Content uploaded by Marko Radanovic
Author content
All content in this area was uploaded by Marko Radanovic on Oct 03, 2022
Content may be subject to copyright.
pp 2010–2028. © Royal Aeronautical Society 2018
doi:10.1017/aer.2018.106
Adaptive aerial ecosystem
framework to support tactical
conflict resolution
M. Radanovic, M.A. Piera and T. Koca
marko.radanovic@uab.cat
Department of Telecommunications and Systems Engineering
School of Engineering
Autonomous University of Barcelona
Sabadell
Spain
ABSTRACT
To support a seamless transition between safety net layers in air traffic management, this
article examines an extra capacity in the generation of the resolution trajectories,
conditioned by future high dense, complex surrounding air traffic scenarios. The aerial
ecosystem framework consists of a set of aircraft services inside a digitalised airspace
volume, in which amended trajectories could induce a set of safety events such as an
induced collision. Those aircraft services strive to the formation of a cost-efficient airborne
separation management by exploring the preferred resolutions and actively interacting with
each other. This study focuses on the dynamic analysis of a decreasing rate in the number of
available resolutions, as well as the ecosystem deadlock event from the identified
spatiotemporal interdependencies among the ecosystem aircraft at the separation manage-
ment level. A deadlock event is characterised by a time instant at which an induced collision
could emerge as an effect of an ecosystem aircraft trajectory amendment. Through
simulations of two generated ecosystems, extracted from a real traffic scenario, the paper
illustrates the relevant properties inside the structure of the ecosystem interdependencies,
demonstrates and discusses an available time capacity for the resolution process of the aerial
ecosystem.
Keywords: Conflict detection; Aerial ecosystem; Resolution manoeuvres; Deadlock event
Received 13 February 2018; revised 20 June 2018; accepted 3 August 2018
THE AERONAUTICAL JOURNAL DECEMBER 2018 VOLUME 122 NO1258 2010
NOMENCLATURE
3D 3-dimensional
4D 4-dimensional
alt altitude
A/C aircraft
ATC air traffic control
ATFM air traffic flow management
ATM air traffic management
C
A
total number of the conflict manoeuvres
CA collision avoidance
CD conflict detection
CNS communication, navigation and surveillance
CP conflict point
CPA closest point of approach
D diameter of protected cylinder
DDR2 demand data repository 2
DST decision support tool
ECAC European Civil Aviation Conference
EDE ecosystem deadlock event
ET ecosystem time
FL flight level
H height of protected cylinder
I number of conflict subintervals
lat latitude
LAT look-ahead time
long longitude
m type of manoeuvre
M number of manoeuvres
MTCD mid-term conflict detection
N
A
number of aircraft
N
STI
number of spatiotemporal interdependencies
NextGen Next Generation Air Transportation System
PM prediction moment
RBT reference business trajectory
saircraft state position
S
A
total number of the ecosystem resolutions
ST surrounding traffic
STCA short-term conflict alert
STI spatiotemporal interdependency
SESAR Single European Sky ATM Research
SM separation management
SSM standard separation minima
t
CP
starting conflict moment
t
DE
deadlock event instant
t
Ek
ending moment of the kth conflict subinterval
t
PM
conflict prediction instant
t
Sk
starting moment of the kth conflict subinterval
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2011
TCAS traffic-alert and collision avoidance system
TM
A
total manoeuvrability of the ecosystem aircraft
TW time window
vaircraft velocity vector
τchecking time rate
1.0 INTRODUCTION
Trajectory deviations in high dense air traffic delimited volumes causes the separation minima
infringements in many air traffic control (ATC) sectors. This reduction generates a com-
plexity in the assigned traffic and increases the workload of air traffic controllers, especially at
the tactical level
(1)
. As a response, several research and development concepts have been
carried out to advance the Communication, Navigation and Surveillance/Air Traffic Man-
agement (CNS/ATM) technology and meet the principal attributes: safety, capacity and cost-
efficiency of operations
(2,3)
. Based on SESAR (Single European Sky ATM Research) and
NextGen (Next Generation Air Transportation System) initiatives
(4,5)
, it is expected a
replacement of the centralised ATC interventions by a distribution of separation management
(SM) tasks, relying on advanced decision support tools (DSTs). This foresees important
changes in the co-operation, situational awareness and functionalities of the overall ATM
system
(6)
.
When a loss of standard separation minima (SSM) between two aircraft occurs, they are
considered being in a conflict. For en-route traffic, the SSM in the horizontal plane takes
5 nautical miles (SSM
H
=5 NM), while in the vertical plane, it is 1000 feet (SSM
V
=1000 ft).
An important aspect of the tactical conflict detection (CD) algorithms is the prediction
moment (PM), that is, a time instant at which the loss of separation is anticipated. Time
measured from this moment until a moment at which two conflicting aircraft reach their
closest point of approach (CPA) is denoted as the look-ahead time (LAT). CPA is an esti-
mated 4D point at the aircraft trajectory, at which a 3D distance between two aircraft in
conflict reaches its minimum value. Depending on the instant at which a separation minima
infringement is predicted (starting conflict moment, t
CP
), aircraft dynamics and trajectory
geometries, the predicted pairwise encounter can be properly handled by an ATC directive,
such as speed, heading or altitude change. In Ref. 7, it is reported how major changes in the
active aircraft manoeuvrability could potentially induce successive conflicts with neigh-
bouring aircraft and pull them in collision avoidance (CA). CA is the last safety net layer
(8)
,
which is fired because the conflicting aircraft following their trajectories, or performing any
feasible manoeuvre, would not preserve the SSM. In these situations, the aircraft separation
falls for other safety requirements and is delegated to the safety systems on-board aircraft
(9)
.
One of such systems is Traffic alert and Collision Avoidance System (TCAS). TCAS, as an
airborne autonomous system, demonstrates excellent performances for pairwise aircraft
encounters but, unfortunately, suffers from a lack of a performance logic owing to well-
reported induced collisions from the surrounding traffic scenarios
(10)
. Moreover, TCAS
resolution advisories are frequently opposite from the ATC procedures and could create a lack
of integration between the SM layer at the tactical level, and the CA layer at the operational
level
(11)
. Thus, new research concepts relating a coherent integration of the full safety net are
essential.
The aerial ecosystem framework relies on the analysis of spatiotemporal interdependencies
between aircraft located in the surrounding traffic of a pairwise conflict that will lead to a
2012 THE AERONAUTICAL JOURNAL DECEMBER 2018
trajectory amendment. By checking the manoeuvrability impact of any aircraft that could be
affected by a pairwise conflict resolution, it is possible to predict an operationally emergent
behaviour of the surrounding traffic and identify a subset of the trajectory amendments that
will not cause a negative domino effect with neighbouring aircraft. At a technological level,
the proposed ecosystem concept
(12)
relies on multi-agent technology
(13,14)
, in which agents
represent a set of aircraft inside a computed airspace volume, with a trajectory-amendment
decision-making capability, whose trajectories are causally involved in the safety event. Each
time a conflict is detected, an aerial ecosystem is initialised with the aircraft involved in a
pairwise conflict, and it engages all the surrounding aircraft whose trajectory segments could
be affected by a trajectory amendment of a conflict aircraft during the LAT. The set of
spatiotemporal interdependencies (STIs) between ecosystem aircraft is analysed in this paper
to evaluate the extra capacity to support a seamless transition between the SM and CA safety
layers. The STI identification is computed timely in advance by applying the proper opera-
tional metrics at specific time instances, preceding the conflict event. The concept supports
the trajectory-based operations by discretisation of the 4D trajectories and considers an
operational environment in the en-route airspace, above FL245, within a LAT of 300 s.
Figure 1 illustrates an example of the ecosystem creation where A/C1 and A/C2, being in
predicted conflict, identify the surrounding traffic (ST) aircraft, namely A/C3 and A/C4, by
applying certain avoidance manoeuvers, m
2
and m
4
(the manoeuver types are further
explained in Section 3).
The ecosystem services enhance co-operative aircraft interactions and resolution decisions
before the conflict evolves into an ecosystem deadlock event (EDE)
(15,16)
. This event is
characterised by a time instant at which an induced collision could emerge as an effect of a
trajectory amendment. EDE depends on the geometric profiles of the ecosystem trajectories,
the aircraft closure rates and performances. A time frame between the ecosystem creation
moment and the EDE instant is used by the ecosystem members to negotiate their conflict
resolutions. This negotiation could be implemented by the multi-agent ontology frame-
work
(17)
, in which each aircraft is enhanced by a self-governed capability to follow its own
Figure 1. Ecosystem creation.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2013
performance model to identify a preferred resolution. This technology provides right support
in the negotiation interactions, aiming to reach a timely resolutions consensus and avoiding
the ATC intervention before EDE, which does not consider the airspace users’preferences.
Developed ecosystem algorithms provide a robust methodology and the functionalities,
intended as an integrated ATC-supporting tool. The tool is to be operable in the en-route
airspace, at the tactical level, for a relatively short-time separation provision. The main out-
comes for ATC are the computation of compulsory resolutions at the different ecosystem
timestamps, a loose measure of the complexity of the ecosystem and a set of feasible resolutions
that could be shared with the airspace users involved in the ecosystem for ranking preferences.
This article examines an extra capacity in the generation of the resolution trajectories when a
time criticality is threatening the decision-making ability of the ecosystem aircraft. This criti-
cality is expressed by the ecosystem EDE that differently occurs in the traffic scenarios with
different complexities. The complexity of those scenarios is based on the concept of aerial
ecosystems. At a tactical level, an aerial ecosystem presents a set of aircraft, having an
autonomous decision-making capability, that are flying inside certain airspace volume, whose
trajectories are causally involved in the safety event-detected conflict. Those aircraft strive to the
formation of a cost-efficient separation management by exploring the preferred resolutions and
actively interacting with each other. This study focuses on computation of a decreasing rate in
the amount of potential resolutions as well as EDE from the identified STIs among trajectories.
EDE is characterised by a time instant at which at a total number of the ecosystem solutions
takes the zero value. Through simulations of two generated ecosystem, extracted from a real
traffic scenario, the paper illustrates the relevant properties inside the structure of the ecosystem
interdependencies and discusses an available time capacity in the resolutions process of the
ecosystem aircraft. The simulation cases of two ecosystems extracted from a real traffic scenario
have been conducted and analysis of the potential resolutions capacity has shown some
operational aspects, but also the limitations. These limitations will be subject to further research
steps through the implementation of the multi-agent systems ontology, as a significant enabler.
In addition to this introductory section, the paper comprises additional four sections.
Section 2 is dedicated to the problem definition. Section 3 describes the computational
framework for identification of the STIs from the ecosystem creation algorithm and analytical
model for a potential resolution capacity (decreasing rate in the number of potential resolu-
tions) and the EDE. Section 4 discusses the simulation results and comments on the time
capacity for both ecosystems, while the concluding remarks and directions for the follow-up
research are given in Section 5.
2.0 PROBLEM DEFINITION
This section illustrates two key aspects of the conflict resolution analysis in the complex
traffic environments. Their justification requires the new quantitative methods to enhance
present safety nets.
2.1 Conflict interval for the pairwise encounter
The tactical level within air traffic flow management (ATFM) is timely framed between two
ATC thresholds: mid-term conflict detection (MTCD), that activates approximately 15 min
before the closest point of approach (CPA) between two aircraft, and short-term conflict alert
(STCA), that is triggered around 120 s before the CPA
(18)
. After STCA, two aircraft in
conflict could potentially enter a CA layer that is characterised by a non-ATC separation
2014 THE AERONAUTICAL JOURNAL DECEMBER 2018
provision, but an autonomous airborne safety system, such as TCAS
(19)
. Therefore, new
research lines are required towards the development of the collaborative and decentralised
tactical aerial system, on which both human behaviour and automation will be fully aligned.
That envisages an operational integration of the safety procedures in such a way that any pair
of aircraft involved in a conflict, together with surrounding aircraft, behave as a stable
conflict-free air traffic system. Furthermore, the integration should be characterised with the
critical information on the feasible resolution trajectories proposed through the development
of the airborne and ground-based DSTs
(20)
.
Figure 2 describes the conflict process between two aircraft, projected in the horizontal and
vertical plane. The conflict between aircraft A/C1 and A/C2 (Fig. 2(a)) starts when they reach
Figure 2. (Colour online) Conflict process for pairwise encounter; transition from SM to CA.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2015
the waypoints CP
1
and CP
2
, and ends when fly over CP’
1
and CP’
2
, respectively. At CPA
1
and CPA
2
, the aircraft come to the shortest conflict distance. The starting conflict moment is
approximately close to the STCA threshold. In most cases, it is detected after STCA (CP-I),
but it can also occur before (CP-II) when the closure rates are lower and the geometric
configuration of trajectories is more complex. Detection of this moment is essential for
coherent transition from separation management to collision avoidance. The conflict interval
ends up at CP’(Fig. 2(b)). A very frequent case in a geometry of the aircraft encounters is that
a CPA instant presents also the starting conflict moment (Fig. 3). In this case, PM is advanced
300 s before the beginning of the conflict.
Therefore, a proper detection of the conflict interval, for the pairwise encounter, is essential
for the ecosystem conflict management. The starting conflict instant must be a referent
moment from which the EDE instant can be computed, depending on the complexity of the
Figure 3. (Colour online) The conflict interval where the CPA and CP moments overlap.
2016 THE AERONAUTICAL JOURNAL DECEMBER 2018
ecosystem scenario, i.e. the number of aircraft and a geometry of their trajectories (Fig. 4). In
this sense, the ecosystem life time (ET) is defined as time difference between EDE and PM. A
longer LAT provides an extra time for the analysis of all STIs and enhances a co-ordinated set
of conflict resolution manoeuvers; however, it also increases considerably the uncertainty in
the trajectories and the amount of unnecessary ecosystem members. On the other hand, a
reduction of the LAT could drastically affect the safety of planned operations. Thus, a LAT of
300 s allows the use of aircraft state variable information to represent the ecosystem trajec-
tories by segments with a low uncertainty.
2.2 Ecosystem evolution and deadlock event
The ecosystem evolution towards a determined EDE is characterised by a continuously
decreasing rate in the number of potential resolutions that could be applied during the eco-
system life time. Figure 5 illustrates the ecosystem evolution in the vertical plane over three
time-windows, TW1, TW2 and TW3. Each subsequent window is a sub-window of the
previous one. TW3 denotes a CA window whose edges present the EDE moment. Aircraft
reaching this ecosystem window on their Reference Business Trajectories (RBTs) would not
be subject to the ATC separation provision, but the TCAS activation. Therefore, any co-
ordinated (co-operative) manoeuvers of the aircraft that would provide a conflict-free eco-
system resolution, with respect to the SSM, must be performed before entering TW3.
Figure 6 shows a theoretical decreasing rate in a number of the conflict-free solutions,
denoted with S(t), over the LAT. The values for S(t) have been taken as an example to
illustrate a higher drop in the amount of solutions occurring until the TW1, and then follow-
up with lower decreasing rate until the TW2. S(t) is approaching to the value “0”when the
ecosystem enters the TW3. The time threshold for entering TW3 presents a CP instant for a
detected pairwise conflict. In most cases, their order of magnitude is higher, that depends on
the manoeuverability discretisation supported by the technology, the ecosystem size (the
number of involved aircraft) and the STI structure among the trajectory segments.
Figure 4. (Colour online) EDE positioning within the LAT and ET determination.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2017
3.0 COMPUTATIONAL FRAMEWORK
This section describes the procedure for the STI identification and its utilisation on the
analytical computation of the EDE. A short reference to the ecosystem generation has been
provided with an aim at introducing the manoeuverability criteria that could be maintained for
the resolutions generation.
3.1 State-based CD and ecosystem creation
For computation of the starting conflict moment, t
CP
, there has been implemented a Eucledean
state-based CD algorithm. As developed in Ref. 21, the algorithm simplifies the methodology
referring to the case of two aircraft in conflict, A and B. Their states can be described by
positions s
A
and s
B
, and their velocity vectors by v
A
and v
B
, respectively. Projections of the
aircraft A position along axes are denoted with s
Ax
,s
Ay
and s
Az
, while projections of its
velocity vector are marked with v
Ax
,v
Ay
and v
Az
, respectively. Each aircraft is surrounded by
an imaginary volume called the protected zone. It defines a minimum separation distance
between aircraft. The protected zone in a 3D space takes a shape of a flat cylinder with
diameter Dand height H. Therefore, the imaginary cylinder around the aircraft A is defined
Figure 6. (Colour online) Rate of change in the number of resolutions.
Figure 5. (Colour online) Ecosystem evolution towards EDE.
2018 THE AERONAUTICAL JOURNAL DECEMBER 2018
by the set of points (x,y,z) satisfying the conditions:
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðxsAxÞ2+ðysAy Þ2
q<D
2...(1)
zsAz
jj
<H
2...(2)
The minimal separation distance is considered here as SSM, which means:
SSMH=D
2=5NM;and SSMV=H
2=1000 ft. In a general context, the protected zone of the
aircraft A is defined as a set of points, P
A
, that satisfy
PA=xs
Ax
kk
<1
2
...(3)
where ∥.∥denotes a norm vector. In the case of the cylinder of diameter Dand height H, the
norm is defined as
x;y;zðÞ
kk
=maxffiffiffiffiffiffiffiffiffiffiffiffiffi
x2+y2
pD;
jzj
H...(4)
Using a norm expression, it can be defined as a distance between aircraft and, as a result, a
loss of the SSM. The distance between the aircraft A and B is defined as
ΔðA;BÞ=jj sAsBjj ...(5)
A and B are in loss of separation if and only if Δ(A, B) <1. One of the assumptions for the
ecosystem creation is a linearity. At the future time instant: t, the state prediction A(t) from the
current position can be expressed as
sAtðÞ=sA+tvA...(6)
vAðtÞ=vA...(7)
CD is a predicted loss of separation between aircraft A and B within LAT. A and B are in
conflict if there is a predicted instant t
CP
, at which an achieved distance between the aircraft
will be less than 1:
ΔðAðtCPÞ;BðtCP ÞÞ <1 ...(8)
Once the conflicts are detected, duration of the conflict intervals is checked by sam-
pling the trajectories with 1-s rate, and comparing the shortest distances between the
points of two trajectories at each instant with the SSM criteria (SSM
H
and SSM
V
).
When the inter-distance exceeds either SSM
H
or SSM
V
(or both) the conflict interval
ends in the moment t
CP
The ecosystem creation procedure has been elaborated in Refs
12 and 22. This algorithm determines all cluster members as surrounding traffic aircraft
for which the loss of SSM with any of two conflicting aircraft would occur if this
aircraft performs a given avoidance manoeuver at any moment during LAT (Fig. 1).
Considerably, the ecosystem creation is a spatiotemporal category as the applied
manoeuver generates conflict subintervals with the surrounding aircraft. Manoeuver-
ability is applied in both horizontal and vertical planes (Fig. 7) using a certain set of
parametric values to identify those surrounding traffic aircraft that should be considered
ecosystem members:
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2019
∙m
1
: left heading change with a deflection angle of +30°;
∙m
2
: right heading change with a deflection angle of −30°;
∙m
3
: climb at a vertical rate of +1000 ft/min, and a flight path angle of +2°;
∙m
4
: descent at a vertical rate of −1000 ft/min, and a flight path angle of −2°.
3.2 STI identification and EDE computation
The STI identification refers to the computation of the time windows for each ecosystem
aircraft, inside which any potential co-operative or non-co-operative, horizontal or vertical,
manoeuver could produce a loss of the SSM. Those windows are subintervals of LAT and the
total number of conflict manoeuvers within each window is obtained as per defined time rate
(by default, 1 s) along each RBT segment. Figure 8 shows an example of the conflict sub-
interval generated using left heading change. Conflict subinterval no. 1 (CI1) denotes
the period in which A/C1 performing a given manoeuver generates continuous conflict with
A/C3.
The number of STIs (N
STI
) between pairs of aircraft is obtained using four types of
avoidance manoeuvers, explained above (m
1
,m
2
,m
3
and m
4
), and one additional, m
0
: RBT
follow-up. m
0
means that an aircraft decides to continue flying over its RBT in a given
moment. In this study, therefore, five types of manoeuvers are counted for, i.e., M=5.
Figure 8. Conflict subinterval for a single RBT applying a deflection angle of +30°.
Figure 7. Identification of two ST aircraft; (a) A/C3 by left heading manoeuver of A/C1 and (b) A/C4 by climb
amendment of A/C1.
2020 THE AERONAUTICAL JOURNAL DECEMBER 2018
However, further research might introduce more manoeuvers (i.e., holding turns, regulated
speed modifications among others) in the analysis. Each interdependency contains one or
more conflict subintervals, and a total number of the conflict subintervals (I) within one
ecosystem must satisfy the following condition:
I≤NAðNA1Þ
2M2...(9)
N
A
denotes the number of ecosystem members, and M
2
is a derived property that presents the
total number of manoeuvering combinations applied to one pair of aircraft. An example of the
STI structure is presented in Table 1. It consists of the STI identifier, a combination of two
interdependent flight identifiers, manoeuvering combination and conflict subinterval. t
Sk
presents the starting instant of the conflict subinterval kfor a pair of the ecosystem aircraft,
while t
Ek
denotes the ending moment, t
Sk
<t
Ek
,k∈[1, I]. One STI for one aircraft pair might
have more conflict subintervals generated due to different manoeuvering combinations.
Figure 9 illustrates an ecosystem example described in Table 1.
From Fig. 4, it can be expressed LAT and ET
LAT =tCPtPM ...(10)
ET =tDEtPM ...(11)
where t
CP
,t
PM
and t
DE
present timestamps of predicted pairwise conflict, prediction moment
and deadlock, respectively. With an objective to compute t
DE
, in this study, the ecosystem
solutions are treated as three-fold:
Table 1
Example of the STI structure
STI_ID Interdependent
aircraft IDs
Manoeuvering
combination
Conflict
subinterval (s)
STI_1 A/C1–A/C2 m
0
–m
0
t
S1
–t
E1
STI_2 A/C1–A/C3 m
2
–m
2
t
S2
–t
E2
STI_3 A/C2–A/C3 m
1
–m
2
t
S3
–t
E3
Figure 9. (Colour online) Spatiotemporal interdependencies for the given ecosystem with three members.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2021
1. Co-operative mechanism: All aircraft involved in an ecosystem resolution amendment
should initialise the manoeuver at the same time instant. Therefore, the resolution
capacity at a certain timestamp, and in terms of the available time until the system
deadlock, is compressed.
2. The resolution manoeuvers must correspond to the avoidance manoeuvers (Table 1) with
a certain discretisation of the heading and vertical rate. Analysis of the ecosystem
solution with a spectrum of the manoeuvering values is out of scope in this paper. The
computation of the deadlock event would require a more thorough state-space-search
technique.
3. The resolution manoeuvers are considered potential, as some of them might be acceptable
and other unacceptable by the airspace user. The paper only analyses a potential time-
space capacity for resolutions, and the acceptability is not considered since it highly
depends on the airspace users’business models.
Nevertheless, the STI algorithm outputs the ecosystem conflict structure for each inter-
dependency, meaning that an ecosystem solution is not possible at a time instant belonging to
any of the conflict subintervals. In other words, any time window in which no inter-
dependency exists is treated as a potential solution interval. Therefore, it can be concluded
that a total number of ecosystem resolutions from the given time instant t, t∈[t
PM
,t
CP
), to t
DE
equals to a difference between a theoretical number of the ecosystem manoeuvers and a total
number of the conflict manoeuvers, succeeding this instant, provided that the manoeuver-
ability checks are performed with the constant time rate. Therefore, the theoretical number of
the ecosystem manoeuvers is defined as
TMAðtÞ=MNA
τðtCPtÞ...(12)
where τpresents checking time rate (1 s, by default). In addition, for ∀t
Sk
∈[t
PM
,t
CP
),
∀t
Ek
∈(t
PM
,t
CP
], a conservative bound of conflict manoeuvers that cannot be flown due to
STIs with surrounding aircraft is computed as
CAðtÞ=MðNA2Þ
τX
I
k=1
½tEkmaxðtSk ;tÞ ...(13)
and the number of potential ecosystem resolutions
SAðtÞ=TMAðtÞCAðtÞ=MðNA2Þ
τ M2ðtCPtÞ X
I
k=1
½tEkmaxðtSk ;tÞ!...(14)
The maximum number of solutions is obtained in the moment of the ecosystem creation, i.e.
t=t
PM
SAmax =MðNA2Þ
τ M2ðtCPtPMÞ X
I
k=1
½tEktPM !...(15)
S
A
(t) is characterised by a decreasing rate in the time evolution. Finally, the deadlock event
occurs when the number of the ecosystem solutions reaches the 0-value, i.e., S(t)=0:
M2ðtCPtDE Þ X
I
k=1
½tEktDE =0 ...(16)
2022 THE AERONAUTICAL JOURNAL DECEMBER 2018
Expression (16) computes the value of t
DE
that corresponds to TW3, illustrated in Figs 5
and 6.
4.0 ANALYSIS OF SIMULATION RESULTS
This section provides relevant results obtained from simulations of two ecosystems. The
traffic scenario used for this purpose was DDR2,
1
s06.m1 model that comprises 4D flight
plans
(23)
. The analysed traffic was dated on 24/08/2017 within the ECAC (European Civil
Aviation Conference) airspace, with the total number of 36,095 flights during the day. Then, a
traffic scenario was created by extracting of this traffic volume over the time interval 16:00–
19:00, and filtering by altitude, above FL245 (the en-route airspace). The scenario counted for
9,698 flights.
From the traffic simulation, 2,237 pairwise conflicts have been identified, and two of them
have been selected for analysis of the ecosystem creation and resolutions generation. The first
ecosystem consists of three aircraft, while the second one was composed of four. Tables 2 and
3provide the structure of the trajectory segments for the aircraft inside both ecosystems
(for simplicity, instead of the flight identifiers –digits –there has been used an abbreviation
“A/C#”). The abbreviations “lat”,”long”and “alt”present 3D spatial co-ordinates: latitude,
longitude and altitude, respectively. The index “1”denotes the co-ordinates of the first 4D
point, while “2”express the second one. Latitude and longitude are expressed in decimal
degrees, altitude in feets and time in seconds. The time values are given in the accumulated
seconds counted from the beginning of day.
For a better understanding of computation of the number of resolutions, the timestamps are
converted to LAT interval (300-s-periods), taking
∙time-1 =t
PM
=0s–for both ecosystems,
∙time-2 =t
CP
=298.00 s –Ecosystem 1,
∙time-2 =t
CP
=218.49 s –Ecosystem 2.
Figures 10 and 11 (a, b and c) graphically describe the ecosystems in 3D (latitude–
longitude–altitude) and 2D projections (longitude–latitude and time–altitude).
The simulation runs have output the main properties related to the STI structure (Table 4)
while Figure 12 describes the potential resolution capacity for Ecosystems 1 and 2.
Looking at the values in Table 4, it can be observed that Ecosystem 1 (Ecosystem ID
column) with three aircraft (N
A
column) generated three interdependencies (N
STI
column),
while Ecosystem 2 with four aircraft generated four interdependencies. The interdependencies
within Ecosystem 1 produced 19 conflict subintervals in total (Icolumn), while in the case of
Ecosystem 2 there were 33. The maximum number of solutions in the moment of Ecosystem
1 creation is 20,990 (S
Amax
column), and, in the case of Ecosystem 2 creation, this number
goes to 50,291 that initially provides more resolution capacity to Ecosystem 2. However, due
to a significantly higher number of the conflict subintervals and their longer durations,
Ecosystem 2 reaches the deadlock moment faster (t
DE
=149.47 s) comparing to Ecosystem 1
(t
DE
=219.17 s). The values for t
DE
are provided with respect to the ET interval (t
DE
column).
1
Demand Data Repository 2 (DDR2) is an extensive ATM database, developed and maintained by EURO-
CONTROL. It contains a variety of traffic data, such as historical, filtered and forecast traffic datasets, as well the
analytical tools and reporting sections. DDR2 is intended for use by the airspace users, the ATC as well as an
academic research.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2023
As the cumulative timestamps, those values correspond to 60,848.17 and 67,128.47 s for
Ecosystems 1 and 2, respectively.
Figure 12(a) describes the resolutions capacity of Ecosystem 1 over its time, which equals
to LAT, i.e., ET =LAT since t
CP
=300 s. In the case of Ecosystem 2 in Fig. 12(b), the
capacity is measured within a time interval of 218 s (ET =218). Obviously, in the latter case,
Table 3
Ecosystem 2 –trajectory segments
Flight
ID
lat-1
(°)
long-1
(°)
alt-1
(ft)
time-1
(s)
lat-2
(°)
long-2
(°)
alt-2
(ft)
time-2
(s)
A/C1 39.660556 –8.246667 25,000.00 66,994.00 40.107527 –8.318823 29,314.71 67,212.49
A/C2 40.401960 –7.906300 40,869.57 66,994.00 40.032700 –8.271550 30,314.71 67,212.49
A/C3 40.303130 –7.782880 37,000.00 66,994.00 40.022940 –8.265890 29,763.55 67,212.49
A/C4 39.515320 –8.196740 26,909.09 66,994.00 39.82233 –7.800140 30,000.00 67,212.49
Figure 10. (Colour online) Ecosystem 1 in (a) 3D projection (latitude–longitude–time), (b) 2D planar
projection (latitude–longitude) and (c) 2D vertical projection (time–altitude).
Table 2
Ecosystem 1 –trajectory segments
Flight
ID
lat-1
(°)
long-1
(°)
alt-1
(ft)
time-1
(s)
lat-2
(°)
long-2
(°)
alt-2
(ft)
time-2
(s)
A/C1 50.498611 8.411389 25,000.00 60,629.00 49.932222 8.774444 32,000.00 60,927.00
A/C2 50.536087 8.527662 33,000.00 60,629.00 50.012420 8.791093 33,000.00 60,927.00
A/C3 50.119104 9.170007 36,000.00 60,629.00 50.273001 8.236522 36,000.00 60,927.00
2024 THE AERONAUTICAL JOURNAL DECEMBER 2018
the starting conflict point does not overlap with the closest point of approach (300 s after the
ecosystem prediction moment) due to the operational factors, such as a relative geometry of
the ecosystem trajectories and aircraft dynamics (position, velocity and flight mode). The
solutions curve in the first case is decreasing at a lower rate with respect to the distribution
(allocation) of the conflict subintervals produced by the aircraft interdependencies, with the
distinction that, after 170 s, the reaming number of solutions drops at a higher rate. However,
Ecosystem 1 still reaches deadlock after approximately 219 s, which is notably earlier from
t
CP
(80 s before). Regarding Ecosystem 2, the structure of the conflict subintervals is quite
specific. The solutions curve is slightly maintained first 50 s, and then decreases at a quite low
rate until 90 s. Because of the fact that frequency and duration of induced conflict subintervals
is dominant after 100 s, the curve showed a drastically negative trend by a drastic drop in the
capacity until the deadlock moment, that occurred approximately after 150 s. Based on the
results presented in Table 4 and illustrated in Fig. 12, it can be concluded that both eco-
systems faced a relatively shorter time in resolution with respect to the available times. The
t
DE
values for both ecosystems are significantly “shifted back”with respect to the t
CP
values,
as the complexities of evolving trajectories close to these instants are significantly increased.
At those moments, no combination of the co-ordinated manoeuvers would remove the SSM
infringement.
To co-operatively resolve the conflicts before collision avoidance, the aircraft are fre-
quently required to align their velocities and adjust their inter-distances for computation of the
Figure 11. (Colour online) Ecosystem 2 in (a) 3D projection (latitude–longitude–time), (b) 2D planar
projection (latitude–longitude) and (c) 2D vertical projection (time–altitude).
Table 4
STI properties
Ecosystem ID N
A
N
STI
IS
Amax
t
DE
(s)
Ecosystem 1 3 3 19 20,990 219.17
Ecosystem 2 4 4 33 50,291 149.47
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2025
moment for triggering the co-operative resolutions. On the contrary, the ecosystem concept
elaborated in this paper extends the time horizon providing more decision capacity at the SM
level. The aircraft are aware of a potential EDE while flying to the CPA, and given a
possibility to interactively negotiate the solutions, not requiring a priori any adjustment in
velocity or heading and following the trajectories as approved. A time frame between the
ecosystem prediction moment and a moment where ATC issues the compulsory directives is
reserved for the ecosystem members to negotiate the system solutions. As indicated in the
paper (Section 1), this negotiation might be implemented through the multi-agent ontology, in
which each aircraft, as an intelligent agent, is enhanced by a self-governed capability to
follow its own performance model, identify a preferred resolution, and try to impose it to
other members. This framework provides the right support in the negotiation interactions,
aiming to reach a timely resolutions consensus and avoiding the ATC directives before the
EDE, which do not always consider the airspace users’preferences.
A decreasing rate between the available resolution capacity and elapsed time, expressing a
potential path in negotiations among the aircraft, indicates that each missed moment in
reaching an agreement among them reduces the number of remaining conflict-free solutions.
Moreover, identification of a higher number of the causally involved aircraft into enlarged
ecosystem volume provides an opportunity for an efficient modelling of the optimal trajec-
tories, usually with the minimal deviations, and not compromising the separation criteria.
5.0 CONCLUSION AND FOLLOW-UP RESEARCH
This article relies on the previous research on the ecosystem creation algorithm, trying to
identify the potential extra capacity in the search space of the conflict-free resolution man-
oeuvres. The main driver in this creation is the state-based CD in a pairwise aircraft encounter
and its time instants, the prediction moment and starting conflict moment. The computational
framework has presented the baseline in the identification of the STIs, expressing the
structure of the conflict subintervals as a product of the potentially combined manoeuvres.
The model has further included the analytical computation of decreasing rate in the amount of
potential resolutions, as well as the ecosystem deadlock event in which this amount has
reached the zero value. The study has shown a significance in providing the time capacity for
a set of certain manoeuvres, at the operational level, when a severity of the conflict situation
occurs very rapidly. A decreasing rate of the available ecosystem resources and an elapsed
time described a potential path in a thorough analysis of resolution dynamics, meaning that
each missed moment in making a resolution agreement induces less number of the conflict-
Figure 12. (Colour online) Decreasing rate in the number the resolutions of (a) Ecosystem 1 and (b)
Ecosystem 2.
2026 THE AERONAUTICAL JOURNAL DECEMBER 2018
free manoeuvres. The results, obtained through analysis of two simulated ecosystems, illu-
strated the cases of the variable resolution capacity that decreases over time at a different rate.
With an increased ecosystem size and diverse trajectory geometries, the interdependencies
structure becomes larger which produces less resolution capacity and a shorter ecosystem
time. Finally, the ecosystem runs out of capacity at a certain time instant, which shows a time-
critical nature of the ecosystem, where timely-advanced decision provides more flexible and
resilient solution.
Further research is considered as multi-directional: analysis of multi-threat conflicts with
respect to the time to CPA, and improving the computational performances. Moreover, an
improvement in the ecosystem resolutions will focus on a multi-agent technology for simu-
lation of the aircraft interactions during the negotiation intervals. That will provide more
reliability in the solutions search space. Another task will be directed towards the generation
of the resolution segments, based on the concept of performance-based operations. The main
objective will be the computation of the tactical waypoints and definition of modelling
elements that could provide smooth transition from the conflict-free amendments. The main
criteria in the selection of the ecosystem solutions will be rather feasibility than optimality.
Nevertheless, the early resolution agreements shall guarantee the smallest deviations from the
RBT segments.
ACKNOWLEDGEMENTS
This research is partially supported by the H2020 Research and Innovation Programme, the
project: “Adaptive self-Governed aerial Ecosystem by Negotiated Traffic –AGENT”(Grant
agreement no. 699313), and the national Spanish project: “Automated Air Traffic Manage-
ment for Remotely Piloted Aircraft Systems”(Ref. TRA2017-88724-R). Opinions expressed
in the article reflect the authors’views only.
REFERENCES
1. LIU, W. and HWANG, I. Probabilistic trajectory prediction and conflict detection for air traffic
control, J Guidance Control and Dynamics, 2011, 34, (6), pp 1779–1789, https://doi.org/
10.2514/1.53645.
2. COOK, A., BELKOURA, S. and ZANIN, M. ATM performance measurement in Europe, the US and
China, Chinese J Aeronautics, 2017, 30, (2), pp 479–490, https://doi.org/10.1016/j.
cja.2017.01.001.
3. GLUCHSHENKO, O. and FOERSTER, P. Performance based approach to investigate resilience and
robustness of an ATM System, Tenth USA/Europe Air Traffic Management R&D Seminar,p7,
2013.
4. SESAR. SESAR-NextGen state of harmonisation, Integrated Communications, Navigation and
Surveillance Conference, ICNS, May 2014, https://doi.org/10.1109/ICNSurv.2014.6820056.
5. ENEA, G. and PORRETTA, M. A comparison of 4D-trajectory operations envisioned for Nextgen and
SESAR, some preliminary findings, 28th Congress of the International Council of the Aero-
nautical Sciences, pp 1–14, 2012.
6. PRANDINI, M., PIRODDI, L., PUECHMOREL, S. and BRÁZDILOVÁ, S. L. Toward air traffic complexity
assessment in new generation air traffic management systems, IEEE Transactions on Intelligent
Transportation Systems, 2011, 12, (3), pp 809–818, https://doi.org/10.1109/TITS.2011.2113175.
7. TANG, J., PIERA, M. A. and NOSEDAL, J. Analysis of induced traffic alert and collision avoidance
system collisions in unsegregated airspace using a Colored Petri Net model, Simulation, 2015, 91,
(3), pp 233–248, https://doi.org/10.1177/0037549715570357.
RADANOVIC ET AL ADAPTIVE AERIAL ECOSYSTEM FRAMEWORK... 2027
8. KOCHENDERFER, M. J. and CHRYSSANTHACOPOULOS, J. P. A decision-theoretic approach to devel-
oping robust collision avoidance logic, IEEE Conference on Intelligent Transportation Systems,
Proceedings, ITSC, 2010, pp 1837–1842, https://doi.org/10.1109/ITSC.2010.5625063.
9. SHEPERD, R., CASSELL, R., THAPA, R. and LEE, D. A reduced aircraft separation risk assessment
model, AIAA Guidance, Navigation and Control Conference, 1997, pp 1–16, https://doi.org/
10.2514/6.1997-3735.
10. BENNETT, S. The 1st July 2002 Mid-Air Collision over Überlingen, Germany: a holistic analysis.
Risk Management, 2004, 6, (1), pp 31–49, https://doi.org/10.2307/3867933.
11. MURUGAN, S. TCAS functioning and enhancements, Int J Computer Application, 2010, 1, (8), pp.
46–50, https://doi.org/10.5120/184–320.
12. RADANOVIC, M., PIERA, M. A., KOCA, T. and NIETO, F. J. S. Self-reorganized supporting tools for
conflict resolution in high-density airspace volumes, Twelfth USA/Europe Air Traffic Manage-
ment Research and Development Seminar, June 2017.
13. PREMM, M. and KIRN,S.Multiagent System Technologies, September 2015, Vol. 9433, Springer,
Cham, https://doi.org/10.1007/978-3-319-27343-3_6.
14. RAMASAMY, S., SABATINI, R., GARDI, A. and KISTAN, T. Next generation flight management system
for real-time trajectory based operations, Applied Mechanics and Materials, 2014, 629,pp344–349,
https://doi.org/10.4028/www.scientific.net/AMM.629.344.
15. LI, Y. On deadlock-free modular supervisory control of discrete-event systems, IEEE Transac-
tions on Automatic Control, 1997, 42, (12), pp 1705–1708, https://doi.org/10.1109/9.650022.
16. MISRA, J. Distributed discrete-event simulation, ACM Computing Survey, 1986, 18, (1), pp 39–65,
https://doi.org/10.1145/6462.6485.
17. HADZIC, M., WONGTHONGTHAM, P., DILLON, T. and CHANG,E.Ontology-Based Multi-Agent Sys-
tems, 2009, Vol. 219, Springer, Berlin, Heidelberg, https://doi.org/10.1007/978-3-642-01904-3.
18. CHALOULOS, G., ROUSSOS, G. P., LYGEROS, J. and KYRIAKOPOULOS, K. J. Mid and short term conflict
resolution in autonomous aircraft operations, 8th Innov Res Work Exhib Proc, December 2009.
19. TANG, J., PIERA, M. A. and GUASCH, T. Coloured petri net-based traffic collision avoidance system
encounter model for the analysis of potential induced collisions, Transportation Research Part C:
Emerging Technologies, 2016, 67, pp 357–377, https://doi.org/10.1016/j.trc.2016.03.001.
20. VELA, P. A., VELA, A. E. and OGUNMAKIN, G. Topologically based decision support tools for
aircraft routing, AIAA/IEEE Digital Avionics Systems Conference –Proceedings, December
2010, https://doi.org/10.1109/DASC.2010.5655530.
21. DOWEK, G. and MUNOZ, C. Conflict detection and resolution for 1,2,…,Naircraft, 7th AIAA
Aviation Technology, Integration and Operations Conference, September 2007, https://doi.org/
10.2514/6.2007-7737.
22. RADANOVIC, M. and EROLES, M. A. P. Spatially-temporal interdependencies for the aerial eco-
system identification, Procedia Computer Science, 2017, 104, pp 242–249, https://doi.org/
10.1016/j.procs.2017.01.131.
23. WANDELT, S. and SUN, X. Efficient compression of 4D-trajectory data in air traffic management,
IEEE Transactions on Intelligent Transportation Systems, 2015, 16, (2), pp 844–853, https://doi.
org/10.1109/TITS.2014.2345055.
2028 THE AERONAUTICAL JOURNAL DECEMBER 2018