Академический Документы
Профессиональный Документы
Культура Документы
December 2010
Dedicated to my dear sister Deepti Muley
Keywords
BRT, busway, capacity, lost time, crowd, transit, public transport, interface, dwell
time.
Bus Rapid Transit (BRT), because of its operational flexibility and simplicity, is
rapidly gaining popularity with urban designers and transit planners. Earlier BRTs
were bus shared lane or bus only lane, which share the roadway with general and
other forms of traffic. In recent time, more sophisticated designs of BRT have
emerged, such as busway, which has separate carriageway for buses and provides
very high physical separation of buses from general traffic.
This research identified new parameters concerning busway station operation, and
through intricate analysis identified the elements and processes which influence the
bus dwell time at a busway station platform. A new variable, Bus lost time, was
defined and its quantitative descriptions were established. Based on these finding
and analysis, a busway station platform bus capacity methodology was developed,
comprising of new models for busway station lost time, busway station dwell time,
busway station loading area bus capacity, and busway station platform bus capacity.
The new methodology not only accounts for passenger boarding and alighting, but
also covers platform crowd and bus lost time in station platform bus capacity
estimation. The applicability of this methodology was shown through demonstrative
examples. Additionally, these examples illustrated the significance of the bus lost
time variable in determining station capacities.
Abstract iii
List of Tables ix
List of Figures xi
Chapter
One 1Introduction 1
1.1 General 1
1.2 Background 1
1.3 Research motivation 2
1.4 Research hypothesis 3
1.5 Research aim and objectives 3
1.6 Scope of this research 3
1.7 Relevance of this research 4
1.8 Thesis outline 5
1.9 Publications from this research 7
11References 157
Appendix 12
Table 2.5 Efficiency of multiple offline linear loading area at bus stops 41
Table 4.4 Bus flow rate and passenger demand classification split 69
Table 5.3 Passenger boarding and alighting during evening peak period 85
Table 5.8 Effect of fare collection policy on boarding time per passenger 92
Table 5.9 Effect of fare collection policy on alighting time per passenger 92
Table 6.8 Assessing normality for loading area 1 (Off-peak period) 105
Table 6.9 Assessing normality for loading area 2 (Off-peak period) 105
Table 6.10 Assessing normality for loading area 3 (Off-peak period) 105
Table 6.11 Assessing normality for loading area 1 (Peak period) 106
Table 6.12 Assessing normality for loading area 2 (Peak period) 106
Table 6.13 Assessing normality for loading area 3 (Peak period) 106
Table 6.15 Descriptive characteristics of bus lost times (Peak period) 116
Table 6.16 Descriptive characteristics of bus lost times (Off- peak period) 116
Table 9.2 Occupancy and blocking rates for loading areas at outbound platform of 148
Mater Hill Busway Station (Afternoon peak period)
Table 9.3 Number of effective loading areas calculation for bus station platform 149
Figure 1.1 Developed framework for busway station platform bus capacity analysis 4
Figure 2.2 Lane configuration of a busway and its station (Brisbane, Australia) 12
Figure 4.1 Camera positions at Mater Hill Busway Station (Outbound platform) 59
Figure 4.5 Boarding and alighting at inbound platform of Mater Hill Busway Station 66
Figure 4.6 Boarding and alighting at outbound platform of Mater Hill Busway
Station 67
Figure 4.7 Number of buses servicing outbound platform of Mater Hill Busway
Station 68
Figure 4.8 Matrix for data mining of passenger demand and bus flow 70
Figure 5.3 Passenger – bus interface duration and its dependent variables 79
Figure 5.5 Distance to loading areas from the waiting area on the busway platform
(Off-peak) 82
Figure 5.7 Variation in bus lost times over platform crowd by loading area 87
Figure 6.3 Bus lost time probability distribution curves (Peak period) 109
Figure 6.4 Bus lost time probability distribution curves (Off-peak period) 110
Figure 6.5 Comparison of peak and off-peak bus lost time probability distribution
curves 111
Figure 6.6 Bus lost time cumulative distribution curves (Peak period) 112
Figure 6.7 Bus lost time cumulative distribution curves (Off-peak period) 113
Figure 6.8 Comparison of peak and off-peak bus lost time cumulative distribution
curves 114
Figure 7.1 Overview of model form for busway platform dwell time estimation 121
Figure 7.2 Effect of bus lost time on dwell time at loading area 1 (Peak period) 124
Figure 7.3 Effect of bus lost time on dwell time at loading area 1 (Off-peak period) 125
Figure 7.5 Effect of bus lost time on dwell time at loading area 2 (Off-peak period) 126
Figure 7.6 Effect of bus lost time on dwell time at loading area 3 (Peak period) 126
Figure 8.2 Variation in busway station bus capacity with boarding load per bus 138
Figure 8.3 Effect of bus lost time on busway station bus capacity 138
Figure 9.1 Trajectory of bus processing at the Mater Hill Busway Station
(Outbound platform) 143
Statement of Original Authorship
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institute. To the best
of my knowledge and belief, the thesis contains no material previously published or
written by another person except where due reference is made.
Signature _____________________
Date _____________________
Completing this PhD has been the most challenging task for me in the journey called
life. In this journey, I own my deepest gratitude to my principal supervisor, Dr.
Jonathan Bunker, and Associate supervisor, Prof. Luis ferreira, for their advice
encouragement, and support.
I would also like to thank my sister and fellow PhD colleague, Miss Deepti Muley, for
her everlasting support in various aspect of my research.
I also greatly appreciate the support, help and expertise received from Queensland
Transport’s Busway Operation Centre, Brisbane and in particular Mr. Jurgen
Pasiezny and Mr. Andrew Haddock.
I am grateful to Mr. Daniel Buntine for his assistance, especially during the data
extraction phases.
My special thanks go to my parents, relatives and siblings for their efforts and best
wishes which have provided me this unique opportunity. Last, but not least, I would
like to thank my wife Shraddha for her understanding, love and support.
Introduction
1.1 General
This chapter establishes the motivation behind this research and defines its aim and
objectives. This is followed by a description of scope and relevance of this research.
This chapter then outlines the structure on this thesis.
1.2 Background
Bus Rapid Transit (BRT) is rapidly gaining popularity with urban designers and
transport planners to address the ever increasing needs for fast and assessable, yet
economical and reliable transport. Busway is one form of bus rapid transit (BRT),
and consists of dedicated roadway infrastructure exclusively for use of buses. It is
designed to provide a very high level of physical separation to buses from general
vehicle and other forms of traffic. A key advantage of a busway is that the bus can
serve suburban communities using local and arterial roads and then enter the
busway to run limited stop or line haul. Busways generally have stations located
further apart than on-road bus stops, and in some cases as far apart as a suburban
rail system. Thus, the busway can provide a premium transit service of quality
approaching that of rail. Other reasons for the increasing popularity of busways
include their simplicity to operate the bus service and flexibility to provide more bus
routes and frequencies when the demand arises.
Some of the well known busway networks include Ottawa’s Transitway opened in
1983, Pittsburgh’s Martin Luther King Jr. East Busway opened in 1983, Brisbane’s
South East Busway network opened in 2001, Auckland’s Northern Busway started in
2008, and Adelaide’s O-Bahn Busway, opened in 1986, which is a guided busway
with its unique specially-built track for buses.
In response to this increase, The Translink Transit Authority (TTA) is increasing bus
service frequencies and routes to increase and maintain the system efficiency. Even
so, such high busway patronage means that its stations have crucial roles to perform
for the smooth operation of its busways. Firstly, a station must accumulate
passengers until their desired services arrive. Secondly, a station should facilitate a
smooth process of boarding and alighting of passengers with their desired buses, so
that buses can be accommodated by the station without any non-service related
delays.
However, waiting passengers on the platform can lead to crowding, which can
interfere in passenger boarding vis-à-vis bus dwell time. Therefore, impacts of
platform crowd on bus dwell time needs to be considered in estimating bus
throughput capacity of a station. Currently there is no methodology available to
analyse the effects of platform crowd on the boarding process and bus dwell time.
The established bus capacity analysis methodology for busway (TRB, 2003) is
primarily based on the operational characteristics of a bus stop adjacent to a bus
only lane, which lacks account for the effects of platform crowding. Thus, there is a
need to develop a busway station bus capacity analysis methodology which can
approximate the operation of a busway station.
The subject of this thesis is to develop a purpose made methodology for busway
station bus capacity analysis. The hypothesis, aim and objectives of this research
are given in the following sections.
“Passenger walking and the prevailing crowd at a busway station platform influence
the bus dwell times”
This hypothesis emphasizes that the traditional bus dwell time models cannot be
used for busway analysis because these modes do not account for accumulation of
passengers at the station platform. A detailed discussion on development of this
hypothesis is presented in Chapter 3.
minimal influence on platform crowd. Hence, the data collection for this study was
limited to a busway platform where boarding passengers were dominant.
Chapter 1 (this chapter) establishes the hypothesis, aim and objectives of this
research. It describes the scope and contributions of this work.
Chapter 2 reviews the existing literature relevant to this research and identifies
the gaps in the area of busway analysis.
Chapter 4 reviews the state of art in data collection and develops a set
methodology concerning collection, extraction and analysis of data for
this research.
Chapter 5 analyses and evaluates the parameters and develops the variables. It
sets the foundation for data analyses.
Chapter 6 establishes the quantitative descriptions of bus lost time and defines
its descriptive characteristics.
Chapter 10 concludes this thesis and identifies the contribution and innovations of
this research. It also provides guidance for further research.
Chapter 2
Literature review
Chapter 3
Research problem development
Chapter 5
Parameter analysis and evaluation
Solutions
Chapter 6
Modelling bus lost time
Chapter 7
Busway station dwell time model
Chapter 8
Busway loading area capacity model
Chapter 9
Busway station efficiency model
Chapter 10
Conclusions
Figure 1.2: Structure of this thesis
Literature Review
2.1 Overview
This chapter reviews the relevant literature in the field of bus dwell time and busway
station bus capacity estimation. The next section first provides an outline of different
forms of bus rapid transit (BRT) systems. Later section 2.3 outlines the classification
of various bus stops and stations and their impact on the transit system is discussed
in section 2.4.
Section 2.5 discusses in detail the role and impact of a busway station. Since this
research targets the effects of passenger walking and platform crowding at a busway
station upon the bus dwell time, a review of studies related to pedestrian flow and
density is also reviewed in section 2.6.
The past research in the area of busway bus capacity analysis is presented in
section 2.7. The chapter closes with section 2.8 identifying the gaps in existing
knowledge of busway bus capacity analysis.
The above definition highlights the operating characteristic of the BRT system - a
bus service which combines suburban door to door service with a high speed line
haul transit of rail transit. A more detailed definition illustrating the design and
implementation of BRT system was given by the TCRP Report 90 -
There are many forms of BRT system in use in different parts of the world. Most
common forms are, but not limited to, exclusive bus lanes, and dedicated busways.
Figure 2.1 shows BRT configuration a) with bus lane and b) with busway. An
exclusive bus lane is a traffic lane reserved for bus use only. This is a relatively
cheaper option however it provides limited improvement in transit speed and
reliability. A busway, on the other hand, can be a fully grade separated exclusively
built rightway for buses and can provide greater improvement in transit speed and
reliability.
A busway usually has a non-overtaking regime where buses are not able to overtake
one another on the corridor. The necessity of overtaking does not generally arise on
the busway corridor since buses operate at common speed on any given section.
However, busway stations are mostly designed with provision of a passing lane to
facilitate buses to overtake stopped buses. This makes the busway station design
and management highly important for smooth operation of the busway corridor.
Figure 2.2 shows Brisbane busway and its station.
When a bus is required to observe a stop it needs to decelerate from its cruise
speed. Similarly, after passenger service, the bus needs to accelerate to get back to
its cruising speed. These deceleration and acceleration from/ to cruise speed also
add delay time to the bus journey time. The deceleration and acceleration time
delay can be mathematically modelled as follows (Changshan and Murray, 2005;
Wirasinghe et.al., 1981).
1 1
0.5 Equation 2.1
Where,
= Total delay time at stop i associated with bus acceleration and
deceleration.
= Bus cruise speed
= Acceleration rate
= Deceleration rate
The door opening and closing time is the time needed to fully open the door from
the closed position and vice-versa. It depends on the mechanical properties of the
bus door. The value for door opening and closing time can be obtained from on-site
observation. In the absence of such data the delay due to door opening and closing
is usually taken as a constant varying between 2 and 5 seconds (TRB, 2003;
Changshan and Murray, 2005)
The Clearance time is the time a bus takes to clear the loading area after
completing the passenger servicing at the stop and making it available for the next
bus. The clearance time can be influenced by the type of loading area (TRB, 2003).
If the loading is on-line, being within the traffic lane, then the clearance time will
equate to the time required by the bus to start up and travel its own length plus the
time for the subsequent bus to pull into the loading area. However, when the loading
area is off-line, or out of the traffic or passing lane, the bus driver needs to find a
suitable gap in the adjacent traffic flow to re-enter into the flow. This re-entry delay to
bus is an additional component of clearance time.
The fourth component of bus stop delay, delay due to passengers boarding and
alighting, is the most important of all delays. Unlike the other three delays, this
delay varies from stop to stop and is highly sensitive to passenger demand. The
delay due to passenger boarding and alighting is commonly assumed as a dual
linear function of the number of passengers boarding, and of those alighting.
Collectively, the delay due to the door opening and closing and delay due passenger
boarding and alighting are referred to in the literature as ‘bus dwell time’. The bus
dwell time at a bus stop can be influenced by four main elements – number of
boarding and lighting passengers, fare collection system, vehicle characteristics
such as number of doors and floorplan, and on board crowd levels.
Bus dwell time has been well established in the literature as a significant factor
causing bus bunching and thereby the reliability of transit service (Rajbhandari et.al.,
2003). Maloney and Boyle (1999) observed that the dwell times at stops on surface
roadways constitute about 7 percent of total time for a bus in service when the bus is
running along with the general traffic. The study also pointed out that the time
manoeuvring out of and into traffic, i.e clearance time, constituted around 7 percent.
In another study, Levinson (1983) found that at a CBD stop the dwell time range
from 20 to 60 seconds and for a non – CBD stop the dwell time ranged between 10
to 15 seconds.
and alighting passengers plus any time required for door opening and closing
operation. At a given stop / station, the dwell time is directly related to passenger
boarding and alighting, fare payment method, vehicle type and size, and in vehicle
circulation. The studies considering these variables for dwell time estimation are
described in the following sub sections.
According to Levinson (1983) dwell time for any bus is directly proportional to the
number of passengers it serves. As one of the earliest studies toward the
understanding of effect of passenger boardings and alightings on the bus dwell time,
Levinson’s study found that each boarding or alighting passenger contributes
between 2.6 to 3.0 seconds towards the bus dwell time (Equation 2.2). For a stop
having predominant number of alighting passengers, the study found that each
passenger adds 1.2 to 1.7 seconds towards dwell time (Equation 2.3).
Where,
N = Number of boarding and alighting passengers
Guenthner and Shina (1983) found that the number of boarding and alighting
passengers can be best described using a negative binominal function. Their study
highlighted that, even though the total dwell time increases, the service time per
passenger deceases as the number of passengers at a stop increases. A logarithmic
model was developed to estimate the dwell time per passenger (Equation 2.4). This
equation yields a maximum bus dwell time when there are 24 passengers. The dwell
time per passenger at this point is 1.2s per passenger. The authors, therefore,
suggested a dichotomised relationship to estimate dwell time based on number of
passengers, according to Equation 2.4 and Equation 2.5.
Where,
N = Number of boarding and alighting passengers
Following these single variable dwell time models, researchers turned to multi-
variable models to estimate dwell times. The modification was made to improve the
accuracy of dwell time estimation. The modified approach considered the number of
alighting passengers and number of boarding passengers as two independent
variables. Vuchic (2005) noted that the dwell time for a bus, where boarding and
alighting take place via different doors, is the maximum of the boarding time and
alighting time, plus a constant time annotated t0 to reflect station standing,
comprising of lost time at the station due to the opening and closing of door, plus
clearance time. Mathematically,
, Equation 2.6
A modified equation for a system where boarding and alighting from all doors is
permitted was also suggested by Vuchic (2005). The equation naturally pertains to
the busiest door.
Equation 2.7
Where,
An identical equation was suggested by the Transit Capacity and Quality of Service
Manual in 1999 (TCRP, 1997) and 2003 (TRB, 2003) (Equation 2.8).
t d = Pa t a + Pb t b + t oc Equation 2.8
Where,
These multivariate dwell time models imply that alighting and boarding occur in
series, and account only for those alighting through the busiest door, normally
presumed to be the front door, assuming that it is the only door available for
boarding. Any passengers alighting through the rear door are neglected in the
standard model, as their activity occurs in parallel to the front door activity, which is
implied to be time-critical.
To determine the probability of a passenger choosing the front door of the bus to
alight, Zhao and Li (2005) suggested a door choice model for alighting passengers
using the data collected from the Broward County Transit (BCT) in Florida, USA. A
utility based binary choice model was proposed to obtain the probability of choosing
the front door by an alighting passenger. The utility function U was defined as
follows.
Where,
P(Y = 1) = Probability of passenger choosing front door to alight bus.
TOTALOFF = Total number of alighting passengers at a given stop.
ONBOARD = Total passengers onboard before bus doors were opened at a
given stop.
TIMEPOINT = Dummy variable for a given stop 1 for time point and 0
otherwise.
AM = Dummy variable, 1 for observation during AM and 0 otherwise.
PM = Dummy variable, 1 for observation during PM and 0 otherwise.
The logit model for the utility function was approximated as equation 2.10.
Subsequently, the probability of an alighting passenger using the front door can be
obtained from equation 2.11.
eU
P(Y = 1) = Equation 2.11
1 + eU
The complexity in fare collection system lead to increase in the boarding time per
passenger as the number of boarding increases at a stop (Guenthner and Hamat,
1988) as opposed to the decrease as expected (Guenthner and Sinha, 1983). Later,
it was noted that the accuracy of the dwell time model can be improved if variable for
fare medium is also considered (Marshall et.al., 1990). An exponential equation was
developed for dwell time calculation considering fare collection medium and bus
induced delay (Equation 2.13). An average service time per passenger was obtained
as 8 seconds (approximately) under complex fare structure.
6.65 .
exp 0.39 0.20 (R2 = 0.71) Equation 2.13
Where,
= Average dwell times (s)
= Total number of boarding and Alighting passengers. (p)
= Fare collection medium
= Bus induced delay
In a recent study, Milkovits (2008) showed that with the 100 percent use of a smart
card medium the bus dwell time could possibly be to reduce by 22.8 percent. It was
also found that the advantage of smart (card) media over magnetic strip card can be
1.5 to 2s.
Fare collection systems influence service time because some media of fare
collection require more transaction time than others. The fare collection method
mainly affects the boarding time per passenger (Guenthner and Hamat, 1988) but in
some cases like smartcard, it could affect the alighting time as well. In the absence
of any field data the Transit Capacity and Quality of Service manual (TRB, 2003)
suggests boarding service time per passenger between 2.25s with no on–board
ticket purchase, and 4.3s when exact change is used for ticket purchase.
Low floor buses have an advantage of reduced dwell time due to reduced boarding
and alighting time per passenger. This is because all types of passengers find it
easier and quicker to board and alight a low floor bus compared to a conventional
bus having raised floor or steps. This is especially true for elderly people, parents
with baby strollers, small children and disabled persons. A saving of nearly 2
seconds per elderly passenger and around 6 seconds for a parent with child was
observed in a study undertaken at St. Albert Transit, US (Liggett, 1992). Similar
observations were made by Levine and Torng (1994). The data from Ann Arbor
Transportation Authority, US was used where the conversional buses and low floor
buses were providing transit service. The study showed a saving of 0.5s per
nondisabled passenger for low floor buses and hold good for both boarding and
alighting times. Later TCQSM (TRB, 2003) also suggested a reduction of 0.5s in per
passenger boarding time if the bus is low floor.
Standees on-board a bus reduce the speed of passenger boarding and alighting.
These standees block the aisles resulting in difficulties for boarding passenger
movement especially if the standees are standing near the front door. On the other
hand if the standees are standing near the rear exit door, the alighting passenger
may use the front door which in turn increases the boarding time. Zografos and
Levinson (1986) found that the passenger service time increases when the bus
operates beyond its seating capacity. Based on the field study the authors
investigated the bus dwell time and its variations with the size of boarding passenger
group and the number of passengers already on the bus. A linear relationship was
established between passenger service time, each boarding passenger’s rank and
number of passengers on board.
Where,
= Service time per boarding passenger (s)
= Rank of boarding passenger
= Number of passengers on board
Most of the studies in the field of bus dwell time have been based on data from
simple bus stops. These studies have been applied to the enhanced bus stop since
these bus stops are basically an improved version of the simple bus stop. However,
busway stations are different than bus stops, especially in terms of the number of
loading areas and therefore the length of station platform. Provision of simultaneous
stopping of buses at a busway station also changes the dynamics of the passenger
service process. At present there appears to be little dedicated study of bus dwell
times at these busway stations.
and Schijns, 2002). Brisbane, Australia, busway stations design is an open platform
type, with large awnings to provide passenger shelter against sun and rain.
Passenger transfer between platforms can be controlled and is only permitted via
overbridge by stair and elevator or lift.
Busway stations are also crucial from the bus operation standpoint. As mentioned
earlier, stations are the only sections in a busway where buses can overtake
stopping buses. Equally, a busway station can easily interrupt the smooth flow of
buses, especially during the peak period when bus queuing often happens at station
entry.
wait for relatively longer for their bus. These waiting passengers lead to the
formation of crowding with pedestrian flow disequilibrium.
The ‘crowd effect’ and unpredictable waiting time for the desired bus can alter the
characteristics of pedestrian flow within the platform area. Although there is little
doubt that the presence of many people in a given platform space causes problems
in terms of reduced walking speed and manoeuvring capacity, it also blocks the
pedestrians’ line of eye sight making the identification of incoming buses difficult.
This increases the reaction time of the pedestrian. Overall it increases the bus dwell
time and delay time.
Passengers are required to take certain decisions and steps after entering the
platform area and before entering the door of their desired vehicle. Passengers in
fact make decisions at three levels, observed Hoogendoorn and Bovy (2004). These
three levels, in the contexts of busways stations, are described below.
Decisions at strategic level: At this level the passenger decides on the activities to
be done after entering the station and the order of their execution. At busway
stations such activities are often limited to seeking information from the information
board and electronic information display unit.
Decision at tactical level: At this level the decision on how to perform the activities
that were decided at the strategic level is taken. Such decisions are basically
dependent upon the prevailing conditions faced by the passenger upon entering the
station platform area. A passenger will not seek information from the information
area should the desired bus being serving the platform. Therefore, it is possible that
a new activity could be decided and some or all activities decided at strategic level
can be discarded at this level. The most important decision at this level is
identification of the probable waiting area and making an assessment upon the most
suitable place. The choice for waiting area depends on the amount and location of
crowding present within the platform area.
Decision at operational level: At this level the pedestrian executes the options
decided at the tactical level. The decisions are mostly related to their walking
behaviour, and route choice to reach at a pre-determined point.
Impact on passengers
Firstly, crowding at a platform can make passengers uncomfortable, because
reduced space availability could induce unwanted behavioural changes in them (Hui
and Bateson, 1991) and could also reduce the individual productivity. Fear of
missing the bus or being stranded at the platform may stress certain passengers.
Secondly, reduced air quality at the station because of vehicular pollution could have
lasting effects (Chertok et.al., 2004; Chan et.al., 2002). On the whole, such
conditions make the journey difficult and unpleasant which may lead to passengers
having a lower perception of public transport.
Though not much literature dealing with the effect of passenger crowding at the
platform is available, a study carried out by Cox et.al. (2006) identifies crowding as a
problem across the British rail network. They emphasised that crowding should be
accepted as a possible threat both to the health of passengers and the transit (rail)
industry. Undoubtedly, the aim of providing stations is to facilitate a comfortable
interface from being a pedestrian to a passenger under the given pedestrian density.
However, there is seen to be breakdown in the smooth interface when there is
crowding at platform. The TCQSM (TRB, 2003) suggested that the restricted and
uncomfortable movement caused by the crowd can be the reason for such
breakdown.
Layover time is the time built into the schedule between arrivals and departures,
used to recover delays and preparations for the return trip. It also includes the driver
rest time. However, when the accumulated delay exceeds the minimum amount of
layover time (the minimum layover time should not be less than the driver rest time),
the service provider is forced to put their reserved vehicle and operator into service
to maintain the schedule.
Tc
F= Equation 2.15
Ah
Where,
= fleet size (vehicles)
= cycle time in minutes
= availability of vehicles
= service headway in minutes
The cycle time in minutes Tc is defined as the time required by a transit vehicle (bus)
for a return journey plus the two layover periods. Keeping travel speed of the bus
constant, which in the case of a busway is quite possible, the journey time increases
and the availability of the bus decreases if the delay at the station(s) increases. This,
according to Equation 2.15, would lead to the requirement of a larger fleet size to
maintain the service. From the economical prospective this is not desirable to the
service provider.
Pedestrian walking behaviour can be divided into two groups: unconstrained and
constrained flow (Antonini, 2005). The unconstrained walking is characterised by
behaviours independent of the presence of other pedestrians. Passengers under
unconstrained walking tend to walk toward their destination, keeping the original
direction and decide free flow acceleration and deceleration to maintain the desired
speed. On the other hand, in a constrained situation pedestrian walking behaviour is
influenced by the interaction with other pedestrians. In the constrained flow situation
pedestrians have a strong tendency to follow the leader and avoid collision with other
people. This pedestrian walking behaviour framework is reproduced in Figure 2.7.
Unconstrained Constrained
q = kv Equation 2.16
The density is measured in passengers per unit area and hence its numerical value
can be a fraction. To avoid the uses of number of pedestrians as a fraction some
researchers use reciprocal of density, space per passenger, M. In such case
equation 2.16 becomes
v
q= Equation 2.17
M
Where,
q = Pedestrian flow (p/m/min)
v = Speed (m/min)
k = Density (p/m2)
M = Space (m2/p)
The relationships between pedestrian speed, density and flow, which are very much
similar to the three fundamental diagrams of vehicle flow, are shown in Figure 2.8.
Similar to the this speed vs. density curve, Kholshevnikov et.al. (2007) presented a
graph between pedestrian speed and pedestrian density, as shown in Figure 2.9,
after reviewing the different studies for pedestrian egress under various situations,
like building, underground station and experimental setup. The figure demonstrates
that pedestrian walking speed is inversely proportional to density. However it is not
just the crowd density which influences pedestrian walking speed, pedestrian age
also does (Ando et.al., 1988). Work presented by Ando (1988) showed pedestrian
walking speed as a function of age under the free flow condition (Figure 2.10).
A
Speed, v
Speed, v
Density, k A/B Flow, q A2/4B
(a) (b)
A2/4B
Flow, q
A/2B
Density, k
(c)
Source: Khisty, 1990
Figure 2.8: Theoretical model of pedestrian flow in single channels; (a) Speed versus Density;
(b) Volume versus Density; and (c) Speed versus Volume.
Walking Speed, V (m/min)
D (person /m2)
Source: Kholshevnikov et.al, 2007
Figure 2.9: Empirical relations between travel and density of pedestrian (door opening).
Buildings: different 1; retail buildings 2,3,4; Sport structure 5; Underground station 6,7,8,9;
Experimental 10,11,12,13,14.
Age
Source: Smith, 1995
Figure 2.10: Walking speed variations as a function of age
The Level of Service (LOS) of transit facilities was divided into six distinct categories,
varying from A to F, based on the pedestrian density. Figure 2.12 illustrates the
characteristics of each LOS and Table 2.1 provides the corresponding threshold
values. LOS A suggests that the pedestrians at a transit facility, such as on
walkways, can freely select their walking speed and would not interfere with each
other. Whereas LOS F represents the situation where walking manoeuvre takes
place in a restrictive environment with frequent and unavoidable contact with others.
LOS A
Walking speeds freely selected; conflict with other
pedestrian unlikely.
LOS B
Walking speeds freely selected; pedestrians respond to
presence of others.
LOS C
Walking speeds freely selected; passing is possible in
unidirectional streams; minor conflict for reverse or
cross movement.
LOS D
Freedom to select walking speed and pass other is
restricted; high probability of conflict for reverse or cross
movements.
LOS E
Walking speed and passing ability are restricted for all
pedestrians; forward moving is possible only by
shuffling; reverse or cross movements are possible only
with extreme difficulty; volume approach limit of walking
capacity.
LOS F
Walking speeds are severely restricted; frequent,
unavoidable contact with others; reverse or cross
movement are virtually impossible; flow is sporadic and
unstable.
Source: TRB, 2003
Figure 2.12: Illustration of walkway level of service
The above stated level of service(s) is applicable only for walkways. The station
platform does not have exactly the same characteristics as walkways, though
walking activity takes place there. The station platform is an area where passengers
walk to and from the bus (transit vehicle) and where passengers wait (in a group) in
anticipation of their desired bus. In walkways, all pedestrians will be in motion,
whereas, at the platform a large number of passengers will be standing. This
requires commencement of walking from being stood and then forcing their way to
the bus door amongst the other standing passengers, when the passenger sees their
desired bus arriving at a loading area. TCQSM provides a different set of criteria for
determining the level of service for platform (queuing) areas. Figure 2.13 lists the
characteristics of each level of service and Table 2.2 presents the corresponding
threshold values.
LOS A
Standing and free circulation through the queuing area possible
without disturbing other within the queue.
LOS B
Standing and partially restricted circulation to avoid disturbing
others within the queue is possible.
LOS C
Standing and restricted circulation through the queuing area by
disturbing others is possible; this density is within the range of
personal comfort.
LOS D
Standing without touching other is impossible; circulation is
severely restricted within the queue and forward movement is only
possible as group; long-term waiting at this density is discomforting
LOS E
Standing in physical contact with other is unavoidable; circulation
within the queue is not possible; queuing at this density can only be
sustained for a short period without serious discomfort.
LOS F
Virtually all persons within the queue are standing in direct physical
contact with other; this density is extremely discomforting; no
movement is possible within the queue; the potential for pushing
and panic exists.
Source: TRB, 2003
Figure 2.13: Illustration of queuing area level of service
While comparing the LOS for walkway and queuing area, it was found that the
walkway LOS used average pedestrian speed as one of the deciding parameter,
while the queuing LOS considered the average inter-person spacing as the deciding
factor. Both the LOS criteria used pedestrian area (space available per pedestrian,
m2/p) as the primary parameter in deciding the LOS; however the threshold values of
pedestrian space for a walkway are higher than the respective values for a queuing
area because walking requires more space to gait.
It has been known for some time that BRT station bus capacity constricts line
capacity (TRB, 2003, Vuchic 2005). Moreover, as the American Transit Capacity and
Quality of Service Manual (TCQSM) notes, BRT station bus capacity is in turn
governed by capacities and efficiencies of individual loading areas (for more
common station configuration of linear loading areas placed in series). The bus
capacity of individual loading areas, in turn, depends on the bus dwell times at these
loading areas. Figure 2.14 provides the overview of calculation for station bus
capacity using current methodology as suggested by TCQSM (TRB, 2003).
Station capacity
Based on these similarities it was considered that for a busway the same formula
can be used for calculation of bus lane capacity. The capacity formula of bus lane
with no overtaking facility provided by Highway Capacity Manual (HCM, 1994) is
given in Equation 2.18.
Where,
3600 ⁄
Equation 2.19
⁄
Where,
Table 2.4 provide the value of ‘z’ corresponding to desired failure fate.
Failure rate z
1.0% 2.330
2.5% 1.960
5.0% 1.645
7.5% 1.440
10.0% 1.280
15.0% 1.040
20.0% 0.840
25.0% 0.675
30.0% 0.525
50.0% 0.000
Source: TRB, 2003
This revised equation was subsequently adopted by the TCQSM (TRB, 2003).
Based on this equation the TCQSM, provided a methodology for the capacity
estimation for a Busway station. This methodology involves uses of bus dwell time
model to first estimate the capacity of each individual loading area and then treats
these capacities with efficiencies of respective loading area to obtain final station
platform bus capacity.
Part 4 of the TCQSM (TRB, 2003), which deals with the bus transit capacity, defined
the dwell time as the average amount of time a bus is stopped at the station platform
to serve passenger movement, including the time required to open and close the
door. The manual suggests three methods for determining the dwell time. One of the
methods is to calculate the dwell time using the mathematical formula as given in
Equation 2.8. The other two methods are field measurement and use default values
respectively. This equation is identical to Vuchic’s equation for dwell time (Equation
2.7). The capacity of individual loading area can be determined using Equation 2.19.
For a platform configuration where more than one Loading Area (LA) is provided, as
in the cases of a busway station platform, the platform bus capacity can be equal to
the summation of the capacities of all the loading areas (Equation 2.20).
Equation 2.20
Equation 2.21
Equation 2.20 provides the theoretical capacity of the platform area. However, where
more than one loading areas are provided, the effective loading area would always
be less than the total integer value of such areas, because of the variation in dwell
time amongst loading areas and the possibility of blocking of front loading area by
preceding loading area/s. Additionally, the effectiveness also depends on the type of
loading area, off-line or on-line (Figure 2.15). The off-line loading area allows buses
to pull out of the line to get on to the loading area, providing overtaking opportunity to
the buses which do not have to stop at the station. It therefore reduces the station
impact on the travel time of certain buses like express bus which does not have the
schedule stop at the station but just have to pass it. On the contrary, at the on-line
loading area, the flow of the traffic gets blocked by the bus serving to the
passengers. Due to this, the number of effective loading areas for an on-line facility
is further reduced. Generally, busway systems have off-line loading facility. The on-
line loading facility can be seen predominantly in the CBD area of the city due to the
scarcity of the road space. For the busway situation where the loading areas are
typically off-line, Table 2.5 provides the variation of efficiency and effective loading
area with increase in the number of physical loading areas. The platform bus
capacity, therefore, is the product of loading area bus capacity and total effective
loading areas of the platform (Equation 2.21)
Table 2.5: Efficiency of multiple offline linear loading area at bus stops
The off-line loading area efficiency factors (Table 2.5) are based on the experience
at the Port Authority of New York and New Jersey’s Midtown Bus terminal (TRB,
2003) published in NCHRP Report 155 (Levinson et.al, 1975).
simple bus stops and bus lanes and is not based on the operation of busway station.
For instance, the present dwell time equation can describe delay caused to a bus at
a simple bus stop but could be less accurate for a busway station where multiple
loading areas are available. The loading area bus capacity relies on the bus dwell
time model, which does not consider appropriate multiple loading operation as well
as influence of platform crowding.
To accumulate the effects of presence of multiple loading areas, the TCQSM (TRB,
2003 based on Levinson et al., 1975) suggested the use of loading area efficiency
factors. However, the suggested efficiency factors are based on the operation of a
bus terminal facility and not based on the operation of busway station facility. The
busway station operation is described in chapter 3.
Based on these finding, Figure 2.16 shows gaps in the present methodology.
Station capacity
Gap in knowledge
Figure 2.16: Gap in busway station bus capacity estimation approach
Specific findings, in relation to busway bus capacity methodology, for this review
include:
The weak link in present methodology is the bus dwell time model, due to its
insensitivity towards passenger walking, platform crowding and multiple
loading areas.
The current dwell time models account for passengers only to the point of
their boarding and alighting times. In fact, the boarding and alighting service
themselves have some gap for local conditions.
The efficiency factors based on terminal operation may not reflect the busway
station operation.
Though TCQSM (TRB, 2003) provides the criteria for level of service for
queuing area, it does not relate the transit vehicle dwell time with a level of
service measure.
Though many models for pedestrian flow study were found in the literature,
these models are mainly stand-alone models and are not integrated with
vehicle/ dwell time models.
Based on these finding the research problem was developed, which is discussed in
the next chapter.
Research Problem Development
3.1 Overview
Previous chapter identified the lack of a purpose built methodology for busway dwell
time and capacity estimation. The shortcomings of using the traditional bus stop
dwell time methodology for busway station operation were described. Therefore it is
necessary to first study the operation of a bus stop and a busway station and identify
the differences between them. To start with, preliminary visual observations were
carried out at some of Brisbane’s busway stations including Cultural Centre, South
Bank, Mater Hill and Eight Mile Plains. Similarly, two arterial road bus stops in
Brisbane, at Coorparoo West and Mt. Gravatt Showground, were observed.
The aim of this exercise is to develop the research problem from the gaps
established in Chapter 2 and identify the parameters needed to model bus dwell time
at a busway station. This chapter describes the analysis of preliminary observations.
Section 3.2 presents the difference established here between busway station and
bus stop vis-à-vis passenger boarding process and operation. The research problem
is conceptualised in Section 3.3. Section 3.4 describes the busway operation at four
levels: platform, vehicle, station, and line. The definition of terms identified in this
research is provided in section 3.5 and section 3.6 concludes the chapter.
The first and second phase of the boarding process are considered here to comprise
the ‘passenger – bus interface’ stage. The passenger – bus interface starts when
the passenger first sees the desired bus and hails the bus operator and/or starts
walking towards a point by anticipating its stopping location. Similarly, the bus driver,
after seeing the hailing passenger, prepares to stop the bus at the closest available
loading area to the “lead stop” loading area. During this course of action, both the
bus operator and the passenger act independently but anticipate each other. The
third and fourth phases of the boarding process are considered here to comprise the
‘passenger – bus interaction’ stage. However, when there are only one or two
passengers boarding the bus, queuing may not occur and the passenger – bus
interaction stage may consist of only boarding.
On the contrary, at a simple bus stop since the passengers often align themselves
before the bus arrival, as discussed in section 3.2.1, the boarding process essentially
consists of initial hailing, queuing at entry door and boarding with walking component
been minimal or negligible. With respect to these actions, Table 3.1summaries the
difference between busway station and bus stop boarding processes.
Where,
O Origin
D Destination
D i Information signage
W Trip route
W Chosen waiting point
i O
Figure 3.2: Origin and destination of a trip segment at platform
Often after entering the platform passengers require to wait for their desired buses to
arrive at the station. These waiting passengers, hence, results into crowding and
affect others walking manoeuvres. Accordingly, the ease in completing walking
segment between waiting (point W) and the bus entry door (point D) depends on the
prevailing passenger density at the platform.
As the passenger density increases it is intuitive to expect that the walking speed
decreases, and therefore the walking time increases. Therefore the time spent by the
passenger to walk from their waiting (point W) to the bus entry door (point D) is likely
to influence the passenger service time for the bus. Higher crowding not only
reduces passengers’ manoeuvrability but also hinders their line of sight to
approaching buses. This increases their reaction time (hailing) to the bus arrival. On
the whole, the delay in walking and the time spent in walking, and therefore, the
dwell times, may vary in a manner somewhat proportional to the number of persons
in the crowd who are encountered by the passenger on the way to the bus door.
Bus Clearance /
STATION Bus to bus interface Bus queuing
Passengers walking
PLATFORM Passenger – Bus interface on platform
At the platform level: The busway station platform has multiple loading areas,
which necessitates passengers to walk to the bus entry door. As explained before,
the requirement of walking induces an interface of passenger and bus where the
passenger walks with a sense of uncertainty about possible stopping point of bus.
The time required in completing this task could result into Lost Time (LT) for a bus,
which ultimately affects its dwell time. The passenger – bus interface ends when the
passenger enters into the queue at the bus entry door or boards the bus, whichever
occurs first. The part of interface which occurs after the bus has stopped at the
loading area and doors are opened results in bus lost time. Figure 3.4 shows a
passenger – bus interface state of a passenger with the desired bus arriving at
loading area 3 (shown in solid circle). Figure 3.5 shows a bus accumulating lost time
because of a passenger – bus interface (shown in dashed circle).
At the vehicle level: Lost time, as the term suggests, forces the bus to occupy the
loading area on the platform waiting for the arrival of its first passenger. Hence the
bus accrues the dwell time in its travel time but without any processing of boarding
service. Therefore the Lost Time (LT) is considered here to be incremental to bus
dwell time. The bus dwell time has previously been considered as the summation of
boarding and alighting times for passengers plus door opening and closing time
(TRB, 2003). Therefore the amount of time a bus spends at the busway platform
could be better represented as –
Equation 3.1
Where,
= Bus dwell time
= Bus lost time
= Passenger processing time
= Door opening and closing time
It is worth noting again that the per passenger processing time i.e. boarding and
alighting time per passenger is sensitive to the fare collection system and policy
(TRB, 2003; Milkovits, 2008).
At the line level: Queuing of buses at the station entry makes platform servicing
simultaneous rather than random. When multiple platform loading areas are used by
buses simultaneously with insufficient clearance space, the buses tend to leave the
station in bunches. Hence they will tend to arrive at downstream stations in bunches.
This process is therefore a compounding one, and here is referred to as station –
station interface. It may be further compounded by any signalised intersections either
to control conflicting bus movements at access intersections, as is the case
immediately outwards of Mater Hill station, or to separate conflicting bus and general
traffic movements on at-grade signalised intersections, as is the case on the South
East Busway two stations inwards of Mater Hill station (and therefore upstream of
the platform under study). Intuitively, bus bunching would tend to improve the
efficiency of loading area utilisation at a busway station, hence increasing bus
capacity. Data in the Transit Capacity and Quality of Service Manual (TRB, 2003)
supports this. However, observations from this study suggest that when buses arrive
together on the station platform, the passenger – bus interface is amplified, mainly
because the passengers have less certainty about the bus arrival sequence and the
stopping location of their desired bus. The increased uncertainty can delay
passengers’ reaction to their desired bus. This may lead to an increase in the lost
time component of dwell time, and hence a reduction in bus capacity.
Passenger – bus interface: The ‘passenger – bus interface’ (IF) is the phase where
the first passenger and the bus driver are involved in a state when they both interact
but perform their respective activities independently.
Bus lost time: The ‘lost time’ (LT) is defined as the time lapse between bus stopping
time and the time first passenger puts his/ her foot at the bus floor.
3.6 Conclusions
Discovery of the passenger walking component, particularly walking the distance
between the passenger’s original waiting point and bus entry door, has enhanced
knowledge of bus dwell time at a busway station. The literature on pedestrian flow
characteristics clearly established that walking speed reduces under high density
situations. Therefore the hypothesis of this research is –
“Passenger walking and the prevailing crowd at a busway station platform influence
the bus dwell times”.
After establishing the hypothesis the next step was to collect that data at busway
station to test the research hypothesis and model the busway station bus dwell time.
The data related to bus dwell time as well as pedestrian flow behaviour, pedestrian
walking, and pedestrian crowd behaviour were collected at the study busway station.
The details of study station selection and data collection methodology are described
in chapter 4.
Data Collection and Processing
4.1 Overview
This chapter presents the data collection methodology used in this research to
capture the effect of bus lost time on bus dwell time. The chapter starts with a
discussion on different data collection techniques used in past research studies. This
was followed by section 4.3 which provides details of data collection technique used
in this research. Section 4.4 outlines the various methodologies applied for the
purpose of this research. The selection criteria of study station and characteristics of
selected busway station are discussed in section 4.5 and section 4.6 respectively.
Section 4.7 details data collection sequence and section 4.8 discusses about
processing of the collected data.
Manual count: Manual counting and field studies are the most common approaches
to collect data related to passenger boarding and alighting from transit vehicles
(Guenthner and Sinha, 1983; Levinson, 1983; Kim, 2004). However, costs involved
in manual collection of data, as observed on Milkovits (2008) limits the number of
observations to a handful of stops, operators, and times of day. Moreover, manual
data collection techniques are laborious and may generate errors (Dueker et.al.,
2004).
Automatic counter: An automated data collection system can provide rich data sets
across time of day, stop, route, operator and other significant but rare events such as
lift operation (Dueker et.al., 2004; Rajbandari et.al., 2003). However this technique
can onle be used where transit vehicles are fitted with appropriate sensors such as
automatic passenger counter (APC), automatic fare counter (AFC) and automatic
vehicle locater (AVL). Maikovits (2008) used automatic passenger counter system
installed on Chicago Transit Authority buses to count boardings and alightings to
study bus dwell times.
Video recording: Past literature shows that vision based techniques such as video
recording is the preferred method of data collection for pedestrian crowd studies. In
this technique a CCTV camera is used to record the events that are later analysed in
the laboratory. Often analysis of video footage was carried out by image processing
algorithms (Davies et.al., 1995) to extract the desired variables. Kilambi et.al. (2008)
presented two methods, one based on heuristic learned during training and another
based on shape models, to estimate the crowd size. Johansson et.al. (2008) used
the video recording technique to study stop-and-go wave and crowd turbulence
phenomena at a high density location.
fixed video cameras and later analysed. Seyfried et.al. (2007) used the experimental
study and video recording to test the validity of pedestrian specify flow concept with
respect to bottleneck situation.
Because of the large numbers of passengers and bus flows at the subject station
platform during the study times, the data collection method needed to be selected
with care. On site manual counting can prove to be very laborious and may be
susceptible to high human error. Literature search on data collection techniques
(Table 4.1) established that video recording technique was most suitable for this kind
of study. Video recording at the station followed by laboratory counting can eliminate
much of the human and machine errors.
The study busway station platform has two CCTV cameras which record platform
activities on a continuous 24 hr basis. These cameras were used for this data
collection effort with the permission and help of TransLink’s Busway Management
Centre (pers. comm. Mr Andrew Haddock and Mr Jurgen Pasiezny). Additionally, a
QUT camera was installed at the front end on the platform to enhance the quality of
data. Figure 4.1 shows the positions of three cameras used in data collection.
QUT Camera 3
CCTV 1 CCTV 2
Figure 4.1: Camera positions at Mater Hill Busway Station (Outbound platform)
queuing time, etc). For measuring bus attributes, the guidelines from the Transit
Capacity and Quality of Service Manual - Part 4: Bus Transit Capacity (TRB, 2003)
were followed –
1. Record the bus route number and its queue ingress time (if any)
2. Record its queue egress time (if any)
3. Record the loading area number on which it is serves passenger.
4. Record the time at which the bus comes to a complete halt.
5. Record the time of full opening of the bus front door.
6. Count the number of alighting passengers separately from the front and rear
door and number of boarding passengers onto the front door.
7. Record the timing for first and last passenger alighting.
8. Record the timing for first and last passenger boarding.
9. Record the time of full closing of the bus front door.
10. Record the time when bus starts leaving the bay.
Since analysis of the passenger – bus interface has not been done before; the steps
of recording the passenger side data were derived from the concept explained in
section 3.3. The following steps were involved in the measurement –
The South East busway in Brisbane, Australia is a16 km corridor with 11 dedicated
stations. Figure 4.3 shows the busway corridor and its stations. Each busway station
has one platform in each inbound (to city) and outbound (from city) direction. The
characteristics of the corridor are such that, during the morning peak, flow of
passengers toward the city is high, contributing to high numbers of passengers
alighting at the inner inbound platforms. This situation reversed during the afternoon
and evening peaks when there are more boarding passengers at the outbound
platforms of the inner stations. Therefore, during the evening peak period, crowding
of boarding passengers can be observed at outbound platforms of the inner busway
stations.
Woolloongabba Station
Buranda Station (3.17 km)
(4.37 km)
Greenslopes Station
(5.97 km)
Based on the above listed desired criteria, three busway stations qualified for this
research study. Table 4.2 provides the characteristics of the outbound platform of
these three candidate stations.
Photo
Platform
60m 60m 45m
length
Number of
bus
Four Three Three
loading
area
Weekday
5000* 2775* 3122*
boardings
Weekday
2215** 829** 785**
alightings
* Total number of boarding on a weekday Source: TransLink, 2007
** Total number of alighting on a weekday Photo by: Jaiswal (2009)
Among the candidate stations, the Cultural Centre Busway Station has highest
number of boardings; however, this station also acts as a terminal station for some
buses. Those buses terminating have dwell times due to alighting passengers only.
Since the aim of this research is to investigate the effect of platform crowding on bus
dwell time, and because a terminating bus will have less influence because of
platform crowd, Cultural Centre Busway Station was rejected. The second highest
boarding load was observed at Mater Hill Busway Station. Moreover, Mater Hill
busway station (outbound platform) also has a dominance of boarding passengers
on the platform. The outbound platform of Mater Hill Busway station was therefore
selected for this study.
In Brisbane, a passenger can board the bus from the front door only. However, an
alighting passenger can use either the front or rear door. During the annual analysis
periods over the three years of the study, four methods or combinations thereof were
available for passengers to validate their journey. These included two manual
means, being onboard ticket purchase from the bus operator or presentation of a
pre-paid paper ticket. The earlier of two automated means was the use of the 10 trip
saver ticket, a magnetic stripe card dipped on entry into one of two readers located
inside of the front door only. The later of two automated means was the use of the
GoCard smart card, with each bus equipped with four readers; two readers inside
the front door for touch-on and touch-off, and two readers inside the rear door for
touch-off only.
In 2009 TransLink introduced the pre-paid platform policy for the outbound platforms
of three innermost busway stations on the South East busway. Under this policy,
during the busy outbound mid afternoon to evening peak period, all passengers must
have a pre-paid ticket or a GoCard to enter the outbound platform of each of the
Cultural Centre, South Bank, and Mater Hill Busway stations, as no on board ticket
purchasing is permitted. The rationale being to minimise bus dwell times and
therefore improve bus capacity and reduce delays. The details of recordings and fare
policies on the recoding days are given in Table 4.3.
Station
Stationentry/
entry/ exist
exit point
point
Bus direction
LA Loading area
INBOUND PLATFORM
Lead Stop
LA3 LA2
Tunnel
Lead Stop
LA2 LA3
Tunnel
OUTBOUND PLATFORM
10 trip magnetic
10 trip magnetic
strip card into front
stripe card into front
door dip readers
door dip readers
(Phasing out)
Fare policy
GoCard smart card
with onboard touch GoCard smart card with
on using readers onboard touch on using
front door only & readers front door only
touch off using & touch off using
readers at front and readers at front and
rear doors rear doors
(Introduced)
Source: TransLink
Alighting Boarding
600
500
Number of passengers
400
300
200
100
0
9-10 am
10-11 am
11-12 am
12-1 pm
9-10 pm
10-11 pm
11-12 pm
5-6 am
6-7 am
7-8 am
8-9 am
1-2 pm
2-3 pm
3-4 pm
4-5 pm
5-6 pm
6-7 pm
7-8 pm
8-9 pm
Time of day
Source: TransLink, 2007
Figure 4.5: Boarding and alighting at inbound platform of Mater Hill Busway Station
Alighting Boarding
900
800
700
Number of passengers
600
500
400
300
200
100
0
5-6 am
6-7 am
7-8 am
8-9 am
10-11 am
11-12 am
1-2 pm
2-3 pm
3-4 pm
4-5 pm
5-6 pm
6-7 pm
7-8 pm
8-9 pm
10-11 pm
11-12 pm
9-10 am
12-1 pm
9-10 pm
Time of day
Source: TransLink, 2007
Figure 4.6: Boarding and alighting at outbound platform of Mater Hill Busway Station
It is noted that a greater, but still moderate proportion of buses were articulated
during the later years on the study.
140
120
100
Number of Buses
80
60
40
20
0
9-10 am
12-1 pm
9-10 pm
5-6 am
6-7 am
7-8 am
8-9 am
10-11 am
11-12 am
1-2 pm
2-3 pm
3-4 pm
4-5 pm
5-6 pm
6-7 pm
7-8 pm
8-9 pm
10-11 pm
11-12 pm
Time of day
Source: TransLink, 2007
Figure 4.7: Number of buses servicing outbound platform of Mater Hill Busway Station
Since it was not feasible to analyse the entire day’s operation, a matrix based
approach was developed to identify the time of day which could produce
homogenous data for analysis. To devolve this matrix, the data collected by
TransLink in November 2007 were examined (Figure 4.6 and Figure 4.7). The bus
arrival rate and passenger demand were divided in three groups - Low, Medium and
High. The criteria for division are given in Table 4.4.
Table 4.4: Bus flow rate and passenger demand classification split
The derived matrix is given in Figure 4.8. On the horizontal axis, the bus arrival rate
in ascending order (from left to right) and their corresponding hours ending are
plotted. On the vertical axis, the passenger demands for the study station platform
their corresponding hours ending are arranged in ascending order (from top to
bottom). Based on this matrix, three different hours of the day, out of total 18 hours
of day’s operation, were selected for analysis. Each selected hour represents
particular flow characteristics as shown in Table 4.5.
No data set was analysed from the period when passenger demand was low,
because of a lack of statistically significant data. In total the data available for
analysis was comprised of 9 hours (March 2007, March 2008 and April 2009). Note
that the morning peak period (8:00 am to 9:00 am) was not considered for analyses,
because it is negligible for boarding passengers. Instead, the evening off-peak
period (6:00 pm to 7:00 pm) was selected to have diverse set of data for analysis.
The selected hours for data analysis are shown in Figure 4.8 with thick borders.
Low
50 7 7
Observed passenger demand (p/hr)
58 22 22
Corresponding hour ending at
60 21 21
101 20 20
Classification split
105 10 10
136 19 19
138 11 11
Medium
152 13 13
Null set
159 14 14
165 12 12
171 8 8
180 9 9
200 15 15
255 18 18
High
For each bus servicing at the platform, its first boarding passenger was tracked from
the time they first reacted to the bus until the time they boarded that bus. A
passenger is considered as boarded when no part of his/ her body is out of the bus.
Furthermore, for each one hour time period, 50 passengers randomly selected from
the platform crowd were observed, from the time when they first reacted to their
desired bus until the time when they boarded their bus.
For this research outbound platform of Mater Hill busway station of Brisbane’s South
East Busway network was selected. The platform has very high amount of boarding
passenger load which leads to platform crowding. This situation is appropriate to
study effect of platform crowd and linear loading area on the bus dwell time and
station operation.
In the next chapter, the data was analysed to evaluate various parameters and
processes identified in chapter 3.
Parameter Analysis and Evaluation
5.1 Overview
This chapter evaluates the two vital parameters of busway station operation. These
two parameters were identified in chapter 3 as: passenger – bus interface, and bus
lost time. The aim of this chapter is to analyse these parameters based on platform
crowding. The outcome of this analysis will be used as inputs for modelling of bus
lost times and bus dwell times at a busway station.
Video recordings were analysed to determine how passenger – bus interface occurs
and to study how platform crowding influences it. For each bus, randomly selected
passengers were manually tracked from the time they first initiated walking after
seeing the desired bus (event 1) to the time when they entered the boarding queue
(event 2). Note that, if they are the first or only passenger, this is just the time when
they boarded the bus. The time stamps of both events were recorded. Therefore,
mathematically the duration of passenger – bus interface can be estimated as –
, ,
Equation 5.1
Where,
= Duration of passenger – bus interface for passenger, i
, = Time stamp of passenger ‘i' first initiated the walking.
, = Time stamp of passenger ‘i' joined the boarding queue at entry door.
The video recordings from March 2008 were used to analyse the passenger – bus
interface. Table 5.1 provides descriptive statistics of two off-peak periods examined
(10:00 am - 11:00 am and 7:00 pm – 8:00 pm). Table 5.2 provides the descriptive
statistics of peak period (3:00 pm - 3:40 pm). During the off peak period, the
observed number of passengers on the platform in a 15 min interval was between 10
and 27. During the peak period, the observed number of passengers on the platform
in a 15 min interval was between 40 and 67.
Note: Off – peak operation; passengers on platform < 30; total analysis time = 120 minutes
Mean 19 s 13 s 11s
^
Max 31s (51s) 25s 38s
Note: Peak operation; # of passenger on platform ≥ 30; total analysis time = 40 minutes
^ In this outlier a passenger was standing at the lead end of the platform and observed the desired
bus stopping at rear loading area (loading area 3). However, before he reached the desired bus at
loading area 3, loading areas 1 and 2 were cleared and the bus driver moved up to loading area 1
causing this passenger to walk back to loading area 1. This resulted in a very high passenger – bus
interface. Such extreme durations were less frequent at the study station and were therefore not
considered further.
Large variations in the duration of passenger – bus interface, both in off-peak period
and peak period, were observed at the platform across all three loading areas. The
passenger – bus interface shows increments in all statistical parameters when the
time period changes from off-peak to peak operation, with the exceptions of
minimum duration of passenger – bus interface for loading area 1 and loading 3.
It was also noted that, during the off-peak period, the minimum duration of
passenger – bus interface was identical for loading areas 1 and 3. However, while
the minimum duration of passenger – bus interface for loading area 1 decreased by
6s, the reduction for loading area 3 was only 1s. In the off-peak the loading area 1
has the higher value of maximum duration of passenger – bus interface compared to
loading area 3. This situation was reversed during the peak period where the
maximum value for duration of passenger – bus interface was observed for loading
area 3.
In order to study the spread of passenger – bus interface, the duration of passenger
– bus interface was averaged over prevailing platform crowding and plotted against
the platform crowd (Figure 5.1).
LA 1 LA 2 LA 3
40
Avg duration of passenger - bus interface (s)
25
20
15
10
0
0 10 20 30 40 50 60 70 80
Passengers on platfrom
The Figure 5.1 provides some vital information about the busway station operation
under different prevailing crowd levels. Firstly, a clear distinction between the low
crowd operation and the high crowd operation can be seen in the figure. The
observed duration passenger – bus interface (averaged) formed two well separated
clusters representing off-peak and peak operations.
a passenger’s walking speed decreases. The passenger can no longer use the
straight path between the waiting point and the bus entry door, but is forced to
undertake a zig-zag manoeuvre to complete the path. This zig-zag manoeuvring also
increases the distance and thereby increases the walking time.
This analysis (Figure 5.1), however, failed to establish any clear mathematical
relationship between passenger – bus interface duration and platform crowding.
Nevertheless, it clearly illustrated the unevenness of the passenger – bus interface
under the high crowd condition. The analysis suggests that the study busway station
platform operates in two different operational paradigms. During the low crowd and
coinciding low bus flow operation, the passenger – bus interface appears to be
relatively stable and varying with a standard deviation of 4s. On the other hand, the
standard deviation is doubled to 8s, during the high crowd and high bus flow
operation. The duration of passenger – bus interface is more unstable and less
predictable. This instability can cause high variability in bus dwell time. This finding
is important as it confirms that platform crowd affects bus dwell time.
The focus of analysis, therefore, moved to low crowd operation to examine the
correlation between duration of passenger – bus interface and platform crowd, if any.
The plot of average duration of passenger – bus interface and platform crowd is
given in Figure 5.2.
20
15
10
0
0 5 10 15 20 25 30
Passengers on platform
The result shows that duration of passenger – bus interface, generally, increases
weakly with an increase in platform crowding. However, the graph suggests that
there is no strong two dimensional relationship between duration of passenger – bus
interface and platform crowding. The R2 values for all three loading areas shows that
linear regression is not able to explain any correlation.
Walking speed
Figure 5.3: Passenger – bus interface duration and its dependent variables
The figure exemplifies that the passenger – bus interface depends on characteristics
of passenger walking, which in turn depending on the platform crowd and distance of
waiting point from loading area.
In the absence of any prior research on passenger – bus interface and its results on
bus dwell time, it is advantageous here to further explore the passenger – bus
interface duration and its influence on bus dwell time using graphical means.
represented on the X axis. The X – X axis, in space, represents the location of the
bus entry door on the loading area. P1, P2, … Pn are the position of the first, second
and nth passenger at time t0 when they first observe their desired bus, B. B’ is the
point in time when bus B reaches a stop at the loading area. P’1, P’2, … P’n, are the
points in time when the first, second and nth passenger reach the bus entry door. The
lines P1-P’1, P2-P’2, and Pn-P’n are the idealised trajectories of respective passengers.
The slope of these lines represents the passenger’s walking speed respectively.
Note that these trajectories may not be straight in practice due to the zig-zag path
discussed previously. The B-B’ line is the trajectory of bus and the slope of line
represents bus speed.
Note that the time lapse between B’ and P’1 is identified here as lost time for the bus,
where no real passenger service occurs. This research defines ‘lost time’ as the time
lapse between when the bus comes to rest on its loading area and the time of
boarding of the first passenger. Note that portions on the passenger – bus interface
duration of 2nd … nth passenger occurs during a period termed here as the
passenger – bus interaction period, IA. Therefore, in this case the dwell time for the
bus is equal to sum of the lost time (LT), passenger – bus interaction (IA) and the
door opening and closing time (toc). The variables in Figure 5.4 are described below:
DT
Space
LT IA
to ’ tc
X B ’
P 1P’2 P’n
X
t0 Time
P1
P2
Pn
B IF1
IF2
IFn
In the time – space diagram, the loading area position represented by X-X is static.
However, the passenger is not. The passenger may change the waiting position due
to the influence of the level of crowding. This would change the duration of
passenger – bus interface and subsequently the bus lost time. Hence, we analysed
the behaviour of waiting passengers to understand the character of the passenger –
bus interface.
dy2 dy3
dy1
Outbound platform
Front Rear
y Where, dy1 > dy2 ; dy1 ≈ dy3
Figure 5.5: Distance to loading areas from the waiting area on the busway platform (Off-peak)
As the crowd at the platform increases, additional passengers first spread toward the
left side of the shaded area (front area of platform) and later sprawl towards left of
the shaded area (rear area of platform). However, observations at the study station
found that passengers outside of the shaded area moved quickly into the shaded
area whenever a space was created. This behaviour is largely driven by their desire
to minimise the walking to their desired bus, which, due to the ‘lead stop’ operation
principle, has a higher probability of using either the loading area 1 or the loading
area 2 compared to loading area 3. For simplicity, the shaded area and its centroid
i.e. y-y line is used as the reference line.
The distance between the bus door on loading area 2 and the shaded area centroid
y-y is less than the distance between the bus door on loading area 1 and the shaded
area centroid (Figure 5.5). Therefore, in the best case scenario, the lost time for any
bus would be lowest when it uses the loading area 2.
From the reference line i.e. y-y line, the passenger walking to loading area 1 walks
nearly in the direction of approaching bus to loading area 1. In contrast, a passenger
walking to loading area 3 walks in the direction opposite of approaching bus to the
loading area 3. These directional differences between passenger and bus alter their
interface and subsequently the lost time for buses on these loading areas. Walking in
the opposite direction of the approaching bus reduces the overlapping component of
the passenger – bus interface, causing higher lost time. This is because the bus
needs to cover less distance to reach that loading area. In contrast, walking in the
direction of approaching bus slightly increases the overlapping component of the
passenger – bus interface, and could cause a slight reduction in lost time. This
phenomenon of passenger – bus interface is better explained in Figure 5.6, which
illustrates a case where the bus B would experience different lost time, due its first
If the bus ‘B’ uses loading area 2 then it will experience the lowest lost time because
the passenger is standing near to the loading area and hence will not be required to
walk large distance to reach the bus entry door. However, if the bus uses loading
area 3 then it will experience largest lost time because of the passenger is required
to walk large distance and the walking will be in the opposite direction of the bus
arrival. On the other hand, the lost time for the bus B at loading area 1 will be lager
that loading area 2, because of the walking distance increases requirement to walk,
but will be less than that of loading 3 because the passenger’s walking is in the
direction that of bus arrival. The variables in Figure 5.6 are described below:
LT1, LT2, LT3 = Bus lost time at loading area 1, 2, and 3 respectively.
P1 = Location of 1st passenger of the bus B.
dy1, dy2, dy3 = Passenger, P1, walking distance to loading area 1, 2, and 3
respectively.
y-y line = Centroid of shaded area (Ref: Figure 5.5)
LT1
LA1
dy1
dy2
P1 y - y line
LA2
LT2
dy3
LA3
Space
B LT3
Time
From the analysis presented in this section and in Section 5.3 we found that the
passenger – bus interface time influences the bus dwell time through a variable
called bus lost time (LT). The analysis illustrated how passenger bus interface
results in bus lost time (Figure 5.4) and explains how loading area is a determinant in
the lost time for a bus (Figure 5.6). In light of these findings, the bus lost time for
individual loading areas is analysed in following section.
For each bus, the time stamps of its stopping at its loading area (event B1), full
opening of its entry door (event B2) and its first passenger placing his/her foot on the
bus floor (event B3) were recorded from the video recordings. Quite often it was
observed that bus driver opens the entry door simultaneously while stopping at the
loading area, resulting in identical time stamps for event B1 and event B2. The bus
lost time, therefore, from Figure 5.4, can be estimated as -
, ,
Equation 5.2
Where,
= Bus lost time for bus, j.
= Time stamp of full opening of entry door for bus, j.
, = Time stamp of first passenger placing his/her foot on the floor the bus, j.
If any alighting occurred through the bus entry door during the bus lost time (LT), the
time consumed by the alighting passengers was deducted from the lost time. The
proportion of such observation was relatively very small throughout the analysis
period. This was mainly due to two reasons; first the study station platform had
predominantly boarding passengers during all study periods, and second the
majority of the alighting passengers used rear door of the bus. The number of
boarding and alighting passenger observed during an half hour afternoon peak
period (3:00 pm to 3:30 pm) at the study busway platform are given in Table 5.3.
Table 5.3: Passenger boarding and alighting during evening peak period
Number of
Number of Number of alighting
Month/ Year boarding
buses passengers
passengers
Front door Rear door
March 2007 51 348 5 16
March 2008 36 475 4 13
April 2009 62 250 11 42
The video recordings from March 2008 and April 2009 were used to analyse bus lost
time. Table 5.4 provides the descriptive statistics of off-peak periods (10:00 am -
11:00 am and 7:00 pm – 8:00 pm). Table 5.5 provides the descriptive statistics of
peak periods (3:00 pm - 3:40 pm for March 2008; 3:00 pm - 4:00 pm for April 2009).
During the off peak periods, the observed number of passengers on the platform in a
15 min interval was less than 30. During the peak periods, the observed number of
passengers on the platform in a 15 min interval was greater than 40.
Min 1s <1s 1s
Max 11 s 10s 16 s
Mean 4s 3s 5s
Count 51 30 14
Note: Off – peak operation; # of passenger on platform < 30; total analysis time = 120 minutes
Mean 4s 3s 5s
# count 61 50 38
Note: Peak operation; # of passenger on platform ≥ 30; total analysis time = 100 minutes
Note that minimum bus lost time, in both the off-peak and peal periods, across all
three loading areas were either 1s or less than 1s. This represents the cases where
the first passenger of the bus happened to be standing right where the bus stopped.
In such cases the first passenger boards the bus almost immediately as soon as it
comes to halt and door opened. Such situations are more common at simple bus
stops where passengers wait for their bus at the signpost for the stop.
Of note, the mean bus lost times for loading areas 1 and 2 were no different between
the off-peak and peak periods. However, the maximum bus lost time for loading area
1 increased in the peak period. On the other hand, loading area 2 experienced a
decrease in bus lost time during the peak period. Loading area 3 however,
experienced an increase in both mean and maximum bus lost times during the peak
periods.
To study the profile of bus lost time between off-peak and peak periods, the mean
bus lost time of each loading area was analysed against platform crowding. Figure
5.7 shows that they have dissimilar changes in their lost time as the crowd level at
the platform increases. For loading area 1 the mean lost time initially increases with
the increase in crowd level. However, with a further increase in crowd level the mean
lost time decreases. The initial increment in the lost time may be due to the crowd
having acted as an obstruction in the path to bus entry door for loading area 1. When
passengers move into the area left of the shaded area (Figure 5.5) this may result in
a decrease in bus lost time for loading area 1.
On the contradictory, the mean lost time for buses using loading area 2 decreases
initially but later increases back. This is reasonable because of the behaviour of
passengers in selecting the area adjacent to loading area 2 (shaded area in Figure
5.5). However, as more passengers tries to fit into this limited space they increases
the obstruction to walking. This results in lost time again increasing back under the
high crowd level.
Whereas, the relatively obstruction free path to loading area 3, during off-peak
period, results in mean lost time for buses on loading area 3 lesser than loading area
1 under the low crowd level condition. However, the mean lost time steadily
increases with platform crowd. This is reasonable for two reasons. First, the
obstructions are likely to increase with crowd level and second, the majority of
passengers must to walk in the direction opposite to the approaching bus.
Peak period onset
9
Off-peak period
Peak period
8
Mean lost time per Bus (s)
7
LA3
6
5
LA1
4
2 LA2
0
10 20 30 40 50 60 70
Number of passengers on platfrom
Figure 5.7: Variation in bus lost times over platform crowd by loading area
The results clearly indicate that bus lost time is not a constant value for the entire
platform. It varies from loading area to loading area and also varies with platform
crowd levels. These observations support the analysis presented in Section 0.
both interact and perform their respective activities based on each other’s position or
action. The passenger – bus interaction activities are passenger boarding and
passenger alighting. The boarding and alighting activities at a busway station
platform are similar to that at a simple bus stop. The factors influencing the
passenger boarding time and alighting times include fare collection method, bus
floorplan and onboard standees. In the literature, for a simple bus stop, the
passenger boardings and/ or alightings are considered as the part of bus dwell time.
Much literature has been found related to the passenger boarding and alighting at a
simple bus stop. A detail survey of literature on existing bus dwell time model was
presented in Section 2.4.1. For the purpose of this research, the simple bus stop
dwell time model suggest by TCQSM (TRB, 2003), was used (Equation 5.3).
t d = Pa t a + Pb t b + t oc Equation 5.3
Where,
This equation implies that alighting and boarding occur in series, and accounts only
for those alighting through the front door. Any passengers alighting through the rear
door are neglected in the standard model, as their activity occurs in parallel to the
front door activity, which is implied to be time-critical (Vuchic, 2005).
Equation 5.4
Where,
As discussed in the section 5.3.2, the passenger – bus interaction time is the time
lapse between first passenger boarding and last passenger boarding. A passenger
was considered boarded when no part of his/ her body is out of the bus. For each
bus, time stamps for its first passenger boarding (event PB,1) and last passenger
boarding (event PB,l) were recorded. Similarly, time stamps for its first passenger
alighting from front door (event PAF,1), from rear door (event PAR,1) and last
passenger alighting from front door (event PAF,l) and from rear door (event PAR,l)
were recorded. The number of boardings and alightings was also recorded.
The boarding time per passenger and alighting time passenger at the study station
was estimated as -
Equation 5.5
Equation 5.6
The video recordings from April 2009 were used to estimate the boarding and
alighting times per passenger. The marginal service time was estimated for each
loading area individually. The descriptive statistics of boarding time and alighting
time are provided in Table 5.6.
Loading Area 2
Boarding time per pax Front 3.0s 7.5s 5.0s 2.3s
Alighting time per pax Rear 3.5s 5.0s 4.0s 0.6s
Loading Area 3
Boarding time per pax Front 1.9s 8.0s 4.9s 4.8s
Alighting time per pax Rear 2.0s 6.0s 3.6s 1.4s
The video recordings from March 2007, March 2008 and April 2009 were used to
study the impact of fare collection system on boarding and alighting times per
passenger. The details of recordings and fare policies on the recoding days are
given in Table 5.7.
Number of
51 36 62
buses
Number of
boarding 348 475 250
passengers
Number of
alighting 21 17 53
passengers
The result highlighted that with removal of on board ticket purchasing, an increased
uniformity in service time per boarding passenger was observed among the three
loading areas on the study station platform. The boarding time was decreased by
approximately 15 percent for loading area 1 and loading area 2, and approximately
40 percent for loading area 3.
Table 5.8: Effect of fare collection policy on boarding time per passenger.
Table 5.9: Effect of fare collection policy on alighting time per passenger.
The highest increases in alighting time were observed for loading area 2 (110%)
followed by loading area 3 (71%) and loading area 1 (55%). Alighting times were
also found to be affected by the bus type; rigid bus and articulated bus. Passengers
alighting from an articulated bus at the study station showed a tendency towards
using the front door of the bus. As there were still few articulated buses used on the
study platform during this study, no statistically significant results could be produced.
5.6.1.3 Findings
This analysis has highlighted that the boarding time per passenger was reduced by a
minimum of 14 percent. This reduction occurred due to various fare collection policy
changes implemented over the analysis period. On the contrary, the alighting time
per passenger increased in excess of 50 percent. The average service time for each
boarding passenger was 4.7s and for each alighting passenger 3.6s. However, more
investigation is required to assess the impact of the smart card system on alighting
passengers, particularly, for a period when alighting passengers are predominant,
such as the morning peak on inbound platforms of inner urban (destination) stations,
for example, the case on the inbound platform at Mater Hill Busway station
In this chapter, the passenger waiting behaviour at the busway station platform was
explained and its influence of passenger – bus interface was studied. By studying
the complex nature of the passenger – bus interface and the relationship with bus
lost time, a sound theoretical framework for busway station bus dwell time was
developed.
In next chapter, the bus lost time model is elaborated. Chapter 7 integrates the bus
lost time variable with existing knowledge base to develop a more comprehensive
busway station bus dwell time model and busway station loading area bus capacity
model. Chapter 8 presents a methodology to estimate busway station loading area
efficiency factors. Chapter 9 propose a comprehensive busway station platform bus
capacity methodology.
Modelling Bus Lost Time
6.1 Overview
Having introduced in preceding chapters the concept of bus lost time and having
analysed it in chapter 5, this chapter discusses the development of stochastic
relationships from the observed bus lost time data set for the study busway station.
The purpose of developing a stochastic model was to generalise the pattern of bus
lost time variations. The generalised bus lost time can be applied to develop an
improved bus dwell time model and bus capacity model for the busway station.
Before estimating the stochastic model for the observed bus lost time it was
necessary to study its frequency density. This analysis was performed using a
histogram technique. Details of analysis are discussed in Section 6.2. Section 6.3
estimates the bus lost time probability distribution function and Section 6.4 fits the
probability distribution curve on observed bus lost time data. As found from the
analysis presented in Section 5.5, the bus lost times depended on the particular
loading area and platform crowd. Further separate probability curves were fitted to
each loading area for each of the peak and off-peak periods. Section 6.5 closes the
chapter and summaries the key contributions to knowledge.
14
12
10
8
Count
6
4
2
0
<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Bus lost time
a) Loading area 1
12
10
8
Count
0
< 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Bus lost time
b) Loading area 2
10
9
8
7
6
Count
5
4
3
2
1
0
<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Bus lost time
c) Loading area 3
Figure 6.1: Off-peak period bus lost time histogram
14
12
10
Count 8
6
4
2
0
<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Bus lost time
a) Loading area 1
20
18
16
14
12
10
Count
8
6
4
2
0
<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Bus lost time
b) Loading area 2
8
7
6
5
Count
4
3
2
1
0
<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Bus lost time
c) Loading area 3
Figure 6.2: Peak period bus lost time histogram
Figure 6.2 shows the bus lost time histogram for the peak period at the same
location. The values for bus lost time observed at the station platform range from
less than 1s to 17s. At loading area 1, the bus lost time had widened range
compared to the off-peak period, with the majority of buses experiencing lost time
between 1s and 7s. On the other hand, at loading area 2, most of the buses had
more concentrated lost times between 1s and 3s. Similar to loading area 1, the
majority of buses at loading area 3 experienced the lost time ranging from less than
1s to 7s.
The resulting histograms indicate that the bus lost time, both in off-peak and peak
period, is not normally distributed for any of the loading areas. Graphically we can
interpret that the histograms are positively skewed which suggests that lognormal
distribution or Weibull distribution could be fitted (Devore, 2008).
Another most widely used distribution, especially for lifetime distributions in reliability
engineering due to its versatility, is the Weibull distribution. The Weibull distribution
can take on the characteristics of other types of distributions, based on the value of
its shape parameter, ‘β’ (Devore 2008). If the log normal distribution fails to represent
the observed data then Weibull distribution will be applied.
The four descriptive parameters which are of particular interest are mean, median,
skewness, and kurtosis. These descriptive parameters describe the shape of the
distribution.
The values of mean and median describe the skewness of the data (Hamburg,
1983). If the mean exceeds the median, the distribution is said to have positive
skewness. If the mean is less than the median then the distribution is said to have
negative skewness. From tables note that, for each loading area and period, the
mean of the observed data is greater than the median. This validates the results of
histogram analysis and concludes that the bus lost time at each loading area is
positively skewed distribution.
The value of skewness measures the asymmetry (skewness) of the distribution. For
a perfectly normal distribution the value of skewness is zero, i.e. perfectly symmetric.
For each loading area the value of skewness was found to be greater than 0, as
expected.
Count 50 50
Count 30 30
Count 14 14
Count 61 61
Count 50 50
Count 38 38
To fit the lognormal distribution, the natural log of the observed data set was taken
and test for normality was performed. The test of normality of log-data is described in
next section.
In this research the normality of observed data set was assessed using histogram
plot, and skewness score. The normality of log-data was assessed by testing null
hypothesis about skewness and kurtosis scores.
To test the null hypothesis the skewness statistic and Kurtosis statistic score
methods involve construction of 95% confidence interval (z0.025 = 1.96) about a
skewness score (or a kurtosis score). The null hypothesis, in this case, is that the
skewness estimate (or kurtosis estimate) is not significantly different from a value of
zero, i.e. the estimated score is from a standard bell shaped normal distribution. If
the value of ‘zero’ is within the 95% confidence interval than the null hypothesis can
be accepted, otherwise reject the null hypothesis.
The calculations for loading area 1, using off-peak data set, are presented here to
demonstrated the null hypothesis testing method. Using the statistic and standard
error values of skewness, the lower and upper bound of its 95% confidence interval
was determined. The standard z value for 95% confidence is 1.96. The lower bound
of skewness is -0.669 (Statistic plus 1.96 times the std. error) and the upper bound
of skewness is 0.653 (Statistic minus 1.96 times the std. error). Hence the 95%
confidence interval for the skewness ranges from -0.669 to 0.653. Since, zero, the
main value of interest lies with the 95% confidence interval, the null hypothesis is
accepted. This means that natural logarithm of observed data set is normally
distributed and therefore it can be concluded that the off-peak bus lost time data set
for loading area 1 is log-normally distributed.
Similarly, the 95% confidence interval of the kurtosis for off-peak data of loading area
1 ranges from -0.669 to 0.653.
The above steps were repeated for other loading areas. The test of normality for
each loading area during off-peak and peak periods are summarised in Table 6.8 to
Table 6.13. Note that the value of zero for skewness and kurtosis falls within their
respective 95% confidence interval, establishing that natural logarithm of observed
data set is normally distributed. Hence, the bus lost time data set for all loading
areas, in off-peak and peak periods was found to be log-normally distributed.
1 ⁄
0
; ; √2 Equation 6.1
0 0
; ; 0 Equation 6.2
Where,
= An event; in this case bus lost time, whose probability of occurrence
is to be calculate
= Mean of In(X)
= Standard deviation of In(X)
X = Variable; in this case observed bus lost time
Note that the parameter µ and σ are not the mean and standard deviation of X but of
In(X). The mean and standard of X can be calculated as
⁄
X eµ
Equation 6.3
⁄
X eµ
1 Equation 6.4
Where,
= Mean of variable X
= Variance of variable X
Figure 6.4 provided the bus lost time probability distribution function (PDF) for the
lognormal distribution curves for loading area 1 and 2 for off-peak time periods. Due
to the insufficient observation during off peak operation of loading area 3 the curve
fitting was not done for it. During off-peak period the loading area 1 and 2 have
identical probability of experiencing a lost time of 2.9s. This value of lost time is less
than that of peak period bus lost time value, which is 4.1s. However, similar to peak
period curves the loading area 2 has higher chances of experiencing a lost time less
than 2.9s in comparison to loading area 1.
The comparison between peak and off peak curves is shown in Figure 6.5. The
figure shows that the off-peak curve for loading area 1 is shifted toward right relative
to the peak period curve. This shift indicates that loading area 1 is more likely to get
higher lost time during low crowd situation. On the contradictory, for loading area 2
the likelihood of a given lost time is fairly same in both periods of operation. In fact,
there are two different movements in probabilities for loading area 2. While the
probability for lost time less than 1.8s and for lost time greater than 5.5s decreased
in peak period, the probability for lost times between 1.8s and 5.5s increased.
These results are consistence with the observations in Figure 5-7 and passenger
behaviours explained in section 5.4.
0.35 LA 1 (Peak)
LA 2 (Peak)
0.30
LA 3 (Peak)
0.25
Probabilty, f(x)
0.20
0.15
0.10
0.05
0.00
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Lost time (x)
Figure 6.3: Bus lost time probability distribution curves (Peak period)
0.35 LA 1 (OP)
LA 2 (OP)
0.30
0.25
Probabilty, f(x)
0.20
0.15
0.10
0.05
0.00
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Lost time (x)
Figure 6.4: Bus lost time probability distribution curves (Off-peak period)
0.35 LA 1 (Peak)
LA 2 (Peak)
0.30 LA 3 (Peak)
LA 1 (OP)
0.25 LA 2 (OP)
Probabilty, f(x)
0.20
0.15
0.10
0.05
0.00
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Lost time (x)
Figure 6.5: Comparison of peak and off-peak bus lost time probability distribution curves
0.9
85th percentile
0.8
0.7
Cumulative probabilty
0.6 LA 1 (Peak)
0.5
LA 2 (Peak)
0.4
LA 3 (Peak)
0.3
0.2
0.1
0
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Bus lost time
Figure 6.6: Bus lost time cumulative distribution curves (Peak period)
0.9
85th percentile
0.8
0.7
Cumulative probabilty
0.6
0.5 LA 1 (OP)
LA 2 (OP)
0.4
0.3
0.2
0.1
0
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Bus lost time
Figure 6.7: Bus lost time cumulative distribution curves (Off-peak period)
0.9
0.8
0.7
Cumulative probabilty
0.6 LA 1 (Peak)
LA 2 (Peak)
0.5
LA 3 (Peak)
0.4
LA 1 (OP)
0.3 LA 2 (OP)
0.2
0.1
0
0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Bus lost time
Figure 6.8: Comparison of peak and off-peak bus lost time cumulative distribution curves
The CDF also demonstrates the variations in upper range of bus lost time with
loading area. For example, during the peak period operation, 85 percent of buses
using loading area 1 may have a lost time value less than or equal to 7.1s. On the
other hand, 85 percent of buses using loading area 2 and loading area 3 may have a
lost time value less than or equal to 4.5s and 7.1s respectively.
Figure 6.7 shows the cumulative distribution function (CDF) of bus lost times for
loading area 1 and 2 for off-peak time period. Similar to peak period, in off-period too
the bus lost times at loading 1 are higher compared to loading area 2.
Figure 6.8 compares the bus lost time curves of peak and off-peak period. For
loading area 2 the cumulative probabilities in peak and off-peak are almost identical.
Though for loading area 1 there is increased probability of higher lost time in off-peak
period.
Overall, this chapter has three key contributions in evolving the knowledge of
busway station bus lost time. It,
1) Established the quantitative descriptions of the busway station bus lost time.
2) Developed a set of loading area centric and time period specific probability
models.
3) Defined the descriptive characterises of busway station bus lost time.
With bus lost time variable quantified in this chapter, the next chapter discusses the
development of busway station bus dwell time model incorporating the bus lost time.
Busway Station Dwell Time Model
7.1 Overview
Previous chapters established the bus lost time variable as a significant parameter
influencing bus dwell time at a busway station platform. Along with the traditional
variables of the number of boarding and alighting passengers and respective
marginal time and door opening and closing time, bus lost time is also a determinant
of dwell time of a bus at a busway station.
This chapter describes the development of the busway station bus dwell time model,
incorporating the concept of bus lost time. Section 7.2 describes the framework of
the model. The mathematical model is presented in section 7.3. Section 7.4
discusses the effects of bus lost time on bus dwell time estimation. Section 7.5
summaries the key features of the new model and discusses model implications.
Section 7.6 closes the chapter.
The number of boarding and alighting passengers and their marginal service
time, during the passenger – bus interaction phase as described in Section 5.6, are
the most traditional and widely used variables in the dwell time estimation. As
discusses in section 5.6, these variables can be modelled using existing bus stop
dwell tide models. For the purpose of this research, the bus stop dwell time model
given by TCQSM, Part 4 (TRB, 2003) was used (Equation 2.5).
The door time is a time consumed by a bus to fully open its doors after the bus has
come to rest and close it back fully after the completion of passenger servicing. The
TCQSM dwell time model also accounts for door time, i.e. door opening and closing
time in its equation (Equation 2.5).
The bus lost time variable results from the complex process between the platform
crowd, passenger walking and the bus itself. As explained in Figure 5.4, the bus lost
time occurs, prior to start of actual boarding, but after the bus has come to rest and
is additive to the traditional variables of the number of boarding and alighting
passengers and door time. The theoretical as well as statistical analyses on bus lost
time, presented in previous chapters, established that the lost time varies across the
platform from loading area to loading area and between peak and off-peak periods.
This implies that the bus dwell time at each loading area varies depending on the
bus lost time at the loading area under the prevailing conditions and is not a uniform
value across the platform length.
Figure 7.1: Overview of model form for busway platform dwell time estimation
Equation 7.1
Where,
= Bus dwell time at nth loading area
; = Number of passenger boarding and alighting respectively
; = Service time per boarding and alighting passenger
respectively (s)
= Bus door opening and closing time (s)
= Bus lost time at nth loading area (s)
This model has bus lost time as an additional term compared to the traditional
Equation 2.5. Note that the bus lost time term is loading area specific. Hence this
bus dwell time model estimates different dwell time values for each specific loading
area.
The outbound platform of the example busway station has three linear loading areas
with an adjacent passing lane. A boarding load of 7 passengers per bus with
boarding service time of 4s and no alighting load for the front entry door was
considered. The estimation was done with an assumed door opening and closing
time of 2s. The mean bus lost time values obtained from Table 6-15 were used in
BSDT equation. Table 7.1 shows the comparison of the results from two methods.
Unlike TCQSM model which estimated a single dwell time value for all three loading
areas, the BSDT method which considered the bus lost time estimated different
dwell time values for each specific loading area depending on the prevailing
operation condition of the busway station platform. For a constant passenger load,
all three loading areas have different bus dwell times, due to their dissimilar bus lost
times. The lowest bus dwell time was found for loading area 2, whereas the highest
dwell time was found for loading area 3. The example demonstrated the effects of
bus lost time on bus dwell time at platform level.
The effects of bus lost time can also be seen at the scale of the individual loading.
The bus lost time is a loading area specific variable and accounts the process
between the bus and the first boarding passenger only. Hence if more passengers
board per bus the proportion of lost time in the total dwell time of bus decreases.
Similarly if the marginal boarding time increases, the share of lost time in the total
dwell time of the bus decreases. To clarify through an example, the bus dwell time
was estimated for loading area 1 with varying boarding load and marginal boarding
time. The influence of bus lost time on bus dwell time gradually decreased with
increase in boarding load. The means that the impact of bus lost time on a bus is
subsidised with increased boarding load.
On the contrary, the influence of lost time increased with a decrease in marginal
boarding time. The inference of this result is that with a decrease in marginal time,
the bus lost time becomes more crucial component in bus dwell time. Figure 7.2
shows the changing influence on bus dwell time at the loading area 1.
50
tb = 2s
40
30
tb = 4s
20
10 tb = 5s
0
1 2 3 4 5 6 7 8 9 10
Boarding load per bus
Figure 7.2: Effect of bus lost time on dwell time at loading area 1 (Peak period)
Similar kind of trend of bus lost time influence on dwell time can be observed for
other loading areas and for peak and off-peak periods. Figure 7.3 to Figure 7.6
present these graphs. It is worth noting here that the above demonstration example
is designed using mean bus lost time value. The results and shape of these graphs
will change with variation in bus lost time value.
No graph for off-peak period of loading area 3 was plotted because no statistical
analysis was done due to the lack of data.
tb = 2s
50
40
tb = 4s
30
20
tb = 5s
10
0
1 2 3 4 5 6 7 8 9 10
Boarding load per bus
Figure 7.3: Effect of bus lost time on dwell time at loading area 1 (Off-peak period)
45
Contribution of lost time in dwell time (%)
40
tb = 2s
35
30
25
tb = 4s
20
15
10
tb = 5s
5
0
1 2 3 4 5 6 7 8 9 10
Boarding load
Figure 7.4: Effect of bus lost time on dwell time at loading area 2 (Peak period)
45
tb = 2s
40
35
30
25 tb = 4s
20
15
10 tb = 5s
5
0
1 2 3 4 5 6 7 8 9 10
Boarding load
Figure 7.5: Effect of bus lost time on dwell time at loading area 2 (Off-peak period)
60
Contribution of lost time in dwell time (%)
tb = 2s
50
40
tb = 4s
30
20
tb = 5s
10
0
1 2 3 4 5 6 7 8 9 10
Boarding load
Figure 7.6: Effect of bus lost time on dwell time at loading area 3 (Peak period)
7.5 Discussion
The bus dwell time is a crucial factor in the design of bus rapid transit system. Along
with some other factors, it determines journey time and adherence of the bus service
with its schedule. Station bus capacity, which is determinant of line bus capacity,
also depends on the bus dwell times. With the recent increase in emphasis of
providing real time information, accurate estimation of bus dwell times at stations
also becomes more important. The new methodology for dwell time estimation has,
therefore, multiple implications while designing a bus rapid transit system. The key
features of BSDT model are –
The finding that the impact of bus lost time on bus dwell time varies with boarding
load could potentially help in the development of new tools to optimise service
frequencies to achieve reduced dwell times. Whereas, the discovery of the
relationship between lost time and marginal boarding time relation is vital to derive
full advantages of the improvement in fare collection methods, such as a smart card
system. The benefits in bus dwell from reduction in marginal service time could be
substantially nullified due to bus lost time.
Another important finding is the viability of bus dwell time across the loading area.
This variability will impact the bus capacities of individual loading area and therefore
the combined bus capacity of the platform.
Busway Loading Area Bus Capacity Model
8.1 Overview
This chapter describes the development of busway loading area bus capacity model.
This new capacity model incorporates the new busway dwell time model, presented
in chapter 7, which in turn incorporated the bus lost time variable, described in
chapter 6.
The next section portrays the approach to model development and discusses the
variables influencing the busway loading area bus capacity. Section 8.3 presents the
busway loading area capacity model. Section 8.4 presents the model for effective
loading area bus capacity and Section 8.5 presents the model for busway station
bus capacity.
Section 8.6 demonstrates calculations for an example case and discusses the
effects of bus lost time on loading area bus capacity. Section 8.7 provides a
discussion on the new model. The chapter closes with section 8.8.
3600
Equation 8.1
A more detailed model for estimation of bus capacity of a loading area depends on
dwell time, clearance time, dwell time variability considering failure rate (TRB, 2003).
Further, this model incorporates, in fairly simplistic terms, the effects of immediately
adjacent signals, which eliminate some of the time available to clear buses. The
Equation 2-19 is reproduced here as Equation 2.19.
3600 ⁄
Equation 8.2
⁄
Where,
The various determinants of busway loading area bus capacity are discussed below.
Equation 8.3
Where,
= Operating margin for passenger load variability (s)
= Standard normal variable corresponding to a desired
failure rate
= Coefficient of variation of bus dwell time
= Clearance time (s)
The Equation 8.3 is based on the assumption that the dwell times are normally
distributed. Table 8.1 reproduces the Table 2.4, which provides the value of ‘z’
corresponding to desired failure fate.
Failure rate z
1.0% 2.330
2.5% 1.960
5.0% 1.645
7.5% 1.440
10.0% 1.280
15.0% 1.040
20.0% 0.840
25.0% 0.675
30.0% 0.525
50.0% 0.000
Source: TRB, 2003
In this research a new bus dwell time model was developed to estimate average
dwell time at busway station which includes a variable for bus lost time. The bus lost
time variable at busway station is loading area specific and depends on the bus and
its first passenger. Unlike kerbside bus stop dwell time, which was assumed as
normally distributed (TRB, 2003), the bus lost time component studied through this
research was found to be log-normally distributed. Therefore, the Equation 8.3 is
modified as Equation 8.4 to make it appropriate for busway station analysis. This
equation is called as passenger service time operating margin. The equation
accounts for the fluctuation in service time due passenger loads between buses and
routes.
, ,
Equation 8.4
Where,
, = Operating margin for passenger service time variability (s) for nth
loading area
= Standard normal variable corresponding to a desired failure rate
, = Coefficient of variation of passenger service time at busway
loading area
= Busway dwell time at nth loading area (s), as defined in chapter
7.
= Average bus lost time at nth loading area (s)
In the above equation the bus lost time component of busway dwell time equation
was removed to calculate the operating margin for dwell time variability. The
remaining component i.e. ( ) is identical to that for the standard kerbside
bus dwell time model of TCQSM (TRB, 2003). The coefficient of variation here,
however, is that of the passenger service time. The operating margin due to the bus
lost time variability will be treated separately due to its different statistical properties
from the marginal service times.
failure rate (TRB, 2003), this research proposes a variable, lost time operating
margin (toml), to account for bus lost time variability by way of failure rate. The lost
time operating margin is a value by which the bus lost time at a given loading area
will increased for the desired failure rate and is additive to the mean bus lost time.
Statistically, the area under the log-normal distribution to the right of point LTi (the
shaded area in Figure 8.1) represents the probability that the lost time for a bus at a
given loading area will be longer than the LTi. Z is the corresponding variates, given
by:
Equation 8.5
Equation 8.6
Note that parameter µ and σ are not the mean and standard deviation of variable,
bus lost time but the log of the variable.
Subsequently, the operating margin for bus lost time can be given by
Equation 8.7
⁄
Equation 8.8
The generalised model for lost time operation margin is therefore given as:
⁄ Equation 8.9
,
Where,
3600
Equation 8.10
, ,
Where,
= Bus capacity of nth loading area (bus/hr)
3600 = Number of seconds in 1 hour
= Bus dwell time at nth loading area (s)
, = Operating margin for passenger load variability (s)
It is noted that this model applies to a loading area away from the influence of
signalised intersection. Equation 8.10 may be enhanced to yield Equation 8.11 when
adjacent signals impact loading area operation. The model is known as BSLC model,
an abbreviation of Busway Station Loading Area Bus Capacity Model.
3600 Equation 8.11
, ,
Where,
= green time ratio (the ratio of effective green time to total traffic signal
cycle length, equals 1.0 for unsignalised streets and bus facilities like
busway without adjacent signal control)
, Equation 8.12
Where,
The discussions and modelling of linear loading areas efficiencies are presented in
chapter 9.
, Equation 8.13
Where,
= Busway station platform bus capacity (Bus/ hr)
coefficient of variation (TRB, 2003), 3 linear loading areas, g/c of 1, 10s clearance
time, and door opening and closing time of 2s. The mean bus lost time obtained from
Table 6-14 and loading area efficiency factors from Table 9.3 were used for revised
capacity design. The loading area efficiency factors for TQCSM method are obtained
from Table 9-1 were used. Table 8.2 shows the comparison of the results from the
two methods. The calculation steps are provided in details in Appendix B.
The above table demonstrated the reduction in station bus capacity value when the
bus lost times were accounted in the busway station bus capacity estimation. For the
given demonstrative example, 13.3 percent reduction in bus capacity was observed.
For the above example, Figure 8.2 shows the variation in busway station bus
capacity station with boarding load per bus, calculated using both methods: TSCQM
and BSLC model. As expected the bus capacity deceases with an increase in
boarding load. This is due to the fact the buses remain at the loading area for longer
with an increase in boarding load. However, the difference in bus capacity estimation
between both methods decreases with an increase in boarding load. This illustrates
the effect of bus lost time on capacity with the increase in boarding load. This finding
is reasonable because, as discussed in past chapters, the lost time for a bus occurs
due to its first boarding passenger only. Other boarding passengers do not contribute
to lost time.
500
450
Busway station capacity (Bus / hr)
400
TCQSM
350
300 BSLC
250
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10
Boarding load per bus
Note: busway station with three linear loading area, no alighting load, tb=4s, failure rate 7.5%, cv=0.6,
tc=10s.
Figure 8.2: Variation in busway station bus capacity with boarding load per bus
40
Effect of bus lost time on station capacity
35 tb = 2s
30
25
20
15
tb = 4s
10
0
1 2 3 4 5 6 7 8 9 10
Boarding load per bus
Note: busway station with three linear loading area, no alighting load, failure rate 7.5%, cv=0.6, tc=10s.
Figure 8.3: Effect of bus lost time on busway station bus capacity
Therefore, as the boarding load increases, the share of lost time in the total dwell
time of the bus decreases, resulting in its lesser influence on capacity.
On the contrary, when marginal service time decreases, the influence of lost time on
capacity increases, as seen from Figure 8.3. This is because, with a decrease in
marginal service time, the share of lost time in the total dwell time of the bus
increases.
Figure 8.4 provides busway station bus capacities for various g/C ratios and
boarding loads, assuming no alighting load from entry door. The figure illustrates the
impact of adjacent signal on loading area bus capacity. As seen from the graph, the
station bus capacity decreased with decrease with g/C ratio.
350
300
g/C = 1
250
g/C = 0.7
Station bus capacity
150
100
50
0
1 2 3 4 5 6 7 8 9 10
Boarding load
Note: busway station with three linear loading area, no alighting load, tb=4s, failure rate 7.5%, cv=0.6,
tc=10s.
Figure 8.4: Estimated busway station capacities
8.7 Discussion
Determining the reliable and accurate loading area bus capacity is absolutely critical
for design of a bus transit system. Specially, in case of a busway system, where its
stations are the major source of delay to buses, the precise estimation of station bus
capacity becomes more important in determining the line capacity and design period.
With the bus lost time accounted for the station bus capacity estimation, the new
methodology better approximates the station operation. The key features of the
BSLC model are,
1) It is sensitive to bus lost time and its variability between loading areas.
2) It is sensitive to the platform crowd and passenger walking to the bus entry
door.
The comparison of station capacities, estimated from TCQSM method (TRB, 2003)
and the BSLC model, presented in this chapter, showed that the busway station bus
capacity reduced when bus lost time was considered. This shows that bus lost time
affects the busway station bus capacity and needs to be considered.
With the knowledge of bus lost time characteristics, better operation and
management policies can be developed. For example, the advantages of a reduction
in marginal service time can be undermined because of bus lost time. Any gain in
bus dwell time might be limited until steps towards reducing bus lost time are made.
Furthermore, a busway line can potentially serve as trunk line in a trunk and branch
type arrangement of public transport network. The buses on the trunk line would be
of high frequency and high speed service connected to feeder services, which
branch out deep into the suburban areas. With the improvement in the
understanding of operation of busway and its stations, a better coordinated service
may be created.
All steps involved in busway station platform bus capacity analysis are compiled in a
worksheet provided in Appendix A.
The next chapter presents the methodology to estimate the loading area efficiency
and effective numbers of loading areas for a busway station platform.
Busway Station Efficiency Model
9.1 Overview
Under a perfectly ideal situation each loading area of a busway (BRT) station
platform should operate without interfering with the others’ operation. However in a
linear arrangement of loading areas this may not be the case, particularly during the
peak period operation when the bus flow is high. Every loading area can potentially
obstruct the bus entry to the platform (as buses are not permitted to slip into
downstream, unoccupied loading areas) and obstruct exiting from upstream loading
areas. This reduces the number of effective loading areas for the station from the
physical number, as noted by TCQSM (TRB, 2003).
Section 9.4 estimates the loading efficiencies based on the experience of the Mater
Hill busway station, Brisbane, Australia. Section 9.5 provides a discussion on loading
area efficiency. Section 9.6 closes the chapter.
The study platform i.e. outbound platform of Mater Hill busway station, Brisbane, has
a total of three loading areas. The station platform reaches its ideal bus capacity
when all the three loading areas are occupied by buses. However, as the buses use
the loading areas on a first come first in basis, the station may reach a non-ideal
capacity when one or both of loading area 1 and 2 are empty but a preceding loading
area is occupied. Loading area 3 cannot experience blocking since it does not have
any predecessor loading area. Figure 9.1 shows bus arrivals and departures
progression observed at the study platform. The ‘red line’ represents buses which
used loading area 1 for passenger servicing. Similarly, the ‘orange line’ and ‘blue
line’ represent buses used loading area 2 and 3 respectively. A dotted line means
bus in queue at station entry, behind the loading area 3. Whereas, a solid line means
the bus is dwelling at a loading area.
The start point of a dotted line represents the entry of the bus in queue. The start of
a solid line represents the arrival of that bus at the loading area. The end point of the
solid line represents the departure of that bus from the loading area after completing
passenger servicing. The length of solid line gives the time spent by the bus at the
loading area, including its dwell time. The length of dotted line gives the time spent
by that bus in the queue. A lack of a dotted line for any bus means that bus it is able
to enter an empty loading area directly upon its arrival at the station platform.
During the non-peak period, bus arrivals at the station platform were random and
most of the buses found an empty loading area at the platform. With the onset of
peak period the flow of bus increased and similarly so their dwell time at the
platform. This caused the loading area to be blocked by buses in its predecessor
loading areas as Figure 9.1 illustrates. Loading area 1 can be seen in this figure
being blocked by either loading area 2, 3, or both. Similarly, loading area 3 blocks
loading areas 2 and 1.
Platoon departure
Platoon arrival
at platform.
There were five different scenarios of inter-loading area blocking observed at the
study platform as shown in Figure 9.2. A loading area which is empty but blocked is
shown in red shading. A loading area occupied a bus is shown in gray shading. An
empty loading area available for bus is shown in white shading.
Scenario 1: All the three loading areas are available for buses to serve the platform.
The available bus capacity at the platform is 100 percent. The platform is performing
under capacity.
Scenario 3: Loading area 2 and 3 are occupied, which results in blockage of loading
area 1. In this scenario, the platform has one third of its capacity blocked and the
remaining two third of its capacity is in utilization. The platform is virtually performing
at full capacity with a loss of one third of its total capacity.
Scenario 4: Loading area 1 and 3 are occupied. However, the bus at loading area 3
is blocking entry to loading 2. In this scenario, like scenario 3, the platform has one
third of its capacity blocked and remaining two third capacity been utilised. The
platform is said to be virtually performing at full capacity with a loss of one third of its
total capacity.
Scenario 6: All the three loading areas are occupied, resulting ideal capacity for the
platform. The platform is performing at full capacity with no loss in capacity.
In Figure 9.2, scenarios 2, 3, and 4 have equal numbers of effective loading areas.
However, in scenario 2, the platform is not at its full capacity. There is a spare
capacity of one loading area. Therefore, arrival of one additional bus at the platform
will not result in formation of a queue at platform entry, i.e. behind loading area 3. On
the other hand, in scenario 3 and 4, the platform is at its full capacity, with no
available loading for any further bus. Arrivals of additional buses at the platform
under these scenarios cause these buses to queue behind loading area 3. This
situation could potentially lead to scenario 5, where loading area 3 blocks the entry
to loading areas 1 and 2. This causes the platform to lose its two third of its total
available loading areas. Under a continuous bus arrival situation, this may result in
bus arrivals at station, arrival at platform, and departure from the platform, in
platoons, as highlighted in the Figure 9.1.
Theoretically, loading area 3 will not be blocked at any time because no bus will
stand in queue if loading area 3 is empty. But a bus on loading area 3 can block the
entry to empty loading area 2. For instance, consider that all loading areas are
occupied by buses and there is a bus queued behind loading area 3. A bus on
loading area 2 moves out of the platform, making this loading area available for the
next bus. But if the bus on loading area 3 is still dwelling at the platform, it blocks the
entry to loading area 2. During data collection, the amount of time that loading area 2
is blocked because of the presence of a bus on its upstream loading area was
recorded as blocked time. The efficiency of loading area 2 is given by Equation 9.1.
,
Equation 9.1
Where,
E2 = Efficiency of loading area 2
T3 = Total time that loading area 3 is occupied during time T
T2,b = Total time that loading area 2 was empty while a bus occupied loading
area 3 during time T
T = analysis period
Similarly a bus on loading area 2 or loading area 3 can block loading entry to empty
loading area 1. The efficiency of loading area 1 of the platform at the station is given
by Equation (4).
, ,
Equation 9.2
,
Where,
E1 = Efficiency of loading area 1
T2,3 = Total time that loading area 3 OR loading areas 2 and 3 are occupied
while there is a queue behind loading area 3 during time T
T1,b = Total time that loading area 1 was empty while a bus occupied loading
area 2 OR loading area 3 OR both loading areas 2 and 3
The number of effective loading areas (Nel) on the three loading area station platform
may then be calculated using Equation 9.3.
9.4 Loading area efficiency factors for Mater Hill Busway Station
Table 9.2 shows the occupancy rate and blocking rate for each loading area during
the half hour analysis time period (T=1800s).
Table 9.2: Occupancy and blocking rates for loading areas at outbound platform of Mater Hill
Busway Station (Afternoon peak period)
1 790s 372s
2 958s 82s
3 915s 0s
Table 9.3 presents the calculation of the number of effective loading areas on the
outbound platform at Mater Hill Station, during the analysis period, T.
Table 9.3: Number of effective loading areas calculation for bus station platform
These results indicate that, during peak period, the outbound platform of Mater Hill
busway station has 2.61 effective loading areas out of the available 3 loading areas.
In other words, the platform has a loss of 13 percent of its total available physical
loading areas. The results welcomely resemble the default values from TCQSM
(TRB, 2003) as shown in Table 9.4.
9.5 Discussion
The loss in the effective number of loading areas on a busway (BRT) platform results
in a reduction in its bus capacity. This reduction amplifies queuing of buses at the
station platform entry. This, in turn, tends to result in a higher level of platooning.
Platooned arrival means the station is forced to function at its full capacity and with
no opportunity to fall back below capacity. It also means that the last loading area
(Loading area 3, in the case of study station) would have increased usage,
subsequently resulting in a rise in blocking of the downstream loading areas.
The arrival of buses in platoon at a station suggests that buses are carrying the
effects of operation of the upstream station to that station. Similarly, the departure of
buses in platoon from that station shows that buses are carrying forward the effects
of that station. This was referred to in Section 3.4 as station – station interface.
Detailed analyses of platoon arrivals and departures were not in the scope of this
research but warrant further investigation. Further research on bus discipline at
stations, passenger ticking and management, including prepaid ticketing and all-door
usage should also be pursued in future.
Next chapter concludes this thesis and discusses the future research opportunities in
the area of bus lost time to enhance the models.
Conclusions
10.1 Overview
This chapter concludes this thesis and summarises the analyses, discussions and
results presented in past chapters. A brief thesis summary is provided in section
10.2, followed by section 10.3 which discusses the contributions of this research to
the existing knowledge and practice. Later, section 10.4 discusses the implications
of the research findings for both theory and practice. Section 10.5 provides the
conclusions of this research work and finally the recommendations for further
research are given in section 10.6.
Based on the gaps identified, the research problem was formalised in Chapter 3. The
chapter studied the operations of the busway station compared with the traditional
kerbside side bus stop and developed a framework of busway station operation. The
framework identified and elaborated various parameters and processes occurring at
four tiers of busway operation (platform, vehicle, station, and line) and their influence
on each another. The analyses lead to the identification of a previously neglected
process, which this study refers in this thesis as passenger – bus interface, and a
new parameter, termed here as bus lost time.
To study the previously neglected process and parameter, and to identify their
influence on busway station operation, pertinent data was collected. Chapter 4
detailed the development of data collection and processing methodologies for this
research. This chapter presented a matrix based concept for data mining to achieve
homogeneous data for analysis.
Four specific models were then developed in this research, one model each to
estimate bus lost time, busway dwell time, busway loading area bus capacity, and
busway loading area efficiency.
Chapter 6 analysed and modelled bus lost time. Stochastic models were developed
for this previously neglected variable, and its descriptive statistics were established.
With the finding of the bus lost time variable and its roles in operation of a busway
station, a new busway dwell time model was developed in Chapter 7, and a new
busway loading area bus capacity model was developed in Chapter 8. This new
formulation of busway dwell time, in addition to passenger boarding and alighting
load and their marginal service times, also account for bus lost time.
The third model developed in this research was busway loading area bus capacity
model, presented in Chapter 8. The new busway loading area bus capacity model
can better approximate the busway loading area operation for two reasons: firstly, it
incorporates the more accurate busway dwell time model and secondly, it accounts
for variation in bus lost times. A new term, lost time operating margin, was defined,
which allows for variability in bus lost time between buses. Steps for calculating
busway station bus capacity were also presented this chapter.
The total bus capacity of a platform with multiple linear loading areas depends on the
efficiencies of each loading areas. Linear loading areas tend to interfere with the
smooth operation of adjacent loading areas. Such interference causes reductions in
efficiencies. Impacts of one linear loading area on another were studied in Chapter 9.
Based on the experience of the study busway platform, loading area efficiency
models were developed and efficiency factors were estimated for linear loading
areas.
In order to demonstrate the influence the bus lost time has on busway dwell time and
busway bus capacity, demonstrative examples were presented. Through these
examples, the changing patterns of dwell time and bus capacity at a busway platform
were illustrated.
From the practice perspective, the main contribution of this research is the
development of a more refined tool for estimating busway station bus capacity. In
order to develop this tool, this research has,
The concept of bus lost time could be applied to develop a two tier real-time
information system for linear loading area busway station. The first tier of the
information would provide the expected arrival time of a bus at the station. Whereas
the second tier of the information would provide the loading area number for that
bus. Such information should be provided to passengers very close in time to the
actual arrival of the bus, in order to reduce bus lost time.
From a practical application prospective, the new methodology for estimating bus
dwell time at a busway station can help transit planners in improving scheduling and
in turn could greatly enhance the travel time reliability. The new methodology, which
is specifically designed for a linear loading area station can help transit planners to
perform robust capacity analysis of future bus transit system. This tool can naturally
also be applied in evaluating an existing station. Since this methodology of capacity
analysis accounts for the impact of passenger crowd at a busway station, growth in
patronage can be better incorporated in future policies; in particular in the design of
service frequencies.
10.5 Conclusions
In conclusion, this thesis found that the traditional approach of bus dwell time
estimation may not be reliable for a busway station. This is because of the
significance of additional complex variables, such as the presence of large crowds,
multiple linear loading areas, multiple bus services, bus queuing and lost times.
These complex variables have been analysed and subsequently, new models for
dwell time estimation and platform bus capacity were presented in this thesis.
A demonstrative example comparing the traditional approach and new approach for
bus dwell time estimation at a busway station showed that the new approach
estimated the varying dwell time values for different loading areas. It showed that
new approach was able to approximate in finer detail the dwell times of a real
busway platform, compared to the single dwell time value for all the loading areas
given by the traditional model.
The comparison of station capacities showed that the bus capacity of a busway
platform reduced with the bus lost time considered. This highlighted the importance
of accounting for bus lost time in deciding the final throughput of the station.
This research suggested the use of lognormal probability distribution curves for
estimation of bus lost time for three linear loading areas. It would be interesting to
see the impact on lost time distribution if an additional fourth loading area is added to
the current set of three. The bus lost time estimation approach needs to be refined
with the help of more case studies to make it applicable to a wide range of linear
loading area station operations.
Furthermore, this research has only considered the bus lost time due to its first
boarding passenger. Additional research work in the area of bus lost time is
necessary to study the impact of subsequent passengers. Specifically, the impact of
platform crowd on subsequent passenger walking time and in turn on bus lost time
ought to be studied.
The results of this research are based on a predominantly boarding platform, with
very little passenger alighting activity. The impact of alighting passengers on bus lost
time and passenger walking on the platform was not fully studied. More studies
based on the stations where alighting passengers have a substantial impact on
station operation are recommended. Specifically, predominantly alighting platform
would be located in the inbound direction at inner urban platforms, and the morning
peak period would be expected to be critical.
Additionally, this research also recommends the development of a full scale micro-
simulation model for the busway station to refine and strengthen the Busway station
platform capacity methodology suggested by this research.
References
Ando, K., Ota, H. and Oki, T. (1988). Forecasting the flow of people, (in Japanese),
RRR, Railway Research Review, 45(8), 8-14.
Blue, V. J. and Adler, J. L. (2001). Cellular automata microsimulation for modeling bi-
directional pedestrian walkways. Transportation Research Part B, 35, 293-312.
Chan, L.Y., Lau, W.L., Zou, S.C., Cao, Z.X. and Lai, S.C. (2002). Exposure level of
carbon monoxide and respirable suspended particulate in public transportation
modes while commuting in urban area of Guangzhou, Chain. Atmospheric
Environment, 36, 5831-5840.
Chertok, M., Voukelators, A., Sheppeard, V. and Rissel, C. (2004). Comparison of air
pollution exposure for five commuting modes in Sydney – car, train, bus,
bicycle and walking. Health Promotion Journal of Australia, 15(1), 63-67.
Cox, T., Houdmont, J. and Griffiths, A. (2006). Rail passenger crowding, stress,
health and safety in Britain. Transportation Research part A, 40, 244-258.
Davies, A. C., Yin, J. H. and Velastin, A. (1995). Crowd Monitoring Using Image
Processing. IEE Electronic and Communications Engineering Journal, 7(1), 37-
47.
Devore, J. L. (2008). Probability and Statistics: For Engineering and the Sciences.
Thomson/Brooks/Cole Inc., Belmont, CA.
Dueker, K., Kimpel, T. and Strathman, J. (2004). Determinants of bus dwell time.
Journal of Public Transportation, 7(1), 21-40.
Fruin, John J. (1987) Pedestrian Planning and Design, Revised Edition, Elevator
World, Inc., Mobile, AL
Guenthner, R. P., and Hamat, K. (1988). Transit Dwell Time Under Complex Fare
Structure, Journal of Transportation Engineering, 114(3), 367-379.
Guenthner, R. P., and Sinha, K. C. (1983). Modeling Bus Delays Due to Passenger
Boardings and Alightings. In Transportation Research Record: Journal of the
Transportation Research Board, 915, 7-13.
Hamburg, M. (1983). Probability and Statistical Analysis for Decision Making. 3rd Ed.
Harcord Brace Jovanovich, Inc., New York, NY.
HCM (1994). Highway capacity manual. Special Report, 209, 3rd ed. Transportation
Research Board, National Research Council, Washington, D.C.
Hui, M.K., and Bateson, J.E.G. (1991). Perceived control and the effect of crowding
and consumer choice on the service experience. The Journal of Consumer
Research, 18(2), 174-184.
Innes, P. (2007). Testing and fixing for normality. Factsheet No. 15, University of
Queensland Business School, Brisbane, Australia.
Jiang, R. and Wu, Q.S. (2005). The moving behaviour of a large object in the crowds
in a narrow channel. Physica A, 364, 457-463.
Johansson, A., Helbing, D., Al-Abideen, H. Z., & Al-Bosta, S. (2008). From crowd
dynamics to crowd safety: A video-based analysis. Advances in Complex
Systems, Vol 11, pp. 479-527.
Kholshevnikov, V. V., Shields, T.J., Boyce, K.E. and Samoshin, D.A. (in press).
Recent development in pedestrian flow theory and research in Russia. Fire
Safety Journal. doi: 10.1016 j.firesaf.2007.05.005.
Kilambi, P., Ribnick, E., Joshi, A., Masoud, O. and Papanikolopoulos, N. (2008).
Estimating pedestrian counts in groups. Computer Vision and Image
Understanding, 110, 43-59.
Kim, Won-Ho (2004). An improved bus signal priority system for networks with
nearside bus stop. PhD dissertation, Texas A&M University.
Kong, D., Gray, D. and Tao, H. (2005). Counting pedestrians in crowds using
viewpoint invariant training, In Proceeding of British Machine Vision
Conference. 5-8 September 2005, Oxford Brookes University, Oxford.
Lam, W., Lee, J. and Cheung, C. (2002). A study of the bi-directional pedestrian flow
characteristics at Hong Kong signalized crosswalk facilities. Transportation, 29,
169-192.
Lee, W.C.K. (2004). An investigation into the dwell time at Mater Hill Busway station.
Bachelor thesis, Queensland University of Technology, Brisbane.
Levine, J. C. and Torng, G. (1994). Dwell time effects of low-floor bus design.
Journal of Transportation Engineering, 120, 914-929.
Levinson, H. S., Adams, C. L. and Hoey, W. F. (1975). NCHRP Report 155: Bus Use
of Highways: Planning and Design Guidelines. Transportation Research Board,
National Research Council, Washington DC.
Levinson, H. S., Zimmerman, S., Clinger, J. and Rutherford, C. S. (2002). Bus Rapid
Transit: An Overview. Journal of Public Transportation, 5(2), 1-30.
Liggett, W.D. (1992). Low-Floor Bus--The St. Albert Experience. paper presented at
the CUTTER Fall Conference in Laval, Quebec.
Marshall, L. F., Levinson, L.C., Lennon, L. C., and Cheng, J. (1990). Bus service
times and capacities in Manhattan, In Transportation Research Record: Journal
of the Transportation Research Board, 1266, 189-196.
Mikovits, M. N. (2008). Modelling the factors affecting bus stop dwell time: Use of
automatic passenger counting, automatic fare counting, and automatic vehicle
location data. In Transportation Research Record: Journal of the Transportation
Research Board, 2072, 125-130.
Park, H. M. (2008). Univariate Anlaysis and Normality Test Using SAS, Stata, and
SPSS. Working paper. The University Information Technology Service (UITS)
Center for Statistical and Mathematical Computing, Indiana University.
Rajbhandari, R., Chien, S. I. and Daniel, J. R. (2003). Estimation of Bus Dwell Times
with Automatic Passenger Counter Information. In Transportation Research
Record: Journal of the Transportation Research Board, 1841, 120-127.
Seyfried, A., Passon, O., Steffen, B., Boltes, M., Rupprecht, T., and Klingsch, W.
(2007). New insight into pedestrian flow through bottlenecks, Transportation
Science, 43(3), 395-406.
Seyfried, A., Steffen, B., Klingsch, W. and Boltes, M. (2005). The fundamental
diagram of pedestrian movement revisited, Journal of Statistical Mechanics:
Theory and Experiment, 10, 1-13.
Shen, L. D., Elbadrawi, H., Zhao, F. and Ospina D. (1998). At Grade Busway
Planning Guide. Center for Urban Transportation Research, Florida
International University, The State University of Florida at Miami, Miami, FL.
Smith, R. A. (1995). Density, velocity and flow relationships for closely packed
crowds. Safety Science, 18, 321-327.
TCRP (1997). Transit Capacity and quality of service. TCRP web document 6,
Transportation Research Board, Washington, D.C.
TCRP (2003). Bus rapid transit volume 2: Implementation guidelines. TCRP Report
90. Transportation Research Board, Washington, D.C.
Teknomo, K., Yasushi, T. and Hajime, I. (2001). Tracking System to Automate Data
Collection of Microscopic Pedestrian Traffic Flow, In Proceeding of The 4th
Eastern Asia Society For Transportation Studies, Hanoi, Vietnam, 3(1), 11-25.
Tyler, N. (2002) Accessibility and the bus system: from concepts to practice. Thomas
Telford. London.
Zhao, F. and Li, M. (2005) Calibration of highway/ transit speed relationships for
improved transit network modelling in FSUTMS. Draft final report, Florida
Department of Transportation, Tallahassee, US.
Zografos, K. G., and Levinson, H. S. (1986). Passenger service times for a No-fare
bus system, Transportation Research Record 1051, 42-48.
Appendix – A
Diagram
Facts
Platform Outbound Inbound Loading area type = S Linear b
Analysis period Peak Off-peak # of loading areas (N) = ________(N ≤ 3)
Input
# boarding Passenger Pb = ____ Parameter, µ, for bus lost time (s) (Table 6-14)
Marginal boarding time tb = ____ Loading area 1 = _________
# alighting Passenger Loading area 2 = _________
Front door Pa = ____ Loading area 3 = _________
Rear door Pa’ = ____
Marginal alighting time ta = ____ Parameter, σ, for bus lost time (s) (Table 6-
14)
Door opening &closing time toc = ____ Loading area 1 = _________
Clearance time tc = ____ Loading area 2 = _________
Failure rate = ____ ~~ Z = ____ Loading area 3 = _________
Green time ratio g/c = ____
Coefficient of variation for Cv,p = ____ Mean lost time (s) (Table 6-15 for peak
passenger service time period & table 6-16 for off-peak period)
Cont…
Calculation steps
Capacity Calculation
Loading area LTn DTn tomp,n toml,n Bn En Bef,n
1 ____ ____ ____ ____ ____ ____ ____
Appendix – B
Bus capacity example application
Question: Let say we need to design a virtual busway station say Mater Hill Busway
Station with 3 loading areas. For capacity calculation, two methodologies are
available –
a) Transit Capacity & Quality of Service Manual (TCQSM) 2003 Method
b) Busway Station Loading Area Bus Capacity Model (BSLC) 2009 Method (this PhD)
Inputs:
1 Boarding passengers per bus Pb = 7
2 Marginal boarding time tb = 4s
3 Alight passenger per bus
Front door Pa = 0
Rear door Pa’ = 0
4 Marginal alighting time ta = 0
5 Door opening and closing time toc = 2s
6 Bus clearance time tc = 8s
7 Failure rate = 7.5%
8 Z value corresponding to the failure rate = 1.44
9 Co-efficient of variation Cv = 60%
10 g/c = 1
11 Effective loading area (TCQSM, 2003) Leff = 2.65
Solution:
i. Calculate bus dwell time using Equation 2.8 (on page 20)
td = 30s
ii. Calculate loading area capacity using Equation 2.19 (on page 38)
3600
iii. Calculate station capacity using Equation 2.21 (on page 40)
i. Calculate bus dwell time using revised dwell time equation, Equation 7.1 (on
page 121)
ii. Calculate operating margin for passenger service time using Equation 8.4 (on
page 132)
, ,
iii. Calculate operating margin for bus lost time using Equation 8.9 (on page 134)
⁄
,
iv. Calculate loading area bus capacity using Equation 8.10 (on page 134)
3600
, ,
v. Calculate effective loading area bus capacity using Equation 8.12 (on page
136)
,
vi. Calculate platform bus capacity using Equation 8.13 (on page 136)
Appendix – C
List of publications
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2010). Influence of
platform walking on BRT station bus dwell time estimation: Australian analysis.
Journal of Transportation Engineering (ASCE), Vol 136, No 12, pp 1173-1179.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2010) Modelling
Bus Lost Time: An Additional Parameter Influencing Bus Dwell Time and Station
Platform Capacity at a BRT Station Platform. In Proceedings 89th Annual Meeting
of Transportation Research board, Washington DC, United States.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Effects of
Fare Collection Policy on Operating Characteristics of a Brisbane Busway Station.
In Proceedings 32nd Australasian Transport Research Forum (ATRF), Auckland,
New Zealand.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Effect of
Passenger Crowding at a Busway Station on Dwell Time. In Proceedings 2nd
Infrastructure Theme Postgraduate Conference, Queensland University of
Technology, Brisbane, Australia.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Modelling the
Relationship Between Passenger demand and Bus Delays at Busway Station. In
Proceedings 88th Annual Meeting of Transportation Research board, Washington
DC, United States.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2008) Measuring
Bus Dwell Time and Platform Crowding at a Busway Station. In Proceedings 31st
Australasian Transport Research Forum (ATRF), Gold Coast, Australia.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2007) Operating
Characteristics and Performance of a Busway Transit Station. In Proceedings 30th
Australasian Transport Research Forum (ATRF), Melbourne, Australia.
• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis. Modelling Bus Lost
Time: An Additional Variable Influencing Bus Dwell Time at a BRT Station
Platform. Journal of Advance Transportation, Wiley InterScience. (Submitted on
19 July 2010).