The Influence of Traffic Structure on Airspace Capacity

Abstract

Airspace structure can be used as a procedural mechanism for a priori separation and organization of en-route air traffic. Although many studies have explored novel structuring methods to increase en-route airspace capacity, the relationship between the level of structuring of traffic and airspace capacity is not well established. To better understand the influence of traffic structure on airspace capacity, in this research, four airspace concepts, representing discrete points along the dimension of structure, were compared using large-scale simulation experiments. By subjecting the concepts to multiple traffic demand scenarios, the structure-capacity relationship was inferred from the effect of traffic demand variations on safety, efficiency and stability metrics. These simulations were performed within the context of a future personal aerial transportation system, and considered both nominal and non-nominal conditions. Simulation results suggest that the structuring of traffic must take into account the expected traffic demand pattern to be beneficial in terms of capacity. Furthermore, for the heterogeneous, or uniformly distributed, traffic demand patterns considered in this work, a decentralized layered airspace concept, in which each altitude band limited horizontal travel to within a predefined heading range, led to the best balance of all the metrics considered.

Full text

The Influence of Traffic Structure on
Airspace Capacity
Emmanuel Sunil1, Jacco Hoekstra1, Joost Ellerbroek1, Frank Bussink2,
Andrija Vidosavljevic3, Daniel Delahaye3 and Dennis Nieuwenhuisen4
1Control and Simulation, Faculty of Aerospace Engineering, TU Delft, Delft, The Netherlands
2Cockpit and Flight Operations, National Aerospace Laboratory (NLR), Amsterdam, The Netherlands
3Laboratory of Applied Mathematics, Ecole Nationale de l’Aviation Civile (ENAC), Toulouse, France
4ATM and Airports, National Aerospace Laboratory (NLR), Amsterdam, The Netherlands
Abstract—Airspace structure can be used as a procedural
mechanism for a priori separation and organization of en-route
air traffic. Although many studies have explored novel
structuring methods to increase en-route airspace capacity, the
relationship between the level of structuring of traffic and
airspace capacity is not well established. To better understand
the influence of traffic structure on airspace capacity, in this
research, four airspace concepts, representing discrete points
along the dimension of structure, were compared using large-
scale simulation experiments. By subjecting the concepts to
multiple traffic demand scenarios, the structure-capacity
relationship was inferred from the effect of traffic demand
variations on safety, efficiency and stability metrics. These
simulations were performed within the context of a future
personal aerial transportation system, and considered both
nominal and non-nominal conditions. Simulation results suggest
that the structuring of traffic must take into account the expected
traffic demand pattern to be beneficial in terms of capacity.
Furthermore, for the heterogeneous, or uniformly distributed,
traffic demand patterns considered in this work, a decentralized
layered airspace concept, in which each altitude band limited
horizontal travel to within a predefined heading range, led to the
best balance of all the metrics considered.
Keywords – airspace structure; airspace capacity; en-route
airspace design; air traffic control; air transportation system
performance
I. INTRODUCTION
The current en-route airspace design is centred around
predefined airways, sectors and ground-based Air Traffic
Controllers (ATCo) [1]. Although enhancements to air traffic
systems and procedures have led to incremental capacity
improvements, the current centralized system architecture has
been widely reported to be nearing saturation levels [2]–[4].
To keep pace with the ever growing demand for air
transportation, it is necessary to investigate novel methods of
organizing and structuring traffic to increase en-route airspace
capacity. However, a fundamental relationship between the
level of structuring of traffic and resulting properties, such as
efficiency and safety, is not well established, and different
studies in this field report seemingly contradictory findings.
Free-Flight researchers, for instance, advocate that higher
densities can be achieved through a reduction of traffic flow
constraints and structure [5]–[7], whereas other studies argue
that capacity would benefit more from a further structuring of
airspace [8], [9]. This dichotomy suggests that airspace
structure and capacity are invariably tied together. The
relationship between these two variables, however, is not well
understood, i.e., does more or less structuring lead to higher
capacity? Or, is there a transition point, where a further
increase in capacity will require a switch from one approach
to the other?
To answer these questions, in this work, the impact of
airspace structure on capacity is investigated as part of the
Metropolis project, a research initiative funded through the
Seventh Framework Programme of the European
Commission. To this end, four airspace concepts, ranging
from a decentralized direct routing concept, to a highly
structured tube network using 4D trajectory-based operations,
are compared using large-scale simulation experiments. The
analysis is performed within the context of a futuristic
personal air transportation system, to enable exploration of
extreme traffic densities that would be impossible to achieve
in the current operational context. The goal of the simulations
is not to arrive at precise capacity estimations for the four
concepts, but rather to consider how the level of structuring
affects capacity. Therefore, the concepts are subjected to
multiple traffic scenarios with heterogeneous demand
patterns, and a relative capacity ranking is performed by
measuring how safety, efficiency and stability metrics vary
with traffic demand. By including rogue aircraft that ignore
concept dependent routing requirements in selected simulation
scenarios, the robustness of the concepts to non-nominal
conditions is also investigated in this study.
This paper begins with descriptions of the four airspace
concepts used to empirically relate airspace structure and
capacity in section II. This is followed in section III with the
setup of two simulation experiments used to compare the
concepts. The results of the experiments are presented and
discussed in sections IV and V respectively. Finally the main
conclusions are summarized in section VI.
This research received funding from the Seventh Framework Programme of
the European Commission under grant number 341508 (Metropolis)

II. AIRSPACE CONCEPTS
To empirically study the structure-capacity relationship,
four en-route airspace concepts of increasing structure, named
Full Mix, Layers, Zones and Tubes, have been defined. This
section describes the conceptual design of the four concepts.
A. Full Mix
The Full Mix airspace concept can be most aptly described
as ‘unstructured airspace’. As demand is often unstructured,
the Full Mix concept assumes that any structuring of traffic
decreases overall system efficiency, and that safety is actually
improved by the spreading of traffic over the available
airspace. Therefore aircraft in the Full Mix concept use direct
horizontal routes, as well as optimum altitudes and velocities,
to minimize fuel usage and other related trip costs.
In Full Mix, traffic separation responsibility is
decentralized to each individual aircraft. As no level of
airspace structure is used to separate potentially conflicting
trajectories, safe separation between aircraft is entirely
dependent on self-separation automation, see section III for
more details. Since Full Mix imposes no restrictions to the
path of aircraft, combined heading, speed and altitude conflict
resolution maneuvers are used.
B. Layers
In this concept, the airspace is segmented into vertically
stacked bands, with each altitude band limiting horizontal
travel to within an allowed heading range, similar to the
hemispheric rule. The resulting vertical segmentation of
airspace is expected to improve safety when compared to the
Full Mix concept, by reducing the relative velocities, and
thereby reducing conflict probabilities, between aircraft
cruising at the same altitude. However, this increased safety
comes at the price of efficiency; while direct horizontal routes
are still possible, vertical flight profiles are dictated by the
bearing between origin and destination, and the corresponding
altitude band with the required heading range. Thus, flights
might not be able to cruise at their optimal altitude, resulting
in higher fuel burn. An exception is made for climbing and
descending aircraft; these aircraft are allowed to maintain
heading while climbing or descending to their destination
altitude.
Figure 1 displays a schematic view of the Layers concept
as implemented in this is research. Here, it can be seen that
that each layer corresponds to a heading range of 45o and has
a height of 300 ft. With these dimensions, two complete sets
of layers fit within the airspace volume used to simulate
traffic, see section III for more details on the experiment
volume. As a result, short flights can stay at low altitudes
while longer flights can improve fuel burn by flying at higher
flight levels. This is expected to mitigate the efficiency drop
of predetermined altitudes in this concept.
The Layers concept also makes use of the same self-
separation automation utilized by Full Mix, albeit with
restrictions on the allowed resolution maneuvers. While
combined heading, speed and altitude resolutions are
permitted for climbing and descending traffic, for cruising
aircraft, resolutions are limited to combined heading and
speed maneuvers for cruising aircraft.
C. Zones
Similar to Layers, the Zones concept separates traffic
based on similarity of travel direction. However, in this case, a
horizontal segmentation of airspace is used to separate traffic
along pre-defined trajectories. In this respect, the Zones
concept somewhat resembles the airway based airspace design
used today.
As a personal aerial transportation scenario is used in this
work, the Zones topology takes into account the layout of
urban environments in the design of its structure, see Figure 2.
Here, two major zone types can be seen: circular and radial
zones. Circular zones are used in a similar way to ring roads
in contemporary cities, while the radial zones facilitate travel
towards and away from the city center, and function as
connections between the circular zones. Additionally, both
zone types are defined to be unidirectional to further aid
traffic separation. As there is no vertical segmentation of
airspace in this concept, optimum altitudes are selected based
on the planned flight distance between origin and destination.
Figure 1: Side view of the Layers concept. Two complete layer sets have been
defined within the airspace volume used to simulate traffic.
Figure 2: Top down view of the Zones topology. Given the personal air
transportation scenario used in this work, the Zone concept is designed to
take into account the layout of a modern city.

The Zones concept also uses self-separation to separate
aircraft flying within the same zone, as well as to assist with
the merging of aircraft between circular and radial zones.
Since the zone topology dictates the horizontal path of an
aircraft, heading resolutions are not allowed for this concept.
D. Tubes
As a maximum structuring of airspace, the fourth concept
implements four-dimensional tubes that provide a fixed route
structure in the air. Here, the aim is to increase predictability
of traffic flows by means of pre-planned conflict free routes.
The tube topology used in this study can be thought of as a
graph with nodes and edges, see Figure 3. The nodes of the
graph are connection points for one or more routes. The edges
are the tubes connecting two nodes. Tubes at the same
horizontal level never intersect, except at the nodes, and are
dimensioned to fit exactly one aircraft in the vertical and
horizontal plane. To provide multiple route alternatives, a total
of thirteen tube layers are placed above each other with
decreasing granularity. This way, short flights profit from a
fine grid at the lowest layer, while at the same time, longer
flights benefit from lengthier straight tubes in higher layers.
Aircraft are only allowed to climb/descend through one tube
layer at a time.
Unlike the other concepts, the Tubes concept uses time-
based separation of aircraft to ensure safety within the
network. This mode of separation dictates that when an
aircraft passes a node, it will ‘occupy’ that node for a
prescribed time interval. Within this occupancy interval no
other aircraft is allowed to pass through that node, and new
flights may only pass through a particular node if the
necessary occupancy interval is completely free. To ensure
that separation at the nodes ensures separation within the
tubes as well, all aircraft within the same tube layer are
required to fly at the same velocity. This prescribed speed
increases with the altitude of the layer to match the decreasing
granularity of the tube network. A major advantage of this
method of separation is that it allows the tube network to be
bi-directional, as the occupancy at a node is independent of
travel direction. This simplifies the design, and enables closer
packing of tubes in the topology.
III. EXPERIMENT DESIGN
Two large-scale simulation experiments were conducted to
compare the four airspace concepts in terms of capacity. This
section describes the design of these two experiments.
A. Simulation Development
1) Simulation Platform
The Traffic Manager (TMX) software, developed by the
National Aerospace Laboratory of the Netherlands (NLR),
was used as the simulation platform in this research. TMX has
been extensively validated and has been used for numerous
ATM related simulation studies. For more information on
TMX capabilities, the reader is referred to [10] .
2) Concept Implementation
Aircraft in the Full Mix concept were programmed to use
the direct horizontal route and the most fuel efficient altitude,
as determined by the APMs. Layers also used the direct
horizontal trajectory. However, altitude was selected based on
the bearing to the destination and the matching altitude from a
predefined list. Additionally, total flight distance determined
the choice between the upper and lower layer sets; flights less
than 22 Nmi used the lower layer set, see Figure 1.
For the Zones concept, the A* path planning algorithm
was used to determine the shortest route over a predefined
horizontal topology, while the most fuel efficient altitude was
chosen. Tubes also employed A* to calculate the shortest
path, but in this case, it was also used to examine whether the
selected path was conflict-free. Here, an instantaneous
planning approach was used whereby the occupancy of each
node along a proposed route was checked at traffic desired
departure times. If any node along a proposed route was found
to be occupied by another flight, the corresponding route was
discarded, and the A* algorithm backtracked to evaluate the
next best solution. If no route could be found, a pre-departure
delay was applied in multiples of 10 seconds up to a
maximum of 30 minutes. After this period, the tube network
was considered to be saturated, and that flight was canceled. A
complete description of the A* algorithm can be found in
[11].
3) Self-Separation Automation
The Full Mix, Layers and Zones concepts relied on self-
separation automation for tactical separation, consisting of
separate Conflict Detection (CD), Conflict Resolution (CR)
and Conflict Prevention (CP) modules. CD was performed
through linear extrapolation of aircraft positions over a
prescribed ‘look-ahead’ time. Once conflicts were predicted,
the Modified Voltage Potential (MVP) algorithm is used for
CR in a pair wise fashion, resulting in implicit cooperative
resolution strategies. Finally, the CP algorithm ensures that
aircraft do not turn into conflicts, in an effort to mitigate
conflict chain reactions. Previous research showed that this
three pronged system was highly effective in solving multi-
aircraft conflicts. For more details, please consult [5].
Figure 3: An example tube topology with three layers of decreasing
granularity. The dashed yellow lines are used to indicate the placement of
nodes above each other. Tubes are bi-directional.

Based on initial test runs, a look-ahead time of 60 seconds,
and separation margins of 0.135 Nmi horizontally and 150 ft
vertically, were selected. Also, aircraft were assumed to have
perfect knowledge of the states of neighboring traffic to focus
exclusively on the structure-capacity relationship.
4) Wind
Wind was modeled as a uniform and time-invariant vector
field with random direction and speed. Wind was deliberately
omitted from the simulation's trajectory planning functions to
study the effect of uncertainties, which could cause deviations
from the planned trajectory, on safety. Thus, the wind used in
the simulation has a similar effect to wind prediction errors in
real life operations.
B. Traffic Scenarios
1) Testing Region
Given the personal air transport scenario, a fictional city
was designed to represent the simulation's physical
environment. To create high density traffic scenarios, a small
portion of the city, with an area of 1600 square Nmi, was used
for traffic simulations, see Figure 4. Here it can be seen that
the city is divided into three major districts: city center, inner
ring and outer ring. To define origin and destination points for
traffic, 1600 ‘PAV-ports’ were evenly distributed over the city
in a grid pattern. Although traffic is simulated over the entire
city, data is only logged between 1650 ft and 6500 ft, as the
focus of this research is on en route airspace design.
2) Traffic Demand
Four demand scenarios of increasing density were used to
compare the concepts, and are defined in terms of
instantaneous traffic demand, see Table 1. These scenarios
were created by setting the average nominal trip time to
fifteen minutes, and rely on assumptions for future population
growth and per capita demand for PAVs, see [12] for more
details.
In addition to traffic volume, it is also necessary to
consider urban traffic patterns. To this end, city blocks were
characterized as either commercial or residential, with a
greater proportion of commercial buildings near the city
center, see Figure 4. This distinction made it possible to
simulate morning rush hour as traffic converging towards
commercial areas of a city. Similarly, evening rush hour could
be simulated as traffic diverging from the city center to
suburban residential areas. Therefore, for each traffic volume,
scenarios with converging, diverging and 'mixed' traffic flows
were created. Also, each scenario had a duration of two hours,
consisting of a forty-five minute build-up period, a one hour
logging period, and a fifteen minute wind-down period.
C. Independent Variables
Two separate experiments were performed; the nominal
experiment and the non-nominal experiment.
1) Nominal Experiment
The nominal experiment focused on the impact of airspace
structure on capacity; although traffic was subjected to a
uniform wind field, no other detriments to aircraft motion
were included. For this experiment, four levels of airspace
structure and four traffic demand scenarios represented the
independent conditions. Six repetitions were performed for
each experiment condition (two repetitions for three traffic
demand patterns). Furthermore, the scenarios were simulated
with and without conflict resolution, resulting in a total of 192
nominal runs.
2) Non-Nominal Experiment
This experiment is aimed at comparing the relative
robustness of the concepts to non-nominal situations. For this
purpose, the four airspace concepts were compared for
simulations with 4, 8, 16, and 32 rogue aircraft. These rogue
aircraft were introduced randomly during the logging hour,
and flew haphazardly through the airspace. Nominal aircraft
were solely responsible for resolving conflicts with rogue
aircraft using its self-separation automation, in all concepts.
Although time based separation is used in Tubes, the self-
separation automation described above is used with speed
resolutions to resolve conflicts with rogue aircraft alone. Once
again, 6 repetitions were performed, with and without conflict
resolution, resulting in a total of 192 non-nominal runs.
D. Dependent Variables
Three categories of dependent variables are used to
compare the concepts: safety, efficiency and stability. The
metrics used to access each category are described below.
1) Safety
Safety metrics focus on the ability of an airspace concept
to maintain safe separation between aircraft. Separation
performance is measured in terms of the number of intrusions
and conflicts. Here, intrusions are defined as violations of
minimum separation requirements, while conflicts are defined
as predicted intrusions, i.e., when two (or more) aircraft are
Figure 4: Map of fictional city used as the simulation physical environment.
Simulation data is logged for the airspace volume between 1600-6500ft
Scenario
Low
Medium
High
Ultra
Instantaneous
Traffic Volume
2,625
3,375
4,125
4,875
TABLE 1: TRAFFIC DEMAND SCENARIOS

expected to violate separation requirements within a
predetermined ‘look-ahead’ time (60 seconds in this research).
Intrusions do not imply collisions. Therefore, in addition
to counting the number of intrusions, it is important to
consider the severity of an intrusion. The severity of an
intrusion is dependent on the path of an aircraft through the
protected zone of another, see Figure 5, and is computed using
the following expression:
  
 


(1)
Here, 
 and 
 are the horizontal and vertical intrusions
that are normalized with respect to the corresponding
minimum separation requirements, while  and are the
start and end times of an intrusion. Using the above relation,
the intrusion severity for the intrusion path shown in Figure 5
is equal to the normalized horizontal intrusion at point ‘A’.
2) Efficiency
The efficiency of the concepts is analyzed using the work
done metric. This metric considers the optimality of an
aircraft's trajectory, and therefore has a strong correlation with
fuel/energy consumption. For each flight, the work done is
computed as:
   

(2)
Here,  and  are the thrust and displacement vectors.
3) Stability
Resolving conflicts may cause new conflicts at very high
traffic densities due to the scarcity of airspace. The stability of
the airspace as a direct result of conflict resolution maneuvers
has been measured in literature using the Domino Effect
Parameter (DEP) [6]. The DEP can be visualized through the
Venn diagram pictured in Figure 6. Here S1 is the set of all
conflicts without resolutions, and S2 is the set of all conflicts
with resolutions, for identical scenarios. Furthermore, three
regions can be identified in Figure 6 from the union and
relative complements of the two sets, with  ,
  and  .
By comparing R3 with R1, the proportion of additional
‘destabilizing’ conflicts that were triggered by resolution
maneuvers can be determined. Thus, the DEP is inversely
proportional to airspace stability, and is defined as [6]:
   
 
 
(3)
IV. RESULTS
The results of the nominal and non-nominal experiments
are presented separately in this section. The effect of the
independent variables on the dependent variables are analyzed
using error bar charts that displays the mean, and the 95%
confidence interval of the mean, for each simulation
condition. Whenever relevant, the effect of CR is also
discussed using separate error bar charts.
A. Nominal Experiment
Over six million flights were simulated during this
experiment. Data from approximately 50% of these flights,
which flew during the logging period, are analyzed for these
results. The analysis begins by considering the traffic volumes
and densities simulated, and the consequent implications on
the analysis of the dependent measures.
1) Traffic Volume and Density
The total traffic volume and average traffic density per
simulation run is displayed are Figures 7 and 8, respectively.
For Full Mix, Layers and Zones, traffic volumes and densities
were fairly similar. However, in both cases, the Tubes concept
deviates significantly from the other concepts. In terms of
traffic volume, Tubes simulated significantly fewer aircraft for
all demand scenarios. This is because Tubes delayed and
cancelled flights if conflict free routes were not available at
scenario specified departure times. Despite the lower traffic
volume, Tubes caused the highest traffic densities. This
inconsistent trend is due to the significantly longer routes of
the Tubes concept (see efficiency metrics), which in turn
increased flight durations and traffic densities.
These differences in traffic volumes and densities for
Tubes need to be taken into account when considering the
other dependent variables. Although Figure 7 suggests that the
Tubes concept has a lower airspace capacity relative to the
other concepts, it should be noted that the figure does not
imply that the other concepts are able to, for instance,
facilitate the higher volumes safely. Therefore, conclusions
with respect to capacity also depend on the other dependent
variables discussed below, and cannot be based purely on the
amount of traffic simulated. Moreover, whenever appropriate,
these metrics are computed relative to the number of flights
simulated to allow for a fair comparison between concepts.
Figure 5: Side view of an intrusion. The red dashed line shows the intrusion
path of an aircraft through the protected zone of another.
Figure 6: The Domino Effect Parameter (DEP) relates the additional conflicts
caused by resolution maneuvers to airspace stability

2) Safety
The number of conflicts and intrusions per flight for all
simulation conditions are displayed in Figures 9 and 10,
respectively. As expected, the number of conflicts and
intrusions increased with traffic demand for all concepts.
Furthermore, the figures also show that the more structured
Zones and Tubes concepts led to significantly higher numbers
of conflicts and intrusions compared to the less structured Full
Mix and Layers concepts.
The effect of tactical CR on the number of safety incidents
is also pictured in Figures 9 and 10. As Tubes did not use
tactical CR, there were no differences between the ON and
OFF conditions. For the other three concepts, the number of
intrusions was considerably reduced with CR ON. However,
the effect of CR on the number of conflicts did not follow the
same trend. For Full Mix and Zones, the number of conflicts
increased with CR ON. This was expected, as resolution
maneuvers increase flight distances and the consequent
probability of encountering other aircraft. However, for the
Layers concept, the opposite was found, with CR ON leading
to a lower number of conflicts. This unusual result is further
analyzed using stability metrics.
It is interesting to note that the Tubes concept, which
aimed at deconflicting flights prior to take-off, resulted in a
very high number of conflicts and intrusions for all scenarios.
This is because the trajectory planning functions used by
Tubes did not take uncertainties, such as wind, into account.
These uncertainties caused aircraft to deviate from their
planned flight paths during the simulation, resulting in a large
number of conflicts due to the tight packing of the Tubes
topology. As no tactical CR was used by Tubes, the conflicts
also resulted in a large number of intrusions.
Figure 11 shows that intrusion severity is not significantly
dependent on traffic demand, and is fairly similar for the three
concepts using tactical CR. This suggests that intrusion
severity is more a function of the selected CR algorithm than
airspace structure. Due to the resolution maneuvers initiated
by the MVP algorithm, intrusion severity was reduced when
CR was enabled for Full Mix, Layers and Zones.
3) Efficiency
Efficiency, measured using the work done metric, is
shown in Figure 12. Here, a positive correlation between work
done and the degree of airspace structure, as well as between
work done and traffic demand, can be seen. The Full Mix
concept led to the lowest work done, closely followed by the
Layers concept. Conversely, the Tubes concept led to the
highest work done, implying that aircraft flew significantly
longer distances in this concept. As conflict resolution
maneuvers increase flight distances, work done was increased
with CR ON (not shown).
4) Stability
The stability of the airspace is analyzed using the DEP, see
Figure 13. A negative DEP implies a net stabilizing effect of
tactical CR whereby conflict chain reactions are outweighed
by those that are solved without pushing aircraft into
secondary conflicts, whereas positive values indicate a large
number of conflict chain reactions, and thus airspace
instability. Figure 13 shows that DEP is consistently zero for
Tubes as it did not use tactical CR. For the other three
concepts, the DEP for the Low demand scenario is similar and
negative. However at higher demand levels, the DEP increases
to positive values for the Full Mix and Zones concepts. This
suggests that the maneuvering room available to solve
conflicts decreases rapidly with increasing airspace density for
these two concepts, making it progressively more difficult to
Figure 7: Total number of flights per run
Figure 8: Traffic density per simulation run
a) Effect of traffic demand
(with conflict resolution)
b) Effect of Conflict Resolution (CR)
Figure 9: Number of conflicts per flight
b) Effect of Conflict Resolution (CR)
Figure 10: Number of intrusions per flight
a) Effect of traffic demand
(with conflict resolution)

avoid intrusions without triggering additional conflicts. This is
particularly true for the Zones concept which experienced a
very large DEP increase between the High and Ultra demand
scenarios.
Although the DEP also increased with demand for Layers,
it remained negative for the range of densities considered in
this work. Thus the Layers concept is more able to prevent
conflict propagation from occurring, and is better at assisting
the MVP CR algorithm in solving the conflicts that do occur,
by reducing conflict angles and relative velocities between
aircraft cruising at the same altitude. This result explains the
reduction of the number of conflicts noted earlier for Layers
with CR ON.
B. Non-Nominal Experiment
As stated earlier, the purpose of the non-nominal
experiment is to compare the relative robustness of the
different airspace structuring methods when subjected to
increasing numbers of rogue aircraft. Since rogue aircraft
primarily affect safety metrics, the following paragraphs
discuss the properties of conflicts and intrusions between
rogue and 2.7 million normal aircraft. Thus only incidents
between rogue and normal aircraft are considered.
Figures 14 and 15 display the number of conflicts and
intrusions per flight with rogue aircraft alone. Here, it can be
seen that increasing the number of rogue aircraft increases the
number of conflicts and intrusions for all concepts. The
figures also show that the Tubes concept is considerably more
affected by rouge aircraft than the other three concepts.
As the trajectories of rogue aircraft were not known in
advance, aircraft in the Tubes concept used the MVP CR
algorithm to avoid conflicts with rogue aircraft alone. Since
Tubes specified the horizontal and vertical flight profiles,
speed resolutions were used. Figure 14b shows that these
resolutions did reduce the number of conflicts with rogue
aircraft for Tubes, as for the other concepts. Similarly, Figure
15b shows the number of intrusions for all concepts (including
Tubes) improved significantly with CR ON. Finally, it is
noted that intrusion severity was unaffected by the number of
rogue aircraft for all concepts, although it did decrease with
CR ON (not shown).
b) Effect of Conflict Resolution (CR)
a) Effect of traffic demand
(with conflict resolution)
Figure 11: Intrusion severity
a) Effect of traffic demand
(with conflict resolution)
b) Effect of Conflict Resolution (CR)
Figure 12: Work done per flight
(with conflict resolution)
Figure 13: Domino Effect Parameter
Figure 14: Number of conflicts per flight
with rogue aircraft
Figure 15: Number of intrusions per flight
with rogue aircraft (with conflict resolution)
Figure 7: Total number of flights per
simulation run
a) Effect of traffic demand
(with conflict resolution)
b) Effect of Conflict Resolution (CR)

V. DISCUSSION
In this work, four concepts of increasing structure, named
Full Mix, Layers, Zones and Tubes, were compared using fast
time simulations to study the influence of traffic structure on
capacity and robustness.
In contrast to previous research, which focused on either
fully structured or fully unstructured concepts, the current
results clearly indicate that the moderately structured Layers
concept led to the best overall performance. Although
unexpected, this result can be explained by the heterogeneous,
or uniformly distributed, traffic demand scenarios used in this
work. For such demand patterns, strict structuring of airspace,
as for Zones and Tubes, increased flight distances and caused
traffic concentrations at intersection points of the predefined
topologies. On the other hand, the vertical structuring used by
the Layers concept separated traffic with significantly
different headings, without constraining the horizontal path of
aircraft. This improved safety and stability by reducing
relative velocities, compared to the unstructured Full Mix
concept, without unduly affecting efficiency metrics.
Therefore, it can be concluded that the optimum level of
structuring is dependent on the traffic demand pattern, and for
heterogeneous demand scenarios, a moderate degree of
structure, as exemplified by the Layers concept, results in the
highest capacity.
For the range of densities considered, the results also show
that a switch between structuring methods is not required to
maximize capacity. In fact, the results indicate the opposite,
with a clear distinction between the two less structured and the
two more structured concepts; while performance degraded
with increasing demand for all concepts, it did so at a higher
rate for Zones and Tubes. Furthermore, the results of the non-
nominal experiment showed that the rigid topology and
preplanned routes used by the Tubes concept reduced its
resilience to the haphazard motion of rogue aircraft, while the
flexible structuring of Full Mix and Layers revealed higher
robustness to non-nominal events.
The poor performance of the Tubes concept stands in
contrast with the positive results of structuring traffic using
pre-defined trajectories found in literature. However, those
‘TBO’ concepts generally used globally optimum trajectories,
based on current airspace status and expected future demand.
The Tubes concept, on the other hand, used an instantaneous
planning approach that selected the shortest available route at
the time of departure, to meet the high flexibility of operation
needed to realize a future personal aerial transportation
system. Regardless, the results of the current study show that
pre-planned trajectories, which are common to both TBO and
Tubes, are negatively affected by uncertainties. In the case of
Tubes, these uncertainties, such as those caused by wind,
made it difficult for aircraft to follow RTAs at waypoints
along a planned route, resulting in a large number of
unintended conflicts and intrusions.
VI. CONCLUSIONS
The results of the simulation experiments suggests that the
structuring of traffic must take into account the expected
traffic demand pattern to be beneficial in terms of capacity.
For the heterogeneous demand patterns used here, a
segmentation into altitude bands with similar headings, as for
Layers, showed safety and stability benefits when compared
to the unstructured Full Mix concept, while the strict
structuring and predefined routing of the Zones and Tubes
concepts only reduced performance. For the traffic densities
considered, no reversal can be observed for this trend.
As a large number of conflicts and intrusions were found
for all concepts, it is recommended to investigate novel
conflict detection and resolution algorithms that cope with the
limited maneuvering room available at extreme traffic
densities. It is also recommended to further investigate the
effects of parameters of the Layers concept, such as heading
range per altitude band, on capacity.
REFERENCES
[1] N. A. Doble, R. Hoffman, P. U. Lee, J. Mercer, B. Gore, N. Smith,
and K. Lee, “Current airspace configuration practices and their
implications for future airspace concepts,” in 8th Aviation Technology
Integration Operations Seminar, AIAA, Anchorage, AK, 2008.
[2] Magill, S., “Effect of Direct Routing on ATC Capacity,” in
Proceedings of the 2nd USA/Europe Air Traffic Management R&D
Seminar, Orlando, 1998.
[3] SESAR Consortium, “The Concept of Operations at a glance,” Single
European Sky, 2007.
[4] Joint Planning and Development Office, “Concept of operations for
the Next Generation Air Transportation System,” FAA, Jun. 2007.
[5] J. M. Hoekstra, R. N. H. W. van Gent, and R. C. J. Ruigrok,
“Designing for safety: the ‘free flight’ air traffic management
concept,” Reliab. Eng. Syst. Saf., vol. 75, no. 2, pp. 215–232, Feb.
2002.
[6] J. Krozel, M. Peters, K. D. Bilimoria, C. Lee, and J. S. . Mitchell,
“System performance characterisitcs of centralized and decentralized
air traffic separation strategies,” in 4th USA/Europe Air Traffic
Management R&D Seminar, Santa Fe, 2001.
[7] M. Ballin, J. Hoekstra, D. Wing, and G. Lohr, “NASA Langley and
NLR Research of Distributed Air/Ground Traffic Management,” in
AIAA’s Aircraft Technology, Integration, and Operations (ATIO)
2002 Technical Forum, 2002.
[8] J. Klooster, S. Torres, D. Earman, M. Castillo-Effen, R. Subbu, L.
Kammer, D. Chan, and T. Tomlinson, “Trajectory synchronization
and negotiation in Trajectory Based Operations,” in Digital Avionics
Systems Conference (DASC), 2010.
[9] T. Prevot, V. Battiste, E. Palmer, and S. Shelden, “Air traffic concept
utilizing 4D trajectories and airborne separation assistance,” in
Proceedings of the AIAA Guidance, Navigation, and Control
Conference, AIAA-2003-5770, Austin, TX, USA, 2003.
[10] F. Bussink, J. Hoekstra, and B. Heesbeen, “Traffic manager: a flexible
desktop simulation tool enabling future ATM research,” in Digital
Avionics Systems Conference, 2005
[11] J. M. Hoekstra, “Designing for safety: the free flight air traffic
management concept,” PhD Dissertation, Delft University of
Technology, Delft, 2001.
[12] E. Sunil, J. Hoekstra, J. Ellerbroek, F. Bussink, D. Nieuwenhuisen, A.
Vidosavljevic, and S. Kern, “Metropolis: Relating Airspace Structure
and Capacity for Extreme Traffic Densities,” in USA/Europe Air
Traffic Management R&D Seminar, Lisbon, 2015.