Analysis of experimental data on ATC workload complexity

Analysis of the

experimental data on

ATC workload

complexity

BEngFinal Project Author: Crstic Mihai Adelin Supervisor(s): Octavian Thor Pleter PhD, PhD, MBA (MBS)

Eng. Rzvan Bucuroiu Eng, Head of Operational Planning Unit (EUROCONTROL) Session 2013

iii | P a g e

Acknowledgements

I would like to thank the ATCOs of Bucureti ACC for their assistance and cooperation during the surveys. I highly appreciate their warm welcome and their complete co-operation during the discussions and data gathering process. I would also like to thank EUROCONTROL experts that provided me with their valuable feedback and expertise. Special thanks should go to my mentors, Professor Octavian Thor Pleter and Rzvan Bucuroiu, who provided me with their guidance and support throughout the research.

TThhiiss ppaaggee ii ss lleefftt bb llaannkk iinntteenn ttiioonnaall llyy ffoorr pprriinn ttiinngg ppuurrppoosseess

v | P a g e

Anti-Plagiarism Declaration I, the undersigned, Crstica Mihai Adelin, student of the University Politehnica of Bucharest, Faculty of

Aerospace Engineering declare herewith and certify that this final project is the result of my own,

original, individual work. All the external sources of information used were quoted and included in the

References. All the figures, diagrams, and tables taken from external sources include a reference to the

source.

Date: Signature:

vii | P a g e

Table of Contents

List of Annexes. ........................................................ I

List of figures............................................................ I

List of tables.............................................................. I

Executive summary

a. Executive summary in English.......................VI b. Executive summary in Romanian............................X

1. Introduction............................................1 1.1 Structure of the document 1.2 Workload........................3 1.3 Complexity.....................3 1.4 Objective........................5

2. What has been done? .................................6 2.1. Introduction......................................6 2.2. Information theory approach to complexity.............................9 2.3. Queuing theory...............................................11 2.4. Latest approaches to Complexity Management.................................12

2.4.1. Algorithmic Approach (EUROCONTROL) .......................................................12 2.4.2. Cognitive Approach (AENA) ............................14 2.4.3. Statistical Approach (DSNA) .................................17 2.4.4. Other methods..................................19

a) Dynamic Density...................................................................................................19 b) COCA Project.......................................................................................................19 c) CAPAN Method...................................................................................................20

2.5. Complexity factors.............................................21

3. ATC complexity model.......................26 3.1. Introduction...................26 3.2. Complexity, Workload and Capacity.............................................28

3.2.1. Complexity raised by tra jectory uncertainty....................................28 3.3. Airspace overview..........................32

3.3.1. Sectorisation................................................34 3.3.2. Vertical Movement......................35 3.3.3. Directional flows.....................35 3.3.4. Sectors selection..............................................37 3.3.5. Sectors load.................................................................................. ..............................38

3.4. Complexity factors and metrics.............................................................43 3.5. Weighting the complexity function - method description......................................................49 3.6. Questionnaires Results.......................50

3.6.1. LRCN19 09 - 10 UTC EC answers analysis...........................50 3.6.2. LRCN19 09 - 10 UTC PC answers analysis................................53 3.6.3. LRKNL16 08 - 09 UTC EC answers analysis.....................53 3.6.4. LRKNL81908 - 09UTC EC answers analysis.....................54 3.6.5. LRKNL8908 - 09UTC EC answers analysis.......................54

3.7. Potential improvement per Key Performance Area...........................................54 3.8. Further development and investigations....................55

4. Conclusions..........................56 5. References.................................................................57

viii | P a g e


I | P a g e

List of Annexes

Annex A - AHP FORM..............................................................................59

Annex B ISA...........................................................................................63

Annex C- Traffic Sample..........................................................................67

Annex D- Weights calculation.................................................................71

List of figures

Figure 1.1. Notional representation of the distinctions between cognitive complexity, perceived complexity, and system complexity ([9], page 20)..............................................................5

Figure 2.1. Jing Xing Activity patterns over time for information processing in the three stages......10 Figure 2.2. Complexity assessment with a neural network (SESAR complexity management in en-

route)..................................................................................................................................18 Figure 2.3. CAPAN workload analysis................................................................................................21 Figure 3.1. Flights intersecting Bucureti FIR as planned on 29JUN2012 (Flight Plan).....................29 Figure 3.2. Flights intersecting Bucureti FIR as flown on 29JUN2012 (Radar data)........................30 Figure 3.3. Bucureti ACC Sectors (NEVAC)................................................................................. ...32 Figure 3.4. Distribution of the flights in the vertical plane Bucureti ACC for 23 JUN 2012.............35 Figure 3.5. Directional Flows (SAAM analysis)..................................................................................35 Figure 3.6. LRKNL16 and LRKNL89 top view (SAAM)...................................................................38 Figure 3.7. Sectors LRKNL16 and LRKNL89 3D view (NEVAC)....................................................39 Figure 3.8. Sector LRKNL16-SAAM analysis - sector load-30 JUN 2013........................................39 Figure 3.9. Sector LRKNL89-SAAM analysis - sector load-30 JUN 2013........................................40 Figure 3.10. Traffic passing through sector LRKNL16-30 JUN 2013(08:00-09:00 UTC)...................40 Figure 3.11. Traffic passing through sector LRKNL89-30 JUN 2013(08:00-09:00 UTC)..................41 Figure 3.14. Sector LRCN19-SAAM analysis - sector load-30 JUN 2013..........................................42 Figure 3.16. Complexity factors............................................................................................................44 Figure 3.17. Entry-exit points complexity.............................................................................................46

List of tables

Table 2.1 STATFOR Traffic forecast...............................................................................................6 Table 2.2 Jing Xing- Metrics of information complexity.................................................................10 Table 2.3 CAPAN workload thresholds..........................................................................................20 Table 2.4 Table of complexity factors (COCA project)....................................................................23 Table 3.1 Sectors dimensions and declared capacity (AIP Romania and EUROCONTROL data)..33 Table 3.2 Flights by country-pairs impacting Bucuresti ACC (23 JUN 2012).................................36 Table 3.3 Sectors LRKNL16 and LRKNL89 Load- SAAM Analysis..............................................39 Table 3.4 Sector LRCN19 Load- SAAM Analysis............................................................................42 Table 3.5 Complexity factors and Metrics.........................................................................................48 Table 3.6 ATCO answers...................................................................................................................50 Table 3.9 Factors weights LRCN19 EC answers...............................................................................51 Table 3.10 Factors weights LRCN19 PC answers...............................................................................53 Table 3.11 Factors weights LRKNL16 PC answers............................................................................53 Table 3.12 Factors weights LRKNL89 PC answers............................................................................54 Table 3.13 Factors weights LRKNL89 EC answers............................................................................54

TThhiiss ppaaggee ii ss lleefftt bb llaannkk iinntteenn ttiioonnaa ll llyy ffoorr pprriinn ttiinngg ppuurrppoosseess

III | P a g e

Glossary of terms and acronyms used

A

AENA Aeropuertos Espaoles y Navegacin Area

AHP Analytical Hierarchy Process

AMAN Arrivals MANager

ANSP Air Navigation Service Provider

ASA Automated Support to ATC

ATC Air Traffic Control

ATCO Air Traffic Controller

ATC-TRG ATC Training

ATM Air Traffic Management

ATS Air Traffic Services

A/C Aircraft

B

BEng Bachelor of Engineering

C

CAPAN CApacityAnalyzer

CM Complexity Management

COCA Complexity and Capacity

CORA COnflict Resolution Assistant

D

DSNA Direction des Services de la Navigation Arienne

E

EC/EXEC Executive Controller

E-TLM Enhanced Traffic Load Monitoring

ESRA EUROCONTROL Statistical Reference Area

EUROCONTROL European Agency for the Safety of Air Navigation

F

FAA Federal Aviation Administration

FDP Flight Data Processing

FDPS Flight Data Processing System

FRA Free Route Airspace

FL Flight Level

H

HMI Human Machine Interface

I

IC Information Complexity

ICAT International Center for Air Transportation

IFR Instrument Flight Rules

ISA Instantaneous Self-Assessmentof workload technique

IV | P a g e

K

KPA Key Performance Areas

L

LoA Letter of Agreement

Lvl Level

M

MIT Massachusetts Institute of Technology

MeSP Meta Sector Planner

MTCD Medium Term Conflict Detection

MUAC Maastricht Upper Area Control Center

N

NASA National Aeronautics and Space Administration

NEVAC Network Estimation Visualization ACC Capacity

O

OLDI On-Line data interchange

P

PC Planning Controller

R

RT Radio Telephony

S

SAAM System for Traffic Assignment and Analysis at Macroscopic level

SESAR Single European Sky Air Navigation Research

SESAR JU SESAR Joint Undertaken

STATFOR Statistics and Forecasts

SYSCO Concept for System Supported Coordination

T

TMS Traffic Management System

U

USA United States of America

V

VFR Visual Flight Rules

VI | P a g e B E n g F i n a l P r o j e c t

Executive Summary

This chapter is thought to be a short summary of the work done by Cirstica Mihai-Adelin, student of

Faculty of Aerospace Engineering, University Politehnica of Bucharest, in the field of ATC workload

complexity, under the close supervision of Octavian Thor Pleter PhD, PhD, MBA (MBS) and Razvan

Bucuroiu Eng, Head of Operational Planning Unit (EUROCONTROL), that represents his BEng Final

Project.

Executive Summary in English

This work addresses the subject of ATC workload complexity. Following an in depth study of a large

number of documents, research projects, articles and publications within this BEng Final Project a set of

complexity factors characterising the sectors belonging to Bucureti ACC have been identified and their

associated metrics have been developed. A method of weighting the complexity factors based on AHP

was elaborated and a first test was accomplished with the help of a number of ATCOs from Bucureti

ACC. The factors and metrics were used to support the construction of a new approach to ATC

complexity and to build an original complexity model.

Workload and Complexity in ATC are among the most studied concepts; numerous studies have been

undertaken in these domains for more than 40 years, basically from the beginning of ATC era itself.

Supported by the continuous and sustained traffic growth (2% per annum) forecasted for the coming

years, air traffic density will increase and so will the complexity. Furthermore, new concepts such as Free

Route Airspace are thought by some to bring further increase in the dynamism and complexity of the

traffic; therefore, new tools need to be developed to assist controllers in dealing with the rising traffic

demand and the new arising situation. Following the aforementioned situation, further investigation needs

to be done in this particular area. The proposed idea aims to bring an added value to the already existing

studies.

In the first chapter you can find the introduction of the document, its structure, based on chapters and sub-

chapters as well as the definition of workload and complexity, since the two are the main subject

examined in this paper.

VII | P a g e B E n g F i n a l P r o j e c t

Further, the main objectives envisaged by this BEng are presented as follows:

OBJ1: Develop forms, gather and analyse experimental data on ATC complexity from a

predefined set of sectors

OBJ2: Study workload from the ATC complexity perspective

OBJ3: Study complexity factors and identify those impacting the sectors in study.

OBJ4: Develop metrics for the factors identified in OBJ3

OBJ5: Build a model of air traffic complexity

OBJ6: Develop an ISA (Instantaneous self-assessment of workload technique) tool

It is worth to say that the above-mentioned objectives were successfully attained the in the content of this

document.

The second chapter of this document provides the reader with an overview of the complexity as seen from

various domains. Among them are information theory approach to complexity, queuing theory etc. as well

as an introduction to the most recent models and studies into the field of ATC complexity such as

EUROCONTROLs algorithmic approach, CAPAN method, COCA Project, and many others. This

chapter is crucial, as it creates a background and a foundation on which to build the original idea

developed by the author, which is to be found in the following chapter.

The third chapter represents the author s` original approach to complexity.

In the beginning of the chapter a number of traffic particularities and sectors characteristics are identified

and presented, using a multitude of tools and sources with the aim to familiarise the reader with the

situation. Further on, a traffic analysis (traffic load) done with SAAM tool, for sectors LRKNL16,

LRKNL89, and LRCN19 and traffic for 30JUN2013 between 08:00 and 09:00 UTC, and between 09:00-

10:00 UTC respectively, is presented. Further in the chapter, the author in cooperation with a number of

ATCOs identified a number of 12 complexity factors thought to be the most influential for the area

serviced by Bucureti ACC and then several metrics have been developed.

As a logical continuation a model of ATC complexity is developed by grouping the12 identified factors in

3 main categories: Traffic characteristics, Sector Characteristics and Dynamicity; a proposal for a linear

complexity function is made. The method used for weighting the factors, based on AHP, is thought to

capture the ATCO perceived complexity, thus making the function a hybrid between subjective and

objective complexity that can prove to be a better estimation of ATCO workload, and therefore, of

capacity.

VIII | P a g e B E n g F i n a l P r o j e c t

After the method was presented to a number of controllers from Bucureti ACC, they were asked to

answer to the questionnaires as from Annex B after finishing their shift (9:00 for KONEL, 10:00 for

NAPOC). After this gathering data session the results have been analysed and the weights of the criteria

have been calculated as you can see in 3.7 and in Annex D.

Even though the correlation between PC ratings and EC ratings is good for most of the factors, the results

stressed out once again that complexity is a very subjective concept, and that the same situation can be

categorised with different levels of complexity, mainly because sometimes ATCOs find it difficult to

discriminate between the relative importance of one factor against each comparison.

The ISA tool developed here with the purpose to catch controllers subjective rating of workload each

every 2 minutes has not been used in the real environment due to a number of safety concerns, however,

for a simulated environment the benefit of such a tool has been recognised by all the respondents.

As a conclusion, one can say that this BEeng Final Project did not only successfully attain all the

preliminary objectives set out in Chapter 1, but went beyond its initial purpose. These findings open the

room for future tests and simulation, in order to validate the concept.

X | P a g e B E n g F i n a l P r o j e c t

Executive Summary in Romanian

Aceast lucrare trateaz subiectul complexitii in activitatea de control a traficului aerian. Ca urmare a

studierii n amnunt a nenumrate documente, proiecte de cercetare, articole si publicaii in cadrul acestei

lucrri de licent au fost identificai o serie de factori ce determin complexitatea in sectoarele apartinnd

ACC-ului Bucureti i pentru fiecare factor in parte s-a dezvoltat cate un mod de msurare. De asemenea

a fost elaborat o metoda de stabilire a ponderei fiecrui factor folosindu-se procesul ierarhizrii analitice

si un prim test a fost realizat cu ajutorul controlorilor de traffic aerian din ACC Bucuresti. Factorii i

metricii mai sus menionai au fost utilizai pentru a ajuta la construcia unei noi abordri asupra

complexitii in cotrolul traficului aerian, i la dezvoltarea unui model de complexitate.

ncrcarea si complexitatea in controlul traficului aerian sunt unele dintre cele mai studiate concepte,

studii in aceste domenii facndu-se de mai mult de 40 de ani, teoretic de la inceputul erei ATC.

Alimentat de creterea continu si susinut a traficului(2% pe an) preconizat pentru viitorii ani,

densitatea traficului si complexitatea acestuia va crete. Mai mult dect att noile concept precum spaiula

erian Free Routesunt vzute ca un potenator al dinamismului i complexitii traficului aerian. De

aceea este nevoie ca noi instrumente sa fie dezvoltate, pentru a ajuta controlorii sa fac fa acestei

creteri a traficului in contextul prezentat.

Ca urmare a situaiei contextuale prezentat mai sus sunt de parere ca nevoia pentru inceperea acestui

studiu a aprut in mod natural, i c idea propus poate aduce o valoarea daugat pentru ceea ce exist

deja.

Primul capitol cuprinde ntroducerea acestui document, structura acestuia, bazat pe capitol i

sub.capitole, precum i definiia volumului de munc i al complexitii. n continuarea primului capitol

sunt prezentate obiectivele prevzute de aceast lucrare de licen, dup cum urmeaz:

OBJ1: Dezvoltarea de formulare, strngerea si analizarea datelor experimentale asupra complexitii

activitii de control a traficului aerian, pe un numr predefinit de sectoare.

OBJ2: Studiul volumului de munc al controlorului din perspective complexitii activitii de control a

traficului aerian

OBJ3: Studiul factorilor asociai complexitii i identificarea celor care influeneaz sectoarele studiate.

OBJ4: Dezvoltarea de metrici pentru factoriii dentificai inOBJ3

OBJ5: Construirea unui model de complexitate a traficului aerian

OBJ6:Dezvoltarea unui instrument ISA (tehnic de auto-evaluare instantanee a incrcrii)

Merit spus ca obiectivele mai sus menionate au fost atinse cu success in cuprinsul prezentei lucrri.

Cel de-al doilea capitol al acestui document furnizeaz cititorului o vedere de ansamblu asupra

complexitii aa cum este vazut din diferite domenii, precum i o introducere asupra celor mai recente

modele si studii in cmpul complexitii in ATC, cum ar fi: abordarea algoritmic propusa de

EUROCONROL, metoda CAPAN, proiectul COCA, i multe altele. Importana acestui capitol const n

XI | P a g e B E n g F i n a l P r o j e c t

crearea unui fundal pe care se poate proiecta idea dezvoltat de autor n cel de-al treilea capitol al acestui

document.

Al treilea capitol reprezint abordarea original asupra complexitii a autorului. n nceputul capitolului,

sunt prezentate numeroasele particulariti ale traficului, i caracteristici a sectoarelor, utiliznd o

multitudine de instrumente (SAAM, NEVAC, etc.) i surse (DDR, AIP, etc.) cu scopul de a familiariza

cititorul cu situaia prezentat.Mai departe se regseste o analiza a traficului(ncrcarea de traffic) din

30.06.2013 ntre orele 8:00 si 9:00 UTC respectiv 9:00-10:00 UTC, pentru sectoarele LRKNL16,

LRKNL89 si LRCN19, folosind instrumentul SAAM. n continuare n cooperare cu un numr de

controlori de traffic, autorul, a identificat un numr de 12 factori care au cea mai mare influen asupra

sectoarelor din Bucureti ACC, factori pentru care au fost dezvoltai metricii afereni.

Continuarea logic a fost dezvoltarea unui model de complexitate.Metoda prezentat n acest document

pentru a stabili ponderea factorilor de complexitate este cea a procesului de ierarhizare analitic prin care

s-a urmrit captarea complexitii subiective aa cum este perceput de controlor. Acest lucru face ca

funcia de complexitate dezvoltat in cuprinsul acestei lucrri s devin un hybrid ntre complexitatea

subiectiv i cea masurat obiectiv, ce poate ajuta la o mai bun estimare a incrcrii controlorului, i a

capacitii de sector.

Dup prezentarea metodei unui numar de controlori de traffic de la Bucureti ACC, acetia au fost rugai

s rspund chestionarelor din Annex B, imediat dupace au ieit din tura de dirijare (9:00 pentru KONEL

i 10:00 pentru NAPOC). Dup strngerea datelor i analizarea rezultatelor au fost calculate ponderile

factorilor de complexitate dup cum sepoate vedea in 3.7 si in Annex D.

Chiar dac corelarea ntre rspunsurile controlorului din pozitia Planning i cele ale controlorului

Executive este bun pentru majoritatea factorilor, rezultatele au subliniat inc odat natura subiectiv a

complexitiii, i faptul c aceai situaie poate fi categorizat ca avnd nivele diferite de dificultate.

Lucrul acesta s-a ntamplat din cauza dificultii controlorilor de a decide importana relativa a unui factor

in faa comparatorului su.

Instrumentul ISA dezvoltat de autor pentru a surprinde auto-evaluarea incrcriicontrolorului la interval

de timp de 2 minute, nu a fost folosit in mediul real de dirijare din motive evidente de siguranta. Pentru

mediul simulate insa, respondenii au recunoscut necesitea unui astfel de instrument.

In concluzie se poate spune c aceast lucrare de licen i-a atins cu succes toate obiectivele stabilite in

capitol 1,iar descoperirile fcute au creat premisele necesare pentru dezvoltri viitoare .

TThhiiss ppaaggee ii ss lleefftt bb llaannkk iinntteenn ttiioonnaa ll llyy ffoorr pprriinn ttiinngg ppuurrppoosseess

1 | P a g e B E n g F i n a l P r o j e c t

1. Introduction

The following document is a review of the research work done by the author throughout the last academic

year in the fields of ATC complexity, workload and capacity.

One of the most challenging tasks of Air Traffic Management both for present and future times is and will

be to efficiently handle the further increasing traffic density and airspace complexity a way that assures

capacity will meet the traffic demand. Despite the numerous benefits brought to the network by

innovative concepts of operations such as Free Routes, these are still perceived by some as a contributor

to further complexity.These issues will be addressed in the content of this document and a preliminary

conclusion will be drawn whether this assumption is valid or not.

From the above, we deduct that new or optimised algorithms and next generation support tools need to be

developed in order to assist the controller and keep his workload in acceptable limits, while increasing

sector capacity, though productivity, and maintaining the safety standards.

Acceptable workload limits can be subject to interpretation, but here, we refer to the CAPAN definition

of workload that sees ATCO workload as the actual working time within one hour, and imposes a 70%

thresholds equivalent to 42 minutes/hour as maximum acceptable workload. Every minute above this

threshold is seen as overload and thought to be unacceptable because it could compromise safety. While

most of the ATCOs prefer the CAPAN definition of workload, complexity reasons are manifold and

factors determining it are both numerous and diverse. Moreover, one might argue that the way complexity

is influenced by these factors varies from one sector to another, and therefore, calibrating a function

applicable for any sector and have in the same time a fair level of accuracy is inappropriate. These factors

must be identified so that proper algorithms can be developed and measures of detecting and recognising

complexity can be created.

This paper aims to develop a reliable method to diagnose complexity, and as a future development

envisages the need to calculate its influence upon controller workload and to identify possible and reliable

resolution strategies. Up to now, various research projects and academic papers have developed strategies

to deal with ATC complexity. These include a variety of methods and functions, and see complexity

either at a macroscopic scale, grand scale resolution (e.g. Sectorisation), or at a microscopic scale,

microscopic scale resolution (e.g. optimal aircraft control).


1.1 Structure of the document

The structure of this document is based on three main pillars, and therefore the document is split into

three main equally important parts.

The first part is devoted to the work that has already been done: concepts, ideas, representative works in

the industry. It includes a brief explanation of some of the concepts, and it is important because it helps to

place this work in the overall ATC complexity picture. In the second part, this document features a

synopsis of the state-of-the art approaches to complexity and controller workload as proposed by several

key players in European ATM (EUROCONTROL, AENA, DSNA), as well as with a breakdown of the

most noticeable approaches (algorithmic approach, cognitive approach, statistical approach). The review

does not address other complexity related topics, such as the use of complexity assessment in airspace

design, for instance. These approaches were chosen because in the author s` view a future combination of

the three can be seen as a major breakthrough from this so intense studied field of ATC complexity.

The third and last part of this document presents the ATC workload complexity issue from the author

personal and original perspective, backed by a thorough study of representative works that had been

undertaken in the industry in the past, as well as by the newest research programs currently under SESAR

validation (Complexity Management in En-route ID: 04.07.01.D13). A complexity model or

complexity function will be presented as a step by step exemplification on a selected sector. This intends

to prove the method applicability and usefulness. Eventually, the paper offers some conclusions that result

from this study, possible benefits by KPA of adapting this method, as well as foreseen ways of

improvement.

In the sequel the author will treat and consolidate the topics of Workload and Complexity, in order to both

give a summary of previous studies, as well as to introduce perspectives and outlooks to possible further

advancements.


1.2. Workload

Workload is defined by Cambridge Advanced Learners Dictionary and Thesaurus as the amount of work

to be done, especially by a particular person or machine in a period of time. As air traffic controllers are

people trained to maintain the safe, orderly and expeditious flow of air traffic in the global air traffic

system, their workload is generally mental, as opposed to physical, in nature, therefore it is not surprising

that a clear definition cannot be identified in the literature. However, the general idea is that workload is

subjective, factors such as aptitude, skill, experience, operating behaviours, and personality traits

(Cognitive complexity in air traffic control A literature review) are the determinants of subjective

workload in ATC. Therefore, we may observe an identical traffic situation to present a reasonable amount

of workload for a controller who is experienced on that particular sector, and yet to overtax a novice or a

controller who has no experience with the sector. A distinction is generally made between taskload (the

objective demands of a task) and workload (the subjective demand experienced in the performance of a

task). In practice, in ATC the workload is generally measured in task duration (time) and studied in close

relationship with the sector capacity. The number of aircraft in the sector that determine a cumulative

task duration up to a certain threshold (refer to Annex 3) within one hour is said to be the capacity of the

sector. The threshold used by the majority of ANSPs in determining the maximum capacity is 70% which

means that a maximum of 42 minutes of working time within one hour is accepted. An interesting link

between workload and complexity is provided by Kopardekar et al. in [7] they argue that air traffic

controller workload is a subjective attribute and is an effect of air traffic complexity .

1.3. Complexity

Despite the fact that the concept of complexity has been intensively studied in a multiple ways across a

variety of domains, in ATC there is still some ambiguity left in the use of terms such as complexity,

sector complexity, and traffic complexity.

As a general term, something that is complex or complexity is defined by Cambridge Advanced Learners

Dictionary, as being The state of having many parts and being difficult to understand or find answer to.

It is more difficult to project the future behaviour of a system with a high number of components than of

one with a fewer number of components.


Terms such as complexity, sector complexity, and traffic complexity are sometimes used interchangeably,

resulting in confusion when attempting to review literature in this field and when undertaking new

research projects [The complexity construct in ATC,1995[8]]. Many definitions of complexity revolve

around the size, count or number of items in an object, FAA sector complexity is defined as the

number of arrivals, departures, en-route aircraft, emergencies, special flights, and coordination associated

with a sector (FAA, 1984;[8]).

Generally, ATC complexity can be defined as an assembly composed of a number of sector and traffic

complexity factors. These factors can be physical aspects of the sector, such as size or airway

configuration, or factors relating to the movement of air traffic through the airspace, such as number of

climbing and descending flights. Some factors cover both sector and traffic issues, such as required

procedures and functions (Grossberg, 1989; Schmidt, 1976; [8])

Furthermore, a distinction can be made between different types of identified complexity. These types of

complexity are: system complexity, cognitive complexity and perceived complexity.

Jonathan M. Histon and R. John Hansman have identified a model that links the three types of complexity

as in Figure 1.1.

The system complexity represents the complexity that is inherent within the real system,

independent of an operator controlling it.

Cognitive complexity is a measure of the complexity of the working mental model that issued to

represent the real system. It is a measure of the complexity of the representation of the real

system that the controller must use to perform the current cognitive task. The perceived

complexity is the apparent complexity, to the controller, of their representation of the real system.

It is thought that the controller can regulate the perceived complexity by adapting the cognitive

tools, such as the working mental mode, used to perform the required cognitive tasks(The impact

of structure on cognitive complexity in air traffic control Jonathan M. Histon and R. John

Hansman, MIT ICAT,2002, [9]).


Figure 1.1 Notional representation of the distinctions between cognitive complexity, perceived complexity, and system

complexity ([9], page 20)

1.4. Objective

The chief goal of this BEng Final Project is to study and analyse experimental data on ATC workload

complexity. A broader set of goals is to identify, develop and evaluate factors related to air traffic control

complexity, to understand the ATC complexity by studying previous research and scientific papers, and

finally, to develop a complexity function to support reliable workload estimations, better than the classical

and simple aircraft count method.

Based on the above description the following objectives can be identified:

OBJ1: Develop forms, gather and analyse experimental data on ATC complexity from a predefined set of

sectors

OBJ2: Study workload from the ATC complexity perspective

OBJ3: Study complexity factors and identify those impacting the sectors in study.

OBJ4: Develop metrics for the factors identified in OBJ3

OBJ5: Build a model of air traffic complexity

OBJ6: Develop an ISA (Instantaneous self-assessment of workload technique) tool

To date, research in air traffic complexity has tended to overlook cognitive aspects of the controllers

task. In the context of this BEng project, the author has recognised the need to incorporate such cognitive

aspects of air traffic complexity. One of the aims of the current research paper is to ensure a model of

traffic complexity that adequately captures those cognitive aspects, as well as the objective aspects

imposed by the traffic situation.


2. What has been done?

This chapter is a short introduction into the work that has been done into the field of ATC complexity as

well as workload and capacity and a summary of the main ideas depicted from the literature sources the

author had under review; reports, scientific articles, book chapters and operational reviews. The

importance of this chapter relies on making an overview of the various complexity approaches that helps

placing this work in the overall picture.

2.1. Introduction

Despite the fact that in 2012 air traffic in Europe registered a drop of around 2.4%, according to

EUROCONTROL Medium Term Forecast (February 2013) in the forthcoming years the growth will be

positive and continuous, so that by 2019 there will be around 11.2 million IFR flights that will have to be

accommodated in the crowded European sky during a year. This corresponds to a 17% increase in traffic

comparatively to 2012.The Long Term Forecast states that in 2030 Europe skies will have to

accommodate 16 to 20million fights yearly, which means as much as double as 2008 traffic which is the

year with the peak of traffic recorded until now with 10million 83 thousands IFR flights.

ESRA08 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AAGR

2019/201

2

IFR Flight Movement

s

(Thousand)

H

9,559 9,924 10,358 10,83

5 11,22

2 11,65

8 12,07

9 3.4%

B 10,083 9,413 9,493 9,784 9,548 9,424 9,689 9,975 10,29

8 10,57

6 10,88

5 11,19

5 2.3%

L

9,276 9,409 9,517 9,690 9,815 9,969 10,13

2 0.9%

Annual Growth

(compared to previous

year)

H

0.1% 3.8% 4.4% 4.6% 3.6% 3.9% 3.6% 3.4%

B 0.4% -6.6% 0.8% 3.1% -2.4% -1.3% 2.8% 3.0% 3.2% 2.7% 2.9% 2.8% 2.3%

L

-2.9% 1.4% 1.2% 1.8% 1.3% 1.6% 1.6% 0.9%

H - High; B - Basic; L - Low

Table 2.1.STATFOR Traffic forecast

Supported by the latest technologies available on the market, ATS will have to develop in the near future

new systems able to assist and enable the ATCOs to deal with the increased forecasted traffic as well as

with the new operating environment, free-route airspace, that is currently one of the trends in Europe

and that will most probably be adopted by all European States at some point in the future. The personal

belief of the author is that in order to assure efficient and safe operations in this arising situation, new

complexity models should be developed, ones that will match these new directions of development.


While technological advancements have no foreseeable limits and in theory are able to assure the

monitoring and to supply with flight data an infinite number of aircrafts at once, the air traffic controller

is a limited entity that can only deal and provide useful information to a small and limited number of

aircrafts at the same time. Thus, we can draw a simple yet conclusive idea: the human, ATCO, is likely to

remain the biggest, if not the only bottleneck, of the ATM system in what concerns capacity.

Traditionally, traffic density has been the single factor most associated with complexity.

However, it is increasingly clear that density by itself is an insufficient indicator of the

difficulty a controller faces. Anecdotal evidence suggests that controllers increasingly

speak not of the difficulty of a given traffic density, but of the associated traffic complexity.

Past attempts to assess complexity (Kirwan et al.,2001) have generally relied on geometric

relationships between aircraft, or on observable physical activity (Pawlak et al., 1996).

Increasingly, it is being recognized that complexity factors can interact (Fracker& Davis,

1990) in nonlinear ways (Majumdar&Ochieng, 2000; Athenes et al., 2002),and that

individual differences between controllers can mean that different controllers respond

differently to the same constellation of complexity factors (Mogford et al., 1994). These are

among the considerations that seem to be driving the search for a better way to describe and

predict complexity as it affects the controller (COCA project).

As the air traffic controller workload is primarily affected by the features of the air traffic (FL change,

Heading change, etc.) and as the ATC sector capacity is commonly estimated only by using the number of

aircraft within that sector for a certain period of time (traffic density), usually 20 minutes or one hour

(NEVAC model), it is straightforward to conclude that the models used to assess the workload, and

therefore the sector capacities, are subjected to a big variety of errors and unknowns and the obtained

values present a considerable lack of precision.

Sector capacities are mainly driven by the air traffic controller workload which in turn is driven among

others by the complexity associated to the traffic he has to monitor. In other words, ATC complexity is

the cause that via controllers workload leads to the effect limited sector capacity and impossibility to

adapt the capacity and provide a safe and orderly flow for all the demand.


With this in mind, it is not surprising that throughout the years the ATC complexity has been the subject

of so many studies and research programs both in Europe, but especially in the United States of America,

where ATC complexity related studies were made since the beginning of the ATC era itself. Despite of all

the work and sustained effort during the years, arguably, no one has managed to come up with a generally

accepted solution yet. One that can be applied in every environment and no one has succeeded yet to find

the right mean of correctly quantify the complexity of the ATC workload.

As mentioned above, the preponderance of the literature on ATC complexity factors originates from the

United States of America, primarily from the National Aeronautics and Space Administration (NASA)

and Federal Aviation Administration (FAA), where research references to ATC complexity and the

associated factors were done nearly 40 years ago (cf. Arad, 1964).

A number of seminal reports and articles on the subject of ATC complexity were produced in recent

years, including several in-depth reviews of complexity factor literature including the work done by the

Complexity and Capacity team (COCA) in the project called Cognitive complexity in Air Traffic Control

initiated by EUROCONTROL.

In the sequel relevant approaches to complexity as seen from different fields will be presented. For data

integrity reasons and to avoid possible errors due to wrong interpretation, upon their availability, the

primary sources were preferred and used where possible. Pertinent references were mostly drawn from

the fields of air traffic management, human factors and computer modelling. As this research is treating

the subject of ATC workload complexity, in this review it absolutely necessary to rely on the theoretical

and empirical work in the related field of complexity and workload assessment and capacity, as they are

all linked together. However, it should be noted that the main purpose of this chapter is not to be an

exhaustive review, but rather a framework in which are identified and synthesised the major theoretical,

empirical and operational perspectives on ATC complexity together with an introductory presentation of

the most interesting approaches relevant to the concept of complexity.

Below you can find some relevant approaches to complexity, as identified in the literature sources under

review and the way various researchers try to solve or explain the matter of complexity, using their own

and particular assumption from the perspective of a certain field.


2.2. Information theory approach to complexity

In the field of information theory complexity has been broadly reviewed and investigated along the years

in a number of scientific papers and researches and the term of information complexity (IC) is

commonly used to describe complexity from the perspective of a system.

One simple, yet straightforward definition of the complexity with regards to the information theory is that

of the minimum or shortest possible description size. Kolmogorov complexity known also as

descriptive complexity is defined as the minimum possible length of a description in some language

(Casti, 1979). For instance, if a description can reach a high degree of compression without losing its

meaning, then it is considered simpler than one that cannot. Considering this definition, the expressions

that are highly ordered appear as simple while random when ma intaining maximal complexity. For

example, consider the following two strings of length 18, the string A=(fiafiafiafiafiafia) is less complex

than the string B=(a2b3jcfsadfa52jj3x) because the former has a short English-language description,

being easily compressed into a description fia 6 times which consist of 11 characters. The second one

has no obvious simple description, other than writing down the string itself, which has 18 characters,

however, this definition corresponds to the difficulty of compressing a representation with little direct

connection to the practical aspects of a functioning organism. Indeed, it is only concerned with the

numeric size factor of complexity.

Jing Xing in his paper Information Complexity in Air Traffic Control Displays[]presents a framework for

decomposing factors of information complexity and a set of metrics to measure ATC display complexity.

Jing Xing information complexity model is based on a combination of three basic factors: quantity,

variety, and relation, factors that are evaluated with the mechanisms of brain information processing at

three stages of information processing: perception, cognition, and action. The metrics of complexity can

be derived by associating task requirements to brain functions.

It is worth to mention that the brain mechanisms are much more complex and that this model is just a

rough approximation of reality as the author argues and depicts in Figure 2.1:

While our three-stage classification of brain functions is coarse, those stages have

intrinsically different mechanisms of information processing. One example would be the

neuronal response patterns overtime, as illustrated in Figure2.


Perceptual neuronal responses (left panel) start following the onset of a stimulus and ends

after the stimulus is no longer present, while the cognitive responses (middle panel)

remain for an extended period without the stimulus, such sustained activity is the

substrate of working memory. On the other hand, neuronal responses in the cortical pre-

motor areas become activated before an action and end after the action plan is executed.

Another example is the way in which information is encoded. Perceptual information

processing is initially performed in a parallel manner. Thousands of visual neurons are

activated by a visual image and they simultaneously encode many pieces of the image.

Thus, the perceptual system offers a relatively broad information band width. On the

other hand, working memory, as the basis of cognitive activities, has a highly limited

capacity. That is, only a few pieces of information can be simultaneously encoded

(Cowan, 2001). Consequently, the information bandwidth of the cognition stage is much

less than that for perception. Finally, the cortical areas that encode action plans are

characterised with one plan at a time, yielding an even narrower bandwidth

(Georgopoulos, Schwartz & Kettner, 1986; Pouget, Zemel, & Dayan, 2000; Xing &

Andersen, 2000).

Perception Cognition Action

Stimulus Action execution Working memory

Perception Cognition Action

Quantity

Variety

Relation

Table 2.1 Jing Xing- Metrics of information complexity

Crutchfield and Young (1989) have broaden and enhanced Kolmogorovs complexity concept by defining

complexity as the minimal size of a model representation of a system that can statistically reproduce the

observed data within a specified tolerance with the name of topological complexity.

Figure 2.1: Jing Xing Activity patterns over time for information processing in the three stages


The following two traffic situations can be considered as an example. In both situations it is involved the

same number of aircrafts, the difference is that in Case A the aircraft are flying on fixed routes with just

one intersection while in Case B the aircrafts are flying off the routes, which can create many potential

conflicts. A controller can build a model of the first case that has two flows of aircraft and one crossing

point, while a model of the second case has to be composed of many flows and crossings. Thus the

topological complexity of Case A is less than Case B.

Stuart Kauffman (1993) defined complexity as the number of conflicting constraints. For example, the

airspace can be made less complex by removing air traffic constraints such as military zones, bad

weather, etc. It should be noted however, that the definition is only concerned with the complexity factor

of structural rules and cannot be accepted as a model for assessing ATC complexity.

2.3. Queuing theory

In queuing theory a model is constructed so that queue length and waiting times can be predicted.

Schmidt (1978) used this theory to analyse ATC workload. The difference between this approach and

other system engineering approaches that consider the human as a comparative function between one state

that is actual and another one that is the desired one is that in queuing theory the accomplishment of a task

and performance is measured in terms of task completion times. Making use of the operational data

collected from Los Angeles en-route controllers, Schmidt managed to derive a mathematical expression

which can be used for predicting the average en-route delay as well as the server occupancy upon the

demand, with a very good accuracy.[2]

Finally, cyclo-matic complexity is a complexity metric that attempts to capture the dependencies between

components. It considers the number of linearly independent loops through a system, with the

assumption that the greater the number of feedback loops, the greater the potential for complex behaviour

(Vikal, 2000, [10]). Vikal (2000) has used cyclo-matic complexity to analyse the apparent complexity of a

flight management system to a pilot.


2.4. Latest approaches to Complexity Management at European

Level

This chapter will introduce some of the latest approaches to complexity Management at European Level.

The first three: algorithmic approach, cognitive approach, and statistical approach, have also been

selected for further analysis and validation within SESAR JU working project: Complexity Management

in-en route. These approaches have been chosen because a future combination of the three can really be

seen as a major breakthrough from this so intense studied field of ATC workload complexity and because

they are relevant for this study, since they express the current trends in European ATM, and after

validation, they will most probably be adopted by many of European ANSPs.

Further on, a brief introduction into CAPAN method (CAPacityANalyser), COCA Project and Dynamic

Density will be done and their relevance for this research will be explained. In the end of this chapter

(2.6) complexity indicators and complexity metrics will be discussed.

2.4.1. Algorithmic Approach (EUROCONTROL)

Following the numerous works undertaken in the past with the aim of providing the ATC with automated

assistance, and also following the experiments with a number of assistance tools like Medium Term

Conflict Detection (MTCD-conflict detection), Arrival MANager (AMAN - arrival sequencing), Conflict

Resolution Assistant (CORA-conflict resolution) EUROCONTROL has initiated the work on Complexity

Management (CM) within their Automated Support to ATC Programme (ASA). As conclusions have

indicated, that support tool cannot work in isolation from the airspace organisation and planning and the

dynamic management of the flows and workload.

This approach is also based on previous work on analysis of Traffic Complexity for off-line sector

analysis (COCA project). Supported by this study it was concluded that by following this thread and

using its output operational concept for complexity management could be established.

The main goal of this project was to product an algorithm based on linear inputs for prediction of

complexity in a volume of airspace. Furthermore, the algorithm uses only data that can be downloaded

from the FDPS which restrict the room for erroneous inputs to be made but in the same time disregards

some of the cognitive aspects.


This algorithm was supported by previous fast and real time simulations that were undertaken for its

validation.

Real-time simulations using the Traffic Management System (TMS) Level 0 set-up were held at the

Maastricht Upper Area Control Centre, and essentially consisted of an integration of the E-TLM

(Enhanced Traffic Load Monitoring) tool (see 6.3.2) in the ATC Training (ATC-TRG) simulator; for the

complexity calculations performed by the E-TLM, both an algorithmic approach and a cognitive model

were run in parallel and were operationally validated.

The E-TLM tool is designed to:

Identify Operational Sectors overloads, given the current sectorisation;

Assess the impact when applying a different sectorisation in terms of sector workload (What-If on

sectorisation);

Calculate the changes needed to perform in the sectorisation in order to get the most appropriate

workload situation in the Operational Sectors for the current traffic (Sector Optimiser);

Provide the user with a list of aircraft contributing to specific workload/complexity problems allowing

him to act on the current situation

The E-TLM is connected to an FDP system in the MUAC simulation environment; a Meta Sector Planner

(MeSP) Controller Working Position connected to this FDP system allowed to make What-If inputs on

flights to assess the impact of trajectory changes on the predicted complexity (What-If on trajectories).

Predicted complexity was calculated based on an algorithmic approach and on a cognitive model (both

run in parallel). The cognitive model approach is further described in 2.5.2. For the algorithmic approach,

workload was calculated using the following model:


Where:

: The number of flights in an OPS Sector is calculated by adding the number of flights in each Basic Sector (nsec) belonging to this OPS Sector. Note: An ASPL contributes to the workload of an OPS Sector if it is the UNDER CONTROL OPS Sector.

: The number of flights in the OPS Sector that are in climb or descent is obtained by summing climbing flights and descending flights in this OPS Sector.

: The number of flights in the OPS Sector that are near the boundary is calculated by adding the number of flights that are near the boundary in a Basic Sector (nsb) belonging to this OPS Sector that are not inside any other Basic

Sector belonging to the OPS Sector

: The number of flights in the OPS Sector that can receive and acknowledge ATC clearances by data link is calculated by adding the number of flights in each Basic Sector (ndl) belonging to this OPS Sector classified as AGDL

flights.

Snorm= 1/ , where is the maximum traffic capacity that can be supported by a certain OPS Sector.

For each pair SGCP (Sector Group Configuration Patterns) and OPS Sector the Maximum traffic Capacity, and Upper and Lower threshold is statically adapted.

: The number of flights affected by a TSA availability change will be computed by adding flights that fulfill the condition: A flight contributes to the Ntsa parameter for an OPS Sector, if at the Report Time the intersection of theactive TSA crossing time interval with the OPS Sector crossing time interval is not null or the flight is inside a

timeinterval around the entry to the affected OPS Sector.

: The number of MTCD conflicts within the OPS Sector is calculated by adding the number of MTCD conflicts in each Basic Sector (nmtcd) belonging to this OPS Sector.

: The number of OAT flights controlled by a GAT OPS Sector (applicable only for GAT OPS Sectors) is obtained by summing all the Military classified flights inside this OPS Sector.

: The number of flights in a holding within the OPS Sector is obtained by summing all flights classified as holding flights inside this OPS Sector. (SESAR-JU Complexity Management in En-route, Step1: Consolidation of

previous studies)

The linearity of the model allowed dividing the workload in different components:

Occupancy workload: Coordination workload:

Conflict workload: - even if seen as linear in my view this is a non-linear

component of the function.

Furthermore, this approach allows specifying the contribution of each aircraft separately to the total

workload, thus allowing sorting on workload contribution; this is seen as an important enabler for theso-

called cherry-picking, where individual flights are chosen for specific actions (level capping,rerouting)

to reduce overall workload (SESAR-JU Complexity Management in En-route, Step1: Consolidation of

previous studies)

2.4.2. Cognitive Approach (AENA)

Aeropuertos Espaoles y Navegacin Area (AENA, Air Navigation branch) is the Spanish Air

Navigation Service Provider (ANSP) designated by the state to provide air navigation services within the

Spanish controlled airspace. The company also plays a leading, active role in all European Union projects

relating to the introduction of the Single European Sky, being the Project Manager for various working

packages and projects within SESAR (Single European Sky ATM Research).


Within AENA there have been developed two different methodologies to estimate the airspace

complexity, both equally important NORVASE methodology and E-TLM cognitive model; however for

the purpose of this study we will focus on the latter.

NORVASE methodology

NORVASE methodology is currently used in AENA during the strategic planning phase to determine

sector capacities and to design airspace sectorisation but its complexity indicators can also be applied

during the tactical/execution phase.

Enhanced Traffic Load Monitoring cognitive model. (E-TLM)

E-TLM is a tool internally developed by AENA created to dynamically adjust sector configurations to the

current, up-to-date traffic situation by means of measuring traffic complexity and reacting to high traffic

complexity situations. It is based on continuous complexity which takes into consideration up-to-date data

rather than historical demand. It measures the sector complexity as a function of the controllers workload

from the present time to few hours in advance by continuously doing Fast-Time Simulations of the

predicted traffic.

The best sectorisation is determined based on a predefined set of possible combinations, avoiding

workloads higher than the maximum acceptable safety level, minimis ing the number of open sectors and

balancing workloads between the operative sectors.

As stated before, the concept behind the E-TLM tool is the prediction of the controller workload for a set

of predefined sector configurations and this information is used to dynamically adjust sector

configurations to the real needs of controllers.

The inputs of this tool are the current and forecasted traffic flows as well as the predefined configuration

of different sectors of an airspace volume under study.

This tool calculates the workload values for each sector in all the possible predefined sector

configurations both for current time and as well as for a matter of hours in the future based on traffic

forecast.


It also incorporates a what-if functionality that displays to the user the optimum sectorisation that might

be used at each moment. Thus, the user will have the right information before the demand peaks arise. So,

he/she can detect in advance the demand peaks and can implement in a short time a change of

configuration to eliminate or smooth these peaks.

The E-TLM tool is comprised of three main components:

Workload calculator module - estimates the workload for every operational controller for

all the sector configurations: the one which is applied at that moment and the rest of the

predefined sector configurations.

Optimiser module - calculates the optimum sector configuration at each moment taking

into account user selected criteria.

HMI (Human Machine Interface) is the interface where all graphical data are shown.

For the purpose of this study only the first functionality is of a real interest and therefore in the next line I

will explain a bit the concept.

The E-TLM tool is integrated together with a fast time simulate or in such a way that from the predicted

traffic demand and the set of predefined sector configurations, it identifies the possible demand resources

from the airspace controller by generating a set of control events that will normally be undertaken by the

controller. The fast time simulator runs each every five minutes in order to make sure that the demand is

updated with all the flight plans that have changed during this period.

For calculating the workload the E-TLM model makes use of two concepts, a temporal workload, which

represents the time spent by the ATC to perform the controlling tasks, and a cognitive workload,

measured through an estimation of the impact of the cognitive channels demanded in each controlling

action.

The breakdown of ATC Actions (basic elements such as thinking, listening, speaking, etc.) in

cognitive channels (resources invoked in the brain) has been obtained from the experience of

human factor experts in the ATC domain.

Interferences between actions, which result in an increase of the controller workload, are

modelled in this module as different basic actions done at a time, so an action done

simultaneously with another one will have a different workload associated to it, depending on

what the second is.


The workload calculator module needs the connection between the ATC Tasks and the detailed

ATC Actions that the controller has to perform along with the cognitive channels used. This

connection is defined through an Operational Concept, which establishes the controller

behaviours according to this environment (Human Machine Interface, support tools, etc.). The

control events break down, within the Operational Concept, each of these events into tasks,

actions and channels, as it is shown in Figure 20. (SESAR-JU Complexity Management in En-

route, Step1: Consolidation of previous studies)

2.1.1. Statistical Approach (DSNA)

DSNA (Direction des Services de la Navigation Arienne) is the French ANSP, the company held and

delegated by the state to provide air navigation services within French controlled airspace. DSNA is also

a big supporter of SESAR objectives, and like AENA it holds the Project Manager position for a set of

researches held under SESAR umbrella.

Internally, DSNA has put the basis of the so called statistical approach to workload complexity, which is

backed by the use of artificial neural networks, algorithms inspired from the biological neurons and

synaptic links way of functioning.

The primary driver of the statistical approach was the need to find out whether and when a control sector

should be split into several elementary modules or on contrary more sectors should be merged together

(that is to assign these sectors to a single working position). One key point in answering these questions is

the finding that splitting and/or merging sectors are statistically related to the controller workload.

According to this principle, the studies hereafter have been led by DSNA prior to SESAR and have

focused on two themes: complexity assessment via the statistical approach and complexity reduction via

airspace reconfiguration proposals.

In order to function, this statistical approach proposed by DSNA makes use of six relevant metrics that

have been determined from previous studies and have been used to train the neural network.


The six metrics used are as follow:

1. Sector volume

2. Number of aircraft within the sector

3. The average vertical speed

4. Incoming flows with time horizons of 15 minutes

5. Incoming flows with time horizons of 60 minutes

6. The number of potential crossings with an angle greater than 20 degrees

As previously stated the complexity assessment algorithm uses a trained neural network with the six

metrics to predict the sector status based on the controller workload. The prediction issued from the

algorithm can fall in one of the three variants: the probability of having a low workload, a normal

workload or a high workload.

Depending on the predicted controller workload the decision of changing the airspace configuration either

by splitting the overloaded sectors and merging the under-loaded one can be taken.

The working principle can be seen below in Figure 2.2).

Figure 2.2 Complexity assessment with a neural network (SESAR complexity management in en-route)

In order to give the most accurate workload prediction, the neural networks parameters are tuned on

historical data from a wide variety of sectors, taking into account the traffic complexity, so it can

generalise to any en-route sector. This is an interesting feature, as the model will also give an indication

of when an elementary sector (which cannot be split) will get overloaded, extrapolating from overloads

occurring in wider control sectors that can be split. Another interesting aspect of the calibration on

historical airspace and air situation data resides in the fact that the model is potentially adaptable to

changes in working methods via recalibration and fine tuning of a few parameters.

To conclude, this method belonging to DSNA is an interesting approach, though further analysis shall be

undertaken.

The Neural

Network

Sector volume V

Number of aircraft within the sector Nb

Average vertical speed avg_vs

Incoming flow with time horizon 15 min F15

Incoming flow with time horizon 60 min F60

Number of potential trajectory crossing with an

angle greater than 20 degrees inter_hori

P high

P normal

P low


2.1.2. Other methods for complexity and workload

a) Dynamic Density

One of the most cited works in the literature when it comes to controller workload, which is at the same

time the first complexity indicator incorporating structural considerations along with the simple number

of aircraft, is the Dynamic Density of Laudeman et al. from NASA. The Dynamic Density is a

weighted sum of the traffic density (number of aircraft), the number of heading changes (> 15 degrees),

the number of speed changes (>0.02 Mach), the number of altitude changes (>750 ft), the number of

aircraft with 3DEuclidean distance between 0-25 nautical miles, the number of conflicts predicted in 25-

40 nautical miles. These factors are summed together using weighting factors that were determined by

showing different traffic scenarios to several controllers. Sridhar from NASA has developed a model to

predict the evolution of such a metric in the near future. Efforts to define Dynamic Density have

identified the importance of a wide range of potential complexity factors, including structural

considerations. However, the instantaneous position and speeds of the traffic itself does not appear to be

enough to describe the total complexity associated with the airspace. Some previous studies have

attempted to include structural consideration in complexity metrics, but have done so only to a restricted

degree. For example, the Wyndemere Corporation proposed a metric that included a term based on the

relationship between aircraft headings and dominant geometric axis in a sector. The importance of

including structural consideration has been explicitly identified in work at EUROCONTROL. In a study

to identify complexity factors using judgment analysis, Airspace Design was identified as the second

most important factor behind traffic volume.

b) COCA Project Complexity and Capacity project (COCA) was a major research project initiated by EUROCONTROL

Experimental Centre (EEC) at the end of year 2000 in the field of ATC Complexity, whose aim was to

provide the Agency with a functional model of ATC cognitive complexity and to describe the relationship

between capacity and complexity by means of accurate performance metrics. Due to various reasons this

work was abandoned but the reports and the studies made within this project are of a real value and in my

personal opinion deserve to be further analysed, since they provided promising results.

A quantitative approach was used to evaluate operational complexity intrinsic to MUAC traffic flows and

airspace environment characteristics. That approach consisted of first defining the complexity metrics

which could best describe the factors contributing to the complexity of MUAC sectors.


Those factors were defined considering both static (sector configuration and specific fixed aspects related

to the airspace environment) and dynamic (e.g. operational behaviour, traffic variability) data. The set of

elicited metrics was systematically evaluated for all MUAC sectors in each sector configuration that

occurred during both data collection phases. The results provided quantitative measurements of the

selected indicators and were used as the basis of the sector I/D cards.

The study proved that the method to calculate workload had, in many cases, a good correlation to the

controllers' reported perceived workload. However, additional work was needed to identify those

indicators and situations where the link was weakest, or missing, and incorporate them in an improved

workload calculation. Further, subjective results underscored the potentially critical role of factor

combinations, especially interactions involving a subset of so precursor factors.

c) CAPAN Method

Capacity Analyzer (CAPAN) is a tool and simulation method developed by EUROCONTROL to evaluate

the ATC sector capacities before or after new airspace organisations, implementation of new route

structures, new sectorisation, new procedures, etc. with the aim of providing possible solutions to

improve ATC Operations and increase sector capacity.

After analysing several CAPAN studies I identified some interesting findings that are of a real interest for

this research. Below, I summarise two of them.

a) Workload thresholds

This thresholds used here are the result of empirical experimentation and were validated and calibrated by

several real time simulation studies, and even conservative, they are generally accepted in the industry.

Threshold Interpretation Recorded working time during 1 hour

70% or above Overload 42 minutes or more

54% - 69% Heavy Load 32 - 41 minutes

42% - 53% Medium Load 25 - 31 minutes

18% - 41% Light Load 11 - 24 minutes

0% - 17% Very Light Load 0 - 10 minutes Table 2.3 CAPAN workload thresholds

The Executive Controller workloads are not directly proportional to traffic demand. It is clear from

(Figure1) that the function EC workload versus Traffic demand is not a linear function. A deeper analysis

shows that EC workload is more dependent on complexity of traffic than on demand.


The second conclusion proves that our initial assumption that workload is not a linear function, directly

dependent to traffic demand holds. On the other hand, the PC controller workload appears in direct

dependency with the number of a/c in the sector (due to various enablers: SYSCO, OLDI).

Figure 2.3 CAPAN workload analyses

Entry rate Executive Controller workload; 70% threshold Planning Controller workload

2.2. Complexity factors

As it was already stated in the beginning of this document complexity reasons are manifold and factors

determining it are both numerous and diverse. Most of ATC complexity studies in the past relied on

previous works and researches, however they have introduced as well new complexity factors or gave

different names to indicators that have previously been identified. Workload and ATC complexity studies

have been undertaken in the last 40 years. Hence, identifying all the complexity factors will be a very

demanding job. While for this study the real importance of some of the factors is low, this effort can be

considered by some as questionable. However, for the purpose of this study the author considers that

having a list of all the factors that are thought to affect ATC complexity is of a real help, as it can be

considered a pool from where the most important factors for this study can be identified.

In a study from University Politehnica of Bucharest, based on an analysis of a large number of

complex or difficult traffic scenarios and how did the ATCOs deal with them, the researchers

identified a number of 9 factors and 45 sub-factors that from were proven to be relevant in defining a

comprehensive function of complexity. The authors of the study introduced the idea of a nonlinear


complexity function and attempts to catch the perceived complexity in a quantitative measure as

accurately as possible.

In order to identify complexity factors several techniques were used. In an effort to identify complexity

factors, Mogford (1993, 1994), though not complementary, used two distinct techniques: direct and

indirect. Direct techniques use the results of verbal reports, questionnaires, and interviews to elicit

complexity factors. Indirect techniques use statistical techniques analysing controller judgments of the

relative complexity of different air traffic situations to determine potential complexity factors. Results

indicated the direct technique was an appropriate and adequate tool for identifying potential complexity

factors (Mogford 1994).

Complexity factors previously identified in the literature have included the distribution of aircraft in the

air traffic situation and properties of the underlying sector. Relevant characteristics of the underlying

sector have included sector size, sector shape, the configuration of airways, the location of airway

intersections relative to sector boundaries, and the impact of restricted areas of airspace. Coordination

requirements have also been identified as a factor in ATC complexity. This factor depends on the design

of the airspace and the structure that defines interactions between controllers in adjoining sectors.

Mogford (1995) has made a survey of the literature relating to complexity in air traffic control where he

presented the factors that by that time were thought to have an influence upon ATC complexity. An

exhaustive listing of various factors identified throughout the literature has been compiled by Majumdar

(2002) and further by the research done within EUROCONTROL COCA Project (2004).

Notice that there were great differences across the literature in how much detail was provided for

complexity factors. In some cases, factors were specified in great detail (including the measurement

methods and units such as total number of flights effecting a heading change of more than 15 degrees,

per hour). In other cases, factors were mentioned without any clear indication of how to capture them

operationally (e.g. Military activity). The occasional overlap between factors (below) reflects this

inconsistency. Details are provided where available.


In the table below you can find a list comprising a total of 108 complexity factors as identified by the

COCA team in [2].

1. Aerodromes, number of airline hubs

2. Aerodromes, total number in airspace

3. Aircraft mix climbing and descending

4. Airspace, number of sector sides

5. Airspace, presence/proximity of restricted airspace

6. Airspace, proximity of sector boundary

7. Airspace, sector area

8. Airspace, sector boundary proximity

9. Airspace, sector shape

10. Airspace, total number of navaids

11. Conflicts, average flight path convergence angle

12. Conflicts, degree of flight path convergence

13. Conflicts, number of aircraft in conflict

14. Conflicts, number of along track

15. Conflicts, number of crossing

16. Conflicts, number of opposite heading

17. Conflicts, total time-to-go until conflict, across all aircraft

18. Convergence, presence of small angle convergence routes

19. Coordination, frequency of coordination with other controllers

20. Coordination, hand-off mean acceptance time

21. Coordination, hand-offs inbound, total number

22. Coordination, hand-offs outbound, total number

23. Coordination, number aircraft requiring hand-off to tower/approach

24. Coordination, number aircraft requiring vertical handoff

25. Coordination, number flights entering from another ATC unit

26. Coordination, number flights entering from same ATC unit

27. Coordination, number flights exiting to another ATC unit

28. Coordination, number flights exiting to same ATC unit

29. Coordination, number of communications with other sectors

30. Coordination, number of other ATC units accepting hand-offs

31. Coordination, number of other ATC units handing off aircraft

32. Coordination, total number LoAs

33. Coordination, total number of handoffs

34. Coordinations, total number required

35. Equipment status

36. Flight entries, number aircraft entering in climb


37. Flight entries, number aircraft entering in cruise

38. Flight entries, number aircraft entering in descent

39. Flight entries, number entering per unit time

40. Flight exits, number aircraft exiting in climb

41. Flight exits, number aircraft exiting in cruise

42. Flight exits, number aircraft exiting in descent

43. Flight Levels, average FL per aircraft

44. Flight Levels, difference between upper and lower

45. Flight Levels, number available within sector

46. Flight time, mean per aircraft

47. Flight time, total

48. Flight time, total time in climb

49. Flight time, total time in cruise

50. Flight time, total time in descent

51. Flight type, emergency / special flight operations, number

52. Flow organisation, altitude, number of altitudes used

53. Flow organisation, average flight speed

54. Flow organisation, complex routing required

55. Flow organisation, distribution of Closest Point of Approach

56. Flow organisation, flow entropy/structure

57. Flow organisation, geographical concentration of flights

58. Flow organisation, multiple crossing points

59. Flow organisation, number of altitude transitions

60. Flow organisation, number of current climbing a ircraft proportional to historical maximum

61. Flow organisation, number of current descending aircraft proportional to historical maximum

62. Flow organisation, number of current level aircraft proportional to historical maximum

63. Flow organisation, number of intersecting airways

64. Flow organisation, number of path changes total

65. Flow organisation, routes through sector, total number

66. Flow organisation, vertical concentration

67. Other, controller experience

68. Other, level of aircraft intent knowledge

69. Other, pilot language difficulties

70. Other, radar coverage

71. Other, resolution degrees of freedom

72. Procedural requirements, number of required procedures

73. RT, average duration of Air-Ground communications

74. RT, callsign confusion potential

75. RT, frequency congestion

76. RT, frequency of hold messages sent to aircraft

77. RT, total number of Air-Ground communications


78. Separation standards (separation/spacing/standards)

79. Staffing

80. Time, total climb

81. Time, total cruise

82. Time, total descent

83. Traffic density, aircraft per unit volume

84. Traffic density, average instantaneous count

85. Traffic density, average sector flight time

86. Traffic density, localised traffic density / clustering

87. Traffic density, mean distance travelled

88. Traffic density, number flights during busiest 3 hours

89. Traffic density, number flights during busiest 30 minutes

90. Traffic density, number flights per hour

91. Traffic density, number of arrivals

92. Traffic density, number of current aircraft proportional to historical maximum

93. Traffic density, number of departures

94. Traffic density, total fuel burn per unit time

95. Traffic density, total number aircraft

96. Traffic distribution/dispersion

97. Traffic mix, aircraft type, jets versus propellers

98. Traffic mix, aircraft type, slow versus fast aircraft

99. Traffic mix, climbing versus descending

100. Traffic mix, military activity

101. Traffic mix, number of special flights (med, local traffic)

102. Traffic mix, proportion of arrivals, departures and overflights

103. Traffic mix, proportion of VFR to IFR pop up aircraft

104. Weather

105. Weather, at or below minimums (for aerodrome)

106. Weather, inclement (winds, convective activity)

107. Weather, proportion of airspace closed by weather

108. Weather, reduced visibility Table 2.4 Table of complexity factors (COCA project)

In 3.5 from the total of 108 complexity factors identified in the literature a number of 12 will be selected

for an in-depth study as they proved to be relevant for this research. Furthermore the metrics for them will

be derived.


3. ATC complexity model

This chapter will present en-details the original approach to complexity of this paper, where the author

had developed a conceptual mathematical function adapted to BucurestiACC Sectors that is thought to be

a better and more precise way to quantify ATC workload through the complexity point of view. As stated

in the first chapter, the complexity in ATC can be equally attributed to traffic characteristics, to

dynamicity of the traffic, to sector characteristics, and not in the last place to the controller abilities

(cognitive complexity). Nevertheless, the general experience from consulting previous studies is that there

are many methods and function that would contribute to a great improvement in Complexity

Management; improvement in ATM performance.

With this in mind and after an in-depth study of all the above mentioned methods and principle to

measure complexity and its additional workload to the controller, in the sequel it will be explain the need

to develop a new method to assess ATC workload complexity by making use of the main benefits of

previous researches.

3.1. Introduction

Nowadays the civil aviation industry in Europe is gradually moving towards a Free Route Airspace

concept of operation which is thought to increase the benefits of all the airspace users and to assist the

accomplishment of the Single European Sky Air Navigation research (SESAR) requirements and

expectations, like achieving an almost double airspace capacity by 2020 compared to 2005, reduced en-

Analysis of experimental data on ATC workload complexity

Documents

Transcript of Analysis of experimental data on ATC workload complexity