Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
E.02.39 EMERGIA D2.2 Report
Emergent Behaviour of Simulation Model
Document information
Project title EMERGIA - Powerful Emergent Behaviour In ATM
Project N° E.02.39.
Project Manager National Aerospace Laboratory NLR
Deliverable Name
Project Plan
Deliverable ID D2.2
Edition V1.0
Task contributors
National Aerospace Laboratory NLR
Abstract
One of the key innovations in SESAR2020+ is the introduction of a strategic Trajectory
Based Operation (TBO) layer. In this report for the first time rare emergent behaviour has
been studied for a ground-based future concept that makes use of both a strategic TBO layer
and a tactical resolution layer. The agent-based model studied has been derived from an
advanced airborne self-separation TBO model for which remarkably positive results have
been obtained through the iFly project. The current report documents the development of the
ground-based version of this so well performing airborne based TBO model, and the
systematic evaluation of this ground based TBO model through rare event Monte Carlo
simulations. This study reveals key challenges that remain to resolved in order for a ground-
based TBO model to accommodate very high traffic demands as safe as the airborne based
TBO model does.
Authoring & Approval
Prepared By
Name & organisation Position / Title Date
Henk Blom / NLR Project leader of EMERGIA
13 May 2014
Bert Bakker / NLR Key Expert in EMERGIA 13 May 2014
Reviewed By
Name & organisation Position / Title Date
Dennis Nieuwenhuisen Leader of WP3 in EMERGIA
7 October 2014
Frank Bussink Contributor to WP3 21 October 2014
Richard Irvine SESAR WP-E project officer
21 November 2014
Approved By
Name & organisation Position / Title Date
Henk Blom / NLR Project leader of EMERGIA
11 December 2014
Approval does not refer to approval by EUROCONTROL / SJU but approval by the project consortium. Document History
Edition Date Status Author Justification
V0.1 13 May 2014 Draft H. Blom 1st draft, inputs from D2.1
V0.2 16 June 2014 Draft H. Blom 2nd
draft, inputs from MSc
V0.3 19 June 2014 Draft H. Blom 3rd
draft, + SESAR ConOps
V0.4 26 June 2014 Draft H. Blom 4th draft, significant edits
V0.5 30 June 2014 Draft H. Blom 5th draft, significant edits
V0.6 22 August 2014 Draft H Blom 6th draft, ATCo activity added
V0.7 3 October 2014 Draft H. Blom 7th draft, 8 a/c added
V0.8 13 October 2014 Draft H. Blom Internal review incorporated
V0.9 17 October 2014 Draft H. Blom Minor final editing cycle
V1.0 11 December 2014 Final H. Blom SESAR WP-E review comments
V1.0u 31 December 2014 Final H. Blom Minor corrections
Page 3 of 96
TABLE OF CONTENTS
ACRONYMS ................................................................................................................................ 5
1 INTRODUCTION .................................................................................................................. 7
1.1 BACKGROUND ......................................................................................................................... 7
1.2 EMERGIA PROJECT ................................................................................................................. 7
1.3 THE OBJECTIVE OF THIS REPORT ................................................................................................. 8
1.4 THE ORGANIZATION OF THIS REPORT ............................................................................................ 8
2 THE NOVEL CONOPS: A3GROUND .................................................................................... 10
2.1 FROM A3 CONOPS TO A3GROUND (A3G) CONOPS ................................................................. 10
2.2 RBT UPDATING AND MTCDR IN THE A3G CONOPS ....................................................................... 11
2.3 TACTICAL RESOLUTION AND STCDR IN THE A3G CONOPS .............................................................. 13
2.4 A3G CONOPS VERSUS SESAR 2020+ CONOPS ........................................................................... 14
3 THE A3G MODEL ............................................................................................................... 21
3.1 A3G MODEL ASSUMPTIONS .................................................................................................... 21
3.2 AGENTS IN THE A3G MODEL .................................................................................................. 21
3.3 ATC GROUND SYSTEM ARCHITECTURE IN THE A3G MODEL......................................................... 23
3.4 INTERCONNECTED LPN’S OF THE ATC SYSTEM ............................................................................. 25
3.5 ATCO AS AGENT IN THE A3G MODEL ...................................................................................... 28
3.6 NEW COMMUNICATION SYSTEMS IN THE A3G MODEL ................................................................ 30
3.7 PILOT FLYING AS AGENT IN THE A3G MODEL ........................................................................... 33
3.8 IMPLEMENTATION AND VERIFICATION OF THE A3G CODE ........................................................... 35
4 MC SIMULATION OF 2 AIRCRAFT ENCOUNTERS ...................................................... 38
4.1 A3G BASELINE PARAMETER VALUES ....................................................................................... 38
4.2 MC SIMULATION RESULTS UNDER A3G BASELINE PARAMETER VALUES ......................................... 40
4.3 A3G SIMULATION RESULTS UNDER A3 BASELINE PARAMETER VALUES ......................................... 41
4.4 ADDITIONAL MC SIMULATION TESTS OF 2 A/C ENCOUNTERS ..................................................... 42
Page 4 of 96
4.5 TEST C: GLOBAL GNSS/GPS ............................................................................................... 44
4.6 TEST D: GLOBAL ADS-B FREQUENCY .................................................................................... 45
4.7 TEST E: GLOBAL ATC UPLINK FREQUENCY ............................................................................. 46
4.8 TEST F: AIRCRAFT GPS RECEIVER ......................................................................................... 47
4.9 TEST G: AIRCRAFT ALTIMETER .............................................................................................. 49
4.10 TEST H: AIRCRAFT ADS-B TRANSMITTER ............................................................................... 50
4.11 TEST I: ATC GROUND SYSTEM CORRUPTED ............................................................................. 52
4.12 TEST J: ATC GROUND SYSTEM FAILURE .................................................................................. 53
4.13 TEST K: GROUND ADS-B RECEIVER ...................................................................................... 54
4.14 TEST L: ATCO-TACTICAL MAXIMUM RESPONSE TIME ................................................................ 55
4.15 TEST M: ATCO-PLANNING MAXIMUM RESPONSE TIME .............................................................. 58
4.16 TEST N: ATC UPLINK TRANSMITTER ...................................................................................... 60
4.17 SELECTED PARAMETER VALUES FOR THE A3G MODEL ................................................................ 63
4.18 INTERPRETATION OF THE 2 AIRCAFT ENCOUNTER RESULTS OBTAINED ......................................... 64
5 MC SIMULATION OF 8 AIRCRAFT ENCOUNTERS ...................................................... 68
5.1 A3G SELECTED AND A3G BASELINE PARAMETER VALUES APPLIED TO 8/AC ENCOUNTERS ................ 68
5.2 A3G BASELINE PARAMETER VALUES, EXCEPT A VERY FAST PILOT RESPONSE .................................. 70
5.3 FINDINGS FOR 8 A/C ENCOUNTERS ........................................................................................ 72
6 RANDOM TRAFFIC SCENARIOS ................................................................................... 73
6.1 MONTE CARLO SIMULATION RESULTS FOR RANDOM TRAFFIC SCENARIOS ...................................... 73
6.2 MTCR AND STCR ACTIVITY FREQUENCIES FOR PILOT CREWS .................................................... 73
6.3 PREDICTED A3G CONOPS MTCR AND STCR ACTIVITY FREQUENCIES FOR PLANNING ATCO .......... 74
6.4 PREDICTED A3G CONOPS MTCR AND STCR ACTIVITY FREQUENCIES FOR TACTICAL ATCO ........... 76
7 CONCLUSION .................................................................................................................. 77
8 REFERENCES ................................................................................................................... 79
Page 5 of 96
ACRONYMS
4D
4 dimensional
a/c
aircraft
A3
Autonomous Aircraft Advanced
A3G
Autonomous Aircraft Advanced Ground
ABMS ACAS
Agent Based Modelling and Simulation Airborne Collision Avoidance System
ADS-B
Automatic Dependent Surveillance-Broadcast
AMFF ANP
Autonomous Mediterranean Free Flight Actual Navigation Performance
ANSP AP/FD
Air Navigation Service Provider Auto Pilot / Flight Director
AOC
Airline Operations Centre
ASAS
Airborne Separation Assistance System
ATC
Air Traffic Control
ATCo ATI
Air Traffic Controller Aeronautical Telecommunication Information
ATM
Air Traffic Management
CAA
Civil Aviation Authority
CAST
Commercial Aviation Safety Team
CD
Conflict Detection
CD&R CDTI
Conflict Detection and Resolution Cockpit Display of Traffic Information
CICTT
CAST/ICAO Common Taxonomy Team
ConOps
Concept of Operations
CPDLC
Controller Pilot Data Link Communication
CR
Conflict Resolution
DAG-TM
The Distributed Air/Ground Traffic Management
DCPN FDPS
Dynamically Coloured Petri Net Flight Data Processing System
FL
Flight Level
FMS
Flight Management System
ft.
foot
GNC
Guidance, Navigation and Control
GNSS
Global Navigation Satellite System
GPS GSHS
Global Positioning System General Stochastic Hybrid System
HMI
Human Machine Interface
ICAO
International Civil Aviation Organization
IPN
Interaction Petri Net
IPS
Interacting Particle System
IRS
Inertial Reference System
JAA
Joint Aviation Authority
LOS
Loss of Separation
LPN
Local Petri Net
MAC
Mid Air Collision
MC
Monte Carlo
Page 6 of 96
MSI
Minimum Separation Infringement
MTC
Medium Term Conflict
MTCD MTCDR
Medium Term Conflict Detection Medium Term Conflict Detection and Resolution
MTCR
Medium Term Conflict Resolution
n.a.
not applicable
NLR
National Aerospace Laboratory
NLR
Nationaal Lucht- en Ruimtevaart laboratorium
Nm
Nautical mile
NMAC OSED P-ASAS
Near Mid Air Collision Operational Services and Environment Description Predictive Airborne Separation Assurance System
PF
Pilot Flying
PN
Petri Net
PNF
Pilot-Not-Flying
R/T
Radio-Telephony
RA
Resolution Advisory
RBT RTA
Reference Business Trajectory Required Time of Arrival
RTCA
Radio Technical Commission for Aeronautics
SA SBT
Situation Awareness Shared Business Trajectory
SDCPN
Stochastically and Dynamically Coloured Petri Net
SES
Single European Sky
SESAR
Single European Sky ATM Research
SMC
Sequential Monte Carlo
SSA
Self-Separating Airspace
SSR
Secondary Surveillance Radar
STC
Short Term Conflict
STCA
Short Term Conflict Alert
STCD STCDR
Short Term Conflict Detection Short Term Conflict Detection and Resolution
STCR
Short Term Conflict Resolution
SWIM
System Wide Information Management
TBO
Trajectory Based Operations
TCP
Trajectory Change Point
TOPAZ WP
Traffic Organization and Perturbation AnalyZer Work Package
Page 7 of 96
1 INTRODUCTION
1.1 Background
Following [SESAR, 2007], the SESAR concept of operations beyond 2020 (SESAR2020+)
involves a series of changes relative to current Air Traffic Management (ATM). Central to these
changes is the paradigm shift that aircraft should fly according to agreed conflict free 4D
trajectory plans which are made known to all actors involved as Reference Business Trajectories
(RBT’s). A big unknown in this RBT framework is how everything works under various kinds of
uncertainty, as a result of which one or more aircraft may not realize their RBT’s. There are
several categories of uncertainty (including unexpected disturbances) that cannot be totally
avoided, such as: Meteorological uncertainties; Data related uncertainties; Human related
uncertainties; and Technical systems related uncertainties.
In principle the SESAR2020+ ConOps has been designed to take care of these kinds of
uncertainty through the possibility to revise 4D trajectory plans, and also to allow air traffic
control to issue tactical flight instructions to pilots if the 4D planning layer has run out of time.
Although these tactical instructions are quite similar to the established way of working by an air
traffic controller, there also are significant differences.
Under SESAR2020+ an air traffic controller is expected to handle significantly more aircraft in
its sector. Therefore the SESAR2020+ ConOps also foresees dedicated tactical decision support
tools for air traffic controllers. The key issue is how to optimize the socio-technical collaboration
between the 4D planning layer and the tactical layer in order to manage air traffic most
effectively while taking into account the various uncertainties.
In conventional ATM, mediumterm planning is provided by the planning controller, flight crews
and their Flight Management Systems (FMS), whereas the tactical loop is formed by the tactical
controller and flight crews. Thanks to decades of evolutionary developments, the collaboration
between these two layers has been optimized. For SESAR2020+ a similar optimization of the
novel 4D planning layer with the tactical layer is needed. Because the collaboration between
these layers involves dynamic interactions between human decision makers, technical support
systems, aircraft evolution, weather and other uncertainties, the combined effects result in types
of emergent behaviours that cannot be predicted from the sum of the elemental behaviours. This
can easily lead to negative emergent behaviours at time scales that remain invisible using
established evaluation techniques.
1.2 EMERGIA project
During large European research projects HYBRIDGE and iFly, innovative complexity science
techniques have been developed and applied to airborne self-separation concepts of operations.
In order to understand and improve the emergent behaviours of SESAR2020+ at multiple time
scales, the EMERGIA project will use these innovative complexity science techniques. This way
EMERGIA aims to dramatically reduce the risks that negative emergent behaviours have to be
Page 8 of 96
repaired at a late stage, at huge operational costs, and will shorten the period needed to optimize
the system architecture and design of SESAR2020+.
The most advanced airborne self-separation concept of operations studied within iFly, makes use
of similar 4D planning and tactical layers as SESAR2020+, though fully airborne. This ConOps
is referred to as the A3 model. Based on rare event Monte Carlo simulations of this A3 model,
conducted within the iFly project, [Blom & Bakker, 2011a,b, 2012] have shown that in an
advanced airborne-self separation TBO concept the 4D planning and tactical layers can work so
well together that this leads to very powerful positive emergent behaviours, even beyond
expectations of the concept developers. As a result of these powerful positive emergent
behaviours, the advanced airborne self-separation concept considered can safely accommodate
very high enroute traffic demands. This raises the question whether these powerful emergent
behaviours can be maintained while moving the 4D planning layer and the tactical layer to the
ground, as is the case with SESAR2020+. The objective of EMERGIA is to answer this research
question [EMERGIA, 2012].
1.3 The objective of this report
The original EMERGIA plan was to address the above formulated research question in three
steps. The first step is to develop a ground-based version of the A3 model (shortly referred to as
A3G model), to compare this to the SESAR2020+ ConOps, and to use the innovative complexity
science techniques to identify the emergent behaviours of this A3G model. The second step is to
compare these emergent behaviours to the powerful positive emergent behaviours of the
advanced airborne self-separation ConOps, and to study the possible improvement of the A3G
model in case of significant difference in emergent behaviours. The third step is to evaluate the
improved A3G model on its emergent behaviours, again by using the innovative complexity
science techniques.
Hence, according to the initial EMERGIA plan, the comparison of the A3G model results versus
the A3 model results would only be done during step 2. However it turned out to be more
practical to follow another approach. The idea behind this new approach is that by changing
appropriate parameter values in the A3G model it should be possible to arrive at the same
emergent behaviour results as those found for the A3 model. This novel approach however
required that a regular comparison between the A3 model and the A3G model was made already
during step 1, rather than delaying such comparison to step 2. Therefore, the current report
presents the results obtained during step 1 as well as the results of the comparison against the A3
model behaviour planned for the first half of step 2.
1.4 The organization of this report
This report is organized as follows. First, in section 2 it is described how the A3 model is
systematically used to develop an ground-based version of it, i.e. the A3G model. Also a
systematic comparison of the A3G model is made with the SESAR2020+ ConOps.
Subsequently, Section 3 presents the systematic development and verification of an agent-based
Monte Carlo simulation model of the A3G model.
The systematic evaluation of the A3G model regarding the feasibility of getting its emergent
behaviour the same as it has been seen for the A3 model, is addressed in sections 4-6. In section
Page 9 of 96
4, for 2 aircraft encounter scenarios it is considered under which A3G model parameter values
the behaviour is the same as it has been observed under the A3 model. Subsequently, in Section
5 it is investigated whether there are additional requirements on the model parameter settings
under the eight aircraft encounter scenarios. Finally, in Section 6, a systematic study is
conducted regarding the task load of pilots and controllers under the A3G model relative to those
under the A3 model. Finally, section 7 draws conclusions.
For sections 2-4, some material has been used from [Nieskens, 2014].
Page 10 of 96
2 THE NOVEL CONOPS: A3GROUND
In subsection 2.1 the novel concept of operations (ConOps) is described; it is a ground-based
version of the A3 ConOps. Subsequently in subsection 2.2 a comparison is made of this novel
A3G ConOps versus the SESAR2020+ ConOps [SESAR-JU, 2013].
2.1 From A3 ConOps to A3Ground (A3G) ConOps
Within the iFly project, NASA’s advanced ConOps [NASA, 2004] has gratefully been used as
starting point for the development of an advanced airborne self separation concept for en-route
traffic under the name A3 ConOps [iFly D1.3, 2010]. This A3 ConOps intentionally addresses
the hypothetical situation of 100% well equipped aircraft. For this A3 ConOps an Operational
Services and Environmental Description (OSED) is also available [iFly D9.1, 2009].
Similar to the SESAR2020+ TBO concept, the A3 ConOps adopts TBO in the sense that each
aircraft maintains a 4D trajectory intent that is shared with all other aircraft. According to
SESAR2020+ terminology [SESAR, 2007], the 4D trajectory intent of an aircraft is referred to
as a Reference Business Trajectory (RBT). However, RBT management in the A3 ConOps is
done by each aircraft itself, without any support from air traffic control at the ground. Each
aircraft is assumed to be equipped with the same dedicated ASAS system which is monitoring
the surroundings and helps the flight crew to detect and resolve conflicts.
The A3G ConOps is an adaptation of the A3 ConOps. The A3G abbreviation is short for
A3Ground. In the A3 ConOps the separation was managed by the aircraft. In the A3G ConOps
the responsibility for separation assurance is moved back to ATC. Hence the 4D trajectory plans
and tactical resolutions are provided by ground-based ATC.
Figure 2.1 gives a graphical presentation of A3 ConOps vs. A3G ConOps. At the left side is the
A3G ConOps where separation is controlled by ATC. At the right side is the A3 ConOps where
the pilots are responsible for self separation.
Figure 2.1: Graphical representation of A3G ConOps (left) and A
3 ConOps (right) (Cuevas, et
al., 2010)
Similar as in [NASA, 2004], A3’s uses two layers in the detection and resolution of potential
conflicts: the RBT layer and the tactical resolution layer. The RBT layer takes care in making
Page 11 of 96
updates of the RBT in case of a medium term conflict. The ASAS support of the RBT layer
consists of a Medium Term Conflict Detection and Resolution (MTCDR) support system. The
tactical layer takes care of resolving short term conflicts. The ASAS support of the Tactical layer
consists of a Short Term Conflict Detection and Resolution (STCDR) support system.
Because the development of proper working MTCDR and STCDR support systems has been a
major effort within the iFly project, and these support systems have proven to work well, the
proposal is to reuse these MTCDR and STCDR support systems, with one major difference: now
they are going to be used as support systems for ATC instead of flight crews. In addition to this,
in the A3G ConOps the ATC system will maintain a database containing all currently active
RBT’s.
For each aircraft, MTCDR supports the controller in identifying 4D trajectories which are
conflict-free (i.e. centre lines stay 5NM or 1000 ft apart) with the currently active RBT’s of
higher priority aircraft over a time horizon of at least 15 minutes. Each time MTCDR detects a
medium term conflict between any of the currently active RBT’s in the ATC system database,
then MTCDR tries to resolve this through determining a new conflict-free 4D trajectory for the
aircraft having lower priority. The priority of an aircraft is primarily determined by the
remaining distance to destination. Conflict-free also means that the 4D trajectory does not create
a new conflict with an RBT of any of the other aircraft that have higher priorities. Upon
acceptance of such new 4D trajectory by the controller, it is uplinked to the appropriate aircraft
and evaluated by the flight crew. Upon acceptance by the flight crew this 4D trajectory plan is
entered into the FMS and downlinked to the ATC system as the aircraft’s new RBT. In the ATC
system this downlinked RBT is then stored in the database of currently active RBT’s.
STCDR provides tactical maneuver support to a controller for conflict resolution with a time
horizon of 3 minutes, at a separation criterion of 5Nm/900ft. When STCDR detects a potential
infringement of these separation criteria, then STCDR proposes tactical resolution maneuvers to
the controller for each of the aircraft involved. The controller can select one of these tactical
resolution maneuvers and subsequently instructs the corresponding flight crew to implement this
tactical maneuver. This tactical maneuver instruction is then also inserted in the ATC database as
a correction to the corresponding RBT.
In the A3 ConOps there is an emergency procedure for the crew in case an aircraft suffers from
technical problems. Within the A3G ConOps however, the pilot has to inform ATC about an
aircraft emergency situation. Subsequently ATC should start to handle this problem. The current
A3G ConOps does not yet describe what ATC should do.
2.2 RBT updating and MTCDR in the A3G ConOps
Similar as in the A3 ConOps, in the A3G ConOps an RBT prescribes multiple waypoints which
can be inserted by the pilot in the FMS, directing the aircraft to its end goal.
In Figure 2.2 the new procedure for RBT updating in the A3G ConOps is presented. In this
procedure the ground-based ATC system and the ATCo are incorporated.
Page 12 of 96
Figure 2.2: RBT updating in the A3G ConOps
The procedure is initiated on the ground by the ATC system. The MTCDR support system of
ATC detects a medium term conflict and will then try to generate a new conflictfree trajectory
based on the available intent information of all aircraft. This conflict-free trajectory is proposed
as candidate RBT update to the planning ATCo (ATCo-P). The ATCo-P will check if the
proposal is OK or not. If the ATCo-P accepts the proposal, then it is submitted to the
corresponding aircraft through the ATC Uplink Transmitter. The Pilot Flying will check the
given RBT update and when approved will insert this in the FMS, and the aircraft will follow
this updated RBT. Finally the aircraft will broadcast the updated RBT from its FMS to the ATC
ground system using ADS-B or ADS-C. Upon reception this received RBT is used to update the
RBT data in the ATC ground system.
The above described procedure for RBT updating may also be used to let the FMS guide an
aircraft back to its initial path after a tactical resolution manoeuvre. In such case the RBT
updating consists of a conflict-free 4D trajectory that brings the aircraft back to its goal.
In the MTCDR support system used within the A3 ConOps, the selected conflict resolution
approach was based on Velocity Obstacles [Fiorini & Schiller, 1998; Abe at al., 2001]. Velocity
Obstacles (also known as Collision Cones) based conflict resolution means that an aircraft stays
away from the set of courses and velocities that lead to a predicted conflict with any other
aircraft. In airborne self-separation research, such Velocity Obstacles approach has been referred
to as Predictive ASAS [Hoekstra, 2001]. At this moment the Velocity Obstacle approach is
limited to horizontal maneuvering only.
Complementary to the choice of Velocity Obstacle based conflict resolution, the following
implementation principles have been adopted for the MTCDR support system:
+ MTCDR detects planning conflicts (5Nm/1000ft) 10 min. ahead and subsequently determines
a 4D trajectory plan that is conflict free over a horizon of 15 min.
+ Aircraft nearest to destination are given priority over other.
+ Aircraft with lowest priority are assumed to make its 4D plan conflict free (15 min ahead) with
all other plans.
+ If there is no feasible conflict free plan then rather than doing nothing, it is better for the
MTCDR to identify a plan that has a minimal undershooting of the 5Nm/1000ft criterion and
does not create a short term conflict.
ATC Ground
System
New 4D
trajectory planPlanning ATCo
Pilot-Flying-i
ATC Uplink
Transmitter
FMS-i
4D
trajectory plan
4D
trajectory plan
OK / Not-OK
ADS-B
transmitter-iRBT a/c-i
RBT a/c-i
OK / Not-OK
a/c receiver-i4D
trajectory plan
4D
trajectory plan
Page 13 of 96
+ Upon approval by the controller a non-conflict-free 4D trajectory plan is uplinked to the
aircraft together with a “Handicap” message. For the flight crew this handicap message means
that the priority of its aircraft has been increased, and that the controller will resolve the
remaining conflicts with the help of those aircraft having now a lower priority. Upon acceptance
by the flight crew, the 4D plan is entered into the FMS, and it is downlinked as the new RBT to
ATC, again together with the Handicap message. This new RBT are stored in the ATC database
together with the Handicap message.
Using the above principles, for each aircraft the MTCDR computes an RBT advisory by
determining a sequence of Trajectory Change Points (TCP’s) with minimum turning angles (to
the left or to the right) such that there are no predicted conflicts remaining with any aircraft
which has higher priority and which is within the MTCDR horizon. If there is no minimum
turning angle possible below a certain value φM, max, then the turning angle below φM, max is
identified which does not create a short term conflict and provides the lowest undershooting of
the minimum spacing criteria of 5Nm/1000ft between the RBT’s. In that case ATC assumes the
corresponding aircraft to be handicapped. As soon as the advised MTCDR advisories have been
accepted by the controller and the pilot, then they are implemented in the FMS and downlinked
to ATC together with the handicap message.
2.3 Tactical resolution and STCDR in the A3G ConOps
When a short term conflict is detected its resolution through RBT updating would take too much
time. Hence a faster tactical resolution process is necessary. Just as in the A3 ConOps a tactical
resolution is based on aircraft states and if available also on intent information. A tactical
resolution consists of an immediate heading change or a height change. In Figure 2.3 the tactical
resolution process as used in the A3G ConOps is presented.
Figure 2.3: Short Term Resolution (STC) process in the A3G ConOps.
The tactical resolution process starts with the detection of a short term conflict by the STCDR
support system of ATC. This STCDR will then automatically determine a possible tactical
resolution in terms of a heading or height change. Because the time horizon is short, this tactical
resolution is open loop, i.e. it does not include a back-to-goal maneuver. The proposed tactical
ATC Ground
System
Proposed
Heading
/ height
change
Tactical ATCo
Pilot-Flying-i
ATC Uplink
Transmitter
Aircraft
Guidance
Heading /
height
change
instruction
Heading / height
change
New heading/height
Manual
Heading /
height
change
a/c receiver-i
Heading /
height
change
ADS-B
transmitter-i
Current
heading /
height
Page 14 of 96
resolution is shown to the ATCo-Tactical (ATCo-T). The ATCo-T verifies the proposed
resolution, and may reject or accept it. If accepted it is sent to the corresponding aircraft through
the ATC uplink transmitter (CPDLC message).
Upon receiving the CPDLC message, the Pilot flying will implement the tactical resolution by
switching the aircraft from FMS to manual (tactical Auto Pilot / Flight Director) mode and
subsequently implement the given heading or height change. Subsequently ADS-B broadcasts
the slowly changing heading or height to the ATC ground system.
Simultaneously with sending the tactical resolution through CPDLC, the ATCo-T inserts the
instructed heading or height change in the ATC ground system. A side-effect of this is that the
actual behaviour of the aircraft will happen with some delay relative to the information in the
ATC ground system. This allows the ATC system to anticipate on the proposed heading change,
because it is already aware of the oncoming heading or height change of the aircraft. By directly
updating the intent information before the aircraft actually has changed its heading, the detection
and resolution of other short term conflicts works more efficiently.
The specific implementation principles adopted for the STCDR support system are at this
moment directed to horizontal maneuvers only:
+ STCDR detects conflicts (5Nm/900ft) 3 min. ahead and subsequently determines a course
change into a direction that is conflict free over a horizon of 3 min. plus 1 min.
+ Short term conflict resolution is also based on Velocity Obstacles approach.
+ When a short term conflict is detected between two aircraft, then agent-based STCDR
identifies two conflict-free tactical maneuver options, one for each aircraft. It is up to the
controller to select one of the proposed tactical maneuver options, and then to instruct this
maneuver to the applicable flight crew, and to enter this as an RBT modification in the ATC
database.
+ If there is no feasible alternative, then rather than doing nothing it is better to choose a tactical
maneuver which minimizes the undershooting of the minimum tactical separation criterion.
+ Upon approval of the crew, the aircraft downlinks its new course, which allows the ATC
system to verify that the instruction has been implemented well.
Using the above principles, STCDR proposes resolution course as the minimum turning angle (to
the left or to the right) such that there are no predicted conflicts remaining with any aircraft and
which is within the short term horizon. If there is no minimum turning angle possible below a
certain value φS, max, then the turning angle below φS, max is identified which provides the lowest
undershooting of the minimum separation criteria.
2.4 A3G ConOps versus SESAR 2020+ ConOps
The SESAR2020+ ConOps [SESAR, 2007, 2012] aims for a Trajectory Based Operation (TBO),
in the sense that aircraft should fly according to agreed conflict-free 4D trajectory plans which
are made known to all actors involved as Reference Business Trajectories (RBT’s).
Well ahead of take-off by an aircraft, its airline will publish a Shared Business Trajectory (SBT).
Before take-off, this SBT is agreed between Airline and ATM, becomes registered as a
Reference Business Trajectory (RBT), and is distributed through System Wide Information
Page 15 of 96
Management (SWIM). After take-off, this RBT is updated and down linked by the pilot to ATC
using ADS-B out, and when it is accepted by both pilot and ATC it will be registered as an
Update in SWIM. From then on, it is an active RBT. Every stakeholder will have access to the
RBTs in SWIM.
If during the flight there is any change or delay (e.g. due to significant wind deviations from the
predictions) then an RBT updating process will be conducted with the active involvement of
controllers and pilots concerned [SESAR, 2007, 2012]. Although there is agreement about the
need for such RBT updating process there are multiple views of how this should be done under
SESAR2020+. The consensus is that when there is sufficient time, then an updated RBT is being
produced by the aircraft concerned. In this case the role of ATC is to timely inform the aircraft
about applicable constraints. Because this exchange and verification of information between
ATC and aircraft crews takes significant time, there also is consensus that ATC should propose
an updated RBT themselves when time is too short.
Because the SESAR2020+ ConOps is a work in progress, it was felt to be most relevant to take
into account SESAR2020+ ConOps developments agreed within SESAR-JU. In consultation
with SESAR-JU, it has been decided that SESAR-JU’s Preliminary OSED_2 report [SESAR-JU,
2013] forms the most up to date reference document for the SESAR2020+ ConOps for use
within EMERGIA.
Regarding ASAS, on page 60 of the project P04.07.02 report [SESAR-JU, 2013] it is explicitly
described that ASAS aspects are out of scope, because other SESAR-JU projects address various
ASAS topics, such as:
+ P04.07.04a “ATSA-ITP Pioneer trials”;
+ P04.07.04b “ASAS-ASEP Oceanic Applications”;
+ P04.07.05 “En Route Trajectory and Separation Management – ASAS Separation (Cooperative
Separation)”;
+ P05.06.06 “ASAS Sequencing and Merging”.
The aim of this subsection is to provide a systematic comparison of the A3G ConOps against this
SESAR2020+ ConOps, as a result of which similarities and differences are identified. This
comparison is organized in three steps.
Step 1 compares their scopes.
Step 2 compares their 4D trajectory layer
Step 3 compares their tactical resolution layer.
Step 1: Comparison of scopes
Table 2.1 gives an overview of the main scoping issues for the two ConOps considered.
Table 2.1 Comparison of scoping issues
Page 16 of 96
Aspect SESAR2020+ A3G ConOps
Airspace En Route and TMA En Route
Traffic demand 1.22x 2010 traffic 3x 2005 traffic ~ 3x 2010 traffic
RBT based operation Yes Yes
RBT equipped aircraft 40% 100%
SWIM Yes Yes
ASAS No ASAS use by pilots No ASAS use by pilots
ACAS Improved TCAS Improved TCAS
The main similarities are RBT approach, SWIM and the no use of ASAS by pilots. The main
differences concern the percentages of fully equipped aircraft, the traffic demands, and the type
of airspace.
The 100% equipment level assumed within the A3G ConOps has its rationale in the objective of
the EMERGIA project. It simply will be unrealistic to expect that the remarkably positive
emergent behaviours identified for the A3 ConOps can be realized with not fully equipped
aircraft. Hence from an EMERGIA project perspective this difference is less relevant, although it
will have posed extra challenges to the designers of the SESAR 2020+ ConOps.
For the higher traffic demand (about a factor 2.5) of the A3G ConOps and the restriction to En-
route airspace it has been shown that an agent based model of the A3 ConOps produces the very
positive emergent behaviour we are looking for in a ground based ConOps model.
Step 2: Comparison of 4D trajectory layers
Table 2.2 gives an overview of the main 4D trajectory layer based issues for the two ConOps
considered.
Table 2.2 Comparison of 4D trajectory layer issues
Aspect SESAR2020+ A3G ConOps
Page 17 of 96
Separation Minima 5 NM / 1000 ft 5 NM / 1000 ft
Time Horizon 25 - 8 minutes 15 – 3 minutes
4D trajectories RBT sharing RBT sharing
Responsible Planning Controller Planning Controller
TRACT Subliminal speed
advisories through CTO’s
Not considered within A3G
Conformance Monitoring
MONA for PC MONA for PC
Conflict Detection MTCD Medium term conflict detection
Conflict Resolution MTCD probing by PC MTCDR based proposals to PC
4D Conflict Resolution Architecture
None; based on mental model of PC
Distributed architecture, i.e. conflict resolution algorithm works concurrently for each aircraft.
4D info to aircraft CPDLC CPDLC
Pilot role Reject OR Accept and implement
Reject OR Accept and implement
4D trajectory downlink ADS-C ADS-B
The main differences in the 4D trajectory layer are: the shorter time horizon of A3G, No
subliminal speed advisories in A3G, and conflict resolution is supported by algorithms instead of
MTCD probing by ATCo-P.
The subliminal speed advisories could very well be integrated in an extended version of the A3G
ConOps, which may be a sound option for an improved next A3G version. In such case it also
would make good sense to increase the upper value of the time horizon for this next A3G version
to the 25 minutes of SESAR2020+.
Regarding the lower value of the time horizon, it is important to notice that according to
[SESAR-JU, 2013] it remains to be determined what the optimal prediction time horizon is to
Page 18 of 96
define the split between the RBT layer and the Tactical resolution layer. From this perspective it
is quite relevant that for the A3 Conops significant experience has been gained regarding this
design aspect. For example in [Meulenbelt, 2012], it has been shown that a decrease of the
splitting value below 3 minutes leads to a deterioration, while an increase above the 3 minutes
does not lead to an improvement. That’s why for the A3 ConOps the splitting value has been set
at 3 minutes in order to give the RBT layer as much time as possible to resolve as many
conflicts as is possible, and thus leaving as few as possible to the Tactical resolution layer.
Hence the same splitting time value of 3 minutes has been adopted for the A3G ConOps.
Regarding the SESAR2020+ MTCD supported resolution by the ATCo-P, there are large
differences with the 4D resolution approach in the A3G ConOps. However, in order to maintain
the powerful emergent behaviour of the A3 model, the specific choice made for the A3G
ConOps follows from the principle in staying as close as is possible to the architecture of the
MTCDR in the A3 ConOps. Moreover, the A3G ConOps aims to accommodate a 2.5 times as
high traffic demand than SESAR2020+. Such factor of 2.5 implies two complementary
challenges: 1) there are far more conflicts to be resolved, and 2) the resolution of each conflict
involves more aircraft and is therefore more complex.
Step 3: Comparison of tactical resolution layers
Table 2.3 gives an overview of the main 4D trajectory layer based issues for the two ConOps
considered.
Table 2.3 Comparison of Tactical layer issues
Aspect SESAR2020+ A3G ConOps
Separation Minima 5 NM / 1000 ft 5 NM / 1000 ft
Time Horizon 8 - 6 minutes 3 - 0 minutes
Type of instructions Closed loop heading/height change OR Open loop
heading/height change
Open loop heading/height change
OR
Back to 4D trajectory clearance
Responsible Tactical Controller Tactical Controller
Surveillance SSR Mode-S and ADS-B out SSR Mode-S and ADS-B out
Conformance Monitoring MONA for TC MONA for TC
Page 19 of 96
Conflict Detection MTCD and STCA Short Term Conflict detection
Conflict Resolution MTCD probing by TC STCDR based proposals to TC
Tactical Conflict
Resolution Architecture
None; based on mental model of TC
Distributed architecture, i.e. conflict resolution algorithm works concurrently
for each aircraft.
Sequence of pair-wise resolutions
TC decides on basis of Safety, Geometry, Queue management
FIFO (of proposed resolutions)
ATCo – Pilot
communication
R/T CPDLC
Pilot role Reject OR Accept and implement
AND Reply
Reject OR Accept and implement
through Control Panel AND Confirm
Insertion of tactical instruction in ATC
system
Simultaneously with R/T message Simultaneously with CPDLC message
The main differences in the tactical layer are: Shorter time horizon for A3G, Open loop type of
tactical instruction under A3G, No use of SSR mode-S for surveillance by A3G, Algorithm
based conflict resolution by A3G, and ATCo-Pilot communication using CPDLC instead of R/T.
The rationale of the shorter time horizon has already been explained before. Related to this
shorter time horizon, tactical instructions always are of the open-loop type, which means that the
back-to-goal aspect can be resolved through an RBT update with support of the ATCo-P.
Regarding the SESAR2020+ MTCD supported resolution by the ATCo-T, there are large
differences with the Tactical resolution approach in the A3G ConOps. However, in order to
maintain the powerful emergent behaviour of the A3 model, the specific choice made for the
A3G ConOps follows from the principle in staying as close as is possible to the architecture of
the Tactical layer in the A3 ConOps. Moreover, the A3G ConOps aims to accommodate a 2.5
times as high traffic demand than SESAR2020+. Such factor of 2.5 implies two complementary
challenges: 1) there are far more conflicts to be resolved, and 2) the resolution of each conflict
involves more aircraft and is therefore more complex.
Page 20 of 96
Regarding R/T messages between ATCo-T and Pilots, it may be demanding to continue this
under a 2.5 higher traffic demand. Hence it seems to make good sense that the A3G ConOps has
switched to CPDLC.
Page 21 of 96
3 THE A3G MODEL
In this section the A3G model is presented. First, the main A3G model assumptions are listed in
subsection 3.1. Next, in subsection 3.2, an overview of the agents in the A3G model is presented,
with the focus on the newly added Local Petri Nets (LPNs). In subsection 3.3 to 3.7 the newly
and adjusted agents are shown in more detail by presenting the structure of their interconnected
LPNs. In subsection 3.8 the phased implementation of the A3G model is presented.
3.1 A3G model assumptions
In developing the A3G model, the following model assumptions have been adopted:
A1. In the A3G model all aircraft are identical and fly at the same level with the same speed.
A2. In the A3G model no emergency situations are modelled.
A3. In the A3G model no SSR radar data is assumed to be available to ATC.
A4. In the A3G model the 4D plan in Flight Data Processing System (FDPS) is considered to be
unreliable when ADS-B messages about the RBT in the FMS are not received.
A5. In the A3G model no ground based navigation support is available, i.e. navigation is based
on Global Navigation Satellite System (GNSS) and Inertial Reference System (IRS) only.
The consequences of these A3G model assumptions shall be taken into account later on when
arguing about the results obtained for the A3G model.
3.2 Agents in the A3G model
This subsection provides an overview of the agents in the A3G model. All agents used in the A3
model are also incorporated in the A3G model. In the A3G model the following agents are
present:
Aircraft-i
Pilot-Flying-i
Pilot-Not-Flying-i
Airborne Guidance, Navigation and Control (GNC) systems-i
Air Traffic Control (ATC) Ground System
Air Traffic Controller (ATCo)
Environment
It should be noticed that this model is an initial one which does not (yet) incorporate Weather,
Airborne Collision Avoidance System (ACAS) or Airline Operations Centre (AOC).
The Petri net formalism supports a compositional specification approach, which means that
first for each agent particular local Petri nets are being developed using agent specific expert
Page 22 of 96
knowledge, and without the need to bother about the connections between the agents. Once this
has been done, the interactions between these local Petri nets are being developed. A listing of
local Petri nets per agent is given in Table 3.1.
Table 3.1: All agents and the corresponding number of LPN's in the A
3G model
Aircraft-i local Petri nets:
o Type
o Evolution mode
o Engine system mode
o Navigation system mode
o Emergency mode
Pilot-Flying-i (PF) local Petri nets:
o State Situation Awareness
o Intent Situation Awareness
o Goal memory
o Current goal
o Task performance
o Cognitive mode
Pilot-Not-Flying-i (PNF) local Petri nets:
o Current goal
o Task performance
Airborne GNC-i local Petri nets:
o Indicators failure mode for PF
o Engine failure mode for PF
o Navigation failure indicator for PF
o ADS-B receiver failure indicator for PF
o ADS-B transmitter failure indicator for PF
o Indicator failure mode for PNF
o Guidance mode
o Horizontal guidance configuration mode
o Vertical guidance configuration mode
o FMS Intent
o Airborne GPS receiver
o Airborne Inertial Reference System (IRS)
o Altimeter
o Horizontal position processing
o Vertical position processing
o Regular Broadcast FMS Intent
o Reguar Broadcast aircraft State
o ADS-B transmitter
o ADS-B receiver
o ATC Uplink receiver
o MTCR/STCR audio alert
ATC Ground System:
Page 23 of 96
o ADS-B ground receiver
o ADS-B receiver mode
o ATC uplink transmitter
o System mode
o State & Intent
o Conformance Monitoring
o Conflict Detection & Management -i
o Resolution Mode -i
o STCR Advisory -i
o MTCR Advisory -i
o Back2Goal -i
Air Traffic Controller:
o Air Traffic Controller
Environment:
o Global Navigation Satellite System (GNSS)
o Global ADS-B ether frequency
o ATC uplink frequency occupied
o Weather
The resulting model comprises 50 different local Petri nets. With the exception of 11 ATC
system and Environment local Petri nets, each local Petri net is copied for each aircraft in the
model. Hence, for N aircraft, there are 39N+11 local Petri nets in the A3G model. Table 3.2
gives an overview of the agents and LPNs that were not in the A3 model.
Table 3.2: LPNs and corresponding agent added to A3 model to obtain A
3G model
Agent Local Petri Net
Airborne GNC systems: communication systems-i ATC Uplink receiver-i
ATC ground system ADS-B ground receiver mode
ATC Uplink transmitter
Back-to-Goal-i
Air Traffic Controller ATCo-Tactical
ATCo-Planning
Environment Global ATC Uplink frequency
3.3 ATC ground system architecture in the A3G model
In this subsection the internal structure of the ATC ground system agent in the A3G model is
presented.
The ATC ground system is designed using the ASAS from the A3 model. The internal LPN
structure of the ASAS remained the same in order to obtain similar results. The agent ASAS
consists of 10 LPN’s. These can be categorised in two tasks:
Surveillance and conformance monitoring systems
Conflict Detection and Resolution advice generation systems
Page 24 of 96
This A3 model structure is re-used in order to make it possible that the A3G model can produce
the same positive emergent behaviour as the A3 model . The resulting architecture of the ATC
ground system in the A3G model is shown in Figure 3.1. ATC-System
States, identity and intents of all aircraft
(LPN 5-10)
ATC-System-Other
i
(LPN 1-4)
ATC-CDR-i
...1k
(# of aircraft)...
ATC-CDR-iATC-CDR-i
Figure 3.1: Schematic overview of the ATC ground system architecture in the A3G model
Instead of relocating each individual ASAS from the air to the ground, there is one ATC ground
system designed in the A3G model. The ATC Ground system consists of two parts. The first part
named ‘ATC system-other’ is modelled only once. The ‘ATC system-other’ is used as a global
surveillance system. It receives the state and intent information of all aircraft. This part consists
of the following LPN’s:
State & Intent all aircraft
Conformance monitoring
Surveillance / ADS-B ground receiver
ATC system mode
ATC Uplink transmitter (new)
ADS-B receiver mode (new)
The second part named ‘Conflict Detection and Resolution (CDR)’ is modelled for each aircraft
independently. In the system there are k number of aircraft flying. The part CDR-i is introduced
k times in the model, for i=1, .., k. The CDR-i part is responsible for detecting conflicts and
generating resolution advices for aircraft-i. The CDR-i consists of the following LPN’s:
Conflict Detection (CD) & Management-i
Resolution mode-i
Intent based STCR advisory-i
Intent based MTCR advisory-i
Back-to-goal checker-i (new)
The following two audio alert LPN’s are removed from the system:
STC Audio Alerting
MTC Audio Alerting
Page 25 of 96
In the A3G model the system automatically generates a conflict free trajectory after a conflict is
detected. So an audio alert is not necessary anymore.
3.4 Interconnected LPN’s of the ATC System
In the A3G model, the ATC system is modelled through the SDCPN depicted in Figure 3.2. First
we describe the LPN’s that are similar to those used in the ASAS agent of the A3 model.
Subsequently we describe the three LPN’s which are completely new, i.e. LPN ATC Uplink
transmitter to send each MTCDR advisory or STCDRinstruction through datalink to the
appropriate aircraft, LPN ADS-B receiver mode sometimes switches from working to not-
working, and LPN Back2Goal-i
The ADS-B information received from all aircraft is processed by the LPN ADS-B ground
receiver. This yields up to date information about the state and intent of all aircraft which are
maintained in the LPN State&Intent. This LPN also maintains other relevant information for
each a/c, such as mode, priority and handicap information.
This information is used by LPN CD&Management-i to detect conflicts of a/c i with any of the
other aircraft. The LPN Resolution Mode-i determines which type of conflict advise should be
provided to aircraft i. The LPN STCDR Advisory-i and LPN MTCR Advisory-i generate
advisories for aircraft i, and show these to the air traffic controller (ATCo).
An MTCR Advisory applies to conflicts with any other aircraft within time horizon of M . It is
determined as the minimum turning angle (to the left or to the right) such that there are no
predicted conflicts left with any aircraft which has higher priority than aircraft i and which is
within reach of the M horizon. If there is no minimum turning angle possible below a certain
value ,maxM , then the turning angle below ,maxM is identified which provides the lowest
underscoring of the minimum spacing criteria of 5Nm and 1000 ft between centrelines. In that
case aircraft i is assumed to be handicapped. As soon as the advised MTCDR advisories and the
corresponding advisories have been accepted by the controller and by the crew of aircraft i, then
these are broadcasted together with a handicap-i message.
An STCDRAdvisory applies to conflicts of a/c i with any other aircraft within time horizon of
S . It is determined as the minimum turning angle (to the left or to the right) such that there are
no predicted conflicts left with any aircraft and which is within reach of the S horizon. If there
is no minimum turning angle possible below a certain value ,maxS , then the turning angle below
,maxS is identified which provides the lowest underscoring of the minimum separation criteria.
Finally, there are the following two complementary LPN’s:
LPN system mode represents whether the ATC system is working, failed, or corrupted
(failed or corrupted mode also influences the resolution LPN’s).
LPN Conformance Monitoring Intent compares for each a/c j whether j’s state
information agrees with j’s intent information. In case a significant difference is
Page 26 of 96
identified, then both Medium Term and Short Term CD&R for each other aircraft is
informed to stop using intent information of aircraft j.
G1
State & Intent
I-kI1
I2
Processing
G2
CD & Management-i
I3
Gx
GQ
I5
D
Working
System Mode
D
Failure
D
D
Corrupted
I7
3 MTC Res
G5
1
No Res
I
Int-Res-i
Resolution Mode-i
G3
2 STC Res
I1
I2
G6
G4
Processing
I4
Int-MTCD-i
Int-STCD-i
2
STCR Advisory
I
STCR Advisory-i
MTCR Advisory
MTCR Advisory-i
Info
G
ADS-B Ground Receiver / Surveillance
I
I-k
Int-Surv-Intent-Update
Int-Surv-State-Update
Info
Info
Conf. Mon. Intent
I
IPN-SI-CM
I6
G
Processing
1
2I
G1
Processing
IPN-CM-SI
IPN-SI-CDMan-i
ADS-B Receiver mode
Working
Not Working
D D
Back to Goal-i
Check B2G
G
Int-CD Man-i
ATC Uplink Transmitter
G
I
Sending
is Sent
I-ATCo-k
Figure 3.2: Complete DCPN specification of the ATC Ground system agent in the A3G model
Back-to-Goal-i
The Back-to-goal-i LPN is modelled for each aircraft separately. It is part of the CDR-i part of
the ATC Ground System. The LPN Back2Goal-i verifies whether the final RBT direction is
aiming for the destination of aircraft i; if this is not the case, then a token is generated to an IPN,
from which the LPN CD&Management is reminded that an appropriate RBT should be
determined for aircraft i by the LPN MTCDR Advisory-i . In such case the RBT should satisfy
the 5NM/1000ft separation criterion; i.e. no undershooting is allowed. This may have as
implication that no feasible RBT is found. In the latter case the LPN Back2Goal-i will keep on
sending reminders to the LPN CD&Management until a feasible 4D plan has been found by the
LPN MTCDR Advisory-i and this has been accepted by the ATCo and the crew, and downlinked
as current RBT through ADS-B and stored in the LPN State&Intent.
Page 27 of 96
In Figure 3.3 a simulation realization shows what happens when after a short term conflict
resolution the aircraft has no longer an intent which leads to its final goal. The magenta parts of
the curve indicate when the aircraft is controlled manually (i.e. not by FMS), meaning in Short
Term Conflict resolution mode. As can be seen after an aircraft comes in STC mode, it will
continue to do so. The aircraft will continue to fly the proposed STC heading change. The back-
to-goal resolution is in the A3 model initiated by the Pilot-Flying after an STCR. In the A3G
model this is automatized. After a short term conflict has occurred the back-to-goal initiates the
check for a back-to-goal resolution. This is done at predetermined times. The resolution is
generated in the ‘CD & Management-i’ LPN.
-1.5 -1 -0.5 0 0.5 1 1.5
x 105
-1
-0.5
0
0.5
1
x 105
0
1
2
3
4
5
6
7
2D paths, Turn=Red, Command Mode=Magenta, PID is 1
m
m
Figure 3.3: Eight aircraft scenario A3G model without back-to-goal checker. Magenta = a/c in STC
Resolution mode, Red = a/c in MTC Resolution mode.
ADS-B receiver mode
Via the ADS-B ground receiver / surveillance receives the ATC ground system the state and
intent information of all aircraft. This is only received if the ADS-B receiver mode is working.
The ADS-B receiver mode is only modelled once. The ADS-B receiver mode has the following
places:
Working
Not Working
The system changes at exponentially distributed times.
ATC uplink transmitter
Page 28 of 96
The ATC Uplink transmitter LPN is part of the ATC ground system. It is only modelled once.
The ATC uplink receiver has two places for the following modes:
Sending
Is sent
The ATC Uplink transmitter the resolution advices from the ATCo’s to the corresponding
aircraft. Due to the auxiliary place ‘Int-ATCo-Uplink-queue’ a queue is possible which is
handled on a first-in-first-out basis. The ATC uplink transmitter sends the resolution advice to
the corresponding aircraft if the ‘Global ATC Uplink frequency’ is working. Otherwise the
transmitter will remain in ‘sending’ mode. The ATC uplink transmitter can only send one
resolution advice at a time.
3.5 ATCo as agent in the A3G model
In this subsection an overview of the Air Traffic Controller (ATCo) agent is presented.
The task of the Air Traffic Controller in the A3G model consists of three steps:
Notice the alert of a resolution advice generated by the ATC ground system after a
conflict is detected.
Confirm if the resolution advice is still viable by checking if the aircraft is still in conflict
in ‘Resolution mode-i’ in the ATC ground system.
Insert the resolution advice in the ATC Uplink transmitter so it can be send to the correct
aircraft.
The steps are combined in one reaction time parameter in the A3G model.
The agent consists of two LPN’s. The ATCo’s tasks are divided in two parts; one being the
tactical part and one the strategic part. The tasks require different responses. The ATCo-Tactical
(ATCo-T) deals with the short term resolutions. The ATCo-Planning (ATCo-P) deals with the
strategic tasks. Strategic tasks contain the medium term conflict resolutions and back-to-goal
advices.
In Figure 3.4 the schematic overview of the agent is given with the external interacting agents.
The ATCo receives resolution advices the ATC Ground system agent. The ATCo inserts the
resolution advice in the ATC Uplink transmitter and with a STCR also directly in the ATC
system. The place before the ATCo has space for multiple resolution advices therefore a queue is
possible. The IPN with place ‘Int-ATCo-Uplink-Queue’ makes it possible to have a queue of
outstanding resolution advices.
Page 29 of 96
G1
G
STC
I-i
MTC
STC Advisory-i
MTC Advisory-i
Int-MTC-i
Int-STC-i
Int-ATCo-Uplink-Queue
ATCo-Tactical
I-i
ATC ground systemState & Intent
I1
G2
2
G
ATC Uplink Transmitter
1
I
Sending
is sent
I-i
I2
Info
Processing
Int-STC-State&Intent-i
G
ATCo-Planning
I-ATCo-k
CD & Management-i
Res Mode-i
Figure 3.4: Schematic overview of the ATCo agents and communicating LPN's
ATCo Tactical (ATCo-T)
The ATCo-T is only modelled once. Due to the auxiliary place ‘Int-STC-i’ modelled for each
aircraft independently a queue is possible. An outbound queue is also possible in the place of the
‘Int-ATCo-Uplink-Queue’ IPN. Resolution advices are handled in a first-in-first-out principle.
The place Int-STC-i can be overwritten, meaning the resolution advice can be updated while the
ATCo is working on it.
Start of the ATCo-T tasks is a short term resolution from STC Advisory-i. The ATCo-T then
validates if the corresponding aircraft is still in conflict in Resolution Mode-i (Res Mode-i). If
the aircraft is still conflict the resolution is accepted. The ATCo will then give a sign to the ATC
Uplink Transmitter to send the resolution advice. Simultaneously he/she inserts the new heading
change directly in the ATC ground system. If the resolution is not viable anymore the ATCo-T
will drop the resolution advice.
Page 30 of 96
ATCo Planning (ATCo-P)
The ATCo-P is modelled only once. Due to the auxiliary place ‘Int-MTC-i’ modelled for each
aircraft independently a queue is possible. An outbound queue is also possible in the place of the
‘Int-ATCo-Uplink-Queue’ IPN. Resolution advices are handled in a first-in-first-out principle.
The place Int-MTC-i can be overwritten, meaning the resolution advice can be updated while the
ATCo-P is working on it. The ATCo-P receives Medium Term Conflict Resolution (MTCR)
advices via the ‘MTC advisory-i’. The back-to-goal advices which are needed after a short term
conflict resolution advice are generated by the ‘CD & Management-i’ LPN in the ATC ground
system.
The task starts with an incoming resolution advice. The ATCo-P checks if the aircraft is still in
conflict in ‘Res Mode-i’ before sending the resolution to the ATC Uplink transmitter, otherwise
the resolution is dropped.
3.6 New communication systems in the A3G model
In this section the newly added ground-air communication LPN’s will be discussed. The new
agents are the ATC Uplink Transmitter, global ATC uplink and the airborne ATC Uplink
receiver-i.
Figure 3.5 shows a schematic overview of the path of a resolution advice from ATC ground
system to the pilot. In Figure 3.6 the full process is shown using the SDCPN structure as used in
the A3G model.
The process in the figures starts with a generated resolution advice in the STC advisory-i, MTC
advisory-i or CD & Management-i. The ATCo checks in ATC resolution mode (Res Mode) if
the resolution is still viable before inserting it in the ATC Uplink transmitter. The ATC uplink
transmitter sends the resolution to the corresponding aircraft, but only if the global ATC uplink
frequency is working. This resolution is received in the airborne ATC Uplink receiver-i. The
received message will generate an audio alert. This audio alert will be picked up by the pilot and
the corresponding task performance will be initiated.
Next the new communication systems will be presented in more detail.
Figure 3.5: Schematic overview of the agents involved in the resolution advice process in the A3G model.
Page 31 of 96
G1
G
STC
I-i
MTC
STC Advisory-i
MTC Advisory-i
Int-MTC-i
Int-STC-i
Int-ATCo-Uplink-Queue
ATCo-Tactical
I-i
State & Intent
I1
G2
2
G
ATC Uplink Transmitter
1
I
Sending
is sent
I-i
I2
Info
Processing
Audio Alert (IPN)
Current Goal PF
Collision Avoidance (C1)
Navigation Vertical (C4)G
interrupt
Conflict Resolution (C3)
Preparation Route Change (C6)
Miscellaneous (C7)
Gsub-
sequent
Emergency Actions (C2)
Navigation Horizontal (C5)
Audioalert
II
I
I
Int-Emergency-Audio
Int-Indicator-Audio
Int-STC-Audio
IInt-MTC-B2G-Audio
Int-STC-State&Intent-i
G
ATCo-Planning
I-ATCo-k
CD & Management-i
2
G
1 Not Working
I
Working
ATC global Uplink
FMS Intent-i
Res Mode-i
Int-Uplink-i
ISI
ATC Uplink Receiver-i (IPN)
I S
IM I M
IB I B
Int-MTC-Audio
Int-Uplink_Rec-i
Figure 3.6: DCPN specification overview of the resolution advice process in the A3G model
Global ATC uplink frequency
The global ATC uplink frequency LPN is part of the environment agent. It is only modelled
once. The global ATC uplink LPN has two places representing the following modes of the
system:
Working
Not working
Page 32 of 96
The global ATC uplink frequency is the frequency used to send the resolution advice from the
ATC ground system to the corresponding aircraft. The switches occur at exponentially
distributed times. The ATC uplink frequency is based on the global ADS-B frequency in the A3
model.
ATC uplink receiver
For the ATC uplink receiver-i on-board of aircraft-i a DCPN model is presented in Figure 3.7.
The ATC uplink receiver model manages a proper reception by and alerting of a pilot for the
three different types of resolution advices from the ATC ground system: Tactical instruction, 4D
plan update proposal, and Back-to-Goal resolution advice. The ATC uplink receiver is modelled
as an Interaction Petri Net (IPN) that aims to imitate the response of the Pilot-Flying in the A3
model. The specifics of this response are presented in Table 3.3.
FMS Intent-i
Int-Uplink-i
ISI
ATC Uplink Receiver-i (IPN)
I
Audio Alert (IPN)
Audioalert
II
I
I
Int-Emergency-Audio
Int-Indicator-Audio
Int-STC-Audio
Int-MTC-Audio
I
Int-MTC-B2G-Audio
S
IM I M
IB I B
ATC UplinkTransmitter
Int-Uplink_Rec-i
Figure 3.7: DCPN specification of the ATC Uplink receiver IPN including communicating LPNs
In Table 3.3 the upper row indicates the current task of the Pilot-Flying. The left column
indicates the type of resolution received. Each of the matrix elements specifies what the pilot
should do. The possible response options for the pilot are: 1) to start a new task, or 2) to locally
save this task and start doing it after the current task is finished. The specifics of Table 3.3 are
captured in the IPN’s in Figure 3.7.
Page 33 of 96
Table 3.3: overview of the pilot response for incoming alerts in the A3G model
Current →
Incoming ↓
MTC task STC task Back-to-Goal task Else
MTC
resolution
Advice
Audio Alert
+
Restart MTC task
-
+
Finish STC task
Save MTCR advice
Audio Alert
+
Start MTC task
Audio Alert
+
Start MTC task
STC
resolution
advice
Audio Alert
+
Start STC task
-
+
Finish STC task
Save STC advice
Audio Alert
+
Start STC task
Audio Alert
+
Start STC task
Back-to-goal
resolution
advice
-
+
Save B2G advice
-
+
Save B2G advice
Audio Alert
+
Restart B2G task
Audio Alert
+
Start B2G task
3.7 Pilot Flying as Agent in the A3G model
In this subsection the Pilot flying agent in the A3G model is presented.
In the A3G model the pilot flying is responsible for the final step in executing the resolution
advice generated by the ATC ground system. The Pilot inserts the resolution advice in the FMS
after which the aircraft will change its heading.
In the A3G model only small adaptations to the Pilot Flying agent are made in comparison with
the A3 model. The internal LPN structure remained the same. In the A3G model the overall
responsibilities of the Pilot-Flying are decreased and taken over by the ATC Ground system. The
pilot is not in the position to initiate a process. All instructions are generated by the ATC ground
system and by the ATCo via ATC Uplink send to the aircraft.
In Figure 3.8 the Pilot Flying agent from the A3G model is presented. Relative to the A3 model,
the changes are only in the Audio Alert IPN and the Task Performance LPN; these are further
explained next.
Audio Alert (IPN)
The Audio Alert is an Interaction Petri Net (IPN). Its goal is to give alerts to the pilot of
incoming events. In the A3G model it is just only used for the incoming resolution advices
coming from the ATC Uplink receiver. Its structure is very basic. The adaptations are made due
to the fact that emergency procedures are not yet implemented in the A3G model. In case of an
emergency this is saved in a separate file, this can be used for analysis.
Page 34 of 96
Current Goal PF
Goal Memory PF
Memory
I
I
I3
Other EmergencySituations
Audio Alert PF (IPN)
Audioalert
II
I
I
Int-PF-Emergency-Audio
Int-PF-Indicator-Audio
I
Int-PF-TP1
Int-PF-TP2
Int-Fail-Ind
I
No Emergency
GG
Emergency
(5x)
(5x)
Int-STC-Audio
Int-STC-Audio
Task Performance PF
Monitoring(T1)
Coordination(T3)
Monitoring &Goal Prioritisation(T6)
G
G
D
Execution(T4)
G
G
Monitoring andDecision(T2)
D
G
Execution Monitoring(T5)
G
D
End Task(T7)
Ginterrupt
Gsub-
sequent
D
Failure Indicators for PF
No FailureIndication
I G
Failure Indication
Working
G G
Not Working
(5x)
Collision Avoidance (C1)
Navigation Vertical (C4)G
interrupt
Conflict Resolution (C3)
Preparation Route Change (C6)
Miscellaneous (C7)
Gsub-
sequent
Emergency Actions (C2)
Navigation Horizontal (C5)
II
Int-B2G-Audio
Int-PF-GM2Int-PF-GM5
Int-PF-GM1
ISI
ATC Uplink Receiver-i (IPN)
I S
IM I M
IB I B
Int-Uplink_Rec-i
Figure 3.8: DCPN specification of the Pilot Flying agent in the A3G model
Page 35 of 96
Task Performance
The internal structure of the Task Performance LPN has not been changed in comparison with
the A3 model. The following tasks are present in the A3G model:
Task Performance Goal 3: Conflict resolution actions for STC and MTC
Task Performance Goal 5: Navigation horizontal actions for Back-to-Goal
Task Performance goal 6: Preparation route changes
Task Performance 7: Miscellaneous.
The Tasks Goals 2 (Emergency Actions) and 4 (Vertical Navigation) are not used in the A3G
model. The A3G model can only cope with horizontal heading changes. Task performance goal
1 is not used in both the A3 model as the A3G model.
The other change is the Pilot-Flying now implements the new resolution advice from the ATC
Uplink receiver instead of from the ASAS as in the A3 model.
3.8 Implementation and verification of the A3G code
The next step is to implement the SDCPN model in the selected programming language, which is
the object oriented Delphi XE3 language, i.e. the same language used for the A3 model
implementation.
The implementation of the A3G model code is done in steps. The motivation behind this
stepwise approach is that it allows conducting an intermediate verification test after each step.
Step 1: Introduce ‘shadow’ aircraft, agent 0
A new agent 0 is introduced. Eventually, this agent 0 will form the ATC system agent. In step
agent 0 is filled with the 4 LPN’s of ‘ATC system-other’ part described in subsection 3.3. Agent
0 receives the state and intent information from all other aircraft through ADS-B downlink.
Verification test 1:
Using the eight aircraft scenario the state and intent information in each aircraft’s ASAS
surveillance part is compared to the state and intent information in agent 0. Code corrections
have been made until this verification test has shown that the state and intent data on the ground
equals the state and intent data in the ASAS systems of the aircraft.
Step 2: Insert CDR-i part to agent0
The Conflict Detection and Resolution (CDR) part of agent 1,..,N is added to agent 0. Hence in
agent 0 the CDR part is separately modelled for each aircraft. Each CDR-i in agent 0 uses the
information of the ATC system to detect conflicts and generate a conflict trajectory for aircraft-i.
Verification test 2:
Using the eight aircraft scenario, the resolution advice generated by agent 0 is compared to the
resolution advice generated by corresponding aircraft’s airborne ASAS. Code corrections have
been made until the outcome of the verification test was positive.
Page 36 of 96
Step 3: Add air-ground uplink
The LPN for ATC uplink is added to the system. The ATC uplink is used to send an agent 0
generated resolution advice to the corresponding aircraft.
Verification test 3:
Using the eight aircraft scenario, the resolution advice received is compared to the resolution
advice generated by agent 0. Code corrections have been made until the outcome of this
verification test was positive.
Step 4: Add an Air Traffic Controller (ATCo)
The Air Traffic Controller agent is inserted between the agent 0 and the ATC uplink. In Figure
3.9 an overview of the model after step 4 is shown.
Verification test 4:
Using the eight aircraft scenario, the resolution advice received is compared to the resolution
advice of generated by agent 0. Code corrections have been made until the outcome of this
verification test was positive.
State & Intent all
aircraft +
system
Agent0
CDR-i
Agent0
CDR-i
Agent0
ATCo ATC Uplink
Save
resolution
advice
Aircraft-i
Save
resolution
advice
Aircraft-i
Figure 3.9: Schematic overview of sending the resolution advice to the aircraft.
Step 5: Using ATC ground resolution advice in the air
So far in the model each aircraft uses resolutions generated by its own ASAS. In this step 5 the
resolution advices received through ATC uplink from the ground are used instead of those from
own ASAS. Due to this step 5, each aircraft will fly according to the resolution advice generated
by agent 0 on the ground.
Verification test 5:
Using the eight aircraft scenario, it has been compared whether the aircraft behaved the same as
in the previous test. Code corrections have been made until the outcome of the verification test
was positive.
Step 6: Delete airborne ASAS
Finally, for each aircraft airborne ASAS is deleted.
Verification test 6:
Page 37 of 96
It has been verified that the simulation results do not change due to the deletion of the airborne
ASAS from the implemented code.
Step 7: Rare event verification
The verification tests conducted in steps 1 through 6 are based on a few simulation runs for the
implemented code. Hence, the positive outcomes of these verification tests do not preclude the
occurrence of rare event differences either due to remaining code errors or due to differences in
rare emergent behaviour of the A3G model relative to the A3 model. In order to get hold on
either type of rare event differences, in the next sections we conduct rare event MC simulations
for 2 and 8 aircraft encounters.
During the rare event simulations for 8 a/c encounters, there appeared to be some unexplained
differences between the behaviour of the A3 model and the A3G model [Nieskens, 2014].
Through conducting additional rare event MC simulations, the causes of these differences have
been identified, and subsequently the necessary improvements in the code have been made and
verified through running additional rare event MC simulations.
The three main improvements that resulted from this rare event verification and improvement
activity are:
- When an MTCR plan is too old in the sense that it includes trajectory changes that
should already have been made by the aircraft, then the pilot will not enter this
plan in the FMS. This condition was not properly implemented in the A3G code.
- After having given an open tactical resolution, ATC determines a back-to-goal
tactical instruction. In doing so an erroneous waypoint and an erroneous distance
calculation was used, as a result of which the back-to-goal instruction could work
counterproductive in some rare cases.
- A pilot receives an audio alert in case of an MTCR uplink, which makes the pilot
stop finishing his current activity, e.g. on implementing an STCR instruction. In
some rare events this could lead to a sequence of instructions rendering a pilot
becoming totally unproductive. In order to avoid this, the pilot does no longer
receive an audio alert when there is a sequence of instructions awaiting.
The proper working of these improvements in the A3G model have been verified through
running additional rare event MC simulations.
Page 38 of 96
4 MC SIMULATION OF 2 AIRCRAFT ENCOUNTERS
The aim of this section is to investigate under which conditions it is possible to get A3G model
rare event MC simulation results for two aircraft encounters as good as obtained for the A3
model. The two aircraft head-on encounter scenario is the same as the one being used for the
Monte Carlo simulation results of the A3 model [Blom & Bakker, 2011a,b].
This section is organized as follows. In subsection 4.1 A3G baseline parameter values for the
A3G model are adopted such that it is sure that the A3G model has the same performance on two
aircraft encounters as the A3 model had with the A3 baseline parameter values. These A3G
baseline parameter values are chosen in a conservative way, i.e. such that for the 2 a/c encounter
scenario, it is sure that the A3G model performs as good as the A3 model does. Subsection 4.2
provides MC simulation results for the A3G model using these A3G baseline parameter values.
In subsection 4.3 the effect of A3G results is shown when A3 baseline parameter values would
be used instead of the A3G baseline parameter values. From this point on, for the A3G parameter
values that differ from the A3 baseline values, extra MC simulation tests are conducted in order
to find out whether there is room for a less conservative value, i.e. somewhere in between the A3
baseline and the A3G baseline values. First, in subsection 4.4 the Monte Carlo simulation Tests
to be conducted are defined. Subsequently, in subsections 4.5 to 4.16 the Monte Carlo simulation
results obtained for these Tests on two aircraft head-on encounter scenarios are presented. In
subsection 4.17, A3G selected parameter values and corresponding simulation results are
summarized, and in subsection 4.18 an interpretation is given of the results obtained for the two
aircraft encounter scenarios.
4.1 A3G Baseline parameter values
In this subsection A3G baseline parameter values for the A3G model are adopted. These A3G
baseline parameter values are adopted in a conservative way in order to be sure that the A3G
model has the same performance on two aircraft encounters as the A3 model had with the A3
baseline parameter values [Blom & Bakker, 2011a]. For the complete list of adopted A3G
baseline parameter values (177 in total) the reader is referred to appendix C. In this subsection
only those parameter values are explained that differ from the A3 baseline parameter values.
These A3G baseline parameter values are shown in Table 4.1. The number in the first column
corresponds to the number in the full parameter list in Appendix C.
The adopted A3G baseline parameter values can be separated in three groups:
The twelve parameters that are coloured white in Table 4.1: These influence the state of
the technical systems. A system failure event has a probability of occurrence and a mean
duration of failure. For the A3G baseline parameter values the mean duration parameters
have not been changed. The ATC global frequency is new, though its function is similar
to the global ADS-B frequency in the A3 model. The value zero for the first two
parameters reflects that this functionality is not implemented in the A3G model. The
other A3G baseline probability values are all set to a probability of failure or Not
Working of 101*10 . This is many orders in magnitude better in comparison with the A3
baseline parameter values.
Page 39 of 96
The seven parameters that have a grey shading in Table 4.1: These are all used by the
newly added ground-based agents. First the location of the ATC ground system is located
in the centre. Second the parameter 2B GT is added to the ATC ground system, because
this was previously done in the A3 model for the pilot. The rest of the new baseline
parameter values are all set to 1 second to simulate almost no delay.
For the remaining 158 parameter values, the adopted A3G baseline parameter values
equal the A3 baseline parameter values.
Table 4.1: Baseline parameter values for the A3G model that differ from baseline A
3 model
# Parameter Explanation
A3
G Baseline
Value
A3
Baseline
Value
2 Fail
Enginep Probability of Engine Failure 0 1/6000
4 down
OESp Probability of Other Emergency Failure 0 1/6000
62 down
SATp Probability of Global GNSS/GPS Not working 101*10
51*10
66 occ
ADS Bglobalp Probability of ADS-B global Occupied 101*10
61*10
69 down
ATC global
Mean duration of Global ATC uplink Occupied
Not Occupied
1 hr. 1hr. 1
70 down
ATC globalp Probability of Global ATC uplink Occupied 101*10
61*10
1
94 down
GNSS RECp
Probability of Airborne GPS receiver Not
Working
101*10
55*10
98 down
Altimp Probability of Airborne Altimeter Not Working 101*10
55*10
111 down
ADS TRMp Probability of ADS-B transmitter Not Working 101*10
55*10
165 ownx Position of ATC ground system [x,y,z] [0,0,0] -
168 corr
ATCsysp Probability of ATC ground system Corrupted 101*10
55*10
2
169 down
ATCsysp Probability of ATC ground system Not working 101*10
55*10
2
170
2B GT ATC ground system, Interval time for Back-to-
Goal Evaluation
20 s - 3
172
,
down
ATC ADS RECp
Probability of ADS-B ground receiver Not
Working
101*10
55*10
4
1 The global ATC uplink is new, but in function the same as Global ADS-B frequency in A3 model
2 The ATC ground system is new, but system mode is a direct copy of ASAS in A3 model
3 In A3 model Back-to-goal resolution advice generation was initiated by Pilot Flying with an interval of 20
seconds until a conflict free back-to-goal advice was generated.
4 ADS-B ground receiver is new but exactly the same as aircraft ADS-B receiver in A3 model
Page 40 of 96
173 Transmit
uplinkT ATC ground Uplink Transmitter duration of
sending resolution to aircraft
1 s -
174 min
ATCo TT ATCo-Tactical minimum response time 1 s -
175 max
ATCo TT ATCo-Tactical maximum response time 1 s -
176 min
ATCo PT ATCo-Planning minimum response time 1 s -
177 max
ATCo PT ATCo-Planning maximum response time 1 s -
4.2 MC simulation results under A3G baseline parameter values
In this subsection the MC simulation result for the A3G model using A3G baseline parameter
values is presented. Similar as in [Blom & Bakker, 2011a], in this scenario two aircraft start at
320 km (178 Nm) from each other. The initial 3D-position has standard deviations of 20m in
longitudinal direction along the Reference Business Trajectory (RBT) centreline, 0.5 Nm in
lateral direction and 20m in height. Both fly straight opposite flight plans at 250 m/s airspeed.
The short and medium term detection and resolution criteria used in the MC simulations are
shown in Table 4.2. The horizontal separation minimum (medium and short term) is 5 Nm.
Table 4.2: Short term and medium term separation criteria for the A3 and A
3G model
Look ahead
time
Resolve ahead
time
Horizontal
separation min
Vertical separation
min
Max Turn
angle
STC 3 min 3 min + 10s 5 Nm 900 ft. 60 degrees
MTC 10 min 15 min 5 Nm 1000 ft. 60 degrees
Figure 4.1 presents the results of one million Monte Carlo simulations of the A3G model using
the A3G baseline parameter values. The simulation results in Figure 4.1 show the same curve as
obtained for the A3 baseline parameter results for the A3 model [Blom&Bakker, 2011a].
Page 41 of 96
Figure 4.1: MC simulation results for two aircraft encounter under the A3G model with A3G baseline
parameter values.
4.3 A3G simulation results under A3 Baseline parameter values
Figure 4.2 presents MC simulation results of the A3G model using the A3 baseline parameter
values for those parameters that coincide with those of the A3G model. As expected, Figure 4.2
shows far less good results than Figure 4.1. Comparison with Figure 4.1 shows the same positive
behaviour during the left part of the curve, though far less good results for the right part of the
curve. The difference can be seen from the 42*10 event probability level, meaning once in
5000 Monte Carlo runs of 2 aircraft encounters.
Page 42 of 96
Figure 4.2: MC simulation results for two aircraft encounters under the A3G model with parameter settings
according to the A3 model baseline parameter values.
4.4 Additional MC simulation Tests of 2 a/c encounters
For those A3G baseline parameter values that differ from the A3 baseline values, extra MC
simulation Tests will be conducted in order to find out whether there is room for a less
conservative value, i.e. somewhere in between the A3 baseline and the A3G baseline values.
These extra Tests are conducted in subsection 4.5 to 4.16. These extra tests are defined in the
current subsection.
An overview of the parameter setting in the additional scenarios is given in Table 4.3. Each of
the tests will be performed with the two aircraft encounter scenario and contains one million rare
event Monte Carlo simulations. In each test only one parameter value is changed with respect to
the A3G baseline parameter values.
Tests A and B have been conducted in sections 4.1 and 4.2 respectively. Tests C-N are
conducted in the remainder of this section.
The A3G model parameter tests C-N can be categorised in two groups. The categorisation is
based on what they influence in the system. The two categories are:
Performance of the technical systems parameters (tests C – K).
Page 43 of 96
Parameters of the new ground based agents and LPN’s, which generate a delay in the
system (tests L – N).
Table 4.3: Overview of the alternative parameter setting test for the two aircraft encounter scenario
Test Parameter Explanation Baseline
Value Test value
A Same A3 Baseline Parameter value setting A3 baseline -
B All A3G Baseline Parameter value setting A3G baseline -
C down
SATp Probability of Global GNSS/GPS Not
working
101*10 51*10
D occ
ADS Bglobalp Probability of Global ADS-B Occupied 101*10 61*10
E down
ATC globalp
Probability of Global ATC Uplink
frequency Occupied
101*10 61*10
F down
GNSS RECp Probability of Aircraft GPS receiver Not
Working
101*10 55*10
G down
Altimp Probability of Aircraft altimeter Not
Working
101*10 55*10
H down
ADS TRMp Probability of Aircraft ADS-B transmitter
Not Working
101*10 55*10
I corr
ATCsysp Probability of ATC ground system
Corrupted
101*10
55*10
J down
ATCsysp Probability of ATC ground System Not
working
101*10 55*10
K ,
down
ATC ADS RECp
Probability of ATC Ground ADS-B
receiver Not Working
101*10 55*10
L max
ATCo TT ATCo-Tactical response time 1s 10 s
M max
ATCo PT ATCo-Planning response time 1s 10 s
N Transmit
uplinkT ATC Uplink transmitter Send time 1s 12 s
Each test will be performed using a series of Monte Carlo simulations, whereby in each test only
one variable is changed compared to the baseline parameter value setting. If the results of the
tested version are the same as the baseline results, the parameter has no substantial (negative)
influence on the system and can therefore be changed.
The goal of tests C - K is to bring these A3G model baseline parameter values closer to the
baseline parameter values used in the A3 model. For the parameters depending on probability,
this means that they are set to a higher probability, or a higher likelihood of failure.
The goal of tests L – N is to investigate the influence of a longer delay on the total system.
Instead of the A3G baseline parameter value of 1 second the ‘delay or response time’ parameters
will be set to higher values.
Page 44 of 96
4.5 Test C: Global GNSS/GPS
The parameter ‘Global GNSS/GPS’ is a parameter setting for the probability of Global
GNSS/GPS not working in the environment agent. If global GPS is down, all aircraft are not able
to use the navigation satellites to determine their position. Aircraft are then left to depend on
their inertial reference system (IRS). The non-baseline test C parameter setting of 1*10-5 is
obtained from the A3 baseline parameter values. In Figure 4.3 the Monte Carlo simulation
results for test C are presented.
Figure 4.3: Monte Carlo simulation results for Test C; the non-baseline Global GNSS/GPS parameter
setting in the A3G model
The curve in Figure 4.3 is the same as in Figure 4.1. A global interruption of the navigation
satellites has no significant effect on the results. This can be explained as follows. Each aircraft
has a second system to determine its position, namely the Inertial Reference System (IRS). The
broadcasted state information of each aircraft will thus still be quite precise. The results of the
Monte Carlo simulation show that the effect of the non-baseline variable is not significant.
Therefore the parameter can be changed to the value used in the A3 model.
In Table 4.4 the outcome of test C is indicated through a green background.
Page 45 of 96
Table 4.4: Non-baseline parameter value setting for the Global GNSS/GPS parameter
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test C
value
Environment GNSS system (GPS/
Nav. Global) /
Satellites
down
SATp
Probability of Global
GNSS/ GPS Not working 101*10
51*10
4.6 Test D: Global ADS-B frequency
The parameter ‘Global ADS-B frequency concerns the probability of global ADS-B frequency
being occupied in the environment agent. Global ADS-B frequency is used to send state and
intent information of the aircraft to the ground. If the ADS-B frequency is occupied, this mean
that the ATC ground system cannot receive the latest intent information of all aircraft. The non-
baseline test D parameter setting of 1*10-6 is obtained from the baseline parameter values of the
A3 model. In Figure 4.4 the Monte Carlo simulation results for test D are presented.
Figure 4.4: Monte Carlo simulation results for Test D; Global ADS-B frequency non-baseline parameter
setting
As can be seen the effect of the non-baseline value is negligible. The effect of a global
interruption of the ADS-B frequency has no significant effect on the total safety of the system.
This effect can be explained by two arguments. Firstly if aircraft intent information is not
received the old information can still be used. Also new intent can still be send up via the ATC
Page 46 of 96
uplink frequency. Secondly the non-baseline parameter value is still very small, an effect of
1*10-6 is hard to detect with only one million simulations.
In Table 4.5 the outcome of test D is indicated through a green background.
Table 4.5: Non-baseline parameter setting for Global ADS-B frequency parameter
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test D
value
Environment Global ADS-B
ether
frequency
occ
ADS Bglobalp
Probability of ADS-B global
Occupied 101*10
61*10
4.7 Test E: Global ATC Uplink frequency
The parameter ‘Global ATC uplink frequency is a parameter setting for the probability of global
ATC uplink frequency being occupied in the environment agent. Global ATC uplink frequency
is used to send the short term and medium term resolution advice from the ATC ground station
to each corresponding aircraft. Although the parameter is new, the test E value is based on the
very similar global ADS-B frequency baseline parameter value used in the A3 model.
In Figure 4.5 the Monte Carlo simulation results for test E are presented; they show a significant
effect. When the ATC uplink frequency is blocked, no aircraft can receive a new resolution
advice. Although this blocking happened only once in the one million simulation runs, its effect
is large when it happens. Therefore a better value is needed. The results in Figure 4.5 also mean
that a factor 100 improvement relative to the A3 baseline value of 61*10 will suffice. Therefore
we conclude as outcome of test E that the frequency of Global ATC Uplink frequency blocking
probability should be 81*10 . The latter value is indicated with green background in Table 4.6.
In Table 4.6 the non-baseline parameter setting and the outcome of test E is presented.
Table 4.6: Non-baseline parameter setting for the Global ATC uplink frequency parameter
Agent LPN Parameter Explanation
Test E
derived
value
A3
G
Baseline
Value
Environment Global ATC
uplink
frequency
down
ATC globalp
Probability of Global ATC
Uplink frequency Occupied 81*10
101*10
Page 47 of 96
Figure 4.5: Monte Carlo simulation results for Test E; Global ATC uplink frequency non-baseline
parameter setting..
4.8 Test F: Aircraft GPS receiver
The parameter ‘Aircraft GPS receiver’ is a parameter setting for the probability of GPS receiver
not working in the own positioning systems agent of the aircraft. If the GPS is not working, the
specific aircraft is not able to use the navigation satellites to determine its position. The aircraft is
then only depending on its inertial reference system (IRS).
In Table 4.7 the non-baseline parameter setting for test F-2 is presented. The test F parameter
setting of 5*10-5 is the A3 baseline parameter value.
Table 4.7: Non-baseline parameter setting for the aircraft GPS receiver parameter
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test F
value
GNC systems:
Own
Positioning
Systems
Aircraft GNSS/
GPS receiver
down
GNSS RECp
Probability of Aircraft GPS
receiver Not Working 101*10
55*10
Page 48 of 96
In Figure 4.6 the Monte Carlo simulation results for test F are presented. For rare events the
results in Figure 4.6 differ from those in Figure 4.1. Therefore the A3G model parameter value
should not be changed to the A3 baseline parameter value.
Figure 4.6: Monte Carlo simulation results for Test F; aircraft GPS receiver non-baseline parameter
setting
Test F-2
In test F the effect of the simulation results were significant but not very large. Therefore a
second test has been performed with the parameter setting and the test F outcome shown in Table
4.8.
Table 4.8: Test F-2: parameter setting for the aircraft GPS receiver; the green background indicates the
outcome of test F-2.
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test F-2
value
GNC systems:
Own
Positioning
Systems
Aircraft GNSS/
GPS receiver
down
GNSS RECp
Probability of Aircraft GPS
receiver Not Working 101*10
61*10
Page 49 of 96
In Figure 4.7 the results of Monte Carlo simulation for test F-2 are presented. The simulation
results are similar to the simulation results in Figure 4.1. This means that changing the GPS
receiver parameter to 61*10 is sufficient for two aircraft encounters.
Figure 4.7: MC simulation results for Test F-2; GPS receiver setting to 10-6
in the A3G model
4.9 Test G: Aircraft altimeter
The parameter ‘Aircraft Altimeter’ is a parameter setting for the probability of Altimeter not
working in the own positioning agent of the aircraft. The A3 model and A3G model only detect
horizontal conflicts. The non-baseline test G parameter setting of 5*10-5 is obtained from the
baseline parameter values of the A3 model. In Figure 4.8 the Monte Carlo simulation results for
test scenario G are presented.
Page 50 of 96
Figure 4.8: Monte Carlo simulation result for Test G; aircraft altimeter non-baseline parameter scenario
As can be seen the effect of the aircraft altimeter not working is negligible. The results are the
same as the baseline parameter results. A failure with the aircraft altimeter has no effect on the
system. In the A3G model all aircraft fly at the same flight level. Therefore an error in the
vertical position calculation is negligible.
In Table 4.9 the outcome of test G is indicated through a green background.
Table 4.9: Non-baseline parameter setting for the altimeter scenario
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test H
value
GNC systems: Own
Positioning Systems
Aircraft
Altimeter
down
Altimp Probability of Aircraft
Altimeter Not Working
101*10
55*10
4.10 Test H: Aircraft ADS-B transmitter
The parameter ‘Aircraft ADS-B transmitter’ is a parameter setting for the probability of the
ADS-B transmitter not working in the communication systems agent of the aircraft. The ADS-B
transmitter only sends the intent information of the aircraft to the ground. In Figure 4.9 the
Monte Carlo simulation results for test scenario H are presented.
Page 51 of 96
Figure 4.9: Monte Carlo simulation results for Test H; aircraft ADS-B transmitter non-baseline scenario.
The results Figure 4.9 are significantly different from those in Figure 4.2. The effect is large; this
can be explained as follows. The aircraft sends own intent information via the ADS-B transmitter
to the ATC ground system. If the aircraft’s ADS-B transmitter is not working the intent
information is not received on the ground. Because this denies 4D trajectory plan verification, in
the A3 model the 4D plan of such an aircraft is considered to become unreliable. The same
approach has been copied in the A3G model. That this unreliability assumption yields far more
problems for the A3G model than it does for the A3 model is because in the A3 model the
aircraft with the failing ADS-B transmitter still has high quality state and intent information from
all other aircraft. Hence this aircraft will continue to provide a very reliable resolution. In the
A3G model however, there is only agent 0 (the ATC system) where all state and intents of
aircraft are used to determine proper resolutions; and this agent 0 is lacking proper intent of the
aircraft with failing ADS-B transmitter.
In Table 4.10 the outcome of test H is indicated through a green background.
Table 4.10: Non-baseline parameter value for the aircraft ADS-B transmitter scenario
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test H
value
GNC: Communication
Systems
ADS-B
Transmitter
down
ADS TRMp
Probability of Aircraft ADS-B
transmitter Not Working 101*10
55*10
Page 52 of 96
4.11 Test I: ATC ground system corrupted
The parameter is the probability of ATC Ground system being corrupted. This parameter is used
in the ATC system mode in the ATC ground agent. When the system is corrupted the system
doesn’t detect conflict and also doesn’t give an indication that the ATC system has a failure.
The ATC ground system agent is newly added to the model, but is originated from the ASAS
agent in the A3 model. The ATC system mode is a direct copy of the ASAS system mode,
therefore the non-baseline test value is based on the A3 model baseline value. In Figure 4.10 the
Monte Carlo simulation results for test scenario I are presented.
Figure 4.10: Monte Carlo simulation results for Test I; scenario ATC ground system corrupted
The results show a significant effect on the total system. The ATC ground system is responsible
for the resolution advice for all aircraft. If the ATC ground system is corrupted the system does
not detect conflicts. When there are no conflicts detected, there is no resolution advice generated.
The aircraft will then continue their path along the given trajectories. Hence, the A3G baseline
parameter value cannot be changed to the test I value. In Table 4.11 the outcome of test I is
presented with a green background.
Table 4.11: Non-baseline parameter values for the corrupted ATC ground system scenario
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test I
value
ATC Ground
System
ATC System
Mode
corr
ATCsysp Probability of ATC ground
system Corrupted
101*10
55*10
Page 53 of 96
4.12 Test J: ATC ground system failure
This parameter concerns the probability of failure of the ATC Ground system, and is used in the
ATC system mode in the ATC ground agent. When the ATC ground system fails the system
doesn’t detect conflict but does indicate that the ATC system has a failure.
The ATC ground system agent is newly added to the model, but is originated from the ASAS
agent in the A3 model. The ATC system mode is a direct copy of the ASAS system mode,
therefore the non-baseline test value is based on the A3 model baseline value.
In Figure 4.11 the Monte Carlo simulation for test J are presented. Figure 4.11 shows that the
effects on the simulations results under the non-baseline parameter setting are significant. The
results are very different from the A3G model baseline parameter results. The effect of the ATC
system failure is comparable to ATC system corrupted in Figure 4.10.
Figure 4.11: Monte Carlo simulation result for Test J; ATC ground system failure scenario
The difference between a system failure and being corrupted is as follows. In both situations the
ATC system doesn’t detect conflicts. Thus there is no resolution advice generated. When the
system has a failure it shows a failure indication, this in contrast to ‘corrupted’ when no
indication is shown. In the A3G model there is only one ATC ground system. There are no back-
up systems present. The effect of failure or corrupted is the same. In view of the significant
effect on the results, the A3G baseline parameter value should not be changed to the test J value.
In Table 4.12 the outcome of test J is indicated by a green background.
Page 54 of 96
Table 4.12: Non-baseline parameter setting for the ATC ground system failure scenario
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test J
value
ATC Ground
System
ATC System
Mode
down
ATCsysp Probability of ATC Ground
System Failure
101*10
55*10
4.13 Test K: Ground ADS-B receiver
In the A3G model the ADS-B ground receiver is a LPN in the ATC Ground agent. It has two
places and can be working or not working. The ADS-B ground receiver is a new LPN in the
model, but is derived from the airborne aircraft ADS-B receiver. Therefore the A3G non-
baseline test value is based on the A3 model baseline value.
The switches between the two modes happens at exponentially distributed times. The ADS-B
ground receiver mode is only connected to the Surveillance LPN of the ATC system. When the
ADS-B ground receiver is not working the Surveillance LPN is unable to receive intent or state
information. The system will then try to use the old information. When the information becomes
too old, the system will delete this. For state information this timeframe is 10 seconds, for intent
information this is 6 minutes. See appendix C, parameters 151 and 152. In Figure 4.12 the
Monte Carlo simulation results for test scenario K are presented.
Figure 4.12: Monte Carlo simulation result for Test K; ADS-B ground receiver scenario
Page 55 of 96
The simulation results in Figure 4.12 are different from the A3G baseline setting results. The
effects on the results in Figure 4.12 are significant. In the A3G model when the ADS-B ground
receiver is down the ATC ground system is not able to receive intent or state information from
any aircraft. In Table 4.13 the outcome of Test K is presented with a green background.
Table 4.13: Non-baseline parameter setting for the ADS-B ground receiver scenario
Agent LPN Parameter Explanation
A3
G
Baseline
Value
Test K
value
ATC
Ground
System
ADS-B ground
receiver mode ,
down
ATC ADS RECp
Probability of ADS-B Ground
receiver Not Working 101*10
55*10
4.14 Test L: ATCo-Tactical maximum response time
The ATCo agent consists of two parts: The Tactical, ATCo-T, which is in charge of the short
term conflicts (STC) resolution advices. Second is the Planning, ATCo-P, which handles all the
medium term conflicts (MTC) and back-to-goal (B2G) resolution advices. In either case, the Air
Traffic Controller has to check if the resolution advice generation by the ATC system is accepted
or not. When the ATCo accepts a resolution advice it is given to the ATC Uplink transmitter.
Because the ATCo does not exist in the A3 model, no reference parameter values can be taken
from the A3 model. For test L a maximum ATCo-T response of 10 s is used.
Figure 4.13: Monte Carlo simulation result for Test L; the ATCo-Tactical scenario
Page 56 of 96
The simulation results in Figure 4.13 are slightly different from the A3G model baseline
parameter results shown in Figure 4.2. The effect of the non-baseline parameter setting for the
Tactical Air Traffic Controller is thus significant. In Figure 4.1, the baseline parameter results of
the A3G model the smallest miss distance obtained was 4.6 Nm, while in Figure 4.13 it is 4.4
Nm. The non-baseline test L value for the ATCo-T response time is a factor 10 in comparison
with the baseline. Although this is a large step the effect is noticeable but small. The ATCo-T
only deals with the short term conflict resolution advice; in this procedure time is an important
factor.
Although the effect is only small on the simulation results the baseline value for the ATCo
response time cannot be changed to the non-baseline value of 10s.
In Table 4.14 the outcome of test L is presented as being undecided yet.
Table 4.14: Non-baseline parameter value for the ATCo-Tactical scenario
Agent LPN Parameter Explanation
A3
G
Baseline Value
Test L
value
ATCo ATCo-Tactical
min
ATCo TT
ATCo-T minimum response
time
1 s 10 s
max
ATCo TT
ATCo-T maximum
response time
1 s 10 s
Test L-2
In the previous test scenario L the effect on the simulation results were significant but not very
large. Therefore an additional test scenario L-2 is performed with the parameter setting shown in
Table 4.15.
Table 4.15: Non-baseline parameter value for the ATCo-Tactical scenario
Agent LPN Parameter Explanation
A3
G
Baseline Value
Test L-2
value
ATCo ATCo-Tactical
min
ATCo TT
ATCo-T minimum response
time
1 s 5 s
max
ATCo TT
ATCo-T maximum
response time
1 s 5 s
Figure 4.14 shows the Monte Carlo simulation results of test scenario L-2. The smallest miss
distance obtained is around the 4.45 Nm. This is still different from the A3G model baseline
parameter result of 4.6 Nm. Also n the lower part of the graph smaller miss distances are
obtained in comparison with the A3G model baseline results in Figure 4.1. The results in the
lower part of the graph depend on the Short Term Conflict resolution capacities of the system.
The ATCo-T plays an important role in solving those conflicts. As the effect of the test L-2 value
is still noticeable the conclusion is that the A3G parameter value should be lower than 5 seconds.
Page 57 of 96
Figure 4.14: Monte Carlo simulation results for Test L-2; ATCo-T response time is 5 seconds in the A3G
model
Test L-3
A third test scenario is conducted. The non-baseline parameter setting in this test is shown in
Table 4.16, including an indication of the outcome of test L-3 through a green background.
Table 4.16: Non-baseline parameter value for the ATCo-Tactical scenario
Agent LPN Parameter Explanation
A3
G
Baseline Value
Test L-3
value
ATCo ATCo-Tactical
min
ATCo TT
ATCo-T minimum response
time
1 s 2 s
max
ATCo TT
ATCo-T maximum
response time
1 s 2 s
Figure 4.15 shows the Monte Carlo simulation results for test scenario L-3. The results in Figure
4.15 are the same as the A3G model baseline parameter results. The response time of the ATCo-
T can therefore be changed to the non-baseline parameter value of 2 seconds.
Page 58 of 96
Figure 4.15: Monte Carlo simulation results for Test L-3 ATCo-T is 2 seconds in the A3G model
4.15 Test M: ATCo-Planning maximum response time
The ATCo-Planning (ATCo-P) deals with the Medium term conflict resolution advisory and the
Back-to-Goal resolution advisory. Both these resolutions consist of multiple waypoints which
lead the corresponding aircraft conflict free to its final goal.
The parameter for ATCo-P response time is divided in a minimum and maximum response time
parameter. This function is not used in this test scenario, but can be used to give boundaries to
the response time. The ATCo is a newly added agent. The non-baseline test parameter of 10
seconds is a factor 10 in comparison with the A3G baseline parameter value. It is expected that
this is enough time for the Air Traffic Controller to check if the resolution advice generation by
the ATC system is conflict free and to accept this. When the ATCo accepts the resolution advice
it is given to the ATC Uplink transmitter.
In Figure 4.16 the Monte Carlo simulation results for test scenario M are presented. The results
are very similar to the A3G model baseline parameter results, but not exactly the same. There is
a small kink which starts at the 10-3 mark. So it can be stated that the non-baseline parameter
setting has a minor effect on the results. The overall results are slightly less than the baseline
parameter results of the A3G model. After the kink in the lower part the results are somewhat
more to the right side of the graph in comparison with the baseline results, which means a
smaller miss distance.
Page 59 of 96
The smallest miss distance during the one million simulations is 4.55 Nm, which is in
comparison to the baseline results negligible. This result is not significant. This ATCo-Planning
only deals with the Medium Term conflict and Back-to-Goal resolution advices. The aircraft is
not in a direct conflict when these resolutions are generated and therefore a longer response time
has almost no effect on the results.
The results of the simulation in Figure 4.16 are almost the same as the A3G model baseline
parameter results and therefore the ATCo-P response time parameter setting can be changed to
10 seconds.
Figure 4.16: Monte Carlo simulation result for Test M; ATCo-Planning response time
In Table 4.17 the outcome of test M is indicated through a green background.
Table 4.17: Non-baseline parameter value for the ATCo-Planning scenario
Agent LPN Parameter Explanation
A3
G
Baseline Value
Test M
value
ATCo ATCo-Planning
min
ATCo PT
ATCo-P minimum response
time
1 s 10 s
max
ATCo PT
ATCo-P maximum response
time
1 s 10 s
Page 60 of 96
4.16 Test N: ATC uplink transmitter
The ATC Uplink transmitter is a newly added LPN in the A3G model. The ATC uplink
transmitter is part of the ATC ground system. Just as there is only one ATC ground system, there
is only one ATC Uplink transmitter. The ATC Uplink transmitter is responsible for the sending
the resolution advices from the ground to the corresponding aircraft.
The non-baseline parameter setting is based on the send duration parameter in the ‘Broadcast
FMS Intent’ LPN of the A3 model. In the A3 model the duration of sending is derived from the
following formula:
Send Send
d Num TimeT T T
The duration for sending is the multiplication of the number of waypoints times the duration for
sending of a waypoint. For the two aircraft scenario a normal resolution advice consist of 4
waypoints. The duration for sending of a waypoint is 3 seconds (see appendix C, parameters 90
and 91), which yields a total of 12 seconds.
In Figure 4.17 the Monte Carlo simulation results for test N are presented.
Figure 4.17: Monte Carlo simulation results for Test N; ATC uplink transmitter
The results in Figure 4.17 are very different from the A3G model baseline parameter simulation
results in Figure 4.1.
In order to better understand the difference between the results obtained for the A3 model and
for the A3G model we compare the delays under both models. Under the A3 model the delay is
Page 61 of 96
largely caused by the decision-making delay of the flight crew. The probability density function
of this delay is presented at the top of Figure 4.18 in the form of a Rayleigh shaped density with
mean value of 5.6 s. Under A3, the flight crew can synchronize the tactical decision-making with
the implementation of this decision.
However, under A3G the flight crew no longer has this tactical decision-making power; this is
now done by the ATCo-T (takes 1 s only in the MC simulation of Figure 4.17). Of course the
ATCo-T does this in an optimal way. Though from that moment on there are two sequential
delays: 1) the delay of the data uplink, and 2) the delay of pilot acceptance and implementation
of the tactical instruction received. The sum of these two delays is pictured through the bottom
pdf in Figure 4.18; this is the same Rayleigh density as the one at the top of Figure 4.19, though
now shifted 12 s to the right, which is the uplink delay in the MC simulation of Figure 4.17. This
has as consequence that the chance to be too late (e.g. the probability of more than 30 s delay) is
many orders of magnitude larger.
Figure 4.18: Top: Rayleigh probability density function with mean delay of 5.6 s.
Bottom: The same Rayleigh probability density function shifted to the right by 12 s.
Page 62 of 96
Thus there are two key differences between A3 and A3G:
- The extra delay by the data uplink increases the probability of being too late by
many orders of magnitude.
- The tactical decision making and the implementation is now split over the ATCo-
T and the crew and is therefore no longer synchronized.
Through investigating MC simulated trajectories that ended in the tail of Figure 4.17, the specific
consequence of being too late has also been investigated. The finding is that in rare occasions
only, a 12 s delay of the uplink transmitter leads to a too late implementation of the STCR in the
aircraft. As a result of such extra delay, the next STCR will be generated by the ATC system. In
the current A3G model this next STCR is passed on by the ATCo, through the uplink transmitter
to the pilot. In some specific rare cases this may lead to an alternating series of left/right
instructions, yielding the tail in Figure 4.17.
The above also explains why that the delay by the uplink transmitter is far more critical than a
delay by the ATCo-T. The choice of the STCR update is being made by the ATCo-T in a way
that is kind of optimal at that very ATCo decision-making moment. That’s why some more delay
by the ATCo also leads to a more optimal decision. However the delays by the uplink transmitter
and by the pilot make that by the time this ATCo decision is implemented it may be far from
optimal in some rare cases. For the 2 a/c encounter scenarios this rare but undesired effect could
be mitigated by limiting the delay of the uplink transmitter to 1 second.
Due to the large effect on the simulation results the ATC uplink transmitter sending duration
parameter should not be changed to the non-baseline parameter setting. In Table 4.18 the
outcome of test N is indicated through a green background.
Table 4.18: Non-baseline parameter value for ATC uplink transmitter scenario
Agent LPN Parameter Explanation
Baseline
A3
G
Test N
value
ATC
Ground
System
ATC Uplink
Transmitter
Transmit
uplinkT
ATC Uplink transmitter
Duration of sending resolution
to aircraft
1 s 12 s
Page 63 of 96
4.17 Selected parameter values for the A3G model
In Table 4.19 an overview of the parameter values is presented. The green coloured column
shows the selected parameter values due to the outcomes of the conducted tests C-N (which aqre
taken from the green indicated outcomes in Tables 4.4 through 4.18. The fourth column shows
the A3G model baseline parameter values as presented in Table 4.1. In the last column the
corresponding baseline parameter value of the A3 model is shown for reference.
Table 4.19: Selected parameter values for the A3G model compared to the A
3G baseline parameter
values and the A3 baseline parameter values.
# Parameter Explanation
A3
G
Baseline
value
A3
G
Selected
value
A3
Baseline
value
62 down
SATp Probability of Global GNSS/GPS Not working 101*10
51*10
51*10 =
66 occ
ADS Bp Probability of ADS-B global Occupied
101*10
61*10
61*10 =
69
occ
ADS B
Mean duration of ATC global Occupied Not
Occupied
1 hr. 1hr. -
70 down
ATC globalp
Probability of Global ATC uplink frequency
Occupied
101*10
81 * 10
-
94 down
GNSS RECp
Probability of Airborne GPS receiver Not
Working
101*10
61*10
55*10 ≠
98 down
Altimp Probability of Airborne Altimeter Not Working 101*10
55*10
55*10 =
111 down
ADS TRMp Probability of ADS-B transmitter Not Working 101*10
101*10
55*10 ≠
165 ownx Position of ATC ground system [x,y,z]
[0,0,0] [0,0,0] -
168 corr
ATCsysp Probability of ATC ground system Corrupted 101*10
101*10
55*10 ≠
169 down
ATCsysp Probability of ATC ground system Not working 101*10
101*10
55*10 ≠
170
2B GT ATC ground system, Interval time for Back-to-
Goal Evaluation
20 s 20 s -
172 ,
down
ATC ADS recp
Probability of ADS-B ground receiver Not
Working
101*10
101*10
55*10 ≠
173 Transmit
uplinkT ATC Uplink Transmitter duration of sending
resolution to aircraft
1 s 1 s -
174 min
ATCo TT ATCo-Tactical minimum response time 1 s 2 s -
175
max
ATCo TT
ATCo-Tactical maximum
response time
1 s 2 s -
176 min
ATCo PT ATCo-Planning minimum response time 1 s 10 s -
177 max
ATCo PT ATCo-Planning maximum response time 1 s 10 s -
Page 64 of 96
In Figure 4.19 the Monte Carlo simulation results for the A3G model with the selected baseline
parameter values as presented in Table 4.19 are shown.
Figure 4.19: Monte Carlo simulation results for the selected baseline parameter values of the A3G model.
The results Figure 4.19 are the same as the A3G model under baseline parameter values. The
results show that some parameters could be slightly changed without negatively affecting the
simulation results. The A3G model selected parameter values are closer to the baseline ones of
the A3 model than the A3G baseline parameter values.
4.18 Interpretation of the 2 aircaft encounter results obtained
In this subsection the results obtained for the two aircraft encounter tests C-N are elaborated.
Also the influence of the A3G model assumptions A1-A5 (see subsection 3.1) is considered. In
Table 4.20 a summary of the A3G model results is presented. The first column indicates the
specific test. The second column represents the key model parameter, which was changed in the
specific test. The last column denotes the effect on the results of the change in the key parameter.
The effects can be none, negligible, significant or large. A significant effect can be
surmountable, but this is not possible for a large effect. In none of the test scenarios was a
positive effect on the results observed.
Table 4.20: Overview of the A3G model results from the two aircraft encounters scenarios.
Test Changed parameter Measurable effect
C Global GPS not working (down) none
Page 65 of 96
D Global ADS-B occupied (down) none
E Global ATC Uplink Frequency occupied (down) large
F Airborne GPS receiver not working (failure) significant
G Altimeter not working (failure) none
H ADS-B transmitter not working (failure) large
I + J ATC ground system mode failure / corrupted large
K ADS-B ground receiver not working (failure) large
L ATCo-Tactical maximum response time significant
M ATCo-Planning maximum response time negligible
N ATC uplink transmitter sending duration large
There are seven model parameters with non-negligible effects:
E. Global ATC Uplink Frequency occupied (down)
F. Airborne GPS receiver not working (failure)
H. ADS-B transmitter not working (failure)
IJ. ATC Ground system mode failure / corrupted
K. ADS-B ground receiver not working (failure)
L. ATCo-Tactical maximum response time
N. ATC Uplink transmitter sending duration
Each of these seven is discussed next, taking into account assumptions A1-A5 adopted for the
A3G model.
E. Global ATC uplink frequency occupied (down)
The global ATC uplink frequency is used in the A3G model to send the resolution advices from
the ATC ground system to the aircraft. If this frequency is occupied then no resolution is send.
The aircraft will continue to fly according to their current conflicted flight plan.
The MC simulation results showed that a very high dependability of 81*10 appeared necessary
to obtain similar results as the A3 model. Because none of the assumption A1-A5 has influence,
this may look like a very high requirement. However it should be taken into account that this
high dependability requirement applies for a mean duration of the occupancy of 1 hr (see #69 in
Table 4.1). At factors 10 or 100 lower mean durations, the dependability requirement may go
down by the same factors. This brings the Global ATC uplink requirements at a practically
manageable level.
F. Airborne GPS receiver not working (failure)
The effect of an airborne GPS receiver failure in the A3G model is different from the effect in
the A3 model. In the A3 model a not working GPS receiver doesn’t have a large effect on the
results, because the own aircraft still resolves all the conflicts. However in the A3G model this is
not the case anymore. The position error of an aircraft increases when the GPS receiver is not
working. In the ATC ground system this difference between the real position of aircraft-i and the
4D RBT-i information results in a not conform RBT. The ATC ground system then drops the
intent information of aircraft-i. When only the state information of aircraft-i can be used, this
results in a short term conflict.
Page 66 of 96
The Monte Carlo simulation results showed that in the A3G model a dependability of 61*10 is
necessary. In the A3 model this safety requirement is 55*10 , which is a factor 50 less good.
One should be aware that in the A3G ConOps ground based navigation support will be well
available. This overrules A3G model assumption A5, and implies that for the A3G ConOps there
likely is no problem in realizing a 50 times higher navigation dependability.
H. ADS-B transmitter not working (failure)
The large effect on the results of the ADS-B transmitter is different in the A3G model with
respect to the A3 model. In the A3 model the situation would be as follows. If the airborne ADS-
B transmitter of aircraft-i fails than other aircraft-k are unable to receive state and intent
information of aircraft-i. Without state and intent information aircraft-k cannot safely resolve the
conflict and thus does nothing. But in the A3 model aircraft-i still receives state and intent
information of aircraft-k and thus aircraft-i can resolve the conflict.
In the A3G model separation is controlled from the ground. If the ADS-B transmitter of aircraft-i
fails, the ATC ground system doesn’t receive the state and intent information of aircraft-i. Hence
no resolution with aircraft-k is possible. Outdated state and intent information is dropped by the
ATC ground system after predetermined times. The ATC ground system is then unaware of the
state and intent of aircraft-i both the medium term as tactical layer are unable to generate a
resolution for aircraft-i or aircraft-k. ‘
The MC simulations results showed that a dependability of 101*10 was necessary to obtain
similar results as the A3 model. In the A3 model the safety requirement for the ADS-B
transmitter is 55*10 . One should be aware that in the A3G ConOps assumption A4 does not
hold true, i.e. an RBT in the ATC system will not so rapidly be considered to be unreliable when
ADS-B transmitter is down. This means that for the A3G ConOps it likely would not be required
that the ADS-B transmitter should realize this very high dependability.
IJ. ATC Ground System failure / corrupted
If the ATC ground system is down or corrupted both the Medium Term TBO layer as the Short
Term layer are not working. This means there are no conflicts detected and no resolution
process. The aircraft will continue flying their current conflicted flight plan (RBT).
In the A3 model the separation is controlled by the ASAS. The dependability for the ASAS is 55*10 in the A3 model. In the A3G model the dependability for the ATC ground systems needs
to be 101*10 , which is a much higher safety requirement. One should be aware that this kind of
very high dependability requirement does not come as a surprise, because similarly high
dependability requirements already apply to current ATC ground system in busy airspace.
K. ADS-B Ground receiver not working (failure)
In the A3G model the ADS-B ground receiver is used by the ATC ground system to receive the
state and intent information of all aircraft. When the ADS-B ground receiver is not working there
is no new information received. In the A3 model when the ADS-B ground receiver of aircraft-i is
down the other aircraft are still capable of resolving the conflicts with the available state and
intent information. In the A3G model when the ADS-B ground receiver is down this applies to
all aircraft. Hence none intent and state information is received. The ATC ground system is then
unable to resolve conflicts, due to outdated information.
Page 67 of 96
The Monte Carlo simulation results showed that the dependability for the ADS-B ground
receiver needs to be 101*10 . One should be aware that in spite of assumption A3 for the A3G
model, in the A3G ConOps SSR Mode-S radars are fully in use, which avoids the very high
dependability requirement for the ADS-B ground receiver.
L. ATCo-Tactical maximum response time
In the A3 model there are no ATCo’s present. Conflict detection and resolution advice is
performed by the airborne ASAS and is directly presented to the Pilot-Flying. In the A3G model
conflict detection and resolution advice is performed by the ATC ground system. The resolution
advice is then first checked by the ATCo before sending it to the corresponding aircraft using the
ATC Uplink transmitter.
For the A3G model the results in Figure 4.13 through 4.15 show that the ATCo-T response
should not take longer than 2 seconds in order to get A3 model simulation results. This 2 seconds
response time requirement is not realistic at all. None of the A1-A5 assumptions influences this.
Though, the good news is that the results in Figures 4.13-4.15 also show that the A3G deviation
from the A3 results is rather small when the ATCo-T response is increased from 2 seconds to 5
or 10 seconds.
N. ATC Uplink transmitter sending duration
The ATC Uplink transmitter is used to send resolution advices generated by the ATC ground
system to the aircraft. For the A3G model to get A3 model simulation results it appears
necessary that the uplink transmission does not take longer than 1 second. None of the A1-A5
assumptions influences this.
In contrast with the finding for the ATCo-T response time above, the A3G deviation from the A3
results is far larger when the ATC uplink transmitter sending duration is increased from 1 s to 12
s. In subsection 4.16 it has been explained that this big difference in sensitivities is because the
ATCo-T is still able to decide optimal at the moment of decicion-making. However, the uplink
transmitter just adds delay, and the pilot no longer can synchronize tactical maneuver selection
with implementation. As a result, in rare cases the optimal decision by the ATCo has become
obsolete by the time it is implemented by the pilot.
Overall finding
To conclude the described 2 aircraft encounter MC simulation results: It is possible to obtain
similar results with the A3G model as obtained for the A3 model. However to obtain these
results the A3G baseline parameter values that were based on the A3 baseline parameter values,
needed to be changed to better values (which are referred to as the A3G selected parameter
values). When taking into account assumptions A1-A5, two key parameters remain for which
unexpectedly high requirements apply:
L. ATCo-Tactical maximum response time (2 seconds)
N. ATC Uplink transmitter sending duration (1 second)
Specifically, the 2 second response time for the ATCo-Tactical differs a lot from current
practice. Also demanding are the requirements on the ATC Uplink speed.
Page 68 of 96
5 MC SIMULATION OF 8 AIRCRAFT ENCOUNTERS
In this Section Monte Carlo simulation results for eight aircraft encounters under the A3G model
are presented and discussed. This is organized as follows. First, in subsection 5.1 eight aircraft
encounter scenarios are presented using the A3G baseline parameter values as well as the A3G
selected parameter values from Table 4.19. Next, in subsection 5.2 the effects reducing of the
pilot’s response time on the simulation results in the eight aircraft encounter scenario are shown.
In Subsection 5.3 the eight aircraft scenario A3G model Monte Carlo simulation results are
compared to eight aircraft baseline results of the A3 model in [Blom&Bakker, 2011a,b].
5.1 A3G selected and A3G baseline parameter values applied to 8/ac encounters
In Figure 5.1 the Monte Carlo simulation results are presented for the eight aircraft encounter
using the A3G selected parameter values.
Figure 5.1: MC simulation results for Eight aircraft scenario under A3G model with A
3G selected
parameters (100 thousand MC runs).
The results of the A3G model under A3G selected parameter values in Figure 5.1 are very
different from the eight aircraft scenario MC results of the A3 model under A3 baseline
parameter values [Blom & Bakker, 2011]. The first part of the curve is the same, but beyond a
Page 69 of 96
probability of 210 , the curve is completely different. The A3G model under the selected
parameter values doesn’t resolve Short Term Conflicts as well as the A3 model does. In Figure
5.2 the MC simulation results are shown for the A3G baseline parameter values. Main difference
is faster responses by ATCo-T (from 2s to 1s) and ATCo-P (from 10s to 1s). Moreover, the
dependability of various technical systems has been improved.
Figure 5.2: MC simulation of eight aircraft encounter under A3G model and A3G baseline parameter
values (135 thousand MC simulation runs).
The results in Figure 5.2 show a significant improvement over the results in Figure 5.1.
Nevertheless there still is a major difference with the rare event MC simulation results obtained
for the A3 model under A3 baseline parameter values. For the part of the curve at left of the 5
Nm point, the A3G model results are similar to those for the A3 model. However, for the part of
the curve at right of the 5 Nm point the results are very different from those for the A3 model.
Because of these large differences identified, some realized MC simulation results in the tail of
the curve in Figure 5.2 have been analyzed on what happens. This showed that the cause of the
problem lies in the tail of the delay by the pilot in implementing an updated MTCR or STCR. In
the A3 ConOps such rare lengthy delay by a pilot may also happen, though then this has a
completely different impact. In the A3 model the choice for a new MTCR or STCR is made by
the pilot on the basis of the actual traffic situation, and then it is immediately implemented in the
FMS or through the mode control panel. However, in the A3G model the choice of the MTCR or
Page 70 of 96
STCR update is being made by the controller, also in a way that is an optimal decision at that
very moment. However from that moment on it takes some time until such MTCR or STCR is
being implemented by the pilot. This means that there occur situations in which the optimal ATC
decision is no longer optimal at the moment of implementation by the pilot. According to Figure
5.2, once in 1000 flight hours this leads to a serious mismatch in optimal timing of the MTCR
and/or STCR decision-making by the ATCo relative to the moment that it is implemented by the
pilot.
This explanation also would mean that in the A3G model this problem can easily be resolved by
shortening the time delay caused by the pilot to a very low value. This very fast pilot response is
tested next.
5.2 A3G baseline parameter values, except a very fast pilot response
In order to see the effect of a rapid reaction by pilots, in this subsection the effect of a very fast
reaction time of the Pilot-Flying is investigated through Monte Carlo simulations. All parameter
values are taken from the A3G baseline values, with the exception of the mean decision delay
time and monitoring duration of pilot flying (parameter numbers 14 and 24 in Appendix C). In
Table 5.1 the parameter values are shown.
Table 5.1: Parameter values for the faster Pilot Flying response times for MTC and Back-to-Goal
resolution advice implementation
Agent LPN Parameter Explanation
A3
G
Baseline
value
Test
eight a/c
value
Pilot-
Flying
Task Performance
Goal 3: Conflict
Resolution (STC &
MTC)
#14
2dT
Mean decision delay time MTC
Monitoring & Decision
Execution
30 s 1 s
Task Performance
Goal 5: Navigation
horizontal actions
(Back2Goal)
#24
Mon
PF
Mean duration of Monitoring
Monitoring & Decision
20 s 1 s
Figure 5.3 shows the Monte Carlo simulation results of the test scenario. This confirms the
expected major improvement relative to the results in Figure 5.2. The curve in Figure 5.3 is
almost as good as the curve obtained for 8 a/c encounters under the A3 ConOps with the A3
baseline parameter values.
In Figure 5.3 a miss distance of 3.8 Nm at the 610 frequency level is obtained for the A3G
model in comparisons to 4.1 Nm at the 610 frequency level for the A3 model.
In spite of this dramatic improvement of the A3G model, there still is some difference with the
curve obtained for the A3 model. However, this difference lies below the 610 frequency level,
which means that the A3G results are not statistically meaningful. In order to reach statistically
meaningful results a factor 10 more MC runs should be conducted with the A3G model.
Page 71 of 96
Figure 5.3: Monte Carlo simulation results A3G baseline parameter values except with very faster pilot
response time (2.1 million MC runs).
Figure 5.4: Comparison of A3G model results (blue curve) versus A
3 model results (black curve).
Page 72 of 96
5.3 Findings for 8 a/c encounters
Figure 5.4 compares the simulation results from Figure 5.3 with similar MC simulation results of
the A3 model for the 8 a/c encounter scenario using the A3 baseline parameter values.
Comparison of this curve with the curve in Figure 5.3 shows that the differences are quite small.
To conclude the described 8 aircraft encounter MC simulation results: It is possible to obtain
similar results with the A3G model as obtained for the A3 model. However to obtain these
results the A3G baseline parameter values that were based on the A3 baseline parameter values,
needed to be changed to better values. When taking into account assumptions A1-A5, four key
parameters remain for which unexpectedly high requirements apply:
L. ATCo-Tactical maximum response time (1 second)
M. ATCo-Planning maximum response time (1 second)
N. ATC Uplink transmitter sending duration (1 second)
8. Pilot maximum response time (1 second)
Specifically, the 1 second response time for the Pilot as well as the ATCo-Tactical and ATCo-
Planning differs a lot from current practice. Also demanding are the requirements on the ATC
Uplink speed.
Page 73 of 96
6 RANDOM TRAFFIC SCENARIOS
6.1 Monte Carlo simulation results for random traffic scenarios
For the A3 model rare event MC simulation results have been obtained by making use of a
Periodic Boundary Condition (PBC). This way it was possible to simulate a very large area
through running rare event MC simulations for eight aircraft only [Blom & Bakker, 2011a,b].
For the A3G model there are two challenges to apply this approach: i) The A3G model asks far
more CPU time than the A3 model; and ii) There is a need to develop a dedicated approach to
applying the PBC to the ATC part of the A3G model.
A straightforward approach towards the latter problem would be to make the PBC area such
large that the number of aircraft in it correspond to the number of aircraft in one ATC sector
under the 3x 2005 high traffic demand. This would ask to use a PBC with 24 or more aircraft,
which is 3x as many as have been used in to simulate random traffic scenarios under the A3
model. This factor 3 extra simply multiplies the already higher CPU load of the A3G model.
Hence it appeared to be infeasible to run MC simulations for random traffic scenarios under A3G
model within the time-frame of this D2.2 report schedule.
However, what we will do is to make a theoretical derivation of the activity loads on the
Planning ATCo, the Tactical ATCo and the pilot crew under the hypothetical assumption that the
A3G model would yield the same aircraft trajectories as the A3 model does. In view of the
results obtained for two and eight aircraft encounters this seems to be a rather optimistic
assumption for the A3G model. Thanks to this optimistic assumption for the A3G model, it
becomes possible to predict pilot crew and ATCo resolution activity frequencies simply by
measuring these frequencies for the A3 model under random traffic demands. This is the
accomplished in the current section.
First, subsection 5.2 provides the A3 measured activity frequencies for the pilot crews. Under the
hypothetical assumption about A3G model, these very same frequencies also apply to the pilot
crews under the A3G model. Next, subsections 5.3 and 5.4 use the results of subsection 5.2 for
the prediction of the activity frequencies for a Planning ATCo and a Tactical ATCo respectively.
6.2 MTCR and STCR activity frequencies for pilot crews
Table 5.1 shows the mean STCR and MTCR conflict resolution activity frequencies of a flight
crew that have been measured during rare event MC simulations of the A3 model. In view of the
hypothetical assumption mentioned above, these frequencies are assumed to also apply under the
A3G model. For 3x high 2005 traffic demand, the STCR/MTCR total amounts about one conflict
resolution activity per five minutes. To put this value into perspective, in current European en
route airspace a flight crew receives on average one executive instruction from ATC per eight
minutes [Eurocontrol, 2014]. Hence the mean frequency of executive activities by flight crew is
under 3x 2005 high traffic demand only 60% higher than the current average in en route
European airspace.
Page 74 of 96
Table 5.1 also shows what happens when the random traffic demand goes up by another
factor 2. Under this 6x high 2005 traffic demand the MTCR and STCR activity frequencies go
also up by a factor three to four, to a total MTCR/STCR crew activity frequency of about 0.7 per
minute.
Table 5.2 shows the effect of changing APN1 by a factor 2 in either direction. A factor two
change in ANP value leads to a 10-25% change in the STCR activity frequencies of the flight
crew. These ANP changes have marginal effect on the MTCR activity frequency.
Table 5.3 shows that a wind prediction error of 30 m/s leads to a significant increase of
STCR/MTCR activities for the crew. The mean frequency increases from about one activity per
five minutes to one activity per two minutes. Because this high increase of conflict resolution
activities applies for a short time only, it is expected that this does not form an unmanageable
problem. Through running additional MC simulations it has been verified that an improvement
from ANP1 to ANP.5 may reduce the total MTCR/STCR activity frequency by some 10%, i.e. to
one activity per two minutes.
6.3 Predicted A3G ConOps MTCR and STCR activity frequencies for Planning ATCo
In view of the results found for two and eight aircraft encounters under the A3G ConOps, in
theory it is possible to let the A3G ConOps do the same from the ground as what the A3 ConOps
is doing from the air. The total activity for each pilot crew is expected to be the same as has been
measured for the A3 ConOps, in 5.1. These results allow us to predict the STCR and MTCR
activity loads for the tactical and planning ATCo by multiplying the A3 measured frequencies by
Table 5.1 Mean frequency of STCR and MTCR activities per aircraft under no wind prediction error
3x & 0 m/s 6x & 0 m/s
Mean MTCR frequency 0.11 per min 0.42 per min
Mean STCR frequency 0.08 per min 0.26 per min
Mean Total frequency 0.19 per min 0.68 per min
Table 5.2 Impact of ANP on the mean frequency of STCR and MTCR activities per aircraft under
3x high 2005 traffic demand and 0 m/s wind prediction error
3x & ANP.5 3x & ANP1 3x & ANP2
Mean MTCR frequency 0.10 per min 0.11 per min 0.11 per min
Mean STCR frequency 0.06 per min 0.08 per min 0.09 per min
Mean Total frequency 0.16 per min 0.19 per min 0.20 per min
Table 5.3. Effect of 30 m/s wind prediction error on mean frequency of STCR and MTCR activities
3x & 0 m/s 3x & 30 m/s
Mean MTCR frequency 0.11 per min 0.15 per min
Mean STCR frequency 0.08 per min 0.40 per min
Mean total frequency 0.19 per min 0.55 per min
Page 75 of 96
the number of aircraft in a A3G sector. The results of such predictions are shown in figures 5.1
and 5.2, for the planning and tactical ATCo respectively.
Figure 5.1 Predicted MTCR activity frequency of a Planning ATCo as a function of the
number of aircraft in a sector. Purple line: 3x 2005 high and no wind error; green line: 3x
2005 high and 30 m/s wind prediction error; red line: 6x 2005 high and no wind error.
Planning ATCo
According to the curve in Figure 5.1, under the A3G ConOps, in a sector of 27 aircraft, a
Planning ATCo has to perform 3 MTCR activities per minute under 3x high 2005 traffic demand
and no wind prediction error. This is a demanding task level, though is expected to be
manageable by a well trained planning controller for the A3G ConOps. Under wind prediction
errors of up to 30 m/s, the MTCR activity frequency goes up to about 4 MTCR activities per
minute. Because this higher load will continue for a short period only, this also seems to be
manageable for an ATCo who is well trained for the A3G ConOps. However when the traffic
demand increases up to 6x high 2005 level, then the MTCR activity frequency goes up to more
than 10 per minute. This is expected to be an unreasonably high load which cannot be safely
managed by a planning ATCo. This means that the Flow Control should assure that local traffic
demands do not really go beyond the 3x high 2005 level.
Page 76 of 96
6.4 Predicted A3G ConOps MTCR and STCR activity frequencies for Tactical ATCo
Figure 5.2 Predicted STCR activity frequency of a Tactical ATCo as a function of the
number of aircraft in a sector. Purple line: 3x 2005 high and no wind error; green line: 6x
2005 high and no wind error; red line: 3x 2005 high and 30 m/s wind prediction error.
Tactical ATCo
According to the curve in Figure 5.2, under the A3G ConOps, in a sector of 27 aircraft, a
Tactical ATCo has to perform about 2.5 STCR activities per minute under 3x high 2005 traffic
demand and no wind prediction error. This is a demanding task level, though is expected to be
manageable by a well trained tactical controller for the A3G ConOps. However, under wind
prediction errors of up to 30 m/s, the STCR activity frequency goes up to more than 10 STCR
activities per minute. This is expected to be an unreasonably high load which cannot be safely
managed by a tactical ATCo. This rapidly increasing STCR activity load under non-ideal
conditions makes the A3G ConOps a less realistic future ATM design.
Page 77 of 96
7 CONCLUSION
One of the key innovations in SESAR2020+ is the introduction of a strategic TBO layer
[SESAR, 2007, 2012]. In this report for the first time rare emergent behaviour has been studied
for a ground-based future concept that makes use of both a strategic TBO layer and a tactical
resolution layer. The development of the Trajectory Based Operations A3G ConOps and model
is described. The A3G ConOps and model are derived from the A3 ConOps and model. Inherent
to this, the A3G model is a hypothetical model which for example does not yet address aircraft
emergency situations.
For the A3 model it was shown that the combination of a strategic TBO layer and a tactical layer
had a significant positive effect on safely accommodating high traffic demands. The objective of
this research was to investigate by rare event Monte Carlo (MC) simulations if the same positive
results could be obtained with the A3G model as with the A3 model. In the A3G model all
separation is controlled from one ground-based Air Traffic Control (ATC) system instead of by
each aircraft separately as in the A3 model. The A3G model was evaluated for three types of
scenarios, namely two aircraft head-on encounter scenarios, eight aircraft head-on encounter
scenarios, and random traffic scenarios with very high traffic demand.
The Monte Carlo simulation results for the two aircraft encounter scenario showed that the A3G
model was capable to deliver the same results as the A3 model. However, this posed high
requirements on the settings of the parameter values of the A3G model. By taking into account
the A3G model assumptions adopted, most of these parameter value requirements appeared to be
practically manageable. However, crucial exceptions have been found regarding the maximum
response times of the tactical ATCo (2 seconds) and of the ATC uplink transmitter (1 second).
The MC simulations results for the eight aircraft encounter scenario showed that in order for the
A3G model to obtain similar simulation results as the A3 model, additional requirements have to
be posed on the maximum response times of the ATCo-Tactical (1 second), the ATCo-Planning
(1 second), and the Pilot (1 second). Under these parameter settings the A3G model was able to
perform like A3 model did for the 8 a/c scenarios.
The key explanation of this need of such short response times is as follows. In the A3 model the
choice for a new MTCR or STCR is made by the pilot on the basis of the actual traffic situation,
and then it is immediately implemented in the FMS or through the mode control panel. However,
in the A3G model the choice of the MTCR or STCR update is being made by the controller, also
in a way that is an optimal decision at that very moment. However from that moment on it takes
some time until such MTCR or STCR is being implemented by the pilot. This means that there
occur rare but non-negligible situations in which the optimal ATC decision is no longer optimal
at the moment of implementation by the pilot. For the 8 a/c encounter scenarios this rare effect
could be avoided by reducing all responses by pilots, ATCo’s as well as the uplink transmitter to
1 second. For the 2 a/c encounter scenario the requirements are far less demanding, though the
rare signs already were noticeable in the MC simulations.
Page 78 of 96
Through running MC simulations with the A3 model for random traffic scenarios, it has been
identified how much conflict resolution activities should be handled by the pilots, by the ATCo-
Planning and by the ATCo-Tactical under very high traffic demands. The results obtained show
that these demands are expected to be manageable by pilots, but seem unmanageable for the
ATCo’s.
In conclusion the rare event Monte Carlo simulations showed that the A3G model studied is not
able to safely accommodate very high air traffic demand as well as the A3 model can.
Page 79 of 96
8 REFERENCES
[Abe & Yoshiki,
2001]
Y. Abe, M. Yoshiki, “Collision avoidance method for multiple autonomous mobile
agents by implicit cooperation”, Proc. IEEE/RSJ Int. Conf. Intelligent Robots and
Systems (IROS 01), pp. 1207–1212.
[Blom & Bakker,
2011a]
Blom, H., & Bakker, G. (2011). iFly Deliverable D7.4: Final Report on Accident Risk
Assessment of the A3 model. Amsterdam: NLR.
[Blom & Bakker,
2011b]
H.A.P. Blom, G.J. Bakker, Safety of advanced airborne self separation under very
high en-route traffic demand, Proc. SESAR Innovation Days, ENAC, Toulouse, 29
November-1 December 2011.
[Blom & Bakker,
2012]
Blom, H., & Bakker, G. (2012). Can airborne self separation safely accommodate
very high en-route traffic demand? 12th AIAA Aviation Technology, Integration, and
Operations (ATIO) Conference. Indianapolis, Indiana: American Institute of
Aeronautics and Astronautics, Inc.
[Blom & Bakker,
2013]
Blom, H., & Bakker, G. (2013). In search of positive emergent behaviour in
Trajectory Based Operations. Proceedings of the Third SESAR Innovation days.
Stockholm.
[Bujorianu, 2004]
M.L. Bujorianu. Extended stochastic hybrid systems. In Proceedings of Hybrid
Systems Computation and Control, Eds. O.Mahler, A. Pnuelli, LNCIS number 2993,
Springer, Berlin, pages 234–249, 2004.
[Cassandras &
Lafortune, 1999]
C.G. Cassandras, S. Lafortune. Introduction to Discrete Event Systems, Kluwer
Academic Publishers, Boston, 1999.
[David & Alla, 1994] R. David, H. Alla. Petri Nets for the modeling of dynamic systems - A survey,
Automatica, Vol. 30, No. 2, pages 175–202, 1994.
[EMERGIA, 2012] EMERGIA project web site http://emergia.nlr.nl
[EUROCONTROL,
2014]
Eurocontrol, Airborne Conflict Safety Forum, Organized by Flight Safety Foundation,
European Regions Airline Association and Eurocontrol, 10-11 June 2014, Brussels,
Findings and conclusions, V1.0, 18th July 2014
[Everdij & Blom,
2003]
M.H.C. Everdij, H.A.P. Blom. Petri nets and hybrid state Markov processes in a
power-hierarchy of dependability models. In Proceedings of IFAC Conference on
Analysis and Design of Hybrid Systems, Saint-Malo Brittany, France, pp. 355–360,
June 2003.
[Everdij & Blom,
2005]
M.H.C. Everdij, H.A.P. Blom. Piecewise deterministic Markov processes represented
by Dynamically Coloured Petri Nets, Stochastics, Vol. 77, pp.. 1–29, 2005.
[Everdij & Blom,
2006]
M.H.C. Everdij, H.A.P. Blom. Hybrid Petri nets with diffusion that have into
mappings with generalised stochastic hybrid processes. . Eds: H.A.P. Blom, J.
Lygeros. Stochastic Hybrid Systems: Theory and Safety Critical Applications, LNCIS
series, Springer, Berlin, July 2006, pp. 31–64.
[Everdij & Blom,
2010]
Everdij, M., & Blom, H. (2010). Hybrid state Petri nets which have the analysis
power of stochastic hybrid systems and the formal verification power of automata
(P. Pawlewski, Petri Nets, Chapter 12 ed.). Vienna: I-Tech Education and
Publishing.
[Everdij et al.,
2006]
M.H.C. Everdij, M.B. Klompstra, H.A.P. Blom, B. Klein Obbink. Compositional
specification of a multi-agent system by stochastically and Dynamically Coloured
Page 80 of 96
Petri Nets. Eds: H.A.P. Blom, J. Lygeros. Stochastic Hybrid Systems: Theory and
Safety Critical Applications, LNCIS series, Springer, Berlin, July 2006, pp. 325–350,
2006.
[Fiorini & Schiller,
1998]
P. Fiorini, Z. Schiller, “Motion planning in dynamic environments using velocity
obstacles”, The International Journal of Robotics Research, Volume 17 (1998), pp.
760–772.
[Haas, 2002] P.J. Haas. Stochastic Petri Nets, Modeling, Stability, Simulation, Springer-Verlag,
New York, 2002.
[Hoekstra, 2001] J. Hoekstra, Designing for Safety, the Free Flight Air Traffic Management concept,
PhD Thesis, Delft University of Technology, 2001.
[iFly D1.3, 2010]
iFly Deliverable D1.3, Autonomous Aircraft Advanced (A3) ConOps, written by
Isdefe (Gustavo Cuevas, Ignacio Echegoyen, José García García), Honeywell (Petr
Cásek, Claudia Keinrath), NLR , and Utartu (Aavo Luuk), January 2010. Available on
iFly website http://ifly.nlr.nl/
[iFly D9.1, 2009]
iFly Deliverable D9.1, Operational Services and Environmental Description (OSED)
of Airborne Self-Separation Procedure (SSEP), by E. Gelnarová, P. Cásek, August
2009.iFly Deliverable D8.1, “A3 ConOps refinement”, Editors: L. Biescas and H.A.P.
Blom, Final draft of October 2011
[Jensen, 1992] K. Jensen. Coloured Petri Nets: Basic Concepts, Analysis Methods and Practical Use,
Vol. 1, Springer, London, UK, 1992.
[Labeau et al.,
2000]
P.E. Labeau, C. Smidts and S. Swaminathan. Dynamic reliability: towards an
integrated platform for probabilistic risk assessment. Reliability Engineering and
System Safety, Vol. 68, pp. 219–254, 2000.
[Meulenbelt, 2012] D. Meulenbelt, Veiligheid van A3 ConOps, Afstudeerscriptie, Hogeschool van
Amsterdam, 12 juli 2012.
[NASA, 2004]
NASA (2004). DAG-TM Concept element 5 en-route free maneuvering for user-
preferred separation assurance and local TFM conformance operational concept
description, AATT Project Milestone 8.503.10, NASA Airspace Systems Program
Office, Washington D.C.
[Nieskens, 2014]
Nieskens, C., Agent-based modelling and Monte Carlo simulation of ground
controlled trajectory-based operations, MSc Thesis, Delft University of Technology,
April 2014.
[Pola et al., 2003]
G. Pola, M.L. Bujorianu, J. Lygeros, M.D. Di Benedetto. Stochastic hybrid models:
an overview with applications to air traffic management. In Proceedings of IFAC
Conf. Analysis and Design of Hybrid Systems (ADHS), Saint-Malo, Brittany, France,
2003.
[SESAR, 2007] SESAR, “Concept of Operation, Sesar definition phase, Task 2.2.2, Milestone 3”,
Report DLT-0612-222-02-00, Version 2.0, October 2007.
[SESAR, 2012] SESAR Concept of Operations Step 2, SESAR-JU Deliverable D65-011, edition
01.00.00, May 2012.
[SESAR-JU, 2013] SESAR-JU, Preliminary OSED_2, Final V2 OSED_1, November 2013
Page 81 of 96
Appendix A. Model Specification Formalism
A.1 Petri Net formalism
For the modelling of accident risk of safety-critical operations in nuclear and chemical industries, the most
advanced approaches use Petri nets as model specification formalism, and stochastic analysis and Monte Carlo simulation to evaluate the specified model, e.g., see [Labeau et al., 2000]. Since their introduction
as a systematic way to specify large discrete event systems that one meets in computer science, Petri
nets have shown their usefulness for many practical applications in different industries, e.g., see [David & Alla, 1994]. Various Petri net extensions and generalisations and numerous supporting computer tools
have been developed, which further increased their modelling opportunities. Nevertheless, literature on Petri nets appeared to fall short for modelling the class of General Stochastic Hybrid Systems (GSHS)
[Bujorianu, 2004] that was needed to model air traffic safety aspects well [Pola et al., 2003].
Cassandras and Lafortune [1999] provide a control systems introduction to Petri nets and a comparison
with other discrete eventmodelling formalisms like automata. Both Petri nets and automata have their
specific advantages. Petri net is more powerful in the development of a model of a complex system, whereas automata are more powerful in supporting analysis. In order to combine the advantages offered
by both approaches, there is need for a systematic way of transforming a Petri net model into an automata model. Such a transformation would allow using Petri nets for the specification and automata
for the analysis. For a timed or stochastic Petri net with a bounded number of tokens and deterministic or
Poisson process firing, such a transformation exists [Cassandras and Lafortune, 1999]. In order to make the Petri net formalism useful in modelling air traffic operations, we need an extension of the Petri net
formalism including a one-to-one transformation to and from GSHS. Everdij and Blom [2003, 2005, 2006, 2010] have developed such extension in the form of (Stochastically and) Dynamically Coloured Petri Net,
or for short (S)DCPN.
Jensen [1992] introduced the idea of attaching to each token in a basic Petri net (i.e., with logic
transitions only), a colour which assumes values from a finite set. Tokens and the attached colours
determine which transitions are enabled. Upon firing by a transition, new tokens and attached colours are produced as a function of the removed tokens and colours. Haas [2002] extended this colour idea to
(stochastically) timed Petri nets where the time period between enabling and firing depends of the input tokens and their attached colours. In [Haas, 2002] and [Jensen, 1992] a colour does not change as long
as the token to which it is attached remains at its place. Everdij and Blom [2003, 2005] defined a
Dynamically Coloured Petri Net (DCPN) by incorporating the following extensions: (1) a colour assumes values from a Euclidean state space, its value evolves as solution of a differential equation and influences
the time period between enabling and firing; (2) the new tokens and attached colours are produced as random functions of the removed tokens and colours. An SDCPN extends an DCPN in the sense that
colours evolve as solutions of a stochastic differential equation [Everdij & Blom, 2006].
This appendix explains how the SDCPN formalism has been used to develop a MC simulation model of the Conservative SESAR2020+ operation. Within the iFly project the same formalism has been used to
develop a MC simulation model of the A3 operation. Similarly as applied with the A3 operation, for the development of a Petri net model of A3G operation, two key challenges have to be addressed: a
syntactical challenge of developing a model that is consistent, complete, and unambiguous; and a semantics challenge of representing the A3G operation sufficiently well. This appendix aims to show the
(S)DCPN formalism that is used to address the syntactical challenge.
A.2 Specification of the developed Petri net model
In using the (S)DCPN formalism [Everdij & Blom, 2003, 2005, 2006] for the modelling of increasingly
more complex multi-agent hybrid systems, it was found that the compositional specification power of Petri nets reaches its limitations. More specifically, the following problems were identified:
Page 82 of 96
1. For the modelling of a complete Petri net for complex systems, a hierarchical approach is
necessary in order to be able to separate local modelling issues from global or interaction modelling issues.
2. Often the addition of an interconnection between two low-level Petri nets leads to a duplication of transitions and arcs in the receiving Petri net.
3. The number of interconnections between the different low level Petri nets tends to grow
quadratically with the size of the Petri net.
Everdij et al. [2006] explained which Petri net model specification approaches from literature solve
problem 1, and developed novel approaches to solve problems 2 and 3. Together, these approaches are integrated into a compositional specification approach for SDCPN, which is explained below.
In order to avoid problem 1, the compositional specification of an SDCPN for a complex process or operation starts with developing a Local Petri Net (LPN) for each agent that exists in the process or
operation (e.g., air traffic controller, pilot, navigation and surveillance equipment). Essential is that these
LPNs are allowed to be connected with other Petri net parts in such a way that the number of tokens residing in an LPN is not influenced by these interconnections. We use two types of interconnections
between nodes and arcs in different LPNs:
Enabling arc (or inhibitor arc)
from one place in one LPN to one transition in another LPN. These types of arcs have
been used widely in Petri net literature.
Interaction Petri Net (IPN)
from one (or more) transition(s) in one LPN to one (or more) transition(s) in another LPN.
In order to avoid problems 2 and 3, high level interconnection arcs have been introduced that allow, with well-defined meanings, arcs to initiate and/or to end on the edge of the box surrounding an LPN [Everdij
et al., 2006]. The meaning of these interconnections from or to an edge of a box allows several arcs or transitions to be represented by only one arc or transition.
A.3 High level interconnection arcs
As an illustration of how high level interconnection arcs avoid duplication of arcs and transitions within an LPN and duplication of arcs between LPNs, we give three examples of these high level interconnection
arcs. See [Everdij et al., 2006] for a complete overview of these high level interconnection arcs.
In the first example, Figure A.1, an enabling arc starts on the edge of an LPN box and ends on a
transition in another LPN box, means that enabling arcs initiate from all places in the first LPN and end on
duplications of this transition in the second LPN. The duplicated transitions should have the same guard or delay function and the same firing function and their input places should have the same colour type.
This high level interconnection arc is not defined for inhibitor or ordinary arcs instead of enabling arcs.
Page 83 of 96
FIGURE A.1: High level enabling arc starts at the edge of an LPN box.
In the second example, Figure A.2, an enabling arc ends on the edge of an LPN box. This means that for
each transition in the receiving LPN a copy of this enabling arc should be in place. Figure A.2 shows an example of this high level interconnection arc. This type of high level arc can also be used with inhibitor
arcs instead of enabling arcs. It cannot be used with ordinary arcs, due to the restriction that the number of tokens in an LPN should remain the same.
In the third example, Figure A.3, an ordinary arc starts on the edge of an LPN box and ends on a
transition inside the same box. This means that ordinary arcs start from all places in the LPN box to duplications of this transition. The duplicated transitions should have the same guard or delay function
and the same firing function and their set of input places should have the same set of colour types. Figure A.3 illustrates how this avoids both the duplication of transitions and arcs within an LPN, and the
duplication of arcs between LPNs.
FIGURE A.2: High level enabling arc ends at the edge of an LPN box.
Page 84 of 96
FIGURE A.3: High level ordinary arc starts on the edge of an LPN box and ends on a transition inside the
same LPN box.
Page 85 of 96
Appendix B. List of A3G parameters and their A3G baseline values
# Agent LPN Parameter Explanation
Baseline
value
1 Aircraft Engine System Fail
Engine Mean duration of Engine Failure
No Engine Failure
1 hr
2 Fail
Enginep Probability of Engine Failure 101*10
3 Aircraft
Emergency mode
down
OES Mean duration of Emergency No
Emergency
1 hr
4 down
OESp Probability of Emergency 101*10
5 Pilot
Flying
Current Goal
goals
PFm Total number of goals PF 7
6 failures
PFm Total Number of failures in case of
‘Emergency actions’ goal for PF
6
7 Goal Memory 3ExMon
PF Mean duration of Execution
Monitoring & Goal Prioritisation
14.7 s
8 3MonGP
PF Mean duration of Monitoring & Goal
Prioritisation End Task
10 s
9 Task Performance
Goal 2:
Emergency
Actions
MD
PF
Mean duration of Monitoring &
Decision Coordination, Duration
parameter of Monitoring & Decision
Execution
10 s
10 Coord
PF Mean duration of Coordination
Monitoring & Decision
5 s
11 ExMon
PF Mean duration of Execution
Monitoring & Goal Prioritisation
20 s
12 MonGP
PF Mean duration of Monitoring & Goal
Prioritisation End Task
10 s
Page 86 of 96
# Agent LPN Parameter Explanation
Baseline
value
13 Pilot flying
(continued)
Task
Performance
Goal 3: Conflict
Resolution (STC
& MTC)
1dT
Mean decision delay time STC
Monitoring & Decision Execution
5.7 s
14 2dT
Mean decision delay time MTC
Monitoring & Decision Execution
30 s
15 Coord
PF Mean duration of Coordination
Monitoring & Decision
∞
16 ExMon
PF Mean duration of Execution
Monitoring & Goal Prioritisation
0
17 3ExMon
PF Mean duration of Execution
Monitoring & Goal Prioritisation
14.7 s
18 3MonGP
PF Mean duration of Monitoring & Goal
Prioritisation End Task
10 s
19 Task
Performance
Goal 4:
Navigation
vertical
TW
PF Duration in Monitoring & Decision 10 s
20 MD
PF Mean duration of Monitoring &
Decision Coordination
∞
21 Coord
PF Mean duration of Coordination
Monitoring & Decision
0 s
22
ExMon
PF
Mean duration of Execution
Monitoring Monitoring & Goal
Prioritisation
20 s
23 MonGP
PF Mean duration of Monitoring &
Goal Prioritisation End Task
10 s
24 Mon
PF Mean duration of Monitoring
Monitoring & Decision
20 s
25 Task
Performance
Goal 5:
Navigation
horizontal
actions
(Back2Goal)
MD
PF Mean duration of Monitoring &
Decision Coordination
∞
26 Coord
PF Mean duration of Coordination
Monitoring & Decision
0 s
27
ExMon
PF
Mean duration of Execution
Monitoring Monitoring & Goal
Prioritisation
20 s
28 MonGP
PF Mean duration of Monitoring & Goal
Prioritisation End Task
10 s
29 Mon
PF Mean duration of Monitoring
Monitoring & Decision
20 s
Page 87 of 96
# Agent LPN Parameter Explanation
Baseline
value
30 Pilot flying
(continued)
Task
PerformancePF
Goal 6: Prepare
Route Change
MD
PF Mean duration of Monitoring &
Decision Coordination
∞
31 Coord
PF Mean duration of Coordination
Monitoring & Decision
0 s
32
ExMon
PF
Mean duration of Execution
Monitoring Monitoring & Goal
Prioritisation
20 s
33 MonGP
PF Mean duration of Monitoring &
Goal Prioritisation End Task
10 s
34
2MD E
PF
Mean duration of
Monitoring & Decision
Execution
10 s
35 Task
PerformancePF
Goal 7:
Miscellaneous
MD
PF Mean duration of Monitoring &
Decision Coordination
∞
36 Coord
PF Mean duration of Coordination
Monitoring & Decision
0 s
37
ExMon
PF
Mean duration of Execution
Monitoring Monitoring & Goal
Prioritisation
20 s
38 MonGP
PF Mean duration of Monitoring &
Goal Prioritisation End Task
10 s
39
2MD E
PF
Mean duration of
Monitoring & Decision
Execution
10 s
40 Task
Performancepf
Mon
PF
Duration parameter of
Monitoring Monitoring &
Decision
20 s
41
TD
PF
Duration parameter of
Monitoring Monitoring & Goal
Prioritization
3 min
Page 88 of 96
# Agent LPN Parameter Explanation
Baseline
value
42 Pilot flying
(continued)
State Situation
AwarenessPF
max FL
PFz SA by PF of maximum FL FL 300
43
min FL
PFz SA by PF of minimum FL FL 70
44 Intent Situation
AwarenessPF
FLISA Intended FL FL 220
45
ClimbVSISA Intended ROC 1500
ft/min
46
ClimbxVSISA Intended ROC expedite 2000
ft/min
47
DescVSISA Intended ROD -2000
ft/min
48
DescxVSISA Intended ROD expedite -3000
ft/min
49 SSAFL SSA minimum FL FL 90
50 Cognitive Mode
goals
PFm Total number of goals of PF 7
51
failures
PFm
Total Number of failures in case
of ‘Emergency actions’ goal for
PF
6
52
,
opp
PF i Mean duration of Oportunistic
mode for F of aircraft i = {1…n}
5 min
53 Pilot not flying Task
PerformancePNF
MD
PNF
Mean duration of
Monitoring & Decision
Coordination, Mean duration of
Monitoring & Decision
Monitoring
5 s
54
Coord
PNF
Mean duration of
Coordination Monitoring &
Decision
2 s
55 ExMon
PNF Mean duration of Execution
Monitoring & Goal Prioritisation
5 s
56 MonGP
PNF Mean duration of Monitoring &
Goal Prioritisation End Task
5 s
57
Mon
PNF
Mean duration of
Monitoring Monitoring &
Decision
5 s
Page 89 of 96
# Agent LPN Parameter Explanation
Baseline
value
59 Environment GNSS system (GPS/
Nav Global) /
Satellites
down
SAT Mean duration of Not Working
Working
½ hr
60 degraded
SAT Mean duration of Degraded
Working
0 s
61 corrupted
SAT Mean duration of Corrupted
Working
½ hr
62 down
SATp Probability of Not working 101*10
63 degraded
SATp Probability of Degraded 0
64 corrupted
SATp Probability of Corrupted 201*10
65 Global ADS-B ether
frequency
occ
ADS Bglobal
Mean duration of Occupied Not
occupied 1 hr
66 occ
ADS Bglobalp
Probability of Occupied
101*10
67 SSR frequency
(1030) ,
occ
SSR FRQ Mean duration of Occupied Not
occupied
0 s
68 ,
occ
SSR FRQp Probability of Occupied
0
69 Global ATC uplink
frequency
down
ATC global Mean duration of Occupied Not
Occupied
1 hr
70 down
ATC globalp Probability of Occupied 101*10
71 Weather magW Horizontal wind magnitude 0 m/s
72 degW Horizontal wind angle 0 deg
Page 90 of 96
# Agent LPN Parameter Explanation
Baseline
value
73 Airborne GNC
systems:
Failure
indicators for
PF
Indicators Failure
mode PF
down
HMI Mean duration of HMI Not Working
Working
0 s
74
down
HMIp Probability of HMI Not Working 0
75 failures
PFm Total number of failures in case of
‘Emergency actions’ goal for PF
6
76 Airborne GNC
systems:
Guidance
Systems
Aircraft Guidance down
GUID Mean duration of Not Working
Working
0 s
77
HMI
downp Probability of Not Working 0
78 Horizontal
Guidance
Configuration
Mode
Vertical Guidance
Configuration
Mode
err
Standard deviation of course error
when LNAV disengaged
0.5 deg
79
Standard deviation on position of
aircraft entering the system, vertical
direction
20 m
80
v
Standard deviation on velocity of
aircraft entering the system, vertical
direction
0.5 m/s
81
3b Noise factor on velocity, vertical
direction
0.1 m/s
82
z
leveld
Boundary Baseline value used to
determine if the aircraft is flying
level or climbing/descending
10 m
83
w
z Standard deviation vertical wind 0 m/s
84
w
z Mean vertical wind 0 m/s
85 Aircraft FMS Intent
Intended
bank Intended bank angle 25 deg
86
Intended
gV Intended groundspeed 250 m/s
87 ANP ANP value 1 Nm
88
Hor
fxCB
Factor for Horizontal Conformance
boundary, i.e., boundary value (in
Nm) is 0.5 Hor
fxANP CB
2x
89
Ver
fxCB
Factor for vertical Conformance
boundary, i.e., boundary value (in
m) is
Ver
fxCB
2x
90 Send
TimeTCP Duration for sending one trajectory
change point (TCP)
3 s
Page 91 of 96
# Agent LPN Parameter Explanation
Baseline
value
91
Send
NumTCP
Number of TCP’s sent belonging to
intent (hence total duration of
sending intent takes
Send Send
Time NumTCP TCP
4
92
Pr io
constd
a/c priority (w.r.t. distance to goal)
is constant within this range (to
avoid continuous switching of
priorities)
10 Nm
93 Airborne GNC
systems: Own
Positioning
Systems
Aircraft GNSS
(GPS) receiver
down
GNSS REC Mean duration of Not Working
Working
500 s
94
down
GNSS RECp Probability of Not Working
101*10
95 Aircraft IRS
down
IRS Mean duration of Not Working
Working
0 s
96
down
IRSp Probability of Not Working 0
97 Aircraft Altimeter
down
Altim Mean duration of Not Working
Working
½ hr
98
down
Altimp Probability of Not Working 101*10
99 Aircraft Horizontal
Position
Processing
IRS
x Standard deviation of horizontal
position error in case of IRS
estimate
0 m
100 1c Covariance of horizontal position
and velocity error in case of IRS
estimate
0 m2
/sec
101
IRS
v Standard deviation of horizontal
velocity error in case of IRS estimate
4 Nm/hr
102
GNSS
x Standard deviation of horizontal
position error in case of GNSS/GPS
working well
20 m
103
GNSS
v Standard deviation of horizontal
velocity error in case of GNSS/GPS
working well
2 m/s
104
,GNSS DC
x Standard deviation of horizontal
position error in case of GNSS/GPS
degraded or corrupted
20 m
Page 92 of 96
# Agent LPN Parameter Explanation
Baseline
value
105
,GNSS DC
v
Standard deviation of horizontal velocity
error in case of GNSS/GPS degraded or
corrupted
10 m/s
106
Aircraft Vertical
Position
Processing
Ver
x Standard deviation of vertical position
error in case of altimeter working well
10 m
107 Ver
v Standard deviation of vertical velocity
error in case of altimeter working well
1 m/s
108
,degrver
x
Standard deviation of vertical position
error in case of altimeter degraded or
corrupted
60 m
109
,degrver
v
Standard deviation of vertical velocity
error in case of altimeter degraded or
corrupted
2 m/s
110 b Noise factor on velocity 0.5 m/s
111 Airborne
GNC:
Communicati
on Systems
ADS-B Transmitter down
ADS TRM
Mean duration of Not Working
Working
½ hr
112
down
ADS TRMp Probability of Not Working
101*10
113 ADS-B Receiver
(not used)
down
ADS REC
Mean duration of Not Working
Working
½ hr
114
down
ADS RECp Probability of Not Working
55*10
115 Regular Broadcast
FMS Intent
IRBT Time interval for regular broadcast of
intent to ground
2 min
116 Regular Broadcast
aircraft state
SRBT Time interval for regular broadcast of
intent to ground
1 s
Page 93 of 96
# Agent LPN Parameter Explanation
Baseline
value
117 ATC
Ground
System
CD &
Management
x
updateT Duration before processing update of state
info
1.5 s
118 I
updateT Duration before processing update of Intent
info
1.5 min
119 STC
predT STC prediction time of potential conflict
3 min
120 MTC
predT MTC prediction time of potential conflict
10 min
121
SODT
Time duration after which Start of Descend
(leaving SSA) will be initiated in case of Nav.
Failure
10 s
122 MTCD
sepATCH Vertical separation used in ATC MTCD
1000 ft
123 STCD
sepATCH Vertical separation used in ATC STCD
900 ft
124 MTCR
resATCR Horizontal resolution distance for ATC
MTCR
5 Nm
125 MTCR
resATCH Vertical resolution distance for ATC MTCR 1000 ft
126 STCR
resATCR Horizontal resolution distance for ATC STCR 3 Nm
127 STCR
resATCH Vertical resolution distance for ATC STCR 900 ft
128 2
max
B Goal Maximum turn angle allowed for flying back
to goal after STCR
90 deg
129 Resolution
Mode
STC
resT
Duration of state-based short term conflict
before ATC ‘switches’ to STC resolution
mode
10 s
130 STC
AlertAgainT
If an STC conflict exist longer than
STC
AlertAgainT , then another alert is generated 30 s
131 MTC
AlertAgainT
If an MTC conflict exist longer than
MTC
AlertAgainT , then another alert is generated 2 min
132
STC
in
If another STC is predicted to occur STC
in
earlier than the existing earliest STC, then
an STC alert is generated
5 s
133
MTC
in
If another MTC is predicted to occur MTC
in
earlier than the existing earliest STC, then
an STC alert is generated
5 s
Page 94 of 96
# Agent LPN Parameter Explanation
Baseline
value
134 ATC
Ground
System
(continued)
ATC Intent
based STCR
advisory
max
res Maximum course change for resolution 60 deg
135
STC
addT
Additional time beyond the Short Term
horizon to avoid the new immediate Short
Term conflicts when doing ST resolution
10 s
136
min
resR
Minimum reduced horizontal separation
value allowed if no horizontal resolution
can be found
100 m
137 CPU
STCRT Time duration to calculate STCR 1 s
138 deg
div Angle used to diverge parallel STCR’s 5 deg
139
div
BoundH
All a/c within div
BoundH height difference are
initially taken into account for divergence
of parallel STCR’s
300 ft
140
divStep
BoundH
Stepwise increase of div
BoundH -value if there
are no a/c within div
BoundH height
difference
100 ft
141 Step
ROTd Step size in course change for finding
short term conflict resolution
0.5 deg
142 ATC Intent
based MTCR
advisory
max
res Maximum course change for resolution 60 deg
143
MTC
addT
Additional time beyond the Medium Term
horizon to avoid the new immediate
Medium Term conflicts when doing MT
resolution
5 min
144 CPU
MTCRT Time duration to calculate MTCR 2 s
145
deg
MTCR Step size in course for finding medium
term conflict resolution
0.5 deg
146
2 max
MTCR
B G Maximum turn angle allowed in ‘back to
goal’ part of resolution
45 deg
147
2 max
MTCR
B Gd Maximum detour distance allowed for
MTCR
15 Nm
148
2
MTCR
B GT Time interval at which a waypoint is
placed to find a path ‘back to goal’
15 s
149
MTCR
Adviz
MTC Resolution ‘starts’ (SOT) at
MTCR
Advizt (to take ATCo response time
and sending duration in account)
20 s
Page 95 of 96
# Agent LPN Parameter Explanation
Baselin
e value
150 ATC
Ground
System
(continued)
ATC State &
Intent all
aircraft
ASAS SI
updateT
Duration before automatic reprocessing
of Info (determine if info has become to
old)
1 min
151 State
dropT Time difference for dropping State info
of other aircraft (i.e. info too old)
10 s
152 Intent
dropT Time difference for dropping Intent info
of other aircraft (i.e. info too old)
6 min
153 ADS BR
ADS-B range (horizontal) ∞ *
154 ATC
Conformance
Monitoring
Intent all
aircraft
CMI
DistTd Time duration bound for horizontal and
vertical distance conformance
2 s
155 CMITd Time duration bound for course
conformance
2 s
156 CMI
VgTd Time duration bound for groundspeed
conformance
2 s
157 CMI
ModeTd Time duration bound for Manoeuvre-
mode conformance
7 s
158 CMI
VTd
Time duration bound for vertical speed
conformance
7 s
159 Course
bound Course conformance bound 5 deg
160 Bound
gV Groundspeed conformance bound
10 m/s
161
,
Bound
LevelV Vertical speed conformance bound when
flying Level
0.1 m/s
162
,
Bound
N LevelV Vertical speed conformance bound when
climbing/descending
2 m/s
163 ATC
Surveillance
(ADS-B ground
receiver )
surv
updateT Duration before ADS-B info update of all
aircraft
1 s
164
occ
Probability that any aircraft is not
received due to ADS-B global occupied or
not
0.5
165 ownx Position of ATC ground system [x,y,z] [0,0,0]
*: It is assumed that SWIM provides an unlimited extension of ADS-B reach without causing any
delay.
Page 96 of 96
# Agent LPN Parameter Explanation
Baseline
value
166 ATC
Ground
System
(continued)
ATC System
Mode
Fail
ATCsys Mean duration of Failure Working 1 hr
167 corr
ATCsys Mean duration of Corrupted Working
1 hr
168 corr
ATCsysp Probability of Corrupted 101*10
169 down
ATCsysp Probability of Not working 101*10
170 Back-to-Goal
evaluation 2B GT
Interval time for Back-to-Goal Evaluation
20 s
171 ADS-B ground
receiver mode
_ ,
down
ATC ADS REC
Mean duration of Not Working
Working ½ hr
172 _ ,
down
ATC ADS RECp
Probability of Not Working 101*10
173 ATC Uplink
Transmitter
Transmit
uplinkT Duration of sending resolution to aircraft
1 s
174 ATCo ATCo-Tactical min
ATCo TT ATCo-T minimum response time 1 s
175
max
ATCo TT
ATCo-T
maximum
response time
1 s
176 ATCo-Planning min
ATCo PT ATCo-P minimum response time 1 s
177 max
ATCo PT ATCo-P maximum response time 1 s