Busway Platform Bus Capacity Analysis

Busway Platform Bus Capacity Analysis
Sumeet Kumar Jaiswal

B.E. (Civil), M. Tech (Transportation)
A thesis submitted for the degree of

Doctor of Philosophy
School of Urban Development
Faculty of Built Environment & Engineering
Queensland University of Technology
December 2010
Dedicated to my dear sister Deepti Muley
Keywords
BRT, busway, capacity, lost time, crowd, transit, public transport, interface, dwell
time.
Sumeet Jaiswal Page i

Abstract
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.
Line capacities of a busway are predominately dependent on bus capacity of its

stations. Despite new developments in BRT designs, the methodology of capacity
analysis is still based on traditional principles of kerbside bus stop on bus only lane
operations. Consequently, the tradition methodology lacks accounting for various
dimensions of busway station operation, such as passenger crowd, passenger
walking and bus lost time along the long busway station platform. This research has
developed a purpose made bus capacity analysis methodology for busway station
analysis. Extensive observations of kerbside bus stops and busway stations in
Brisbane, Australia were made and differences in their operation were studied. A
large scale data collection was conducted using the video recording technique at the
Mater Hill Busway Station on the South East Busway in Brisbane.
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
Sumeet Jaiswal Page iii

examples. Additionally, these examples illustrated the significance of the bus lost
time variable in determining station capacities.
Sumeet Jaiswal Page iv

Contents
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
Two 2Literature Review 9

2.1 Overview 9
2.2 Bus Rapid Transit System 9
2.2.1 BRT defined 9
2.2.2 Busway defined 11
2.3 Bus stop/ station classification 12
2.3.1 Simple stop 12
2.3.2 Enhanced stop 14
2.3.3 Dedicated station 14
2.3.4 Intermodal Terminal or Transit Centre 15
2.4 Role and impact of bus stop/ station 16
2.4.1 Bus dwell time 17
2.5 Busway station 23
2.5.1 Role of busway station 24
2.5.2 Passenger flow at a busway station 24
2.5.3 Platform crowd 26
Sumeet Jaiswal Page v

2.6 Crowd density and walking speed 28

2.6.1 Pedestrian speed-density-flow relationship 29
2.6.2 TCQSM method 32
2.7 Busway platform bus capacity 36
2.7.1 Development of Bus capacity model 37
2.7.2 Revised capacity model 38
2.7.3 TCQSM methodology of capacity calculation 39
2.8 Gaps in the knowledge 41
Three 3Research Problem Development 45

3.1 Overview 45
3.2 Difference in bus stop and busway station operation 45
3.2.1 The size 46
3.2.2 The demand 46
3.2.3 The passenger boarding process 47
3.3 Problem conceptualisation 48
3.4 Busway operation 49
3.5 Definition of terms 52
3.6 Conclusions 53
Four 4Data Collection and Processing 55

4.1 Overview 55
4.2 State of art in relevant data collection technique 55
4.3 Technique used for data collection in this study 58
4.4 Research methodology 59
4.4.1 Data collection methodology 59
4.4.2 Data extraction methodology 59
4.4.3 Data analysis methodology 60
4.5 Selection of study station 61
4.6 Characteristics of Mater Hill Busway Station 64
4.6.1 Passenger flow at station 66
4.6.2 Bus flow at station 67
4.7 Sequence of data collection 68
4.8 Data processing 71
4.9 Chapter close 71
Sumeet Jaiswal Page vi

Five 5Parameter Analysis and Evaluation 73
5.1 Overview 73
5.2 Measuring platform crowd 73
5.3 Passenger - bus Interface 73
5.3.1 Discussion of passenger - bus interface 78
5.3.2 Time - Space Diagram 79
5.4 Passenger behaviour while waiting 81
5.5 Bus lost time 84
5.6 Passenger - bus interaction 87
5.6.1 Effect of fare collection policy 90
Six 6Modelling Bus Lost Time 95

6.1 Overview 95
6.2 Bus lost time histogram 95
6.3 Probability distribution curve fitting 99
6.3.1 Assessing normality 103
6.3.2 Null hypothesis testing 104
6.4 Log-normal distribution curves for bus lost time 107
6.4.1 Log-normal probability distribution function curve 108
6.4.2 Log-normal cumulative distribution function curve 115
6.4.3 Descriptive characteristics of busway station bus lost time 115
Seven 7Busway Station Dwell Time Model 119

7.1 Overview 119
7.2 Model framework 119
7.3 Busway station bus dwell time model 121
7.4 Example application 122
7.5 Discussion 127
Eight 8Busway Loading Area Bus Capacity Model 129

8.1 Overview 129
8.2 Approach to busway loading area bus capacity model 129
8.2.1 Busway dwell time 130
8.2.2 Dwell time variability 130
Sumeet Jaiswal Page vii

8.2.3 Failure rate 131

8.2.4 Operating margin due to passenger service time variability 131
8.2.5 Lost time variability 132
8.3 Busway loading area bus capacity model 134
8.4 Effective bus capacity of loading area 135
8.5 Busway station platform bus capacity 136
8.6 Example application 136
8.7 Discussion 139
Nine 9Busway Station Efficiency Model 141

9.1 Overview 141
9.2 Loading area blocking 141
9.2.1 Existing approach 146
9.3 Approach to loading area efficiency factor calculation 147
9.4 Loading area efficiency factors for Mater Hill Busway Station 148
9.5 Discussion 149
Ten 10Conclusions 151

10.1 Overview 151
10.2 Summary of this thesis 151
10.3 Contributions of this research 153
10.4 Implications of this research 153
10.5 Conclusions 154
10.6 Recommendations for future work 155
11References 157
Appendix 12
A 13Busway Station Platform Bus Capacity Analysis Worksheet 163
B 14Bus Capacity example application 165
C 15List of Publications 169
Sumeet Jaiswal Page viii

List of Tables
Table 2.1 Pedestrian level of service on walkways 34
Table 2.2 Levels of service for queuing area 34
Table 2.3 Similar type of bus lanes and busway 37
Table 2.4 Failure rates and corresponding ‘z’ values 39
Table 2.5 Efficiency of multiple offline linear loading area at bus stops 41
Table 3.1 Boarding process at a bus stop and at a busway station 48
Table 4.1 Data collection techniques used in past studies 56
Table 4.2 Candidate busway station (outbound) platforms 63
Table 4.3 Fare collection policies at Mater Hill Busway Station 65
Table 4.4 Bus flow rate and passenger demand classification split 69
Table 4.5 Characteristics of analysis time 69
Table 5.1 Duration of passenger – bus interface during off-peak period 74
Table 5.2 Duration of passenger – bus interface during peak period 75
Table 5.3 Passenger boarding and alighting during evening peak period 85
Table 5.4 Bus lost times (LT) during off-peak periods 85
Table 5.5 Bus lost times (LT) during peak period 86
Table 5.6 Descriptive statistics 90
Table 5.7 Fare collection policies and observations at study station 91
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 6.1 Descriptive statistics of loading area 1 (Off-peak period) 100
Table 6.4 Descriptive statistics of loading area 1 (Peak period) 101
Sumeet Jaiswal Page ix

Table 6.7 Methods for testing normality 103
Table 6.8 Assessing normality for loading area 1 (Off-peak period) 105
Table 6.11 Assessing normality for loading area 1 (Peak period) 106
Table 6.14 Statistical parameters of bus lost time curves 115
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 7.1 Example demonstration 123
Table 8.1 Failure rates and corresponding ‘z’ values 131
Table 8.2 Example demonstration 137
Table 9.1 Efficiency factors provided by TCQSM 146
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
Table 9.4 Comparison of loading area efficiency results 149
Sumeet Jaiswal Page x

List of Figures
Figure 1.1 Developed framework for busway station platform bus capacity analysis 4
Figure 1.2 Structure of this thesis 6
Figure 2.1 BRT configurations and their passenger transit facilities 11
Figure 2.2 Lane configuration of a busway and its station (Brisbane, Australia) 12
Figure 2.3 A kerb side simple bus stop 13
Figure 2.4 An enhanced BRT stop 13
Figure 2.5 Mater Hill Busway Station 14
Figure 2.6 Transit centre with sawtooth arrangement of loading areas 15
Figure 2.7 Framework for pedestrian walking behaviour 29
Figure 2.8 Theoretical model of pedestrian flow in single channels 31
Figure 2.9 Empirical relations between travel and density of pedestrians 31
Figure 2.10 Walking speed variations as a function of age 32
Figure 2.11 Pedestrian speed on walkways 32
Figure 2.12 Illustration of walkway level of service 33
Figure 2.13 Illustration of queuing area level of service 35
Figure 2.14 Steps to calculate station bus capacity 36
Figure 2.15 Examples of loading area 41
Figure 2.16 Gap in busway station bus capacity estimation approach 42
Figure 3.1 Concentration of passenger crowding 47
Figure 3.2 Origin and destination of a trip segment at platform 48
Figure 3.3 Different levels of busway operation 49
Figure 3.4 A passenger – bus interface phase at a busway station 50
Figure 3.5 A bus lost time phase at a busway station 50
Figure 4.1 Camera positions at Mater Hill Busway Station (Outbound platform) 59
Sumeet Jaiswal Page xi

Figure 4.2 Processing of data – from collection stage to analysis stage 61
Figure 4.3 Brisbane’s South East Busway route map 62
Figure 4.4 Configuration of Mater Hill Busway Station 65
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 5.1 Variation in passenger – bus interface 76
Figure 5.2 Variation in passenger – bus interface during off-peak period 78
Figure 5.3 Passenger – bus interface duration and its dependent variables 79
Figure 5.4 Time – space diagram 81
Figure 5.5 Distance to loading areas from the waiting area on the busway platform
(Off-peak) 82
Figure 5.6 Effect of loading area on bus lost time 83
Figure 5.7 Variation in bus lost times over platform crowd by loading area 87
Figure 6.1 Off-peak period bus lost time histogram 96
Figure 6.2 Peak period bus lost time histogram 97
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
Sumeet Jaiswal Page xii

Figure 7.5 Effect of bus lost time on dwell time at loading area 2 (Off-peak period) 126
Figure 8.1 Log-normal density curve 133
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 8.4 Estimated busway station bus capacities 139
Figure 9.1 Trajectory of bus processing at the Mater Hill Busway Station
(Outbound platform) 143
Figure 9.2 Inter-loading area blocking scenarios and associated numbers of

effective loading areas. 145
Sumeet Jaiswal Page xiii

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 _____________________
Sumeet Jaiswal Page xiv

Acknowledgements
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.
I also acknowledge the financial support and assistance provided to me by the

research portfolio and school of urban development.
I am grateful to my colleagues at Bitzios Consulting who always have provided me a

comfortable working environment.
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.
Sumeet Jaiswal Page xv

Introduction
Chapter One
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.
Sumeet Jaiswal Page 1

1.3 Research motivation

The South East Busway in Brisbane, Australia has experienced an exceptional
growth in patronage since its opening in 2001. In the first 6 months of its operation,
the number of passengers grew by 40 percent or by more than 450,000, giving a
daily average patronage of 58,000. Over the first 3.5 years there has been an 88
percent increase in patronage for the busway (Hensher, 2007). In a report published
by the Translink Transit Authority (2009), the South East Busway between Brisbane
Central Business District (CBD) and Eight Mile Plains carried more than 150,000
passengers per day, with sections of busway carrying 18,000 passengers per hour
during the peak. Further to this, there is strong evidence that background patronage
will continue to grow.
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.

Introduction
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.
1.4 Research hypothesis

Hypothesis:
“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.
1.5 Research aim and objectives

The driving aim of this research is the development of a reliable and robust bus
capacity analysis methodology for a busway station. In order to achieve this aim,
additional objective are defined -
1. Understand operation of a busway station and study the passenger movement
on the platform.
2. Identify and investigate the parameters affecting bus dwell time at a busway
station platform.
3. Develop a robust busway dwell time model.
4. Assess the impact of busway dwell time on busway platform bus capacity.
5. Assess the impact of bus – bus interference on platform bus capacity
1.6 Scope of this research

As stated earlier, the motivation behind this research is to study the impact of
platform crowd on bus dwell time at a busway station. The crowding happens only
due to passengers on the platform waiting for arrival of their desired buses, so that
they can board the service. The alighting passengers, on the other hand, quickly
move out of the platform and hence, do not form a part of the standing crowd.
Furthermore, the dwell time of a bus serving only alighting passengers will have

minimal influence on platform crowd. Hence, the data collection for this study was
limited to a busway platform where boarding passengers were dominant.
1.7 Relevance of this research

Lack of a dedicated analysis methodology for a busway station could lead to
inaccurate design of the system, such as over estimation of station platform bus
capacity. Incomplete knowledge of factors influencing the busway station operation
can potentially lessen the advantages of a policy improvement, such as smartcard
fare system. This research, by studying a working busway station on Brisbane’s
busway has developed a methodology which accounts for the various dimensions of
busway station operation.
Various dimensions of busway station operation, such as passenger – bus interface

and bus lost time due to passenger walking along the long platform, and bus – bus
interference due to multiple linear loading areas, have been identified and defined.
Based on these dimensions, a framework for busway station bus capacity analysis
methodology has been developed. Figure 1.1 shows the framework of this
methodology. The stepwise procedure for this methodology is given in Appendix A.
Estimate bus lost time

Chapter 6
Estimate busway dwell time

Chapter 7
Estimate busway loading

area capacity
Chapter 8
Estimate busway loading

area efficiency
Chapter 9
Estimate busway station

platform capacity
Chapter 8
Figure 1.1: Developed framework for busway station platform bus capacity analysis

Introduction
1.8 Thesis outline

Figure 1.2 shows the structure of this thesis. This thesis can be divided into three
parts, namely concerning the development of the research problem, setting the
approach towards solutions, and development of solutions.
Development of research problems are discussed in Chapters 1 to 3. This is followed

by the approach used for solving these problems, Chapters 4 and 5. Chapters 6 to 9
develop the solutions for these problems. Chapter 10 syntheses all the analysis
results and findings, and concludes this thesis.
A brief outlines of each chapter is given below –
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 3 develops the research problems and identifies the parameters

influencing busway station platform operation. It sets the directions for
data collection.
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 7 describes development of busway dwell time model
Chapter 8 describes development of busway loading area bus capacity model

Chapter 9 describes development of busway loading area efficiency model
Chapter 10 concludes this thesis and identifies the contribution and innovations of
this research. It also provides guidance for further research.
Development of research problem

Chapter 1
Introduction
Chapter 2
Literature review
Chapter 3
Research problem development
Approach for solution

Chapter 4
Data collection and processing
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

Introduction
1.9 Publications from this research

This research has lead to publication of one refereed journal paper and six refereed
conference papers. Additionally, one more paper is under review for journal
publications. The complete list of papers is given in Appendix C.

Literature Review
Chapter Two
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.
2.2 Bus Rapid Transit System
2.2.1 BRT defined

The Federal Transit Administration (FTA) of the United States of America defines
BRT as,
“A rapid mode of transportation that can provide the quality of

rail transit and the flexibility of buses.” (Levinson et.al., 2002).
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 -
“BRT is a flexible, rubber-tired form of rapid transit that

combines stations, vehicles, services, running ways, and ITS
elements into an integrated system with a strong identity.”
(TCRP, 2003)
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.
Photo by: Jaiswal (2009)

a. Exclusive bus lane with bus stop

Literature Review

b. Dedicated busway with busway station
Figure 2.1: BRT configurations and their passenger transit facilities
2.2.2 Busway defined

A BRT system is a service that operates on bus lanes or other transitways in order to
achieve high speed of transit. A busway is a special roadway infrastructure designed
for exclusive use of buses (FTA, 2008). A busway is different to the other BRT
treatments, such as bus lanes and bus priority schemes which are more limited in
their scope. A busway will usually have its own right of way, physically separated
from general traffic. With the greatest level of separation from general traffic and
road intersections, a very high speed bus transit is possible on a busway. For
example, posted speed limits are 90 km/h on certain sections of Brisbane’s South
East busway. Similarly, buses on Canada’s Ottawa transitways have a speed limit of
70 to 90 km/h between stations (Wikipedia, 2009a).
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.

Figure 2.2: Lane configuration of a busway and its station (Brisbane, Australia)
2.3 Bus stop/ station classification

There are four types of bus station (FTA, 2004). Depending on their function in the
system they vary in size and amenities. Their brief descriptions are given below. As
noted by the US Federal Transportation Authority, transit stations and their amenities
provide comfort to the passenger and therefore also help in attracting more
patronage.
2.3.1 Simple stop

These stops are simplest of all four types. Such a stop consists of a shelter and
printed information display. Figure 2.3 shows a simple bus stop. These stops
principally cater service to a relatively low level of passenger demand and bus
routes.

Literature Review

Figure 2.3: A kerb side simple bus stop

Figure 2.4: An enhanced BRT stop

2.3.2 Enhanced stop

An enhanced stop is the upgraded version of simple stop with enhanced shelter,
usually designed for the BRT line to distinguish it from other stops. These stops
usually integrate special amenities such as glass wall, water outlet, trash bin and/ or
provision of pay phone. Figure 2.4 shows an enhanced stop.
2.3.3 Dedicated station

Dedicated stations are the high end stations designed specially for a BRT system.
Designs of these stations usually have multiple loading areas for several buses to
stop simultaneously. Their design includes all-weather shelter for passengers,
lighting, level passenger platform boarding and alighting for speedy movement and
high quality passenger information facilities. Amenity wise they may have benches,
water outlet, pay phone and ticket vending machine and food kiosks. Figure 2.5
shows a dedicated busway station on a Brisbane busway corridor.

Figure 2.5: Mater Hill Busway Station

Literature Review
2.3.4 Intermodal terminal or transit centre

An intermodal terminal or transit centre is designed to facilitate the transfer activities
of passengers between different modes, such as rail or ferry and/ or terminus of bus
services. These facilities incorporate a host of amenities such as waiting areas,
benches, water outlets, lighting, pay phones and so on. Usually their platforms have
level boarding and alighting design to facilitate comfortable passenger movement
between bus and platform. They have the provision of more than one loading area,
commonly in sawtooth arrangement to allow smooth and independent movements of
buses in and out of their loading area. Figure 2.6 shows a transit centre provided in a
close proximity to a shopping center.

Figure 2.6: Transit centre with sawtooth arrangement of loading areas

2.4 Role and impact of bus stop/ station

Transit stations are gateways to the network. In fact, these stops and stations are the
only points where customers access the transit service. However, scheduling a bus
to observe a stop or station inflates its journey time due to the delay caused by the
stop. The time spent by a bus at a stop to serve its passengers is commonly known
as dwell time. However, in reality the dwell time is not the only component of bus
delay due to stop (Changshan and Murray, 2005). The stop delay can be divided into
four major components, comprising of delay associated with deceleration and
acceleration, delay due to door opening and closing, delay associated with bus
clearance time, and delay due to passenger boarding and alighting.
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)

Literature Review
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.
2.4.1 Bus dwell time

The Transit Capacity and Quality of Service Manual (TRB, 2003) defines bus dwell
time as the amount of time a bus spends at a stop or at a station to serve boarding

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.
2.4.1.1 Passenger boarding and alighting
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).
5.0 2.75 Equation 2.2
4.0 1.5 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.
5.0 1.2 In N ≤ 23 Equation 2.4

Literature Review
1.2 N ≥ 24 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,
ts = dwell time at station
to = station standing time
b ' and a ' = number of boarding and alighting passengers respectively

τ b and τ a = respective boarding and alighting per passenger
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,
td = Average dwell times (s)
Pa = Alighting passengers per bus through the busiest door (p)
ta = Alighting passenger service time (s/p)
Pb = Boarding passengers per bus through the busiest door (p)
tb = Boarding passenger service time (s/p)
t oc = Door opening and closing times (s)
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.
U = f (TOTALOFF , ONBOARD , TIMEPOINT , AM , PM ) Equation 2.9
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

Literature Review
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.
U = 0.0363TOTALOFF − 0.0213ONBOARD − 0.8389TIMEPOINT + 0.4098 AM + 0.6777 PM

Equation 2.10
eU
P(Y = 1) = Equation 2.11
1 + eU
2.4.1.2 Fare payment method
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.
8.07 . (R2 = 0.67) Equation 2.12
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.
2.4.1.3 Vehicle type and size
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.
2.4.1.4 In-vehicle circulation
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

Literature Review
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.
1.56 0.16 0.09 Equation 2.14
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.
2.5 Busway station

Busway stations can be classified as dedicated stations and can be designed to
facilitate several buses to stop simultaneously. A busway station has linear off-line
loading areas, to allow buses to overtake the stopping bus (FTA, 2004; Rathwell and
Schijns, 2002). A high level of amenities for passenger comfort and convenience is
generally provided. The station can be equipped with map and real time information
system, security and lighting arrangement, benches, pay phone and water outlet.
However, the station design concept can also be driven by local conditions. For
example, the transitway stations in Ottawa, Canada are provided with heated,
enclosed on-platform waiting area due to the harsh winters of the region (Rathwell

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.
2.5.1 Role of busway station

The significance of a busway station or, every transit station, substantially increases
when we see transport as a service. A transport service is different from other forms
of services in a sense that it cannot be stored. That is, it must be consumed when
the service is generated or the service must only be generated when the demand
arises. Practically, achieving such equilibrium for mass public transport system is
very hard, if not impossible. Since busway stations are provided with large spacing in
between stations and together with the fact the busway are the corridors with high
passenger demands, a busway station attracts larger volumes of passengers than a
kerb side bus stop. And therefore, in addition to facilitate a smooth interface between
the passenger and bus, another role which the busway stations have to perform is to
accumulate the demand until the service is provided to them.
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.
2.5.2 Passenger flow at a busway station

Although the platform length of a busway station is greater than for a kerbside bus
stop, it creates a difficult situation to organise the passenger flow. The characteristic
of pedestrian flow at the busway station is rather complicated due to the fact that
there is no channalisation opportunity available to the passengers. During the peak
hours, the high but near continuous inward passenger flow at the platform entrance
points suffers a sudden breakdown in flow after entering the platform area. This is
because the outward flow of passenger is not on the basis of ‘as you come, so you
go’. While some passengers might receive their desired bus promptly, some have to

Literature Review
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.

2.5.3 Platform crowd

Passenger crowding at the station platform can be predominantly observed at transit
stations located in or near the Central Business District (CBD) of the city, because of
higher activity. Such activity generates high number of trips from or to other parts of
the city. Crowds can be made up of either a majority of waiting passengers – who
want to board a transit service – or a majority of alighting passengers from buses
that have arrived at the station platform. It is worth noting, at this point, that crowding
because of waiting passengers is likely to have greater impact on the platform bus
capacity than that of crowding because of the alighting passengers, though both
could lead to a reduction in corridor capacity depending on specific circumstances.
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.

Literature Review
Impact on service provider

Cost of transit operation is a major concern for service providers (IAPT, 2007;
Vuchic, 2005). Journey speed, representing travel time, is one of the major attributes
which represents operational effectiveness of a transit network. Under the busway
scenario the delay to the bus at stations is a major factor in reducing its journey
speed. Delays, such as excess dwell times, excess clearance time, time spent in
queue, can be correlated with passenger crowding at the platform area of the station.
Moreover, at each station buses accumulate further delays to their running time
causing a reduction in the available layover time and /or vehicle productivity at the
terminal.
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.
Equation 2.15 is a mathematical formula suggested by Tyler (2002) to determine the

fleet size required to provide a schedule service.
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.
2.6 Crowd density and walking speed

Understanding pedestrian walking behaviour is a complex task. Some researchers
have tried to explain this behaviour with the analogy of ants (Nishinari et.al., 2006),
while some have used a lattice-gas model to study the pedestrian walking (Jiang and
Wu, 2005). Nevertheless, the majority of studies in the field of pedestrian flow have
pointed out that pedestrian walking speed reduces with increased density. The
increase in density decreases the human-usable space causing interference to
others, resulting in a decrease in walking speed (Osaragi, 2004). Further, it was
found that the pedestrian walking as a group of three or more tend to be slower than
when walking individually or in couples (Tarawneh, 2001). Beside this, if the
pedestrian walking is in opposite directions or crossing the path of each other, there
could be further reduction in their walking speeds (Smith, 1995).
Pedestrian movement is individualised and dynamic. For a given condition the

walking behaviour of different pedestrian can vary. Moreover, a pedestrian may
behave differently every time he/she faces a situation even if the situation is
constant. This may be because humans have the tendency to learn from their past
experience and make necessary changes in their action in rapid successions.
Pedestrians are capable of changing speed more quickly when gaps arise and can
accelerate to full speed from a standstill or decelerate to avoid collision (Blue and
Adler, 2001). Seyfried et.al. (2005) in their study on pedestrian flow reinvestigated
the fundamental relationship between the density and pedestrian speed. Analysing
the single-file movement of pedestrians they observed a linear relationship between
the speed and the inverse of the density. Pedestrians adapt their speed not only to
the person immediately in front, but to the situation further ahead, the study
concluded. Daamen et.al. (2005) described pedestrian flow in congestion using a
fundamental diagram – the relationship between speed, flow and density. Form their
experimental study, they found that on the boundary of the congested region,
pedestrians may walk in nearly free flow conditions. However, inside the congested
regions the low speeds with high density were observed.

Literature Review
The reduction in speed can be caused by close proximity of other pedestrians.

Pedestrians may like to keep certain distances from others. But keeping such a
distance in an environment of a busway platform is difficult. Often after seeing their
bus arriving at the platform, pedestrians walk hastily without anticipating the possible
movement of others, by squeezing through the crowd. However, not all pedestrians
prefer to do so, usually the older pedestrians want to keep some constant distance
and avoid physical touch with others (Gerin-Lajoie and Richards, 2006).
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.
Pedestrian walking behaviour
Unconstrained Constrained
Keep Towards Free flow Collision Leader

direction destination acc/dec avoidance follower
Source: Antonini and Bierlaire, 2005
Figure 2.7: Framework for pedestrian walking behaviour
2.6.1 Pedestrian speed-density-flow relationship

The fundamental of pedestrian flow theory states that the flow (or volume) is a
product of speed and density and their inter-relationship can be approximated by a
parabolic curve which is similar to motor vehicle flow (Khisty, 1990).

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).

Literature Review
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.

Walking Speed, V (m/min)
Age
Source: Smith, 1995
Figure 2.10: Walking speed variations as a function of age
2.6.2 TCQSM method

TCQSM (TRB, 2003 based on Fruin, 1987) highlighted pedestrian density as the
most significant factor in influencing pedestrian walking speed and represented the
relationship between them graphically as in Figure 2.11.
Walking speed (m/min)
Pedestrian space (m2/p)

Source: TRB, 2003
Figure 2.11: Pedestrian speed on walkways

Literature Review
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

Table 2.1: Pedestrian level of service on walkways
Expected Flows and Speeds

Pedestrian Avg. Speed, Flow per unit width, v
LOS v/c
Space (m2/p) S (m/min) (p/m/min)
A ≥ 3.3 79 0-23 0.0-0.3
B 2.3-3.3 76 23-33 0.3-0.4
C 1.4-2.3 73 33-49 0.4-0.6
D 0.9-1.4 69 49-66 0.6-0.8
E 0.5-0.9 46 66-82 0.8-1.0
F < 0.5 < 46 variable variable
Source: TRB, 2003
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.
Table 2.2: Levels of service for queuing area
Average Pedestrian Area Average Inter-Person Spacing

LOS (m2/p) (m)
A ≥ 1.2 ≥ 1.2
B 0.9-1.2 1.1-1.2
C 0.7-0.9 0.9-1.1
D 0.3-0.7 0.6-0.9
E 0.2-0.3 < 0.6
F < 0.2 Variable
Source: TRB, 2003

Literature Review
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.

2.7 Busway platform bus capacity

Various BRT line bus capacities have been reported within a range of 200 bus/h to
450 bus/h, translating to a range of 10,000 p/h to 25,000 p/h. However, BRT stations
themselves might not be able to handle such high bus and passenger throughputs,
essentially becoming the line constrictions. Further, when a BRT station serves more
routes, passenger handling on the platform and therefore the passenger/bus
interactions become more complex, which may affect the line capacity. Design of a
Busway (BRT) system, therefore, involves the estimation of station bus throughput
capacity in order to estimate bus line capacity.
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).
Boarding and Boarding and

alighting load alighting service time
Bus dwell time
Loading area capacity Loading area efficiency
Station capacity
Source: Based on TRB, 2003

Figure 2.14: Steps to calculate station bus capacity

Literature Review
2.7.1 Development of bus capacity model

A comprehensive documentation of methodologies for at-grade busway bus capacity
estimation was presented by Shen et.al. in 1998. The report discussed the principle
behind the adaptation of bus lane capacity methodology from busway capacity
estimation. The different bus lane configuration was co-related with different busway
configurations. Table 2.3 shows different types of bus lanes configurations and
corresponding busway facility. Type 1 bus lanes are similar to a busway with no
overtaking facility, where both cannot pass other buses servicing in the station area.
A type 2 bus lane is similar to a busway with overtaking facility, where buses can
pass other buses servicing at station area. A type 3 bus lane has no similar busways
as busways consist of only two lanes, one in each direction, and buses on busway
are not allowed to use the opposite direction to overtake other buses.
Table 2.3: Similar type of bus lanes and busway
Type Bus lane configuration Similar Busway

1 No use of adjacent lane for buses to pass Busway with no
other buses, right-turn queue and other bus overtaking facility
lane obstructions
2 Partial use of adjacent lane for buses to Busway with overtaking
pass other buses, right-turn queue and facility
other bus lane obstructions
3 Full use of adjacent lane for buses to pass N/A
other buses, right-turn queue and other bus
lane obstructions
Source: Shen et.al, 1998
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.
⁄ 3600 Equation 2.18

⁄

Where,
= Buses per hour per channel per berth (loading area)

= Bus dwell time at stop (s)
= Clearance time (s)
= Reductive factor to account for variations in dwell times and arrival
= Number of effective berths (loading areas)
= Effective green time per cycle (s)
= Cycle length (s)
2.7.2 Revised capacity model

The initial bus capacity model presented in HCM was revised to better incorporate
dwell time variability of a bus stop (TCRP, 1997). A parameter called coefficient of
variation, , was introduced to account for the fluctuation in bus dwell time. The
fluctuation may arise because of the non-uniform passenger demand for different
buses and different routes. Another parameter Z corresponding to the failure rate
was incorporated with the equation. The failure rate sets how often a bus should
arrive at a stop only to find all loading areas occupied, and is assumed to be
normally distributed. Additional parameters, coefficient of variable and failure rate,
were used in combination of bus dwell time and are termed as operating margin.
This revised capacity model suggests that the (average) dwell time at the loading
area is the fundamental in determining its capacity.
3600 ⁄
Equation 2.19
⁄
Where,
Bl = Loading area bus capacity (bus/h)
3600 = Number of seconds in an hour

g C = Green time ratio (the ratio of effective green time to total traffic signal
cycle length; 1.0 for unsignalised streets and bus facilities like busway
without adjacent signal control)
tc = Clearance time (s)

Literature Review
= Standard normal variate corresponding to a desired failure rate

= Coefficient of variation of dwell time
Table 2.4 provide the value of ‘z’ corresponding to desired failure fate.
Table 2.4: Failure rates and corresponding ‘z’ values
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.
2.7.3 TCQSM methodology of capacity calculation

Transit Capacity and Quality of Service Manual (TRB, 2003) is one of the most
comprehensive documents created out of a range of research studies dealing with
transit design and service.

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

Literature Review
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
Loading Cumulative No. of

Efficiency
area effective loading
%
number areas
1 100 1.00
2 85 1.85
3 80 2.65
4 65 3.25
5 50 3.75
Source: TRB, 2003
(a) Off-line loading area (b) On-line loading area

Figure 2.15: Examples of loading area
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).
2.8 Gaps in the knowledge

Single most important finding of this literature search is that there is no methodology
specifically developed for busway capacity estimation. The present methodology
given by TCQSM (TRB, 2003) (Section 2.7.3) was designed from the principles of

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.
Boarding and Boarding and Busway station operation

alighting load alighting service time characteristics
Bus dwell time
Loading area capacity Loading area efficiency
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.

Literature Review
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.
General findings for this review include:
Study investigating the effect of busway platform crowding on passenger

service time vis-à-vis on vehicle dwell time has not been found, though
relevant studies related to airport terminals have been published.
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
Chapter Three
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.
3.2 Difference in bus stop and busway station operation

Although a busway station and a standard kerbside bus stop facilitate passengers
boarding and alighting, there are considerable differences between these facilities.
The most visible difference is in their size. Busway stations are longer and with
multiple loading areas compared to shorter suburban bus stops usually with single or
sometime two loading areas.

3.2.1 The size

At a typical bus stop, typically one marked loading area is provided; hence
passengers know the stopping position of the bus entry door with higher certainty
and are generally positioned within half a bus length of the front door. Waiting
passengers, therefore often align themselves accordingly well before the bus comes
to a stop. Hence the passenger walking to bus entry door plays a minimal role in the
bus dwell time at the bus stop. The dwell time of a bus at a bus stop can therefore be
generally been defined as a function of passenger demand and service time per
passenger, separately for alighting and boarding. However, at least three loading
areas are provided at busway stations in Brisbane. This creates uncertainty in
passengers’ minds about the loading area to be occupied by the desired bus and
hence requires passengers walk to the bus entry door after the bus comes to a halt
at the loading area. This therefore makes passenger walking more significant in the
bus passenger servicing process.
3.2.2 The demand

Another important factor that differentiates a busway station from a bus stop is the
number of bus routes servicing the station. In Brisbane, a typical bus stop serves
between one and five bus routes. However, for example the Mater Hill Busway
station, being a mainline station, serves over 40 separate routes including a number
of Bus Upgrade Zone (BUZ) high frequency spine services. Since the number of
routes at a typical bus stop is far less, the passenger route groups are far less
diverse. Hence at a bus stop with more uniformity in passengers’ directional
behaviour the passengers are of a common route group. However, at a busway
platform the number of bus routes can be very high. Therefore the passenger route
groups at the busway station can be much higher, leading to crowding. Such
platform crowd density at a busway station platform acts as an obstruction to the
passenger’s walking path (TRB, 2003) and also obstructs the passenger’s line of
sight, resulting in a longer passenger – bus interface. Figure 3.1 compares observed
passenger densities at a bus stop and a busway station.

a) Bus stop b) Busway station platform

Source: Jaiswal (2009)
Figure 3.1: Concentration of passenger crowding
3.2.3 The passenger boarding process

Observing passenger boarding activities at busway stations and bus stops, the
passenger servicing process can be divided into four distinct phases. The first phase
is the initial reaction of hailing when the passenger first sees their desired bus. The
second phase consists of walking to the bus entry door. The third phase consists of
queuing at the entry door and fourth phase is boarding the bus.
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.

Table 3.1: Boarding process at a bus stop and at a busway station
Action Bus stop Busway Station

1 Initial reaction to bus / hailing Yes Yes
2 Walking Minimal Can be substantial
3 Queuing Can occur Can occur
4 Boarding Yes Yes
3.3 Problem conceptualisation

The difference between boarding a bus at a bus stop and boarding a bus at a
busway station is, therefore, the requirement of a substantial amount of walking on
the busway platform to reach the bus entry door. The nature of this walking can be
illustrated by considering the typical path of passengers at the busway platform from
when they arrive until when they board their desired bus (Figure 3.2).
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.
3.4 Busway operation

The sequential impact of the passenger boarding process across the busway
operation can be considered at four levels in a hierarchy as shown in the Figure 3.3.
The figure shows the flow of effects from platform level to network level.
Level Process Effect
LINE Station to station interface Bus Bunching
Bus Clearance /
STATION Bus to bus interface Bus queuing
VEHICLE Passenger – Bus interaction Boarding & Alighting
Passengers walking
PLATFORM Passenger – Bus interface on platform
Figure 3.3: Different levels of busway operation
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).

Figure 3.4: A passenger – bus interface phase at a busway station

Figure 3.5: A bus lost time phase at a busway station

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 station level: With a busway station platform designed to have an

arrangement of off-line linear loading areas, buses stop at the platform to service
passengers one after the other in close proximity. In some cases this may lead to
insufficient space for buses to leave a loading area after boarding and alighting
passengers have been processed and doors closed. This results in an increase in
clearance time for buses under these circumstances. Such increase in clearance
times coupled with increased dwell times reduces station bus capacity, and affects
queuing of incoming buses. Typical busway line design includes an off-line loading
area lane and passing lane in each direction at stations. When bus queues build up
at a station, the through lane upstream will become blocked. Any through buses,
which are not scheduled to service the station, can hence be delayed along with
buses servicing the station. This then influences bus route travel times.

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.
3.5 Definition of terms

Three new parameters were identified which explains the distinct operation of
busway and its stations. These parameters are technically defined below –
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.
Passenger – bus interaction: The ‘passenger – bus interaction’ (IA) is defined as

the phase where the first and subsequent passengers and bus driver are involved in
a state when they both interact and perform their respective activities based on each
other’s position or action.

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
Chapter Four
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.
4.2 State of art in relevant data collection technique

This research needed to examine all variables which may explain the effects of
platform crowding, influence of walking to bus entry door, and effects of multiple
loading areas, on delay experienced by buses. Therefore the data required was
multi-dimensional: bus flow and passenger flow, both in time and space. With the
aim of developing an appropriate data collection and processing technique, a review
of techniques used in the past studies related to pedestrian flow and bus stop
operation was performed. Table 4.1 gives the overview of collection techniques used
in past studies.

Table 4.1: Data collection techniques used in past studies

Data Facility Technique used Author
Boarding and Bus stop Manual count Guenthner and Hamat,
alighting 1988
Bus stop Manual count Kim (2004)
Bus route Automatic Rajbandari (2003)
passenger counter
Pedestrian Railway station, Video recording / Davies et.al., 1995

Crowd London, UK image processing
Walkway Video recording / Kilambi et.al, 2008
image processing
Walkway Video recording / Kong et.al. 1995
Neural network
Place of worship Video recording / Johansson et.al., 2008
image processing
Pedestrian Walkway Video recording / Antonini and Bierlaire,

walking manual processing 2005
Road crossing Video recording / Antonini and Bierlaire,
manual processing 2005
Signalised Time-lapse Lam et.al, 2002
crossway photography
Walkway Video recording / Teknomo et.al., 2001
image processing
Pedestrian flow Narrow Lab experiment Seyfried et.al., 2007

behaviour bottleneck
Narrow Lab experiment Daamen (2004)
bottleneck
Four-directional Lab experiment Daamen (2004)
crossing flow
The data collection techniques in Table 4.1 are now explained.
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.
Time-lapse photography: Rime lapse photography is where a camera, typically in

a fixed position, automatically records a sequence or series of photos with a set time
interval between each image. Individual images may then be analysed in the
laboratory. This technique was first used to investigate the gap acceptance
behaviours of pedestrians (Dipletro and King, 1970). Lam et.al. (2002) used time
lapse technique to study bi-directional pedestrian flow characteristics at a signalized
crosswalk facility in Hong Kong.
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.
Experimental: In real life situations pedestrian flow characteristics may be

influenced by factors which are beyond the investigator’s control. To isolate the
effects of such factors on pedestrian walking behaviours, Daamen (2004) carried out
walking experiments in a controlled environment to study uni-directional flow, bi-
directional flow, crossing flow and bottlenecks. The experiment was recorded using

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.
4.3 Technique used for data collection in this study

The aim of this observation is to gather evidence of how bus dwell times at a busway
platform are influenced due to platform crowding and associated factors. It is
therefore important to record the attributes explaining the approach of the passenger
to the bus entry door. It is equally important to record the passenger flow, in order to
relate it to crowd formation and identify the location of crowding at the platform.
Observations related to the buses, such as service time at the platform, door
opening and closing time, and so on are also required to establish the density –
dwell time relation.
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)
4.4 Research methodology

The methodology for this research was divided into three parts: data collection, data
extraction and data analysis. Figure 4.2 illustrates connections between these
methodologies
4.4.1 Data collection methodology

The following methodology was used to collect the data.
1. Identify a suitable busway station platform.

2. Establish camera angles to achieve the best possible view of passenger
movements on the platform.
3. Conduct video recordings.
4.4.2 Data extraction methodology

The purpose of data collection via video recording was to capture all the activities
occurring at the busway station platform area. These recordings were therefore
manually analysed in the laboratory to extract the key attributes explaining the
passenger – bus interface phenomena. The data extraction methodology was
designed to mine the video footage for bus attributes (i.e. bus dwell time, bus
queuing time and bus clearance time) and passenger attributes (i.e walking time,

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 –
1. Select a passenger on the platform and tag it (say xi).

2. Record the time of passenger xi’s reaction to desired bus.
3. If passenger xi needed to queue at the bus entry door, due to the boarding
passengers in front, or any passenger alighting from the bus, then record the
queue entry time for that passenger.
4. Record the time when passenger xi boarded the bus.
5. Record the number of passengers on the platform who were crossed or
passed by that passenger xi
6. Record the total number of passenger present on the platform during this
time.
4.4.3 Data analysis methodology

1. Analyse bus dwell times with respect to the loading area.
2. Analyse passenger walking times with respect to the loading area.
3. Analyse passenger – bus interface and its effect on bus dwell time.
4. Develop a bus lost time profile for platform.

Data collection Data extraction Data analysis

On-site In-lab In-lab
Bus data Dwell time
Video recording Passenger Walking Time
Passenger data Platform Bus Capacity
Figure 4.2: Processing of data – from collection stage to analysis stage
4.5 Selection of study station

For the purpose of this research the busway station platform should meet following
desired criteria:
1. Presence of crowd at platform.

2. Variation in passenger demand thought out the day.
3. Dominance of boarding passengers during at least one of the peak period.
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.

Cultural Centre Station Queen Street Station

(0.67 km) (0.00 km)
South Bank Station

(1.67 km)
Mater Hill Station

(2.47 km)
Woolloongabba Station
Buranda Station (3.17 km)
(4.37 km)
Greenslopes Station
(5.97 km)
Holland Park West Station

(8.57 km)
Griffith University Station

(10.77 km)
Upper Mt Gravatt Station

(13.38 km)
Note - Number in bracket shows the
Eight Mile Plains Station distance of the respective station
(16.00 km) from Queen Street station.
Figure 4.3: Brisbane’s South East Busway route map
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.

Table 4.2: Candidate busway station (outbound) platforms
Culture Centre South Bank Mater Hill
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.

4.6 Characteristics of Mater Hill Busway Station

From Brisbane city, Mater Hill station is the fourth station on the 16 km long South
East busway corridor. Mater Hill Busway station now has three signed and striped
loading areas on its outbound platform as shown in Figure 4.4. Infrequently some
bus operators pull up very close to the dwelling bus ahead of them, thereby creating
a transient fourth loading area. This was much more prevalent prior to the signing
and striping of three loading areas, which occurred midway through this overall
project. This station is located inwards of the confluence of four common-lines; the
mainline South East Busway, the Woolloongabba spur common-line, the Pacific
Motorway access ramps common-line, and the Annerley Road common-line. Hence,
there is a considerable level of passenger demand for this station.
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
Figure 4.4: Configuration of Mater Hill Busway Station
Table 4.3: Fare collection policies at Mater Hill Busway Station
Month/ Year March 2007 March 2008 April 2009
On board ticket On board ticket

purchase purchase
Pre-paid ticket Pre-paid ticket Pre-paid ticket
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)
Pre-paid platform policy
Source: TransLink

4.6.1 Passenger flow at station

Mater Hill Busway station users include school and university students, hospital
employees, patients and visitors, and employees of surrounding businesses. The
peak for inbound platform occurs in morning time and mostly comprise of alighting
passengers (Figure 4.5). The alighting passengers quickly move out of the platform
and hence inbound platform experiences no significant crowding. The outbound
platform has its peak in evening time (Figure 4.6). The outbound platform is more
dominated by boarding passengers. It experiences two non-continuous evening peak
periods (Lee, 2004) with in the peak period. The first afternoon peak occurs following
release of students from two adjacent schools, and the second, evening peak occurs
when hospital, university and other workers commute home.
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
Figure 4.6: Boarding and alighting at outbound platform of Mater Hill Busway Station
4.6.2 Bus flow at station

Figure 4.7 shows the scheduled bus flow rate through the study station. This does
not include unscheduled buses such as blank run to depot or terminus (usually in the
counter peak direction) and some express bus services which do not observe the
study station (although much services normally bypass this section of busway).
Buses servicing the study station can be divided in four groups in accordance to their
floorplan –
a. High floor rigid buses

b. Low floor rigid buses
c. High floor articulated buses
d. Low flow articulated buses
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
Figure 4.7: Number of buses servicing outbound platform of Mater Hill Busway Station
4.7 Sequence of data collection

Video footage of the outbound platform of Mater Hill Busway Station was captured at
a similar time of year in three successive years; March 2007, March 2008 and April
2009. This was to ensure that the data collection was free from noise due to school
or university vacation periods, public holidays etc. Each video data set was captured
on a Wednesday, being a typical midweek day. These data collection activities were
carried out with the assistance of the Translink Transit Authority’s Busway
Operations Centre in Brisbane. The passengers on the platform were unaffected by
the video data collection as permanent security cameras were used. These cameras,
mounted on the ceiling of the busway platform awning, record the movements of
passengers on the platform on a 24hr / 7 day basis.
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
Group Bus arrival rate Passenger demand

Low < 40 bus/ h < 100 p/h
Medium 40 to 80 bus/ h 100 to 200 p/h
High >80 bus/ h > 200 p/h
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.
Table 4.5: Characteristics of analysis time
Time of the day Flow characteristics Time period

10:00 am to 11:00 am Medium bus frequency and Morning off-peak
Medium passenger demand
03:00 pm to 04:00 pm High bus frequency and Evening peak
High passenger demand
07:00 pm to 08:00 pm High bus frequency and Evening off-peak
Medium passenger demand
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.

Observed bus arival rate (bus/hr)

2 28 29 34 38 42 47 66 74 75 75 75 76 81 83 85 109 115 121
Corresponding hour ending at
6 7 24 23 22 21 20 8 12 11 13 14 10 15 9 19 16 17 18
Classification split
Low Medium High
0 6 6
28 24 24
39 23 23 Null set
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
325 17 Null set Null set 17

776 16 16
Note: 6 mean hour ending at 6:00 am, similarly, 17 means hour ending at 5:00 PM
Figure 4.8: Matrix for data mining of passenger demand and bus frequency

4.8 Data processing

For each hour of recording, footage was generated from each of three cameras.
These footages were then processed and synchronised in the laboratory using
Microsoft Movie Maker software. Later, each of the three footages were played
simultaneously on separate computers and manually examined to extract variables
in accordance with the methodologies described in section 4.4.
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.
4.9 Chapter close

This chapter described the developments of methodologies for data collection,
extraction and analysis. Video recording technique was chosen for the purpose of
data collection, as it provided the opportunity to revisit the footage multiple times and
extract the variable as and when needed.
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
Chapter Five
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.
5.2 Measuring platform crowd

For this study the crowd at the study platform was measured as the average number
of people waiting on the platform in a 15 minute interval. Crowd density was not used
as a variable because it was not consistent across the platform. For instance, the
crowd density at middle of the platform under low crowd conditions can be same as
that of crowd density of the entire platform area under high crowd conditions.
Therefore the use of density may not properly reflect the total crowd level variation
between periods. A detailed discussion on passenger behaviour and crowd
concentration on the station platform is presented in Section 5.4.
5.3 Passenger – bus interface

By definition, the duration of passenger – bus interface is made up of time taken by
passengers to reach the bus entry door from their waiting position having hailed the
bus (Ref: Figure 3.2).
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.
Table 5.1: Duration of passenger – bus interface during off-peak period
Duration of Loading area 1 Loading area 2 Loading area 3

Passenger – bus interface
Min 7s 2s 7s
Mean 16s 11s 14s
Max 29s 18s 22s
Std dev 3.2s 3.2s 4.3s
Boarding pax observed 70 27 18
Total boarding pax 165 106 38
Total bus services 57 36 15
Note: Off – peak operation; passengers on platform < 30; total analysis time = 120 minutes

Table 5.2: Duration of passenger – bus interface during peak period
Duration of Loading area 1 Loading area 2 Loading area 3

Passenger – bus interface
Min 1s 3s 6s
Mean 19 s 13 s 11s
^
Max 31s (51s) 25s 38s
Std dev 9s 4.4s 9s
Boarding pax observed 32 22 32
Total boarding pax 149 174 200
Total bus services 12 15 15
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)
Lower spread High spread

35
Low crowding High crowding
30
25
20
15
10
0
0 10 20 30 40 50 60 70 80
Passengers on platfrom
Figure 5.1: Variation in passenger – bus interface
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.
Secondly, considerable spread in passenger – bus interface is noticeable with

change in crowd level. It can be observed that the spread in the duration of
passenger – bus interface was lesser when the crowd at the platform was relatively
small. As the crowding increased, the passenger – bus interface became more non –
uniform and the spread considerably larger. This means, under increased crowd
conditions, while some passengers have a reduced passenger – bus interface others
have very high interface duration. Low interface duration occurs when the bus stops
close to the passenger’s waiting position, which means passenger walks a shorter
distance. High duration of interface may occur because, as the crowd size increases,

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.

LA 1 LA 2 LA 3 Linear (LA 1) Linear (LA 2) Linear (LA 3)

R² = 0.0364 R² = 0.1096 R² = 0.0521
Avg duration of passenger - bus interface (s) 25
20
15
10
0
0 5 10 15 20 25 30
Passengers on platform
Figure 5.2: Variation in passenger – bus interface during off-peak period
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.
5.3.1 Discussion of passenger – bus interface

The inability to draw a robust mathematical relationship between passenger – bus
interface and platform crowd, and the number of passengers encountered en route,
indicates significant complexity amongst all variables. The duration of passenger –
bus interface depends on the passenger’s walking distance to cover, speed and
straightness of walking path. The straightness of the passenger’s walking path,
indeed, depends on the number of other passengers in the path. However, number
of passenger encounters depends on platform crowd density. Crowd also influences
the passenger’s choice of waiting point. The walking distance for passenger, on its
part, is subjective to its waiting position on the platform relative to the loading area.
Moreover, the relative positions of loading area (used by bus) and waiting point, in
space, decide the passenger’s walking direction with respect to the direction of bus
arrival. The effects of walking direction on passengers bus interface is discussed

latter in Section 5.4. Therefore, the duration of passenger – bus interface is a

function of variables having complex interconnections. Figure 5.3 attempts to present
the connections between these variables in the simplest manner possible.
Platform crowd Waiting point Loading area
Number of passenger Walking distance Walking direction w.r.t.

encountered bus arrival direction
Walking speed
Passenger – bus interface duration
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.
5.3.2 Time – space diagram

Observations made at the study busway station platform indicate that only the
duration of passenger – bus interface of the first passenger impacted bus dwell time.
Because of the simultaneous interfaces of all boarding passengers with the desired
bus, all boarding passengers except the first boarding passenger overlap their
respective passenger – bus interface duration with that of first passenger’s interface.
In addition to this, some part of the first passenger’s interface duration overlaps with
the time taken by the bus to reach the loading area. The above observations can be
better understood using the time-space diagram shown in Figure 5.4. The distance in
space along the line of the platform is represented on the Y axis and the time is

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 = Bus dwell time

LT = Bus lost time
t0 = Time when passenger(s) first see the desired bus (say at point B in
space)
P1, P2,.., Pn = Location of 1st, 2nd,…, nth passenger in the space at time t0. IF1,
IF2,…,IFn represent their respective passenger – bus interface
duration.
IA = Duration of passenger – bus interaction
to = Door opening time
tc = Door closing time

DT
Space
LT IA
to ’ tc
X B ’
P 1P’2 P’n
X
t0 Time
P1
P2
Pn
B IF1
IF2
IFn
Figure 5.4: Time – space diagram
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.
5.4 Passenger behaviour while waiting

The bus arrival procedure at the study station, like most stations on Brisbane’s
busway network, follows the ‘lead stop’ operation principle. Under this principle, an
arriving bus should use the lead loading area of the platform. Only when the lead
stop is not available, the bus can use the next available loading area. This means
that, at the study station platform (Figure 5.5), if all the loading areas are empty, the
bus should use loading area 1 to provide service to passenger/s. Only if the front
loading area is occupied or blocked then the next loading area (i.e. loading area 2)
should be used. This created a relatively higher chance that a bus would use the
loading area 1 or 2 over loading area 3 (Table 5.1). This leads to passengers tending
to wait between the middle of lead loading area (loading area 1) and loading area 2
to wait for their desired bus (Figure 5.5). Under the off-peak, low crowd condition,
passengers were observed to wait within this area.

Direction of bus movement

y
Loading area 1 Loading area 2 Loading area 3
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

passenger P1 standing on reference line y – y, depending on the loading area it uses

to serve passengers.
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
Figure 5.6: Effect of loading area on bus lost 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.
5.5 Bus lost time

In this research, bus lost time is defined as the time lapse between bus stopping and
the time its first passenger placed his/her foot on the bus floor.
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.
Table 5.4: Bus lost times (LT) during off-peak periods
Bus lost time Loading area 1 Loading area 2 Loading area 3
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

Table 5.5: Bus lost times (LT) during peak period
Bus lost time Loading area 1 Loading area 2 Loading area 3
Min <1s <1s <1s
Max 16s 10s 17s
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.
5.6 Passenger – bus interaction

This research defines the passenger – bus interaction time as a phase where the
first and subsequent passengers and bus (driver) are involved in a state when they

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).
Therefore, the passenger – bus interaction (IA) can be mathematically represented

as –
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.

Table 5.6: Descriptive statistics
Door Minimum Maximum Mean Std.

deviation
Loading Area 1
Boarding time per pax Front 2.5s 6.5s 4.2s 2.5s
Alighting time per pax Rear 2.6s 8.0s 3.4s 1.4s
Loading Area 2
Loading Area 3
5.6.1 Effect of fare collection policy

During the course of this study, fare collection methods at the study station were
changed. TransLink’s Go Card, a smart card fare collection system, was introduced
between March 2007 and March 2008. Between March 2008 and April 2009 on
board ticket sales were abolished, only during the peak period between 2:30pm and
6:00pm (TransLink Web site). These changes provided a unique opportunity to study
the effects of fare policy changes on marginal service times. However, a detail study
of fare collection policy impact was not originally in the scope of this research. A
short analysis was carried out, findings of which are presented in this section.
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.

Table 5.7: Fare collection policies and observations at study station

Month/ Year March 2007 March 2008 April 2009
Time Period Afternoon Peak Afternoon Peak Afternoon Peak
Analysis time 3:00PM to 3:30PM 3:00PM to 3:30PM 3:00PM to 3:30PM
On board ticket On board ticket

purchase purchase
Pre-paid paper ticket Pre-paid paper ticket Pre-paid paper ticket
10 trip magnetic strip

10 trip magnetic stripe
Fare policy card
card
(Phasing out)
GoCard smart card

GoCard smart card
(Introduced)
Pre-paid platform policy
Number of
51 36 62
buses
Number of
boarding 348 475 250
passengers
Number of
alighting 21 17 53
passengers
5.6.1.1 Passenger boarding time

Table 5.8 gives the observed boarding time per passenger across all loading areas
of the study station platform for the March 2007, March 2008 (transition period), and
April 2009. Data from the transition period showed an initial increase in average
boarding times for all loading areas. However, later the average boarding time
decreased for all loading areas. The initial increase in service time could be
attributed to the inexperience of users in using GoCard and/or the mixture of the
Magstripe, GoCard and operator as cashier systems in place.
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.
Loading March March 2008 change April 2009 change from

area 2007 2008 from 2007 2009 2007
1 4.8s 6.8s +42% 4.2s -14%
2 5.9s 6.0s +2% 5.0s -15%
3 8.1s 8.9s +10% 4.9s -40%
5.6.1.2 Passenger alighting time

Passengers using the GoCard are required to touch off their cards to a card reader
before alighting the bus to facilitate the accurate fare calculation for their trip. Other
passengers are not required to transact at the time of alighting from the bus. Similar
to touch on while boarding, touch off also requires passengers to place the card less
than 10 cm distance from the card reader and steady for one or more seconds. This
inevitably has led to an increase in alighting time per passenger, over the previous
system used in 2007 where magstripe card holders were not required to transact on
alighting from the bus. Table 5.9 gives the average alighting time per passenger
observed at the study station for the three analysis periods.
Table 5.9: Effect of fare collection policy on alighting time per passenger.
Loading March March 2008 change April 2009 2009 change

area 2007 2008 from 2007 from 2007
1 2.2s 2.0s -9% 3.4s +55%
2 1.9s 2.0s +5% 4.0s +110%
3 2.1s 2.1s 0% 3.6s +71%
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
5.7 Chapter close

This chapter evaluated the new parameters of busway operation that were identified
in chapter 3. Intricate analyses of these parameters were performed, in which
relevant elements and processes related to passenger waiting behaviour and bus
dwell time at the busway station were identified. The analysis highlighted some
relationships between parameters and elements that effect the functioning of the
busway station platform.
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
Chapter Six
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.
6.2 Bus lost time histogram

Figure 6.1 shows the measured bus lost time histogram for the off peak period at the
Mater Hill busway station outbound platform. The values for bus lost time observed
at the station platform range from less than 1s to 16s during this period. The
frequency plot shows that at loading area 1, most of the observed lost times were 1s
to 4s. Similarly at loading area 2, the majority of buses experienced a lost time of 1s
to 3s. However, at loading area 3, the buses experienced a wide range of lost times
with no clustering, with maximum value of lost time observed as 16s.

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).
In probability theory, a log-normal distribution is a probability distribution of a random

variable whose logarithm is normally distributed. A variable might be modelled as
log-normal if it can be thought of as the multiplicative product of many independent
random variables each of which is positive (Wikipedia, 2009b). Since the bus lost
time is indeed a function of complex and dynamic relationships of passenger
walking, bus stopping, and crowd density, the log-normal distribution fitting was
therefore first considered. To avoid syntax error, (log of 0 is infinite), all values of bus
lost time less that 1s was clustered in one group of 0.5s instead of 0s. This
assumption is reasonable because in reality the lost time of 0s for a bus is highly
unlikely due to the fact that passenger takes some time to react to bus stopping and
door opening.
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.
In next section, statistical analyses are presented to fit a probability distribution

curve.

6.3 Probability distribution curve fitting

SPPS (originally, Statistical Package for the Social Sciences), was used for the
analysis and curve fitting. Table 6.1 to Table 6.3 provides the descriptive statistics of
off-peak data set of each loading area. Table 6.4 to Table 6.6 provides the
descriptive statistics of peak data set of each loading area.
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.

Table 6.1: Descriptive statistics of loading area 1 (Off-peak period)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 50 50
Mean 4.280 0.400 1.245 0.093
95% confidence Lower bound 3.476 1.058

interval for mean Upper bound 5.084 1.433
Median 3.000 1.098
Variance 8.002 0.436
Std. deviation 2.828 0.660
Skewness 1.038 0.337 -0.008 0.337
Kurtosis 0.073 0.662 -0.682 0.662
Count 30 30
Mean 2.883 0.363 0.859

Median 2.000 0.691
Skewness 1.911 0.427 -0.136 0.427
Kurtosis 4.823 0.833 0.303 0.833

Count 14 14
Mean 7.500 1.142 1.820 0.192

Median 7.500 2.013
Variance 18.269 0.519
Skewness 0.541 0.597 -1.144 0.597
Kurtosis -0.309 1.154 1.980 1.154
Table 6.4: Descriptive statistics of loading area 1 (peak period)
Count 61 61
Mean 4.066 0.415 1.095 0.107

Median 3.000 1.098
Variance 10.512 0.697
Skewness 1.692 0.306 -0.362 0.306
Kurtosis 3.497 0.604 -0.229 0.604

Table 6.5: Descriptive statistics of loading area 2 (Peak period)
Count 50 50
Mean 2.850 0.244 0.888 0.0823

Median 3.000 1.098
Skewness 2.020 0.337 -0.307 0.337
Kurtosis 6.140 0.662 0.575 0.662
Table 6.6 Descriptive statistics of loading area 3 (Peak period)
Count 38 38
Mean 4.868 0.636 1.259 0.141

Median 4.000 1.386
Variance 15.401 0.758
Skewness 1.515 0.383 -0.440 0.383
Kurtosis 2.223 0.750 -0.097 0.750

Kurtosis is a measure of peakedness of the distribution. Like skewness, the value of

kurtosis for a perfectly normal distribution is 0. A positive kurtosis value means the
distribution is too pointed. A negative value means the distribution is too flat. Higher
kurtosis means more of the variance is due to infrequent extreme deviations, as
opposed to frequent modestly sized deviations (Wikipedia, 2009c).
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.
6.3.1 Assessing normality

There are two ways of testing normality; graphical method and numerical method
(Park, 2008). Table 6.7 provided an overview of these methods. Graphical methods
visualize the distributions of random variables a theoretical distribution (e.g., the
standard normal distribution). These methods are quick and easy to interpret.
Numerical methods evaluate summary statistics such as skewness and kurtosis, or
conduct statistical tests of normality such as Shapiro-Wilk test.
Table 6.7: Methods for testing normality
Graphical Methods Numerical Methods

Descriptive Stem-and-leaf plot, Skewness,
box plot, Kurtosis
dot plot,
histogram
Theory-driven P-P plot, Shapiro-Wilk test,
Q-Q plot Shapiro- Francia test,
Kolmogorov-Smirnov test (Lillefors test),
Anderson-Darling/Cramer-von Mises tests
Jarque-Bera test,
Skewness-Kurtosis test
Source: Park, 2008
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.

6.3.2 Null hypothesis testing

Skewness measures the vertical displacement of a distribution from a standard bell
shaped normal distribution. Whereas kurtosis measures the horizontal displacement
of a distribution from a standard bell shaped normal distribution. When an observed
data superimposes perfectively on the standard bell shaped normal distribution, its
skewness statistic and kurtosis statistic equals to zero. Therefore value of zero is the
primarily value of interest in assessing the normality of the data (Innes, 2007).
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.
Table 6.8: Assessing normality for loading area 1 (Off-peak period)

Skewness 1.038 0.337 -0.008 0.337
95% confidence Lower bound 0.377 -0.669
interval for skewness Upper bound 1.699 0.653
Kurtosis 0.073 0.662 -0.682 0.662
95% confidence Lower bound -1.225 -1.980
interval for Kurtosis Upper bound 1.371 0.616

Skewness 1.911 0.427 -0.136 0.427
Kurtosis 4.823 0.833 0.303 0.833

Skewness 0.541 0.597 -1.144 0.597
Kurtosis -0.309 1.154 1.980 1.154

Table 6.11: Assessing normality for loading area 1 (Peak period)

Skewness 1.692 0.306 -0.362 0.306
Kurtosis 3.497 0.604 -0.229 0.604

Skewness 2.020 0.337 -0.307 0.337
Kurtosis 6.140 0.662 0.575 0.662

Skewness 1.515 0.383 -0.440 0.383
Kurtosis 2.223 0.750 -0.097 0.750

6.4 Log-normal distribution curves for bus lost time

This section presents the log-normal probability distribution function (PDF) curve and
cumulative distribution function (CDF) curve of bus lost time for each loading area
under each time period. The standard log-normal PDF is given by Equation 6.1. The
log-normal CDF is given by Equation 6.2.
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
Std. deviation of X = X Equation 6.5

Where,
= Mean of variable X
= Variance of variable X
6.4.1 Log-normal probability distribution function curve

Figure 6.3 provides the bus lost time probability distribution function (PDF) for the
lognormal distribution curves for each loading area for peak time periods. From the
peak period curves, it was noted that all three loading areas have nearly same
probabilities of experiencing a lost time of 4.1s. However loading area 2 has a higher
probability of experiencing lost time of less than 4.1s. Whereas, the loading area 3,
relatively to loading area 1 and 2, has lesser chances of experiencing a lost time less
than 4.1s. On the contrary, the probability of loading area 3 getting a lost time larger
than 4.1s is relatively higher in comparison to loading area 1 & 2.
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)
Note: LA = loading area

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)
Note: LA = loading area; OP = Off-peak

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)
Note: LA = loading area

0.9
85th percentile
0.8
0.7
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
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 curve parameters µ and σ are given in Table 6.14.
Table 6.14: Statistical parameters of bus lost time curves
Loading area Period µ σ

1 Peak 1.095 0.834
2 Peak 0.888 0.582
3 Peak 1.259 0.870
1 Off-Peak 1.245 0.660
2 Off-Peak 0.859 0.650
6.4.2 Log-normal cumulative distribution function curve

Figure 6.6 provides the cumulative distribution function (CDF) of bus lost times for
each loading area for peak time period. The CDF highlights the effect on bus lost
times due to difference in direction of motion of a passenger and the approaching
bus. For a given probability, the lost time for loading area 3 is always higher than that
for loading area 1. Theoretically the CDF for the loading area 1 and 3 should overlap
or near to each other, because the walking distance to both loading area is nearly
same from the shaded area. However, the CDF of loading area 3 is right shifted with
respect to loading area 1. This is because a passenger requires walking in direction
opposite to approaching bus to loading area 3 as against the same directional
motional for the loading area 1. This directional discrepancy resulted in a lesser
slope to the CDF for loading area 3 compared to loading area 1.
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.
6.4.3 Descriptive characteristics of busway station bus lost time

This section presents the descriptive characterises of generalised bus lost time, such
as mean, mode, median and 85th percentile values. In statistics, the mean, mode,
and median are also called as the measures of central tendency which describes the
scores in a distribution. Table 6.15 and Table 6.16 provide central tendency and 85th
percentile values of bus lost time for all three loading areas for peak and off-peak
period respectively.
Table 6.15: Descriptive characteristics of bus lost times (Peak period)
Loading area Loading area Loading area

1 2 3
Mode of lost time 1.6 2.0 2.0
Median lost time 3.0 2.4 3.5
Mean lost time 4.2 2.9 5.2
th
85 percentile lost time 7.1 4.5 8.8
Table 6.16: Descriptive characteristics of bus lost times (off-peak period)
Loading area Loading area Loading area

1 2 3
Mode of lost time 2.3 2.0 -
Median lost time 3.5 2.4 -
Mean lost time 5.3 3.6 -
85th percentile lost time 7.0 4.8 -
6.5 Chapter close

This chapter presented the development of probability distribution models for bus
lost times. Statistical analysis of bus lost time established that the busway station
bus lost time is a positively skewed, log-normally distributed variable. A set of five
log-normal distribution curves are developed which approximates the observed data

sets. These probability models can be used to develop a micro-simulation based

model to study busway station operation. Furthermore, understanding of bus lost
time can be applied in improving the methodology of busway station bus capacity
analysis.
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
Chapter Seven
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.
7.2 Model framework

Chapter 2 had identified the inadequacies in applying the existing bus stop dwell
time models for busway station analysis. These inadequacies arise due to their lack
of sensitivity towards certain characteristics of the busway station platform operation,
such as the time spent by passengers walking to bus entry door, platform crowding
and multiple loading areas. To achieve a reliable bus dwell time estimation
methodology for a busway station, the busway station bus dwell time model should
account for the characteristics of the busway station in addition to the elements of
passenger service times. Figure 7.1 provides an overview of variables determining
the bus dwell time at a busway station platform.
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 loading area characteristics perpetuate the requirement of passenger walking

to the bus entry door to board the desired bus. The passenger station demand
creates platform crowding, which in turn impacts passenger walking. Analysis
presented in section 5.3 established that passenger walking to the bus entry door
and the crowd encountered en route causes lost time for the bus.
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.

Station demand Loading area Number of boarding and

(Pax per hour) characteristics alighting passenger (Pa, Pb)
Platform crowding Walking distance
Bus door passenger

processing time (ta, tb)
Passenger – bus
interface (IF)
Bus lost time (LT)
Door time (toc)
Bus dwell time (DT)
Figure 7.1: Overview of model form for busway platform dwell time estimation
7.3 Busway station bus dwell time model

Equation 7.1 presents the bus dwell time model for dwell time estimation at a
busway station with multiple linear loading areas. The model is known as BSDT
model, an abbreviation of Busway Station Bus Dwell Time Model.
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.
7.4 Example application

To study the impact of the bus lost time variable on bus dwell time estimation, the
BSDT model and the bus dwell time model given by TCQSM (TRB, 2003) were
applied to an example busway station. The latter model does not account for the bus
lost time in its methodology. The difference in estimated dwell time values obtained
from both the models will therefore be due to the influence of the bus lost time. The
effect of bus lost time was analysed at two levels: at the platform level and at the
individual loading area level. For this demonstrative exercise, the peak period
operation of the station was analysed.
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.

Table 7.1: Example demonstration
Bus dwell time

Method Loading Loading Loading
area 1 area 2 area 3
TCQSM model
30s 30s 30s
(without bus lost time)
BSDT model
34.2s 32.9s 35.2s
(with bus lost time)
% change +14.0% +9.7% +17.3%
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.

Contribution of lost time in dwell time (%) 60
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
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

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

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 –
1) The new model is sensitive to multiple linear loading area operation.

2) It is also sensitive to the platform crowd and passenger walking to bus entry
door, through the bus lost time variable.
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.
Another important application of the proposed dwell time estimation methodology is

in the area of development of bus arrival algorithms; for example, the real time
information systems and bus priority signal design. The proposed dwell time
methodology, when incorporated with the bus arrival algorithms could greatly
improve the accuracy as the results will be based on the prevailing conditions of the
upstream stations. Additionally, this new methodology can help transit planners in
improving the scheduling of service timetable and this in turn could greatly enhance
the travel time reliability.

7.6 Chapter close

This chapter presented a bus dwell time model for a busway station. The new model
is called as BSDT model. The BSDT model considers the bus lost time variable in
the dwell time calculation and therefore reflects more accurately the delay to a bus at
a busway station. Bus dwell time is an important parameter in determining the bus
capacity of a loading area. In the following chapter a refined model for loading area
bus capacity estimation is presented. The new loading area bus capacity model
incorporates the new dwell time model incorporating the bus lost time variable.

Busway Loading Area Bus Capacity Model
Chapter Eight
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.
8.2 Approach to busway loading area bus capacity model

Loading area bus capacity, by definition, means the number of buses that are able to
provide service at that loading area in a given period of time, say an hour. It
incorporates the time each bus dwells, and time it take to clear the loading area. In
simplistic terms, the loading area bus capacity, B, may be estimated by -
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,
Bl = Loading area bus capacity (bus/h)
3600 = Number of seconds in an hour

g C = Green time ratio (the ratio of effective green time to total traffic signal
cycle length; 1.0 for unsignalised streets and bus facilities like busway
without adjacent signal control)
tc = Clearance time (s)
= Standard normal variate corresponding to a desired failure rate

= Coefficient of variation of dwell time
The various determinants of busway loading area bus capacity are discussed below.
8.2.1 Busway dwell time

Dwell time was earlier defined in chapter 7 as the average amount of time a bus
stopped at a loading area at the busway station to provide the service to its
passengers. This includes the time required for door opening and closing, and the
time lost by the bus due to the delay in arrival of its first passenger.
8.2.2 Dwell time variability

The concept of dwell time variability is used by Transit Capacity and Quality of
Service Manual (TRB, 2003) for its kerbside loading area bus capacity model. It
accounts for the consistency (or lack thereof) of dwell times among buses using the
loading area. The dwell time variability in combination with failure rate are used to
provided operating margin for the bus operation.

8.2.3 Failure rate

A bus cannot always find the subject loading area empty loading area upon its
arrival. The rate at which such condition occurs is called the failure rate (TCRP,
1997, TRB, 2003).
8.2.4 Operating margin due to passenger service time variability

The equation to estimate the operating margin based upon dwell time variability at
kerbside bus stop as given by Transit Capacity and Quality of Service Manual (TRB,
2003) is
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
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.
Table 8.1: Failure rates and corresponding ‘z’ values
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.
8.2.5 Lost time variability

Not every bus at a given loading area experiences the same amount of lost time.
The bus lost time fluctuates with the waiting position of its first passenger. Similar to
the concept of operating margin to account the bus dwell time variability bt way of

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.
Bus lost time operation margin

Density
Mean lost time
LTi Bus lost time, LT
Figure 8.1: Log-normal density curve
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 8.5 may be rearranged to yield Equation 8.6
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

Replacing the mean lost time ( ) with equation 6-3

⁄
Equation 8.8
The generalised model for lost time operation margin is therefore given as:
⁄ Equation 8.9
,
Where,
, = Lost time operating margin for nth loading area (s)
= Standard normal variable corresponding to a desired

failure rate
& = Lost time probability curve parameters for nth loading area
8.3 Busway loading area bus capacity model

From indicative logic, the bus capacity of nth loading area of a busway station is
given by:
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)
, = Operating margin for bus lost time variability at nth loading

area (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)
8.4 Effective bus capacity of loading area

Both Equation 8.10 and Equation 8.11 estimate the maximum bus capacity of the
loading area. However, a linear loading area may potentially reduce the effective bus
capacity of other loading areas because it may block the bus entry to those loading
areas. Any time duration for which a loading was blocked, lessens availability of
effective time in which it can provides any actual service to the passengers.
Therefore the effective loading area bus capacity is, therefore, a product of
maximum bus capacity and its efficiency factor and can be calculated as

, Equation 8.12
Where,
, = Effective bus capacity of nth loading area (Bus/ hr)

= Bus capacity of nth loading area (Bus/ hr)
= Efficiency factor of nth loading area
The discussions and modelling of linear loading areas efficiencies are presented in
chapter 9.
8.5 Busway station platform bus capacity

The effective capacity of the busway station platform will be equal to the summation
of effective bus capacities of all variable loading areas, and can be given by
Equation 8.13.
, Equation 8.13
Where,
= Busway station platform bus capacity (Bus/ hr)
, = Effective capacity of nth loading area (Bus/ hr)
= Total number of loading areas available at busway station platform
8.6 Example application

The significance of the bus lost time is demonstrated by estimating the station bus
capacity for the same example busway station described in Section 7.4. The details
of the example busway station are redescribed here. The Transit Capacity and
Quality of Service Manual (TRB, 2003) methodology was used for base calculations
and the results were compared with the new methodology considering bus lost time
which is presented earlier in this chapter. For this purpose, a boarding load of 7
passengers per bus with boarding service time of 4s and no alighting load was
considered. The estimation was done with assume failure rate of 7.5%, 60%

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.
Table 8.2: Example demonstration
Bus dwell time

Station
Method Loading area Loading area Loading area Capacity
1 2 3
TCQSM
30s 30s 30s 143 bus/hr
(without bus lost time)
BSLC
34.2s 32.9s 35.2s 123 bus/hr
(with bus lost time)
% change +14.0% +9.7% +17.3% -13.9%
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
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
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
200 g/C = 0.5
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.
8.8 Chapter close

This chapter presented a loading area bus capacity model for busway station. The
new model is called as BSLC model, an abbreviation for Busway Station Loading
Area bus Capacity model. The BSLC model incorporated the new busway dwell time
model (BSDT) and the bus lost time variable.
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
Chapter Nine
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).
This chapter describes the development of a methodology for estimating busway

loading area efficiency. Section 9.2 first discussed the operation of Mater Hill
Busway station. Later the approach for loading area efficiency calculation was
described in section 9.3.
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.
9.2 Loading area blocking

Non-uniform bus dwell time across different loading areas of a busway platform may
lead to blocking of a loading area by its predecessor loading area/s. The Non-
uniformity in bus dwell times may happen due the variation in bus lost time,
illustrated in figure 5-6 and the number of passenger boardings and alightings
between loading areas (TRB, 2003).

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
Start of afternoon peak period

Platoon arrival from platform.
at station.
Platoon arrival
at platform.
Loading area 3 blocking

loading areas 2 & 1.
Loading area 1 blocked by loading area 2
0 100 200 300 400 500 600 700 800

Time (s)
Figure 9.1: Trajectory of bus processing at the Mater Hill Busway Station (Outbound 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 2: Loading area 2 is occupied, which results in blockage of loading area 1.

In this scenario, the platform has one third of its capacity blocked. Of the remaining
two thirds of its capacity, one third is in utilization and other third is available for
servicing. The platform is still 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 5: Loading area 3 is occupied and is blocking loading areas 1 and 2. In

this scenario the platform has two third of its total capacity blocked and only one third
utilised. The platform is said to be virtually performing at full capacity with a loss of
two third 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.

Scenario Diagram Nel

1
3
Outbound platform

2
2
Outbound platform

3
2
Outbound platform

4
2
Outbound platform

5
1
Outbound platform

6
3
Outbound platform
Note: Nel = Number of effective loading areas

Figure 9.2: Inter-loading area blocking scenarios and associated numbers of effective loading
areas.

These blockings of loading areas result in reduction in their efficiencies causing a

decline in effective number of loading areas. The right most column of Figure 9.2
provides the number of effective loading areas (Nel) available at the platform at the
time under each scenario.
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.
9.2.1 Existing approach

The TCQSM (TRB, 2003) noted that each additional loading area provided linearly at
the platform has a reduced contribution to the total effective number of loading
areas. Table 9.1 provides the efficiency factors given by the TCQSM. However,
these suggested efficiency factors are based on the operation of a bus terminal
facility (TRB, 2003; Levinson et.al., 1975) and not based on the operation of busway
station facility.
Table 9.1: Efficiency factors provided by TCQSM
loading area Efficiency Cumulative # of

% Loading Areas
3rd 1 1
2nd 0.85 1.85
1st 0.80 2.65
Source: TRB, 2003

9.3 Approach to loading area efficiency factor calculation

Loading area efficiency is a dynamic factor, sensitive to the bus flow rate. It can
change from time to time. The loading area efficiencies can be higher under a low
bus flow condition compared to high bus flow condition. Therefore, an afternoon
peak period was analysed to determine conservative values.
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.
1.0 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)
Loading area Occupied Time Blocked Time
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
Time preceding Time loading area Loading

Loading Cumulative
loading areas empty while Area
Area Loading Areas
occupied preceding occupied Efficiency
3rd NA NA 1.00 1.00
2nd T3 = 814s T2,b = 82s 0.90 1.90
1st T2,3 = 1244s T1,b = 362s 0.71 2.61
Note: Analysis period, T = 1800s
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.
Table 9.4: Comparison of loading area efficiency results
Number of available Mater Hill Busway TCQSM’s Cumulative Loading

loading area station Areas for Comparison
1 1 1
2 1.9 1.85
3 2.61 2.65
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.
9.6 Chapter close

This chapter presented a methodology to estimate the efficiencies of linearly placed
loading areas at a busway station platform. The estimated efficiencies were used in
the demonstrative example for station platform bus capacity estimation presented in
section 8.6.
Next chapter concludes this thesis and discusses the future research opportunities in
the area of bus lost time to enhance the models.

Conclusions
Chapter Ten
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.
10.2 Summary of this thesis

This thesis presented an intricate analysis of busway station operation and identified
various parameters and processes related to passenger waiting behaviour and bus
dwell time. Chapter 1 established the aim and objectives of this research and the
scope of this thesis. Having set up the research goals, an extensive literature survey
was conducted to study the existing knowledge in the area of busway bus capacity
analysis. The findings of the literature survey were presented in Chapter 2. The
literature search found that no methodology available for busway station bus
capacity analysis which addresses all of the mechanisms considered here to be
important.
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

Conclusions
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.
10.3 Contributions of this research

From the knowledge prospective, this research identified for the first time the phase
of passenger – bus interface which occurs at a busway station. This insight then lead
to the most important contribution of this research, identification of bus lost time as
an important parameter of influence to bus dwell time. Subsequently, a new bus
dwell time model and loading area bus capacity model were developed for a busway
station.
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,
i. Developed a framework of busway operation.

ii. Identified for first time the interface between passenger and bus.
iii. Established descriptive characteristics of the busway station bus lost time.
iv. Established the effectiveness of linear loading areas in busway station bus
capacity, under local conditions.
10.4 Implications of this research

Though this research used a particular busway station platform as its case study, the
findings of this research may be applied to any station with multiple linear loading
areas. The results from this research have applications both for research and
practice. For research, the knowledge of the passenger – bus interface can be
applied in better understanding passenger movements in the platform area. This can
be used in developing better platform designs to reduce bus lost time.

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.

Conclusions
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.
Identifying the variables governing the operation of a busway station is absolutely

necessary to explore a busway system’s full potential. The demonstrative examples
have shown that bus lost time at a platform is one such variable, which manipulates
the operational capacity of a busway station.
10.6 Recommendations for future work

This research has provided a comprehensive analysis of a multiple linear loading
area busway station and has provided many insight of busway station functioning.
However, further research is required to test the new approach for other busway
station configurations and modes of operation. Consequently, future research is
required in a number of directions, in order to make the results of this research more
robust and versatile.
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
References
Ando, K., Ota, H. and Oki, T. (1988). Forecasting the flow of people, (in Japanese),
RRR, Railway Research Review, 45(8), 8-14.
Antonini, G. and Bierlaire, M. (2005). Capturing interactions in pedestrian walking

behavior in a discrete choice framework, Technical Report RO-050921,
Operations Research Group ROSO, Institute of Mathematics, Ecole
Polytechnique F´ed´erale de Lausanne, CH-1015 Lausanne, Switzerland.
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.
Changshan, W. and Murray, A. T. (2005). Optimizing public transit quality and

system access: the multiple-route, maximal covering/shortest-path problem.
Environment and Planning B: Planning and Design, 32, 163-178.
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.
Daamen, W. (2004). Modelling Passenger Flows in Public Transport Facilities. PhD

Dissertation, Department Transport & Planning, Delft University of Technology,
Netherland.
Daamen, W., Hoogendoorn, S. and Bovy, P. (2005). First-Order Pedestrian Traffic

Flow Theory. In Transportation Research Record: Journal of the Transportation
Research Board, 1841, 120-127.
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.
Dipletro, C. M. and King, L. E. (1970). Pedestrian Gap-Acceptance. In Highway

Research Record 308, HRB, National Research Council, Washington, D.C.,
1970, pp. 80-91.
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
FTA (2004). Characteristics of bus rapid transit for decision-making. Department of

Transportation, Federal Transit Administration, Report number FTA-VA-26-
7222-2004.1. Washington D.C.
FTA (2008). Bus rapid transit. Department of Transportation, Federal Transit

Administration, Washington DC. Last accessed November, 2009 at
http://www.fta.dot.gov/research_4240.html
Gerin-Lajoies, M. and Richards, C.L. (2006). The circumvention of obstacles during

walking in different environmental context: A comparison between older and
younger adults. Gait & Posture, 24(3), 364-369.
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.
Hensher, D. A. (2007). Sustainable public transport systems: Moving towards a

value for money and network-based approach and away from blind
commitment. Transport Policy, 14, 98-102.
Hoogendoorn, S. P. and Bovy, P. H. L. (2004). Pedestrian route-choice and activity

scheduling theory and models. Transportation Research Part B, 38, 169-190.

References
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.
IAPT (2007). International Association of Public Transport. Last accessed October,

2007 at http://www.uitp.org/
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.
Khisty, C.J. (1990). Transportation Engineering – An introduction Prentice-Hall, New

Jersey.
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. (1983). Analyzing Transit Travel Time Performance. In

Transportation Research Record: Journal of the Transportation Research
Board, 915, 1-6.
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.
Maloney, M. and Boyle, D. (1999) Components of travel time on the Glendale

Beeline bus network. In Transportation Research Record: Journal of the
Transportation Research Board, 1666, 23-27.
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.
Nishinari, K., Sugawara, K., Kazama, T., Schadschneider, A. and Chowdhury, D.

(2006). Modelling od self-driven particles: Foraging ants and pedestrians.
Physica A, 372, 132-141.
Osaragi, T. (2004). Modeling of pedestrian behaviour and its applications to spatial

evaluation. In Proceedings of the Third international Joint Conference on
Autonomous Agents and Multiagent Syatems, Washington, USA. IEEE
Computer Society, 836-843.
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.

References
Rathwell, S. and Schijns, S. (2002). Ottawa and Brisbane: Comparing a Mature

Busway System with Its State-of-the-Art Progeny. Journal of Public
Transportation, 5(2), 163-182.
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.
Tarawneh, M. S. (2001). Evaluation of pedestrian speed in Jordan with investigation

of some contributing factors. Journal of Safety research, 32(2), 229-236.
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.
TransLink (2007). Unpublished Report (http://www.transport.qld.gov.au/)
TransLink (2009). South East Busway: Rochedale to Springwood. Last accessed

November, 2009 at http://www.transport.qld.gov.au/
Translink web site: www.translink.com.au
TRB (2003). Transit Capacity and Quality of service Manual. Transportation

Research Board of the National Academics, Washington, D.C. Last accessed
February, 2007 at http://www.trb.org/.

Tyler, N. (2002) Accessibility and the bus system: from concepts to practice. Thomas
Telford. London.
Vuchic, A. R. (2005). Urban Transit: Operations, Planning, and Economics. John

Wiley and Sons, Inc., New Jersey.
Wikipedia (2009a). Ottawa Rapid Transit. Last accessed November, 2009 at

http://en.wikipedia.org/wiki/Ottawa_Rapid_Transit
Wikipedia (2009b). Log-normal Distribution. Last accessed November, 2009 at

http://en.wikipedia.org/wiki/Log-normal_distribution
Wikipedia (2009c). Kurtosos. Last accessed November, 2009 at

http://en.wikipedia.org/wiki/Kurtosis
Wirasinghe, S. C. and Ghoneim, N. S. (1981). Spacing of bus-stops for many to

many travel demand. Transportation Science, 15, 210-221.
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
Appendix – A
BUSWAY STATION PLATFORM BUS CAPACITY ANALYSIS WORKSHEET
Diagram
Space for 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)
Loading area efficiency (Table 9-3) Loading area 1 LT1 = _________

Loading area 1 E1 = ______ Loading area 2 LT2 = _________
Loading area 2 E2 = ______ Loading area 3 LT3 = _________
Loading area 3 E3 = a 1.0 t
Cont…

Cont…
Calculation steps
1. Busway dwell time

(s) Equation 7-1
2. Operating margin for passenger service time

, , (s) Equation 8-4
3. Operating margin for bus lost time

⁄
, (s) Equation 8-9
4. Loading area bus capacity

3600
(bus/ hr) Equation 8-10
, ,
5. Effective bus capacity of loading area

, (bus/ hr) Equation 8-12
6. Platform bus capacity
, (bus/ hr) Equation 8-13
Capacity Calculation
Loading area LTn DTn tomp,n toml,n Bn En Bef,n
1 ____ ____ ____ ____ ____ ____ ____
2 ____ ____ ____ ____ ____ ____ ____
3 ____ ____ ____ ____ ____ ____ ____
Platform bus capacity (Bs) = _________ (bus/hr)

Appendix- B
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
Additional Inputs for BSLC model:

12 Loading area efficiency (Table 9-3)
Loading area 1 E1 = 0.71
13 Parameter, µ, for bus lost time (Table 6-14) =
Loading area 1 µ1 = 1.095
14 Parameter, σ, for bus lost time (Table 6-14) =
Loading area 1 σ1 = 0.834
15 Mean lost time (Table 6-14) =
Loading area 1 LT1 = 4.2

16 Co-efficient of passenger variation Cv,p = 60%
Solution:
a) By TCQSM, 2003 method
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

Bl = 54 bus per hour
iii. Calculate station capacity using Equation 2.21 (on page 40)
Bs = Bl * Ne = 54 * 2.65 = 143 bus/hr
b) By BSLC, 2009 method
i. Calculate bus dwell time using revised dwell time equation, Equation 7.1 (on
page 121)

Loading area Dwell time

For loading area 1 = 34.2s

Appendix- B
ii. Calculate operating margin for passenger service time using Equation 8.4 (on
page 132)
, ,
Loading area Operating margin for

passenger service
iii. Calculate operating margin for bus lost time using Equation 8.9 (on page 134)
⁄
,
Loading area Operating margin for

bus lost time
iv. Calculate loading area bus capacity using Equation 8.10 (on page 134)
3600

, ,
Loading area Bus capacity

For loading area 1 = 47 bus/hr
v. Calculate effective loading area bus capacity using Equation 8.12 (on page
136)
,
Loading area Effective bus capacity

vi. Calculate platform bus capacity using Equation 8.13 (on page 136)
= 123 bus per hour

Appendix- C
Appendix – C
List of publications
Refereed journal paper publication
• 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.
Refereed conference paper publication
• 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.

Journal paper (Under review)
• 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).

Busway Platform Bus Capacity Analysis

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Busway Platform Bus Capacity Analysis

Загружено:

Авторское право:

Доступные форматы

Busway Platform Bus Capacity Analysis

Sumeet Kumar Jaiswal

A thesis submitted for the degree of

School of Urban Development

Faculty of Built Environment & Engineering

Queensland University of Technology

Sumeet Jaiswal Page i

Line capacities of a busway are predominately dependent on bus capacity of its

Sumeet Jaiswal Page iii

Sumeet Jaiswal Page iv

Two 2Literature Review 9

Sumeet Jaiswal Page v

2.6 Crowd density and walking speed 28

Three 3Research Problem Development 45

Four 4Data Collection and Processing 55

Sumeet Jaiswal Page vi

Six 6Modelling Bus Lost Time 95

Seven 7Busway Station Dwell Time Model 119

Eight 8Busway Loading Area Bus Capacity Model 129

Sumeet Jaiswal Page vii

8.2.3 Failure rate 131

Nine 9Busway Station Efficiency Model 141

Ten 10Conclusions 151

A 13Busway Station Platform Bus Capacity Analysis Worksheet 163

B 14Bus Capacity example application 165

C 15List of Publications 169

Sumeet Jaiswal Page viii

Table 2.1 Pedestrian level of service on walkways 34

Table 2.2 Levels of service for queuing area 34

Table 2.3 Similar type of bus lanes and busway 37

Table 2.4 Failure rates and corresponding ‘z’ values 39

Table 3.1 Boarding process at a bus stop and at a busway station 48

Table 4.1 Data collection techniques used in past studies 56

Table 4.2 Candidate busway station (outbound) platforms 63

Table 4.3 Fare collection policies at Mater Hill Busway Station 65

Table 4.5 Characteristics of analysis time 69

Table 5.1 Duration of passenger – bus interface during off-peak period 74

Table 5.2 Duration of passenger – bus interface during peak period 75

Table 5.4 Bus lost times (LT) during off-peak periods 85

Table 5.5 Bus lost times (LT) during peak period 86

Table 5.6 Descriptive statistics 90

Table 5.7 Fare collection policies and observations at study station 91

Table 6.1 Descriptive statistics of loading area 1 (Off-peak period) 100

Table 6.2 Descriptive statistics of loading area 2 (Off-peak period) 100

Table 6.3 Descriptive statistics of loading area 3 (Off-peak period) 101

Table 6.4 Descriptive statistics of loading area 1 (Peak period) 101

Sumeet Jaiswal Page ix

Table 6.5 Descriptive statistics of loading area 2 (Peak period) 102

Table 6.6 Descriptive statistics of loading area 3 (Peak period) 102

Table 6.7 Methods for testing normality 103

Table 6.14 Statistical parameters of bus lost time curves 115

Table 7.1 Example demonstration 123

Table 8.1 Failure rates and corresponding ‘z’ values 131

Table 8.2 Example demonstration 137

Table 9.1 Efficiency factors provided by TCQSM 146

Table 9.4 Comparison of loading area efficiency results 149

Sumeet Jaiswal Page x

Figure 1.2 Structure of this thesis 6

Figure 2.1 BRT configurations and their passenger transit facilities 11

Figure 2.3 A kerb side simple bus stop 13

Figure 2.4 An enhanced BRT stop 13

Figure 2.5 Mater Hill Busway Station 14

Figure 2.6 Transit centre with sawtooth arrangement of loading areas 15