24th INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES

RUNWAY CAPACITY PLANNING SUPPORTED BY DYNAMIC

PROGRAMMING
R.A.A. Wijnen
Faculty of Aerospace Engineering
Delft University of Technology
Keywords: Strategic Airport Planning, Dynamic Programming

Abstract

Airport planning nowadays is becoming more

and more difficult because of the increasing complexity of the airport system and the different and
often conflicting objectives of the stakeholders
involved in the planning process. TU Delft Airport Development Center (TUD-ADC) is therefore working on decision-support systems that
enable decision advisers to conduct an analysis
of the entire airport system.
One of those systems is the Airport Business
Suite that provides a user with a user interface
for analyzing the airport operation and business
in order to match future demand and capacity.
Another system that is currently being developed is the Holistic Airport Resource Management and Optimization System (HARMOS),
which has more or less the same goal as ABS
but is based on another philosophy. HARMOS
employs mathematical optimization techniques
in order to support the airport planning process
more efficiently.
One of the problems that a decision support
system for airport planning should be able to
address is the runway expansion problem, i.e.
where and when to expand the runway system.
This paper shows how both ABS and HARMOS
could be used for analyzing this problem. Due
to the status of HARMOS software development
part of the analysis will only be described conceptually 1 .
1 At

the moment of writing the final results could not

The ABS analysis process of the runway expansion problem makes clear that ABS flexibility is
kind of a disadvantage in this case. A lot of user
input is required for setting up the so-called peak
day and yearly cases in ABS. Besides that, the
user has to interact heavily with the ABS in order
to run these cases, evaluate and post-process the
results. Only after having done all of that, a user
could start making comparisons between different runway expansion plans.
The philosophy of HARMOS enables a user
to focus more on the problem at hand and the
users objective, i.e. what would be the best (in
terms of costs) investment plan for runway expansion. By turning the runway expansion problem into a dynamic programming problem, the
analysis and comparison process can be highly
automated which makes identification of the best
plan very efficient.
1 Introduction
Airports these days are more and more becoming a business. In order to survive an airport
must operate efficiently under a set of boundary
conditions imposed by (inter)national authorities.
Therefore, an airport must develop strategies in
order to stay in business in the near and far future. However, due to the increased complexity
of the airport system and its environment, developing strategies is not that easy. Apart from the
yet be produced; therefore these will be presented at the
conference itself.

R.A.A. WIJNEN

fact that future conditions for the airport are unknown, a multitude of aspects has to be considered. An airport decision maker also has to acquire insight into the business process and objectives of the other airport stakeholders, such as the
airlines and air traffic service provider, in order
to asses the impact and acceptability of his/her
strategic plans.
Current practice when developing strategic
plans is to (1) model the current airport operation (needed as a baseline for making comparisons) (2) conduct a study that forecasts future
conditions, (3) make an estimation of an airports
future performance (qualitatively and/or quantitatively), (4) identify the bottlenecks in the airport system, (5) define a strategic plan, i.e. which
policies are used and what investments need to
be planned, that deals with the bottlenecks, and
finally (6) evaluate the effectiveness of the strategic plan. After that, the plan has to be discussed
with the other airport stakeholders in order to
reach consensus about the proposed changes to
the airport system, which might result in repeating (part of) the process described above. This
process involves a huge amount of work since
every part of the job is done by different parties
and spans a considerable time. This motivates
the need for decision-support systems for airport
planning.
TUD-ADC is therefore working on decisionsupport systems for airport strategic planning.
At the moment, TUD-ADC is developing two
different types of decision support systems,
namely the Airport Business Suite (ABS) which
is based on the concept of non-linear navigation
and HARMOS which uses mathematical optimization techniques.
The objective of this paper is to describe how
both decision support systems are used for analyzing plans for runway investments.
Runway expansion of Amsterdam Airport
Schiphol (AAS) will be used as an example, since
just recently a limited cost-benefit analysis [1]
has been conducted that investigated the social
benefits for expanding the airport with a sixth
and/or seventh runway. It appeared that, although

Amsterdam Airport Schiphol has opened a new

runway early 2003, total runway capacity will not
be sufficient to match future traffic demand.
The contents of this paper is as following.
Section 2 first gives some more information
about the cost-benefit analysis just mentioned; after that the problem definition that is used in this
study is presented. Section 3 presents the Airport Business Suite and shows how it can be used
for analyzing runway expansion plans. In section
4 the philosophy and design of HARMOS is introduced. The other part of this section is used
to describe the analysis of the runway expansion
problem with HARMOS. The problem analysis
itself is presented in section 5. Conclusion and
recommendations are presented in section 6.
2 Problem Description
This section gives a detailed description of the
problem. First some background information is
given, followed by a detailed problem statement
and a description of the scenarios that have been
used in the analysis.
2.1 Background Information
Because preliminary studies for AAS showed
that the new runway system (called 5P) is not capable of handling future traffic demand, the possibilities for expanding the runway system with
a sixth or seventh runway are to be investigated.
Three variants for runway expansion, depicted in
Figure 1, have been investigated in [1] and will
also be used here. The first variant, called 6PK,
is a runway system with a total of 6 runways with
an additional runway parallel to the existing 0624 runway (Kaagbaan); the second, called 6P, has
also 6 runways with an additional runway located
in between the 01L-19R (Zwanenburgbaan) and
recently added 18-36 runway (Polderbaan); and
the third, called 7PK, has a total of 7 runways
where both runways from the other two runway
systems are present.

Runway Capacity Planning supported by Dynamic Programming

course on the decision that was made in the previous period. If the previous decision was (1) then
there are four possibilities, namely:
1. Keep the current runway system (5P);
2. Add a runway parallel to the Kaagbaan
(6PK);
Figure 1: Variants for runway expansion

2.2 Problem Statement

When traffic demand increases, an airport needs
to take action to make sure that service levels
dont decrease. One such actions is expanding
the airside capacity in terms of runways. In order to do this efficiently and economically, two
issues have to be dealt with, namely, (1) where
to put the new runway and, (2) when should the
runway(s) become operational. The first question
is related to noise, costs and operational considerations; the second is related to expected economic growth and the associated growth in traffic
demand. The airport management needs to deal
with both issues such that investments in runway
capacity are timely conducted so that future traffic demand is matched with capacity at minimum
costs and acceptable noise impact.
So, for the runway capacity planning, a
decision-maker has to figure out which runway
should be build at what time for a given planning
period. Letss assume that there are three distinct
periods in time where such decisions should be
made. For the first period (2003-2010), there are
three possibilities, namely:
1. Keep the current runway system (5P);
2. Add a runway parallel to the Kaagbaan
(6PK);
3. Add a runway in north-south direction
(6P).
For the second period (2010-2020), there are a
different number of possibilities, depending of

3. Add a runway in north-south direction

(6P);
4. Add both runways simultaneously (7PK).
If decision (2) had been taken in the first period,
there are two possibilities, namely:
1. Keep the runway system (6PK);
2. Add a runway in north-south direction
(7PK).
If decision (3) had been taken in the first period,
the possibilities are:
1. Keep the runway system (6P);
2. Add a runway parallel to the Kaagbaan
(7PK).
For the third and final period (2020-2030), the
process is similar. A overview of the decision
tree just described is given in Figure 2.
As can be seen from Figure 2, there are a lot
of possibilities for runway expansion. All these
possibilities or investment plans need to be evaluated. After this evaluation, the best plan can be
selected for implementation. The next two sections, 3 and 4 will therefore describe how decision support systems could be used for this evaluation process. Before that, first the scenario that
will be used as input for the evaluation process
will be discussed.
2.3 Scenarios
Uncertain future conditions have been taken into
account by using three different scenarios that describe economic growth and technological development at a different pace.
3

R.A.A. WIJNEN

CAPACITY
# flights
per hour
7PK

7PK

6P

6P

6P

6PK

6PK

6PK

5P

5P

5P

150

130

123

5P
105

TIME
2003

2010

2030

2020

Figure 2: Alternative investment plans for runway capacity expansion

riving and departing flights is presented graphically in 3.

120

departures
arrivals

100
80
60
number of flights

The planning period that is considered spans

about thirty years, from 2003 until 2030. During
this period three distinct times, i.e. 2003, 2010,
and 2020, for making decisions about investments in runway capacity have been set. So, airport performance needs to be evaluated in 2003,
2010, 2020 and 2030. For this performance evaluation, forecasts of economic and technological
developments (collectively called a scenario) is
needed.

40
20
0
-20
-40
-60
-80

2.3.1 Economic Growth

-100
-120

Future traffic demand depends on developments

in economic growth, which are rather unpredictable. Therefore three different economic
growth scenarios will be used as input for the
planning problem, ranging from limited (2 %) to
above average (7 %) growth of traffic demand.
The economic growth percentage is used to
extrapolate a flight schedule for the 20th peak
day at Schiphol. This flight schedule has been
extracted from the time table for the summer of
2003, which was published by Schiphol at their
website. The baseline demand distribution for ar-

10

12

14

16

18

20

22

time [hr]

Figure 3: Graphical representation of converted flight

schedule

Figure 3 clearly shows the peaks in the demand

distribution for the arriving and departing flights.
These are the result of the system of flight banks
that is used by the hub carrier, Royal Dutch
Airlines (KLM). The total number of flights
amounts to 1050 for this particular day.

Runway Capacity Planning supported by Dynamic Programming

2.3.2 Technological Developments

The next decades, new technology for ATC/ATM
ground systems as well as airborne equipment
will become available. As a result, airspace
and airport capacity will increase. New ground
tools will enable an air traffic controller to handle more aircraft in a sector than is possible today. Improved radar technology and navigation
equipment will increase tracking accuracy and
decrease navigational errors. As a consequence,
separation standards can be adjusted. Introduction of new technologies makes it possible to
loosen the separation standards while maintaining the same safety level.
For this study it is not exactly required to
know what technology will be introduced in the
future. It is just assumed that new technology
will be introduced that will loosen current separation standards. For each time period in the scenario, a different separation standard matrix will
therefore be specified.
3 Airport Business Suite
This section first describes the Airport Business
Suite and its philosophy; after that a brief description on using the ABS is given.
3.1 Overview
The Airport Business Suite [5] has been designed
as a set of integrated tools that allow the information needed for strategic decision-making to be
synthesized in an efficient, effective, and consistent manner.
The only part of the ABS that will be discussed here in some more detail is the model system of ABS. A graphical representation is shown
in Figure 4. The set of models to be included in
the ABS resulted from a division of the airport
business process, which resulted in five sets of
models.
3.1.1 Model 1: Demand for airport capacity
The first model generates a flight schedule based
on present-day flight schedule data or manual input. A future flight schedule is created by select-

Figure 4: The ABS Model System

ing a forecast year and an economic growth percentage.

3.1.2 Model 2: Supply of airport capacity
Capacity and delay are two of the principal measures of performance of an airport system. To assess these important performance characteristics,
a comprehensive set of tools, covering all relevant airside and landside elements of the airport
system, has been implemented as modules within
Model 2 of the ABS.
For capacity calculations the Federal Aviation Administration (FAA) airfield capacity
model has been included. A network of queuing modules is implemented for predicting delays
due to congestion in the various airport elements.
The ABS Model 2 also incorporates a terminal capacity module but since the focus of this
study is on runway capacity, the terminal model
wont be discussed any further.
ABS also enables a user to fairly automatically generate input for the Integrated Noise
Model (INM), a noise contour calculation program.
3.1.3 Model 3: Matching demand and supply
The output provided by Model 2 is of critical importance for the ABS user. Often the initial output of Model 2 will not show satisfactory performance. For example, capacity or quality limitations may be exceeded, preventing the airport
from accommodating the desired demand. However, Model 3 lets the user tune the airports performance by altering the desired flight schedule,
5

R.A.A. WIJNEN

the airports design, and/or the airports operations. The function of Model 3 is to provide the
tools to enable the user to conduct this tuning
process. Through the ABS flexible Graphical
User Interface (GUI), the airport strategic planner
can test a wide range of alternative approaches to
matching supply with demand.
3.1.4 Model 4: Airport turnover
This sub model (nr. 4) estimates airport revenues
and costs, and - being the difference between the
two - the profits (or losses). At this moment, this
model is reimplemented to make it more generic.
3.1.5 Model 5: Investments and operational
costs
In a capital-intensive business like an airport, investments play a central role. Investments are
therefore covered by this sub model. Just like
the previous model, this model is being reimplemented.
3.2 Using the ABS
An overview of the analysis process with the
ABS is presented in this section. The ABS is
based on working with two type of cases, a socalled peak day case and a yearly case.

Peakday case This is a collection of data related to one single day. For this particular day a
flight schedule, runway use scheme (i.e. a specification of how the runway configurations are
used), and weather type needs to be specified. A
user has to prepare this data beforehand by using the one of the editors that the ABS GUI has
available. After doing this, the appropriate data
can be selected from drop down lists as shown in
Figure 5. When everything is set up, a peak day
case can be run. By running a peak day case, capacity and delay computations are performed for
the entire day. Finally, one can evaluate the output by viewing several graphs.
Yearly case This is a collection of peak day
cases and it is used to evaluate the annual airport performance (such as noise impact, capacity
coverage). A yearly case therefore describes how
each of the runway configurations defined in the
peak day case are used throughout the year. For
each of the peak day cases (i.e. runway configurations) the usage percentage needs to be specified. The runway configurations and the associated usage percentages are used to generate an
input file for the INM so that the noise impact can
be evaluated.
4 HARMOS
Here, the HARMOS decision support system will
be presented. After that the analysis of the runway expansion problem with HARMOS will be
described.
4.1 Overview

Figure 5: Selecting datasets via the ABS GUI

The key element of HARMOS is the employment

of mathematical optimization techniques for the
determination of the most efficient policies for
(1) management of the airport operation and (2)
allocation of resources (both financial and infrastructural). A primary aim is to make the system
truly easy-to-use and robust for the decisionmakers actually relying on it in strategic airport
planning studies. The easy-to-use requirement
also entails that for using the decision-support
6

Runway Capacity Planning supported by Dynamic Programming

system no specific knowledge or training with respect to the employed optimization techniques is
necessary.
4.1.1 Functional description
The HARMOS system can best be explained by
discussing the Use Case Diagram in Figure 6,
which displays the functionality of the system at
the highest level.

3. Decision adviser: this is the person that

advises the people that make the actual decisions. As advisers they need to be able
to test different strategies (defined by their
decision makers or themselves) for developing the airport for different futures. They
also need to be able to clearly present different strategic plans to the decision makers so that the Decision Maker (DM) can
make comparisons.
Seven different actions are to be performed by the
HARMOS system, i.e.:
1. Setup and customize system: the development team needs to be able to customize
HARMOS for a specific user (stakeholder)
based on their requirements;

Figure 6: HARMOS: High-level Use Case Diagram

Figure 6 shows the HARMOS system, its users,

and the actions that it should be able to perform.
Three type of users have been identified, namely:
1. Development team: these are the people
that are developing the HARMOS software
(thats us). HARMOS will consist of a
generic part and a customized part. An
airport or other stakeholder that decides to
use HARMOS for strategic planning will
most likely need user-specific functionality. Therefore the system should be build
such that functionality for customizing the
software is included;
2. Expert: this is a person that has specific
knowledge of (a part of) the airport system.
This user will interact the most heavily
with the system since experts are needed
for populating the system with data, tuning
of models, and validating the various computations done by the system;

2. Setup and model airport: an (domain) expert or a team of experts needs to be able to
setup and model the airport. i.e. data collected about the physical layout of the airport and the operations at the airport should
be stored in and managed by HARMOS;
3. Evaluate current airport characteristics
and performance: HARMOS should provide the users with an overview of the airport characteristics and its performance.
This action is a quick way of getting an
overview of the airport layout, operations
and performance. It is used by both the experts and the decision advisers but in different ways. Domain experts will probably
only focus on that part of the airport system (detailed look) that is related to their
expertise. Decision advisers will need to
get a general overview of the system and
its performance (birds view);
4. Define and implement policies: strategic
planning involves the exploration of different policies and their associated effects on
the system. It is therefore essential that
HARMOS provides functionality to define
various different policies;
7

R.A.A. WIJNEN

5. Define and Setup Investments: another

aspect of of strategic planning is making
the proper investments at the correct time.
This use case provides this functionality;
6. Setup and evaluate (part of) strategic
plan: a strategic plan consist of a set
of policies to be implemented for the future together with investments in the airports infrastructure. Given a certain scenario (economical developments, etc.), the
strategic plan will result in a certain future performance. A user of HARMOS
should therefore have the functionality to
setup and evaluate different strategic plans;
7. Present different cases to DM and get
feedback: depending on the combination
of the scenario and the strategic plan, the
future airport will look and perform different. Decision makers need to determine
how to go forward with the airport development. Therefore the decision adviser needs
to be able to present him or her with a clear
overview of the combinations of strategic
plans and future scenarios that have been
setup and evaluated during action number
6). A decision maker will then be able
to select the best way forward based on
his/her feeling about a certain future scenario.
4.1.2 Software implementation
HARMOS itself is implemented in Python using
wxPython for the GUI. Python is an object oriented language for which a lot of packages for
interfacing with other code already exist. Examples are:
The python COM package which makes it
possible to control and run Microsoft COM
objects (also known as Active X components);
Packages for connecting to a Database
Management System (DBMS). The package for dealing with the Postgresql DBMS

is used by HARMOS, since the Postgresql

system is used for storing airport data;
Packages for running C or Fortran code.
These are e.g. used for running lp_solve 2 ,
which is linear programming code written
in C . The F2Py package will be used for
running our NOISSH tool ([6], [4], [3]),
which is Fortran code for optimizing noise
abatement procedures.
4.2 Using HARMOS
At the moment of writing this paper, development of HARMOS is an ongoing activity. Currently, HARMOS is able to run and control the
input parameters of the ABS model 2. This has
been accomplished by using the python COM
package as interface to the ABS model 2 (ABS
has a modular design so all sub models have been
implemented as COM objects).
Like already mentioned, the philosophy of
HARMOS is to use mathematical optimization
to enable a more efficient analysis. How this actually works for the runway capacity expansion
problem will now be explained.
4.2.1 Dynamic Programming
The problem of determining the best plan
for investments in the runway system can be
turned into a dynamic programming problem [2].
Graphically, this is presented in Figure 2, where
the various options for expansion are shown on
a time line for four discrete times, namely 2003,
2010, 2020 and 2030. The vertical axis denotes
the average capacity of the runway system in the
number of flights per hour. For each moment in
time (stage), a decision has to be made about expanding the runway system or not. If one decides
to expand the runway system, the location of the
new runway has also to be chosen. For Amsterdam Airport Schiphol, this gives a total of 13 investment plans based on the three options given
in Figure 1.
2 http://groups.yahoo.com/group/lp_

solve/

Runway Capacity Planning supported by Dynamic Programming

An investment plan will be evaluated in terms

of total costs, i.e. the sum of the delay costs and
capacity costs. The best plan is the one that has
the lowest total costs over the entire planning period (2003-2020). So, in order to be able to determine the optimal investment plan, the costs of
each the 13 investment plans need to be determined in order to evaluate the first objective of
the optimization problem.
4.2.2 Objective evaluation
The airport management would like to plan their
investments in the period 2003-2020 such that
total costs are minimized. These costs are considered to be the sum of delay costs and capacity costs. Delay costs are directly related to the
(mis)match between capacity and demand. If demand exceeds capacity, delay will be incurred.
The average annual delay depends on the way the
runway system can and will be used throughout
the year. Capacity costs are the cost of investment capital and the operational and maintenance
cost. HARMOS will include modules for evaluating these costs.

5.1.1 Setting up projects and peak day case

Four ABS projects (called 5P, 6P, 6PK and 7PK)
have been defined, one project for each runway
system. For each of those projects a number of
peak day cases respectively yearly cases need to
be defined. In order to cover the runway configurations that are used throughout the year, eight
peak day cases have been defined. These peak
day cases describe how the runways of that particular runway system are used (i.e. the runway
configuration).
5.1.2 Running a peak day case
After setting up the case by creating and selecting the appropriate datasets (see also section 3.1),
a case can be run. Running a case will start the
computation of the capacity and delay for the airside and landside. When it is finished, the results
can be evaluated by looking at several graphs.
Here, the capacity and demand graphs for the
arrivals and departures are the most interesting
since these provide information about the average
delay. One of those (many!) graphs is presented
in Figure 7.

5 Problem analysis
This section describes how ABS and HARMOS
are actually used for analyzing the runway expansion problem.
5.1 Using the ABS
In ABS the following steps for providing quantitative data needed for evaluating the runway expansion problem, need to be taken:
1. Setting up projects and peak day cases;
2. Running peak day cases and evaluating its
output;
Figure 7: Example of peak day case output

3. Setting up yearly cases;

The following subsections will describe each of
these steps in more detail.

The graph shows the demand for either arrivals or departures in the form of bars per 15
minutes. The runway capacity is plotted with a
9

R.A.A. WIJNEN

red line. The average delay is indicated as text

in the lower left corner of the graph. Note, that
for every case, the average delay for the departures and arrivals needs to be collected for postprocessing.
5.1.3 Setting up yearly cases
The next step is using the previously setup and
processed peak day cases, to build a number
of yearly cases since here the interest is in annual performance characteristics, like the annual
average delay and noise impact. The yearly
cases specify how those runway configurations
are used throughout the year. For the 5P project
four yearly cases are needed, since the 5P system
needs to be evaluated for the year 2003, 2010,
2020 and 2030. For the 6P and 6PK project three
yearly cases are needed, since the 6P and 6PK
system needs to be evaluated for the years 2010,
2020 and 2030. For the 7PK project, two yearly
cases are needed since the 7PK system needs to
be evaluated for the years 2020 and 2030.
For each of those yearly cases, the usage percentages for each of the runway configurations
(defined in the peak day case) need to be assigned.
5.1.4 Conclusion
The previous sections described how an ABS
project is setup, populated with peak day cases
and yearly cases. Here, the baseline project (5P)
has been used as an example. Figure 8 shows a
screen shot of ABS with this project opened. The
left side of this figure shows the project tree for
this project, containing 24 peak day cases (8 peak
day cases per yearly case) and four yearly cases
(2003, 2010, 2020, 2030). Each of these cases
needs to be run and after that the output needs
to be evaluated and collected for post-processing
and comparison.
In order to account for technological developments, the correct input data for the capacity
model also needs to be specified before running
a case. In the current version of the ABS, this
needs to be done manually by editing a text file
that specifies the separation standards to be used.

Since there are three variants for runway expansion, three additional ABS projects (6P, 6PK,
7PK) have been setup in order to be able to analyze these expansion options. Although tools
for capacity, delay, and noise impact are integrated in the ABS, the activity of setting up the
projects, analyzing and comparing all the results
is an elaborate task. Note that comparing the different runway investment plans (delay costs versus capacity costs) did not even start yet.

Figure 8: Screen shot of Airport Business Suite

The creation of yearly cases in each project requires a user to interact heavily with the GUI.
Another thing that adds to user workload is the
fact that for each yearly case, peak day cases with
the appropriate data need to be created. Therefore, it might be beneficial to try to automate
some of the above mentioned tasks. That is exactly what has been done in HARMOS, which
will be discussed in the next section.
5.2 Analysis with HARMOS
The actual analysis with HARMOS could not be
conducted yet, since the GUI is still being implemented. At the moment HARMOS is however
able to import data from ABS projects. After doing that the peak day cases defined in the ABS
project are available for processing within HARMOS, which will be explained in some more detail in the next sections.
10

Runway Capacity Planning supported by Dynamic Programming

5.2.1 Setting up yearly cases

A HARMOS user can define its own yearly case
by selecting peak day cases (i.e. runway configurations) and assigning runway usage percentages. Because HARMOS uses the python COM
package for controlling the ABS model 2, only
one yearly case per runway system (5P, 6P, 6PK,
7PK) needs to be setup. When this is done, the
yearly cases can be processed for all years of interest (2003, 2010, 2020, 2030) by passing the
appropriate flight schedule as an argument to the
ABS model 2 COM object. This is already a
enormous time-saver because only four yearly
cases have to be setup instead of 12 yearly cases
when working with ABS.
Besides that, setting the correct input data for
accounting for technological developments has
also been automated in HARMOS by rewriting
the text file with separation standards automatically.
5.2.2 Running the cases and processing its output
After setting up a yearly case, by selecting a number of peak day cases and assigning usage percentages, HARMOS will allow a user to run the
ABS model 2 for all peak day cases subsequently.
Because HARMOS is controlling ABS model 2
through its runtime arguments, there is no need
to run peak day cases one at a time like in ABS.
After each individual run, the output data for that
peak day case is processed immediately so that
when the calculation session ends all annual information (here: average delay) is available.
5.2.3 Optimal Planning
A dynamic programming algorithm is currently
being implemented as a HARMOS module. Such
a module enables a user to determine the best investment plan based on his/her objectives. The
objective is to minimize the overall costs. The
overall costs are the sum of delay costs and capacity costs. Delay costs are directly related to
the (mis)match between capacity and demand.
Capacity costs are the cost of investment capital
and the operational and maintenance cost.

Each decision, i.e. stick with the current runway system or invest in a new runway, has costs
associated with it. The first decision leads to increasing delay cost which may or may not be
acceptable. The second decision results in additional capacity becoming available which will
decrease delay cost but - next to the cost of the actual investment - increases operational and maintenance cost. By turning this decision problem
into a dynamic programming problem it will be
very easy to determine the best investment plan.
5.2.4 Conclusion
Like mentioned before, HARMOS incorporates
mathematical optimization techniques for conducting airport performance analysis and planning. With respect to runway capacity planning it
turns out such an approach makes this task much
more efficient too because the analysis tasks to
be performed (done manually with ABS) can be
highly automated.
Performing the analysis with another scenario
is therefore very easy and does not require a user
selecting new datasets everywhere (as it is the
case in ABS).
6 Conclusions and recommendations
This paper presented two kind of decisionsupport systems for strategic airport planning.
The ABS, which is based on the philosophy of
non-linear navigation and HARMOS which is
based on the philosophy of mathematical optimization. For both decision-support systems
the support for analyzing the runway expansion
problem has been investigated. Based on this preliminary investigation, the following can be concluded:
ABS flexibility turns out to be a disadvantage in this case. Because of heavy data requirements, a lot of user input and interaction is needed with the system even before
one can actually start making comparisons
between different runway expansion plans.
More automation in ABS is needed before
a user can efficiently analyze this problem;
11

R.A.A. WIJNEN

A strong point with respect to ABS is its

modular design. The modular design made
it possible to use the capacity and delay
model of the ABS in HARMOS;
The optimization approach that is used in
HARMOS (i.e. turning the runway expansion problem into a dynamic programming
problem) is expected to make determination of the best investment plan much more
efficient. Implementation efforts are still
ongoing but it already became clear that
this approach enables a highly automated
analysis process. With HARMOS the focus can be more on the problem itself instead of on the analysis process that is
needed for the problem.

huis, and A.R.C. de Haan. The Airport Business

Suite: A Decision Support System for Airport
Strategic Exploration. In Proceedings of the 3rd
AIAA Aviation Technology, Integration and Operations Forum, November 2003.
[6] Roland A. A. Wijnen and Hendrikus G. Visser.
Optimal departure trajectories with respect to
sleep disturbance. Aerospace Science and Technology, 7(1):8191, 2003.

An interesting addition to the problem definition

is to include a second objective. Besides minimal cost, the chosen investment plan could also
be required to have minimal noise impact during
the planning period. Such a problem will be explored, in terms of requirements for the HARMOS system and the optimization techniques,
during future research.
7 Literature
[1] Martin Koning, Eugene Verkade, and Jacco
Hakfoort. Gevolgen van uitbreiding Schiphol.
Een kengetallen kosten-batenanalyse (in Dutch).
Technical report, Netherlands Bureau for Economic Policy Analysis, April 2002.
[2] David G. Smith. A Methodology for time-staged
capacity expansion sensitive to short-run pricing
and operational policy optimization. Engineering
and Process Economics, 4:371384, 1979.
[3] Hendrikus G. Visser and Roland A. A. Wijnen.
Optimization of noise abatement departure trajectories. Journal of Aircraft, 38(4), July-August
2001.
[4] Hendrikus G. Visser and Roland A. A. Wijnen.
Optimization of noise abatement arrival trajectories. The Aeronautical Journal, 107(1076):607
615, 2003.
[5] H.G. Visser, R.A.A. Wijnen, R.T.H. Chin, W.E.
Walker, N.A. Lang, J. Keur, U. Kohse, J. Veld-

12