Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Performance and Design of Taxi Services at Airport Passenger Terminals
David Carvalho Teixeira da Costa
Dissertação para obtenção do Grau de Mestre em
Sistemas Complexos de Infraestruturas de Transportes (CTIS)
Júri
Presidente: Prof. Dr. José Manuel Viegas
Coordenador: Prof. Dr. Richard de Neufville
Co-coordenador: Prof. Dr. Rosário Macário
Vogal: Prof. Dr. João Claro
Outubro, 2009
2
Acknowledgements
Throughout this academic adventure, there were many times when I wondered on the actual
relevance of the result of my work or my ability to finish it. It’s only normal to lose faith at certain
moments during the writing of a thesis, but I never lost mine, fortunately. Self-motivation is a key
instrument for researchers who are forced to semi-isolate themselves and dive into the complex work
they perform, and I’ve learned that very well. Although one should always try to think ahead,
regardless of the difficulties, it was not always easy to identify that line of thought that eventually
drives your work to success, and the help of our supervisors is like a light in the dark. Their role in this
process was fundamental, not only to guide and keep motivation running, but also to teach, and most
certainly to learn as well.
I’d like to thank Professor Richard de Neufville for his rich and enlightening coordination, his
unyielding patience and his faith on my capacity to actually work with him on this topic. For me, it was
an honor to work with such a brilliant academic mind and to be able to tap into several hidden truths of
engineering. Through his classes, through his coordination and teachings and through his own unique
perspective, I have surely become a better engineer, better prepared for my professional life. I’d also
like to thank Professor Rosario Macário, which has co-supervised my thesis and has always been
there whenever I needed closer guidance on several issues, especially regarding the regulation and
institutional parts.
This work implied a substantial effort to plan and execute data gathering initiatives. For several
days, I had to travel to the airport and spend some hours observing how queues behaved. For this I
needed help and help I got. Teixeira’s help in this phase was magnificent and I thank him for the
courage, friendship and sacrifice of standing by my side at the airport, in the middle of August, during
almost a full day. In an environment where the observed people did not like to be observed, a
significant effort had to be made to disguise my presence. Unfortunately, this was not always easy - a
very rich experiment for me. Also, I’d like to thank Mauro, my tireless friend and colleague, who has
spent the last year fighting for the same objectives as myself. For all the sleepless nights spent at IST,
for all the obstacles we’ve overcome…we make a great team and you know I admire you!
To all my friends and family, my ex-colleagues at Engimind and Lisbon Municipality and my
professors at CTIS, my huge thank you. Couldn’t have done it without you!
3
Abstract
The main objective of this dissertation is to analyze, through the consideration of a case-study -
Portela Airport - the current operational and regulatory design options in systems of taxi service
provision at airport passenger buildings, and propose, based on its performance levels, alternative
schemes and possible interventions that can improve the existing services.
On the regulatory side, the methodology chosen to pursue these objectives was based on the
systematic analysis of the involved stakeholders, their institutional roles and power-sharing
mechanisms. On the operational side, an extensive data collection effort was performed and used to
calibrate a simulation model which represents system behavior. Both of these analyses were then
subject to a scenario-building process, in order to test different stimulus for both perspectives.
As main conclusions, it must be stated that the current taxi service system at Terminal 1 is not
able to adequately cope with peak-hour solicitations and offer good quality of service to passengers at
these times. Queues are a fundamental part of the problem and their behavior must not be diluted in
average-based analysis that do not expose the frailties of the system at peak-hours, some of them
intensified by seemingly small exogenous factors such as police coordination or taxi maneuvering
needs. They may also be a key part of the solution, as slight physical rearrangement of queues or
service areas can lead to greatly improved service as regards queue length, delays and reliability.
ANA and Lisbon Municipality should thus behave proactively to face this problem.
Key Words:
Taxi Services
Queuing Systems
Airports
Regulation
Simulation
Simul8
4
Resumo
O objectivo desta dissertação é analizar, através da consideração dum caso de estudo –
Aeroporto da Portela – as actuais opções de design operacional e regulatório em sistemas de táxi em
Terminais de Aeroportos, e propor, baseado nos níveis de performance, esquemas alternativos e
possíveis intervenções que permitam melhorar os serviços existentes.
Do lado regulatório, a metodologia escolhida para alcançar estes objectivos foi baseada na análise
sistemática dos stakeholders, seus papeis institucionais e mecanismos de partilha de poder. Do lado
operacional, um esforço extensivo de recolha de dados foi efectuado e usado para calibrar um
modelo de simulação que representa o comportamento sistémico. Ambas as análises foram sujeitas a
um processo de construção de cenários, de forma a testar diferentes estímulos.
Como principais conclusões, sublinha-se que o actual serviço de táxis no Terminal 1 não é capaz
de adequadamente lidar com solicitações de período de ponta e oferecer boa qualidade de serviço
aos passageiros. As filas de espera são parte fundamental do problema e o seu comportamento não
deve ser diluído em análises baseadas em médias que não expõem as fragilidades do sistema nos
períodos de ponta, intensificadas por factores aparentemente pequenos e exógenos como
coordenação policial ou necessidade de manobras dos taxis. As filas de espera podem ser parte da
solução, já que ligeiras reorganizações destas ou das áreas de serviço podem levar a grandes
melhorias no que toca à sua dimensão, tempos de espera e fiabilidade. A ANA e o Município de
Lisboa devem comportar-se proactivamente para enfrentar este problema.
Palavras-chave:
Serviços de Táxi
Sistemas de Filas de Espera
Aeroportos
Regulação
Simulação
Simul8
5
Table of Contents
Acknowledgements ............................................................................................................. 2
Abstract ................................................................................................................................ 3
Resumo ................................................................................................................................. 4
Table of Contents ................................................................................................................. 5
Table of Figures ................................................................................................................... 7
Acronyms ............................................................................................................................. 9
Chapter 1 – Introduction .................................................................................................... 10
1.1. Objectives and motivation ......................................................................................... 10
1.2. Thesis Structure ........................................................................................................ 11
1.3. State of the Art .......................................................................................................... 12
1.3.1. Sector Framework ........................................................................................... 12
1.3.2. Literature Review ............................................................................................ 14
1.3.2.1. Regulation ................................................................................................ 14
1.3.2.2. Modeling, Queuing Theory and Simulation ............................................... 16
Chapter 2 – Problem definition and methodology ........................................................... 22
2.1. Regulatory and Institutional Issues ........................................................................... 22
2.1.1. Problem definition ........................................................................................... 22
2.1.2. Methodology ................................................................................................... 24
2.2. Operational Issues .................................................................................................... 26
2.2.1. Problem definition ........................................................................................... 26
2.2.2. Methodology ................................................................................................... 28
2.2.3. Field Data Collection Plan ............................................................................... 30
Chapter 3 – Case Study – Portela Airport ......................................................................... 36
3.1. General background ................................................................................................. 36
3.1.1. Introduction ..................................................................................................... 36
3.1.2. The airport ...................................................................................................... 37
3.2. Analysis of the Taxi Service at Terminal 1 ................................................................ 40
3.2.1. Regulatory and Institutional Context ................................................................ 40
3.2.1.1. General Market Characteristics ................................................................ 41
3.2.1.2. General Licensing .................................................................................... 42
3.2.1.3. Regulation ................................................................................................ 44
3.2.1.4. Main Stakeholders and bargaining power ................................................ 45
3.2.1.5. Institutional framework ............................................................................. 49
3.2.1.6. Overview summary .................................................................................. 50
3.2.2. Operational Context ........................................................................................ 53
3.2.2.1. Spatial description of the system .............................................................. 54
6
3.2.2.2. Analysis of the Collected Data ................................................................. 58
3.2.2.3. Simulation Model ..................................................................................... 67
3.3. Scenario Building ...................................................................................................... 75
3.3.1. Regulatory and Institutional Policy Actions ...................................................... 75
3.3.1.1. Policy Actions Analysis ............................................................................ 75
3.3.1.2. Policy Actions Evaluation ......................................................................... 78
3.3.2. Operational Scenarios ..................................................................................... 80
3.3.2.1. Scenario Analysis .................................................................................... 80
3.3.2.2. Scenario Evaluation ................................................................................. 86
Chapter 4 – Conclusions and Proposals .......................................................................... 88
4.1. Main Conclusions ..................................................................................................... 88
4.1.1. Regulations and Institutional Framework ......................................................... 89
4.1.2. Operational Framework ................................................................................... 91
4.2. Intervention Proposals and Suggestions for Future Research ................................... 94
Bibliography ....................................................................................................................... 96
Web Sites ............................................................................................................................ 98
Annexes .............................................................................................................................. 99
I. Literature Review .................................................................................................... 100
II. Field Data ............................................................................................................... 107
III. Distribution Fitting - Inter-arrival and Service Times ................................................ 117
7
Table of Figures
Figure 1 – Two examples of Private Service Paratransit services, adapted from (Cervero, 1992) ....... 14
Figure 2 - Expected delay for four different levels of service capacity at an Airport: R1= capacity is 80
movements per hour; R2 = 90; R3 = 100; R4 = 110 (Odoni, 2007) ...................................................... 18
Figure 3 - Delay versus Utilization ratio (ρ) and confidence limits evolution (left) and Dependence on
Variability (Variance) of Inter-Arrival Times and of Service Times (right), adapted from (Odoni, 2007)19
Figure 4 – Graphical representation of cumulative arrivals and departures from a queue (Newell, 1982)
............................................................................................................................................................... 20
Figure 5 – Graphical representation of departure times (Newell, 1982) ................................................ 21
Figure 6 – General Institutional Framework of Airport Taxi Services .................................................... 23
Figure 7 – Components of a Basic Queuing Process at an Airport Taxi Stand ..................................... 27
Figure 8 – Passengers Traffic by Month (Source: Annual Traffic Report – ANA, 2008) ....................... 30
Figure 9 – Passenger Traffic by Day of the Week (Source: Annual Traffic Report – ANA, 2008) ........ 31
Figure 10 – Arriving Passengers during the Easter Holidays week in 2006 (Source: ANA, 2006) ....... 31
Figure 11 – Hourly distribution of Passengers and Taxis at the Airport Arrivals taxi stand (Source:
ANA, 2006)............................................................................................................................................. 31
Figure 12 – Sample Size parameters for the Service Times and Inter-Arrival Times for Groups ......... 34
Figure 13 – Data Collection Scheme ..................................................................................................... 34
Figure 14 – Terminal 2 location (Source: Virtual Earth) ........................................................................ 38
Figure 15 – Blueprints of the future New Lisbon Airport at Alcochete (Source: www.naer.pt) .............. 38
Figure 16 – Three of the main Tourism destinations from Portela Airport (marker A), in the Lisbon
Metropolitan Area, upper left: Chiado, upper right: Belém, bottom: Cascais (Source: Google Maps) .. 39
Figure 17 – Institutional framework regarding Portela’s taxi service system ......................................... 50
Figure 18 - Schematic classification of taxicab regulatory systems (Schaller, 2007) – Lisbon case, in
blue ........................................................................................................................................................ 51
Figure 19 – Lisbon Airport - Terminal 1 (Source: Google Earth) ........................................................... 55
Figure 20 – Taxi Service Organization at the Arrivals of Terminal 1 ..................................................... 56
Figure 21 – System configuration at the Arrivals Taxi Stand ................................................................. 57
Figure 22 – Histogram for Inter-Arrival Times for Groups ..................................................................... 59
Figure 23 – Histogram for Group Size ................................................................................................... 59
Figure 24 – Comparison between Group Size Proportions from the different measurements .............. 60
Figure 25 – Histograms for Service Times ............................................................................................ 61
Figure 26 – Service Time averages and standard deviations ............................................................... 61
Figure 27 – Main conflicts that can justify delays and differences in service time distributions ............ 63
Figure 28 – Arrival Curve and Exit Curve based on the collected data ................................................. 64
Figure 29 – Queue Length evolution ..................................................................................................... 65
Figure 30 – In-Queue Waiting Time evolution ....................................................................................... 65
Figure 31 – Main results for Queue Length, In-Queue Waiting Time and Arrival/Service Ratio ........... 66
8
Figure 32 – Final System Configuration for the current situation at Portela’s Arrivals Taxi Stand ........ 70
Figure 33 – Final result comparison between the Empirical Process and the Simulation Model .......... 73
Figure 34 – Queue Evolution for the Empirical (above) and Simulation (below) Methods .................... 74
Figure 35 – SWOT analysis for Policy Action I – Introduction of Taxi Sharing ...................................... 79
Figure 36 - SWOT analysis for Policy Action II – Introduction of a Special Airport Fleet and Concession
changes .................................................................................................................................................. 79
Figure 37 - SWOT analysis for Policy Action III - Market segmentation and other changes to the
Departures Taxi Stand ........................................................................................................................... 79
Figure 38 – System configuration for Scenario I – Extra Service Lane, 2 extra servers ....................... 80
Figure 39 – Main SIMUL8 results for the Scenario I system configuration ........................................... 81
Figure 40 - System configuration for Scenario II – One service lane, multiple servers ......................... 82
Figure 41 – Main SIMUL8 results for the Scenario II system configurations ......................................... 83
Figure 42 - System configuration for Scenario III – 2 queues, Special Service Type ........................... 84
Figure 43 - Main SIMUL8 results for the Scenario III system configuration .......................................... 85
Figure 44 – Results for the main Queue Size indicators ....................................................................... 86
Figure 45 – Results for the main Queuing Time indicators ................................................................... 86
Figure 46 - Schematic classification of taxicab regulatory systems (Schaller, 2007) .......................... 101
Figure 47 - Key characteristics of entry-related policies - adapted from (Schaller, 2007) ................... 102
Figure 48 – Customer Wait Time versus Number of Taxis (Li, 2006) ................................................. 105
Figure 49 – Mean customer demand by time of the day (Curry, 1977) ............................................... 106
Figure 50 – Collected Inter-Arrival Times ............................................................................................ 114
Figure 51 – Collected Service Times ................................................................................................... 116
Figure 52 – Exponential Theoretical Distribution fitting to the Inter-Arrival Times experimental
distribution ............................................................................................................................................ 117
Figure 53 – Goodness of fit and descriptive statistics summary for the Inter-Arrival Times............... 117
Figure 54 – Lognormal theoretical distribution fitting to the Service Times experimental distribution. 118
Figure 55 - Goodness of fit summary and descriptive statistics for the Service Times ....................... 118
Figure 56 - Lognormal theoretical distribution fitting to the Service Times experimental distribution
(Scenario II).......................................................................................................................................... 119
Figure 57 - Goodness of fit summary and descriptive statistics for the Service Times (Scenario II) .. 119
9
Acronyms
ASAE - Food and Economic Safety Authority
ANA - Aeroportos e Navegação Aérea
ATL - Lisbon Tourism Association
CAA - Civil Aviation Authority
CAP - Professional Aptitude Certificate
DECO - Consumer's Defense Association
FIFO - First-In First-Out
GIS - Geographical Information system
GPS - Global Positioning System
ID - Identification
IMTT - Institute for Mobility and Land Transportation
LIFO - Last-In First-Out
NAL - New Lisbon Airport
OECD - Organisation for Economic Co-operation and Development
SWOT - Strengths, Weaknesses, Opportunities, Threats
TAP - Transportes Aéreos Portugueses
TGV - Train à Grande Vitesse (High Speed Train)
TX - Texas
UK - United Kingdom
U.S. - United States
10
Chapter 1 – Introduction
1.1. Objectives and motivation
Airports are nowadays multimodal, multi-service platforms, with intense non-aeronautical activities
that cover several different industrial, economical and social areas. Taxi services are a fundamental
piece of the transportation diversity an airport requires, in order to become attractive and efficient.
They provide quick, comfortable, door-to-door service alternatives to a lot of passengers who wish to
travel within, to and from the “airport city” and, in some cases, they are even the only transport means
to connect the airport with the populated areas.
The main objective of this dissertation is to analyze, through the consideration of a case-study -
Portela Airport - the current operational and regulatory design options in systems of taxi service
provision, namely at airport passenger buildings, and propose, based on its performance levels, new
and alternative schemes and possible interventions that can improve the existing services.
This issue is often subject to intuitive thinking and political pressure when building or re-designing
of the system, many times leading to inefficiencies, excessive waiting times and/or low-quality of
service. The way taxi services are organized at the airport taxi stands also impacts the quality of
service and thresholds for efficiency gains at the terminals themselves. Reliability of service is a key
feature for passengers boarding flights at that terminal and availability and lower waiting times at the
taxi stand are important to arriving passengers who wish to quickly get to their destination after a long
trip. Flexibility is an important element to include in this framework, since airports are nowadays
dynamic infrastructures, which are often forced to grow or re-adapt to changing air transport market
conditions and, like most transportation hubs, are subject to significant peak-hour solicitations. This
aspect should be taken into account in the analysis of fleet dimensioning, regulatory aspects, service
types, queue design, depot and queue capacity options, etc. Most features of this analysis will be
addressed by the implementation of a simulation model for the queuing system at an Airport Terminal
taxi stand, based on real data collected in situ.
In recent years, Portuguese society has been intensely discussing the construction of a new
International Airport for the city of Lisbon. After some political turmoil regarding the future location for
this important infrastructure, the government’s decision of building at Ota changed to Alcochete,
mainly due to public opinion pressure. Recently, the first drafts and initial details of the project have
been exposed to the public eye, but there is still some way to go until the first brick is in place and the
first aircraft lands at Alcochete. This major project, still in its early stages of development, opens a
window of opportunity to contribute towards the brainstorming and discussion on the design options of
the taxi service at the new airport and provides an excellent candidate for a case study – the soon-to-
be-replaced, Portela Airport. This opportunity, coupled with the almost absence of acknowledged
studies on airport taxi service performance and design in scientific literature, also provides this theme
with a substantial level of relevance, yet framed within a reasonable complexity context.
11
One of the main difficulties of this endeavor is related to the complexity of the relations between the
several elements that compose the system. Although modeling can provide some interesting answers
on performance indicators and operational optimization, there is the need to analyze the current
overarching regulatory and institutional structure that also influences and limits the thresholds for
gains of efficiency. This multidisciplinary nature, coupled with the inherent complexity of an airport
infrastructure, can make the problem boundaries become blurry and spread the focus of the effort,
intensifying it towards unsustainable levels. Careful focus on the critical elements of this issue and
objective-oriented time management can help minimize this risk. The other main difficulty is related to
a common bane of all researchers – data availability. Little to no recent information can be found on
taxi service indicators on airport passenger terminals and taxi companies either do not usually keep a
record of their service performance, or are very rigid about releasing it. In fact, the sector is
traditionally entrenched, very uneasy about allowing in-depth research on their operational status quo.
Early identification of reliable and critical information sources and preparation of a structured data
collection plan can function as mitigation measures regarding this risk.
The main stakeholders of this effort are firstly taxi owners, drivers and taxi operators as a whole, in
the sense that greater efficiency, and lower queuing lengths and waiting times for passengers and
taxis increases demand and number of trips - consequently revenues - and decreases operational
costs. Airport management companies also benefit from better and more efficient taxi services at their
terminal’s curbside, for increased connectivity, less probability of delays for passengers catching
flights (reliability) and less queuing problems (space efficiency) – consequently less complaints.
Passengers are also confronted with less waiting times, more service flexibility and eventually money-
saving effects, deriving from the potential implementation of different taxi service types such as the
shared-taxi, for example. Finally, there are positive externalities for society in general caused by
possible decreases in congestion and modal share increases for taxi versus private cars, due to
higher occupation rates and greater service efficiency and availability for airport passengers.
1.2. Thesis Structure
This dissertation is based on a main structure of four chapters.
Chapter One provides the general framework for the development of the thesis, starting by
identifying its objectives, opportunity and relevance. It also builds a knowledge basis for the
subsequent analysis, through background research and literature review, framing it in the regulatory,
institutional and operational context of the sector. By gathering all relevant information on previous
studies, theoretical bibliography and emerging trends, a strong knowledge background is created and
the analysis becomes richer in content.
Chapter Two is dedicated to the problem definition and the justification of the proposed
methodology for the analysis. A Data Collection Plan is built and the field work procedures are
described along with all assumptions and simplifications.
12
Chapter Three relates to the description of the Case Study – Portela. It defines the specific
institutional, regulatory and operational setting, analysis on the collected data, model characterization
and assumptions, presentation of results, scenario building and consequent discussion (SIMUL8
Software will be used as a simulation tool for this system). It is a central chapter in this dissertation as
it allows the overview of the system mechanics and defines frameworks for performance testing.
Chapter Four summarizes all the major findings of the experiment, analyses its implications,
discusses possible limitations of the study and proposes interventions for improvement of the system.
1.3. State of the Art
1.3.1. Sector Framework
Taxis are seen as a flexible, fast and convenient transportation mode, although generally
expensive for everyday commuting trips. They have long been part of the transportation mix of almost
every medium-large sized city around the World, providing everyday direct point-to-point service on
request, complementing other transport modes in the fulfillment of market needs. Taxis are also a key
feeder system to bus and rail systems, solving the “last mile” problem and a vital back-up system to
Paratransit in peak traffic hours, for example (Hartman, 2007).
This mode remains highly regulated across many countries, through instruments such as entry
restrictions, numerical limitations, permit systems, medallion systems, average pricing and exclusive
contracts. All of these concepts can be found in the literature review section and bibliographical
sources present in this study (see Annex I). This heavy regulatory environment has been challenged
in many ways, either during the early 1980’s in the U.S. or a bit later in Europe, namely in England
(Schaller, 2007), and more recently in Japan (Flath, 2006). The arguments pro liberalization are not
that different from those present at other transportation mode’s discussions, namely economic
efficiency, lower prices through open competition and technological/service innovation (OECD, 2007).
Regulation supporters claim, however, significant market failure risks, associated with asymmetry of
information, cross-subsidization, externalities, excessive market penetration by small independent
operators and economies of scope and scale within certain market contexts that can lead to
uncompetitive conditions (Schaller, 2007).
This transportation service gains special relevance when coupled to existing high-demand nodes
like hospitals, monuments, certain shopping areas, hotels, convention centers, train stations or
airports. Its “service profile” is highly compatible to the traditionally higher willingness to pay of
passengers with trip urgency, high comfort and safety needs or economic power, such as hospital
patients and visitors, shoppers, businessmen or tourists (La Croix, 1991) (Curry, 1977). Recent
surveys (Cardon, 2007) show that about 50% of the urban mobility customers prefer the taxi to travel
to the Airport, making it an almost specialized airport feeder service. Airports - and train stations, to a
certain degree - are a special case, in the sense that deplaning passengers are usually medium-high
income individuals and families (usually carrying luggage) that mostly travel for business and leisure.
13
These typically are either unfamiliar with the transportation system of the destination city or search for
a quicker and direct trip to their hotel due to travel fatigue and luggage, generally privileging taxis as
their primary choice of transportation (La Croix, 1986) (La Croix, 1991).
London City airport, for example, had about 64% of its surveyed passengers travelling on business
related motives in 2006 and about 78% of passengers can be considered middle-upper class income
individuals - According to UK’s Socio-Economic Classification System (ABC System) (CAA, 2006). In
2007, Brussels Airport also registered 34% of answers for the same motive and 44% were tourists1 -
office workers (32%), followed by management positions (22%) were the main professions among the
surveyed passengers. At Lisbon Airport, in 2000, a survey concluded that about 42% of passengers
were tourists and 28% traveled due to business motives (FCG-Parsons, 2002), while another source,
in 2006, states that 56% of the passengers of this airport had traditionally middle-high income
professions (ANA, 2006). The percentages of taxi users at the airport terminals are 44% for the
London case, 20% for Brussels and 38% for Lisbon (FCG-Parsons, 2002).
Also, according to other recent surveys (Cardon, 2007), among the main strengths of the taxi are
Driver’s behavior during trip (all surveyed countries), Driver’s accommodation and friendliness
(France), Route proposed/taken by the driver (England) and Travel time (Netherlands). Among the
main pointed weaknesses are Cost of the trip (France, England and Portugal) and Driver’s
accommodation and friendliness (Portugal). Regarding regulation, people differ about the expected
interventions, with the exception of customer costs, which everyone thinks should be lowered.
Competition is also one of the main concerns for the taxi sector. Traditionally, buses, trains and
private cars were the most significant taxi competitors for urban mobility services, but certain market
segments, such as businessmen or tourists remained loyal taxi customers. Many alternative service
types for urban mobility have emerged in recent years, similar to taxis in the sense that they are
flexible and do not follow fixed routes or schedules - this service “class” is called Paratransit. Perhaps
the most relevant and known Paratransit service is the shared-taxi, very popular in some countries,
but some others exist, including at airports.
These alternative service types may be adaptable to taxi services, integrating the business models
of the taxi operators; complementary, during peak hours mostly; or assume a strong competing
position, especially at airports. In any case, the analysis of these alternative schemes of offering
flexible mobility services can shed light on the concept of taxi services themselves. Many alternative
mobility options (carsharing, etc.) are often incorporated in the company structures of large and known
transport operators, such as bus companies, for example. The same can be said about the taxi
operators, which can learn from successful experiences in order to adapt to specific or evolving
conditions. It is certainly worthwhile to mention the possible implementation of such schemes –
focusing on shared-taxis and dial-a-ride vans (Figure 1) - within the taxi service system in this study,
especially at airport terminals, where diversity of needs is constant.
1 http://www.brusselsairport.be/en/news/newsItems/paxprofile
14
Shared-ride services, which include shared-taxis, respond immediately to travel requests made by
phone or street hail and for this they charge a premium. These modes are generally heavily regulated,
but can offer several benefits, such as direct cost savings, peak-load shedding – especially where
congestion charging schemes exist – and off-peak specialized curb-to-curb services to senior citizens,
disabled persons, and the poor. (Cervero, 1992) Dial-a-Ride vans or shuttles are shared-taxis with
greater capacity. These services - such as Super Shuttle or Prime Time in the U.S. - became a fast
growing business in terms of the shared-ride airport ground transportation market, especially in
deregulated environments, with open competition on taxis and buses (Cervero, 1992).
Figure 1 – Two examples of Private Service Paratransit services, adapted from (Cervero, 1992)
1.3.2. Literature Review
1.3.2.1. Regulation
Taxi regulation, including airport taxi contractual arrangements, has been a hot topic of discussion
in many developed countries for the better part of the last half of the 20th century. Like with many other
modes of transport, the regulatory environment decisively influences the operational performance
levels, the level of service and the market exploitation degrees of freedom. This aspect is fundamental
in the analysis of any airport taxi service system, since the interactions between the several inter-
dependent actors are complex and sometimes competing. Also, the general aspects of taxi regulation
within cities and regions might sometimes be inadequate to the airport’s operational micro-
environment, due to the specificity of the targeted market – the airport passengers.
Like many discussions on regulation and deregulation of other transport modes and sectors of
economic activity, the taxi market motivates strong disputes on economic and social efficiency, equity
and welfare maximization. On one side, fundamentals of economic theory, supporting free market
benefits such as lower prices, innovation and higher level of service, deriving from increased
competition, supported by relatively good experiences in other sectors and other modes of
transportation. On the other, imperfections in practice that many times lead to market failures, which
call for regulation (Schaller, 2007).
Liberalization supporters base their reasoning on the claim that restrictions on entry to the taxi
industry constitute an unjustified restriction on competition, while also allowing for regulatory capture.
This means that large transfers from consumers to producers might occur, along with associated
economic distortions and corresponding deadweight losses. Furthermore, this perspective defends
that no solid proof exists on the claim that equity is better promoted through the implementation of
15
entry restrictions; on the contrary, higher prices and lower availability affect lower income taxi service
consumers (DeVany, 1977) (OECD, 2007).
The availability argument is also very strong on the part of the deregulation supporters. Some
authors go as far as stating that “Studies have found that travelers are more sensitive to the
availability of taxis than they are to travel times, speeds, or almost any other service features. Where
taxis are given unrestricted freedom to ply their trade, the quality of’ urban transportation has generally
improved.” (Cervero, 1985). Numerical limits on taxis are at the center of this argument for risk of low
availability, which also focuses on the excessively high prices for medallions and permits, which have
emerged as a profitable secondary-market, due to the scarcity of new permit issuing initiatives. There
is also a strong belief that there is no economic justification to the restrictions imposed on alternative
service types, such as shared-ride and dial ride, preventing them from competing for parts of the
transit market largely monopolized by other transport operators (Frankena, 1984).
Pro-regulation supporters often point to significant risks of market failure to defend regulatory
measures on market entry access and quality of service. Among the most argued market
imperfections are the significant economies of scale and scope, which distort competition on some taxi
market segments, cross-subsidization between geographical areas and operational periods,
information asymmetry, negative externalities and oversupply (Schaller, 2007) (La Croix, 1986) (La
Croix, 1991). Some of the pro-liberalization supporters also admit to the need for certain regulation,
assuming that some of the potential market failures provide a credible theoretical rationale for certain
types of regulations, including fare ceilings, prohibition of trip refusals and regulations dealing with
vehicle safety and liability insurance (Frankena, 1984). Quality-based regulation is also seen as a
necessary complement to a desirably open entry policy, in order to maintain and support the benefits
that it generates. Competing interests between producers and users and diversity of demand patterns
across cities are also often recognized as significant issues of regulation (OECD, 2007).
A third perspective has also emerged, based on the idea that, instead of a simple choice between
regulation and deregulation, a spectrum of entry policies should be adopted. This relates to the
contextual dependence of the taxi sector regarding regulatory and operational conditions, which differ
from country to country and even between regions, as can be seen from (Gallick, 1987), (Cairns,
1996) and (Flath, 2006). According to this perspective entry restrictions and policies have different
impacts on different regulatory and economic environments and thus should be analyzed case by
case (Schaller, 2007).
Airport taxi arrangements specifically are also subject to many of the abovementioned
considerations. The contractual types present at airport taxi stand concessions are a relevant issue,
since different arrangements can lead to different impacts on service access and efficiency.
Categorization is usually defined into three types: Exclusive Contract, where a single taxi company is
granted the privilege to solicit passengers leaving the airport; Permit System, when a government
agency issues a limited number of permits to selected taxi operators to provide service and Open
System, in which any licensed taxicab in the metropolitan area is allowed to solicit passengers at the
airport. There seems to be no optimal arrangement for all situations, because there are significant
trade-offs between the criteria that airport authorities or government entities find relevant, as possible
16
concessionaires. Political and economic conditions are not geographically homogenous and often
dictate the weights of each of these criteria. (La Croix, 1986) (La Croix, 1991)
1.3.2.2. Modeling, Queuing Theory and Simulation
One of the key aspects of this study is the modeling of a taxi service stand at an airport terminal, as
a way to test the performance of several design alternatives – more information on the reviewed
studies regarding modeling, queuing and simulation can be found in Annex I. Queuing Theory is a
fundamental piece of the knowledge background of this thesis, so a first look at the basic queuing
concepts is fundamental for contextualization. The main structure of a queuing system is composed of
three basic elements, each characterized by a set of attributes (Valadares Tavares, et al., 1996):
Source or Population – which generates clients arriving to the system, and is characterized by:
Population Dimension – infinite or finite. Infinite population - when the probability of
occurrence of a new arrival on a certain time interval is not influenced by the current number
of clients already in the system. Finite population - when the current number of clients in the
system can be a significant part of the population.
Arrival Dimension – usually divided into Individual arrivals or Group arrivals.
Arrival Control – Controllable or Incontrollable arrivals. Controllable arrivals – when the
arrival process can be limited and predicted through some mechanism (such as different days
for posting an application for the university). Uncontrollable when there is no obvious way of
limiting or predicting the arrival flows (arrivals at a Hospital).
Arrival Distribution – Usually described through the inter-arrival times distribution, which can
be constant, with pre-defined time intervals or random, following experimental-based or
probabilistic distributions.
Arrival Rate – Usually represented through the symbol λ, indicates the average number of
clients that arrived per unit of time. When the arrival rate does not vary, it can be considered
independent of the system state. On other situations, the arrival rate may vary according to the
current number of clients in the system, assuming the symbol λn.
Client Attitude – Usually classified as Patient and Impatient. Patient clients wait in queue until
entering service, whichever queue length or waiting time they experience. Impatient clients
give up and leave the queue after some waiting time or do not join at all from perception of
long waiting times and queuing.
Queue – Intermediate “storage space” element that is characterized by:
Number of Queues – There can be a Single Queue – when there is only one queue for all
servers, or a Multiple Queue – when there is at least one queue per server.
Queue Capacity – can be Infinite, when the maximum capacity is very large, when compared
to the number of elements that usually constitute it; Finite, when the queue can only contain a
smaller, limited number of clients.
Queue Discipline – The most frequent type of queue discipline is called FIFO (First In First
Out), in which clients enter service by the order at which they arrived in queue. There are
17
other situations where queue discipline is based on some relevant attribute of the clients,
randomly or in a LIFO (Last In First Out), for example.
Service – is the system element in which a client is processed, and is characterized by:
Service Configuration – translates the way service is organized, as a function of the number
of servers (or channels) in parallel or the number of service stages. Several combinations of
these two factors can exist. In the case of multiple service stages, each originating a queue,
the system can be considered as a Queuing Network.
Service Dimension – Simple or Multiple. Similarly to Arrivals, with individual or group service.
Service Time Distribution – Similarly to the Arrival Distribution, service time can be
described by a distribution of service times or number of clients served per time unit. It can
also be constant or random, just like the Arrival case.
Service Rate – is usually represented by the symbol µ, and represents the average number of
clients that can be serviced per server and time unit. Similarly, if the service rate is dependent
of the system state (number of clients in the system), then it is usually represented as µn.
The classification of queues is usually based on the following criteria:
X / Y / Z / W
X – Inter-Arrival Times Distribution
Y – Service Times Distribution
Z – Number of parallel servers
W – Other system characteristics such as limited queue capacity (K) or finite population (N) –
when these are left blank or have the ∾ symbol, the system has no additional restrictions.
Among the most common measures of performance are (Valadares Tavares, et al., 1996):
Average Queue Length (Lq)
Average Number of Clients in the System (L)
Average Waiting Time in Queue (Wq)
Average Waiting Time in the System (W)
Average Occupation Rate of the Service (% of time service is occupied)
Other indicators can also provide detailed information on the functioning of the system:
Pn = probability of n elements existing in the system (queue+service)
P(n ≥ k) = ∑ ��∾��� = probability of k or more elements existing in the system
P(Wq=0) = probability of Queue Waiting Time being zero
Nomenclature on the main measures of performance is usually based on the following symbols:
λ – Arrival Rate
1 / λ – Average inter-arrival time
µ – Service Rate
M – Negative Exponential Distribution
G – Unspecified Distribution
D - Deterministic
18
1 / µ – Average Service Time of a server
ρ = λ / mµ – Utilization Ratio, m = number of servers in the queuing system
Characterization of Arrival and Service distributions is usually done through a sequence of tasks:
1. Collect and describe information through histograms and sample parameters;
2. Infer the population parameters from the sample parameters;
3. Adjust a theoretical distribution to the experimental histogram, choosing it in order to
adequately describing the phenomenon.
Most Queuing Theory focuses on the stationary analytical methodologies for calculating values for
many of the abovementioned variables and measures of performance. The term Stationary or
Equilibrium is used when the system oscillates around an average situation, with the distribution of the
queue length being independent of time. It thus considers that arrival and service rates are relatively
regular and/or constant. Also, a queuing system will be able to reach a long-term equilibrium - “steady
state” - in its operation, only if ρ < 1 remains true, on the long run. The dynamic behavior of queues
(see example of Figure 2) is characterized by the following aspects (Odoni, 2007):
Expected delay changes non-linearly with changes in the arrival rate or the service capacity;
The closer the arrival rate is to service capacity, the more sensitive expected delay becomes to
changes in the arrival rate or the service capacity;
The time when peaks in expected delay occur may lag behind the time when demand peaks;
The expected delay at any given time depends on the “history” of the queue prior to that time;
The variance (variability) of delay also increases when the arrival rate is close to capacity.
Figure 2 - Expected delay for four different levels of service capacity at an Airport: R1= capacity is 80 movements per hour; R2 = 90; R3 = 100; R4 = 110 (Odoni, 2007)
Situations where the queue length tends to be infinite, because the arrival rate exceeds service
capacity or where arrival rates significantly vary according to the time of day (such as in many
transportation systems with traditionally high peak-hour demands), correspond to Transient States.
These situations cannot be modeled through the normal Stationary Equilibrium Equations and are
19
analytically complex. Such cases are often dealt with through the use of Simulation such as in (Curry,
1977), (Días Esteban, 2008) and (Cao, 2003). Simulation methodologies can effectively deal with the
peaks in arrival patterns; offer the freedom of using arbitrary distributions for the service time and
arrival patterns; dynamically test alternative schemes, quantifying the changes and offer animation to
support the communication at both management and operational levels (Joustra, 2001).
The fundamental relations that compose of the Stationary Equilibrium Equations are (Valadares
Tavares, et al., 1996):
L = λ . W
Lq = λ . Wq
W = Wq + 1 / µ
L = Lq + λ / µ
From the main equilibrium relationships one can deduct several expressions for the several types
of queuing systems, based on the birth and death processes. For queuing systems with Negative
Exponentially distributed inter-arrival times, any type of service time, one server and infinite queuing
capacity - M/G/1 system (Odoni, 2007):
Figure 3 - Delay versus Utilization ratio (ρ) and confidence limits evolution (left) and Dependence on Variability (Variance) of Inter-Arrival Times and of Service Times (right), adapted from (Odoni, 2007)
For queuing systems that reach steady state the expected queue length, expected delay and the
corresponding standard deviation are proportional to: 1 / (1 - ρ). Thus, as the arrival rate approaches
the service rate the average queue length, average delay and corresponding variability increase
rapidly - a large standard deviation implies unpredictability of delays. As can be seen from Figure 3,
there is a high sensitivity of delay at high levels of utilization, close to the maximum capacity (ρ = 1).
Expected delay increases at lower levels of utilization, in presence of high variability. (Odoni, 2007)
Expected delay is not the only factor that increases exponentially as the utilization ratio approaches
1. Unreliability is also a key issue concerning queuing systems and is equally affected by this ratio. By
Variability increases
σt – standard deviation of service times
E(t) = expected value for service times
Confidence Limits Confidence Interval
20
considering that most Arrival and/or Service rates may usually be described as Poissonian processes,
where variance is equal to the mean (E(X) = Var(X) = λ), then as average delays increase, so does
the variance of delays. Variance is an important measure of variability, which decisively influences the
confidence limits of the mean, greatly increasing the size of the interval of possible values as ρ
approaches 1. This means that as ρ increases, the unreliability also increases and the mere
occurrence of the higher limit values can jeopardize the entire functioning of the system due to
intolerable delays. When congestion occurs (ρ ≈ 1), delays can be tolerable (lower limit), or severe,
causing a complete breakdown of the system (upper limit), theoretically making waiting time infinite.
Although traditional queuing theory, which mainly focuses on stationary queuing processes, is
useful for conceptual contextualization and background, a different approach must be followed for the
case of Transportation Systems such as the Taxi. The Taxi demand at the airport is similar to that of
other transportation modes, conditioned by sudden peak-period increases, thus effectively making the
arrival rate be dependent on the time factor – Transient behavior. To measure, analyze and model
transient behavior, one must focus on the peak-hour evolution of queues using specific
methodologies, such as the ones described in (Newell, 1982), for example.
This method requires a lot of field data, for which at least two observers should exist, in the case of
the simplest situation with one queue and one service point. One of the observers should be placed
upstream of the service point to record the arrival times and identity of each customer that passes him.
The second observer is placed at the server to record the times and identity of the customers entering
the server and possibly a third observer just downstream of the server to record the times at which the
identified customers leave the server. Assuming an initial empty system, the arrival and departure
times of each individual customer are recorded and, if represented sequentially on a graph, can form
cumulative arrivals and departures curves (Newell, 1982). Having:
tj = Arrival time of customer number j
A(t) = cumulative quantity to arrive by time t, with tj ≤ t ; A-1
(x) = tj , for j-1 < x < j
tqj* = time customer number j leaves the queue and enters the service
tqj = time of the jth departure from queue (independently of queue discipline)
Dq(t) = cumulative quantity to leave the queue by time t
Dq*(t) = cumulative quantity to leave the queue by time t considering queuing discipline
Figure 4 – Graphical representation of cumulative arrivals and departures from a queue (Newell, 1982)
21
If one draws both A(t) and Dq(t) on the same graph (Figure 4), the curves cannot cross because,
for any t, the number of customers which have left cannot exceed the number which have arrived. The
vertical distance between the two curves at any time, representing the number of customers who have
arrived but have not yet left the queue, is Q(t).
Q(t) = A(t) – Dq(t) = quantity in the queue (Queue Length)
Assuming a FIFO discipline, tqj = tqj* and Dq*(t)=Dq(t), because customers will be sequentially
served according to the order they have arrived. If we draw A(t) and Dq* on the same graph (Figure 5),
the horizontal distance from A(t) to Dq* is the time which the jth
customer spends in queue:
Wj = tqj*- tj ≥ 0 = in-queue waiting time for customer j (this is also equal to the area of the
rectangular strip between A(t) and Dq*).
Figure 5 – Graphical representation of departure times (Newell, 1982)
If we were to place a third observer downstream of the server, he could record the times at which
customers left the server. From this we can define:
tsj = ordered time at which customer j leaves the service (in FIFO tsj = tsj *);
Ds(t) = cumulative number of customers to leave (in FIFO Ds(t) = Ds*(t))
A description of the server should at least define a relation between the curves Dq* and Ds*, i.e.,
between the tqj* and tsj*. The times each customer will be in service are given,
sj = tsj *- tqj*, for all j
The iterative process of finding Dq* and Ds* can easily be followed for most service systems in
queuing applications. If, for example, the server is a single-channel server, and for a given arrival time
distribution A(t), service times sj and queue discipline:
1. tsj *= tqj*+ sj
2. If the queue discipline is FIFO: tqj + 1 = max (tj + 1 ; tsj) = max (tj + 1; tqj + sj)
3. Within the conditions of points 1. and 2., wj +1 = max (0; wj + sj – (tj + 1 –tj))
The two main gross proprieties one may wish to calculate for queuing systems are the average
waiting time in queue for a set of n customers or the average queue length over some period of time.
The average time in queue for customers j+1 to j+n inclusive is defined as:
< Wk > = 1 / n * (∑ ����� ), with Wk = tqk*- tk
The average queue length is defined as (Newell, 1982):
< Q(t) > = (1 / (b-a)) * � ����. ���
� , for some time interval (a,b)
22
Chapter 2 – Problem definition and methodology
2.1. Regulatory and Institutional Issues
2.1.1. Problem definition
There are several issues beyond the purely operational context which influence the ability of an
airport taxi service system to perform according to certain standards of quality. These are mostly
related to the regulatory and institutional framework that encompasses the operation.
Airport taxi stands are often under the responsibility of airport authorities or local government
branches, who concession the service to one or more companies under different types of contractual
arrangements, according to specific interests – collect rents and limiting rent-seeking by taxi
operators, provide good quality of service to its clients and be politically balanced. (La Croix, 1991)
The regulatory environment conditions the way these are designed and structured, sometimes
imposing significant restrictions on the level of customization that these contracts may require.
Regulation can also exclude the possibility of introducing alternative service types, such as shared-
ride transportation schemes (shared taxi) (Frankena, 1984) and directly influence operational
conditions, such as the case with unrestricted access of all taxi drivers to airport stands.
Some airport taxi stands are completely free access points for licensed taxi drivers/owners, others
can only be accessed by permit-owning companies or individuals, and some others are only open to
companies with exclusive contracts. Within these three main types of arrangement, there may be
significant differences in service regulations regarding safety and technical requirements for vehicles,
professional qualifications for drivers and especially overall quality of service standards. The way the
different institutions involved in this context are articulated can also prove to be redundant or
inefficient, with significant consequences on the system itself, namely if there are disputes and/or need
to promote changes, such as new pricing mechanisms or alternative service types.
Sometimes the airport stand is not designed, planned or subject to operational changes with the
desirable level of participation from the airport authority, adding another barrier to the efficient
management of a fundamental land-side airport service – there are situations where this has been the
responsibility of general metropolitan transport authorities, city councils, municipalities or other state-
entities. This situation increases the difficulties to the airport managers in promoting quality-of-service-
related changes to the system, by adding new bureaucracy layers to the process, having to achieve
full consensus with these entities first. It often also means that this sensitive taxi service point may be
within the jurisdiction of a large generalist entity – sector-related body, city or state-wide Departments
of Transportation and planning, etc. – which might focus less on the airport.
On a regulatory perspective, interesting discussions have been developing for several decades on
market characteristics and the question of liberalization. Pro-liberalization arguments are based on the
belief that supposed unjustified restrictions on competition - seen as a mechanism that incentives
lower prices, innovation, better quality
and cause economic distortions (OEC
points to significant market failures su
high consumer search costs imposed b
(La Croix, 1991). Full liberalization of a
oversupply, unchanged long taxi queu
resulting in low quality of service (Scha
Focusing on the institutional frame
the presence of a regulatory body, w
city-wide and connected to a larger
transportation in general, roads, urba
function for individuals and companies
and stand dimensioning and responsib
urban/transportation planning entities a
by the airport or city authorities and the
Figure 6 – General In
In sum, taxi service at airports is su
interests, some with conflicting objectiv
other transportation systems, the bala
main objective of regulation and trans
achieve, especially because the invo
seems to decisively influence the bes
specific analysis on the main stakeh
know which problems emerge, to what
changes and interventions we can imp
Regulator
Licensing Entity
Planning Entity
Concessionaire
Operator(s)
23
ity of service and availability – significantly increase
ECD, 2007). On the other hand, the opposition to
such as imperfect information of the customers or c
d by the First-In First-Out discipline of taxis, as basis
f access to airport stands has also often yielded bad
euing, city-wide unbalances, rent-seeking and diffic
haller, 2007).
ework (see Figure 6), we can generally say that the
whose scope might be national, regional, state, me
r sector of Transportation or Urban Planning, suc
ban mobility or urban development. The professio
es is mostly performed by Transport Authorities. The
sibility over the taxi stands themselves can be divide
and airports. These stands are sometimes subject t
he operators are individual drivers, taxi companies or
Institutional Framework of Airport Taxi Services
subject to strong influences from many stakeholders
tives, such as taxi operators and airport authorities. A
lance between equity, efficiency and sustainability s
nsport policy. At airport taxi stands, this balance is
volved agents may have conflicting interests and
est course of action for each situation. A more in-
holders, regulation and institutional mechanisms is
at extent this aspect of the system can be improved
prove it.
Passengers
s
Usually National, Regi
Metropolitan or City
Transport Authoriti
Usually Transport Autho
Usually Local/Regional Tr
Planning Departments
Airport Authoritie
Usually Airport or Region
Authorities
Licensed individual driv
taxi companies
e inefficiencies
to liberalization
corresponding
is for regulation
ad results, with
fficult oversight
there is always
metropolitan or
ch as surface
ional licensing
he design, fleet
ded among the
t to concession
or both.
rs with specific
. As with many
should be the
is not trivial to
d local context
-depth, case-
is required to
d and by which
Regional,
ity-wide
orities.
uthorities.
al Transport
ents and
rities
gional/City
drivers or
ies
24
2.1.2. Methodology
In order to analyze the institutional and regulatory impacts on system performance and service
quality, it is necessary to focus on some crucial elements of this framework, which condition the way
the system is conceived, managed and operated. Some of these elements are very different from
airport to airport and so this analysis should be done on a case-by-case basis, according to local
specificities. The methodology for this regulatory and institutional analysis is to be centered on the
following steps:
1. Analysis of the general market characteristics and regulatory environment, namely the
types of existing or possible contractual arrangements and access to profession.
2. Identification of the main stakeholders and their specific interests and bargaining power.
3. Identification of the hierarchical relationships between the involved institutional agents and
the sharing of responsibility and power among institutions.
Regulations vary according to location of the airport, nature and size of the market and the socio-
economic model of the country or region. Usually, regulatory presence in this context is under the form
of restrictions to market access or numerical caps on the taxis serving a specific location or area.
Other interventions of regulatory bodies focus on restrictions to alternative service types, which are
also limited in many locations - so shared-ride services are not allowed to compete for passengers –
technical and professional requirements for vehicles and drivers and market structure - large
companies versus individual owners (Schaller, 2007). Different access schemes must be analyzed for
adequacy with local contexts, and usually divide into three types:
Exclusive contracts are contracts which only allow a single company to operate taxi services at
the terminal. These contracts provide greater flexibility than the Permit system in face of
fluctuating demand for services. It is, however, less politically balanced, administratively costly
and prone to rent-seeking, in the sense that it excludes competition and requires a contract
between the Airport and the Taxi Operator, which obviously increases contract management and
enforcement costs.
Permit systems are contracts that allocate permits to certain selected taxi operators to provide
service. These are preferred in cases where service quality is less important and the exclusive
contract lacks political support. If demand is relatively stable, the system becomes more
sustainable for permit holders and quality of service might remain at good levels, while being
politically acceptable. Disadvantages are mainly related with monitoring costs, which are higher
than in the exclusive contract system.
Open systems allow any licensed taxicab in the region to provide service at the airport. These
situations, if not subject to some form of local or global control on the number of taxi licenses,
promote low quality of service and excessive supply. This arrangement can, however produce
reasonable improvements on the elimination of excess rents and reducing administrative costs,
while fostering healthy competition and ensuring stable availability of service.
25
The main stakeholders are usually composed by the taxi companies/drivers, passengers, airport
management and competition. Other secondary stakeholders can also be found, as described below.
Taxi drivers and companies wish to maximize their profit by ensuring cost reductions produced by
typically longer trips from the airport to the city centers and less waiting times at the terminal,
increasing the number of trips per working day. Taxi companies are sometimes also target of
complaints by supposedly trying to cut back on costs by ignoring maintenance and cleanliness of the
vehicles, which has since become a special requirement in some airport taxi concession contracts.
The bargaining power they exercise is high, namely in the political field, where they are considered
very influential. They are a small and homogenous group, which increases their ability to effectively
organize and unite under a set of common goals, but are still big enough to have a significant weight
in the ballot box (La Croix, 1991).
Passengers usually want to minimize waiting time for empty – and clean - taxis and to be served by
polite and experienced drivers. Issues with driver friendliness, geographical knowledge and
professionalism, service reliability and prices are often the main problems that passengers identify
when asked on how to improve taxi services (Cardon, 2007). The passengers are simultaneous clients
of the airport authorities and taxi companies, when picking up a taxi at the curbside of an airport
passenger building. This characteristic makes them very influential politically (namely locally and
regionally) and increases their weight with the airport administration, who also has a lot of other clients
– airlines, commerce and services - depending on the steadiness and growth of passenger demand.
Airport Authorities want a reliable, fast and comfortable complementary surface transportation
mode, which can serve their clients with the minimum possible delay and inconvenience, in order to
avoid complaints. They are interested in offering good connectivity options to the passengers so they
retain the perception of good service quality and return to this airport for future trips – competition
among airports is nowadays fierce, with the advent of low cost airlines and emergence of secondary
airport hubs. The airport authorities sometimes contract out with taxi companies for the provision of
taxi services at their terminals, ensuring availability and quality of service, but just like with many other
contracts between principal and agent, hard bargaining and conflicts often come into play (La Croix,
1991). These entities are maybe the most pressured parties in the whole of the transaction scheme.
Competitors are always major stakeholders in this context. Other modes, such as buses, trains and
even rent-a-car companies are interested in maintaining or increasing their market share of airport
passengers who need land transportation and choose their services. They can use their own
bargaining power to influence the transportation authorities and political agents to invest on the
improvement of their operational conditions and even change or diversify their service types to directly
compete with the taxi sector.
Other stakeholders are, for example, hotels and other similar services, for whom airport taxi service
is a major feeder system for potential clients (Schaller, 2007); companies who wish to have their
collaborators arrive quickly at their business meetings and workplaces; direct airport clients such as
airlines, commerce and service providers who want passengers flows to keep growing and cities and
municipalities in general, who benefit from better airport accessibility.
26
The several agents involved in the several aspects of the operation of a taxi service such as this
are connected by links of hierarchy, power-share, complementarities and cooperation. As described
earlier, there is the need to identify the main entities responsible for the regulation, planning, licensing,
concessioning and operating and analyze if the corresponding responsibilities are well defined, power
and independence is well allocated and jurisdiction boundaries are clearly set.
2.2. Operational Issues
2.2.1. Problem definition
Airport operations management often focuses on land-side logistics and operations on the in-
terminal functions: Check-in services, luggage claim, security facilities, customs, information systems,
elevators and escalator systems, emergency and evacuation plans, etc. Despite the inherent
importance of these factors in efficiently managing and processing passenger flows, curb-side
operations also significantly influence the airport performance as a multi-service, multi-modal hub.
Intermodal connectivity is essential for the completion of the last segment of an airport passenger’s
trip. Taxis are one of the most used transportation modes to travel to and from the airports (Cardon,
2007). An efficient taxi service at the airport terminals is crucial to the overall passenger’s perception
of airport service quality. Even with high efficiency levels regarding in-terminal operations, an airport
terminal’s performance is handicapped without good connectivity, because people will be forced to join
long queues for ground transportation after a tiresome trip. The main concern for airport managers
shouldn’t be getting the passengers quickly to the exterior, but rather thinking of transferring the
passengers quickly to the next transportation mode that will take them to their final destination in the
fastest and most comfortable way.
A taxi service at an airport terminal is often subject to sudden spikes in demand levels at certain
peak periods throughout the day, mostly due to the concentration of scheduled arrivals of flights during
the morning and evening. Because this demand is time-dependent and can grow very fast at peak
hours, daily average rates of arrival and service times are not representative of the real solicitations
that the system is subject to. Passengers often experience long waiting times when queues build up at
a high-paced rate, because of queuing and service restrictions which prevent the system from
instantly responding at similar speed, accumulating delays. This situation can often be caused by a
poor design and/or management of the system, which does not seem to predict some of the variability
of the characteristics of demand and service. Moreover, the high number of taxis that concentrate at
the airport are also incurring in efficiency losses, by being parked, sometimes for several hours during
off-peak times, waiting for customers. At its core, these systems are prone to significant inefficiencies,
whether because of queuing delays for customers at peak hours or long waiting times for taxis
queuing at off-peak. Often there appears to be a mismatch between supply and demand mechanisms.
Although not usually a topic of research and in-depth studies, taxi services at airports can provide
very interesting challenges on a queue-modeling perspective. Because of the nature of the service
(Figure 7) - queues of taxis feeding qu
these systems are often in transient st
based on stationary considerations for
of these systems, namely during peak
that delays in queues are also subje
because of unstable rate of arrival for
periods, the rate of arrival increases an
arriving flights. This incoming flow also
service times, downstream of the queu
pieces of luggage or composed of sm
coordinate and board a taxi or a set
These two factors directly influence q
delays.
Queue and service configuration is
queues are not organized, signaled an
the service area. The number of queu
depending on the service type and
elements for an efficient operation, and
spatial distribution and allocation of re
abovementioned variability of demand
reasonable level of service can be m
queues and service areas and better s
taxi parking spots at the service area o
and off-peak periods can influence the
Figure 7 – Components o
Po
pu
lati
on
Airport
Passengers
Qu
eu
e
Arrival Process
27
ueues of passengers – and especially high variabilit
state during large portions of the day. Traditional qu
or arrival and service rates is not adequate to depict
ak hours, where these rates often show fast growth
ject to significant variation according to the time of
r groups and individuals and also a variable service
and becomes unstable, based on incoming flows from
lso has a group structure that is not irrelevant, main
eue. Groups, depending on size and type, (especia
small children or elderly/disabled people) tend to ta
et of taxis and thus may significantly increase the
queuing time and length, increasing probability of
is also an important element to consider regarding th
and disciplined, people are not efficiently and order
eues and their maximum capacity in space is also
number of service points. Service areas are, of
nd their configuration in terms of number of servers (
resources (taxis) influence the delays people are su
nd and its impacts on the system’s capacity to fun
mitigated by improving and adapting the configu
r synchronize their interaction. The number and disp
or the flexibility to change the way this area is organ
e service times and queuing times of taxis and passe
of a Basic Queuing Process at an Airport Taxi Stand
Qu
eu
e
Taxi Passenger
Queue
Se
rvic
e M
ech
an
ism Taxi Servic
Area
Queue Discipline
Queue configuration
Service Process
Queue System
ility in demand,
ueuing theory,
ct the behavior
th. This means
of day, mainly
e rate. At peak
om successive
inly in terms of
ially with many
take longer to
e service time.
f “unexpected”
this problem. If
erly directed at
o a key issue,
of course, key
s (taxi spaces),
subject to. The
function with a
uration of the
sposition of the
anized at peak
sengers.
ervice
Exit
28
Having the abovementioned factors into consideration, the operational issues surrounding the taxi
service stands at airports are not at all trivial. In fact, this mode of transportation is of key importance
to airports worldwide and many passengers often complain about long waiting times and queuing at
the curbside of airport passenger buildings. Changing the configuration of the elements of the queuing
system can potentially yield better results, by allowing flexibility between high and low demand periods
and improving level of service. In order to test alternative schemes for queue and service configuration
or different service types, there is need to focus on the actual behavior of these queues, define quality
indicators and build a basic model of the system, to serve as reference for performance evaluation.
2.2.2. Methodology
The airport taxi stand operational setting can be evaluated and analyzed through quality and
performance indicators of the queuing system, such as average waiting time for passengers or
maximum queue length. These can be measured or estimated through a series of methods, some of
which usually require a lot of field data collection and processing, which feed into mathematical
models built to mimic reality. These are used to test different scenarios and system configurations, in
search for better solutions, while presenting results on the basic system’s performance.
The main identified steps that constitute the methodology concerning the operational context are:
1. Identification of the problem of queuing at Airport Taxi Stands (Section 2.2.1)
2. Literature review and general queuing theory research (Section 1.3.2.2) ;
3. Identification of a suitable and real case-study ;
4. Preliminary in situ observations of system behavior ;
5. Elaboration of a Data Collection Plan ;
6. Test data collection procedures ;
7. Collect relevant data ;
8. Compile and analyze the collected data ;
9. Build basic queuing simulation model ;
10. Test and validate the basic simulation model based on the collected data ;
11. Perform scenario building and testing ;
12. Results analysis and conclusions.
Similarly to many other transportation services, queues at airport taxi stands show a significant
transient behavior, regarding the arrival rates for queues of passengers, especially at peak hours. This
process is characterized by irregularity and instability in the arrival rates, which also have an impact on
the queue length and waiting times. This means that analytical methods, by which general Queuing
Theory usually addresses queuing systems problems, while useful for insight, are not advisable as a
main analysis tool, especially because they assume stationary behavior of queues, with regularity of
arrival and service rates. The analytical formulation of transient solutions that has been developed is
relatively complex and mathematically burdening, which, in this context, opens the window for
simulation, a powerful tool that can adequately model variability, while being less time-consuming.
29
While simulation is important, it is more relevant when connected to a real case in an experimental
space that can be modeled, discussed and tested. In order to apply these principles of queuing and
simulation to a real situation, there is need to first of all, find a suitable case study. Given the resource
and time limitations of this academic effort, the obvious choice should be the Arrivals Taxi Stand, at
Portela Airport, in Lisbon. The proximity, past familiarity and relevance of this airport – when taking
into account the new projected Lisbon International Airport - make it a suitable candidate for a more in-
depth analysis.
In order to build a simulation model that mimics this specific situation, one must first try to
understand the behavior of the system in reality. This calls for careful observations in situ, but also
research into what is understood of this kind of system. The latter is present in Chapter 1, where a
strong literature review on simulation and queuing theory is built. The former is essential, not only to
know about queues in general, but to be able to model the observed queue correctly.
Some details, such as the influence of police or the effects of queuing in front of terminal exits,
formation of secondary queues, etc, can only be perceived by direct observation during significant
periods of the day, namely at peak-times. These preliminary observations can be highly valuable in
order to avoid mistakes and predict difficulties during the real data collection efforts. While doing these
observations, several different methods were tested for collecting relevant information on queuing and
some sampling data was also collected. This allowed the estimation of the recommended sample size,
mainly based on the observed standard deviation of the sample. This preparatory phase involved
communication and collaboration with the airport authorities, namely ANA (Aeroportos de Portugal) to
facilitate measurement efforts at the terminal, allowing easier interaction and increased acceptability of
the actors present at the operational area (Taxi drivers, Portway employees and Police). ANA also
provided relevant data on a survey they performed in 2006, when the fleet size of the taxi contingent
at the Airport was being studied.
A Data Collection Plan was built, in order to organize the time, resources and methods to be used
in the data collection procedures. The Taxi Stand at the Arrivals Hall was chosen as the basis for this
planning phase. Some portions of this Plan were developed iteratively, as different methods for
collecting data and some conceptual considerations were being tested in the field. Some of these
conceptual considerations were linked to the measurement of queue length and in-queue waiting time
for passengers, which eventually had to be obtained indirectly through the consideration of other
collected data, related to Arrival and Service Times. Independence among Arrival and Service Times
was assumed as a basis for separate individual observations of these factors, due to the inherent
limitations of the author in mounting a simultaneous multi-observation point scheme. This plan - shown
below - was a valuable tool in terms of identification of the main objectives and preparation for the
main difficulties involved in field work efforts.
30
2.2.3. Field Data Collection Plan
Field Data Collection Plan
Performance and Design of Taxi Services at Airport Passenger
Terminals
Summary
I. Determining the best time period for the data collection
II. Determining the target parameters for observation
III. Determining the required sample size of the collection
IV. Determining the best data collection scheme
I. Determining the best time period for the data collection
In order to capture relevant data for the modeling of the queuing system, the best period for
measurements and observations should be the busiest times of the year and of the day. This is when
the maximum solicitations on the system occur, when the system is most vulnerable to delays or clear
signs of inefficiencies.
After consulting ANA’s Annual Traffic Report and its 2006 survey on taxi usage at the terminal –
performed during the Easter Holidays - we reach the conclusion that:
The month with the most passenger traffic is August (almost 1.600.000 passengers), closely
followed by July, with about 1.400.000 passengers (see Figure 8). These aggregate values refer
to arrivals and departures, but it is assumed that August is still one of the busiest – if not the
busiest – months of the year, namely due to the traditional seasonal increase of foreign tourism
during the Summer. (In 2008, the peak-hour traffic of passengers occurred on the 10th of August,
between 8 and 9 am, with a total of 4.926 passengers handled.)
Figure 8 – Passengers Traffic by Month (Source: Annual Traffic Report – ANA, 2008)
The weekly distribution of passenger traffic is close to uniform (Figure 9). Regarding arriving
passengers, it had an average of about 20.300 passengers/day during the week of the 11th to the
17th of April, 2006 – see Figure 10.
31
Figure 9 – Passenger Traffic by Day of the Week (Source: Annual Traffic Report – ANA, 2008)
Figure 10 – Arriving Passengers during the Easter Holidays week in 2006 (Source: ANA, 2006)
The busiest hours are between 8 am and 11 am in the morning (200 to 250 passengers/hour),
between 3 and 5 pm (about 200 passengers/hour) and at 10 pm (286 passengers/hour) – see
Figure 11.
Figure 11 – Hourly distribution of Passengers and Taxis at the Airport Arrivals taxi stand (Source: ANA, 2006)
This leads to the conclusion that the best periods for collecting data are between July and
August, at any day of the week, in the morning between 8 am and 11 am and in the
afternoon, from 3 pm to 5 pm and close to 10 pm.
18.038
21.944 21.622 20.56418.868 19.256
22.265
0
5.000
10.000
15.000
20.000
25.000
11Tue 12Wed 13Thu 14Fri 15Sat 16Sun 17Mon
Arriving Passengers from 11 to 17 April 2006
0
50
100
150
200
250
300
350
7:0
0
8:0
0
9:0
0
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
0:0
0
1:0
0
táxi
s/p
asse
ng
ers
passengers
taxis
32
II. Determining the target parameters for observation
After discussion about the key target parameters for observation, the conclusion is that the model
requires information on:
Arrivals (Times/Rates) – both of groups and individuals. Individuals determine queue length and
average waiting time and groups relate to the number of taxis. This will allow the estimation of an
Inter-Arrival Time Distribution.
Comments:
- The group distribution and composition is to be determined at different peak-times of the day,
in order to verify if assumptions on time-of-day independence and low daily variability are
admissible.
Tasks:
- Register types of “sets” of people arriving at the queue, characterized by the number of
perceivably “related” people in each set (n); When n = 1, individual set, when n = 2, 3, 4…
group set.
- Register Arrival Times (for different “sets”) – Inter-Arrival times will be calculated as the
difference between arrival times of subsequent groups.
Queue Lengths and In-Queue waiting times – Estimating the length of the queue line and
passenger waiting times.
Comments:
- An indirect method, based on individual measurement of Arrival and Service Rates and
Times, was chosen due to the difficulty in measuring arrival patterns and service patterns
simultaneously. Queue length and in-queue waiting times during a specific period of time are
dependent on the real-time arrival and service of passengers. This could be measured by
having observation points at the start and end of the queue, measuring inter-arrival and
service times for individuals/groups simultaneously during that specific period. However,
service can be provided by up to four different “servers” at each moment, with significant
unpredictability in the behavior of the agents involved (driver discussions, police intervention,
luggage issues, children, etc) and even with two observation points, data collection becomes
very complex.
33
Service (Times/Rates) – Measuring the time between a taxi parking in a service space and the
next taxi parking in the same space, available for service.
Comments:
- The existence of four different service parking spots for taxis (2x2 disposition, Figure 13),
subsequent increased complexity and few human resources lead to the consideration of
measurement of individual server service times. This means that individual measurements of
servers must be focused on the total time between service availability, which represents the
effects of some interdependence between servers. Total number of observations should be
divided among the different servers in order to determine if service times are similar.
Tasks:
- Measure time between occupation of the service spot by two consecutive taxis in order to
determine “service duration”, including “empty time”.
III. Determining the required sample size of the collection
In order to be considered a statistically significant representation of a wider group of occurrences,
the collected sample must feature a certain minimum number of observations. This sample size is
determined so that the maximum difference between the sample mean and the population mean is
estimated to be within a certain interval, according to a specific level of confidence. The following
formula allows the calculation of the recommended sample size:
is known as the critical value, the positive value that is at the vertical boundary for the area of
in the right tail of the standard normal distribution – alpha is defined as 0,05, which corresponds,
in this case, to a Zα/2 of 1,96.
is the population standard deviation - estimated through pilot tests.
is the sample size.
E is the maximum difference between the observed sample mean x and the true value of the
population mean µ - considered 10% of the sample mean.
Based on the preliminary data collection effort, and using the abovementioned formula for
estimating sample size and a confidence level of 95%, we reached the following conclusions (see
Figure 12):
Sample size for Inter-Arrival Times for Groups is 351 observations.
Sample size for Service Time is 88 observations.
34
Service Times (seconds) Inter-Arrival Times for Groups (seconds)
Z 1,96 Z 1,96
σ 32 σ 22
E(%) 10% E(%) 10%
E 6,7 E 2,3
n (sample size) 88 n (sample size) 351
Sample Average 67 Sample Average 23
Figure 12 – Sample Size parameters for the Service Times and Inter-Arrival Times for Groups
IV. Determining the best data collection scheme
After defining the main target indicators, a good set of observation points and duty division must be
assured. The consideration of two separate measurements for time intervals, one at the start of the
queue and another at the end indicates the need for two distinct observation points. The observer at
the end of the queue must also be able to register the group size, based on his perception of such.
Measurements will take place at different peak hours of different days. This will allow comparison
between the different peak hours and days, mitigating possible unexpected unknown anomalies and
patterns, and will facilitate the collection of a larger variety of observations. This discontinuous
measurement of service times and inter-arrival times is based on the assumption that peak-hour
service times are independent from the day, different peak-hour periods of the day, and arrival rates.
Similarly, so should be inter-arrival times. As a possible collection scheme, the position of the
observation points is sketched on Figure 13.
Figure 13 – Data Collection Scheme
2 1
4 3
35
After understanding the way the system works and planning the collection of data, a series of
measurements were performed, namely:
On Wednesday, 5th
of August – Experimental collection of preliminary data in order to
determine sample size. Inter-arrival times and corresponding group sizes were registered from
8 to 10 a.m. and service times for the inner row server positions (servers 1 and 3) were
registered from 10:20 to 11:20 a.m. Additional observations were made, from 2:30 to 3 p.m. in
an attempt to measure queue length and in-queue waiting times, but later discarded.
On Thursday, 13th
of August – Inter-Arrival times for Groups and corresponding group size
were registered from 9 to 10 p.m.
On Thursday, 27th
of August – Service Times and corresponding taxi occupation on the inner
row servers (servers 1 and 3) were registered from 9 to 10 p.m.
On Monday, 14th
of September – Service Times on the outer row servers (servers 2 and 4)
were registered from 9 to 10 p.m.
The collected data was then compiled, formatted and processed, in order to estimate key
performance indicators, namely queue length and in-queue waiting time (see Annex II). For this,
peak-period conditions were isolated and analyzed and, based on a set of assumptions and
simplifications, these parameters were determined. As discussed in Chapter 3, these assumptions
allowed to indirectly estimate queuing characteristics that through in situ measurements would be
highly resource-consuming, and thus difficult to execute, in the context of this thesis.
A basic queuing simulation model was built using the SIMUL8 software. This model is analyzed in
detail in Chapter 3, but its main structure is composed of a “work” entry point, a queue for passengers,
four servers and a “work” exit point. In this case, “Work” can be described as passengers or groups,
whose effect is also considered in the model. Reneging, balking and jockeying were not considered.
The parameters that characterize the processes which connect these entities, such as the arrival
process, the service process or the routing of “work” within the system were estimated and modeled
based on the raw collected data.
After introducing data into the system and calibrating it, the simulation is tested to determine the
validity of the model. This validity check compares the logic and similarity of results and behavior of
the simulation and data collected in situ. The model should reasonably approximate reality and mimic
witnessed field behavior.
Scenario building and alternative testing is a key step in the methodology scheme of this thesis.
The main objectives of this study are not only to describe the current behavior of the system, but also
to propose alternative schemes and test them against the status quo. In order to do this, the
simulation model is run and adapted to different operational contexts such as different numbers of
servers, different restrictions, etc. The results are then analyzed and confronted against the current
situation, in order to determine which improvements could be promoted and how. Chapter 3 further
develops this procedure as the Portela case study is subject to an in-depth analysis.
36
Chapter 3 – Case Study – Portela Airport
3.1. General background
3.1.1. Introduction
This chapter has the intention of providing an in-depth look at a specific situation in reality,
identifying the stated problem on a case-study, contextualizing, modeling and analyzing it. The choice
of the case-study, Portela Airport, was based on three criteria:
Proximity – Portela Airport is located in Lisbon, relatively close to the city center, and this
coincides with the location where this thesis is developed. This allows easy access to the
location, and facilitates observations and understanding of the system.
Familiarity – Portela Airport is generally known to most Lisbon citizens, including the author,
which is relatively familiarized with the problems at the taxi stand, accessibilities, involved
entities, history of the airport and recent developments/expansions, either due to personal
experience or through the media in general.
Relevance – This aeronautical infrastructure is the most important in the country, serving as a
major gateway for Europe and as the main TAP (Transportes Aéreos Portugueses) hub, which
allows it to control a significant share of the market of trips to South America, especially Brazil.
The future construction of the new Lisbon International airport also brings new interest to the
study of the current operational conditions, in hope of correcting past mistakes and
inefficiencies. The taxi system at Portela is therefore a key element of connectivity for Lisbon’s
citizens and foreign visitors.
The methodology presented in Chapter 2 is applied to this case-study, presenting the main results
of the analysis, carried out according to the different steps of the regulatory and operational
“checklists”. Following this reasoning, the status quo is the obvious focus of attention at this point.
Objectively defining the several elements at play in present is fundamental for modeling of the system
and understanding the several agent dynamics involved in its management. This procedure
simultaneously allows the establishment of a basis for scenario-building, looking at possible
alternatives for future improvements and testing the performance and adequacy of those options.
Just like with other academic endeavors, simplifications and assumptions on certain system
characteristics are present, and promptly justified, to the maximum possible extent. Many of them are
a reflection of the size and scope of this thesis; others are related to the randomness and
unpredictability of many of the human and system behaviors encountered. Most of these
considerations derive from in situ observations and general perception of the system itself, from also
having been an active airport and taxi passenger. Regardless of the cause, they should be
contextualized with the nature of the thesis itself, intended as a meaningful contribution to further
understanding this transportation problem, not an exhaustive and error-free analysis of an airport taxi
system.
37
3.1.2. The airport
The Lisbon International Airport, also known as Portela Airport, is the biggest and most important
airport in Portugal, serving the nation’s capital, Lisbon. It has two crossed runways, with 3800m and
2400m, respectively, and two passenger terminals, Terminal 1 for International flights and Terminal 2,
recently built, to handle domestic flights. The airport is run by state-owned ANA (Aeroportos de
Portugal), which is also responsible for the operation of most airport facilities in the country.
In 2008, Portela processed about 13,6 million passengers, 1,5% more than in 2007, following a
steady but slow growth trend, since 2003. (ANA, 2008) Also at Lisbon Airport, in 2000, a survey
concluded that about 42% of passengers were tourists and 28% of the passengers traveled due to
business motives (FCG-Parsons, 2002), while another source, in 2006, states that 56% of the
passengers of this airport had traditionally middle-high income professions. (ANA, 2006) The
percentage of taxi users at Terminal 1 was 38% in 2000. (FCG-Parsons, 2002)
Terminal 1’s services and operating processes are distributed along three levels (Level 2, 3 and 4).
Arrivals are placed on Level 2, and international departures on Levels 3 and 4. The specific services
located at these levels can be identified as following (Días Esteban, 2008):
Level 2
Arrivals (international)
Information stands
Baggage lost and found
Access to parking
Lounge concessions
Level 3
International departures
Shuttle to Terminal 2
Police
Post office
Lost properties
Commercial area
Level 4
International departures
Ticketing offices
Access to boarding gates
The processes of managing arrival and departure flows are segregated by levels but whereas the
arrivals take place on the front side of the building, the departures entrances are located on its lateral
side. Both areas also feature a significant presence of shops and other commercial services.
38
Terminal 2 (Figure 14) was inaugurated in 2007 as part of ANA’s expansion plan for Portela
Airport. This secondary terminal is designed to process domestic flights and is a central piece of the
objective setting that ANA defined for the future: sustain the present and expected traffic growth until
2017 (planned date for the inauguration of the new Lisbon Airport), by increasing the airport capacity
to 40 movements per hour, with 15 to 17 million passengers per year; increase the levels of comfort,
safety and quality of the service provided; create more opportunities for the non‐aviation businesses
(expanding commercial areas) and consolidate the airport as a national hub. (Carvalho, 2008)
Figure 14 – Terminal 2 location (Source: Virtual Earth)
Also relevant is the construction of the New Lisbon Airport (NAL), in the area of Alcochete (Figure
15) on the south bank of the river, which has recently been approved and is currently in the tendering
process. It will replace Portela airport in 2017, as the main aeronautical infrastructure of the country,
with capacity to handle up to 100 aircraft movements/hour, located at about 48 km from Lisbon. This
airport will also be served by a TGV link and standard trains in addition to taxis and buses. Portela’s
survival as an airport infrastructure after the inauguration of the NAL is still under discussion.
Figure 15 – Blueprints of the future New Lisbon Airport at Alcochete (Source: www.naer.pt)
There are a few well-known and traditional tourism destinations in the Lisbon Metropolitan Area
(see Figure 16). The main ones are the downtown and historical part of the city, located at about 7 km
from the airport, such as: Marquês do Pombal (17 minutes away - source: Google Maps), Chiado (21
minutes) and Belém (24 minutes); new cosmopolitan areas such as Parque das Nações (12 minutes);
Terminal 1
Terminal 2
39
and two places outside of the Lisbon Municipality: Sintra (34 minutes) and Cascais (37 minutes), both
to the West. One interesting fact is that most of these destinations, including the last two, are
approximately within 30 minutes and 35 km drive of the airport, unlike many European and U.S.
airports, which are sometimes located significantly far from the city center.
Figure 16 – Three of the main Tourism destinations from Portela Airport (marker A), in the Lisbon Metropolitan Area, upper left: Chiado, upper right: Belém, bottom: Cascais (Source: Google Maps)
Currently, Terminal 1 is directly served by six Carris (Lisbon’s main Bus and Light metro
transportation company) bus routes - buses nº 5, 22, 44, 83, 208 and 745 - with a ticket price of 1,4 €
per trip; the Aerobus route, which connects the airport to the city center every 20 minutes with a valid
daily ticket in all of Carris network for 3,5 €; the Aeroshuttle, which connects the airport to several main
multimodal transport stations in the city (Entrecampos, Oriente and Sete Rios), also costing 3,5 €, and
a taxi service, which has a minimum flag price of 2 € and outside of the city has a price of 0,46 €/km;
luggage or animal transport is charged at 1,6 € extra.
There is also a special pre-paid taxi service in Lisbon called Taxi Voucher. This service is available
to passengers arriving at Portela Airport who wish to travel by taxi. The service operates with vouchers
on sale at the Turismo de Lisboa counter, located in the Arrivals Hall of Terminal 1. The price of the
voucher depends on the distance of the trip and on the type of service required: normal or
personalized (in the latter, the driver is trained to speak foreign languages and acts as tourist guide).
The client pays a fixed fee according to the destination area. Prices for normal service of this type
range from 14,61 € to 25,36 € for trips within Lisbon and 48,60 € to Sintra and Cascais.
40
Regarding Terminal 2, there are only three ways for a passenger to get there: taxi, free shuttle from
Terminal 1 or private vehicle (paying for parking). This represents a risk if, for example, the passenger
has very little time to get to Terminal 2 and wishes to quickly pick up a taxi at Terminal 1. This
becomes a problem because of two factors: Long queuing at the Taxi Stand - if on peak hours - and
taxi driver’s resistance to making the trip. Excessive queuing is a common problem to all taxi
passengers at Portela during peak hours. Taxi drivers at the airport stand also wait, sometimes for
several hours, for a service they perceive as lucrative because of typically longer trips downtown. This
makes short trips such as from Terminal 1 to Terminal 2 undesirable and taxi drivers might pose
serious problems to this request – complaints about trip refusals or rudeness of drivers regarding short
trips are frequently heard among taxi passengers.
3.2. Analysis of the Taxi Service at Terminal 1
3.2.1. Regulatory and Institutional Context
The taxi service at Terminal 1 is the central point of analysis of this thesis. There are several
aspects to mention regarding the status quo of this service, from basic queuing elements to the higher
hierarchy of institutions responsible for the adequate functioning of the system. In this section, a
systematic analysis of the regulatory and institutional context is performed, according to Chapter 2’s
proposed methodology:
1. Analysis of the general market characteristics and regulatory environment, namely the
types of existing and possible contractual arrangements and access to profession.
2. Identification of the main stakeholders and their specific interests and bargaining power.
3. Identification of the hierarchical relationships between the involved institutional agents and
the sharing of responsibility and power among institutions.
The purpose of this section of the thesis is not to do a full in-depth analysis of regulations and
legislation concerning taxi services in Lisbon or to create a whole new institutional design, but rather to
highlight the main restrictions and variables that integrate the system overarching the operational
environment. The analysis of licensing requisites, regulations, hierarchies and power-sharing
mechanisms among participating entities allows for the identification of eventual incoherencies and
redundancies in the system, possibly opening the window for the testing of options for improvement.
Although the aspects involving institutions and other participating agents are a key aspect in this
context, they can assume a degree of complexity that goes beyond simple considerations of hierarchy
and power-share, due to some of their inherent opacity. An effort to listen to every actor’s position
pertaining to this taxi service’s performance was made, from the side of the Airport Authority, the Taxi
drivers and the Municipality technicians. Several of these informal inputs are present in the author’s
perspective and criticism of the whole institutional network of relationships, from what was perceived
of the logic of confronting the several collected views.
41
3.2.1.1. General Market Characteristics
As mentioned earlier, airport taxi stands are often seen by taxi companies and owners/drivers as
profitable service locations. In Lisbon, and despite the proximity of the airport to the city center, this
seems to be no exception. Distance, and to a certain degree, time, although always considered as
crucial income factors for drivers at the airport, are often accompanied by the luggage factor, which
allows for the charging of an extra fee of 1,60 €, and tips, which tourists are known to give, traditionally
more often than regular city locals. Another crucial element, that apparently makes long waiting times
in the taxi queue at Portela worthwhile, is the presence of a large supply of customers, mainly at peak
hours. Many of these customers, due to their high value of time and comfort needs, do not even
consider using other alternatives to the taxi, and often face the waiting time in the taxi passenger
queue as annoying, but unavoidable, compared to the perspective of taking a crowded bus downtown.
The main tourist destinations - and many other main passenger destinations - have been identified
(see section 3.1.2) but the core of this apparently profitable market is contained on the arriving flows of
passengers at the terminal, not so much on disperse demand for trips back to the airport. The size of
the market is closely related to the flows of deplaning passengers at Portela and the share of these
passengers that usually choose the taxi service as their primary transportation mode.
As previously mentioned, in 2000, a survey concluded that about 42% of passengers were tourists
and 28% traveled due to business motives (FCG-Parsons, 2002), while another source, in 2006,
states that 56% of the passengers of this airport had traditionally middle-high income professions.
(ANA, 2006) Also, in 2008, Portela processed about 13,6 million passengers, 1,5% more than in 2007,
following a steady but slow growth trend, since 2003. (ANA, 2008) The percentage of taxi users at
Terminal 1 was 38%, in 2000 (FCG-Parsons, 2002), later estimated at 30% in a 2006 study, by ANA.
These percentages, coupled with the fact that air passenger flows at Portela have not diminished
for a long time, means that Demand is not only relatively high and constant, mainly at peak times,
during certain periods of the day and of the year, but also that the customer characteristics themselves
are adequate for the transportation segment under analysis. It is common to associate taxi services to
middle-high income professions, business travelers and tourists, since it is a more direct, comfortable
(and expensive) way of travel, which obviously correlates with many airport passenger profiles.
On the supply side, taxis are highly abundant at the airport, gathering at the taxi parking facility,
about fifty meters from the terminal and queuing along the access road, on a segregated lane,
sometimes for several hours, according to some taxi drivers. The taxi queues are long and constant
throughout the active daily operational period of the airport. There are two taxi stands at Portela, one
in front of the Arrivals hall and another at the Departures hall. This secondary stand has long been a
topic of some discussion among some taxi drivers with respect to fairness of competition.
External competition for passengers is also present, mainly in the form of buses, special pre-
booked car services and rent-a-car companies. The first service type is mainly provided by Carris,
through the existence of six normal bus routes and two shuttle-like service routes. Special pre-booked
42
car transportation services are available for passengers who make reservations prior to boarding for
Lisbon or have a company agreement or business endeavor which allows them to access this type of
service. Rent-a-car companies, also abundant at Portela are an option for those wishing to travel from
the airport to the city, with increased freedom and immediate availability. The share of the market for
buses was, in 2000, about 7% (FCG-Parsons, 2002) while the share for rental cars was 10%.
Access to the Lisbon airport taxi stand is free to any licensed taxi, company and/or taxi driver in
Lisbon. The only two restrictions on service access are the maximum parking capacity of the stand
and the need to be a licensed taxi operator of a licensed taxi vehicle, the latter by the Lisbon
Municipality. Lisbon Municipality is the regulator of this service, determining the location of the stands,
their dimension and parking capacity and controlling the total number of issued licenses. There is a
regulation document (see 3.2.1.3), issued by the Municipality regarding all taxi services within city
limits, which defines service types, market access rules, vehicle requisites, etc. We can thus say that
the type of arrangement present in Lisbon is a kind of mix between an open system and a wider permit
system, in the sense that it is open to any licensed driver, but only Lisbon-registered taxis may solicit
service within city limits. There is no known specific contractual arrangement between the City, the
Airport and/or the sector-related associations/companies for the imposition of any other service access
restrictions; therefore, there is no actual permit system at the Airport.
3.2.1.2. General Licensing
Licensing of Taxi companies and individual entrepreneurs
Taxi services in Portugal can only be supplied by registered commercial companies or individual
entrepreneurs – in case of using a single vehicle in their fleet. These entities are subject to licensing
requirements, demanded by IMTT (Instituto da Mobilidade e dos Transportes Terrestres – Institute for
Mobility and Land Transportation). (IMTT, 2009)
Registration requires mandatory licensing, issued and renewable, with a maximum validity of five
years, constrained to the fulfillment of the following requisites:
Competence and ethical integrity of the administrators, managers or directors, in case of
companies and the license owner, in case of individual entrepreneurs;
Professional capacity, same as above;
Financial capacity: 5.000 € in the beginning of the activity and 1.000 € per licensed vehicle, at
renewal.
Legal framework consists of Decree-Law n.º 251/98, 11th of August, altered by Law n.º 156/99,
19th of September, Law n.º 106/2001, 31
st of August and Decree-Law n.º 41/2003, 11
th of March;
Order n.º 8894/99, 5th of May.
43
Licensing of Taxi drivers
The professional exercise of taxi driving and transportation services is restricted to the ownership
of a Professional Aptitude Certificate (CAP – Certificado de Aptidão Profissional), issued by the IMTT,
which also certifies the corresponding professional training courses. This renewable certificate is valid
for five years. Requisites for issuing of the certificate are:
Age between 18 and 65 years, minimum education level, Portuguese language mastership,
and drivers license (type B);
Having successfully concluded a Type I professional training course (minimum 17 years
old), certified by the IMTT (550 hours) - through training, or;
Having successfully concluded a Type II continuous training course (need to have 2 years
of experience in driving automobiles), certified by the IMTT (200 hours) - through
Professional Experience complemented by training, or;
Ownership of a license that enables the exercise of the taxi driver profession, issued less
than five years ago, in the European Union, or another country in case of reciprocity
agreements, just as long as the professional training is equivalent to the requisites of
Portuguese Law – through Title Equivalency.
Requisites for the renewal of the certificate are:
Having concluded a continuous training course of minimum 20 hours, if the person
exercised the profession for at least 36 months during the CAP’s period of validity;
If the previous condition is not verified, the minimum training duration is 50 hours.
Legal framework consists of Decree-Law n.º 263/98, 19th of August, republished by Decree-Law n.º
298/2003, 21st of November; Portaria n.º 788/98, 21
st of September, republished by Portaria n.º
121/2004, 3rd
of February.
Licensing of taxi service vehicles
The companies that are licensed by the IMTT to provide taxi services can register vehicles for taxi
transportation. These vehicle licenses are issued by the municipalities, restricted to public tendering,
within fixed contingents (numerical limits) with a periodicity of 2 years, and cease to be valid with the
end of the commercial license’s validity. They are sequentially numbered and identified with the
corresponding municipality, and then fixed to the vehicle through small plates. If the vehicle is
replaced or the ownership of the vehicle is changed by license transfer, the number of the license will
remain the same, even if there is a new issuing of the license. Once issued, the owner contacts the
IMTT, in order to integrate that vehicle in the global company license.
Legal framework consists of Portaria n.º 277-A/99, 15th of April, altered by Portaria n.º 1318/2001,
29th of November, by Portaria n.º 1522/2002, 19
th of December, and by Portaria n.º 2/2004, 5
th of
January; Order n.º 8894/99, 5th of May.
44
3.2.1.3. Regulation
Regulation of Lisbon’s Taxi services is the responsibility of the Lisbon Municipality, which has
created, in 2002 the “Regulation for the Exercise of the Activity of Taxi Services in the Municipality of
Lisbon” (CML, 2003). There are some interesting articles in this document, worthy of highlight.
Article 5th (Chapter III, Section I) of this regulation states that taxi services can only be performed
by vehicles with national plates, with a maximum capacity of 9 seats, including the driver’s, equipped
with a taximeter and driven by licensed drivers with a professional aptitude certificate (CAP).
Regarding Market Organization (Chapter III, Section II), article 7th mentions the types of services
allowed, namely that taxi services are a function of the travelled distance and waiting times or:
a) By the hour, as a function of Service Duration;
b) According to the itinerary, as a function of established prices for certain origin-destination trips;
c) By contract, as a function of a written agreement with a duration of more than 30 days.
The parking regime (Article 8th) speaks of the permission to solicit and pick up passengers
anywhere in the public road circulation network, except at less than fifty meters of a taxi stand, as long
as there is a visible vehicle parked there. It also states that the utilization of taxis within a taxi stand is
made according to the order in which these are parked (First-In First-Out). There is also the possibility
of the Municipality altering the locations of the stands, in the context of its traffic ordnance
competencies.
The fixation of contingents, numerical limits on the number of taxis serving the city, is defined by
the Lisbon Municipality for the whole Municipal area, according to article 9th. This will be periodically
reviewed with a frequency not less than two years and preceded by a consultation of the
representative sector-related entities. For this procedure, the global needs of taxi services in the
municipal area will be taken into consideration.
Article 11th (Chapter IV) states that the attribution of taxi service licenses is done through an open
public tendering process, where IMTT-licensed companies or individual entrepreneurs can participate.
The different criteria for this attribution are explicit in the article 19th of this regulation, where the
following factors are considered, by decreasing order of importance and preference:
a) Location of the Social Headquarters is in the Lisbon Municipality;
b) Number of years without having been awarded a license in a tender;
c) Number of years of sector-related activity;
d) Age of the Social Headquarters.
Article 28th (Chapter V) speaks about the mandatory service provision. It states that taxis should be
at the disposal of the general public, in accordance to the parking regime that is attributed to them,
and they cannot refuse service, except when:
45
a) Service requires the circulation on very difficult or inaccessible terrain and/or places where the
driver perceives significant danger to himself, the vehicle or the passengers;
b) Services solicited by people with dangerous-like suspicious behavior.
Article 35th (Chapter VI) states that the monitoring entities, regarding the respect for the regulations
present in this document, are the IMTT, the Lisbon Municipality and the general Police forces.
Finally, article 37th identifies the main fines to be applied in case of failure to comply with the
refereed regulation (offenses) – generally monitored by police - attributes the processing of these
offenses to the Municipality and the application of fines to the Mayor of Lisbon. These infractions and
sanctions are then communicated to the IMTT.
3.2.1.4. Main Stakeholders and bargaining power
When analyzing the spatial, operational, commercial and institutional influence of a taxi stand at
an airport, there are many stakeholders to account for. Taxi services in this context are often viewed
as more than just another road transportation mode. They are the preferred mode of transportation of
many traditional airport passengers and a critical intermodal link that allows the completion of an also
critical last segment of a long trip. Sometimes the taxi stands - important interfaces between air
transport and a fast, door-to-door, comfortable mobility service - are under the direct or indirect
responsibility of more than one institution. This may cause serious problems of coordination and
hierarchy definition, resulting in conflicts over liability, regulatory power, responsibility, funding,
implementation, monitoring, etc. It is important to analyze these several stakeholders and their
perceivable specific interests, because any proposal of policy/regulatory change will impact them in
different ways, to which each will react with proportional determination.
Because of the importance and scope of this service on a city-level context, there are many actors
other than the ones that actually complete the main transaction of this system: the taxi
drivers/companies and the airport passengers/visitors/employees. The third main stakeholder in this
process is the Airport itself, namely the airport authority or the airport manager/operator company,
which is interested in an effective, comfortable and reliable way of offering connectivity to its
passengers. By ensuring good quality of service for taxis at their terminals, the airport is also
improving its capacity to attract customers, to increase interest in its services/commercial concessions
and to improve efficiency in the processing of passenger flows within the terminal itself and at its
curbside. There are many other actors in this context, such as the Municipalities, National Transport
Regulators, Land Transport Competition, Hotels, Police forces, Tourism Associations, etc. All of these
entities and their specific interests will now be analyzed in further detail.
At Portela, there are some aspects that, as mentioned before, are related to the specificity of the
location and market context. Some of the stakeholders might differ from airport to airport according to
the market structure, the importance of the airport, the size of the city and of the taxi service supply,
regulations, etc. In order to clarify this important list of interested parties to this service, the following
stakeholders were identified:
46
Taxi companies, who usually hire or are composed of a collective of taxi drivers, and individual
entrepreneurs (which are often the drivers also), offer an individualized road transportation
service, with customizable choice of destination and route.
Interests: Taxi drivers want to maximize the profit on their service by transporting
passengers going to distant locations, preferably with luggage and, if possible, awarding
generous tips. They also want to minimize the waiting time in queue or “empty time” and be
assured of having customers when it is their time to pick up passengers.
Influence: The bargaining power and influence in the political decision making process of
the taxi sector in Lisbon is, similarly to other cities, perceived as relevant and significant. The
image of taxis blocking the city’s main roads in protest is a politician’s nightmare and
lobbying to stop or delay competition from the Metro at the airport has been frequently
rumored to have been exercised by the taxi sector somewhere in the past.
Inter-stakeholder relationships: The taxi sector, its associations, taxi federation,
companies and drivers are often criticized by passengers, because of alleged low quality of
service and price gauging. ANA, assuming its passengers perspective, also seems to
somewhat agree that quality of service and overall reputation of Lisbon (and Portuguese)
taxi drivers is not what it should be, and thus thinks taxi services at its terminal should be
significantly improved.
Taxi passengers, who are usually deplaning airport passengers, but can also be airport
employees or visitors, solicit transportation services that best adapt to their traditionally high
value-of-time, and thus choose the taxi as the transportation mode to complete their journey.
Interests: Taxi passengers want a reliable, fast, door-to-door and comfortable transportation
mode. They also want taxi drivers to be knowledgeable, friendly and professional, driving
clean, comfortable and safe vehicles. High prices are, of course, always the center of a lot of
complaints, as passengers obviously want to minimize the cost of their trip. Time and
availability is also an important factor for taxi passengers, who wish to minimize the waiting
time in queue and to have an empty taxi, ready to serve them.
Influence: Passengers are always influential in any transportation context, because they
constitute the demand for trips, which generates revenues through tariffs. Airport taxi
passengers are common customers to the airlines, airports, taxi companies and secondary
city businesses such as hotels, museums, shopping malls, etc, thus highly influential, both
with the politicians and the companies that serve them.
Inter-stakeholder relationships: Passengers often complain about taxi drivers, mainly
because of lack of vehicle hygiene and comfort, driver’s friendliness, geographical
knowledge, price gauging and even trip refusals. Recent studies made by the Portuguese
Consumer’s Defense Association (DECO) concluded that there are many taxi drivers at the
Arrivals stand that fool tourists into paying extra fees for trips, among other very serious
47
accusations and behaviors. (Pereira, 2009) This and many other statements made by
customers show that there is a feeling of general distrust between passengers and taxis.
The Airport Manager, in this case, ANA – Aeroportos de Portugal, which is the entity responsible
for the management of the entire airport infrastructure. This company, similarly to many other
airport operators, has concessions to commercial, logistic and service companies within the limits
of the airport, and especially at its terminals.
Interests: As the airport manager and looking at the specific situation of the curbside
operations, this entity wishes to efficiently process passenger flows, promote or offer good
connectivity options to its clients and minimize passenger complaints, which are often
directed at ANA itself, despite its weak intervention power. It also wishes to increase its
influence and participation on the management and planning process of such a critical
curbside service at its own terminal, currently very limited.
Influence: ANA is an important public-owned company, and politically influential, namely at
the central government level. The perception is that this significant influence does not
extend at the same degree to the municipal level, namely with the regulator, the Lisbon
Municipality. This influence also does not seem to extend to the taxi sector.
Inter-stakeholder relationships: There are some conflicts between ANA and the taxi
sector, as seen for example, with the taxi parking issue, which was moved to a different
location by pressure from ANA, some years ago, and caused protests and strikes from the
taxi drivers. ANA is also the target of some criticism from its customers because
passengers sometimes attribute some of the responsibility of bad taxi service to the airport
manager.
The Lisbon Municipality, which, in Portela’s case, is the market regulator for all taxi services
within the administrative limits of Lisbon. It is also responsible, like many other municipalities, for
most of the general urban planning aspects, from land use to transport planning. This includes
the responsibility to plan the location and capacity of all taxi stands, including at the Airport.
Interests: As a major public administration entity at the municipal level, the interests of
Lisbon Municipality are that the general public has access to transportation both in quality
and quantity of options – right to mobility - and that business thrives with responsibility
within city boundaries – economic development. This means that as a regulator of a market
such as the taxi service, it has to balance the rights and duties from agents on both the
supply and demand sides. Politically, satisfying passengers and taxi drivers without
ostracizing too much any of them is the main objective, which will eventually translate into
more votes for the political party in power.
Influence: As the main regulator and planning entity, the Lisbon Municipality’s influence in
the context of this service is naturally very high, although this point is more important to
48
other stakeholders, who pressure the municipality for decisions or interventions in their
favor.
Inter-stakeholder relationships: The Lisbon Municipality has assumed a role of pivoting
and managing of different interests. The weighted combination of those interests and the
alignment with the City’s own interests are always a priority, therefore, the several issues
that may come up should be dealt with diplomacy and a conflict-avoidance attitude. The
Lisbon Municipality is often criticized for having a heavy bureaucratic structure that
sometimes does not allow it to decide and intervene in due time.
Competition, for example, under the form of buses, rent-a-car companies, and in a near future,
the metro, is also a constant presence around Terminal 1, offering alternative mobility services to
airport passengers, based on other market segments and service type preferences.
Interests: All of these alternative modes of transportation are interested in maintaining
and increasing their share of the large flows of the typically high value-of-time passengers
that airlines usually transport. For this they also try to influence the Lisbon Municipality
and government transportation-related entities to allow them to keep or earn the right to
operate and compete with the taxi companies at the airport.
Influence: Carris is a very well-known and influential company in Lisbon’s public
transport context as the main incumbent company for surface transportation in the city. It
has a strong social impact and reputation, as well as strong political leverage. The metro
only recently approved the construction of the expansion of the red line from Oriente to
the airport, despite relative general consensus on its usefulness in the past. It is seen as
a very efficient, reliable and fast transportation mode, and will surely be a very strong
competition for taxis, if Portela is to keep its aeronautical infrastructure after the NAL.
Inter-stakeholder relationships: Not much is known about the real degree of
competition among taxis and other public transportation, but Carris – the main current
competition – has a very close relationship with the Lisbon Municipality and with most
Lisbon citizens. Tourists often travel on Carris as well, although recently, standard Carris
buses have ceased to allow the transport of typical air-travel hand luggage in their
vehicles. Carris is also seen by ANA as a partner for the improvement of curbside
transportation options, but their service quality is currently considered as insufficient, for
the passenger segment that usually travels to and from the airport.
The IMTT (Institute for Mobility and Land Transportation) is the main land transportation
regulator in Portugal. It is responsible for the licensing of transportation activities and service
providers, such as transportation companies or individual entrepreneurs. Its interests are mainly
those of a general transportation regulator such as promoting mobility to people all over the
country within certain quality, safety and price standards. Its influence in the taxi sector in Lisbon,
especially at the airport taxi stand is limited, as this is mostly responsibility of the municipalities. It
49
may have some influence on the licensing requirements of companies and drivers in general, but
it theoretically does not have the power of intervening in specific situations, such as this one.
The Police is, in this case, more than just the traditional main security force at an airport terminal.
It is responsible for the monitoring of the taxi service conditions, from the respect of the driver’s
behavior code to the vehicle’s safety requirements, according to the regulations of both Lisbon
municipality and general road regulations. It represents the main monitoring entity in this process,
although many consider ASAE (Food and Economic Safety Authority) eligible for monitoring
duties as well. The police have also given a significant coordination contribution to the operational
context, by persuading taxi drivers to avoid entering conflicts, resolving disputes and making an
effort to enforce the First-In First-Out discipline, near the passenger queue.
Other economic activities, such as Hotels, Convention Centers, Museums, traditional
commerce, shopping malls, and even business companies also benefit from an efficient and
reliable taxi service at the airport terminal. This is a direct feeder system for hotels and can
indirectly benefit other activities such as tourism and leisure, but also provide quicker connections
for travelling business employees, coming from abroad. Among the secondary interested parties
in a respectable taxi system at Portela is the Lisbon Tourism Association (ATL), a private but
publicly-managed company, which has the function of promoting tourism in Lisbon. The influence
of these entities is not as big as the ones directly involved, but it is, nonetheless, very real.
3.2.1.5. Institutional framework
After defining the main stakeholders involved in this context, it is important to know who is
responsible for which main basic functions, such as regulation, licensing, etc. and if those entities
have the conditions to exercise those roles efficiently and effectively. This last aspect is very important
to understanding the system, namely if the involved agents are prepared to perform these functions in
terms of independence, competence and intervening power. If overlapping of duties exists, this can
also pose a serious problem of coordination and significantly increase bureaucracy. The main roles of
the entities involved in the institutional framework of this system are relatively well defined (Figure 17),
with the exception of the planning process, where there are questions on the degree of participation
that ANA should have on the planning process of the taxi stands at Portela.
The regulator is the Municipality of Lisbon, empowered by law to define market access rules,
licensing requirements for vehicles, allowed service types, numerical limitations on parking regimes
and licenses for taxis, and stand location and general planning. The sharing of the licensing power is
clearly defined, with vehicle registration being the responsibility of the Municipality and the general
driver/company licensing belonging to the IMTT. On a financing perspective, the Lisbon Municipality
seems to be able to maintain good levels of independence from the main actors (passengers, taxis
and airport) and the national transport regulator, IMTT since it does not directly depend on neither of
them.
The main planning entity is also the
ANA. This aspect is important, since it
of inadequate planning or slow interven
This is in fact assumed by ANA as one
able to have a much faster and sig
operation of taxis at the curbside of its
The conceding entity is the Lisbon
service, taxi companies and drivers,
transport operators. ANA does not col
fees the City collects from taxi drivers
values for the valid licensing period of
The supervising and monitoring a
Municipality of Lisbon, who processe
overarching entity, which registers the
relevant behaviors of taxi drivers servi
end of the journey. This is usually
exchanged on account of provided ser
should this transaction be manipulated
accused of price gauging to extract gre
Figure 17 – Institutional
3.2.1.6. Overview sum
After the systematic analysis of the
at Lisbon Airport, it may be useful to
the perspective of the three basic elem
Regulator
Licensing Entity
Planning Entity
Conceding Entity
Operator(s)
50
the Lisbon Municipality, with the participation and co
it is mainly the airport manager who will suffer the d
entions from the Lisbon Municipality on the taxi stand
ne of the major downsides of the current framework. I
ignificant participation in the management and pla
ts Terminal.
n Municipality, which opens the taxi stands to the ope
, which are financially independent also, unlike
ollect any rents from this concession and the taxes
rs are not enforced by passenger or trip, but rather
f time.
authorities are mainly police, on a more operation
es the witnessed offenses to regulations and the
ese offenses at a central level. Currently it is not ea
ving at the airport, in the most critical segment of the
ly the moment when the transaction is made, an
ervice and only with some luck will there be a police
ted. This is also where most complaints focus, with
reater profits from gullible tourists and out-of-town pa
al framework regarding Portela’s taxi service system
ummary
e regulatory and institutional framework of the airpor
o summarize the main factors that might generate d
ments of transport policy: equity, sustainability and e
Passenger
ss
Lisbon City Municipalit
IMTT – Institute for Mobili
Land Transportation and Li
Municipality (vehicles)
Planning Department of
Lisbon City Municipality a
ANA – Aeroportos de Port
Lisbon City Municipalit
Licensed individual drivers
Lisbon Taxi companies
collaboration of
direct impacts
nds at Portela.
. It wants to be
lanning of the
perators of the
e many public
s and licensing
r through fixed
ional level, the
e IMTT, as an
asy to monitor
he taxi trip: the
and money is
ceman nearby,
h drivers being
passengers.
ort taxi service
discussion on
efficiency.
pality
ility and
d Lisbon
cles)
t of the
lity and
Portugal
ality
vers and
nies
51
There is an open system-like arrangement at Lisbon airport for taxis wanting to solicit service at
Terminal 1’s stands. This system has only two main basic restrictions, besides the mandatory IMTT
professional licensing of companies and drivers: the taxis that want to service the airport must be
Lisbon-registered, licensed vehicles and the maximum number of taxis serving any taxi stand is limited
to the corresponding parking regime (street hail or dispatch service) and spaces available. There is
also a global numerical limit to the contingent serving the municipal area of Lisbon. Thus, Lisbon’s
system can be qualified as Type C in Schaller’s classification of taxi regulatory systems (Figure 18).
Figure 18 - Schematic classification of taxicab regulatory systems (Schaller, 2007) – Lisbon case, in blue
This system has promoted sufficient demand for the airport and simultaneously ensured relatively
balanced taxi services in the city. The restriction to the parking capacity is an important measure to
avoid oversupply at certain stands and the open nature of the access to the airport stand also serves
to politically balance the taxi sector, creating competition and equal access rights. The fact that
demand is relatively stable at Portela - being the main airport in Portugal and Lisbon, and slowly
growing in annual traffic flows – also allows for this system to keep providing reasonably steady
results in terms of service request. It is seen by taxi drivers as a profitable and reliable market, for
which they are willing to wait several hours to enter. Although prices are set for the whole municipality
and taxi drivers and companies have little incentive to innovate or improve on regular service, there is
room for alternative service types and exploitation of different market segments.
The secondary taxi queue in front of Portela’s Departures Hall has been target to some criticism on
behalf of some of the traditional Arrivals taxi drivers, on account of the less waiting times in queue of
their colleagues and the possible demand-reducing effect that this has on the Arrivals stand,
prolonging their own waiting time for service. The existence of this stand can be seen as a diminishing
element of equity and efficiency, at first glance and based on some of these claims. But is just one
stand really better than two? Without more data on this second stand, it’s hard to be definite in taking
conclusions, but there are some system behaviors that might give a clue on what the answer might be.
If viewed from a homogenous service type perspective, one of the stands seems to be taking a
share of the other’s demand, basically competing against each other, but the long and persistent
queuing at the Arrivals seems to show, especially at peak times, that waiting times for taxis at the
Arrivals are more related to the service area configuration than to lack of demand. Since both stands
show similar behavior during the day and the Departures stand has about 25% of the Arrivals demand
52
(as we will see further ahead), transferring this demand would not significantly speed up the service at
the Arrivals, only increase the size of queues of passengers, which would take longer to serve. As a
whole, and looking from an efficiency point of view, closing down the Departures stand would probably
lead to the substantial increase of peak-hour passenger traffic at the Arrivals, while also increasing the
in-queue waiting time for passengers. This would allow for additional demand to exist, so more taxis at
the Arrivals would be serving, but it would be at the expense of the waiting time for passengers.
Equity is also questioned, but in the author’s opinion, with unfounded arguments. There is no real
restriction on where and when a taxi driver can park, waiting for service, other than driving a Lisbon
licensed taxi and having a free parking space at the waiting area. This means that the market’s
alleged hotspots, such as the Departure Stand, are accessible to everyone, everywhere, just as long
as they get there first, and the decision to wait more at the airport than in any other city taxi stand is
entirely up to the drivers, based on their own business perspective. Another key element to disarm this
claim is that the Departures stand is, based on the perception from many of the gathered opinions,
relatively unknown to most airport passengers. This means that there is a good possibility that a big
share of the demand at the Departures stand is based on the market segments that are more
familiarized with the airport, namely airport employees and frequent Lisbon Airport passengers, such
as businessmen and Lisbon citizens. These experienced passengers are arguably less lucrative than
the Arrivals clients, because they usually have little to no luggage, are probably travelling to nearby
destinations, such as the city center (business) and Lisbon suburbs (returning home), and some of
them are known by taxi drivers to be traditionally less generous on the tip, unlike occasional tourists.
Sustainability of the service and current regulatory and institutional design is threatened by the
certain future introduction of extra land transport competition for airport passengers, namely the Metro,
which is due to begin operation in December 2010 (source: Metropolitano de Lisboa). This is
somewhat offset by the fact that a new airport is going to be built, with the inauguration date foreseen
for 2017, which means Portela will receive less passengers (if any at all) and the taxi companies will
transfer their business focus to the new airport, avoiding this competition.
Regarding regulation and taking into account the usual complaints about price gauging, lack of
driver friendliness, professionalism and hygiene made by airport passengers, stronger and more
persuasive monitoring and penalty systems should be implemented. Recently, sector-related
associations have met to discuss possible changes to the airport service system, namely to improve
the image tourists and Lisbon citizens have of the taxi service in Portugal. Some of these changes are
focused on the creation of a local monitoring commission, formed by ANA and the taxi associations, in
order to better manage and control the stands at Portela and quickly resolve conflicts and other
issues. These proposals are still currently under review by state and municipal entities.
ANA wishes to be an active and influential actor in the planning of the airport stands, not only on a
spatial perspective, as to how the stand interacts with the other Terminal functions, but also on the
flexibility of the service to handle peak-hour passenger traffic, the stand size and mechanisms by
which taxis pick up passengers, interaction with police and taxi drivers, etc. It also argues it should
53
have more intervention power when certain changes are justified by low quality of service conditions.
Waiting for the Lisbon Municipality to listen and to act might cause significant nuisance to ANA as an
airport operator, responsible for maintaining efficient curbside operations.
The taxi drivers and companies usually do not have an easy relationship with customers, in the
sense that general mistrust is installed on account of certain past and recurrent behaviors and
problems. ANA mostly takes the passenger perspective as their own, looking at current taxi services
as insufficient and problematic, mainly because it is one of the main recipients of complaints made by
airport passengers. The Lisbon Municipality is considered to be a highly bureaucratic entity, with a
very heavy institutional structure, which often slows the decision making process and blocks
immediate changes. Although this is recognized as a problem, mainly by ANA, the Municipality’s role
of regulator is not questioned, as this responsibility is, by law, theirs to bear.
The taxi stands are made accessible for operation by the Lisbon Municipality to all licensed drivers
and companies who wish to service the Lisbon area - ANA does not collect rents from the taxi
companies and drivers. This means that every taxi stand is under the direct responsibility of the
municipality, including the special airport taxi stand. Inefficiencies and lack of perspective may emerge
from this situation because a large, bureaucratic and multi-function entity, in charge of managing and
planning of dozens of taxi stands might not pay sufficient attention the specificity of the airport stand.
There are several aspects about Portela’s taxi stands that differ from those of regular stands
downtown, and “normalizing” these stands can be very prejudicial to the system.
One very important actor in this context is the Police, which has mainly been seen as a security
force, at airports. In Portela’s case, the Police forces are also in charge of coordinating the taxis
towards their service spots, at the Arrival stands. They persuade taxi drivers to avoid conflicts and
respect the first-in first-out regime, integrated in every taxi stand in Lisbon. This role could arguably be
performed by another entity, one which could be internalized into ANA’s structure with a strong taxi
association contribution or even co-management, releasing the Police for more important duties.
3.2.2. Operational Context
After looking at the institutional and regulatory context, a closer look at the operational mechanisms
is necessary, to be able to better sustain the arguments for possible changes and base new proposals
on concrete quantitative data and measurable impacts. The study of the operational context is based
on the several steps of the methodology presented in Chapter 2:
1. Identification of the problem of queuing at Airport Taxi Stands ;
2. Literature review and general queuing theory research ;
3. Identification of a suitable and real case-study ;
4. Preliminary in situ observations of system behavior ;
5. Elaboration of a Data Collection Plan ;
6. Test data collection procedures ;
54
7. Collect relevant data ;
8. Compile and analyze the collected data ;
9. Build basic queuing simulation model ;
10. Test and validate the basic simulation model based on the collected data ;
11. Perform scenario building and testing ;
12. Results analysis and conclusions.
Some of these steps have already been inherently addressed at specific points throughout this
document, integrated in the framework definition of the problem and in the research effort that
preceded this more detailed analysis on the subject. Step 1 - Identification of the problem of queuing
at Airport Taxi Stands, has been discussed both on the regulatory and operational perspectives in
Sections 2.1.1 and 2.2.1 of Chapter 2, immediately before the methodology sequence proposals. Step
2 - Literature review and general queuing theory research, is present in Chapter 1, where some
important bibliographical sources are reviewed as the basis for the theoretical background and on the
state-of-the-art of this subject. Step 3 - Identification of a suitable and real case-study is addressed in
the introductory section of this Chapter, where the main reasons for choosing Portela as the focus of
analysis are presented and justified. Steps 4 through 7, related to the data collection procedures were
also subject of detailed analysis in the Field Data Collection Plan, present in Section 2.2.3 of Chapter
2. The analysis of this topic is therefore developed from Step 8 - Compile and analyze the collected
data onwards, preceded by a spatial description of the system, for contextualization.
3.2.2.1. Spatial description of the system
As mentioned earlier, Terminal 1 has not one, but two taxi stands, which for a terminal of its size,
can be strange to most foreign – and also many domestic - passengers, who normally have no
information on the existence of this secondary taxi stand. This stand is located on the lateral side of
the Terminal, in front of the Departures entrance (see Figure 19). This taxi stand, according to a study
by ANA, in 2006, during Easter week, has about 25% of the demand registered at the Arrivals Taxi
Stand and a lower average occupancy rate of 1,69 passengers/taxi, compared to the 2,06
passengers/taxi at the Arrivals. This study, conducted in order to determine the operational needs in
terms of space and taxi numbers, also shows similar distribution of passenger demand during the day.
This taxi stand is known to serve mainly airport/airline employees and the few people who know of its
existence and regularly use the airport. This secondary supply point has been a topic of discussion
and some turmoil among some taxi drivers, related to fairness of competing for the same type and
source of passengers, while having to wait less time in queue.
Initially, this secondary stand was considered as a possible case study, parallel to the Arrivals
stand, which is clearly the most relevant and problematic, in terms of queuing lengths and waiting
times. The time and resources needed for the field data collection and the effort involved in analyzing
this data, coupled with the need - and choice - to do a more in-depth analysis on the Arrivals stand
lead to a less detailed analysis of the operational characteristics of this stand. Regardless of not
55
focusing on its specific system mechanics, its influence and impacts on global taxi service at the
terminal are still considered in a wider context.
The taxi stand for the Arrivals has also been subject to measurements in 2006, included in the
previously referred ANA study. This study concluded that about 30% of the deplaning passengers
used the taxi service, an average of 87 taxis/hour were present and the maximum solicitations
occurred at peak hours during the morning (9-11 a.m.), afternoon (3-5 p.m.) and night (10-11 p.m.).
This stand will be the main focus of analysis due to its size and relevance on Terminal 1’s curbside
operations context and due to the well known problems of excessive queuing, price gauging, trip
refusals and other related complaints usually made by taxi passengers, at Lisbon Airport. This stand
processed, on Easter week of 2006, a total of about 22.500 passengers, with an average of
approximately 3.200 passengers/day and 180 passengers/hour.
Finally, there is an exclusive taxi parking lot, free of charge, built in 2003, about fifty meters from
the Terminal, where taxis form long queues waiting for service at the main Arrivals stand. It has a
capacity for about 150 to 200 taxis, a leisure/waiting room, food and drink machines, bathrooms and a
security system, ensured by police. This capacity plus the maximum queue length from the parking
facility to the Terminal restricts the number of taxis that can serve the airport at any given moment.
Initially taxis waited for service at P1, a large parking facility located next to the Arrivals Stand. This
transfer to the new parking lot was faced, at the time, with protests and even strikes by the taxi drivers.
Figure 19 – Lisbon Airport - Terminal 1 (Source: Google Earth)
The front side of the building (Arrivals) has four main entrances (see Figure 20), one located at the
far end of the building front, one near the taxi service area and another two near the entrance of the
taxi passenger queue. The taxis form a queue originating from the nearby parking lot, along a
segregated lane of the access road, splitting into two lanes as they run along the curbside of the
building. This segregated access lane has two “conflicts” with the road access to P1 (entrance and
exit), a general parking lot with a capacity of 300 vehicles, located at right side of the Arrivals, on the
Arrivals Taxi Stand
Departures Taxi Stand
Taxi Parking Lot
P1
56
same level. This may sometimes cause some delays on the taxi routing towards the service area, as
taxis might have to wait for other vehicles to be able to enter the main road. There is a strong police
presence at the curbside, including at the taxi stand itself, not only to monitor illegal parking of private
cars, but also to coordinate and help discipline the queuing and service of taxis. Also at the curbside,
several bus stations are present, mainly serving Carris (general surface transportation operator in
Lisbon). A future metro link – the response to a long-awaited claim of airport passengers and Lisbon
citizens in general - is currently under construction, connecting the airport to the Metro red line.
Figure 20 – Taxi Service Organization at the Arrivals of Terminal 1
As mentioned earlier, there are four main terminal entrances/exits, which lead to crosswalks,
immediately in front of these entrances/exits. There is a passenger queue which extends between two
of these exits, signaled with a “TAXI” sign, at its entrance, near the far right terminal exit (see Figure
21). This queue is composed of three “snake-like” corridors, managed by police, which, similarly to
check-in or security check queuing systems, open more “corridors” as more people arrive. These three
corridors were visually observed at peak-hours and estimated to have an approximate maximum
capacity of 30 people/each, in a total delimited in-queue capacity of about 90 people.
Observations at peak-hours have confirmed occasional formation of secondary queues, originating
from the two closest terminal exits, which occupy the small space between the terminal and the
queue, and several queues going beyond the defined space for the taxi passenger waiting area.
These fast-growing queues are often in conflict with three of the four terminal entrances, expanding
beyond the bars defining the waiting space, obstructing them, along with most of the inner curbside
area of the terminal. This phenomenon is recurrent at peak times and may cause serious problems of
pedestrian congestion, conflicts between passengers from different queues and even safety and
security problems, related to emergencies and evacuation procedures.
Taxis are always present in significant numbers, almost at any time of day, parked along the
curbside of the terminal in a single row which splits into two closer to the passenger queue. This taxi
queue extends to the parking lot located more than fifty meters away from the Terminal, along a
segregated lane of the main access road. Supply of service does not seem to represent a restriction,
as the taxi parking facility rarely empties during airport working hours, and taxis keep coming in to join
Service Area
Bus Stops
Parking Entrance
Terminal Entrances
Taxi Queue
Passenger Queue
Parking Exit
57
the queue. Despite great number of taxis, supply is intermittent closer to the passenger queue, at the
service area itself, namely due to service area characteristics. This service area is configured so that it
is possible, in optimal conditions, the simultaneous loading of a maximum of four taxis at a time.
It has been observed in situ that the system closer to the passenger queue is composed of two
parallel taxi rows or lanes (inner Row A and outer Row B – see Figure 21) and service is usually
restricted to the four front parking positions. Server 4 does not seem to show equal behavior, in terms
of service times, to the other servers and is sometimes empty due to a conjunction of factors that will
be analyzed with greater detail further on. There are some conflicts between taxi flow and the
pedestrian flows from cross-walks. First-In First-Out discipline is supposedly present both at the
passenger queues and taxi queues. Due to the splitting of the single original segregated lane into two
service lanes/rows, taxis are ordered to stop, advance or bypass other taxis parked at the inner Row A
by a police element, always present at the stand, especially at peak-times. This somewhat arbitrary,
error-prone and random-like selection and organization is sometimes contested by taxi drivers and
discussions with the police element and among themselves are recurrent.
Figure 21 – System configuration at the Arrivals Taxi Stand
1 2
3 4
Row A Row B
Taxi Service Spots
Passenger Queue
Queue Entrance
Crosswalks
Queue Exit
58
3.2.2.2. Analysis of the Collected Data
The field data collection effort was essential for the understanding of the system behaviors and
quantification of its functioning. The available data on taxi service systems at airports is either very
difficult to access or largely incomplete or non-existent. The fact is that most of the literature on taxi
service systems at airports is more focused on the regulatory aspects than on the quantitative
performance measures of the service, especially relating to passenger queues. To add to this lack of
information on the functional parameters of the system, there are also two other factors that contribute
for the inexistence of detailed and relevant data on taxi stands. Both of these factors are related to the
analysis performed on taxi services, as it is the case with the measurements made by ANA in 2006.
These measurements focus on the supply side, such as the adequate fleet size serving the airport
versus the total number of passengers that require service at a given time of the day, than actually on
key demand characteristics such as in-queue waiting times and queue lengths. Another aspect that
discourages the extended use of the little data that is available is the fact that many times it also
focuses on averages. ANA’s measurements of total number of taxis and passengers at the arrivals
and departures stands were made by counting these elements on an hourly basis, throughout the
active period of the airport’s operation. These average values are opaque to the extreme behaviors
that occur at peak hours, during which queues rapidly form and the system is flooded with arriving
passengers, significantly diminishing during the following hours. At these times queue lengths and
waiting times can increase almost exponentially, causing several problems in terms of curbside space
and passenger discomfort. This problem is at the core of this study, and a different approach, based
on the observation of peak-hour behaviors, has been chosen.
The collected data at Portela (see Annex II) was based upon the need to have more information on
three key aspects: the arrivals of passengers, namely the inter-arrival times and the group
composition; the service of taxis, namely the service times, including the “empty time” between
availability of service; and the passenger queue evolution, namely the queue length and the in-queue
waiting time. The methodology and reasoning used for the identification of these key aspects and for
the measurements of the relevant indicators is present in the Data Collection Plan, on Chapter 2. After
compiling and processing this data, it is now useful to present some of the main results of the analysis,
in order to better justify the choices for the simulation model and better quantify system behavior.
Arrival of Passengers
The arrival of passengers to the queue was subject to measurements at two different moments in
time, each on a separate day and peak-hour:
Wednesday, 5th
of August, from about 8 to 10 a.m., in a total of 275 observations, although
some of these were later excluded from the main analysis because, from about 8 to 9 a.m., peak
hour conditions were not fully observed.
Thursday, 13th
of August, from about 9 to 10 p.m., in a total of 234 observations.
59
During both of these periods, two indicators were measured, processed and compiled, for which
the main histograms are shown below:
Inter-Arrival Times (for Groups) - see Figure 22;
Group Size (based on the perception of the observer) – see Figure 23.
Figure 22 – Histogram for Inter-Arrival Times for Groups
Figure 23 – Histogram for Group Size
There are some interesting conclusions to draw from this data on the arrivals of passengers. The
first interesting aspect of this analysis is that the majority of the inter-arrival times are below 10
seconds and about 70% of all inter-arrival times are below 20 seconds. The mean for the total
observed inter-arrival times is 17 seconds and the standard deviation is 15 seconds. This means that
individuals or groups arrive at a very high rate, consistent with a peak-hour situation. Another relevant
conclusion is that the majority of the people soliciting a taxi were composed either by a single person
(36%) or a group of two people (41%), usually couples. Although less frequent, groups of 3 (15%) and
4 persons (6%) are still relevant for the group structure of this arrival stream, especially because
above 3 people, groups tend to take longer to coordinate and divide (themselves and their luggage)
among taxis, which have to agree on going to the same destination. Once more, it should be noted
201
102
5730
14 9 6 6
0,00%
20,00%
40,00%
60,00%
80,00%
100,00%
0
50
100
150
200
250
≤10 20 30 40 50 60 70 80 90 100 110 120 More
Fre
qu
en
cy
Time (seconds)
Frequency Cumulative %
153
175
62
27
5 2 10,00%
20,00%
40,00%
60,00%
80,00%
100,00%
0
50
100
150
200
1 2 3 4 5 6 7 8 9 More
Fre
qu
en
cy
Size of Group (persons)
Frequency Cumulative %
that this consideration of group sizes
considered group behavior.
The measurement of group sizes,
service area, also allows for the estima
peak-hour demand. It is therefore imp
order to more safely make the assu
independent of the time of measure
comparison between proportions of
differences between the two days were
groups of 3 people. These differences
of the abovementioned assumptions.
were the average values between the t
Figure 24 – Comparison between
Service Times for Taxis
Taxi service times were subject of m
Wednesday, 5th
of August, from
Thursday, 27th
of August, from a
Monday, 14th
of September, from
The first two measurement sets w
Server 1 and 3 and the last one was do
Initially, the servers were assumed
distribution for measurement purposes
server’s state (busy back servers bl
measured as the sum of the time it ta
passenger(s) to board the taxi. The
39,3%42,9%
10,2%
37,6% 35,0%
17,9%
0,0%
10,0%
20,0%
30,0%
40,0%
50,0%
Singles Groups of
2
Groups o
3
Proportion of Total (
60
es is based on the observer’s visual perception of
s, besides giving clues to the possible increase in d
ation of an approximate number of taxis that are req
portant to have measured this factor at different p
umption that these proportions are more or less c
rement, simplifying the modeling of the system ar
f group sizes was made, (see Figure 24) and th
re 7,9% regarding the groups of 2 people and 7,7%
s were not considered as relevant enough to underm
s. The values used for the posterior modeling of g
e two data sets.
en Group Size Proportions from the different measurem
f measurement on three different occasions:
m about 10 to 11 a.m., in a total of 73 observations.
about 9 to 10 p.m., in a total of 94 observations.
om about 9 to 10 p.m., in a total of 75 observations.
were done on the inner-Row A servers (see Figure
done on the outer-Row B servers, namely Server 2 a
ed as independent and relatively similar in terms of
es. Because server availability is sometimes conditio
block front servers, for example) the service tim
takes for a taxi to reach the service spot and the tim
e fact is, as explained in the Data Collection P
5,8%1,1% 0,4% 0,0% 0,0%
7,9%
6,4%1,7% 0,9% 0,4% 0,0%
ps of Groups of
4
Groups of
5
Groups of
6
Groups of
7
Groups of
8
tal (5-8-2009) Proportion of Total (13-8-2009)
f what can be
delays at the
required for the
peak-hours in
constant and
arrivals. So, a
the maximum
regarding the
rmine the basis
group arrivals
ements
re 21), namely
and 4.
of service time
tioned by other
me had to be
ime it takes for
Plan, that the
0,4%,0% 0,0%
s of Groups of
9
simultaneous loading of four server
technique (stopwatch method), especi
The service area close to the queue
children, elderly people, police, trolleys
in such a way as to not interfere, distra
with the few human resources availab
as the relevance of data collected indiv
Based on these assumptions, the i
observations. After some time, anothe
that the inner row and the outer row
suspicion became stronger as the an
being conducted and the model test
behavior. This motivated the third set
credibility of these suspicions. The serv
Figure 2
After analyzing the service times an
Unit
Server
Average
(se
con
ds)
Standard Deviation
Global Row average
Global Row Standard Deviation
Figure 26 – Serv
0
2
4
6
8
10
12
14
16
≤20 30 40 50 60 70 80 90
Fre
qu
en
cy
Time
Server 1 Server 2
Server 3 Server 4
61
ers increases the complexity of the employed m
cially if the number of observers is small, such as i
ue exit can be a very confusing stage of taxi drive
s, etc. The observation method also had to be relati
tract or disturb the main actors in any way. These fac
ble to the author imposed some simplification assum
ividually, one server at a time.
inner-row servers were chosen as the focus of the
her service area tendency became increasingly pre
w could be behaving differently, in terms of service
analysis on the queue lengths and in-queue waitin
sted, and results were not totally in accordance
t of observations done on servers 2 and 4, in order
ervice time histograms of all servers are plotted on Fi
25 – Histograms for Service Times
and building the histograms, the following values wer
Row A Row B
Server 1 Server 3 Server 2
66,8 73,5 67,6
33,2 35,5 30,0
70,0 76,7
34,4 37,3
rvice Time averages and standard deviations
100 110 120 130 140 150 160 170 180 190 200 210 220
ime Intervals (Seconds)
measurement
it is the case.
ivers, luggage,
tively discrete,
actors, coupled
umptions, such
e service time
resent, namely
ice times. This
ting times was
to witnessed
er to verify the
Figure 25.
ere calculated:
w B
Server 4
90,4
43,2
,7
,3
20 230
62
These values (see Figure 26) coupled with the information on the service times distributions of the
different servers, show that server 2 does not evidence any significant difference regarding the
averages, standard deviations or time distributions of the inner-row (Row A) servers. Server 4
however, is indeed different, judging from the collected data, in the sense that its average values for
service times are about 23% higher than the average of the remaining servers and about 33% higher
than the average for Server 2. The measurements on Servers 2 and 4 were made under identical
peak-hour conditions to the remaining observations on the other Servers, without any additional
perceivable interference or special circumstance. This difference can be explained by a conjunction of
factors that were witnessed in every observation period throughout the data collection (Figure 27):
Splitting of the single queue lane into two rows, close to the curbside of the Terminal –is
perhaps the most important factor that influences Server 4’s availability and may generally be
increasing service times for all servers. The taxis form a single queue on the inner-row and then
must bypass the front taxis to occupy service positions 2 and 4. This movement is often done
under the supervision of the police agent that is usually coordinating the operation, trying to
maintain First-In First-Out queue discipline. The problem resides in the reluctance of some drivers
to bypass their colleagues because they perceive that the service points upfront are almost
available and don’t want to lose their position or because of their unawareness that there is an
available service spot further ahead in the other lane (line of sight issues). The fact is that with
people crossing the road, timing of policemen authorization, other taxis maneuvering and the long
queuing, this movement is slowed down, especially for Server 4, since the front Server 2 is
frequently occupied by the third taxi on Row A.
Coordination and authorization of police – as mentioned above, there is always at least one
police element present at the curbside to monitor and help coordinate the taxis to position
themselves to pick up clients at the service area. This agent is sometimes slow in his action to
authorize the taxis on the back to move to the front by bypassing the inner-row taxis. This is due
to the complexity of the surrounding environment and the several factors this element has to be
aware of, such as the way passengers are treated and distribute themselves, the spatial positions
where taxis park, the conflicts among taxi drivers, etc. Also, when he does call taxis at the back to
move along, they sometimes react slowly or don’t pay immediate attention.
Passengers crossing Row A and other difficulties – In order to reach server 4, passengers
have to cross Row A and some of the area reserved for Server 1 and 2. This usually means that
both passengers and taxis have to be cautious in their movement, slowing down the process. It is
also the Server farthest from the queue exit, and the line of sight from there to Server 4 can
sometimes be blocked by other taxis, depending on the exact parking position of the other taxis.
This coupled with the carrying of luggage, existence of children or elderly people or a big group
can increase the time it takes to complete the service cycle.
63
Figure 27 – Main conflicts that can justify delays and differences in service time distributions
Queue Length and In-Queue Waiting Time
In the Data Collection Plan, the reasons for not directly measuring queue length and in-queue
waiting time are presented. They are based on the complexity of rigorously observing the functioning
of the system, at two separate observation points in real time, especially the service area, for which
more people would be required. Nevertheless, if we consider that inter-arrival times and service times
are independent - a fair assumption, because usually taxis do not significantly increase their time
efficiency, just because more people arrive – we can measure them separately, at different times.
Based on this assumption, we can then “fit” both time distributions and create arrival and service flows
during a specific period of time, corresponding to individual group arrival and departure instants (to
and from the queue). Although these measurements took place at different days at different hours,
they both represent peak-hour conditions, so the behavior of the system at any of those moments
would most likely be similar to the assumed behavior.
For the Empirical estimation of queue length and in-queue waiting time, an indirect method was
used, based on the methodology suggested by (Newell, 1982). This method consists in observing the
arrival-to-queue and exit-from-queue instants of the groups/individuals and plotting the cumulative
arrival instants/cumulative arrived passengers curve and the cumulative exit instants/cumulative exited
passengers curve on the same chart. After this process, the queue length in terms of number of
people, is measured at each instant by the vertical distance between the two curves (if the cumulative
number of passengers is on the vertical axis) and the in-queue waiting time is determined by
measuring the horizontal distance between the same two curves.
In order to adapt the different arrival times and service times distributions to the same dataset and
capture the system behavior at a period of highest solicitation, some simplifications, assumptions and
decisions were made, as can be seen below.
Conflict
Row A
Taxi Queue
Row B
Police element
Terminal Exits
Passenger Queue
Crosswalks
Service Area
64
The chosen duration for this “sample period” was 1 hour, due to the fact that most of the
measurements were taken over periods of 1 hour and that normally peak-times at Portela last for
about 1 to 2 hours maximum.
For the arrival flow of 1 hour, the busiest two half-hours were chosen sequentially, from the total
data available on inter-arrival times – 9:30 to 10 a.m. of the 5th of August (218 people) and 9:30 to
10 p.m. of the 13th of August (312 people).
The observed pairs (arrival time; group size) were preserved, including by order of measurement.
The service times were divided by Server, according to the observed data, but some additional
values were generated for Server 2 and 4, through a random number generation function, based
on one of the theoretical distributions that better fitted the data – the Lognormal distribution.
Because Server 4 shows relatively higher service times compared to other servers, the generated
values for Server 2 were multiplied by a factor of 1,3 for Server 4. These additional values had to
be generated because of the lower amount of observations done on these last two servers, which
didn’t allow for the representation of their contribution to the service rate during this whole hour.
These service times were considered to have been experienced by the arriving groups, following
a First-In First-Out discipline, meaning each group would exit the queue in the order they arrived.
In order to know at which time each group exited the queue, the “cumulative service times” (in
other words, the instants at which a server would be available and the client leaves) were
determined by server and then compiled as a whole for the system, sorted by increasing value.
Then, the arrival times were compared to these service instants and the maximum of both values
was chosen - this is the exit time of that group. If the value is equal to the arrival time, it means
the passenger didn’t have to wait at all and had an empty cab ready to pick him up immediately.
The two curves were compared by means of linear interpolation, in order to find the common
values for the two axis and determine the queue length and in-queue waiting time distributions
and their corresponding average, standard deviation and maximum values.
Figure 28 – Arrival Curve and Exit Curve based on the collected data
3566; 530 4777; 530
0
100
200
300
400
500
600
0 600 1200 1800 2400 3000 3600 4200 4800
Cu
mu
lati
ve
nu
mb
er
of
pa
sse
ng
ers
Cumulative Elapsed Time (Seconds)
ARRIVAL
SERVICE
Queue Length
In-Queue Waiting Time
65
This chart (Figure 28) allows a more detailed view on the queue evolution regarding two key
indicators, the queue length (persons) and in-queue waiting times (seconds). During the first 25
minutes, the system would be capable of relatively handling the arrival flows without creating any
significant queue of people. After this, the arrival flows rapidly grow at a much higher pace than the
service, which keeps relatively constant – this is consistent with service time independence
assumptions. These queues expand as the elapsed time increases, reaching its peak by the end of
the considered period, for a maximum of 130 people and about 20 minutes in-queue waiting time.
There are some possible explanations for this behavior. After analyzing the two half-hours that
constitute the arrival flow, we reach the conclusion that the first half hour (measured in the morning of
the 5th of August) was “weaker” than the second half hour (measured in the evening of the 13
th of
August), in terms of passenger arrivals – a significant difference of 94 people. This does not
necessarily mean that morning and evening peak-hours are substantially different in terms of arriving
flows. It can be caused by common flight delays that slightly alter the composition of the “half-hour
peak periods” or even that the measured period consisted of the first growth phase of the peak-hour,
during which the arrival rate is rapidly growing but does not necessarily reach its maximum value.
These different paces allow for a sustainable 25-minute service, after which long queuing inevitably
appears, reaching a peak value and eventually dissipating as the peak-period then comes to an end.
Figure 29 – Queue Length evolution
Figure 30 – In-Queue Waiting Time evolution
130
0
20
40
60
80
100
120
140
0 600 1200 1800 2400 3000 3600
Qu
eu
e L
en
gth
(N
um
be
r o
f P
eo
ple
)
Cumulative Elapsed Time (Seconds)
1211
0,0
300,0
600,0
900,0
1200,0
1500,0
0 600 1200 1800 2400 3000 3600
In-Q
ue
ue
Wa
itin
g T
ime
(S
eco
nd
s)
Cumulative Elapsed Time (Seconds)
66
The queue length (Figure 29) and in-queue waiting times (Figure 30) evolutions were also plotted,
in order to better observe the queue behavior over time. Both indicators show somewhat similar
evolution, which makes sense because usually as queues grow bigger, so does the waiting time
experienced in them. Also to be noted is the fact that for a certain interval of this experiment period –
from 33rd
to the 50th minute of elapsed time - the queue seems to be stabilizing at about 60 people,
with an in-queue waiting time of about 500 seconds (8 minutes).
The final results of these calculations for the real data are shown in Figure 31. One important
aspect that needs to be underlined is that the reasoning for the determination of queue length and in-
queue waiting times is subject to a substantial degree of error. This happens mainly because there
was need to adapt inter-arrival times and service times of 4 different servers, all taken during different
measurement periods, and attempt to “fit” them together. The choices, assumptions and
simplifications made to achieve this - as mentioned at the start of this analysis - can produce some
discrepancies regarding some of the indicators that can be extracted from this process, but hopefully
adequately represent the witnessed field behavior. The results seem to be in accordance with what
was perceived during the several in situ observations.
Queue
Length
(persons)
In-Queue
Waiting Time
(seconds)
Average 37 315
Standard Deviation 37 321
Maximum value 130 1211
Figure 31 – Main results for Queue Length and In-Queue Waiting Time
Figure 31 proves that queuing at Portela can become very problematic. If the arrival rate of
passengers is as aggressively high as it was during this peak-hour period (especially the second half-
hour) and service rate is relatively constant (as it seems to be the case throughout the entire
measurement period), a maximum of 130 people will be queuing by the end of the hour. This arrival
rate tends to significantly decrease after the peak-time is over, slowly decompressing the queue
during the following hour or two. But the fact that so many people, with their luggage and trolleys, can
be concentrated at the curbside waiting for approximately 20 minutes for a taxi, should be considered
bad service, to say the least. The average queue is about 37 people-long, but this can be somewhat
deceiving - as most averages usually tend to be – on account of the first half-hour period, which is
characterized as significantly lighter than the following one. This slower rate of arrivals lowers the
average queue size because no queues are formed until approximately the 25th minute, after which it
grows very fast. If we average the queue size for people who indeed queue at all - considering only
the values after the 25th minute – then the average queue size jumps to 60 people, which is even more
consistent with the observed reality at Portela. If we consider that the observed maximum capacity of
each of the corridors of the passenger queue is about 30 people, then by the end of this period, we
would have the entire queue completely filled with passengers and an extension of about another
67
corridor-long of queuing to the exterior of this delimited space. This queue would probably form in front
of the two terminal exits/entrances to the right of the Terminal façade and/or eventually a secondary
queue could also form, coming from the left Terminal entrance, closer to the passenger queue exit.
Similar phenomenon has been witnessed during the measurement process, causing serious problems
of pedestrian congestion and passenger impatience and discomfort.
3.2.2.3. Simulation Model
The simulation model is a very important part of this study, in the sense that it allows the
manipulation of the system characteristics, such as arrival and service rates, number of servers,
queues, queue discipline, introduction of new bottlenecks, routing of passengers, time-dependent
behaviors, etc. This model is the instrument through which different possible system configurations
and special conditions are tested, with almost immediate results on many performance indicators,
without entering into very complex mathematical analytical considerations. The simulation model is
essential to learn about the sensitivity of system behavior to certain changes in its elements.
The building of the simulation model structure is a key stage of this process. The assumptions
about the required data and the translation from reality to the simulation background must be done
with caution and adequate detail, in order to avoid building a very complex simulation model that does
not mimic field behavior correctly. For the construction of this model, SIMUL8 software was chosen.
SIMUL8 is a computer package for Discrete Event Simulation from SIMUL8 Corporation. It is
frequently used in the modeling of industrial processes or services such as hospitals, repair shops,
gas stations, etc., focusing on queuing systems. This choice was based on three main reasons:
The software explicitly considers randomness and variability, namely through the possibility of
modeling arrival and service rates with recourse to several known theoretical distributions, or
even external ones (from EXCEL, etc.).
It is user-friendly and visually simple, focusing on the main elements of a queuing system (work
entry point, work centers/servers, queues and work exit points), allowing for an intuitive interface
with the inherent complexity of such a system, promoting a fast learning curve.
It is easily accessible through a free-license for educational purposes, thus available at any time.
Typical SIMUL8 objects might be work items, queues or work centers (servers). The work items
may be physical entities such as manufactured goods, which could be held in a storage area before
being processed on a machine, or they may be virtual work items such as telephone enquiries, which
are held in a virtual queue before being processed by an operator. These work items can also have
several attributes based on a label system. When the structure of the model has been built, the
software can perform a series of trials in order to statistically describe system performance. Many
other object attributes may be defined, such as the inter-arrival time distributions or the service time
distributions, routing discipline to other objects, queue capacity, operating shifts, breakdown
probabilities, etc. Statistics of interest may be average waiting times, average queue lengths,
utilization of work centers or resources, etc. (Días Esteban, 2008)
68
Model building
During the transition from the conceptual model, built from the observed reality, to the simulation
model, there are significant simplifications that must be considered. Some of these simplifications are
a result of the impossibility to perfectly and accurately model certain real behaviors or system
elements and the need to balance complexity with obtaining credible results on key indicators. The
main objective of this simulation model is to mimic the observed queue behavior, so the focus should
be on the arrival and service rates at the queue. Examples of this are the arrival flow, which even
though originating from four possible terminal exit points, is modeled as a single stream of passengers
or the effect of the crosswalks that is assumed to be included in the service time distribution values.
Simul8 uses specific building blocks to represent queuing system elements. Each of these building
blocks has a set of attributes and properties that model the way it accomplishes its function. Before
presenting the simulation system, it may be useful to define and describe these building blocks.
Work Entry Point
This object is a work items generator, a source of entities to be processed by the system, or in this
case, potential taxi passengers coming from the Terminal at the Arrivals. The Work Entry Point is
characterized by an arrival pattern. This arrival pattern of the work items can be controlled in order to
follow a scheduled arrival pattern (deterministic behavior) or a particular probability distribution
(stochastic behavior). In this case, the Work Entry Point generates groups of passengers according to
an inter-arrival time distribution, as will be discussed in more detail further ahead.
Work Center
The work centers represent servers, machines and other processing elements that perform a
certain function or job on the work item(s) they receive, that lasts a certain period of time and may
require the use or consumption of a certain amount of resources. These work centers may also be
modeled to assume certain “Routing In” rules, especially if there is more than one origin and certain
“Routing Out” disciplines, especially if there is more than one destination. The time that the work
center requires to perform a job may be described by probability distributions. In this case, there was
no need to use resources as we assume the taxi supply as basically infinite during the simulation
period. The Work Center object is therefore the taxi service position, which when available means
there is an empty taxi parked there, ready to serve. Work Centers may also be performing virtual
functions, serving as proxies for the modeling of such factors as group size, etc. In this case, there are
four main work centers, representing the taxi service spots.
69
Queue
Here the work items, in this case the passengers, are held while they are waiting to be processed
(picked up by a taxi). The idea of queues is similar to the storage areas in a manufacturing system,
virtual queues in call centers, etc. It is possible to define such parameters for queues as capacity,
shelf life of items and service discipline of storage (FIFO, LIFO, etc.) In this case, the queue object
represents the delimited queuing space at the curbside of the Terminal where three snake rows are
defined by steel bars. There are no capacity constraints because the queue often expands outside this
perimeter without any type of restriction, the passengers are assumed as patient and do not give up
and leave – so no shelf life – and the queue discipline is FIFO.
Work Exit Point
This object signals the exit of the work item from the system. It corresponds to the taxi leaving the
curbside and entering the main access road to the Terminal.
Routing is another important concept in the building of the simulation model. Routing is the
definition of the way the work item travels between objects and through which of the possible paths
within the system. In this case, the Work Entry Point is connected to the Queue object which in its turn
is connected to each of the four Work Centers and finally each of these is connected to the Work Exit
Point. The Work Centers actively search for passengers at the Queue and dispatch them to the Work
Exit Point as soon as the job is done. Routing In discipline at the Work Centers is based on “Priority”,
which consists in taking the top item of the list and collecting it – the passenger or group of
passengers that is at the front of the queue.
SIMUL8 features several possible measures of performance. The most relevant are:
For a Work Entry Point, the basic measure of performance is Number of Work Items Entered.
For a Queue object, there are two main indicators to be extracted: the Number of work items in
storage, basically equivalent to queue length (also available in graphic form, showing evolution
throughout the simulation period) and Queuing Time, either for all work items or only for those
who queued (also available in graphic form, namely a histogram). For each of these two
indicators, there are analytical results in the form of average, minimum, maximum and, for the
Queuing Time, also standard deviation.
For a Work Center, the computed measures of performance are the Number of Work Items
(currently in the center, completed, average, minimum and maximum) and the percentages of
time the Work Center was waiting for a work item, blocked, stopped or working.
For a Work Exit Point, the main indicators are the Number of Work Items Completed and Time in
the System. This last measure of performance can be represented by a histogram.
70
In general, when using SIMUL8 we can also obtain two different sets of results: the set of results at
the end of a run and the set of results of a trial. The set of results at the end of a run represents what
happened during the conducted run, under the form of several indicators - in other words these are the
relevant values that were registered over the length of the conducted single run. The set of results of a
trial demonstrates a result summary on average for a conducted trial as well as the level of variability
over the simulation experiment. In the result summary, the variability is assessed by computation of
confidence intervals. Each run is characterized with a proper set of random numbers. It is called
“Random Sampling”. Every random sampling yields a different set of results. (Días Esteban, 2008)
Figure 32 – Final System Configuration for the current situation at Portela’s Arrivals Taxi Stand
Based on the building block types, their possible attributes and on the relationships between them,
the basic system configuration was iteratively built and tested until stabilizing on a final setup. (Figure
32) In order to reach this last setup, several conceptual questions were asked about the way to better
model the arrival and service of passengers, especially the intention to model groups. Other issues,
related to the service capacity and “routing in” discipline of servers were also addressed:
Modeling Groups – this aspect, as mentioned earlier, has a relevant impact on the system,
either because groups tend to take more time in coordinating and boarding a taxi or a set of taxis
or because they also condition the number of taxis that are required. In order to model this aspect
in SIMUL8 environment, work item label called “Batch Size” was created – SIMUL8 has a
different definition for “groups” – and implemented, at the Work Entry point, a routine that
attributes a value to Batch Size based on a customizable distribution, which the author chose to
71
be the one on Figure 24. This means that each work item (group) that enters the system will have
a variable Batch Size value (group size) based on the distribution of group size taken from field
observations. After this procedure, there was need to create two more virtual (processing time is
zero) intermediate work centers in order to adequately process the groups. The first Work Center
entitled “Identify Group” creates a unique ID number for each work item (still a group, at this
stage), in order to be able to distinguish this group and every future subdivision/multiplication of it,
in terms of group members. The second work center, entitled “Disaggregate”, has the function of
disaggregating the group into a number of passengers that equals the Batch size label value,
each with the ID of the corresponding original group. It also creates another label called “Unity”
which equals 1, to serve as proxy for counting contents of servers later on. This way, after exiting
“Disaggregate”, work items will travel in batches to the queue, appearing there as several
independent work items, each with the ID of the group they originate from. This allows
determining actual queue length while still modeling the arrival of groups upstream.
Taxi capacity and group handling – For maximum taxi capacity, and although on some
occasions – when there is a lot of luggage or special circumstances – this is not the case, the
author chose four passengers per taxi. This means the servers, which actively pickup work items
from the queue, had to coherently choose the “ID type” and number of passengers that respected
the restrictions on the maximum capacity of the taxi and being part of the same group (even if it is
a “group” of 1 person). For this, some Visual Basic programming was introduced on the “Routing
In” rules and some options such as “Use Label Batching” (to choose up to maximum number of
items at a time - four) and “Batch By Type” (for choosing work items according to same label - ID)
were also selected. These restrictions and rules allowed for the correct modeling of a taxi pickup
system. Servers only pick up passengers that come from the same group, from 1 to a maximum
of 4 each time, so a group of 7 will be divided as 4 on one server and 3 on the other. Each server
does not collect passengers from other groups to combine with existing ones, even if there is
room, so if a server has 3 passengers from group with ID X it does not collect the single
passenger with group ID Y even though it theoretically still has capacity to do so.
Results on Work Entry Point and Work Exit Point (Work Complete) – Work Entry Point will
show the total number of group entities that entered the system while the Work Exit Point (Work
Complete) will show the total number of passengers that exited the system. This object can
present results which are disaggregated by label, for example. In this case, and having not
activated that option, Work Complete will be registering individual number of passengers
processed, and not groups.
Model parameters and assumptions
After building the main basic model structure, the relevant parameters for system and object
behavior and processing had to be introduced, in order to calibrate the model to resemble reality. For
this, the collected field data was statistically analyzed, namely the inter-arrival times distribution and
72
the service time distributions for the servers. The software used for distribution fitting was EasyFit. The
following conclusions and assumptions were made:
The simulation duration was set to 3600 seconds (1 hour), approximately equivalent to the 3566
second-period (59 minutes and 26 seconds) which the analysis on the collected data was based
on, basically representing a whole peak-hour of operation.
For the definition of the inter-arrival times distribution, the observations used for the analysis of
the most relevant collected data were chosen (see Section 3.2.2.2). The statistical analysis of this
dataset resulted in a good fit to the Exponential distribution (λ=0.07263) (see Annex III), based
on a dataset with an average value of 14 seconds (SIMUL8 often asks for average or standard
deviations, not distribution parameters).
For the definition of the service times, two alternative approaches were considered. One was
based on modeling Server 1, 2 and 3 based on a single data set of service times (combining all
the observations), fitted to a single distribution, equally used on all three and treat Server 4
independently. The other was to gather all four different server service time observation sets and
try to fit the aggregated results into a single theoretical distribution. The second approach was
considered as more realistic because of the low number of observations done on Server 4, and
for simplification purposes. The chosen theoretical distribution was the Lognormal distribution
(µ=4.1983; σ=0.46905), based on a dataset with average value of 74 seconds and a standard
deviation of 36 seconds (see Annex III).
Resources were not considered in this simulation. This derives from the fact that the supply of
taxis at Terminal 1’s curbside is highly abundant and constant during the operational period
during which the airport is open, especially at peak-hours, when long queuing of taxis is verified.
Having this into consideration, resource usage and availability can be considered as infinite, for
modeling purposes, and therefore, server availability is only restricted to the service time, which
considers the “empty time” as well as the actual servicing time.
Queue discipline is set to First-In First-Out (FIFO) and the queue’s routing out protocol states that
passengers walk primarily to the front Servers 1 and 2 and only then to Servers 3 and 4, by
default. This is different from the methodology used for the analytical processing of the collected
data, which considered as a routing out discipline the first available server, and may yield some
small discrepancies, mainly in terms of queue length and waiting times.
Walking distances and times were set to zero, which means that the trajectories through which
passengers walk up to the queue, coming from the arrivals area inside the Terminal, are instantly
travelled, thus not considered relevant for the problem at hand.
The model does not consider Reneging, Jockeying or Balking. Reneging is the passenger
behavior of joining a queue, waiting for some time and giving up eventually due to intolerable
delay, Jockeying is the behavior of queue switching and Balking is the discouragement of joining
the queue at all due to the perception of long queuing and/or queuing times.
Crosswalks and other realistic system elements and objects such as police, trolleys, and even
group behavior at the service area, etc. are not explicitly considered. Measured service times can
represent, to a certain extent, some of these system interactions, namely delays on service.
73
Model results and validation
Model validation is an important phase of the simulation model building process. It increases the
degree of legitimacy of the model results, conferring credibility and serving as a proof of similarity
between reality and the model. In order to test the level of approximation between simulation models
and reality, analysts must validate their model, for which there are many possible techniques and
formal mathematical processes, present in the Literature. However, the author chose a more
generalist and less-systematic view of this process, based upon the consensual and general idea that
if the model presents a similar trajectory and results to those verified in reality, it may be considered
valid. Some bibliography suggests that validation should be supported on this idea, based on the
analyst’s perspective of reality and model (Valadares Tavares, et al., 1996) while other authors
specifically approach this issue with elaborate mathematical methods. The chosen approach for this
model is the first one, in which we compare the evolution of the main indicators for queue behavior
and final average, standard deviation and maximum results for both situations. The sampling and
distribution fitting process should also contribute for a fair approximation to reality. Within this context,
the main results for the basic simulation model, described above, are also presented.
Before running the simulation model, and in any of the trials performed with this software, the Trial
Calculator function of SIMUL8 was used, to determine the optimal number of runs for the trial, based
on the required precision of the confidence limits around the estimate of the mean for the simulated
results. The main measures of performance for queues were selected to fit this criterion (average,
standard deviation and maximum queue length and waiting time). Required precision was set at 5%
of the mean, this means that the confidence limits (95%) will each be within this percentage of
estimate of the mean. The resulting recommended number of runs was 128.
Low 95% range Average Result High 95% range
Queue
Length
(persons)
In-Queue
Waiting
Time
(seconds)
Queue
Length
(persons)
In-Queue
Waiting
Time
(seconds)
Queue
Length
(persons)
In-Queue
Waiting
Time
(seconds)
Queue
Length
(persons)
In-Queue
Waiting
Time
(seconds)
Average 37 315 60 426 63 445 66 465
Standard Deviation 37 321 - 253 - 264 - 275
Maximum value 130 1211 125 873 130 907 136 940
Empirical Process Simulation Model
Figure 33 – Final result comparison between the Empirical Process and the Simulation Model
The results shown on Figure 33 refer to the common indicators than can be extracted from SIMUL8
and from the Empirical Process. There are other possible indicators that SIMUL8 can calculate, but
they are more important during the scenario building phase than for validation, so only the comparable
ones are presented. The maximum values for the queue length of both sources, on average, are equal
(130 people) and the maximum values for In-Queue Waiting Time are separated by less than 25% (5
minutes). The average values for queue length are less similar but the averages of the waiting time
are only about 30% (2 minutes) apart. Standard deviations for waiting time are also similar (difference
of 56 seconds). The discrepancies can be explained not only from all the simplifications the Empirical
74
method considers, but also from the different routing out discipline, considered for the queue (first
available server versus prioritized routing for Server 1 and 2). Also, the fact that the Empirical process
represents a specific situation (equivalent to a single model run) can also mean that queue indicators
may be assuming values that are below or above average, such as it is the case with the first 20-25
minutes of the Empirical period, where the arrivals are clearly less intensive – no queuing. The total
number of Groups that entered the SIMUL8 system was 257 on average, which is very close to the
272 that the measurements show. This means that the modeled arrival stream is similar. The total
average number of taxis that completed service in the simulation was about 190, similar to the
average registered by ANA on Easter week, in 2006, of 180 taxis/hour.
Figure 34 – Queue Evolution for the Empirical (above) and Simulation (below) Methods
Queue evolution is also important to assess if the SIMUL8 system can reasonably mimic the
witnessed behavior. From Figure 34, some similarities can be found between queue evolution from the
Empirical Process and from one random run of the system. While it is true that queuing starts at
different times in both charts, it is also true that once it starts, it has similar development. It starts by a
intermittent evolution, rapidly rising immediately after and relatively stabilizing close to the interval
between 50 and 80 people. Then it rises again and reaches 130 close to the end of the period.
Overall, and despite some expected discrepancies, main indicators and evolution seem to point to
similar results and witnessed field behavior appears to be adequately represented by the SIMUL8
Model. The arrival stream, the maximum queue lengths and waiting times and queue evolution seem
to reasonably resemble reality at the Arrivals Stand.
130
0
20
40
60
80
100
120
140
0 600 1200 1800 2400 3000 3600
Qu
eu
e L
en
gth
(N
um
be
r o
f
Pe
op
le)
Cumulative Elapsed Time (Seconds)
75
3.3. Scenario Building
After describing the system on a regulatory/institutional and operational perspective, a wider view
of the system should be considered, namely for the future. Following this reasoning, and with the final
objective of promoting and justifying intervention proposals, a number of policy actions and scenarios
were built, in order to test system behavior to the introduction of different stimulus. The complexity of
modeling certain conditions restricts the possibility of equally testing the same action on both
perspectives, so there may be differences between the actions on the regulatory/institutional and
operational views.
3.3.1. Regulatory and Institutional Policy Actions
3.3.1.1. Policy Actions Analysis
Policy Action I - Introduction of Taxi Sharing
The introduction of taxi sharing or collective taxis is an interesting topic for discussion. Although
many benefits can be obtained from this service type, such as less externalities from less moving taxis
and individual trips and greater space and time efficiency, from having taxis with higher occupation
rates, there are other factors that might reduce some of the attractiveness of this policy measure.
At Portela, for example, supply of taxis is constant and abundant, so only rarely does shortage of
taxis exist. In fact, one of the main problems taxi drivers complain about is the long waiting times in
queuing for service, due to the large numbers of taxis that park at the airport stand. This measure,
although it could allow for more transport efficiency in terms of space, energy and time, including at
the passenger queue, would require two relatively major interventions, as described below.
Firstly, at the operational level, a new, GPS/GIS-based management system would have to be
installed at the airport (and in taxis), in order to group passengers according to destination and other
relevant characteristics. For such a service to be efficient, reservations should be done in advance or
service would have to be relatively fast, in order not to create a new passenger queue, probably inside
the Terminal. Such intervention would create a second layer of service but this could be mitigated by
pre-flight reservations or a separate pick-up location, such as the Departures Stand, for example.
The second intervention would have to focus on the regulatory framework, namely the building of a
coherent and fair pricing scheme, according to destination, distance, etc and other market rules. The
collective taxi rules should provide taxi drivers with incentives to adopt this service type, when
compared to the traditional street hail/taxi stand or dispatch services. This last intervention may
become problematic, in the sense that taxi drivers will be very reluctant to allow this artificial decrease
in the demand, which will surely mean that individually, fewer taxis will be needed to handle the usual
demand, or waiting times in queue for service will significantly increase.
76
Politically, if no special provisions are adopted to compensate the taxi sector, this measure will be
strongly opposed, so it will not definitely be risk-free for politicians. Socially, people will surely take
some time to adapt and some risk of low participation might also exist, because this kind of service is
not common in Portugal and because some of the success of this mode is conditioned to certain
country-specific socio-behavioral patterns.
In sum, although politically and socially challenging and despite some possible lack of operational
interest at Portela, this service type should be analyzed for viability, especially at the NAL, because of
the economic and social benefits that it can provide. The construction of the new airport in Alcochete
can be seen as a window of opportunity to evaluate the implementation potential of this service.
Policy Action II - Introduction of a Special Airport Fleet and Concession changes
With the construction of the NAL, the regulatory context for airport taxi services is bound to change,
because the new airport stand will no longer be located in the Lisbon Municipality, but rather at Montijo
or Benavente, depending on the final design. This change in regulatory environment can be seen as
an opportunity, and raises the question on the hypothesis of creating a special fleet to serve the airport
exclusively, namely through a permit system. This contingent would possibly feature additional or
special requirements on driver professional training and vehicle characteristics. ANA becoming the
concessionaire for this service at the terminal of the NAL can also be considered as a possible action,
especially if the airport is to be viewed as an independent city-like infrastructure such as an airport-
city, increasingly decoupled from the influence of municipal and regional power.
Both of these decisions would not be trivial to implement, of course. The introduction of restrictions
on access to the airport stand would imply taking some of the current installed freedom back from taxi
companies/drivers, and this kind of political mechanism is bound to be met with charges of market
discrimination. Taxi drivers who are more reluctant in investing or simply cannot afford it would be
pushed out of a profitable and accessible market without apparent compensation. A permit system
would most likely be the contractual and regulatory basis for this system.
On the level-of-service perspective, special trained drivers, with improved communication and
driving skills, formally linked to the airport stand would not only be more “passenger-friendly”, reducing
the major complaints on poor driver behavior, but also more accountable, because the taxis that serve
the airport would be registered and identifiable. Moreover, if some investment was made on taxi
vehicles, such as GPS systems, air conditioned, newer vehicles, etc. trips would also be cleaner, safer
and more comfortable for airport passengers. But level of service is not limited to the actual trip itself, it
also has to do with the pre-taxi trip conditions, namely the queuing and waiting problem, which of
course should be improved and seen as a key part of an integrated service package. Also, these
improvements would cost money and time to the drivers and companies, who would have to gain
something from doing so – increasing trip fees could be a strong possibility. Service availability can
become a problematic issue when a permit system is implemented. The necessary number of permits
77
to ensure full service coverage throughout the day must be carefully determined, and numerical limits
should account for unexpected high solicitations such as major cultural, political or sport events, etc.
On an institutional and regulatory level, the introduction of a permit system introduces greater
participation on the part of the concessionaire, in the sense that most of the monitoring on the
compliance with requirements becomes its responsibility, and a possible bad service image is shared.
The regulator role would have to become either the Municipality of Montijo/Benavente’s responsibility
or shifted back to the IMTT. The IMTT would probably be better at regulating this specific transport
service than a relatively small municipality, because of the greater influence, resources and power it
yields. Administratively, the operation would become more complex and costly, because permit
owners are not Airport employees and thus cannot be easily discharged or penalized for violating
company rules, adding a layer of bureaucracy where none existed.
New pricing schemes would possibly have to be implemented, on account of greater investment,
greater quality of service and need for incentives to taxi drivers to compete for the permits – prices for
this service would probably have to be higher than for standard city service, although this should be
confirmed through further studies. These permits should only allow an airport taxi to solicit passengers
at the NAL Terminal, in order not to create a monopoly and discriminate remaining drivers. Another
special taxi stand could eventually be built or concessioned somewhere in the city of Lisbon, in order
to create a kind of shuttle system to avoid empty roundtrips, but every taxi should be allowed to drive
passengers to the airport, regardless of their origin. This secondary stand would have to be carefully
planned in order not to excessively tap into the rest of the taxi demand in that area, despite the higher
price of service. Many issues are raised on account of the introduction of a permit system, namely
political ones related to competition. Healthy competition for taxi services should be preserved and this
measure should be seen as a way to modernize and improve the quality of service provided by the
fleet serving at the airport and not a way to create an exclusive and inaccessible monopoly. This
implies that, much like the municipal permit system, these special permits are publicly tendered
according to relevant and objective criteria, to ensure transparency and fairness of market access.
If the new airport development model is to be based on the concept of an airport-city, then
additional independence could be promoted for airport authorities to extend and manage their services
and businesses. This opens the door for the transfer of ownership from the municipality to the Airport
Operator – ANA - who is not satisfied with current taxi service at Portela and wishes to have a greater
influence on the definition and planning of this important curbside service.
Policy Action III – Market segmentation and other changes to the Departures Taxi Stand
The Departures Stand can be seen as an extension of the Arrivals Stand, at the service of the
small share of Portela’s passengers who know and are familiar with it. As mentioned earlier, this
stand is currently a topic of discussion among some taxi drivers and companies, on account of internal
competition issues. Also, as time goes by, more passengers become acquainted with this faster taxi
service spot, creating the risk of some of the queuing from the Arrivals transferring to that Stand.
78
Directing service at the Departures Stand to target certain airport taxi market segments is an
interesting option to explore. This would imply market segmentation, in the sense that different kinds
of service could be offered at the two Taxi Stands. Shared Taxi has already been mentioned as a
possibility, but the consideration of high-quality service types should also be equally interesting and
relevant. This taxi stand is known to traditionally serve airport employees and frequent Lisbon Airport
passengers, such as local businessmen, for example. It is a much smaller taxi stand than the Arrivals
and is naturally perceived as offering faster service, because of small or inexistent queuing
phenomenon. Such characteristics, coupled with the fact that the Stand already exists – so no
jurisdictional or spatial problems – provide an adequate context for the introduction of new and special
service types. Lisbon regulations on taxi services also allow for the creation of different service types,
with different pricing schemes. This would allow pre-arranged reservations or on-site agreements
between clients and operators. Such a system would probably require an investment on a small
management structure, probably located on in-terminal facilities, to process or register reservations for
deplaning passengers.
Implementing this measure would of course be politically challenging as other drivers do not like to
see colleagues earning more money and waiting less for service than they do. The taxi sector can
probably react by arguing that there cannot be first class taxi drivers and second class taxi drivers and
some earning more than others. Competition for parking spots at the Departures Stand should
significantly intensify, increasing the risk of becoming aggressive and tensions can rise among drivers.
Taxi drivers, associations and companies will also probably oppose restrictions on access to a pre-
existent largely free-access, market, fearing for their own market share. This opposition should be
especially intense if special service or vehicle requirements are introduced.
But the access to a richer market niche may not come without investment in vehicle conditions and
driver’s professional skills, all of which also cost money and time. This type of measure, if introduced,
should also be seen as an incentive to innovation, professionalization and modernization, not a
discriminatory measure.
Very high value-of-time passengers, such as businessmen could thus have an alternative service
type from which to choose, tailored for their market segment, for which they would surely be willing to
pay more than for a regular taxi service. They would be basically paying for less queuing, less waiting
times, more comfort and personalized service - all the characteristics that a higher-class passenger
searches for in a transportation service.
3.3.1.2. Policy Actions Evaluation
There are many ways to analyze policy actions or alternatives in terms of their main characteristics.
The following analysis does not intend to be an exhaustive study on every risk and impact of the
measures contained in each of these actions. The intention is not to choose one action as the best or
to score them according to some value scale, but rather to assess the pros and cons of their existence
in a possible future. The aim is identifying the main characteristics and potentials of each of these
79
actions, in order to be aware of the differences between them and the consequences that each of
them enclose. SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis is a useful tool for
diagrammatically representing these characteristics (Figure 35; Figure 36 and Figure 37).
Str
en
gth
s
Increased Spatial and Energy Efficiency
Less Congestion Externalities
Less Environmental Externalities
Decrease in passenger queuing Weakn
esses
Lack of experience and knowledge of
Portuguese passengers
Taxi Drivers resistance to reduction of
Demand for individual taxis
Investment and operational costs of the
GPS/GIS-based management system.
Op
po
rtu
nit
ies
Introduction of the Collective Taxi on a
Competitive environment with intense demand for
taxi services
Existence of the Departures Stand for possible
implementation of the service
Th
reats
Lack of passenger interest due to social
habits
Lack of operational interest for
transportation operators
Strikes and boycotts from taxi drivers
and associations
Figure 35 – SWOT analysis for Policy Action I – Introduction of Taxi Sharing
Str
en
gth
s
Safer, cleaner and more comfortable taxis
Friendly, knowledgeable and professionally
trained drivers
Increase in Airport Operator’s responsibility and
intervention power
Weakn
esses
Significant decrease in competition levels
Increased administrative costs in
monitoring and regulatory enforcement
Possible increase of taxi fares
Op
po
rtu
nit
ies Increase in service quality while ensuring more
driver accountability
Investment in a new and modern taxi service
fleet and driver skills
Integration of Taxi Services into the Airport’s
concessions
Increased independence and strengthening of
the airport-city development model
Th
reats
Political issues related to the introduction
of service access restrictions where none
existed
Service availability issues, if correct
number of permits is not adequately
determined
Protests from passengers due to higher
taxi fares
Figure 36 - SWOT analysis for Policy Action II – Introduction of a Special Airport Fleet and Concession changes
Str
en
gth
s
Creation of a new service type, increasing
diversity and options for airport passengers
Take advantage of the high willingness to pay of
high value-of-time passengers to segment the
market
Weakn
esses
Impact on regular Departures Stand
users, who might be forced to join the
longer queues at the Arrivals Stand.
Creation of a second layer of service,
inside the Terminal, increasing
administrative costs.
Op
po
rtu
nit
ies The Departures Stand already exists and is,
from a regulatory perspective, an authorized
area for taxi services
Incentive taxi drivers to modernize their vehicles
and improve their professional skills by possibly
introducing special service or vehicle
requirements
Th
reats
Political issues related to the perception
of driver and passenger discrimination.
Increase in taxi driver tensions and
possibility of aggressive competition, due
to higher profitability of the Departures
Stand.
Figure 37 - SWOT analysis for Policy Action III - Market segmentation and other changes to the Departures Taxi Stand
80
3.3.2. Operational Scenarios
3.3.2.1. Scenario Analysis
Scenario I – Increase the number of service lanes/rows and servers
One of the changes that could be promoted at Portela’s Arrivals Stand would be to enlarge the
inner-curbside area to allow for three lanes of taxis to serve simultaneously at peak hours, instead of
the current two. Physically this would imply the creation of a special pavement structure that would
uneven the road at the service area and crosswalks to ensure passenger safety in crossing the taxi
lanes to the farthest servers. This extra lane could become an important flexibility option, in the sense
that, during off-peak periods, it could function as a free maneuvering lane for easier bypassing and
during peak-hours it could become an extra service lane. Police would need to be more active in their
role to ensure passenger safety and taxi coordination, which would probably imply the increase on the
number of police agents. The modeling of this situation consisted in adding two extra servers to the
service area, with the assumption that they follow service time distributions that are similar to those of
the remaining servers (Figure 38).
Note: any of the following results on queue waiting times are in seconds.
Figure 38 – System configuration for Scenario I – Extra Service Lane, 2 extra servers
81
Low 95% range Average Result High 95% range
Work Entry Point Number Entered 255,98 257,05 258,13
Work Complete Number Completed 485,44 487,77 490,09
Queue for Taxi Stand Arrivals
Average Queue Size 5,77 6,08 6,38
Maximum Queue Size 27,3 28,04 28,77
Items Entered 503,16 505,57 507,99
Average Queuing Time 40,16 42,15 44,14
St. Dev. Of Queuing Time 41,85 43,32 44,8
Maximum Queuing Time 159,08 163,61 168,13
Taxi Stand Arrivals 1 Working % 90,58 90,84 91,1
Waiting % 8,9 9,16 9,42
Taxi Stand Arrivals 2 Working % 90,07 90,34 90,61
Waiting % 9,39 9,66 9,93
Taxi Stand Arrivals 3 Working % 80,03 80,54 81,05
Waiting % 18,95 19,46 19,97
Taxi Stand Arrivals 4 Working % 79,84 80,36 80,88
Waiting % 19,12 19,64 20,16
Taxi Stand Arrivals 5 Working % 89,1 89,38 89,65
Waiting % 10,35 10,62 10,9
Taxi Stand Arrivals 6 Working % 87,84 88,12 88,4
Waiting % 11,6 11,88 12,16
Figure 39 – Main SIMUL8 results for the Scenario I system configuration
From these results (Figure 39) we can conclude that a major service improvement occurs when we
introduce 2 extra servers into the system. Maximum queue length drops from 130 people to 28, on
average, while maximum waiting time also drops from approximately 15 minutes to about 3 minutes.
Average waiting time would be less than a minute, compared to the almost 7 minutes and a half of the
base situation. Average queue length would also drop from about 63 people to 6 people. One
interesting fact is that all servers occupy a very high percentage of the time working, which means the
system is working at close to full capacity, almost as if demand is meeting supply on very similar
proportions, with a good level of synchronization but risking increases in queuing.
In this scenario, a total of about 250 taxis would complete service during the simulated hour.
The introduction of a third lane would definitely solve the queuing problem at peak-hours, as
service availability and system capacity basically increase by 33%. This measure would however be
physically challenging to implement, as the inner-curbside space is limited and scarce and increasing
it would force a significant investment on civil works and traffic re-routing. By expanding this taxi
service area, the airport would also be reducing the outer-curbside space, located in front of the bus
stops, which is reserved for buses and private cars to pick up passengers. Safety issues should also
be considered, in terms of passengers crossing the service area to the farthest servers.
This change could also be interpreted as the simple adding of an extra server to the two existing
lanes (loading of 3 taxis at a time in each lane), but this would possibly increase the issues with the
bypassing and maneuvering of vehicles, and also change service times. Back row taxis that take
longer to exit would now be blocking two parking positions instead of one, more maneuvering and
service priority conflicts would emerge, less line of sight for taxis and people, more distance to cover
for passengers to get to the back servers, etc.
82
Scenario II – Different service area configuration (single lane, multiple servers)
One scenario that is worth exploring is changing the configuration of the service area. This change
would aim at mitigating queuing based on the reduction of service times, - instead of increasing
capacity - in order to have a faster and more fluid service stream. This reduction - assuming the
impossibility of increasing the speed at which passengers move and board the taxi or the speed at
which the taxi driver loads the luggage on the trunk - would have to focus on the time lost during
maneuvering, bypassing or being blocked by front/back servers. Having this into consideration, a
system where taxis would not be forced to wait for the front servers to empty would be the obvious
choice. The tested configuration (Figure 40) features an inner service lane, loading several taxis at a
time, and a free lane, allowing taxis to safely and quickly bypass front colleagues, similar to other
systems, such as the one implemented at La Guardia Airport, in New York. Such a system at Portela
would require a relevant increase in inner-curbside space, especially the queuing area, which would
have to be reorganized – the exit should be transversal to the current one, effectively exiting the
queue in the corridor’s direction. A larger (also delimited) “buffer” zone would have to exist near the
queue exit extending for a significant portion of the curbside between the crosswalks, in order to allow
room for luggage/passengers to board the taxis. This could lead to the reduction from 4 servers to 3,
depending on the spatial disposition and design of the new system, therefore both situations were
tested.
Figure 40 - System configuration for Scenario II – One service lane, multiple servers
83
3 Servers 4 Servers
Low 95%
range Average Result
High 95% range
Low 95% range
Average Result
High 95% range
Work Entry Point
Number Entered 254,71 257,45 260,18 255,89 257,02 258,16
Work Complete Number Completed 382,94 386,51 390,09 481,16 483,36 485,56
Queue for Taxi Stand Arrivals
Average Queue Size 53,89 56,72 59,54 10,13 10,66 11,19
Maximum Queue Size 113,65 118,89 124,13 34,26 35,28 36,3
Items Entered 498,33 504,49 510,65 503,09 505,63 508,18
Average Queuing Time 381,95 399,6 417,25 70,51 73,97 77,43
St. Dev. Of Queuing Time
226,18 235,91 245,63 55,69 57,68 59,66
Maximum Queuing Time 783,35 813,42 843,5 206,73 213,11 219,49
Taxi Stand Arrivals 1
Working % 98,84 98,99 99,15 94,04 94,28 94,52
Waiting % 0,85 1,01 1,16 5,48 5,72 5,96
Taxi Stand Arrivals 2
Working % 98,54 98,71 98,87 93,77 94,01 94,25
Waiting % 1,13 1,29 1,46 5,75 5,99 6,23
Taxi Stand Arrivals 3
Working % 97,81 98,06 98,31 89,69 90,09 90,5
Waiting % 1,69 1,94 2,19 9,5 9,91 10,31
Taxi Stand Arrivals 4
Working % - - - 89,22 89,63 90,04
Waiting % - - - 9,96 10,37 10,78
Figure 41 – Main SIMUL8 results for the Scenario II system configurations
In face of a different kind of system dynamic at the service area, there was need to model service
times differently. The service times that this kind of configuration would generate are mostly linked to
the period that comprises of the loading of luggage and boarding of passengers, and the “empty”
times are less relevant. This happens because the parking, maneuvering, bypassing and blocking of
taxi services are much less frequent due to the existence of the free lane. For the modeling of this
situation, the author used a sample of this type of service time, taken on the 20th of August, from 9 to
10 p.m. from Servers 1 and 2, which was later dismissed because it did not account for the “empty
time” parcel, which was considered necessary for the modeling of the current service. These service
times were also fitted to the Lognormal distribution (µ=3.8543; σ=0.48449) (see Annex III), from a
sample with an Average of about 53 seconds and Standard Deviation of 29,5 seconds.
The SIMUL8 model produced some interesting results, as can be seen on Figure 41. For the 3
server situation, the average queue length drops to 57 people, while the maximum value also
decreases to 119 people. The average waiting time is about 6 minutes and the maximum waiting time
is 13 minutes. This is a slight improvement regarding the 4 server disposition of the base case, which
also raises the doubt on the degree of efficiency of the current configuration when compared to other
similar ones. Regarding the 4 server scenario, there are clear improvements on average queuing
length and time (11 people and about 1 minute) and maximum values (35 people and about 4
minutes). This is clearly one of the best solutions for the service configuration, in terms of simulation. It
would also probably be less cost intensive in terms of civil works and space than the 6 server version
(2 service lanes); safer for passengers - who would not need to cross the service area - and faster for
taxis, which would be able to freely bypass their colleagues on completion of the boarding phase.
In this scenario, a total of about 200 taxis would complete service during the simulated hour, for the
3 server situation and 247 for the 4 server situation.
84
Scenario III – Multiple queues and service segmentation
The inner-curbside at Portela, although not very wide in terms of dimensions, is relatively long,
which allows for the possibility of creating a secondary and smaller queue near the exit of the inner-
curbside road, on the far left end of the Terminal’s façade. This area is far enough from the main
passenger taxi queue to allow for the taxis that serve there to easily bypass any possible one-lane
queuing that might form here. Of course this queue would have to be limited to one service lane, or
else it could block the exit of taxis from the normal queue.
It would be interesting to take this opportunity and maybe promote a different service type, namely
for special class of service, pre-booked or simply faster, more expensive but also with less queuing
and waiting times. This would relieve some of the Demand from the main queue, namely the high-
income and high-value-of-time passengers such as businessmen, for example, helping reduce the
heavy queuing there. At the same time, it would diversify the service offer at the Terminal. On a
regulatory note this service would have to be subject to some degree of access control. One way to
ensure equity in access would be to either restrict service to a certain parking capacity, like at the
Arrivals Stand (first to arrive to the spot gets the right to queue, if there is room available) or to
randomly raffle from the Arrivals Stand’s parked taxis (according to an entry ID or similar) a certain
number to sequentially serve at that stand. Physical access to the stand could either be through a new
built-in access lane or through the normal segregated lane, if the taxi service was designed similarly to
Scenario II configuration.
Figure 42 - System configuration for Scenario III – 2 queues, Special Service Type
85
For the service time distribution at the Special Queue, the author used the previously mentioned
values for service time on Scenario II that exclude “empty” time, since taxis are free to maneuver and
bypass other colleagues. The focus is once again on how much time it takes to load a taxi. Based on a
2000 survey, 28% of the Portela passengers are visiting on business motives, so the passenger types
were divided into type 1 (business class) and type 2 (normal), through use of a label called “passenger
class”, which assumes value 1, 28% of the times. The routing out of the Disaggregate work center is
based on the value of this label. If this value is 1, the item will go to the special queue, and if it is 2, it
will go to the normal queue. System configuration is represented on Figure 42.
Low 95% range Average Result High 95% range
Work Entry Point Number Entered 255,93 257,02 258,11
Work Complete Number Completed 480,34 482,47 484,6
Special Queue
Average Queue Size 0,52 0,54 0,57
Maximum Queue Size 8,36 8,56 8,76
Items Entered 139,69 140,93 142,17
Average Queuing Time 12,86 13,4 13,94
St. Dev. Of Queuing Time 22,32 23,03 23,74
Maximum Queuing Time 94,71 97,37 100,02
Queue for Taxi
Stand Arrivals
Average Queue Size 8,92 9,39 9,86
Maximum Queue Size 29,96 30,87 31,79
Items Entered 362,62 364,73 366,83
Average Queuing Time 85,53 89,68 93,84
St. Dev. Of Queuing Time 67,85 70,26 72,68
Maximum Queuing Time 247,43 255,01 262,59
Taxi Stand Arrivals 1 Working % 93,31 93,58 93,85
Waiting % 6,15 6,42 6,69
Taxi Stand Arrivals 2 Working % 92,81 93,08 93,85
Waiting % 6,65 6,92 7,19
Taxi Stand Arrivals 3 Working % 88,35 88,8 89,25
Waiting % 10,75 11,2 11,65
Taxi Stand Arrivals 4 Working % 87,76 88,21 88,67
Waiting % 11,33 11,79 12,24
Taxi Stand Arrivals 5 Working % 53,64 54,1 54,56
Waiting % 45,44 45,9 46,36
Taxi Stand Arrivals 6 Working % 53,1 53,58 54,06
Waiting % 45,94 46,42 46,9
Figure 43 - Main SIMUL8 results for the Scenario III system configuration
The results of the simulation model (Figure 43) show a significant decrease in values for average
and maximum queuing length and waiting times at the normal queue. On average, about 10 people
would be waiting for 1,5 minutes and the maximum solicitation would occur with the queuing of 31
people who would wait for about 4,5 minutes. The servers at the normal queue show high working
percentages, which show that demand and supply rates are also increasingly similar. As for the
special queue, the average queue is almost zero and the maximum queue would be about 9 people,
waiting for about 2 minutes, consistent with a higher-quality service. The two servers introduced at this
stand would be working for about 50% of the time. Items entered on the special queue account for
about 27,9% of the total, which coincides with the defined percentage of business-related passengers.
In this scenario, a total of about 247 taxis would complete service during the simulated hour.
86
3.3.2.2. Scenario Evaluation
Unlike with the regulatory policy actions, the performance of different operational scenarios is more
objective and comparable. Besides presenting and summarizing the main results, there is an obvious
interest in assessing the impacts of the different operational changes in the queuing at Portela. The
different scenarios generated the following results in terms of the main queue performance indicators
(Figure 44 and Figure 45):
Figure 44 – Results for the main Queue Size indicators
Figure 45 – Results for the main Queuing Time indicators
Scenario I, which corresponds to the increase in number of lanes and servers, is clearly the one
that produces the best improvements, with a huge decrease in all the indicators. Adopting this
measure would mean passenger queuing problems at Portela would cease to exist. However, this is
clearly the most capital intensive decision, and the limited space availability can also jeopardize the
viability of this possible intervention. Safety problems with the interaction of passengers and a larger
and wider service area are also somewhat difficult to solve.
6
57
11 9
63
0,5
28
119
35 31
130
9
0
20
40
60
80
100
120
140
Scenario I Scenario II.A Scenario II.B Scenario III Base Scenario Scenario III -
Special Queue
Qu
eu
e L
en
gth
(N
um
be
r o
f p
eo
ple
)
Average Queue Size Maximum Queue Size
42
400
74 90
444
13
164
813
213255
905
97
0
100
200
300
400
500
600
700
800
900
1000
Scenario I Scenario II.A Scenario II.B Scenario III Base Scenario Scenario III -
Special Queue
Tim
e (
Se
con
ds)
Average Queuing Time Maximum Queuing Time
87
Scenario II.A - one service lane, 3 servers – would marginally improve service speeds at the
Arrivals Stand, contributing for a slight but insufficient decrease in queuing lengths and times. It is
clearly the less expensive and difficult operational scheme because it would imply minor changes to
the service area, although forcing a substantial reorganization of the queuing area.
If queuing is to be clearly reduced with recourse to reorganization of the service area, it should go a
step further and perhaps consider room for 4 servers loading simultaneously – Scenario II.B. This
option would have to be studied for physical feasibility, because space is scarce at the curbside of
Terminal 1, but, it found viable, should definitely be considered as one of the best options, according
to this simulation analysis. It achieves reductions similar to the ones gained from introducing extra
lanes and servers, while probably costing a lot less to implement.
Regarding Scenario III, results show a very interesting reduction, similar to the one of Scenario I
and II.B on every queue-related indicator. The introduction of a secondary queue for “business” (or
special) class passengers is an interesting way of dividing taxi demand and creating a new set of
servers, while also taking advantage of the willingness to pay for differentiated services. The queue
results show almost zero average queuing and a maximum of 9 people, who would wait for 13
seconds on average and about 1,5 minutes maximum. However, this option is not without downsides
and difficulties. On the operational view, this secondary stand should require a second road access
that crosses the curbside from the main road to the inner lanes, further ahead of the normal queue.
This may be physically challenging and constitutes a possible new conflict with pedestrians, buses,
private cars and other taxis. Choosing to serve this queue through the existing segregated lane would
probably lead to delays and confusion both upstream and downstream.
On a regulatory perspective, this new stand would have to be framed within a new context, such as
the service by contract. Pre-booked arrangements or reservations can be used to minimize passenger
processing times and establish a new service type that is compatible with the current ones allowed by
law. Issues regarding the access of taxi drivers to this stand can also emerge, increasing the
complexity of this scenario.
88
Chapter 4 – Conclusions and Proposals
4.1. Main Conclusions
The first main conclusion of this study is that the current taxi service system at Terminal 1 is not
able to adequately cope with peak-hour solicitations and offer good quality of service to passengers at
these times. Observations have confirmed rapidly growing queues of deplaning passengers that often
expand beyond the delimited space for queuing, making people excessively wait for a supposedly
faster transportation service. There are many “small” factors that seem to be increasing service times,
the most relevant of which is probably the two-by-two server disposition, resulting from the splitting of
the single queue into two service lanes, possibly increasing service times for all servers, especially for
Server 4’s case (33% more than the remaining server’s average). The taxi service system at Portela,
namely the articulation between the queuing and service areas could be significantly improved, even
with relatively simple interventions, such as the ones suggested on the Scenario Analysis section.
The way of analyzing and designing the operation of the airport taxi service cannot be based on the
limited observation of the supply and demand quantities per hour or on average. Queues are a
fundamental part of the problem and their actual behavior must not be diluted in aggregate numerical
counts that do not expose the frailties of the system at peak-hours, which is exactly when passengers
experience the worse service quality. Regardless of other exogenous factors, a large supply of taxis
and large supply of passengers or a low average queue length throughout the day does not mean a
balanced and reliable operation exists, or that level of service is good – as can be seen at Portela. The
interface between passengers and taxis is a bottleneck, which can be narrower or allow greater flows
of service, according to the design of the actual system. The mismatch between supply and demand is
directly proportional to the lack of flexibility of the system to handle peak-hour conditions. The entities
responsible for planning and designing the airport taxi stands must perform a detailed, peak-hour
focused analysis on queue behavior; otherwise they study the system on the wrong average-based
scale, completely missing the main issue, embedded on the system’s peak-period behavior.
But queues are not only a very important part of the problem; they may be the key part of the
solution. As shown in this study, physical rearrangement of queues can lead to greatly improved
service as regards queue length, waiting times and reliability. In fact, many of these changes can be
promoted without need to alter taxi regulations or queue discipline rules, just by simply reorganizing
the space reserved for queuing or service. Many of the scenarios tested in this thesis prove that small
changes to service or queuing mechanisms can significantly decrease queue length and waiting times
for passengers at peak-periods. A proactive attitude from the involved agents regarding this issue,
promoted by ANA and followed through by the Lisbon Municipality with the collaboration of the taxi
companies could drastically improve performance at Portela. This small investment could increase the
number of taxi trips per day, lower the waiting times for taxis and passengers and reduce complaints.
89
As this study points out, several secondary operational factors, sometimes exogenous to the queue
itself, can also greatly influence the performance of the system, as can be seen at Portela, and
possibly in all airports, in general. These apparently small issues can lead to drastic losses of
efficiency in service and are often overlooked by analysts during the transition from field observations
to the model building process, ignored among simplifications and assumptions on system behavior.
Police presence and coordination at the taxi stand, crosswalks, distance and line of sight issues,
bypassing maneuvers, secondary queues, driver conflicts, etc. can significantly increase service times
for the whole system, consequently increasing delays for passengers and taxis.
On a more regulatory and institutional perspective, one can effectively conclude that each case is a
case, with regards to airport taxi stands. There are many factors that can justify substantial
differences, such as regulatory framework, taxi market structure, institutional power-sharing network,
airport accessibilities, size, influence and location, Terminal spatial constraints, queue/service area
design and planning and even socio-cultural factors, etc. This leads to the conclusion that there is no
general optimal solution for taxi systems at airports, and each should be analyzed in detail in order to
evaluate current/expected quality of service and possible alternative system structures. Another
important issue is that, in order to effectively analyze such a system and propose alternative designs
based on efficiency indicators, one must consider not only the quantitative measures of performance,
but also the overarching network of power and inter-dependence of the involved stakeholders.
Proposing operational changes without considering the possible political and regulatory impacts is
ignoring most of the risks and requirements of a fundamental phase of every operations research
project – Implementation (Odoni, et al., 1981).
The airport taxi is also many times subject to different operational, regulatory and competitive
conditions from taxis of other service points, mainly because it serves a very specific type of
passenger – the airport passenger. The airport passenger flows are usually composed of a significant
share of foreign and/or medium-high income, high value-of-time, luggage-carrying travelers, which are
going to cover a significant distance – assuming that many airports are located at some distance from
the city centers, unlike some other mode’s terminal stations. These characteristics are appealing to
most taxi drivers and companies in the sense that a considerable share of their profit is based on
distance, followed by luggage and tips. In most airports where the distance to the city center is big,
taxis can achieve lower operational costs while profiting on the distance/time-dependent fare.
However, airport taxi passengers are also more sensitive to quality of service, either through
availability of taxis or reliability and comfort of the trip – taking a taxi after a long journey by plane,
usually carrying luggage, must be as comfortable and direct as possible. These characteristics not
only transform the airport taxi stand into a profitable taxi hotspot, but should also lead to the increase
of the level of service requirements at this location. This last part, unfortunately, is not necessarily true.
4.1.1. Regulations and Institutional Framework
Regarding airport taxi services, there is still a wide variety of opinions on how the markets should
behave and which restrictions should apply. Although discussion is still very much focused on the
90
general tendency of the market design – regulation versus deregulation – there is some evidence that
these options are best discussed on a case-by-case basis. Experiences with regulation and
deregulation in the U.S. (Schaller, 2007) seem to show that the same formula does not yield the same
results in different airports and regions and that a spectrum of policies, rather than a single path
choice, can prove more useful. This is coupled with the variety of different contextual conditions the
airport is subject to, which may not favor the theoretical optimum, in the sense that it may happen that
not all of the assumed system and market behaviors are verified.
The institutional framework is also an important issue. Airport Authorities, regulators, intervening
government agencies, the taxi sector, passengers, competition and other direct or indirect
stakeholders such as hotels, businesses or commerce form a complex network of influence, interests
and bargaining power. Sometimes the roles and jurisdiction of each of the involved actors may not be
clearly defined or institutional power, responsibility and independence may not be adequately
distributed. This can lead to conflicts, redundancies, liability mazes and inefficiencies, which slow
down, block or tamper with the duties and intervention potential of key agents, such as the Regulator
(regulatory capture). This might result in decreases in service quality or endanger efficiency, equity
and/or sustainability of the system, undermining the operational performance as well, regardless of its
physical design.
Regarding the case-study, Portela – an open-like system, with main access restrictions based on
the parking capacity of the stand - the analysis on the regulatory and institutional framework resulted
on the following main conclusions:
The system has promoted sufficient demand for the airport and simultaneously ensured relatively
balanced taxi services in the city. The fact that demand is relatively stable at Portela - being the
main airport in Portugal and Lisbon, and slowly growing in annual traffic flows – also allows for
this system to keep providing reasonably steady results in terms of taxi service request.
The restriction to the parking capacity is an important measure to avoid oversupply at certain
hotspot stands – such as the Airport - and the open nature of the access to the airport stand also
serves to politically balance the taxi sector, creating competition and equal access rights.
Although prices are set for the whole municipality and taxi drivers and companies have little
incentive to innovate or improve on regular service, there is room for alternative service types and
exploitation of different market segments, possibly under different contractual and pricing
arrangements. Taxi sharing introduction does not seem to be a politically (possibly also
operationally, at least at Portela) attractive option.
Claims on unfair competition and efficiency decreases deriving from the existence of a
Departures Taxi Stand do not seem reasonable or justified. Equally free-access conditions,
queuing and service system behaviors on both stands and possible lower profit effects of
traditional Departures Stand passengers can help demystify this claim.
Innovation, greater service monitoring and improvements to service quality are needed at the
airport, in light of the complaints and consensually poor image the taxi sector has with tourists
and even Lisbon citizens. A permit system which filters airport stand access to vehicles and
91
drivers with certain minimum requirements and rewards investment in new vehicles, equipments
or professional skills could be an interesting option, namely at the NAL.
Sustainability of the service is threatened by the certain future introduction of extra land transport
competition for airport passengers, namely the Metro, due to begin operation in December 2010,
although significant delays for this kind of project are frequent, around 2 or 3 years. This is
somewhat offset by the fact that a new airport is going to be built, with the inauguration date
foreseen for 2017, which means Portela will receive less passengers (if any at all) and the taxi
companies will transfer their business focus to the new airport, avoiding this competition.
ANA’s role in the planning and management process of the taxi stands at Terminal 1’s curbside is
limited and somewhat unclear. This may cause significant nuisance to the Airport Operator in
case taxi service conditions deteriorate or continue to indirectly produce customer complaints,
directed at ANA itself. The regulator, Lisbon Municipality, although clearly defined as such, is
seen as a highly bureaucratic institution, with a size and scope of responsibilities that may not be
compatible with the needs of this specific situation.
The construction of the NAL will bring new regulatory and political issues, since the Lisbon
Municipality should, theoretically, cease to become the regulator/owner of the Airport taxi stand.
These issues can force the transport authorities to change the responsibility of the regulator role
to the Montijo/Benavente Municipality or shift it upstream to the IMTT, eventually awarding the
concession of airport taxi stands to ANA. This issue will possibly create a new institutional design
and new types of relationship links between the involved agents.
In Portela’s case, the Police forces – one of the system agents - are also in charge of
coordinating taxis towards their service spots, at the Arrival stands. They persuade taxi drivers to
avoid conflicts and respect the first-in first-out regime, integrated in every taxi stand in Lisbon.
4.1.2. Operational Framework
The first general conclusion that needs to be underlined in this context is that any analysis based
on averages taken from hourly or daily counts of people and taxis, such as the ones done by ANA in
2006, are not adequate for the modeling of taxi queuing systems. These average values are opaque
to the extreme behaviors that occur at peak hours, during which queues rapidly form and the system is
flooded with arriving passengers, significantly diminishing during the following hours. At these times
queue lengths and waiting times can increase almost exponentially, causing several problems in terms
of curbside space and passenger discomfort. Most of the general Queuing Theory focuses on the
stationary analytical methodologies for calculating relevant indicators and measures of performance.
The term Stationary is used when the system oscillates around an average situation, with the
distribution of the queue length being independent of time and the arrival rate not exceeding the
service rate. Like in many transportation problems, these conditions frequently do not exist in this
situation, as queues for taxis at airports are frequently in transient state, with time-dependent (peak-
hours) arrival rates. This means steady-state queuing theory is not appropriate to model this kind of
system. This problem is at the core of this study, and a different approach, based on the observation
of peak-hour behaviors and recourse to a computer-based simulation model, has been chosen.
92
But the analysis of the operational performance of a taxi service system at the curbside of an
Airport Terminal must be seen as more than just building a simulation model. There are numerous
aspects of the real functioning of a system that cannot be modeled and are frequently overlooked by
analysts, in their quest for simplification and standardized methods for obtaining fast results. Factors
such as secondary queue formation, police influence, line of sight issues and disturbances upstream
of the system, etc. can undermine many of the assumptions one can assume during modeling.
Portela’s system proved, more than once (for example, in the service times’ case), that it required
careful and persistent in situ observations in order to be fully understood and adequately modeled.
The data collection procedures are a key step of the analysis, in the sense that if the source of
information is not reliable, any model we build will produce misleading results, no matter how complex
or rigorous it is. Collecting data is not easy and this became clear during this study. Field observations
are often prone to surprises, interruptions and other unexpected difficulties, and the time and effort
required to observe and register a set of data is also very important for those measuring. In order to
minimize lost time, prevent difficulties and ensure the relevance of the collected data, the elaboration
of a Data Collection Plan was essential, even if developed iteratively. It decisively helped to clearly
define when, where, how and what is to be measured, using which resources and techniques.
The analysis of the operational framework at Portela leads to the following conclusions:
There are some conflicting points between the taxi queue and other system elements. Firstly, the
segregated access lane crosses two entry and exit roads of the parking facility located next to
Terminal 1’s Arrivals. Secondly, two of the Terminal’s crosswalks are frequently conflicting with
taxi maneuvering as flows of people coming from the Terminal have to cross the road to go to the
bus stops or to be picked up by private vehicles. Both of these situations can cause some delay
on service as taxis are sometimes temporarily blocked from entering the service area.
Observations at peak-hours have confirmed occasional formation of secondary queues,
originating from the two closest terminal exits, which occupy the small space between the
terminal and the queue, and several queues going beyond the defined space for the passenger
waiting area. These fast-growing queues are often in conflict with three of the terminal entrances,
expanding beyond the bars defining the waiting space, obstructing them, along with most of the
inner curbside area of the terminal. This phenomenon is recurrent at peak times and may cause
problems of pedestrian congestion, conflicts between passengers from different queues and even
safety and security problems, related to emergencies and evacuation procedures.
Supply of service does not seem to represent a restriction, as the taxi parking facility rarely
empties during airport working hours, and taxis keep coming in to join the queue. Despite great
number of taxis, supply is intermittent closer to the passenger queue, at the service area itself,
namely due to service area characteristics. This service area is configured so that it is possible, in
optimal conditions, the simultaneous loading of a maximum of four taxis at a time.
Analysis on the collected data shows some interesting results. The majority of the group inter-
arrival times are below 10 seconds and about 70% of all inter-arrival times are below 20 seconds.
Also, the majority of the people soliciting a taxi were composed either by a single person (36%) or
93
a group of two people (41%), usually couples. Average Queue Length and Waiting Time prove
that queuing at Portela can become very problematic. If the arrival rate of passengers is as
aggressively high as it was during the considered peak-hour period and service rate is relatively
constant (as it seems to be the case throughout the entire experiment period), an average of 37
people will be queuing for about 5 minutes and a maximum of 130 people will be queuing by the
end of the hour, waiting for about 20 minutes for a taxi. Average Service time for all servers is
little over 1 minute. However, Server 4 shows different behavior, in the sense that its average
values for service times are 23% higher than the average of the remaining servers and about
33% higher than the average for Server 2. This may occur due to a conjunction of factors that
involve bypassing maneuvers, line of sight issues, police coordination and pedestrian conflicts.
Scenario I, which corresponds to the increase in number of lanes and servers, is clearly the one
that produces the best improvements, with a huge decrease in all the queuing indicators.
Adopting this measure would mean passenger queuing problems at Portela would cease to exist.
However, this is clearly the most capital intensive decision, and the limited space availability can
also jeopardize the viability of this possible intervention. Safety problems with the interaction of
passengers and a larger and wider service area are also somewhat difficult to solve.
Scenario II.A - one service lane, 3 servers – would marginally improve service speeds at the
Arrivals Stand, contributing for a slight but insufficient decrease in queuing lengths and times. It is
clearly the less expensive and difficult operational scheme because it would imply minor changes
to the service area, although forcing a substantial reorganization of the queuing area.
The 4 server option of Scenario II.B would have to be studied for physical feasibility, because
space is scarce at the curbside of Terminal 1, but, it found viable, should definitely be considered
as one of the best options, according to the simulation. It achieves reductions similar to the ones
gained from introducing extra lanes and servers, while probably costing a lot less to implement.
Regarding Scenario III, results show a very interesting reduction on every queue-related
indicator. The introduction of a secondary queue for business-class passengers is an interesting
way of dividing taxi demand and creating a new set of servers, while also taking advantage of the
willingness to pay for differentiated services. The queue results show almost zero average
queuing and a maximum of 9 people, who would wait for 13 seconds on average and about 1,5
minutes maximum. However, this option is not without downsides and difficulties, namely a
complex physical implementation and political issues related to fairness of access.
In sum, the taxi service system at Portela is not providing a good level of service to its
passengers, who often wait for excessive periods of time, having to join long queues, at peak
hours. The service area disposition, coupled with many other small operational issues is clearly
contributing to the lack of system’s capacity to cope with peak-hour solicitations. The service area
disposition is restricting of taxi movement and service spot occupation and the secondary factors
represent relevant time efficiency drains in the system. This situation is causing many problems
of pedestrian congestion, passenger discomfort and fatigue, taxi driver impatience and conflicts,
etc. Scenario analysis has proven that relatively simple and small changes to queuing or service
area configuration can yield drastic performance improvements without major regulatory changes.
94
4.2. Intervention Proposals and Suggestions for Future Research
After reaching the abovementioned conclusions, the author organized a set of proposals for
possible future interventions or decisions regarding the taxi system at Portela, or at the NAL.
The taxi system at Portela should be closely monitored for service quality and compliance with
price, service and vehicle regulations. For this, the author proposes the creation of a new
monitoring entity, integrated into the company structure of ANA, with the co-management or
active participation of taxi companies/associations. The only way to improve the image of the Taxi
sector among visiting tourists or airport passengers in Portugal is to tighten the monitoring
process, especially at the airport stand. This has already been recognized by all relevant and
participating agents, which seem to agree on this necessity.
Among the initiatives that this newly formed entity could promote are the mandatory installation of
automatic receipt machines and/or a GPS-based system in taxis, a detailed information panel on
the several pricing schemes, close to the passenger queue (located so as to not disturb queuing)
and random periodic inspections to ensure greater transparency and service quality.
This new entity could also have a coordinating and management role at the stand, effectively
replacing police in that function, releasing them for more relevant security roles.
The creation of differentiated service types should be considered, namely shared-taxis and
business-class taxi services, possibly taking advantage of the characteristics of the Departures
Taxi Stand as a suitable area for implementation. This would also imply the creation of a special
permit system, accessible through open public tendering, subject to minimum technical
requirements for taxis and drivers. Such as system would help foster investment and innovation
on newer and safer vehicles and driver professional training, while synchronizing higher service
quality with higher willingness to pay and comfort requirements. Special focus should exist on the
definition of the numerical limits on permits, based on expected demand and competition levels.
In the event of considering a permit system for the NAL’s taxi stand, another special taxi stand
could eventually be built or concessioned somewhere in the city of Lisbon, in order to create a
kind of shuttle system to avoid empty roundtrips, but every taxi should be allowed to drive
passengers to the airport, regardless of their origin. This secondary stand would have to be
carefully planned and located in order not to create a local monopoly and excessively tap into the
rest of the taxi demand in that area, despite the higher price of service.
A change in the concessionaire role should be studied for feasibility and efficiency, especially at
the future Lisbon Airport, in the south bank of the Tagus River. Following a airport-city logic of
development, the NAL’s taxi stand should become a concession of ANA, as a publicly owned
company and airport operator, increasingly decoupled from municipal and regional power. This
may or may not extend to the rest of the airports in the country, depending on the case.
Lisbon Municipality and municipalities in general are also seen as heavy and bureaucratic
institutions that may not act or focus adequately on the airport stand, as regulators. This, coupled
95
with the fact that the Montijo/Benavente Municipality has a smaller influence, power and resource
pool, should raise the question of the possibility of attributing this role to the IMTT.
The operational design of the taxi system at Portela should be changed, in order to achieve better
level of service, both at the queue for passengers and taxis, relating to queue length and waiting
times. The most cost-efficient of the studied alternatives seems to point to the change in the
queue configuration, to allow more room for the boarding phase and to the service area, with one
single service lane, a free lateral lane and four parking service spots (Scenario II.B).
The new queue area should be designed or even moved so that the main Terminal exits are
always clear of obstacles and people. This is a security and safety issue that cannot be ignored,
independently of peak-hour solicitations.
The road access to the parking facility (P1) should be moved to the lower level, close to the taxi
parking facility, in order to mitigate the effects of the conflicts between taxis and private vehicles,
currently present closer to the Terminal’s curbside.
The existing crosswalks should be slightly elevated in order to ensure passenger safety when
crossing in front or immediately behind the main taxi service area. While performing observations,
the author witnessed several dangerous situations involving passengers and taxis.
The taxi service system at the NAL (or the intervention at Portela) should be planned and
designed according to a methodology that explicitly considers peak-hour variability, and does not
focus on average behaviors. The best way to ensure a cost-effective way of building such a
system is to consider flexibility in design, such as the possibility to open an extra free lateral lane
or transform it into a service lane, depending on solicitations.
New service types such as shared-taxis and business class taxi services should require a small
management structure and a GPS-GIS system in order to process, pool and register clients. This
structure should be designed in order to avoid secondary in-terminal queues, promoting pre-flight
reservations and quick and easy taxi identification and boarding. Promoting the taxi voucher can
also help this objective.
Following this reasoning, the warning of new competition from the Metro, foreseen for late 2010,
should be a focus of attention and a call for innovation and increased service quality on behalf of
taxi companies as operators and Lisbon Municipality, as a regulator. Defining new taxi parking
limitations to take into account possible decreases of demand or investing in new, differentiated
service types should be seriously considered in this context.
The next steps in research for this topic should focus on three main vectors. First, there is an
inherent need to further quantify the benefits of the different regulatory and operational designs. On
the institutional and regulatory side, a more in-depth economic analysis could be performed on the
main market impacts of the several policies and on the operational side, perfected methods for
collecting and analyzing data from different case-studies should be invested upon. Secondly, it would
be interesting to study the value of flexibility in these kinds of systems, as a main driver for efficiency,
based upon real options analysis, for example. And finally, a more systematic view on the several
existing types of contractual arrangements and market conditions on other airports would also prove
useful for a better perception of what the world-wide patterns of airport taxi service provision are.
96
Bibliography
ANA Aeroporto de Lisboa - Procura e Capacidade // Sessão de Apresentação do Plano de Expansão do Aeroporto de Lisboa. - Lisboa : [s.n.], 2006.
CAA Civil Aviation Authority CAA Passenger Survey 2006. - [s.l.] : Civil Aviation Authority, 2006.
Cairns, Robert D. and Catherine Liston-Heyes Competition and regulation in the taxi industry // Journal of Public Economics. - London : Elsevier, 1996. - 59. - pp. 1-15.
Cao Y., Nsakanda, A.L. & Pressman, Irwin A Simulation Study of the Passenger Check-in System at the Ottawa International Airport. - [s.l.] : SCSC, 2003. - pp. 573-579. - ISBN: 1-56555-268-7.
Cardon, Nicolas The place of the taxi through urban mobility:its practices, positioning & potential for expansion // City on the move. - Lisboa : Institut de la Ville en Mouvement, 2007.
Cervero, Robert Deregulating urban transportation. - [s.l.] : Cato Journal, 1985. - pp. 219–237.
Cervero, Robert Fostering Commercial Transit: Alternatives in Greater Los Angeles // Reason Magazine. - 1992. - 146.
Cooper, James Ground Transportation, Airports and External Regulation Conflict, a worldwide question?. - 2004. - Presentation to the Transportation Research Board.
Corgan Associates Inc. Innovations for Airport Terminal Facilities. - Washington, D.C. : Transportation Research Board, 2008.
Curry, Guy L. , Arthur De Vany and Richard M. Feldman A queueing model of airport passenger departures by taxi // Transportation Research. - [s.l.] : Pergamon, 1977. - Vol. 12. - pp. 115-120.
Darshan Santani Rajesh Krishna Balan and C Jason Woodard Spatio-temporal Efficiency in a Taxi Dispatch System. - Singapore : [s.n.], 2007.
DeVany, Arthur S. Alternative ground transportation systems for Dallas/Fort Worth Airport. - [s.l.] : Texas A&M University, 1977.
Días Esteban, Pedro J. Check-in process at Lisbon Airport - Event-based Simulations and Collaborative Design. - Lisboa : Instituto Superior Técnico, 2008.
FCG-Parsons Plano Director de Referência de Desenvolvimento Conceptual do Aeroporto. - [s.l.] : NAER, 2002.
Flath, David Taxicab regulation in Japan // The Japanese and International Economies. - Raleigh : Elsevier, 2006. - 20. - pp. 288–304.
Frankena, M.W., Pautler, P.A. An Economic Analysis of Taxicab Regulation. - Washington, DC : Federal Trade Commission, 1984.
Gallick, Edward C. and Sisk, David E. A reconsideration of Taxi regulation // Journal of Law, Economics and Organization. - [s.l.] : Oxford University Press, 1987. - 1 : Vol. 3. - pp. 117-28.
97
Hai Yang Min Ye , Wilson H. Tang , S.C. Wong Regulating taxi services in the presence of congestion externality // Transportation Research. - Hong Kong : Elsevier, 2005. - Part A. - 39. - pp. 17–40.
Hai Yang Yan Wing Lau, Sze Chun Wong and Hong Kam Lo A macroscopic taxi model for passenger demand, taxi utilization and level of services // Transportation. - Hong Kong : Kluwer Academic Publishers, 2000. - 27. - pp. 317–340.
Hartman, Ron Improving Public Transport by Integrating Taxi Services // Taxi International Conference. - Lisbon : [s.n.], 2007.
Horn, Mark E. T. Fleet Scheduling and dispatching for demand-responsive passenger services // Transportation Research. - Canberra : Pergamon, 2002. - Part C. - 10. - pp. 35-63.
Hyunmyung, Kim Jun-Seok Oh and R. Jayakrishnan Effect of Taxi Information System on Efficiency and Quality of Taxi. - Washington, D.C. : Transportation Research Board, 2004.
Joustra, Paul E. and Dijk, Nico M. Van Simulation of Check-In at Airports // 2001 Winter Simulation Conference. - 2001.
K.I., Wong S.C., Wong, Hai Yang , J.H. Wu Modeling urban taxi services with multiple user classes and vehicle modes // Transportation Research. - Hong Kong : Elsevier, 2008. - Part B. - 42. - pp. 985–1007.
La Croix James Mak and Walter Miklius Airport taxi service regulation: An analysis of an exclusive contract. - [s.l.] : Martinus Nijhoff Publishers, Dordrecht, 1986.
La Croix James Mak and Walter Miklius Evaluation of alternative arrangements for the provision of airport taxi service. - Manoa : [s.n.], 1991.
Li, Sonny Multi-Attribute Taxi Logistics Optimization. - [s.l.] : MIT, 2006.
Newell, Gordon Applications of Queuing Theory. - [s.l.] : Chapman and Hall, 1982.
Odoni, Amedeo R. Airside Congestion Slides // Airport Systems Planning, Design, and Management Course. - [s.l.] : Massachusetts Institute of Technology, 2007.
Odoni, Amedeo R. e Larson Richard C. Urban Operations Research. - New Jersey : Prentice-Hall, 1981.
OECD Competition Committee Taxi Services: Competition and Regulation. - [s.l.] : OECD, 2007.
P.C. Productivity Commission Regulation of the Taxi Industry. - Canberra : Ausinfo, 1999.
Schaller, Bruce A Regression Model of the Number of Taxicabs in U.S. Cities // Journal of Public Transportation. - 2005. - 5 : Vol. 8. - pp. 63-78.
Schaller, Bruce Entry controls in taxi regulation: Implications of US and Canadian experience for taxi regulation and deregulation // Transport Policy. - New York : Elsevier, 2007. - 14. - pp. 490–506.
Toner, Jeremy P. An Econometric Approach to Demand, Supply and Service Quality in the Taxi Industry. - Leeds : Institute of Transport Studies, University of Leeds, 1991.
Toner, Jeremy P. The Demand for Taxis in Leeds and the Value of Time. - Leeds : Institute of Transport Studies, University of Leeds, 1991.
98
Valadares Tavares e Rui Oliveira Isabel Themido, Nunes Correia Investigação Operacional. - Lisboa : McGraw-Hill, 1996.
Web Sites
http://www.ny.com/transportation/taxis/
http://www.ville-en-mouvement.com/taxi/uk/articles.htm
http://www.cm-lisboa.pt/
http://www.springerlink.com/
http://www.sciencedirect.com/
http://www.b-on.pt/
http://www.antral.pt/
http://www.ana.pt/
http://www.askmelisboa.com/
http://www.citywidetaxi.ca/airport.html
http://www.imtt.pt/
http://www.brusselsairport.be/
http://www.lcacc.org/
http://www.taxi-library.org/stands.htm
http://www.airportbusiness.com/
http://www.simul8.com/
http://pt.wikipedia.org/
http://www.metrolisboa.pt/
http://www.transportesemrevista.com/
http://www.deco.proteste.pt/
http://www.taxiblog.co.uk/
http://www.caa.co.uk/
http://singaporepublictransport.blogspot.com/
http://jn.sapo.pt/paginainicial/interior.aspx?content_id=1155172#AreaComentarios
http://www.isixsigma.com/library/content/c000709a.asp
http://diario.iol.pt/noticia.html?id=138571&div_id=4071
https://www.washingtontimes.com/news/2006/jan/18/20060118-095014-4844r/print/
99
Annexes
100
I. Literature Review
Regulation
There are many studies on the adequacy of regulation on the taxi market, where questions about
market access, efficiency and sustainability are extensively debated, and different points of view can
emerge from the bibliography that can be found on the subject. On one side, fundamentals of
economic theory, supporting free market benefits such as lower prices, innovation and higher level of
service, deriving from increased competition, supported by relatively good experiences in other
sectors and other modes of transportation. On the other, imperfections in practice that many times
lead to market failures, which call for regulation. (Schaller, 2007)
Liberalization supporters base their reasoning on the claim that restrictions on entry to the taxi
industry constitute an unjustified restriction on competition, while also allowing for regulatory capture.
This means that large transfers from consumers to producers might occur, along with associated
economic distortions and corresponding deadweight losses. They also defend that no solid proof
exists on the claim that equity is better promoted through the implementation of entry restrictions; on
the contrary, higher prices and lower availability affect lower income taxi service consumers.
Regarding reform strategies, the main proposals state that immediate implementation of open entry
policies can be politically challenging, but necessary, because a slow, staged approach will most likely
lead to a stalled or reversed reform process.(OECD, 2007)
Some studies also argue that there is no persuasive economic rationale for some of the most
important regulations. These defend that restrictions on numerical limits for companies and vehicles
and on minimum fares waste resources and impose a disproportionate burden on low income people.
They also take a strong pro-liberalization stand on alternative service types, supporting that there is no
economic justification for regulations that restrict shared-ride, dial ride, and jitney service from
competing for parts of the transit market largely monopolized by bus and subway operators.(Frankena,
1984) This last discussion is in line with the perspective and claim that regulation can also inhibit
innovation and creation of alternative service types.
The availability argument is also very strong on the part of the deregulation supporters. Some
authors go as far as stating that “Studies have found that travelers are more sensitive to the
availability of taxis than they are to travel times, speeds, or almost any other service features. Where
taxis are given unrestricted freedom to ply their trade, the quality of’ urban transportation has generally
improved.” Availability of cab service would also improve, even in low-density areas, as ‘‘small taxi
companies and private individuals who are currently denied entrepreneurial freedom’’ will be able to
service ‘‘marginal markets abandoned by large fleets’’. (Cervero, 1985) Numerical limits on taxis and
companies are at the center of this argument for risk of low availability, which also focuses on the
excessively high prices for medallions and permits, which have emerged as a profitable secondary-
market, due to the scarcity of new permit issuing initiatives.
Pro-regulation supporters often point to significant risks of market failure to defend regulatory
measures on market entry access and quality of service. Among the most argued market
imperfections are the significant economies of scale and scope, which distort competition on some taxi
101
market segments, cross-subsidization between geographical areas and operational periods,
information asymmetry, negative externalities and oversupply. (Schaller, 2007) Some econometric
studies on the taxi sector conclude that Price regulation is necessary but not sufficient to produce
equilibrium in a simple model of the taxi industry, also stating that the authority can improve upon price
regulation by regulating the number of taxis as well. However, most conclusions stay away from the
idea that this topic can be generalized for all cities and all regulatory environments (Cairns, 1996)
(Gallick, 1987) (Flath, 2006). This contextual dependence perception is present in other studies which
point towards a spectrum of entry policies, rather than a simple choice between regulation and
deregulation (see Figure 46). The authors state that “The effects of entry policies depend on market
characteristics. Open entry has had negative effects on the availability and quality of cab service when
implemented in cities with a large number of cab stand and street hail trips.” (Schaller, 2007) This
supports the theory that entry policy impacts significantly differ according to market specificities.
Figure 46 - Schematic classification of taxicab regulatory systems (Schaller, 2007)
As far as the specific airport taxi sector is concerned, little literature is actually available. However,
there are some interesting studies that show that the market regulation mechanisms can be
substantially different according to the city and airport. Schaller’s 2007 study of 43 cities and counties
in the United States and Canada concludes that in total, 21 of the 43 locations do not allow
unrestricted access to pick up at the airport stands, while 20 of them allow free access to the airport
terminal services to any taxi, with minimum requirements/restrictions. An interesting fact is that all
locations of Type A regulation system (see Figure 46) have restrictions on entry to the airport taxi
service and only one of the Type B locations (Dallas, TX) allows free access to it, although dependent
on the compliance of cabs with “specified requirements”. This might be related to the fact that there
are no numerical limits to taxis in general, in these 10 locations - 5 of type A and 5 of type B - which
could cause severe problems of oversupply and long queuing at the “profitable” airport stands and
unbalances in taxi service provision elsewhere in the region. In type C and D locations, general
numerical limits are usually in place, but the existence of airport access restrictions is varied, despite a
slight predominance of absence of significant barriers to entry, probably due to the existence of
controlled total number of taxis and need to politically compromise with existing operators (Figure 47).
102
Figure 47 - Key characteristics of entry-related policies - adapted from (Schaller, 2007)
Other airport taxi markets have been studied in the past. In his research, La Croix and his
colleagues defended a very strong pro-regulation attitude for taxi markets, focusing on the airport
stands. In a period – beginning of the 1980’s - where exclusive contracts for taxi services at airports
were being questioned for supposed social and economic inefficiencies and deregulation was being
adopted by many airport authorities, the pressure on airports that kept an exclusive contract with a
103
single taxi company increased. La Croix’s study points two key reasons for arguing that a decrease in
regulation would not promote price competition: Firstly, large numbers of arriving passengers lack
adequate information regarding local taxi fares, the best route to their destination, and the precise
distance to their destination. Secondly, the FIFO queue discipline imposed upon waiting taxis raises
the cost of consumer search, essentially restricting passengers to examining taxis in the order they are
lined up in the queue, that is, taxis must be examined sequentially, contrary to usual consumer search
procedure where choices are made over options that are displayed simultaneously (La Croix, 1986).
The contractual types present at airport taxi stand concessions are also a relevant issue, since
different arrangements can lead to different impacts on service access and efficiency. Categorization
is usually defined into three types: Exclusive Contract, where a single taxi company is granted the
privilege to solicit passengers leaving the airport; Permit System, when a government agency issues a
limited number of permits to selected taxi operators to provide service and Open System, in which any
licensed taxicab in the metropolitan area is allowed to solicit passengers at the airport. Discussion on
the benefits and disadvantages of the several arrangements are frequent in the bibliography, with
arguments on both sides. In his analysis of the exclusive contract to provide taxi service at the
Dallas/Fort Worth Airport, DeVany concluded that objections to an open, competitive system cannot
be sustained "in light of the inefficiency of the present exclusive system, and what is known or can be
predicted about a competitive system." He also suggests that the choice of a single taxi operator by
airport administrators is due to their preference for a simple life at the expense of economic efficiency.
(DeVany, 1977) Additional studies by La Croix, aimed at countering DeVany’s conclusions, analyzed
specific contractual arrangements at some U.S. airports in terms of level of service, price fairness,
revenues and deadheading, in order to assess the adequacy of market access restrictions. According
to La Croix, Airport authorities wish to design a contract which collects rents and limits rent-seeking by
taxi operators, provides better quality of service than elsewhere in the metropolitan area and is
politically balanced. These objectives are more or less achieved depending on the type of
arrangement that is implemented (La Croix, 1991):
Exclusive contracts are described as better at minimizing rent-seeking, reducing administrative
costs and provide greater flexibility than the Permit system in face of fluctuating demand for
services. It is, however, less politically balanced, in the sense that it excludes competition.
Permit systems are preferred in cases where service quality is less important and the exclusive
contract lacks political support. If demand is relatively stable, the system becomes more
sustainable for permit holders and quality of service might remain at good levels, while being
politically acceptable. Disadvantages are mainly related with monitoring costs, which are much
higher than in the exclusive contract system.
Open systems are seen by the authors as prone to rent-seeking, low quality of service and high
administrative cost problems. This kind of system is supposedly bound to be chosen in situations
where quality of service is not a priority and/or there is a significant political leverage on behalf of
a large number of taxi operators.
104
Other complementary documents, such as the Commission Research Paper, by the
Commonwealth of Australia (P.C., 1999), the presentation on Ground Transportation Regulation and
Airports, by James Cooper (Cooper, 2004), papers on Taxi demand, supply, quality of service and
Value of Time, by Jeremy P. Toner (Toner, 1991) or a Congestion Externalities study, by the
University of Hong Kong (Hai Yang, 2005), can also prove to be insightful.
Modeling, Queuing Theory and Simulation
Several studies relate to the modeling of transportation systems, such as air and road traffic,
pedestrian behavior and network planning, for example. This study focuses on a specific part of the
taxi (and road) network and also relates, somewhat, to the “logistic” chain of an airport system, namely
its land side operations. Studies on airport taxi stand performance and design are almost inexistent,
and the specificity of the infrastructure and market context is so high, that comparisons with taxi
studies in general provide only contextual perceptions on taxi service types and network effects. Still,
several documents proved interesting, namely those that deal with queue modeling and simulation,
central issue at airport terminals, especially regarding key customer interface services such as check-
in systems, for example.
Taxi service modeling in general has produced a lot of bibliography. It is often though Network-
based models and optimization that several studies approach the system mechanics of taxi services.
In 2004, Hyunmyung Kim and his colleagues modeled a taxi service system in urban areas
considering taxi drivers' knowledge on the transportation network, in which the taxi drivers’ passenger
seeking behavior is modeled, based their expected travel time and expected waiting time. The
intention was to assess the effect of a taxi information system on efficiency and quality of taxi services.
Through the articulation between a Stochastic Network and Demand model and an Inductive Learning
model, the study concludes that despite taxi drivers’ improved knowledge on network condition from
their experience, the operational efficiency and the quality of taxi service may be not improved. The
taxi information system helps drivers efficiently seek passengers and reduces unnecessary travel,
providing a benefit equivalent to increasing the number of taxis by 20% in terms of quality of taxi
service (Hyunmyung Kim, 2004).
Another study, of 2008, models urban taxi services in congested networks to the case of multiple
user classes, multiple taxi modes, and customer hierarchical modal choice. The model is based on a
set of assumptions on Taxi movements in a road network, Customer and taxi waiting times, Cost
structures, Taxi service time constraints, Behavior of vacant taxi drivers and Hierarchical logit mode
choice. Although complex in nature, the model seems to be an interesting tool for the planning and
evaluation of different policy options and scenarios for urban taxi services, mainly due to this added
flexibility of defining different users and services within the same transportation sector. This ability
makes the model applicable to a wide range of taxi problems, such as the modeling of accessible taxis
for providing special services to handicapped passengers, and luxury taxis with better services and
facilities for affluent customers. (K.I. Wong, 2008)
105
Computer-based Simulation models are also used on Sonny Li’s study of Multi-Attribute Taxi
Logistics Optimization. This computer model tests various attributes that affect logistic optimizations
for taxi services. In particular, the effect of taxi fleet size, the quantity of hotspots, and the
concentrations of customers at hotspots are analyzed in detail using the model. The metric of interest
includes the customers' wait time, taxi revenue, and costs of operations. Model outputs (see Figure 48
for an example) consist of Total Customer Wait Time (Generic or Hotspot), Total number of customers
(Generic and Hotspot), Taxi Idling Time, Taxing Time, Taxi Pickup Time and Taxi Courier Time.
Figure 48 – Customer Wait Time versus Number of Taxis (Li, 2006)
Among the main conclusions of this study, special relevance to the fact that the results of the
experiments indicate that as the number of taxis increases, the customer wait time will decrease and
the average revenue of each taxi will increase; this will be true as long as customer demand is greater
than or equal to the supply of taxis. As main aspects towards the improvement of efficiency, measures
such as the incorporation of real-time demand and current traffic conditions into taxi's dispatching
system as well as adoption of a GPS-GIS system to fleet management are mentioned. (Li, 2006)
The difficulty of analytically handling transient behaviors has lead to the widespread use of
simulation software, as tools for obtaining approximate results while being less time consuming. Two
of the considered studies relate to this. Yuheng Cao and his colleagues presented a simulation of the
check-in system at the Ottawa International Airport, where significant amount of data was collected
and used to define the inputs to a simulation model. The performance variables taken as output are
the average waiting time in queues, the maximum waiting time in queues, the average queue length,
the maximum queue length, and the distribution of passengers waiting times in queues. The data
inputs for the model were determined through data collection and interviews with airport managers.
For the arrival patterns, data was gathered during both the morning and afternoon peak periods. For
the service rate, sample data was randomly collected at different times using stopwatch methodology.
The data gathered were: passenger party size; the number of pieces of luggage; passenger
destination; and the flight number. This paper is important, not only because it deals with time-
dependent arrival/service rates in queuing systems, but also because it focuses on the data collection
procedures that “feed” the queue modeling parameters, describing the statistical methods that allow
for a successful calibration of the model. These first steps are crucial when building a simulation,
especially in the absence of pre-existent and reliable data (Cao, 2003).
106
Paul E. Joustra and Nico M. Van Dijk also published, in 2001, an interesting paper whose main
purpose is to describe why simulation is necessary to evaluate check-in processes (queues). The
authors believe that classical queuing theory principles are not adequate for modeling check-in
services, namely because “(…)these formulas represent so-called steady state situations. For the
check-in process this would imply that the arrival rates of passengers are constant during long periods
of time. This is clearly not the case with check-in arrival patterns. In contrast peakedness and
variability is the major concern for planning.” Despite this, they find Queuing Theory valuable as a
way to support verification and validation of a simulation model, defining experiments and analyzing
results. Main identified simulation advantages are: the fact that it can deal with the peaks in arrival
patterns; offer the freedom of using arbitrary distributions for the check-in processing time and arrival
patterns; dynamically test alternative check-in schemes, quantifying the changes and offer animation
to support the communication at both management and operational level. (Joustra, 2001)
Finally, Guy L. Curry, Arthur De Vany and Richard M. Feldman worked on a queuing model of
airport passenger departures by taxi and bus competition, in 1977, producing a study which offers a
closer perspective into transient behavior, coupled with taxi services. The authors built a model, based
on a set of assumptions and notation, which they then proceed to test, by constructing a simulation
and calibrating it from sample data collected at Dallas airport. After registering daily cyclic fluctuations
in demand over a five day period, the authors used multiple period analysis with a step function for the
mean arrival rate (see Figure 49). A further simplification was made by assuming that the transient
behavior is short lived relative to the interval size for the step function. This assumption allowed the
use of the steady state results computed with the parameters appropriate to the individual periods.
Figure 49 – Mean customer demand by time of the day (Curry, 1977)
As with regulation, there are many other studies pertaining to modeling of taxi services or similar
individualized vehicle services optimization, such as the paper by Hai Yang et al, (Hai Yang, 2000),
Darshan Santani et al (Darshan Santani, 2007), or work by Mark Horn, (Horn, 2002). A very
interesting approach, based on SIMUL8 software (also used in this thesis) is done on airport check-in
services simulation by Pedro Díaz, (Días Esteban, 2008). Finally, two very interesting sources of
information: one related to a statistical modeling of the factors that influence number of taxicabs in US
cities, by Bruce Schaller, (Schaller, 2005) and the other speaks about recent innovations of airport
terminal facilities, which include taxi services, by Corgan Associates for the Airport Cooperative
Research Program, (Corgan Associates Inc., 2008).
107
II. Field Data
108
Measurement took place from 8 to 10 a.m. 05-08-2009
Measurement took place from 9 to 10 p.m. 13-08-2009
Hour of the
day
Inter-Arrival
Times for
groups
Inter-Arrival Times
for Groups
(seconds)
Number of
people
Cumulative
Time
Cumulative
number of
people
Hour of the
day
Inter-Arrival
Times for
groups
Inter-Arrival
Times for Groups
(seconds)
Number of
people
Cumulative
Time
Cumulative
number of
people
8:03:25 0:03:25 205 2 205 2 21:00:10 0:00:10 10 2 10 2
8:04:30 0:01:05 65 1 270 3 21:00:34 0:00:24 24 2 34 4
8:06:59 0:02:29 149 2 419 5 21:00:55 0:00:21 21 3 55 7
8:07:00 0:00:01 1 1 420 6 21:01:20 0:00:25 25 1 80 8
8:07:02 0:00:02 2 5 422 11 21:02:17 0:00:57 57 1 137 9
8:07:40 0:00:38 38 4 460 15 21:02:41 0:00:24 24 1 161 10
8:08:05 0:00:25 25 1 485 16 21:02:43 0:00:02 2 1 163 11
8:08:07 0:00:02 2 1 487 17 21:02:47 0:00:04 4 1 167 12
8:08:44 0:00:37 37 2 524 19 21:03:00 0:00:13 13 2 180 14
8:09:27 0:00:43 43 1 567 20 21:03:38 0:00:38 38 1 218 15
8:09:34 0:00:07 7 2 574 22 21:04:15 0:00:37 37 1 255 16
8:09:51 0:00:17 17 1 591 23 21:04:22 0:00:07 7 1 262 17
8:13:08 0:03:17 197 1 788 24 21:04:35 0:00:13 13 1 275 18
8:15:09 0:02:01 121 1 909 25 21:04:52 0:00:17 17 2 292 20
8:17:15 0:02:06 126 1 1035 26 21:05:15 0:00:23 23 1 315 21
8:17:37 0:00:22 22 1 1057 27 21:05:40 0:00:25 25 1 340 22
8:21:26 0:03:49 229 1 1286 28 21:06:08 0:00:28 28 2 368 24
8:24:18 0:02:52 172 1 1458 29 21:06:10 0:00:02 2 1 370 25
8:24:40 0:00:22 22 1 1480 30 21:06:55 0:00:45 45 3 415 28
8:25:48 0:01:08 68 1 1548 31 21:07:48 0:00:53 53 4 468 32
8:26:31 0:00:43 43 1 1591 32 21:08:17 0:00:29 29 1 497 33
8:27:08 0:00:37 37 2 1628 34 21:08:30 0:00:13 13 3 510 36
8:28:00 0:00:52 52 1 1680 35 21:08:35 0:00:05 5 4 515 40
8:28:14 0:00:14 14 1 1694 36 21:08:46 0:00:11 11 2 526 42
8:28:28 0:00:14 14 2 1708 38 21:08:54 0:00:08 8 2 534 44
8:28:56 0:00:28 28 5 1736 43 21:09:21 0:00:27 27 1 561 45
8:29:38 0:00:42 42 3 1778 46 21:09:33 0:00:12 12 1 573 46
8:31:40 0:02:02 122 2 1900 48 21:09:37 0:00:04 4 4 577 50
8:32:37 0:00:57 57 2 1957 50 21:10:15 0:00:38 38 3 615 53
8:33:34 0:00:57 57 2 2014 52 21:10:22 0:00:07 7 1 622 54
8:34:24 0:00:50 50 1 2064 53 21:10:28 0:00:06 6 2 628 56
8:34:36 0:00:12 12 2 2076 55 21:10:59 0:00:31 31 1 659 57
8:35:53 0:01:17 77 2 2153 57 21:11:09 0:00:10 10 2 669 59
8:36:06 0:00:13 13 1 2166 58 21:11:16 0:00:07 7 5 676 64
8:36:10 0:00:04 4 1 2170 59 21:11:22 0:00:06 6 4 682 68
8:37:11 0:01:01 61 1 2231 60 21:11:41 0:00:19 19 2 701 70
8:37:17 0:00:06 6 2 2237 62 21:11:47 0:00:06 6 3 707 73
109
8:37:57 0:00:40 40 1 2277 63 21:12:03 0:00:16 16 2 723 75
8:39:14 0:01:17 77 2 2354 65 21:12:27 0:00:24 24 1 747 76
8:39:32 0:00:18 18 1 2372 66 21:12:31 0:00:04 4 2 751 78
8:40:59 0:01:27 87 1 2459 67 21:12:37 0:00:06 6 1 757 79
8:41:32 0:00:33 33 1 2492 68 21:12:42 0:00:05 5 3 762 82
8:41:49 0:00:17 17 3 2509 71 21:12:48 0:00:06 6 2 768 84
8:42:18 0:00:29 29 3 2538 74 21:13:09 0:00:21 21 3 789 87
8:43:23 0:01:05 65 4 2603 78 21:13:13 0:00:04 4 3 793 90
8:44:23 0:01:00 60 2 2663 80 21:13:19 0:00:06 6 1 799 91
8:44:39 0:00:16 16 1 2679 81 21:13:34 0:00:15 15 1 814 92
8:44:57 0:00:18 18 1 2697 82 21:14:13 0:00:39 39 2 853 94
8:45:48 0:00:51 51 2 2748 84 21:14:17 0:00:04 4 2 857 96
8:47:12 0:01:24 84 1 2832 85 21:14:22 0:00:05 5 1 862 97
8:47:14 0:00:02 2 1 2834 86 21:14:31 0:00:09 9 2 871 99
8:47:53 0:00:39 39 1 2873 87 21:14:37 0:00:06 6 1 877 100
8:47:57 0:00:04 4 9 2877 96 21:15:20 0:00:43 43 1 920 101
8:48:46 0:00:49 49 3 2926 99 21:15:28 0:00:08 8 1 928 102
8:49:10 0:00:24 24 3 2950 102 21:15:34 0:00:06 6 2 934 104
8:49:59 0:00:49 49 1 2999 103 21:15:38 0:00:04 4 1 938 105
8:50:29 0:00:30 30 1 3029 104 21:15:51 0:00:13 13 2 951 107
8:50:31 0:00:02 2 1 3031 105 21:15:59 0:00:08 8 1 959 108
8:50:50 0:00:19 19 2 3050 107 21:16:21 0:00:22 22 3 981 111
8:51:03 0:00:13 13 2 3063 109 21:16:35 0:00:14 14 3 995 114
8:51:31 0:00:28 28 2 3091 111 21:16:39 0:00:04 4 3 999 117
8:52:06 0:00:35 35 4 3126 115 21:16:50 0:00:11 11 1 1010 118
8:52:42 0:00:36 36 1 3162 116 21:17:20 0:00:30 30 2 1040 120
8:52:50 0:00:08 8 2 3170 118 21:17:46 0:00:26 26 2 1066 122
8:53:16 0:00:26 26 1 3196 119 21:17:52 0:00:06 6 1 1072 123
8:53:17 0:00:01 1 1 3197 120 21:18:09 0:00:17 17 1 1089 124
8:53:50 0:00:33 33 6 3230 126 21:18:22 0:00:13 13 2 1102 126
8:53:59 0:00:09 9 3 3239 129 21:19:21 0:00:59 59 1 1161 127
8:54:21 0:00:22 22 1 3261 130 21:19:48 0:00:27 27 3 1188 130
8:54:37 0:00:16 16 1 3277 131 21:20:30 0:00:42 42 1 1230 131
8:54:39 0:00:02 2 2 3279 133 21:21:04 0:00:34 34 2 1264 133
8:55:01 0:00:22 22 3 3301 136 21:21:37 0:00:33 33 1 1297 134
8:55:21 0:00:20 20 2 3321 138 21:21:46 0:00:09 9 1 1306 135
8:55:50 0:00:29 29 1 3350 139 21:22:16 0:00:30 30 2 1336 137
8:56:00 0:00:10 10 2 3360 141 21:22:20 0:00:04 4 1 1340 138
8:56:12 0:00:12 12 1 3372 142 21:22:35 0:00:15 15 3 1355 141
8:56:55 0:00:43 43 2 3415 144 21:22:50 0:00:15 15 2 1370 143
8:57:42 0:00:47 47 1 3462 145 21:23:01 0:00:11 11 1 1381 144
8:59:03 0:01:21 81 1 3543 146 21:23:05 0:00:04 4 1 1385 145
110
8:59:24 0:00:21 21 3 3564 149 21:23:52 0:00:47 47 1 1432 146
8:59:35 0:00:11 11 1 3575 150 21:24:05 0:00:13 13 1 1445 147
8:59:42 0:00:07 7 4 3582 154 21:24:16 0:00:11 11 1 1456 148
8:59:52 0:00:10 10 2 3592 156 21:24:21 0:00:05 5 1 1461 149
8:59:56 0:00:04 4 2 3596 158 21:24:46 0:00:25 25 2 1486 151
9:00:28 0:00:32 32 2 3628 160 21:24:59 0:00:13 13 1 1499 152
9:01:15 0:00:47 47 2 3675 162 21:25:15 0:00:16 16 1 1515 153
9:01:36 0:00:21 21 2 3696 164 21:25:24 0:00:09 9 2 1524 155
9:02:45 0:01:09 69 1 3765 165 21:25:45 0:00:21 21 2 1545 157
9:02:46 0:00:01 1 1 3766 166 21:25:49 0:00:04 4 1 1549 158
9:02:47 0:00:01 1 1 3767 167 21:26:20 0:00:31 31 2 1580 160
9:03:58 0:01:11 71 1 3838 168 21:27:31 0:01:11 71 1 1651 161
9:04:06 0:00:08 8 1 3846 169 21:28:30 0:00:59 59 1 1710 162
9:04:07 0:00:01 1 2 3847 171 21:28:49 0:00:19 19 2 1729 164
9:04:08 0:00:01 1 1 3848 172 21:29:54 0:01:05 65 1 1794 165
9:04:09 0:00:01 1 1 3849 173 21:30:02 0:00:08 8 3 1802 168
9:05:22 0:01:13 73 1 3922 174 21:30:07 0:00:05 5 2 1807 170
9:05:24 0:00:02 2 3 3924 177 21:30:24 0:00:17 17 4 1824 174
9:05:40 0:00:16 16 3 3940 180 21:30:27 0:00:03 3 2 1827 176
9:05:52 0:00:12 12 3 3952 183 21:30:42 0:00:15 15 2 1842 178
9:06:04 0:00:12 12 2 3964 185 21:31:07 0:00:25 25 2 1867 180
9:06:05 0:00:01 1 1 3965 186 21:31:12 0:00:05 5 2 1872 182
9:06:15 0:00:10 10 3 3975 189 21:31:36 0:00:24 24 2 1896 184
9:06:16 0:00:01 1 2 3976 191 21:31:42 0:00:06 6 3 1902 187
9:07:19 0:01:03 63 3 4039 194 21:31:50 0:00:08 8 3 1910 190
9:07:27 0:00:08 8 2 4047 196 21:32:00 0:00:10 10 1 1920 191
9:07:28 0:00:01 1 2 4048 198 21:32:12 0:00:12 12 2 1932 193
9:07:59 0:00:31 31 3 4079 201 21:32:18 0:00:06 6 2 1938 195
9:08:39 0:00:40 40 2 4119 203 21:32:23 0:00:05 5 1 1943 196
9:09:25 0:00:46 46 2 4165 205 21:32:27 0:00:04 4 1 1947 197
9:10:08 0:00:43 43 2 4208 207 21:32:42 0:00:15 15 2 1962 199
9:10:39 0:00:31 31 2 4239 209 21:32:59 0:00:17 17 4 1979 203
9:10:41 0:00:02 2 2 4241 211 21:33:10 0:00:11 11 1 1990 204
9:10:42 0:00:01 1 1 4242 212 21:33:19 0:00:09 9 1 1999 205
9:11:20 0:00:38 38 2 4280 214 21:33:22 0:00:03 3 3 2002 208
9:11:42 0:00:22 22 2 4302 216 21:33:35 0:00:13 13 2 2015 210
9:11:44 0:00:02 2 1 4304 217 21:33:45 0:00:10 10 4 2025 214
9:12:02 0:00:18 18 3 4322 220 21:33:54 0:00:09 9 3 2034 217
9:12:04 0:00:02 2 1 4324 221 21:34:00 0:00:06 6 3 2040 220
9:12:41 0:00:37 37 2 4361 223 21:34:14 0:00:14 14 3 2054 223
9:13:23 0:00:42 42 2 4403 225 21:34:20 0:00:06 6 1 2060 224
9:13:26 0:00:03 3 1 4406 226 21:34:29 0:00:09 9 1 2069 225
111
9:14:40 0:01:14 74 2 4480 228 21:34:40 0:00:11 11 1 2080 226
9:14:43 0:00:03 3 2 4483 230 21:34:48 0:00:08 8 6 2088 232
9:15:17 0:00:34 34 3 4517 233 21:35:02 0:00:14 14 2 2102 234
9:15:19 0:00:02 2 1 4519 234 21:35:35 0:00:33 33 3 2135 237
9:15:36 0:00:17 17 2 4536 236 21:35:52 0:00:17 17 2 2152 239
9:15:59 0:00:23 23 3 4559 239 21:35:57 0:00:05 5 1 2157 240
9:16:08 0:00:09 9 5 4568 244 21:36:02 0:00:05 5 1 2162 241
9:17:08 0:01:00 60 4 4628 248 21:36:07 0:00:05 5 1 2167 242
9:17:15 0:00:07 7 2 4635 250 21:36:20 0:00:13 13 1 2180 243
9:17:17 0:00:02 2 2 4637 252 21:36:53 0:00:33 33 3 2213 246
9:17:38 0:00:21 21 4 4658 256 21:37:19 0:00:26 26 2 2239 248
9:18:08 0:00:30 30 2 4688 258 21:37:37 0:00:18 18 2 2257 250
9:18:40 0:00:32 32 1 4720 259 21:37:56 0:00:19 19 3 2276 253
9:18:50 0:00:10 10 1 4730 260 21:38:18 0:00:22 22 1 2298 254
9:18:56 0:00:06 6 4 4736 264 21:38:40 0:00:22 22 1 2320 255
9:19:08 0:00:12 12 4 4748 268 21:38:46 0:00:06 6 1 2326 256
9:19:46 0:00:38 38 2 4786 270 21:39:23 0:00:37 37 2 2363 258
9:20:15 0:00:29 29 1 4815 271 21:39:30 0:00:07 7 3 2370 261
9:20:53 0:00:38 38 2 4853 273 21:39:37 0:00:07 7 3 2377 264
9:21:21 0:00:28 28 2 4881 275 21:39:50 0:00:13 13 3 2390 267
9:22:24 0:01:03 63 2 4944 277 21:41:08 0:01:18 78 1 2468 268
9:22:56 0:00:32 32 1 4976 278 21:41:18 0:00:10 10 1 2478 269
9:23:18 0:00:22 22 1 4998 279 21:41:29 0:00:11 11 1 2489 270
9:23:33 0:00:15 15 1 5013 280 21:41:38 0:00:09 9 1 2498 271
9:23:52 0:00:19 19 1 5032 281 21:42:33 0:00:55 55 3 2553 274
9:24:05 0:00:13 13 1 5045 282 21:42:57 0:00:24 24 1 2577 275
9:24:08 0:00:03 3 1 5048 283 21:43:02 0:00:05 5 2 2582 277
9:24:36 0:00:28 28 2 5076 285 21:43:09 0:00:07 7 4 2589 281
9:24:56 0:00:20 20 1 5096 286 21:43:15 0:00:06 6 4 2595 285
9:24:57 0:00:01 1 1 5097 287 21:43:33 0:00:18 18 5 2613 290
9:25:41 0:00:44 44 2 5141 289 21:43:50 0:00:17 17 1 2630 291
9:25:44 0:00:03 3 1 5144 290 21:44:12 0:00:22 22 3 2652 294
9:25:57 0:00:13 13 2 5157 292 21:44:52 0:00:40 40 2 2692 296
9:26:20 0:00:23 23 1 5180 293 21:44:55 0:00:03 3 2 2695 298
9:26:36 0:00:16 16 1 5196 294 21:45:15 0:00:20 20 2 2715 300
9:26:44 0:00:08 8 1 5204 295 21:45:17 0:00:02 2 1 2717 301
9:27:59 0:01:15 75 2 5279 297 21:45:42 0:00:25 25 1 2742 302
9:28:47 0:00:48 48 1 5327 298 21:45:57 0:00:15 15 2 2757 304
9:29:09 0:00:22 22 1 5349 299 21:46:06 0:00:09 9 1 2766 305
9:29:20 0:00:11 11 2 5360 301 21:46:21 0:00:15 15 2 2781 307
9:29:48 0:00:28 28 2 5388 303 21:47:10 0:00:49 49 2 2830 309
9:29:55 0:00:07 7 1 5395 304 21:47:15 0:00:05 5 2 2835 311
112
9:30:16 0:00:21 21 2 5416 306 21:47:22 0:00:07 7 4 2842 315
9:30:48 0:00:32 32 1 5448 307 21:47:28 0:00:06 6 1 2848 316
9:31:45 0:00:57 57 1 5505 308 21:47:35 0:00:07 7 1 2855 317
9:32:50 0:01:05 65 2 5570 310 21:47:38 0:00:03 3 3 2858 320
9:32:58 0:00:08 8 2 5578 312 21:47:46 0:00:08 8 2 2866 322
9:33:45 0:00:47 47 1 5625 313 21:47:53 0:00:07 7 2 2873 324
9:34:23 0:00:38 38 4 5663 317 21:47:58 0:00:05 5 2 2878 326
9:34:47 0:00:24 24 2 5687 319 21:48:04 0:00:06 6 2 2884 328
9:35:14 0:00:27 27 1 5714 320 21:48:10 0:00:06 6 1 2890 329
9:35:28 0:00:14 14 4 5728 324 21:48:15 0:00:05 5 1 2895 330
9:35:51 0:00:23 23 2 5751 326 21:48:20 0:00:05 5 1 2900 331
9:35:58 0:00:07 7 2 5758 328 21:48:48 0:00:28 28 1 2928 332
9:36:08 0:00:10 10 1 5768 329 21:48:53 0:00:05 5 2 2933 334
9:36:18 0:00:10 10 1 5778 330 21:49:04 0:00:11 11 2 2944 336
9:36:47 0:00:29 29 1 5807 331 21:49:59 0:00:55 55 3 2999 339
9:37:03 0:00:16 16 2 5823 333 21:50:11 0:00:12 12 2 3011 341
9:37:14 0:00:11 11 2 5834 335 21:50:18 0:00:07 7 1 3018 342
9:37:48 0:00:34 34 4 5868 339 21:50:21 0:00:03 3 2 3021 344
9:38:25 0:00:37 37 1 5905 340 21:50:25 0:00:04 4 4 3025 348
9:38:31 0:00:06 6 2 5911 342 21:50:32 0:00:07 7 2 3032 350
9:38:38 0:00:07 7 3 5918 345 21:50:40 0:00:08 8 4 3040 354
9:38:43 0:00:05 5 2 5923 347 21:50:44 0:00:04 4 1 3044 355
9:39:03 0:00:20 20 3 5943 350 21:50:53 0:00:09 9 2 3053 357
9:39:17 0:00:14 14 3 5957 353 21:51:07 0:00:14 14 3 3067 360
9:39:26 0:00:09 9 2 5966 355 21:51:08 0:00:01 1 1 3068 361
9:39:46 0:00:20 20 3 5986 358 21:52:16 0:01:08 68 3 3136 364
9:40:21 0:00:35 35 1 6021 359 21:52:28 0:00:12 12 2 3148 366
9:41:15 0:00:54 54 1 6075 360 21:52:46 0:00:18 18 2 3166 368
9:41:25 0:00:10 10 2 6085 362 21:52:59 0:00:13 13 2 3179 370
9:41:34 0:00:09 9 1 6094 363 21:53:03 0:00:04 4 3 3183 373
9:41:51 0:00:17 17 2 6111 365 21:53:12 0:00:09 9 2 3192 375
9:42:03 0:00:12 12 1 6123 366 21:53:16 0:00:04 4 2 3196 377
9:42:24 0:00:21 21 2 6144 368 21:53:24 0:00:08 8 4 3204 381
9:42:44 0:00:20 20 2 6164 370 21:53:31 0:00:07 7 3 3211 384
9:42:59 0:00:15 15 4 6179 374 21:53:40 0:00:09 9 2 3220 386
9:43:08 0:00:09 9 2 6188 376 21:53:48 0:00:08 8 6 3228 392
9:43:19 0:00:11 11 2 6199 378 21:54:07 0:00:19 19 3 3247 395
9:43:27 0:00:08 8 2 6207 380 21:54:23 0:00:16 16 2 3263 397
9:43:29 0:00:02 2 1 6209 381 21:54:29 0:00:06 6 2 3269 399
9:43:39 0:00:10 10 2 6219 383 21:54:38 0:00:09 9 2 3278 401
9:43:49 0:00:10 10 3 6229 386 21:54:43 0:00:05 5 1 3283 402
9:43:55 0:00:06 6 1 6235 387 21:54:54 0:00:11 11 3 3294 405
113
9:43:58 0:00:03 3 2 6238 389 21:55:08 0:00:14 14 1 3308 406
9:44:19 0:00:21 21 1 6259 390 21:55:22 0:00:14 14 2 3322 408
9:44:39 0:00:20 20 3 6279 393 21:55:31 0:00:09 9 4 3331 412
9:45:08 0:00:29 29 4 6308 397 21:55:38 0:00:07 7 2 3338 414
9:45:14 0:00:06 6 2 6314 399 21:55:44 0:00:06 6 1 3344 415
9:45:39 0:00:25 25 3 6339 402 21:55:52 0:00:08 8 3 3352 418
9:46:08 0:00:29 29 1 6368 403 21:55:59 0:00:07 7 1 3359 419
9:46:10 0:00:02 2 1 6370 404 21:56:07 0:00:08 8 2 3367 421
9:46:18 0:00:08 8 1 6378 405 21:56:12 0:00:05 5 4 3372 425
9:46:23 0:00:05 5 2 6383 407 21:56:22 0:00:10 10 3 3382 428
9:46:31 0:00:08 8 2 6391 409 21:56:34 0:00:12 12 2 3394 430
9:46:38 0:00:07 7 2 6398 411 21:56:42 0:00:08 8 3 3402 433
9:46:58 0:00:20 20 2 6418 413 21:56:49 0:00:07 7 5 3409 438
9:47:08 0:00:10 10 2 6428 415 21:56:57 0:00:08 8 2 3417 440
9:47:38 0:00:30 30 2 6458 417 21:56:58 0:00:01 1 1 3418 441
9:48:27 0:00:49 49 1 6507 418 21:57:09 0:00:11 11 1 3429 442
9:48:38 0:00:11 11 2 6518 420 21:57:16 0:00:07 7 2 3436 444
9:48:40 0:00:02 2 1 6520 421 21:57:20 0:00:04 4 3 3440 447
9:48:57 0:00:17 17 1 6537 422 21:57:33 0:00:13 13 2 3453 449
9:49:04 0:00:07 7 1 6544 423 21:57:48 0:00:15 15 2 3468 451
9:49:32 0:00:28 28 2 6572 425 21:58:15 0:00:27 27 5 3495 456
9:49:43 0:00:11 11 2 6583 427 21:58:25 0:00:10 10 2 3505 458
9:49:45 0:00:02 2 2 6585 429 21:58:34 0:00:09 9 3 3514 461
9:49:48 0:00:03 3 2 6588 431 21:58:50 0:00:16 16 1 3530 462
9:49:58 0:00:10 10 1 6598 432 21:59:08 0:00:18 18 2 3548 464
9:50:07 0:00:09 9 3 6607 435 21:59:15 0:00:07 7 7 3555 471
9:50:11 0:00:04 4 1 6611 436 21:59:52 0:00:37 37 3 3592 474
9:50:18 0:00:07 7 1 6618 437 21:59:57 0:00:05 5 1 3597 475
9:50:47 0:00:29 29 2 6647 439 21:59:59 0:00:02 2 2 3599 477
9:50:59 0:00:12 12 3 6659 442
Average 15 2,04
9:51:10 0:00:11 11 1 6670 443
Std. Deviation 14 1,11
9:51:21 0:00:11 11 2 6681 445
Total
477 3599 477
9:51:45 0:00:24 24 2 6705 447 9:51:52 0:00:07 7 2 6712 449 9:52:02 0:00:10 10 2 6722 451 9:52:08 0:00:06 6 2 6728 453 9:52:15 0:00:07 7 2 6735 455 9:52:27 0:00:12 12 2 6747 457 9:52:31 0:00:04 4 2 6751 459
114
9:52:48 0:00:17 17 2 6768 461 9:53:03 0:00:15 15 2 6783 463 9:53:08 0:00:05 5 1 6788 464 9:53:34 0:00:26 26 1 6814 465 9:53:41 0:00:07 7 2 6821 467 9:53:48 0:00:07 7 1 6828 468 9:53:52 0:00:04 4 2 6832 470 9:53:55 0:00:03 3 2 6835 472 9:53:58 0:00:03 3 2 6838 474 9:54:08 0:00:10 10 2 6848 476 9:54:14 0:00:06 6 2 6854 478 9:54:45 0:00:31 31 4 6885 482 9:54:58 0:00:13 13 2 6898 484 9:55:08 0:00:10 10 4 6908 488 9:55:24 0:00:16 16 3 6924 491 9:55:32 0:00:08 8 1 6932 492 9:55:42 0:00:10 10 4 6942 496 9:56:07 0:00:25 25 2 6967 498 9:56:37 0:00:30 30 1 6997 499 9:57:26 0:00:49 49 3 7046 502 9:57:43 0:00:17 17 2 7063 504 9:57:55 0:00:12 12 1 7075 505 9:57:58 0:00:03 3 2 7078 507 9:58:12 0:00:14 14 2 7092 509 9:58:22 0:00:10 10 2 7102 511 9:58:27 0:00:05 5 2 7107 513 9:58:38 0:00:11 11 1 7118 514 9:58:47 0:00:09 9 2 7127 516 9:58:48 0:00:01 1 2 7128 518 9:59:08 0:00:20 20 2 7148 520 9:59:16 0:00:08 8 2 7156 522
Average 23 1,90
Std. Deviation 22 1,04
Total
522 7156 522
Figure 50 – Collected Inter-Arrival Times
Observation # Server 1 Server 3 Observation # Server 2 Server 4
Legend: 5-8-2009 sample
1 43 57 1 61 96
27-8-2009 sample
2 116 32 2 66 136
14-9-2009 sample
3 26 77 3 41 183
4 19 42 4 40 83
(All values are in seconds)
5 30 126 5 99 59
6 35 66 6 135 96
7 36 90 7 35 87
8 78 40 8 49 103
9 27 63 9 85 110
10 59 72 10 60 130
11 97 55 11 75 39
12 27 106 12 38 78
13 26 27 13 45 42
14 36 123 14 57 79
15 36 51 15 94 42
16 41 70 16 94 31
17 59 80 17 76 87
18 66 35 18 41 107
19 106 98 19 70 74
20 48 60 20 50 51
21 42 75 21 21 87
22 59 153 22 99 60
23 62 72 23 89 122
24 94 41 24 65 221
25 71 141 25 60 50
26 58 128 26 66 114
27 65 138 27 30 138
28 33 72 28 45 107
29 32 54 29 103 51
30 46 53 30 45 50
31 73 91 31 132
32 51 30 32 35
33 80 45 33 52
34 92 50 34 146
35 66 77 35 105
36 56 117 36 46
37 143 47 37 54
38 110 29 38 41
39 59 79 39 48
40 58 66 40 80
41 63 127 41 67
42 44 122 42 60
43 42 48 43 130
44 41 41 44 69
45 43 52 45 43
46 58 54
47 40 43
48 52 55
49 116 108
50 80 155
51 50 56
52 59 44
53 76 73
54 31 182
55 26 67
56 36 48
57 62 32
58 60 98
59 40 85
60 64 30
61 57 20
62 147 70
63 160 79
64 57 59
65 76 51
66 38 50
116
67 35 153
68 108 49
69 140 44
70 73 68
71 73 62
72 105 58
73 48 107
74 50 67
75 174 63
76 87 57
77 103 81
78 73 153
79 103 59
80 55 83
81 67
82 97
83 70
84 75
85 154
86 69
87 77
Figure 51 – Collected Service Times
117
III. Distribution Fitting - Inter-arrival and Service Times
Figure 52 – Exponential Theoretical Distribution fitting to the Inter-Arrival Times experimental distribution
Figure 53 – Goodness of fit and descriptive statistics summary for the Inter-Arrival Times
118
Figure 54 – Lognormal theoretical distribution fitting to the Service Times experimental distribution
Figure 55 - Goodness of fit summary and descriptive statistics for the Service Times
119
Figure 56 - Lognormal theoretical distribution fitting to the Service Times experimental distribution (Scenario II)
Figure 57 - Goodness of fit summary and descriptive statistics for the Service Times (Scenario II)