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 Busway Platform Bus Capacity Analysis 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 Busway Platform Bus Capacity Analysis
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 5.7 Chapter close 93
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 6.5 Chapter close 116
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 7.6 Chapter close 128
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 Busway Platform Bus Capacity Analysis
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 8.8 Chapter close 140
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 9.6 Chapter close 150
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 6.2 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 Busway Platform Bus Capacity Analysis
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 6.8 Assessing normality for loading area 1 (Off-peak period) 105
Table 6.9 Assessing normality for loading area 2 (Off-peak period) 105
Table 6.10 Assessing normality for loading area 3 (Off-peak period) 105
Table 6.11 Assessing normality for loading area 1 (Peak period) 106
Table 6.12 Assessing normality for loading area 2 (Peak period) 106
Table 6.13 Assessing normality for loading area 3 (Peak period) 106
Table 6.14 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 Busway Platform Bus Capacity Analysis
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 7.4 Effect of bus lost time on dwell time at loading area 2 (Peak period) 125
Figure 7.5 Effect of bus lost time on dwell time at loading area 2 (Off-peak period) 126
Figure 7.6 Effect of bus lost time on dwell time at loading area 3 (Peak period) 126
Figure 8.1 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 Busway Platform Bus Capacity Analysis
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 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 2 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
Sumeet Jaiswal Page 3 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 4 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
Sumeet Jaiswal Page 5 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 6 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.
Sumeet Jaiswal Page 7
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
Sumeet Jaiswal Page 9 Busway Platform Bus Capacity Analysis 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
Sumeet Jaiswal Page 10 Literature Review
Photo by: Jaiswal (2009) 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
Sumeet Jaiswal Page 11 Busway Platform Bus Capacity Analysis
and management highly important for smooth operation of the busway corridor. Figure 2.2 shows Brisbane busway and its station.
Photo by: Jaiswal (2009) 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.
Sumeet Jaiswal Page 12 Literature Review
Photo by: Jaiswal (2009) Figure 2.3: A kerb side simple bus stop
Photo by: Jaiswal (2009) Figure 2.4: An enhanced BRT stop
Sumeet Jaiswal Page 13 Busway Platform Bus Capacity Analysis
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.
Photo by: Jaiswal (2009) Figure 2.5: Mater Hill Busway Station
Sumeet Jaiswal Page 14 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.
Photo by: Jaiswal (2009) Figure 2.6: Transit centre with sawtooth arrangement of loading areas
Sumeet Jaiswal Page 15 Busway Platform Bus Capacity Analysis
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)
Sumeet Jaiswal Page 16 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
Sumeet Jaiswal Page 17 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 18 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,
t = dwell time at station s t = station standing time o b ' and a ' = number of boarding and alighting passengers respectively τ τ = respective boarding and alighting per passenger b and a
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).
Sumeet Jaiswal Page 19 Busway Platform Bus Capacity Analysis
Equation 2.8 td = Pata + Pbtb + toc
Where,
t = Average dwell times (s) d P = Alighting passengers per bus through the busiest door (p) a t = Alighting passenger service time (s/p) a P = Boarding passengers per bus through the busiest door (p) b t = Boarding passenger service time (s/p) b t = Door opening and closing times (s) oc
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
Sumeet Jaiswal Page 20 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 + 0.4098AM + 0.6777PM 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
Sumeet Jaiswal Page 21 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 22 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.
Equation 2.14 1.56 0.16 0.09
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
Sumeet Jaiswal Page 23 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 24 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.
Sumeet Jaiswal Page 25 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 26 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,
Sumeet Jaiswal Page 27 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 28 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).
Sumeet Jaiswal Page 29 Busway Platform Bus Capacity Analysis
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 = M Equation 2.17
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).
Sumeet Jaiswal Page 30 Literature Review
A v v Speed, Speed,
Density, k A/B Flow, q A2/4B (a) (b)
A2/4B q Flow, Flow,
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.
(m/min) Walking Speed, V
2 D (person /m ) 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.
Sumeet Jaiswal Page 31 Busway Platform Bus Capacity Analysis (m/min) Walking Speed, V
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.
) m/min ( eed p s g Walkin
2 Pedestrian space (m /p) Source: TRB, 2003 Figure 2.11: Pedestrian speed on walkways
Sumeet Jaiswal Page 32 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
Sumeet Jaiswal Page 33 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 34 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.
Sumeet Jaiswal Page 35 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 36 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 ⁄
Sumeet Jaiswal Page 37 Busway Platform Bus Capacity Analysis
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)
Sumeet Jaiswal Page 38 Literature Review
td = Average dwell times (s) = 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.
Sumeet Jaiswal Page 39 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 40 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 Photo by: Jaiswal (2009) 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
Sumeet Jaiswal Page 41 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 42 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.
Sumeet Jaiswal Page 43
Research Problem Development
Chapter Three
Research Problem Development
3.1 Overview Previous chapter identified the lack of a purpose built methodology for busway dwell time and capacity estimation. The shortcomings of using the traditional bus stop dwell time methodology for busway station operation were described. Therefore it is necessary to first study the operation of a bus stop and a busway station and identify the differences between them. To start with, preliminary visual observations were carried out at some of Brisbane’s busway stations including Cultural Centre, South Bank, Mater Hill and Eight Mile Plains. Similarly, two arterial road bus stops in Brisbane, at Coorparoo West and Mt. Gravatt Showground, were observed.
The aim of this exercise is to develop the research problem from the gaps established in Chapter 2 and identify the parameters needed to model bus dwell time at a busway station. This chapter describes the analysis of preliminary observations. Section 3.2 presents the difference established here between busway station and bus stop vis-à-vis passenger boarding process and operation. The research problem is conceptualised in Section 3.3. Section 3.4 describes the busway operation at four levels: platform, vehicle, station, and line. The definition of terms identified in this research is provided in section 3.5 and section 3.6 concludes the chapter.
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.
Sumeet Jaiswal Page 45 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 46 Research Problem Development
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.
Sumeet Jaiswal Page 47 Busway Platform Bus Capacity Analysis
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 Figure 3.2: Origin and destination of a trip segmentO 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
Sumeet Jaiswal Page 48 Research Problem Development
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
Sumeet Jaiswal Page 49 Busway Platform Bus Capacity Analysis
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).
Photo by: Jaiswal (2009) Figure 3.4: A passenger – bus interface phase at a busway station
Photo by: Jaiswal (2009) Figure 3.5: A bus lost time phase at a busway station
Sumeet Jaiswal Page 50 Research Problem Development
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.
Sumeet Jaiswal Page 51 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 52 Research Problem Development
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.
Sumeet Jaiswal Page 53
Data Collection and Processing
Chapter Four
Data Collection and Processing
4.1 Overview This chapter presents the data collection methodology used in this research to capture the effect of bus lost time on bus dwell time. The chapter starts with a discussion on different data collection techniques used in past research studies. This was followed by section 4.3 which provides details of data collection technique used in this research. Section 4.4 outlines the various methodologies applied for the purpose of this research. The selection criteria of study station and characteristics of selected busway station are discussed in section 4.5 and section 4.6 respectively. Section 4.7 details data collection sequence and section 4.8 discusses about processing of the collected data.
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.
Sumeet Jaiswal Page 55 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 56 Data Collection and Processing
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
Sumeet Jaiswal Page 57 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 58 Data Collection and Processing
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,
Sumeet Jaiswal Page 59 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 60 Data Collection and Processing
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.
Sumeet Jaiswal Page 61 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 62 Data Collection and Processing
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.
Sumeet Jaiswal Page 63 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 64 Data Collection and Processing
Stat Stationion entry/ entry/ existexit point p oint Bus direction L A Loading area
INBOUND PLATFORM
Lead Stop L A 3 L A 2 Tunnel
Lead Stop Tunnel L A 2 L A 3
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
Sumeet Jaiswal Page 65 Busway Platform Bus Capacity Analysis
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
400
300
200
Number of passengers 100
0 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 9-10 am 12-1 pm 9-10 pm 10-11 am 11-12 am Time of day 10-11 pm 11-12 pm Source: TransLink, 2007 Figure 4.5: Boarding and alighting at inbound platform of Mater Hill Busway Station
Sumeet Jaiswal Page 66 Data Collection and Processing
Alighting Boarding 900 800 700 600 500 400 300 200 Number of passengers 100 0 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 9-10 am 12-1 pm 9-10 pm 10-11 am 11-12 am Time of day 10-11 pm 11-12 pm Source: TransLink, 2007 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.
Sumeet Jaiswal Page 67 Busway Platform Bus Capacity Analysis
140
120
100
80
60
Number of Buses 40
20
0 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 9-10 am 12-1 pm 9-10 pm 10-11 am 11-12 am Time of day 10-11 pm 11-12 pm Source: TransLink, 2007 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
Sumeet Jaiswal Page 68 Data Collection and Processing
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.
Sumeet Jaiswal Page 69 Busway Platform Bus Capacity Analysis
Observed bus arival rate (bus/hr) 2282934384247667475757576818385109115121 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 06 6 28 24 24 39 23 23 Null set
50 7Low 7 58 22 22 60 21 21 101 20 20 105 10 10 136 19 19 138 11 11 152 13 13 Null set 159 14 14 Medium 165 12Classification split 12 171 8 8
180Corresponding hour at ending 9 9
Observed passenger (p/hr) demand 200 15 15 255 18 18 325 17 Null set Null set 17 High 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
Sumeet Jaiswal Page 70 Data Collection and Processing
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.
Sumeet Jaiswal Page 71
Parameter Analysis and Evaluation
Chapter Five
Parameter Analysis and Evaluation
5.1 Overview This chapter evaluates the two vital parameters of busway station operation. These two parameters were identified in chapter 3 as: passenger – bus interface, and bus lost time. The aim of this chapter is to analyse these parameters based on platform crowding. The outcome of this analysis will be used as inputs for modelling of bus lost times and bus dwell times at a busway station.
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
Sumeet Jaiswal Page 73 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 74 Parameter Analysis and Evaluation
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.
Sumeet Jaiswal Page 75 Busway Platform Bus Capacity Analysis
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 Lower spread High spread 35 Low crowding High crowding 30
25
20
15
10
5
0
Avg duration of passenger -of passenger duration Avg (s) interface bus 0 1020304050607080 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,
Sumeet Jaiswal Page 76 Parameter Analysis and Evaluation
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.
Sumeet Jaiswal Page 77 Busway Platform Bus Capacity Analysis
LA 1 LA 2 LA 3 Linear (LA 1) Linear (LA 2) Linear (LA 3) R² = 0.0364 R² = 0.1096 R² = 0.0521 25
20
15
10
5
0
Avg duration of passenger -of passenger duration Avg (s) interface bus 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
Sumeet Jaiswal Page 78 Parameter Analysis and Evaluation
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
Sumeet Jaiswal Page 79 Busway Platform Bus Capacity Analysis
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 th ’ and n 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) st nd th P1, P2,.., Pn = Location of 1 , 2 ,…, n 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
Sumeet Jaiswal Page 80 Parameter Analysis and Evaluation
DT
Space LT IA to ’ ’ ’ ’ tc B P 1P 2 P n X 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.
Sumeet Jaiswal Page 81 Busway Platform Bus Capacity Analysis
y Direction of bus movement Loading area 1 Loading area 2 Loading area 3
dy2 dy3 d y1 Outbound platform Front Rear Where, d > d ; d ≈ d y y1 y2 y1 y3 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
Sumeet Jaiswal Page 82 Parameter Analysis and Evaluation
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. st P1 = Location of 1 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 L A 1
dy1 dy2
P1 y - y line L A 2
LT2
dy3
L A 3
B LT 3 Space Time
Figure 5.6: Effect of loading area on bus lost time
Sumeet Jaiswal Page 83 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 84 Parameter Analysis and Evaluation
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
Std dev 2.7s 2.0s 4.3s
Count 51 30 14
Note: Off – peak operation; # of passenger on platform < 30; total analysis time = 120 minutes
Sumeet Jaiswal Page 85 Busway Platform Bus Capacity Analysis
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
Std dev 3.3s 1.5s 4.0s
# 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.
Sumeet Jaiswal Page 86 Parameter Analysis and Evaluation
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.
9
8
7 Peak period Peak period LA3 6 period Off-peak Peak period onset Peak period 5 LA1 4
3
2 LA2 Mean lost time per Bus (s) lost time per Bus Mean 1
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
Sumeet Jaiswal Page 87 Busway Platform Bus Capacity Analysis
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).
Equation 5.3 td = Pata + Pbtb + toc
Where,
t = Average dwell times (s) d P = Alighting passengers per bus through the busiest door (p) a t = Alighting passenger service time (s/p) a P = Boarding passengers per bus through the busiest door (p) b t = Boarding passenger service time (s/p) b t = Door opening and closing times (s) oc
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
Sumeet Jaiswal Page 88 Parameter Analysis and Evaluation
Where,
t = Average dwell times (s) d P = Alighting passengers per bus through the busiest door (p) a t = Alighting passenger service time (s/p) a P = Boarding passengers per bus through the busiest door (p) b t = Boarding passenger service time (s/p) b t = Door opening and closing times (s) oc
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.
Sumeet Jaiswal Page 89 Busway Platform Bus Capacity Analysis
Table 5.6: Descriptive statistics
Door Minimum MaximumMean 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 Boarding time per pax Front 3.0s 7.5s 5.0s 2.3s Alighting time per pax Rear 3.5s 5.0s 4.0s 0.6s
Loading Area 3 Boarding time per pax Front 1.9s 8.0s 4.9s 4.8s Alighting time per pax Rear 2.0s 6.0s 3.6s 1.4s
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.
Sumeet Jaiswal Page 90 Parameter Analysis and Evaluation
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.
Sumeet Jaiswal Page 91 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 92 Parameter Analysis and Evaluation
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.
Sumeet Jaiswal Page 93
Modelling Bus Lost Time
Chapter Six
Modelling Bus Lost Time
6.1 Overview Having introduced in preceding chapters the concept of bus lost time and having analysed it in chapter 5, this chapter discusses the development of stochastic relationships from the observed bus lost time data set for the study busway station. The purpose of developing a stochastic model was to generalise the pattern of bus lost time variations. The generalised bus lost time can be applied to develop an improved bus dwell time model and bus capacity model for the busway station.
Before estimating the stochastic model for the observed bus lost time it was necessary to study its frequency density. This analysis was performed using a histogram technique. Details of analysis are discussed in Section 6.2. Section 6.3 estimates the bus lost time probability distribution function and Section 6.4 fits the probability distribution curve on observed bus lost time data. As found from the analysis presented in Section 5.5, the bus lost times depended on the particular loading area and platform crowd. Further separate probability curves were fitted to each loading area for each of the peak and off-peak periods. Section 6.5 closes the chapter and summaries the key contributions to knowledge.
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.
Sumeet Jaiswal Page 95 Busway Platform Bus Capacity Analysis
14 12 10 8
Count 6 4 2 0 < 11 2 3 4 5 6 7 8 9 10111213141516 Bus lost time a) Loading area 1
12
10
8
6 Count 4
2
0 < 112345678910111213141516 Bus lost time b) Loading area 2
10 9 8 7 6 5
Count 4 3 2 1 0 <112345678910111213141516 Bus lost time c) Loading area 3 Figure 6.1: Off-peak period bus lost time histogram
Sumeet Jaiswal Page 96 Modelling Bus Lost Time
14 12 10 8 6 Count 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 8 Count 6 4 2 0 <1123456789101112131415161718 Bus lost time b) Loading area 2
8 7 6 5 4
Count 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
Sumeet Jaiswal Page 97 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 98 Modelling Bus Lost Time
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.
Sumeet Jaiswal Page 99 Busway Platform Bus Capacity Analysis
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
Table 6.2: Descriptive statistics of loading area 2 (Off-peak period)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 30 30
Mean 2.883 0.363 0.859
95% confidence Lower bound 2.140 0.616 interval for mean Upper bound 3.626 1.102
Median 2.000 0.691
Variance 3.960 0.424
Std. deviation 1.990 0.650
Skewness 1.911 0.427 -0.136 0.427
Kurtosis 4.823 0.833 0.303 0.833
Sumeet Jaiswal Page 100 Modelling Bus Lost Time
Table 6.3: Descriptive statistics of loading area 3 (Off-peak period)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 14 14
Mean 7.500 1.142 1.820 0.192
95% confidence Lower bound 5.032 1.404 interval for mean Upper bound 9.968 2.236
Median 7.500 2.013
Variance 18.269 0.519
Std. deviation 4.274 0.721
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)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 61 61
Mean 4.066 0.415 1.095 0.107
95% confidence Lower bound 3.235 0.881 interval for mean Upper bound 4.896 1.309
Median 3.000 1.098
Variance 10.512 0.697
Std. deviation 3.242 0.834
Skewness 1.692 0.306 -0.362 0.306
Kurtosis 3.497 0.604 -0.229 0.604
Sumeet Jaiswal Page 101 Busway Platform Bus Capacity Analysis
Table 6.5: Descriptive statistics of loading area 2 (Peak period)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 50 50
Mean 2.850 0.244 0.888 0.0823
95% confidence Lower bound 2.359 0.723 interval for mean Upper bound 3.341 1.054
Median 3.000 1.098
Variance 2.982 0.339
Std. deviation 1.726 0.582
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)
Raw data (X) Log [X]
Statistic Std. Error Statistic Std. Error
Count 38 38
Mean 4.868 0.636 1.259 0.141
95% confidence Lower bound 3.578 0.973 interval for mean Upper bound 6.158 1.545
Median 4.000 1.386
Variance 15.401 0.758
Std. deviation 3.924 0.870
Skewness 1.515 0.383 -0.440 0.383
Kurtosis 2.223 0.750 -0.097 0.750
Sumeet Jaiswal Page 102 Modelling Bus Lost Time
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.
Sumeet Jaiswal Page 103 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 104 Modelling Bus Lost Time 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)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error 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
Table 6.9: Assessing normality for loading area 2 (Off-peak period)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error Skewness 1.911 0.427 -0.136 0.427 95% confidence Lower bound 1.074 -0.973 interval for skewness Upper bound 2.748 0.701 Kurtosis 4.823 0.833 0.303 0.833 95% confidence Lower bound 3.190 -1.330 interval for Kurtosis Upper bound 6.456 1.936
Table 6.10: Assessing normality for loading area 3 (Off-peak period)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error Skewness 0.541 0.597 -1.144 0.597 95% confidence Lower bound -0.629 -2.314 interval for skewness Upper bound 1.711 0.026 Kurtosis -0.309 1.154 1.980 1.154 95% confidence Lower bound -2.571 -0.282 interval for Kurtosis Upper bound 1.953 4.242
Sumeet Jaiswal Page 105 Busway Platform Bus Capacity Analysis
Table 6.11: Assessing normality for loading area 1 (Peak period)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error Skewness 1.692 0.306 -0.362 0.306 95% confidence Lower bound 1.092 -0.962 interval for skewness Upper bound 2.292 0.238 Kurtosis 3.497 0.604 -0.229 0.604 95% confidence Lower bound 2.313 -1.413 interval for Kurtosis Upper bound 4.681 0.955
Table 6.12: Assessing normality for loading area 2 (Peak period)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error Skewness 2.020 0.337 -0.307 0.337 95% confidence Lower bound 1.359 -0.968 interval for skewness Upper bound 2.681 0.354 Kurtosis 6.140 0.662 0.575 0.662 95% confidence Lower bound 4.842 -0.723 interval for Kurtosis Upper bound 7.438 1.873
Table 6.13: Assessing normality for loading area 3 (Peak period)
Raw data (X) Log [X] Statistic Std. Error Statistic Std. Error Skewness 1.515 0.383 -0.440 0.383 95% confidence Lower bound 0.764 -1.191 interval for skewness Upper bound 2.266 0.311 Kurtosis 2.223 0.750 -0.097 0.750 95% confidence Lower bound 0.753 -1.567 interval for Kurtosis Upper bound 3.693 1.373
Sumeet Jaiswal Page 106 Modelling Bus Lost Time
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
µ ⁄ X e Equation 6.3
µ ⁄ X e 1 Equation 6.4
Std. deviation of X = X Equation 6.5
Sumeet Jaiswal Page 107 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 108 Modelling Bus Lost Time
0.35 LA 1 (Peak)
LA 2 (Peak) 0.30 LA 3 (Peak)
0.25
f(x) 0.20
0.15 Probabilty, Probabilty,
0.10
0.05
0.00 0.512345678910111213141516171819 Lost time (x)
Figure 6.3: Bus lost time probability distribution curves (Peak period)
Note: LA = loading area
Sumeet Jaiswal Page 109 Busway Platform Bus Capacity Analysis
0.35 LA 1 (OP)
LA 2 (OP) 0.30
0.25
f(x) 0.20
0.15 Probabilty, Probabilty,
0.10
0.05
0.00 0.512345678910111213141516171819 Lost time (x)
Figure 6.4: Bus lost time probability distribution curves (Off-peak period)
Note: LA = loading area; OP = Off-peak
Sumeet Jaiswal Page 110 Modelling Bus Lost Time
0.35 LA 1 (Peak) LA 2 (Peak) 0.30 LA 3 (Peak) LA 1 (OP) 0.25 LA 2 (OP)
f(x) 0.20
0.15 Probabilty, Probabilty,
0.10
0.05
0.00 0.512345678910111213141516171819 Lost time (x)
Figure 6.5: Comparison of peak and off-peak bus lost time probability distribution curves
Note: LA = loading area; OP = Off-peak
Sumeet Jaiswal Page 111 Busway Platform Bus Capacity Analysis
1
0.9 th 85 percentile
0.8
0.7
0.6 LA 1 (Peak)
0.5 LA 2 (Peak) 0.4
Cumulative probabilty Cumulative probabilty LA 3 (Peak) 0.3
0.2
0.1
0 0.512345678910111213141516171819202122232425 Bus lost time
Figure 6.6: Bus lost time cumulative distribution curves (Peak period)
Note: LA = loading area
Sumeet Jaiswal Page 112 Modelling Bus Lost Time
1
0.9 th 85 percentile
0.8
0.7
0.6
0.5 LA 1 (OP) LA 2 (OP) 0.4
Cumulative probabilty Cumulative probabilty 0.3
0.2
0.1
0 0.512345678910111213141516171819202122232425 Bus lost time
Figure 6.7: Bus lost time cumulative distribution curves (Off-peak period)
Note: LA = loading area; OP = Off-peak
Sumeet Jaiswal Page 113 Busway Platform Bus Capacity Analysis
1
0.9
0.8
0.7
0.6 LA 1 (Peak) LA 2 (Peak) 0.5 LA 3 (Peak) 0.4 LA 1 (OP)
Cumulative probabilty Cumulative probabilty 0.3 LA 2 (OP)
0.2
0.1
0 0.512345678910111213141516171819202122232425 Bus lost time
Figure 6.8: Comparison of peak and off-peak bus lost time cumulative distribution curves
Note: LA = loading area; OP = Off-peak
Sumeet Jaiswal Page 114 Modelling Bus Lost Time
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.
Sumeet Jaiswal Page 115 Busway Platform Bus Capacity Analysis
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 85th 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
Sumeet Jaiswal Page 116 Modelling Bus Lost Time
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.
Sumeet Jaiswal Page 117
Busway Station Dwell Time Model
Chapter Seven
Busway Station Dwell Time Model
7.1 Overview Previous chapters established the bus lost time variable as a significant parameter influencing bus dwell time at a busway station platform. Along with the traditional variables of the number of boarding and alighting passengers and respective marginal time and door opening and closing time, bus lost time is also a determinant of dwell time of a bus at a busway station.
This chapter describes the development of the busway station bus dwell time model, incorporating the concept of bus lost time. Section 7.2 describes the framework of the model. The mathematical model is presented in section 7.3. Section 7.4 discusses the effects of bus lost time on bus dwell time estimation. Section 7.5 summaries the key features of the new model and discusses model implications. Section 7.6 closes the chapter.
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
Sumeet Jaiswal Page 119 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 120 Busway Station Dwell Time Model
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, th = Bus dwell time at n 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) th = Bus lost time at n 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
Sumeet Jaiswal Page 121 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 122 Busway Station Dwell Time Model
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.
Sumeet Jaiswal Page 123 Busway Platform Bus Capacity Analysis
60
50 tb = 2s
40
30 tb = 4s
20
10 tb = 5s Contribution of lost time in dwell (%) of lost Contribution 0 12345678910 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.
Sumeet Jaiswal Page 124 Busway Station Dwell Time Model
60
t = 2s 50 b
40
tb = 4s 30
20
tb = 5s 10 Contribution of lost time in dwell (%) of lost Contribution
0 12345678910 Boarding load per bus Figure 7.3: Effect of bus lost time on dwell time at loading area 1 (Off-peak period)
45
40
tb = 2s 35
30
25
tb = 4s 20
15
10 tb = 5s
Contribution of lost time in dwell (%) of lost Contribution 5
0 12345678910 Boarding load Figure 7.4: Effect of bus lost time on dwell time at loading area 2 (Peak period)
Sumeet Jaiswal Page 125 Busway Platform Bus Capacity Analysis
50
45 tb = 2s 40
35
30
25 tb = 4s
20
15
10 tb = 5s
5 Contribution of lost time in dwell (%) of lost Contribution 0 12345678910 Boarding load Figure 7.5: Effect of bus lost time on dwell time at loading area 2 (Off-peak period)
60
tb = 2s 50
40
tb = 4s 30
20
tb = 5s 10 Contribution of lost time in dwell (%) of lost Contribution
0 12345678910 Boarding load Figure 7.6: Effect of bus lost time on dwell time at loading area 3 (Peak period)
Sumeet Jaiswal Page 126 Busway Station Dwell Time Model
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.
Sumeet Jaiswal Page 127 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 128 Busway Loading Area Bus Capacity Model
Chapter Eight
Busway Loading Area Bus Capacity Model
8.1 Overview This chapter describes the development of busway loading area bus capacity model. This new capacity model incorporates the new busway dwell time model, presented in chapter 7, which in turn incorporated the bus lost time variable, described in chapter 6.
The next section portrays the approach to model development and discusses the variables influencing the busway loading area bus capacity. Section 8.3 presents the busway loading area capacity model. Section 8.4 presents the model for effective loading area bus capacity and Section 8.5 presents the model for busway station bus capacity.
Section 8.6 demonstrates calculations for an example case and discusses the effects of bus lost time on loading area bus capacity. Section 8.7 provides a discussion on the new model. The chapter closes with section 8.8.
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
Sumeet Jaiswal Page 129 Busway Platform Bus Capacity Analysis
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)
td = Average dwell times (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.
Sumeet Jaiswal Page 130 Busway Loading Area Bus Capacity Model
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 = Clearance time (s)
The Equation 8.3 is based on the assumption that the dwell times are normally distributed. Table 8.1 reproduces the Table 2.4, which provides the value of ‘z’ corresponding to desired failure fate.
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
Sumeet Jaiswal Page 131 Busway Platform Bus Capacity Analysis
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,
th , = Operating margin for passenger service time variability (s) for n loading area = Standard normal variable corresponding to a desired failure rate
, = Coefficient of variation of passenger service time at busway loading area th = Busway dwell time at n loading area (s), as defined in chapter 7. th = Average bus lost time at n 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
Sumeet Jaiswal Page 132 Busway Loading Area Bus Capacity Model
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
Sumeet Jaiswal Page 133 Busway Platform Bus Capacity Analysis
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,
th , = Lost time operating margin for n loading area (s)
= Standard normal variable corresponding to a desired failure rate
th & = Lost time probability curve parameters for n 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 , ,
Sumeet Jaiswal Page 134 Busway Loading Area Bus Capacity Model
Where,
th = Bus capacity of n loading area (bus/hr) 3600 = Number of seconds in 1 hour th = Bus dwell time at n loading area (s)
, = Operating margin for passenger load variability (s) th , = Operating margin for bus lost time variability at n loading area (s)
= Clearance time (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
Sumeet Jaiswal Page 135 Busway Platform Bus Capacity Analysis
, Equation 8.12
Where,
th , = Effective bus capacity of n loading area (Bus/ hr) th = Bus capacity of n loading area (Bus/ hr) th = Efficiency factor of n 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) th , = Effective capacity of n 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%
Sumeet Jaiswal Page 136 Busway Loading Area Bus Capacity Model
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.
Sumeet Jaiswal Page 137 Busway Platform Bus Capacity Analysis
500
450
400 TCQSM 350
300 BSLC 250
200
150
100 Busway station capacity (Bus / hr) capacity (Bus / Busway station 50
0 12345678910 Boarding load per bus
Note: busway station with three linear loading area, no alighting load, tb=4s, failure rate 7.5%, cv=0.6, tc=10s. Figure 8.2: Variation in busway station bus capacity with boarding load per bus
40
35 tb = 2s 30
25
20
15
tb = 4s 10
5 Effect of bus lost time on station capacity 0 12345678910 Boarding load per bus
Note: busway station with three linear loading area, no alighting load, failure rate 7.5%, cv=0.6, tc=10s. Figure 8.3: Effect of bus lost time on busway station bus capacity
Sumeet Jaiswal Page 138 Busway Loading Area Bus Capacity Model
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
200 g/C = 0.5
150
Station bus capacity 100
50
0 12345678910 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.
Sumeet Jaiswal Page 139 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 140 Busway Station Efficiency Model
Chapter Nine
Busway Station Efficiency Model
9.1 Overview Under a perfectly ideal situation each loading area of a busway (BRT) station platform should operate without interfering with the others’ operation. However in a linear arrangement of loading areas this may not be the case, particularly during the peak period operation when the bus flow is high. Every loading area can potentially obstruct the bus entry to the platform (as buses are not permitted to slip into downstream, unoccupied loading areas) and obstruct exiting from upstream loading areas. This reduces the number of effective loading areas for the station from the physical number, as noted by TCQSM (TRB, 2003).
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).
Sumeet Jaiswal Page 141 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 142 Busway Station Efficiency Model
Platoon departure from platform. eriod Platoon arrival p at station. eak p
Platoon arrival at platform.
Loading area 3 blocking
Start of afternoon 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)
Sumeet Jaiswal Page 143 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 144 Busway Station Efficiency Model
Scenario Diagram Nel
1 Loading area 1 Loading area 2 Loading area 3 3 Outbound platform
2 Loading area 1 Loading area 2 Loading area 3 2 Outbound platform
3 Loading area 1 Loading area 2 Loading area 3 2 Outbound platform
4 Loading area 1 Loading area 2 Loading area 3 2 Outbound platform
5 Loading area 1 Loading area 2 Loading area 3 1 Outbound platform
6 Loading area 1 Loading area 2 Loading area 3 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.
Sumeet Jaiswal Page 145 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 146 Busway Station Efficiency Model
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 ,
Sumeet Jaiswal Page 147 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 148 Busway Station Efficiency Model
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 nd 2 T3 = 814s T2,b = 82s 0.90 1.90 st 1 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.
Sumeet Jaiswal Page 149 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 150 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
Sumeet Jaiswal Page 151 Busway Platform Bus Capacity Analysis
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
Sumeet Jaiswal Page 152 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.
Sumeet Jaiswal Page 153 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 154 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
Sumeet Jaiswal Page 155 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 156 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.
Sumeet Jaiswal Page 157 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 158 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.
Sumeet Jaiswal Page 159 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 160 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/.
Sumeet Jaiswal Page 161 Busway Platform Bus Capacity Analysis
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.
Sumeet Jaiswal Page 162 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…
Sumeet Jaiswal Page 163 Busway Platform Bus Capacity Analysis
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)
Sumeet Jaiswal Page 164 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 Loading area 2 E2 = 0.90 Loading area 3 E3 = 1.00 13 Parameter, µ, for bus lost time (Table 6-14) = Loading area 1 µ1 = 1.095 Loading area 2 µ2 = 0.888 Loading area 3 µ3 = 1.259 14 Parameter, σ, for bus lost time (Table 6-14) = Loading area 1 σ1 = 0.834 Loading area 2 σ2 = 0.582 Loading area 3 σ3 = 0.870 15 Mean lost time (Table 6-14) = Loading area 1 LT1 = 4.2 Loading area 2 LT2 = 2.9 Loading area 3 LT3 = 5.2
Sumeet Jaiswal Page 165 Busway Platform Bus Capacity Analysis
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 For loading area 2 = 32.9s For loading area 3 = 35.2s
Sumeet Jaiswal Page 166 Appendix- B
ii. Calculate operating margin for passenger service time using Equation 8.4 (on page 132)
, ,
Loading area Operating margin for passenger service For loading area 1 = 25.9s For loading area 2 = 25.9s For loading area 3 = 25.9s
iii. Calculate operating margin for bus lost time using Equation 8.9 (on page 134) ⁄ ,
Loading area Operating margin for bus lost time For loading area 1 = 5.7s For loading area 2 = 2.7s For loading area 3 = 7.2s
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 For loading area 2 = 50 bus/hr For loading area 3 = 45 bus/hr
v. Calculate effective loading area bus capacity using Equation 8.12 (on page 136)
,
Loading area Effective bus capacity For loading area 1 = 33 bus/hr For loading area 2 = 45 bus/hr For loading area 3 = 45 bus/hr
vi. Calculate platform bus capacity using Equation 8.13 (on page 136)
,
= 123 bus per hour
Sumeet Jaiswal Page 167
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.
Sumeet Jaiswal Page 169 Busway Platform Bus Capacity Analysis
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).
Sumeet Jaiswal Page 170