on December 1,
a day for a total count of daily boardings and alightings by 1969. In 1975, the agency began operating Fareless Square in
stop. The resulting data are used primarily within the agency downtown Portland, 2 years before the opening of the tran-
to provide an order of magnitude type of comparison of pas- sit mall. The Banfield light rail line began service in 1986,
senger activity by stop. and the light rail system now has four lines. TriMet serves
an area with a population of 1.3 million and operates 526
Overcoming the various issues with APC data is a chal- peak buses and 81 peak light rail vehicles. Annual ridership
lenge. As is the case at many small and medium-sized is 101.6 million (2006).
agencies, limited staff resources affect what can be done.
APCs are just one of many ITS-related projects that Metro Over the past 25 years, the agency's IT staff has sought
has implemented in recent years, and staff has had lim- to develop new and innovative applications of APC data.
ited opportunity to dedicate the time needed to iron out A fairly large IT staff and a strong analytical staff that has
the problems. Planning staff recognizes the benefits of worked together with APCs for a long time have been two
OCR for page 36
36
positive factors in TriMet's success. The addition of a CAD/ approximately 80% of the raw APC data being used for analy-
AVL system several years ago improved the passenger sis. A recent study indicated that APC counts are accurate at a
counting program by greatly enhancing and simplifying the confidence level of 95% and an error of +5%.
ability to identify bus stops.
Previously, TriMet would balance loads by block, but now
TriMet is a good example of an agency that developed its the program adjusts loads at the trip level. Boardings and
own analytical software, mainly because nothing of that sort alightings are averaged, and adjustments are made at stops
existed when it first implemented the APC system. The in- with the greatest passenger activity. Loads are zeroed out at
house program is driven by block-level and stop-level data, the end of the line if the bus lays over for at least 5 minutes.
which are then matched to the schedule and the bus stop. The This condition allows for non-zero end loads on through
bus stop matching program can identify and flag a location trips. The balanced loads are used only for load-related anal-
with no matching stop and assign the data to the nearest bus yses; stop-level analyses use the actual APC counts.
stop. This capability was developed in the early days when
stop matching relied on odometer data and continues to be In-house programs also address other data anomalies. For
useful in many circumstances today. example, if a bus lays over before its designated layover loca-
tion, running time and load data are not registered correctly.
Approximately 75% of the bus fleet is equipped with side- TriMet developed a routine to use time-stamp information
mounted APCs. The integrated APC/AVL system collects from the AVL system to correct the data. Another routine
data every time a bus passes a stop or opens its doors. Data associates boardings at a layover point with the following
upload occurs in the garage through PCMCIA (Personal trip and alightings with the previous trip. IT staff has devel-
Computer Memory Card International Association) cards oped these and similar routines to address the data oddities
that are removed from the buses by operators and placed that have been observed over 25 years.
into a networked computer by station management person-
nel. This older technology has its challenges. It is difficult All APC and AVL data reside in an Oracle database
to obtain new 1-megabyte cards. Today 1-gigabyte cards are maintained by the IT department. Scheduling and payroll
common, but the data transfer takes much longer with the data are integrated into this database. The open nature of
larger cards. The data retrieval and processing automatically the database provides access to a wide variety of data and its
happens every night and data are available the next day. inclusiveness encourages innovative analysis. TriMet is cur-
rently examining factors contributing to absenteeism among
Overhead APC units are installed on light rail vehicles. The bus and light rail operators and is able to explore a num-
APC database treats rail like bus, except that adjustments are ber of hypotheses related to loads and on-time performance
made if only one car in a two-car train is equipped with APCs. through the database.
The Oracle database contains all the scheduling information
needed to make these adjustments. APCs transmit data wire- Use of APC data for NTD reporting is a topic of great
lessly, but sometimes the overnight transmission does not hap- interest. TriMet keeps a separate database to calculate loads
pen and data are not available the next day. The next generation as an input to the passenger mile calculations required for
of APC for the bus fleet is likely to use overhead beams. NTD. The reason is that the standard validation program
adjusts for negative loads, which biases the load estimate
TriMet did use a program that randomly assigned APC- upward. TriMet works with James Strathman from Portland
equipped vehicles to specific blocks, but no longer does so. State University to ensure that any errors in APC boardings
The Service and Performance Analysis manager will check and alightings are random. FTA approved TriMet's use of
periodically that all blocks have been scheduled for APC APC data for NTD in 1986, but has since changed its pro-
within a given pick. All new buses have APCs, but only cedures and requires annual validation of APC data, with a
about 30% of the oldest buses in the fleet are APC-equipped. minimum sample of 100 trips. FTA granted TriMet a waiver
All buses collect AVL data, therefore arrival and departure for its most recent NTD report, but the agency is conducting
times at stops are always available. Specific requests for an manual validation this year. TriMet would prefer to see a
APC bus assignment are occasionally made, usually for a periodic validation requirement of every 3 to 5 years instead
low-frequency low-ridership route that is under consider- of an annual requirement for NTD reporting.
ation for discontinuation or service adjustment.
Maintenance of APC units is the responsibility of elec-
The overnight data processing includes automated programs tronic technicians in the maintenance department. Origi-
to ensure that the beams are working correctly, to match data nally, when TriMet had less than 60 APCs and no AVL,
to schedules, and to validate boarding and alighting totals. The TriMet had one project manager/technician to perform the
validation program checks that boarding and alighting totals equipment maintenance and data programming/processing
differ by no more than 10% and adjusts for negative loads. requirements. With the increasing number of electronic com-
Any suspect data are removed from the system, resulting in ponents on the buses, additional technicians have been hired
OCR for page 37
37
and the APC system is only one of their responsibilities. The department in conjunction with data users. Lessons learned
maintenance director understands the importance of APCs include the following:
for the agency, but repairs can take longer than expected.
· Check and validate all data.
TriMet is very satisfied with the performance of its APC · Develop accurate and reasonable techniques to balance
system. The primary benefits are a large amount of statisti- loads. Load data are often the weak link in an APC
cally valid ridership data, greater confidence in the accuracy system.
of the data, and no need for ride checkers. · Develop routines to address end-of-trip data. Consider
where and when to zero out loads and handle other
TriMet has such a long history (25 years) with APC data anomalies that can arise at the end of the trip.
that it has been able to address all the problems encoun- · Realize that a close working relationship with the infor-
tered. For example, the agency noted that it took about a year mation technology department is essential. TriMet's
to work out all issues in the integration of APC and AVL success was aided by a large IT department that has
databases. The system works very well and there is a high developed and refined data processing and reporting
degree of confidence in using the data. Because of this his- procedures.
tory and the current state of affairs, TriMet was hard pressed · Use the open architecture of the agency's Oracle
to answer the question of what it would go back and change database and integrate with the computerized
if it could. scheduling software to access a wide variety of
data. TriMet relies on this extensive database as
TriMet provides an excellent example of successful APC it continues to develop innovative analytical and
implementation using internal software developed by its IT reporting techniques.
