Managing airport operations using simulation

MS Fayez
1
*, A Kaylani
1
, D Cope
1
, N Rychlik
1
and M Mollaghasemi
2
1
Productivity Apex, Inc., Orlando, FL, USA; and
2
University of Central Florida, Orlando, FL, USA
Airport terminals have dramatically changed after September 11th, primarily due to the tightened security measures.
These changes had a major impact on passenger arrival patterns, passenger ows, space allocation, processing times,
and waiting times. In turn, it impacted a terminals performance, levels of service, and the overall passenger experience.
Airport planners and decision makers required a decision support tool that can quickly evaluate the impact of the often
changing security regulations and the decisions to counterpart these changes on the airports level of service. The
intellectual focus of this paper is to present the methodology and the generic tool that will quantify and assess passenger
ow in airport terminal functional areas and relate these requirements to the airports key performance indicators and
level of service.
Journal of Simulation (2008) 2, 4152. doi:10.1057/palgrave.jos.4250030
Keywords: simulation; modelling; Arena; airport; passenger ow; levels of service
1. Introduction
As any traveller can attest to, airline terminal operations
have changed drastically over the last decade. The most
visible of these changes has been the tightened security
measures adopted after September 11th and the subsequent
changes in security-checking procedures. However, many
other less visible changes have also had drastic effects on the
management and planning of terminals. Most notably is the
adoption of new ticketing technology, such as the addition
of e-ticketing capabilities, new ticketing rules, and remote
check-in options. Each of these changes has had a major
impact on passenger ows, space allocation, processing
times, waiting lines, and waiting times. In turn, it impacted
the terminals performance, levels of service (LOS), and the
overall passenger experience. For example, the deployment
of e-ticketing machines has signicantly decreased ticketing
times and pushed passengers more quickly to security, which
accordingly increased security checkpoint lines and waiting
times. Also, remote check-in has affected the arrival patterns
of passengers where passengers arrive much closer to ight
time than before, therefore changing queuing patterns across
the terminal functional areas.
With all these changes, it is difcult for terminal operators
to gauge the effects of these changes on the ow of
passengers through the terminal. In fact, it might be virtually
impossible for operators to predict how the system will be
affected when these changes occur simultaneously and at a
highly dynamic pace. As a result, terminal operators have
been placed in a reactive position that led to lower
performance, LOS, and physical space availability at a given
time. The situation will be worse if the many other changes
on the horizon are considered, such as the common-use
ticketing counters, known traveller programmes, and
partnerships with hotels and resorts that provide remote
check-in and baggage-handling services for passengers.
The need for predictive capabilities has become vital to the
decision-making process of airport terminal planners and
operators. In this paper an airport terminal decision support
system methodology is presented: a simulation environment
that will provide airport terminal operators with a decision
support tool and the capability to analyse the end-to-end
airport terminal operations. The tool will enable them to
quantitatively predict and compare the impact of the new
procedures, regulations, and their counterpart decisions on
the terminals performance, LOS, and the overall passenger
experience. In particular, the decision support tool and its
underlying methodology presented in this paper are intended
to provide terminal operators with the capability to assess
the terminals LOS performance measures such as the space
allocation per passenger at different functional areas, the
various passenger waiting times at different processing
points, and the availability of resources to handle various
services for passengers and their luggage.
In addition, the decision support tool can be used to
evaluate the impact of several factors on the terminal
operational performance. Factors can include passenger
arrival/departure patterns and trends, capacities, and avail-
ability of resources, and also, new technologies, new
regulatory and security requirements, natural disasters such
as hurricanes or ice storms, public transportation factors
*Correspondence: MS Fayez, 12689 Challenger Pkwy, Suite 130, Orlando,
FL 32828, USA.
E-mail: sfayez@productivityapex.com
Journal of Simulation (2008) 2, 4152 r 2008 Operational Research Society Ltd. All rights reserved. 1747-7778/08 $30.00
www.palgrave-journals.com/jos
that impact the terminal, key equipment failures, routine
maintenance schedules, and variability and randomness
inherent in the whole system.
2. Background
The intellectual focus presented in this paper was to develop
and apply scientically sound methodologies to quantify and
assess ow and related LOS in airport terminal functional
areas. Terminal capacity and space allocation in passenger
terminals are among the critical aspects of airport design as
well as operation. Its importance also increases as we
consider the LOS targets and ever-increasing passenger
volumes of the nations airports. LOS is a commonly
accepted metric for assessing the performance of airport
terminal functional areas under different scenarios (Davis
and Braaksama, 1987; Correia and Wirasinghe, 2004). It has
been developed for functional areas in which (or through
which) passengers will wait (or ow). Current LOS
standards use several levels; each level dictates a target
customer service level starting from excellent to unaccep-
table. Levels of service standards developed in the 1970s did
not consider the total time that the passenger requires or
utilizes a given space and were always designed for peak
hour operations (Correia and Wirasinghe, 2004). Addition-
ally, the standards did not address the subjective nature of
the passenger experience (eg how satised or comfortable did
the passenger feel during a particular segment of the terminal
experience). In general, airport planners specify two or three
target LOS, based on the period where higher LOS is
expected during normal periods, and lower LOS during
congested and peak hours. Airports utilize general guidelines
to determine capacities, allocate space, assign resources, and
decide on other aspects of terminal design towards target
LOS. These guidelines have been used for years, and
improvements in specic airport LOS were primarily
achieved through the diligent efforts of innovative and
forward-thinking airport planners and operations managers.
Although the majority of airports share a common
objective regarding the safe and efcient movement of
passengers via aviation services, airports are generally
different and in some instances unique in their passenger
terminal design. Airports can be generally classied accord-
ing to their size, measured in Million Annual Passengers,
number of terminals, and number of gates. This classica-
tion generally indicates the airports capacity for handling
passengers, baggage, aircrafts, and other supporting ele-
ments such as inter-modal transportation systems and
miscellaneous services offered. Airports can also be classied
according to the type of passengers the airport handles,
including origin and destination travellers or connecting
travellers. Airports generally have different layouts and
appearance, different settings for ticketing and security
checkpoints, different passenger processing and information
processing technologies deployed with varying levels of
maturity, and differing passenger in-terminal circulation
systems such as escalators and elevators.
These differences greatly contribute to the dynamic nature
of the interaction and relationship between passengers and
functional areas and between inter-dependent functional
areas. These differences will also greatly impact the LOS and
the allocation of capacity and passenger space. Therefore,
standards that were designed and targeted to t every airport
were at worst impractical and at best questionable due to
uncertainty arising out of site-specic levels of variability.
Because these guidelines were static, relied heavily on
averages, and did not take into consideration the stochastic
or the dynamic factors previously discussed, they could not
be adapted for every situation or airport. Therefore, targeted
overall LOS might not have been consistently achieved or at
best achieved within certain functional areas but not
achieved on others. The traditional way of one solution ts
all might be hampering airports from achieving their unique
best possible level of services economically.
3. Literature review
To provide a point of reference for the presented methodol-
ogy, it will be useful to briey discuss prior attempts
published in the literature to advance the practice regarding
terminal space allocation based on scientic methods. Some
of these practices take into consideration airport differences,
unique settings, and features, including the uctuation and
pattern in airport passenger ows through the terminal
functional areas. Some of the related research will be briey
discussed and contrasted to the approach we are adopting in
this paper.
The active research in this area started in the early 1960s
(Cox and Smith, 1961; Lee, 1966), when queuing models
were developed for enplanement passenger check-in resulting
in the determination of average service and wait times. This
study was useful in the design/redesign of efcient ticketing
functional areas, including optimum capacity and space
allocation. Subsequently other queuing models were devel-
oped and used to analyse not only ticketing areas but the
entire departure lounge in an airport terminal taking into
consideration economic implications (Mckelvey, 1988;
Odoni and Neufville, 1992a, b). Other similar studies
considered stochastic analysis and included other functional
areas such as corridors, seating, immigration, well wishers,
greeters, etc (De Neufville and Grillot, 1982). However, no
record was found for successful attempts to consider all the
functional areas or the end-to-end passenger experience, that
is, from airport entry to departure or from ight arrival to
exiting the airport.
In the 1990s other models were developed, considering
passenger socialization and next-logical movement patterns
to determine optimal space requirements. As discussed,
42 Journal of Simulation Vol. 2, No. 1
queuing theory is currently one of the most frequently used
analytical methods for space programming in passenger
terminals (see Gilliam, 1979). Queuing theory is, in essence,
the mathematical study of waiting lines (or queues). There
are several related processes, such as arriving at the back
of the queue, seizing physical space within the queue, waiting
in the queue, and being served by the server, for example
ticketing agent, at the front of the queue. One method to
solve queuing problems involves analytical methods which
make simplifying assumptions such as the following: arrival
times are random and the time between arrivals is distributed
exponentially, service times are also distributed exponen-
tially, the queue is of rst come rst served type, and there
are no signicant interdependencies, among other assump-
tions. These assumptions compromise the accuracy of the
results because the real-world situation is altered to
accommodate these assumptions. When decisions and/or
passenger space allocations are based solely on queuing
theory, there is a high probability of either lowering the
customer service levels due to a smaller space allocated per
passenger or incurring unnecessary cost due to the extra
space allocated for queuing areas. Hence, our approach
while not abandoning queuing theory in its totality will
instead utilize more advanced and appropriate methods. In
particular, we will use simulation modelling and analysis,
which will be discussed in more detail later in this section.
Another noteworthy approach previously utilized in
passenger terminal analysis is the timespace concept (Parizi
and Braaksma, 1992). Here the airport terminal functional
areas were considered as timespace zones where passengers
owed or waited and required different spaces, than they
currently occupy for a certain period of time. However, once
again timespace does not consider several critical variables,
and certain highly complex variables are totally ignored,
hindering the achievement of accurate results and thus the
real or full benets of the analysis undertaken (Benz, 1987).
Another approach, and the methodology we are embra-
cing, is the use of simulation modelling and analysis, which
provides a signicantly higher measure of accuracy for
studying passenger ow in an airport terminal. Contrasting
simulation modelling and other analytical methods, simula-
tion can be considered as one of the best-t solutions for
modelling and analysing a passenger terminal for capacity
analysis under varying situations, taking into consideration
LOS and several cost factors. Since simulation modelling
can, by nature, model deterministic as well as stochastic
systems, it takes into consideration the variability and the
dynamic features of the system modelled, allowing analysis
of the end-to-end passenger experience within the airport
terminal at a varying level of details. However, our early
results of using simulation were not as successful as one
might have expected. In retrospect it was determined that
since simulation is a highly advanced analytical method,
requiring special skills, early practitioners were not fully
schooled in the basics of the underlying sciences and
therefore the results reected the lack of understanding of
the rigour required. The input and output analysis requires a
signicant knowledge of statistics and experimental design.
From our perspective, and agreed upon by practitioners
within the domain, simulation modelling and analysis is a
very powerful tool, which when correctly used can be the
most practical, accurate, and useful alternative. Our simula-
tion methodology described in this paper provides a model
that generically mimics airport operations, including all the
functional areas. The model can be tailored to any airport
and can be populated with data for any period of time (ie
peak or non-peak). The simulation can be executed (run) for
any period from 1 h of operation to 1 year of operation,
selected days of the week, and selected periods of the year,
reporting, numerically and graphically, performance metrics
such as various LOS and space occupied. For example, a
certain functional area performance or LOS can be reported
for a given time of day, day of week, month, etc.
4. Airport terminal operations simulation using AirSim
The simulation methodology we are presenting in this paper
was implemented in an easy-to-use end-to-end airport
terminal simulation tool called AirSim. AirSim was devel-
oped by our team at Productivity Apex, Inc. to meet the
emerging needs of airport planners and operations man-
agers. AirSim is a generic airport simulation modelling and
analysis tool that can be used to analyse the current state of
the airports functional areas, key processing elements, and
their performance or LOS. It enables assessing and
predicting the LOS at different functional areas under
operational parameters and conditions such as changing
ight schedules, higher passenger trafc, routine mainte-
nance of functional areas, and variable resource schedules. It
can also consider atypical changes, such as the event-driven
implementation of new regulations, or new technologies that
drastically alter functional areas processing scheme or
passenger ow patterns.
Because of the highly variable nature of airport opera-
tions, it was determined that a exible and extensible discrete
event simulation model would be required to provide
meaningful information in a timely and efcient manner.
This understanding of the challenges facing airport
operators and the processes underlying this system acted as
the catalyst for the development of the tool AirSim. The
simulation logic modules and templates were the product of
extensive on-site observation and brainstorming sessions
with airport subject matter experts. First, conceptual ow
diagrams for the overall passenger ow and each of the
major processes were created and the key passenger
attributes for each process were identied. Then simulation
logic modules were developed to represent these processes
and to collect critical attributes and data from the user.
Modules were developed to model the items shown in
MS Fayez et alManaging airport operations using simulation 43
Table 1. (These modules were tested and validated in real
airport projects.) The modules described in Table 1 provide
the set of essential building blocks that can be used to
quickly build simulation models for a wide range of airport
congurations. Also, these modules provide the required
exibility to conduct what-if analysis for evaluating the
changes in regulations and passenger patterns, and the
actions taken to preserve and improve the level of service at
the airport terminals.
AirSim also includes modules and templates that model
the airport of the future, that is, remote checking from hotel,
home, etc, which result in a different passengers arrival
patterns to the airport. These modules have been tested and
validated in Orlando International Airport (OIA). OIA is
one of the rst to implement remote checking in conjunction
with Disney.
Creating these reusable, generic templates and modules
has greatly reduced the modelling effort, allowing the
modellers to rapidly construct discrete event simulation
models for any airport with any design or layout; a 2-week
project is possible. Existing CAD drawings of the focus
airports layout can be imported as a background and the
modules are placed in the appropriate areas to create an
accurate representation of the system. However, this is an
optional feature that might not be used unless it is necessary.
The second component of AirSim is a generic graphical user
interface (GUI). This GUI is designed to allow the user to
design what-if scenarios, populate and run the model, and
view results. The screenshot shown in Figure 1 provides a
representative version of the GUI.
The GUI is composed of two frames: the left frame is the
model browser and the right frame is the attributes and
the corresponding data of each operational scenario. Once
the different operational scenarios have been developed, the
GUI allows users to populate and run multiple scenarios
with a single press of a button.
Output reporting is enhanced as the output reports can be
customized to only show the LOS for a particular functional
area or the whole airport, eliminating the time-consuming
task of searching through the myriad of available statistics
provided. An example of the graphical output is shown in
Figures 2 and 3. Figure 2 shows the number of passengers
Table 1 Module and templates needed for constructing airport models
Module or templates Description
Enplanement passenger
creation
Creates the enplanement passengers (and accompanying well wishers) from a time-based arrival
schedule. Assigns initial passenger attributes such as group size, international/domestic, number
of bags, and servicing airline. This module allows for modelling passenger arrival behaviour
distribution (ie arrive time before ight).
Deplanement passenger
creation
Creates the deplanement passengers (and accompanying well wishers) from a time-based arrival
schedule. Assigns initial passenger attributes such as group size, number of bags, and servicing
airline. Models the deplanement processes including customs.
Ticketing areas and lobby
(Manned ticketing, electronic,
kiosks, and remote)
Contains logic of a generic ticket area (including common use) that might be composed of a
combination of e-ticket machines with manned or unmanned stations, and manned ticket
counters. Passengers are smartly routed to respective facilities based on passenger characteristics
(remote check in passengers bypass ticket counters, etc). This module also provides the ability to
model agent schedules.
Curbside The Curbside Logic represents the vehicular and passenger trafc ow throughout the curbside
areas. It also represents the curbside check in process for passengers.
Security screening checkpoints Generic representation of detailed security checkpoint composed of a combination of regular
lanes and employee lanes. Include preceding queuing area, individual queues, bag drop off,
scanning, and bag pick up. Each lane is modelled individually with unique characteristics such as
schedules, failure probabilities, and stafng.
Baggage claim Generic representation of baggage claim belt representing the process of passenger bag pick up.
EDS (Explosives Detection
system) machines for baggage
screening
Generic representation of detailed EDS machine units and groups. Each EDS machine unit can
be modelled individually or within a group with unique characteristics such as schedules, failure
probabilities, and stafng.
Airport access and ow Passengers are smartly routed (based on their characteristics) throughout the airport to mimic
real passenger behaviour. Includes, entrances, exists, horizontal ow (corridors, moving
sidewalks), vertical ow (elevators, escalators, stairways), waiting areas, and other services
(car rental service desks, food court, restrooms, etc).
44 Journal of Simulation Vol. 2, No. 1
waiting in the line of an airline ticketing over a full
operational day. On the other hand, Figure 3 shows the
sq. ft. per passenger in this particular ticketing area, where
the total area is 1000 sq. ft. Figures 4 and 5 display similar
information during the peak periods. Airport terminal
operators can predict each functional area LOS and
compare it to their target levels. Figure 5 shows the target
LOS lines pertaining to space allocation. This comparison
might result in questioning the total space allocated for this
particular area to sustain a target LOS; other alternative
decisions might be the increasing number of servers to
process passengers, which in turn will decrease the number
of passengers waiting at any given time. However, by using
the simulation environment, the decision makers can design
a set of experiments to evaluate several decisions pertaining
to these different scenarios and implement the decision that
brings the most operationally feasible solution.
As shown in Figures 3 and 5, the sq. ft. occupied per
passenger changes over time, which results in varying LOS
for passengers based on the time of day. AirSim can aid
decision makers to assess changes in passenger LOS as space
allocated to each functional area is changed. For instance,
one scenario could allocate more space for the check-in
period, for example 1500sq. ft. instead of 1000 sq. ft. as
shown in Figure 6. In this scenario, the space per passenger
with the 1500 sq. ft. is higher and the lowest space LOS
achieved is LOS B (20 sq. ft./passenger) versus the lowest
achieved with 1000 sq. ft. is LOS C (15 sq. ft./passenger).
AirSim will quantify that change per passenger and will
compare it to the LOS targets; also the total area can be
dened differently based on the time of the day to maintain a
target space LOS at any time of the day, that is, peak or
non-peak.
Parameters other than total area allocated can be
changed, for example by changing the ticketing agents
schedule during the peak period, that is, increasing the
number of agents will also enable processing more
passengers per hour, thus decreasing the number of
Figure 1 Screenshot of AirSim GUI.
MS Fayez et alManaging airport operations using simulation 45
passengers waiting, and consequently reducing space
requirement or increasing the space LOS if space remained
xed. The decision maker can run many scenarios and
generate the output side by side for comparison.
Simulation modelling will require data in order to be able
to populate and run the simulation model and analyse the
airport performance and LOS. Data should be collected to
sufcient depth to provide statistically valid sampling for
0
10
20
30
40
50
60
70
80
2
:
0
0

A
M
6
:
1
2

A
M
6
:
5
2

A
M
7
:
1
9

A
M
8
:
0
1

A
M
8
:
3
2

A
M
1
0
:
3
8

A
M
1
1
:
0
3

A
M
1
1
:
4
3

A
M
1
2
:
1
3

P
M
1
2
:
3
9

P
M
1
:
0
4

P
M
1
:
2
7

P
M
2
:
0
7

P
M
2
:
3
3

P
M
3
:
2
6

P
M
3
:
5
9

P
M
4
:
4
7

P
M
5
:
1
1

P
M
5
:
4
7

P
M
6
:
2
2

P
M
7
:
0
2

P
M
9
:
0
0

P
M
Time of the day
A
v
e
r
a
g
e

n
u
m
b
e
r

o
f

p
a
s
s
e
n
g
e
r
s

w
a
i
t
i
n
g

i
n

l
i
n
e
Figure 2 Number of passengers waiting in an airline ticketing line over a full operational day.
0
200
400
600
800
1000
1200
2
:
0
0

A
M
6
:
1
5

A
M
6
:
5
8

A
M
7
:
3
9

A
M
8
:
2
0

A
M
9
:
4
6

A
M
1
0
:
5
8

A
M
1
1
:
3
8

A
M
1
2
:
1
3

P
M
1
2
:
4
3

P
M
1
:
1
0

P
M
1
:
4
2

P
M
2
:
1
9

P
M
3
:
1
1

P
M
3
:
5
2

P
M
4
:
4
5

P
M
5
:
1
5

P
M
5
:
5
1

P
M
6
:
3
2

P
M
7
:
1
5

P
M
Time of the day
S
q
.

f
t
.

p
e
r

p
a
s
s
e
n
g
e
r
Figure 3 The sq. ft. per passenger in an airline ticketing line over a full operational day.
46 Journal of Simulation Vol. 2, No. 1
input to the simulation model. Based on our experience, the
data have to be collected during a busy time of the year
during a peak period at the terminal. At a minimum the data
should cover a full operational day for a sufcient number of
hours before the rst peak period through number of hours
after the last peak period. The peak periods should be
selected based on normal ight arrival and departure
schedules. However, it was proven that collecting data for
the full week renders more accurate results and analysis.
From a simulation and statistical perspective, it is more
0
10
20
30
40
50
60
70
80
1
0
:
4
6

A
M
1
0
:
5
8

A
M
1
1
:
0
8

A
M
1
1
:
2
7

A
M
1
1
:
4
3

A
M
1
1
:
5
6

A
M
1
2
:
0
9

P
M
1
2
:
2
1

P
M
1
2
:
3
1

P
M
1
2
:
4
2

P
M
1
2
:
5
4

P
M
1
:
0
3

P
M
1
:
1
1

P
M
1
:
1
8

P
M
1
:
3
4

P
M
1
:
5
2

P
M
2
:
0
8

P
M
2
:
1
7

P
M
2
:
3
0

P
M
2
:
4
8

P
M
3
:
1
3

P
M
3
:
3
0

P
M
3
:
4
5

P
M
Time of the day (Peak Period)
A
v
e
r
a
g
e

n
u
m
b
e
r

o
f

p
a
s
s
e
n
g
e
r
s

w
a
i
t
i
n
g

i
n

l
i
n
e
Figure 4 Average number of passengers waiting in an airline ticketing line during the peak period.
0
10
20
30
40
50
60
70
80
90
100
1
0
:
4
6

A
M
1
0
:
5
8

A
M
1
1
:
0
8

A
M
1
1
:
2
7

A
M
1
1
:
4
3

A
M
1
1
:
5
6

A
M
1
2
:
0
9

P
M
1
2
:
2
1

P
M
1
2
:
3
1

P
M
1
2
:
4
2

P
M
1
2
:
5
4

P
M
1
:
0
3

P
M
1
:
1
1

P
M
1
:
1
8

P
M
1
:
3
4

P
M
1
:
5
2

P
M
2
:
0
8

P
M
2
:
1
7

P
M
2
:
3
0

P
M
2
:
4
8

P
M
3
:
1
3

P
M
3
:
3
0

P
M
3
:
4
5

P
M
Time of the day (Peak Period)
S
q
.

f
t
.

p
e
r

p
a
s
s
e
n
g
e
r
LOS A
LOS B
LOS C
LOS D
LOS E
Figure 5 The sq. ft. per passenger in an airline ticketing line during the peak period with 1000 sq. ft. total area.
MS Fayez et alManaging airport operations using simulation 47
technically correct to collect over a week period, which
means a larger sample size. The data collected should
capture passenger ow (enplanement and deplanement),
functional areas processing times (ticketing, security screen-
ing, facility usage, etc), waiting times, passenger transfer
times between functional areas, capacities, resources sche-
dules (ticketing agents, number of functioning security
checkpoints, number of elevators, etc), functional areas
physical space (dimensions, layout, etc), horizontal transfer,
vertical transfer, and passenger behaviours (time passengers
arrive in advance of the ight, % passengers who check in
bags, % passengers who use e-ticketing, etc). Also, the data
0
10
20
30
40
50
60
70
80
90
100
1
0
:
4
6

A
M
1
0
:
5
8

A
M
1
1
:
0
8

A
M
1
1
:
2
7

A
M
1
1
:
4
3

A
M
1
1
:
5
6

A
M
1
2
:
0
9

P
M
1
2
:
2
1

P
M
1
2
:
3
1

P
M
1
2
:
4
2

P
M
1
2
:
5
4

P
M
1
:
0
3

P
M
1
:
1
1

P
M
1
:
1
8

P
M
1
:
3
4

P
M
1
:
5
2

P
M
2
:
0
8

P
M
2
:
1
7

P
M
2
:
3
0

P
M
2
:
4
8

P
M
3
:
1
3

P
M
3
:
3
0

P
M
3
:
4
5

P
M
Time of the day (Peak Period)
S
q
.

f
t
.

p
e
r

p
a
s
s
e
n
g
e
r
LOS A
LOS B
LOS C
LOS D
LOS E
Figure 6 The sq. ft. per passenger in an airline ticketing line during the peak period with 1500 sq. ft. total area.
Figure 7 Model animation for security checkpoint lanes.
48 Journal of Simulation Vol. 2, No. 1
should include feedback from passengers on their perception
of the time spent in a specic functional area and their
comfort level as related to the number of people in the
specic functional area.
The data should consider the different airport functional
areas for enplanement and deplanement passengers. For
enplanement, it should include curbside ticketing/curbside
check-in, ticketing, well wishers, security screening check-
points, horizontal and vertical circulation, and gate areas/
holding rooms. For deplanement, it should include hold
room exit, horizontal and vertical circulation, exit from
secured to non-secured areas, meeters and greeters, baggage
claim/carousel, and inner curb.
These data have to be collected according to a data
collection plan. The data collection plan will provide
a procedural approach for collecting the data, validating
the data, analysing the data, and transforming the data
to a suitable format for the simulation. The following
steps can be used to accomplish the data collection
process:
1. Interview airport operators and subject matter experts,
and collect preliminary data on airport design and layout,
busiest time of the year, peak periods, and other
information related to functional areas, circulation, etc.
2. Develop a list of data required and submit to airport
operators or subject matter experts.
3. If data are available from airport operators or subject
matter experts, go to step 5.
4. If data are not available, a facility visit is crucial, develop
data collection forms, assemble a data collection team,
and collect the data. The data will be collected using the
following techniques:
5. Time and motion study.
6. Questionnaires targeting passengers, airline managers,
security personnel, other key airport personnel.
7. Analyse the data, check for outliers, consistency check,
etc.
8. Model the data for simulation model input.
Data in dynamic systems, such as airports, are stochastic and
contain numerous sources of randomness. During data
collection, the project team should not only be concerned
with what data to collect, but how to collect it to capture the
randomness and to ensure statistical validity. Rigorous
statistical methods should be applied to the data to ensure
the selection of valid input modelsthat is, probability
distributions that accurately mimic the behaviour of the
random input processes driving the airport under study.
Examples of sources of randomness are the inter-arrival rate
of passengers to the airport, check-in method selected,
ticketing times at various counters, security check time, and
walking times between functional areas. Once the input data
are collected, analysed, and validated from a statistical point
of view, it is used to populate AirSim through the GUI and
will run simulation models for various scenarios for a given
airport. The output from the simulation model will be
analysed to draw conclusions for each functional area in the
airport under study. This analysis will produce tables and
graphs related to the airport performance and LOS, similar
to the graph shown in Figure 5 and 6. Several simulation
scenarios will be developed to provide more in-depth
analysis and insights on airport performance and the
different functional areas LOS under several settings. The
next section will describe a detailed case study of implement-
ing the methodology in an airport.
5. Customizing AirSim for an airport
The primary objective of this effort was to provide airport
ofcials with a tool that would allow them to predict the
effect that changes such as the increased passenger volume,
the implementation of new programmes such as those
discussed in the previous section, and changing security
processes will have on passenger ow within the terminal. In
order to customize AirSim to meet this objective, the rst
step was to determine the appropriate level of details and
scope of the model. Based on information solicited from the
customer, the following airport areas were identied as
critical:
Security CheckpointsDue to the continuously changing
nature of the TSAs (Transportation Security Adminis-
tration) security policies, it was critical that the tool
provide enough detail to allow for the sub-processes
within security, the number of lanes, the schedules for
each lane, and the types of passengers served by each lane
to be dynamic.
Ticket Stations and LobbyThis was an area of great
interest to OIA because of the effect that technological
advances such as remote check-in and policy changes such
as common use ticketing counters have had and will
continue to have on the processes associated with
checking in passengers. Because the airlines make the
determination to adopt new initiatives or technologies, it
was decided that the model must provide an adequate
level of detail and enough exibility to specify these
parameters for each airline.
Baggage ScreeningIt was determined that the model
had to provide the capability to model not only the typical
baggage handling ow for each airline but must be able to
accommodate the newly introduced EDS (Explosives
Detection System) machines as well as atypical baggage
ows because of special programmes like the Disney
Magical Express.
Landside Terminal CurbsIn order to provide insight into
and allow for better management of the ever increasing
trafc ows surrounding the airport, the model was
designed to include the curbside pickup and dropoff of
MS Fayez et alManaging airport operations using simulation 49
passengers by both private and commercial vehicles. Users
can manipulate parameters in the curb such as arrival
patterns, pickup/dropoff times and curbside resources.
Based on these requirements, a model was developed
utilizing the AirSim simulation logic modules along with a
minimal number of standard Arena logic modules to
accurately model the ow of passengers through the
system. Furthermore, model animation was developed
(see Figure 7 for an example) to reect the level of detail in
the simulation model. The animation was developed to
Figure 8 Waiting time in security under various operational scenarios.
Figure 9 Number of passengers in the security queue under various operational scenarios.
50 Journal of Simulation Vol. 2, No. 1
accurately represent the terminal design, including ticket-
ing station, security checkpoints, elevators, people
movers, facilities, etc, and allow airport authorities to
visualize the passenger ow through the terminal as well
as evaluate it numerically and graphically based on the
model outputs.
Once completed, the customized model was veried and
validated by comparing the results for specic dates to video
footage of various queues throughout the airport on those
dates as well as detailed scrutiny by subject matter experts.
The model is being used to support day-to-day operations
to identify potential bottlenecks based on scheduled
passengers and resources, and is also being used to provide
information for larger strategic decisions. OIA recently went
through renegotiations with Disney regarding the future of
the Disney Magical Express, and the model was used to
evaluate the effect that the increased passenger loads would
have on queues and congestion of the terminal side where
Disney operates. The results of this study helped OIA and
Disney in developing a timeline that Disney would need to
move its operations from the A-side to the B-side of the
terminal in order to accommodate the projected passenger
increases as well as moving one of the sub-processes that
acted as a bottleneck from the rst oor to the second in
order to reduce congestion and increase passenger ow.
The model is also being used to evaluate what effect the
addition of private baggage handlers will have on the
security checkpoint queues. The addition of these baggage
handlers could potentially free some TSA employees to assist
with security checkpoints, thereby decreasing wait times. In
short, AirSim has provided the airport authority at OIA
with the ability for the rst time to visualize and quantify the
effects that changes to processes within the system will have
on passenger ow.
6. Case study
In this section, the simulation methodology will be
implemented for OIA. The implementation was developed
in a multi-phased project over the course of 3 years. The
OIA ranked 24th in the world and 14th in the US in 2004 in
terms of passenger volume, with roughly 31 million
passengers served. OIA is one of the fastest growing airports
in the nation, with passenger volume expected to increase
dramatically in the coming years. Unlike most of the other
major airports, OIA is primarily an origindestination
airport with only about 7% of the total volume made up
of transit passengers. OIA also has a number of special
programmes in place that make its operations unique. It
serves as the site for a pilot study of the known traveller
programme, in which biometric verication is used to
identify frequent travellers in an effort to reduce security
checkpoint processing times. OIA has also partnered with
Disney to provide the Disney Magical Express Service,
which provides free airport shuttle service, luggage delivery,
and airline check-in for Disney hotel guests. Due to the
success of this programme, a number of other resorts are
considering the implementation of similar programmes.
Because of this anticipated growth as well as the adoption
of programmes and services like these, that OIA recognized
the need for a decision support tool that would allow airport
authorities to proactively plan for these changes.
As the national security measures progress, the security
check procedures conducted by TSA change as well.
Planners and operators nd it very difcult to keep the
LOS within their pre-set target levels. OIA planners used the
simulation tool AirSim to re-plan and re-allocate resources
in the security check areas to be able to maintain the LOS at
the target levels. The airport authority believed that the
waiting time in the queue and the number of passengers
waiting in the queue will be signicantly altered with the
new security procedures implemented. AirSim was used
to evaluate and quantify the two security procedures.
The management at OIA dened several operational
scenarios to be simulated for the security checkpoints. These
scenarios are:
The baseline scenario (current state).
The new security procedures scenario.
Opening one more security lane.
Opening two more security lanes.
The varying parameters between the different scenarios and
the output key performance indicators are shown in Table 2.
The output includes: the average waiting time in the queue
and the number waiting in the queue. The security check time
data were collected and tted into an empirical continuous
probability distribution (designated by CONT(X)).
Table 2 Simulation scenarios
# Scenario # Security
lanes
Security check time Output: Average
waiting time in security
lines (min)
Output: Average
number of passengers
waiting in queue
1 Baseline 6 CONT(X) 49.24 600
2 New security procedure 6 CONT(X) TRIA (0, 0.2, 0.4) 111.08 1235
3 One extra security lane 7 CONT(X) TRIA (0, 0.2, 0.4) 60.39 715
4 Two extra security lanes 8 CONT(X) TRIA (0, 0.2, 0.4) 29.20 367
MS Fayez et alManaging airport operations using simulation 51
A graphical comparison for the waiting time in the
security check queues and the number of passengers waiting
in the queue over time of the four scenarios are shown in
Figures 8 and 9.
As shown in Figures 8 and 9, the new security procedure
signicantly alters the waiting time in the queue and the
number of passengers waiting. Adding an extra security
checkpoint will position the LOS closer to the baseline; this
scenario should be marked as the least acceptable scenario
when scheduling security agents and planning for the
queuing area. Adding two extra security checkpoints will
even make the LOS levels better than baseline levels.
6. Conclusion
In this paper we presented a simulation methodology that
assists airport planners and decision makers to analyse
airport terminals. The simulation methodology is delivered
in a software tool called AirSim that can be used by non-
simulation experts. The tool is a one-stop interface, where
the user can input data, create scenarios, design experiments,
dene required output, dene output charts, and run the
simulation experiments. The simulation methodology aims
at enabling airport planners and managers to measure and
predict key performance indicators related to airport
terminal operations, in terms of efciency and customer
service level.
AirSim is a generic tool that can be used to model and
simulate a wide range of airport congurations; it was
designed to be easily customized and implemented to a
particular airport. In AirSim, the simulation logic was
dened in modules; each module represents a function at the
airport. Using the methodology and modules presented in
this paper, the development time to implement a simulation
model for a given airport is signicantly shortened.
The tool presented was implemented in OIA, and has been
successfully used for several years by the airports planners.
To demonstrate the value of the methodology, we presented
several scenarios in this paper. These scenarios were built to
evaluate the airport LOS: in particular, the waiting time of
passengers in security lines, the number waiting in the queue,
and the space allocated per passenger.
References
Benz GP (1987). Application of the timespace concept to a
transportation terminal waiting and circulation area. Transport
Res Rec 1054: 1622.
Correia A and Wirasinghe SC (2004). Evaluating level of service at
airport passenger terminals: Review of research approaches.
Transport Res Rec 1888: 16.
Cox DR and Smith WL (1961). Queues. John Wiley & Sons: New
York, 180p.
Davis D and Braaksama JP (1987). Level-of-service standards for
platooning pedestrians in transportation terminals. ITE J 57:
3135.
De Neufville R and Grillot M (1982). Design of pedestrian
space in airport terminals. J Transport Eng 108: 87102.
Gilliam RR (1979). An application of queuing theory to airport
passenger security screening. Interfaces 9: 117123.
Lee AM (1966). Applied Queuing Theory. St. Martins Press
Company: New York.
Mckelvey FX (1988). Use of analytical queuing model for airport
terminal design. Transport Res Rec 1199: 411.
Odoni AR and De Neufville R (1992a). Passenger terminal design.
Transport Res Part A: Pol Pract 26(1): 2735.
Odoni AR and De Neufville R (1992b). Walking distance
minimization for airport terminal congurations. Transport
Res Part A: Pol Pract 26(1): 5974.
Parizi MS and Braaksma JP (1992). Dynamic capacity of airport
enplaning curbside areas. Transportation Research Record
No. 1373, Airport Landside Planning and Operations,
pp 816.
Received 7 May 2007;
accepted 24 September 2007
52 Journal of Simulation Vol. 2, No. 1