This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant number n o 341508 (Metropolis) METROPOLIS – Urban Airspace Design Work Package 3: Development & Metrics Definition (D3.2) Document author(s) D. Delahaye (ENAC), A. Vidosavljevic (ENAC), E. Sunil (TUD), J. Hoekstra (TUD), J. Ellerbroek (TUD), and Roalt Aalmoes (NLR) Responsible Partner ENAC Dissemination Level PU Public PP Restricted to other programme participants (including the Commission Services) RE Restricted to a group specified by the consortium (including the Commission Services) CO Confidential, only for members of the consortium (including the Commission Services) X
49
Embed
METROPOLIS – Urban Airspace Designhomepage.tudelft.nl/7p97s/Metropolis/downloads/... · 2015. 5. 26. · (Metropolis) METROPOLIS – Urban Airspace Design Work Package 3: Development
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
This project has received funding from the
European Union’s Seventh Framework Programme
for research, technological development and
demonstration under grant number no 341508
(Metropolis)
METROPOLIS – Urban Airspace Design
Work Package 3: Development & Metrics Definition (D3.2)
Document author(s) D. Delahaye (ENAC), A. Vidosavljevic (ENAC), E. Sunil (TUD), J. Hoekstra (TUD), J. Ellerbroek (TUD), and Roalt Aalmoes (NLR)
Responsible Partner ENAC
Dissemination Level
PU Public
PP Restricted to other programme participants (including the Commission Services)
RE Restricted to a group specified by the consortium (including the Commission Services)
CO Confidential, only for members of the consortium (including the Commission Services) X
Metropolis
2
The page intentionally left blank
Metropolis
3
Document information table
Contract number: ACP3-GA-2013-341508
Project title: METROPOLIS
Project Co-ordinator: Delft University of Technology
This document contains the results of work package 3 Development & Metrics Definition. It provides
review of the metrics that are used as performance indicator for evaluation of concepts for urban
airspace design of Metropolis.
Such urban environment is characterised by high density and variety of air vehicles that requires a
structure that differs significantly from what happens today in order to accommodate traffic in a
safe and efficient manner. To have better understanding of alternatives, four extreme concepts that
has been design in WP2 are evaluated under different scenarios of Metropolis growth (WP1). To
discover the limits of concepts and to compare them in term of effectiveness of the delivered
services, a set of performance indicators (metrics) has been defined.
This document is based on a review of literature of existing and research of new possible metrics
adapted to high traffic volume in urban airspace. There are many metrics available, which might be
applicable for future urban air traffic. Extensive research of such metrics has been performed. Some
metrics are refined/combined to fit the requirements and some new are investigated.
The following sub-categories of performance metrics have been identified:
• Organization (complexity) metrics,
• Efficiency metrics,
• Environmental metrics,
• Safety metrics,
• etc.
In WP3 the metrics that are relevant to determine the quality of the different urban airspace
concepts are defined and elaborated. Societal demand as output of the WP1 and concept definition
as output of the WP2, are used as reference for metric definition. The metrics are described in a
common way, which captures all strong and weak points of the individual concepts and also enables
direct concept comparison. The metrics will be implemented in the batch simulations (WP 4).
Metropolis
10
2 Critical feature of transportation system
The focus of Metropolis research project is to investigate future urban transport organization on the
city 50+ years into the future. Future urban environment is characterised by high density and variety
of air vehicles that requires a structure that differs significantly from what happens today in order to
accommodate traffic in a safe and efficient manner. To have better understanding of alternatives,
four extreme concepts that have been design in WP2 [1] (Figure 2.1) are evaluated under different
scenarios of Metropolis growth (WP1) [2].
Proposed concepts of urban airspace design differ in the terms of structure and control involved. To
discover the limits of concepts and to compare them in term of effectiveness of the delivered
services, a set of performance indicators (metrics) has been defined.
In this chapter, performance review process of the current ATM system will be discussed followed by
the different approaches to measure performance of the future urban transportation.
Metropolis
11
Full mix
In this design, all vehicles share the same airspace, without any structure or non-physical constraints, in which via a prescribed airborne separation assurance algorithm, supported by automation, the vehicles avoid each other while flying an optimal route. UAVs and PAVs are mixed. This is a static airspace design, which does not require adjustments based on demand.
Layers
In this design, every altitude band corresponds to a heading range in a repeating pattern. The aim is to allow maximum freedom of routing while lowering the relative speeds, facilitating the separation and increasing the safety. A limit to the ceiling of UAVs will be an option on this design. This is a static airspace design, which does not require adjustments based on demand.
Zones
Based on the principle of airspace design today, different zones for different types of vehicles, speed ranges as well as global directions have been defined to aid the separation by the structure of the airspace. UAVs and PAVs each have their own zones and are mostly, if not completely, separated. A dynamic adjustment of zones based on demand or observed densities, is an option with this design.
Tubes As a maximum of structuring of airspace, tubes have been defined to provide a fixed, but dense, route structure. Different directions, speeds and vehicle types will use different tubes ensuring safety by separating potentially conflicting traffic. UAVs and PAVs each have their own tubes and are completely separated. A dynamic adjustment of zones based on demand or observed densities, is an option with this design.
Figure 2.1: Airspace Concept Designs [1]
Metropolis
12
2.1 Background
In order to ensure the effective management of the European Air traffic management system,
Performance Review Commission (PRC) has been established in 1998 by the EUROCONTROL
Member States. PRC with support of Performance Review Unit (PRU) every year produce
Performance Review Report (PRR) as an assessment of Air Traffic Management in the Europe during
given calendar year. PRR evaluate performance and assesses to what extent agreed target are met
including recommendations for the improvement of the performance in the future.
In the 2000, based on the research and analytical work of PRC [3], International Civil Aviation
Organization (ICAO) realized the need to create a performance framework for the purpose of
enhancing safety and efficiency in the air navigation system. Following performance areas have been
identified: safety and security, capacity, predictability, flexibility, efficiency, access and equity, cost
effectiveness, global Interoperability, environment. Figure 2.2 shows some of the currently
monitored performance areas.
Figure 2.2: Key performance areas [3]
Safety. In the context of aviation, safety is defined as “the state in which the possibility of harm to
persons or of property damage is reduced to, and maintained at or below, an acceptable level
through a continuing process of hazard identification and safety risk management. [4]”
Aviation is very complex system involving many different actors, technologies and procedures.
Humans and human-built systems are prone to the errors and no system can be absolutely safe.
Therefore safety must be continuously monitored and assess in order to managed and mitigate
identified risks. Today, safety is measured by the number of accidents, serious incidents and
incidents categorised by the cause and contributing factors.
Capacity. The objective of the ATS is to “provide sufficient capacity to accommodate the demand in
typical busy hour periods without imposing significant operational, economic or environmental
penalties under normal circumstances. [5]” To effectively determine future capacity requirements, it
is necessary to monitor current capacity performance. Key indicator of capacity performance is
average ATFM delay per flight, that is calculated as ratio between the total ATFM delay and the
number of flights in a defined area over a defined period of time [3].
Metropolis
13
Efficiency. It addresses the operational and economic cost-effectiveness of a flight from a single-
flight perspective. Therefore it may be represented by the ratio of the cost of ideal flight to the cost
of procedurally constrained flight or as measure of deviation from ideal route (user preferred route).
Today horizontal route inefficiency is measured as a ratio between lengths of the flight trajectory to
the corresponding Great Circle distance. At airport level, inbound flight inefficiency is measured as
additional delay in ASMA1 area due to airborne holding, metering and sequencing of arrival flights;
and pre-departure delay, additional taxi-out time as a measure of outbound flight inefficiency.
Environment. Future ATS has to be environmentally sustainable. Increase of the system capacity in
respond to future growth, has to go along with corresponding increase of efficiency, flexibility and
predictability to ensure that there is no adverse impacts on the environment. Today environmental
impact is measured mainly through CO2 emission.
Cost-effectiveness. Every improvement in the service quality or performance of the ATM involves
additional costs to airspace users. To be cost-effective this costs need to be balanced with interests
of the community. It is measured as a number of controlled flights per Air traffic control officer’s
work hour (ATCo-hour productivity), en-route ANS costs per service unit, etc.
2.2 Categories of performance metrics
Future urban transport, as today air transport system, has to ensure safe, efficient and expeditious
movement of air vehicles. Due to high risk involved in urban transport, safety is a main issue.
However, from the user perspective, flight efficiency is the most important as every deviation from
user preferred routes will involve additional costs to the users. The Air Traffic Management (ATM)
community has been recently faced with a new goal: reducing the impact of ATM on the
environment. Having in mind that the number of vehicles in the future urban transport will be
significantly higher than today’s volume of air transport, this issue will be more critical.
In order to compare proposed concepts of the future urban airspace design, following categories of
performance metrics have been identified.
Organizational metrics. This category of performance metrics aims to identify how the structure
involved in the concept of urban airspace design influence the complexity of the traffic situation.
Comparing complexity of traffic situations, we can implicitly compare how difficult is to control a
given situation which is crucial in the safety-critical system. Measuring the robustness, that is
strongly related with traffic complexity, will allows us to determine how much system is invariant to
changes in the initial conditions and also external influences. .
Operational metrics. From the airspace users and passenger perspective the most important is the
impact of concept on the operational performance of the flights. In the future urban transport
system, most of the people will drive their private air vehicles and safety, punctuality, efficiency, and
the cost of the flight would be equally important for them. Safety metrics focus on the ability of an
airspace concept to maintain safe separation between aircraft. Efficiency is measure of deviation
from the user preferred route in space and time.
1 ASMA (Arrival Sequencing and Metering Area) is the airspace within a radius of 40NM around an airport. [3]
Metropolis
14
Environmental metrics. As already mentioned, impact of aircraft on the environment becomes
increasingly important and will be the main restricting factor for the future growth of the airline
industry. The main environmental issues related to traditional aircraft, that we have today, such as
emissions/pollution, noise pollution, and third party risk will remain equally or more important for
the future urban transport. The new concept of future urban transport, new vehicles, high traffic in
city areas will raise a new issues such as shadow flickering (similar to wind turbines) or light pollution
in case of nightly operations, privacy concerns, distraction, and effects on the biotope.
2.3 Concept evaluation process
Although the identified metrics could be used to evaluate future urban transportation system, the
main goal of Metropolis project is to compare different concepts of urban airspace design in term of
effectiveness of delivered services. Therefore, the analysis will focus on comparing and illustrating
how good or bad given concepts performs regarding each metric; and not comparing metrics results
against given targets. Quality of metrics results, as well as, form of the results (qualitative or
quantitative form) may influence the use of given metrics in concept evaluation. Furthermore,
different concepts may score the same way for a given metrics, making it irrelevant for final
evaluation of the concepts, as they have little influence on the specific metrics.
The different results, measured by metric, against different concepts will demonstrate which
concept performs better and suggest directions for further research to investigate the reasons and
measures that could improve results.
Metropolis
15
3 Organization (complexity) metrics
This chapter describes organization (complexity) metrics that are used to compare different
concepts of urban airspace design. Different concepts use different ATM strategies, ranging from
central to fully autonomous control (human operator or an automatic process); therefore intrinsic
measurement of the complexity that is independent of the system in charge of the traffic is
considered here. Future urban transport, like today’s air transport, is a safety critical system and it is
required to be robust. Robustness is strongly related with traffic complexity and it will be analysed in
this chapter.
3.1 Scope and definition
Future urban transport is a safety critical system and it aims to provide safe flow of air vehicles
before making it efficient and expeditious. Maintaining safe separation between vehicles and with
other obstacles is imperative for the system. When a future conflict is detected, a resolution process
is launched which, in certain situations, may generate new conflicts. This interdependency between
conflicts is linked to the level of mixing between trajectories. In addition, uncertainty with respect to
positions and speeds increases the difficulty of predicting future trajectories. The difficulty to control
a system depends on both its sensitivity to initial conditions and interdependency of conflicts. As an
example, Figure 3.1 presents three traffic situations classed according to increasing level of difficulty
as a function of the level of predictability and of inter-dependency between trajectories.
Figure 3.1: Three traffic situations classed by increasing order of complexity
Situation on the left in Figure 3.1, do not present any difficulties as the relative distance between the
aircraft will be maintained, at least for the immediate future. The middle situation, which presents a
significant risk of conflict, is easy to manage as the same direction order must simply be given to all
of the aircraft in order to place them into safe roundabout trajectories. Finally, the most difficult
situation is presented on the right because it is not ease predictable and have a high level of
interdependency between trajectories.
Aim of this sub-category of performance metrics is to measure traffic complexity as an intrinsic
measurement associated with a traffic situation. This measure has to be independent of the system
in charge of the traffic and to solely dependent on the geometry of trajectories. Measuring and
comparing complexity of the resulting traffic situations of different airspace concepts we can
implicitly compare how difficult is to control given system. Measuring the robustness, that is strongly
Low sensitivity
No conflict
Easy situation
High sensitivity
Conflicts with no interaction between solutions
Average situation
High sensitivity
Potential conflicts with
interaction between solutions
Difficult situation
Metropolis
16
related with traffic complexity, allows us to determine how well changes in the initial and
environmental conditions can be absorbed by the structure of the airspace concept or how much
system is invariant to those changes.
3.2 Metrics
Research into air traffic complexity metrics has attracted considerable attention in recent years.
Proposed models can be grouped into two groups: the first one focused on the air traffic control
officer (ATCo) workload, and the second one focus on traffic complexity using automatic conflict
resolution algorithms.
The first group of models has objective to model the control workload associated with given traffic
situations. The main approaches are as follows. In model based on traffic level [6], the workload is
defined as the proportion of control time (duration of control actions taken to resolve conflicts) over
an hour. Queue-based model [7] consider control sector as a system supplying service and queuing
theory is used to determine a maximum acceptable arrival rate for a sector. Model based on
airspace structure [8] estimate the capacity of a sector is based solely on its structure (flight levels,
routes, route intersections, etc.). In the context of operational control, the ideal would be to find a
metric which precisely measures the level of mental effort of the controller that is needed to
manage a certain situations. In NASA a set of traffic characteristics (number of changes in direction,
changes in speed, changes in altitude, etc.) and the workload experienced by a controller have been
studied, and Dynamic density model [9] [10] [11] is built as a weighted sum of traffic complexity
factors. However listed models are not generalized and are linked to studied sector structure and
sensitive to controllers used to infer the model.
Other approaches [12] [13] model the complexity of a traffic situation using automatic conflict
resolution algorithms, for which we measure the number of trajectory modifications required in
processing a given situation. In the same way as before, these methods are highly dependent on the
type of algorithm used to resolve conflicts.
The goal of WP3 of the Metropolis project is to compare developed concepts for the future urban
airspace design. The concepts proposed in WP2 of the project differs in the level of structure and
way how system is managed and controlled. For this reason, previously listed approaches to
measure traffic complexity are not suitable as it is necessary to use an intrinsic traffic complexity
metric that is only linked to trajectory structure, and not to the system used to process them. In the
rest of this chapter, some of existing complexity metrics linked to trajectory structure are studied.
3.2.1 Geometrical approaches [14]
These metrics are calculated at a given instant using the positions and speed vectors of airplanes
present in the chosen geographical zone. Each of these geometrical metrics exhibits a particular
characteristic associated with the complexity of the situation.
3.2.1.1 Proximity metric
Proximity metric is used to characterize the geographical distribution of aircraft in the given volume
of airspace. It allows us to identify spatial zones with high levels of aggregation in relation to their
Metropolis
17
volume. Thus, for a constant number of airplanes in a sector, proximity is used to distinguish
whether these aircraft are distributed homogeneously or in the form of clusters. In the example in
Figure 3.2, on the left five airplanes are well distributed across the sector, while in the situation
represented on the right, the same five airplanes are aggregated in a reduced spatial zone.
Figure 3.2: Two situations of spatial distribution of airplanes
Example in Figure 3.3 presents an artificial traffic situation with four groups of eight aircraft placed
on a square. Top left in the figure shows pure conflict situation where all airplanes converge to the
same point. In the top right and bottom left situation airplanes are organized in the roundabout-like
pattern and parallel tracks respectively. Finally, bottom right figure represent completely random
situation. For each point in the space, the average level of proximity is calculated, considering the
airplanes in the vicinity of the point, and the map of proximity is shown in Figure 3.3. The result
shows that proximity indicator is able to identify areas where airplanes aggregate, but is unable to
distinguish between situations using speed vectors. The two situations at the bottom of the figure
are represented in the same manner, despite the fact that the situation on the right is much more
difficult to manage.
Figure 3.3: Proximity map
Using proximity metric, one can easily (in quantitative manner) identify moments of low traffic
density and moments of clustering. However proximity does not take into account orientation of the
Sector
Metropolis
18
speed vectors and therefore not take into account evolution of the current situation. This
consideration led to develop a convergence indicator.
3.2.1.2 Convergence metric
The convergence indicator is used to quantify the geometric structure of the speed vectors of
airplanes present in a sector. Thus, for identical proximity values, the convergence indicator allows
us to distinguish between converging and diverging aircraft.
When a dense zone has been identified, the zone may be characterized using the rate of
convergence of the aircraft present in this area. This indicator is higher the closer the aircraft and the
faster the convergence. Thus, in the example shown in Figure 3.4, the convergence indicator is used
to provide an unambiguous classification of the eight situations. Each situation corresponds to two
aircraft, for which the relative distance is constant (higher in the top four cases) and the relative
speed varies from strong divergence to strong convergence (from left to right). In the case of
divergence, the indicator will be null, and for convergences, it will be increasingly high as the relative
distance diminishes and the relative speed increases.
Figure 3.4: Converging/Diverging flights example
Same example with the artificial situation involving four groups of eight aircraft (as for proximity
map) is used for the test and resulting convergence map is shown in Figure 3.5. From this figure, it is
clear that only the two non-organized situations (pure conflict and random situation) are identified
by the indicator.
Metropolis
19
Figure 3.5: Convergence map
3.3 Measurement and evaluation
Complexity metrics using previous geometrical approaches can be computed during post-processing
for every concept of the urban airspace design. To do so, it is necessary to log trajectories of all flight
in given simulation scenarios including information of vehicle type (UAV, PAV, etc.).
Trajectories of the future flying vehicles, as airplane trajectories, are objects belonging to the space
with infinitive dimension. It is represented as a mappings from a time interval [a, b] to a state space
E with E either ℜ3 or ℜ6 depending on whether speed is assumed to be part of flying vehicle state or
not. Traditional way for representing trajectories is to use order list of trajectory samples in the 4D
space. Each trajectory sample or position vector contains information of trajectory position (3D) and
time (T). As additional information, each vector position may be associated with velocity vector as a
vehicle intention. However, this information can be extracted from position vectors assuming
uniform motion between two consecutive vehicle’s positions. This assumption is correct if time
discretization step is small enough (5-10 seconds) and the given model represents good
approximation of the real situation.
Smaller time discretization step induces a higher precision of trajectory representation.
Unfortunately small time discretization requires a lot of memory to log the data and highly depends
on size of simulated scenario. As in previous case, we can use approximate tools using linear uniform
motion between two consecutive trajectory positions.
Having position and velocity vectors for all time steps and all flight in the given scenario, complexity
metrics can be computed using the following methods.
3.3.1 Proximity calculation
For the given time and for each vehicle under consideration, we open a spatial weighting window
centred on that vehicle, making it the reference vehicle. We then calculate the relative distances of
Metropolis
20
the other vehicle from the reference vehicle in order to calculate weighting coefficients using a
spatial window using following formula:
������ = ��� � (3.1)
where � is a parameter fixed by the user and ��� is the normalized distance separating vehicle � from vehicle �. Normalized distance is used as the measurement of relative distances between pairs
of vehicle due to the fact that separation norms are not the same in the horizontal and vertical
planes and depend on the type of air vehicles. Consequently a classical Euclidian notion of distance is
not suitable and normalized distance between two vehicles � and � is calculated:
In reality, the risk associated with the convergence of a pair of vehicle also depends on the relative
distance between vehicles. We must therefore simultaneously account for the speeds and relative
distances of each pair of vehicle. One possible form of a complexity metric using convergence
indicator associated with a vehicle � is given below:
3,"�# = λ $ −*�� ∙ �5∙�� ��/7� 89 (3.7)
where λ and α are weighting coefficients.
The complexity of the given traffic situation at a given time is then calculated using the sum of the
convergence of the vehicle present in that zone for the given time.
3.3.3 Robust extension of the metrics
The approaches presented so far use noiseless observations, allowing us to generate instantaneous
metrics. Due to possible change in initial conditions (delay) and external issues (wind, disruptions,
regulations, etc.), the stochastic aspect of observations need to be taken into account in order to
generate reliable (robust) metrics. To do this, trajectory observations, computed through simulation
using a set of flight plans, are affected by noise, particularly in the temporal dimension.
In the context of stochastic process theory, this phenomenon is known as clock shifting: “the
trajectory continues to conform to the flight plan in the spatial dimension, but the position of the
vehicles on the trajectory may be subject to significant deviations in the temporal dimension [14]”.
iP
jP
Vi
Vj
ijd
Metropolis
22
As shown in the Figure 3.7, vehicle trajectories are extended in the future by using uncertainty
segment.
Figure 3.7: Trajectory time shifting
Using the distribution of these time shifts it is possible to produce realistic sets of observations and
to compute convergence indicator for these trajectory segments (not considering time difference).
Finally, resulting complexity metrics values for this trajectory segments can be aggregated in order
to obtain robust metric.
To include robustness when calculating convergence metric, for the given time and for the given
reference vehicle, we open a spatiotemporal window centred on that vehicle (Figure 3.8).
Figure 3.8: Spatiotemporal window for the reference vehicle ;<===> and time ? In the given spatiotemporal window we compute the convergence indicator between reference
vehicle and some other vehicle taking into account all possible pair of observations of those vehicles.
Red lines in the Figure 3.8 indicate all possible pair of observations between vehicles � and �. Convergence indicator associated with a vehicle � with respect to vehicle � at a given time ) is
computed as an time averaging of the convergence over all pairs of observations () − ∆) A B A ) +∆)) and is given by:
where *��C�C and ���C�C represent variation of relative distance and normalized distance of vehicle � at
the time B� and vehicle � at the time B�.
In the same manner as before, convergence indicator associated with a vehicle � is computed as the
sum over all pairs of vehicles in the spatiotemporal window:
3,"�# = λ$3,"�#�� (3.9)
3.3.4 Concept evaluation
The presented method is used to compute complexity of the traffic situation at a given instant of
time t, and it may be used to compare complexity of different traffic situations. However, the goal of
organizational metrics is to evaluate and compare complexity of different concepts of urban airspace
design in Metropolis. It is therefore necessary to compare the overall complexity of the resulting
traffic situations using given concepts.
Using presented method, it is possible to compute the evolution of the traffic situation complexity
for the whole simulation period. The graph in Figure 3.9 allows us to quantify the level of complexity
as a function of time, and identify easily the moments of low complexity (the night or beginning of
simulated period) and the moments of high complexity.
Figure 3.9: Evolution of traffic complexity as a function of time
However, comparison of complexity evolution for different scenarios, is generally not easy task.
Figure 3.10 presents an easy situation, which undoubtedly shows that one of compared concepts
performs better than other one in the terms of traffic complexity.
Metropolis
24
Figure 3.10: Comparison of concept’s traffic complexity – easy situation
Contrary, Figure 3.11 presents a situation where it is not possible to give simple answer which
concept is better without further examination.
Figure 3.11: Comparison of concept’s traffic complexity – hard situation
Several individual indicators can be used to compare concept complexity. Table 3.1 shows resulting
values of some indicators for two concepts presented in Figure 3.11. The maximum value of traffic
complexity over time, taken as indicator of concept complexity, is not suitable as it doesn’t take into
account time distribution of complexity. Using single maximum value of complexity, orange concept
performs worse in this example. The average value of traffic complexity over time, on the other
hand, might favour concepts with high complexity for a smaller time periods than the concepts
which result in traffic situations of moderate complexity for longer time periods. Using average
complexity value only, blue concept in the example performs worse compared to orange concept
(although both concepts have similar average complexity value). Sum of complexity over time
individually might not be sufficient as well. For that reasons, the concept’s complexity in the
Metropolis project is compared using weighted sum of maximum complexity value and sum of
complexity values over time (3.10).
Metropolis
25
3, = α ∙$3,")#E + I ∙ maxE 3,")# (3.10)
Table 3.1: Concept evaluation results
Max. value Avg. value Sum Complexity
(α= 0.1, β=0.5 )
Blue concept 18 10.2 153 24.3
Orange concept 28 9.7 145 28.5
Metropolis
26
4 Operational Metrics
This chapter describes metrics that aim to measure the operational differences between the
airspace concepts considered in the Metropolis project. Here, operational metrics are defined in
terms of:
1. Safety
2. Stability
3. Efficiency
4. Capacity
It should be noted that all the metrics described in this chapter can be computed separately for PAV-PAV, PAV-UAV and UAV-UAV cases (if required). Furthermore, safety and stability related metrics are to be logged/computed only in the experiment area (green trapezium) of the city (to alleviate CPU load during the simulation), while the efficiency and capacity related metrics are computed for the entire simulation area.
4.1 Safety
In this project, four concepts are compared as a means of providing separation through different
airspace structures. Hence, safety related metrics focus on the ability of an airspace concept to
maintain safe separation between aircraft. Separation assurance performance can be measured in
terms of loss of separations and conflicts. Loss of separations, also termed as intrusions, occur if the
minimum separation requirements are violated, i.e., if an ‘intruder’ aircraft enters the protected
zone of an ‘ownship’ aircraft. On the other hand, conflicts are defined as predicted intrusions, i.e., if
the track of an intruder is expected to pass through the ownship protected zone when both aircraft
trajectories are extrapolated over a pre-defined ‘look-ahead’ time. In the following subsections,
metrics associated with these two safety related occurrences are explained.
4.1.1 Loss of Separations
Consider the hypothetical scenario in Figure 4.1 where the path of an intruder aircraft (red dashed
line) through the protected zone of the ownship (grey area) is pictured. Here, the intruder
approaches the ownship with a very low minimum vertical distance (point A) and a very low
minimum horizontal distance (point B). But as these two extremes do not occur simultaneously, it
can be concluded that the example displayed below represents relatively low safety risk to the
ownship.
From the above discussion, it is clear that both the number as well as the severity of protected zone
intrusions must be taken into account when analyzing loss of separations.
Metropolis
27
Figure 4.1: Vertical cut-out view (Z-Y axis) of a loss of separation. The red dashed line indicates the path of an intruder
through the ownship protected zone (grey area). Note that the horizontal and vertical intrusions, MNand MO, are
measured from the boundary of the ownship protection zone.
Severity of Loss of Separations As explained above, the severity of a loss of separation/intrusion is dependent on the path flown by
an intruder aircraft through the ownship protection zone. This severity can be computed using the
following expression:
PQRSTUT7�E/ = maxEWEX Ymin\]�")#, ]U")#_` (4.1)
Here, ]ab and ]ac represents the horizontal and vertical intrusions that are normalized with respect to
the horizontal and vertical protection zone dimensions2, while )9 and )d are the start and end times
of an intrusion. In eq. 4.1, the part within the square brackets implies that the extent of intrusion for
each point flown within the ownship protection zone is equal to the minimum of the instantaneous ]ab and ]ac. From this, the total severity can be determined as the maximum intrusion of all the
individual points that make up the complete intrusion trajectory. Using this logic, the loss of
separation severity for the intrusion trajectory pictured in Figure 4.1 is equal to the normalized
horizontal intrusion at point A.
PQReTUT7�E/ is a metric that is computed during post-processing for each loss of separation that is
experienced during a simulation run. To do so, the following data needs to be logged during a loss of
separation with high a frequency (e.g. 0.5 [Hz]):
1. Scenario file and concept id 2. Runtime 3. Aircraft id 1 4. Aircraft id 2 5. Aircraft type 1 (PAV, UAV, etc.) 6. Aircraft type 2 7. Current horizontal distance [NM] (absolute value) 8. Current vertical distance [NM] (absolute value)
In addition to computing PQReTUT7�E/, recording the above data set for each loss of separation allows
for a graphical comparison of the intrusion paths for the different airspace concepts, see Figure 4.2.
2 Note that intrusions are measured from the boundary of the ownship protection zone, see Figure
4.1
Metropolis
28
Such comparisons may reveal qualitative trends in the trajectories used by different airspace concepts to avoid collisions.
Figure 4.2: Four example trajectories through the protection zone. Such graphical comparisons may reveal qualitative
trends that are not easily visible through numeric metrics.
As mentioned earlier, PQReTUT7�E/ is computed separately for each loss of separation incident.
However, an aggregate metric is needed such that different airspace concepts can be compared for a particular traffic volume scenario. For this purpose, the average loss of separation severity, PQRfffffeTUT7�E/, can be computed during post-processing as: PQRfffffeTUT7�E/ = ∑PQReTUT7�E/hije (4.2)
Here, hije is the total number of loss of separations during a simulation run. This can be logged as a ‘running total’ every time a loss of separation is encountered during the run.
Rogue Aircraft Rogue aircraft are used within the Metropolis project to test the robustness of the different airspace
concepts to non-nominal situations. Rogue aircraft fly over the city without respecting the airspace
structure, nor do they try to avoid other traffic. They can be considered synonymous to small
stochastically moving ‘no-go’ areas that other aircraft have to avoid. Therefore, the number and the
average severity (eq. 4.2) of the loss of separations experienced by the rogue aircraft is a good
indicator of the ability of an airspace concept to absorb unexpected events.
4.1.2 Conflicts
The second set of safety related metrics relates to the number of conflicts. The number of conflicts is
mainly affected by the traffic scenario, however, airspace concepts with a low level of structure are
expected to lead to more secondary conflicts. Furthermore, the number of conflicts is a good
indication of the load on the tactical conflict resolution algorithm (MVP) for each concept.
Similar to loss of separation metrics, it is necessary to evaluate the severity of a conflict. As conflicts
reflect a lower safety risk than loss of separations, it has been decided to classify the severity of a
conflict into two categories that are based on the predicted time to closest point of approach, )klm:
1. )klm < 60[r] 2. )klm < 30[r]
In TMX, the first conflict category is denoted as an ‘amber alert’, which requires the aircraft with
lower priority in a conflict pair to maneuver to solve the conflict. The second conflict category is
denoted as a ‘red alert’, and requires both aircraft in a conflict pair to maneuver to solve the conflict.
Percentage of Aircraft in Conflict
Metropolis
29
The percentage of aircraft experiencing a conflict can be computed online as a measure of the
Here n is the total number of aircraft landing at a particular runway during the logging hour.
During post-processing, the average and minimum intervals can be computed for each runway. This
data can be used to cluster runways in terms of (pre-defined) arrival interval ranges (by means of a
cluster bar chart). In this way, it is possible to investigate whether a particular airspace concept leads
to unacceptable arrival sequences that cannot be sustained in reality.
4.3.5 Summary of Efficiency Related Metrics
Table 4.2 below summarizes the safety related metrics proposed for Metropolis:
Table 4.2: Summary of efficiency related metrics
Metric Description
Route Efficiency Ratio of direct distance and actual distance flown
Metropolis
33
Relative Delay Absorption capability
Comparison of trip times for 12 standardized trajectories
Departure Delay Time interval between the spawning and take-off of an aircraft
Arrival Sequencing Time interval between two successive arrivals at a particular airport
Note that all efficiency metrics require parameters to be computed and logged online, using which
the aggregate metrics listed in Table 4.2 can be computed during post-processing.
4.4 Capacity
4.4.1 Influence of Safety and Efficiency Metrics
One of the main research goals of the Metropolis project is to investigate the airspace structure-
capacity relationship. Four traffic scenarios of increasing traffic demand have been defined to study
this relationship. Similar to other transportation systems, airspace capacity is difficult to define
explicitly, however, it is clear that that both safety and efficiency must be considered when
evaluating the structure-capacity relationship. Therefore, it is proposed that capacity be measured
indirectly by considering the relationship of the safety and efficiency metrics with respect to the
(prescribed) demand of the four traffic scenarios.
To better illustrate this rationale, consider the following simple fictional example where the average
conflict percentage for two airspace concepts are plotted against scenario number (i.e., against
traffic demand):
Figure 4.4: Fictional example of how capacity can be inferred indirectly from safety and efficiency metrics. Here the
relationship between the average conflict percentage and traffic demand/scenario is illustrated. The gradient with
respect to traffic demand may reveal capacity limits.
Such plots may reveal how the safety and efficiency metrics are affected by traffic demand for the
different airspace concepts. By reviewing the plots of several metrics in unison, as well as evaluating
the trends between metrics (for instance between conflict percentage and the airspace busyness
index), it may be possible to qualitatively study the effect of structure on capacity. Furthermore, by
analyzing the gradient of the safety and efficiency metrics with respect to demand, it may be
possible to detect capacity limits. For instance in Figure 4.4, the large increase of the conflict
percentage for 'concept 2' between scenarios 3 and 4 may indicate that a capacity limit exists
Metropolis
34
between these two scenarios. However, it should be noted that the limited number of scenarios
used in this project may make it difficult to arrive a conclusive capacity limit for all concepts, as
illustrated for 'concept 1' in Figure 4.4.
4.4.2 Traffic Density vs. Traffic Demand
Another way to evaluate capacity is to measure the extent to which traffic density matches the pre-
defined traffic demand for each scenario. It is possible that for high demand scenarios, the
departure metering used to prevent conflicts during take-off may limit the maximum number
aircraft that can enter the airspace. The ratio between the number of aircraft that took-off and the
number of spawned aircraft during the logging hour can be used to measure this relation:
� = h�yE��T|ddh�y�����
(4.10)
A running total of the number of 'actual' take-offs can be used to log h�y�����XXduring the logging
hour. As mentioned above, the number of aircraft spawned during the simulation hour, h�y�����, is
known in advance of the simulation.
If the ratio between density and demand is below a prescribed threshold (for example 75%), or if
there is a significant reduction of the ratio between two scenarios, a capacity limit may be identified.
Metropolis
35
5 Environmental metrics
5.1 Introduction
The impact of aircraft movements on the environment has always played an important role in the
acceptance of new airports, new air routes, and new aircraft types. It is therefore expected that the
introduction of a complete new concept, such as the setting of UAVs/PAVs in a metropolis, will have
significant impact on the environment of the city and the environment involved. In Figure 5.1, an
overview is given of some of the most important concerns with respect to PAV, and UAV operations
within a city. These impacts can be split into two categories: First, the main environmental concerns
around traditional aircraft, such as emissions/pollution, noise pollution, and third party risk. These
are equally present for PAVs and UAVs. Second, other concerns may also become important such as
shadow flickering (similar to wind turbines) or light pollution in case of nightly operations, privacy
concerns, distraction, and effects on the biotope. For this latter category, the impact on the society
is still not clear and requires a timely process to assess the real impact, similar to other
environmental questions raised with the introduction of other “new” transportation technologies,
such as the train, the automobile, and the (traditional) aircraft.
For the first category, the emission concern may be of less concern in 2040 as the vehicles are
smaller than traditional aircraft, and there is significant development of sustainable energy sources
now already, that may resolve emission issues in 2040. Furthermore, the total flying time in a
scenario relates strongly with the energy use for an aircraft and can be used as indicator for the
emission/energy use as well.
Both noise pollution and third party risk is therefore the most promising candidate to measure for
the METROPOLIS scenarios. Both environmental impact relate strongly to the technology that is to
be developed, and will therefore deviate from the impact that is expected in 2040. But the
differences between the scenarios that are tested can be determined, as long as the parameters for
the model are the same. Third party risk is probably the most challenging impact to be considered,
Figure 5.1: Environmental concerns related to Personal and unmanned aerial vehicles.
Metropolis
36
as there is little literature available for these aircraft, and therefore chosen to be measured in more
detail for METROPOLIS.
5.2 Scope and definition
For all of the metrics mentioned in this chapter, the focus is on comparing the different scenarios,
not on finding the right absolute value or score for each scenario: The future scenario of Metropolis
with its aircraft types and other assumptions are too futile to predict the right value or score, but
comparing the scenarios with each other creates an equal level playing field.
It is also assumed that traffic cannot migrate from one type of aircraft to another, as defined in the Metropolis definition report [2]. So a number of persons that plan to take a PAL-V to work cannot switch to taking the AW609 ‘bus’ instead, resulting in a decrease of PAL-V movements and an increase in AW609 movements.
5.3 Third party risk metric
5.3.1 Introduction
One of the main concerns is the safety for the people on the ground, also known as the third party
risk (TPR). In the present context, third party risk is the risk of people on the ground involuntarily
exposed to an aircraft accident. Those people on ground are defined as third parties. By definition,
people who are on board of an aircraft (air crew and passengers) and people who work within an
airfield terrain are considered as first party or second party, respectively.
As mentioned above, different futuristic aerial vehicles will operate alongside with each other. For
the personal air vehicles and unmanned aerial vehicles it is envisioned that the amount of traffic
The following metrics will not be considered for evaluating the scenarios, but are worth to
mention as they may play an important role in the acceptance of any of the Metropolis
scenarios:
• Privacy concerns: the current debate on the acceptance of civil unmanned vehicles is strongly
influenced by privacy concerns and the risk of eavesdropping by providing a camera (or
microphone) to a UAV.
• Shadows flickering: The effects of shadow flickering of wind turbines cause some of these
energy sources to be turned off during certain time of day. This effect can be equally
disturbing to civilians living or working underneath certain aircraft routes. The exact impact
of shadow flickering is not yet known as the Metropolis concept has not been adopted yet.
• Light pollution: Similar to shadow flickering is the disturbance of low-flight aircraft emitting
lights for navigational or other purposes during the night.
• Effects on biotope: although large cities are usually not considered to be national parks, there
can still be a variety of animals, mostly birds, which can be affected by the increase of low-
flying aerial movements.
• Distraction: Just as animals, people can be distracted as well by the metropolis concept.
Especially around schools, hospitals or some other areas where concentrated work is done,
this can be important to take into account.
Metropolis
37
could be significant. With their capabilities to operate within confined air space, it is likely the traffic
takes place in or in the close proximity of built-up areas. In order to understand the magnitude of
risk that people on the ground are exposed to, there is a need to quantify the third party risk due to
the traffic of those aerial vehicles.
The NLR has done research on TPR around airports [15] and developed a TPR model that is now part
of Dutch law to determine safety zones around airports [16] in the Netherlands. The principles of
this model can be used to determine the TPR in METROPOLIS. However, the specific procedures of
METROPOLIS aircraft must be considered, and in particular the cruise phase of flight, which is
exempted from consideration in the third party risk around an airport, should be taken into account
as well. This requires a different approach in modelling third party risk and in developing a TPR
model suitable for the METROPOLIS scenarios.
5.3.2 Assumptions and considerations for TPR model
Risk factor related to (human introduced) hazards depends on the compensating benefit [17]: the
higher the societal benefit, the higher the risk that is accepted for a lethal accident. At this moment
the benefit of using drones or personal air vehicles is low and not considered beneficial, so the
safety criteria for these vehicles will be high. Perception of safety plays an important role for UAV
operations [18]. It could be expected that when the benefits of improved delivery times or reduced
travel times become more obvious, the safety criteria may reduce, but this is not certain.
The main difference between the traditional TPR model used in the determination of risk around
airports, and the TPR model in Metropolis is that for the traditional model the risk during cruise
phase of a flight is not considered while in Metropolis it should be taken into account. Because the
flights of PAV/UAV concentrate over the city area where the population density is high, the cruise
phase of flight should be included in risk modelling and determination.
Secondary effects caused by accidents such as an aircraft crashing into a gas station causing huge
fire or explosion shall not be taken into account in this model. Only the risk of direct causalities on
ground caused by aircraft crashes is considered relevant.
For TPR, the following kind of hazards can be discerned [19]:
• Mid-air collision with other aircraft;
• Other hazards, such as system failure, weather, bird strike, terrorism, human error. For
these kinds of error a specific analysis per type of hazard is considered not necessary, since a
general estimation of the total hazard per aircraft type is assumed to be sufficient.
The risk of mid-air collision is described in other metrics, See section 4.1.1 on page 26,, and
therefore not specifically considered for the third party risk. Instead of analysing the mid-air
collision risk, an estimated mid-air collision risk shall be taken into account when calculating the risk
in cruise phase.
The following factors have influence on the TPR model:
• Probability of an accident during landing or take-off. This probability is expressed per flight
stage.
Metropolis
38
• Probability of an accident during cruise and the chance. This probability is expressed per
flight hour.
• Location of the crash area after malfunction (it depends on whether the aircraft is fixed wing
or not). There should be some estimation of the location of the crash area. Also the chance
that an aircraft crashes into a building and leads to casualties in the building or on the
ground beneath should be taken into account.
• Counter measures to prevent crashing (e.g. different routes, back-up systems, etc.)
The routes that the aircraft take are influenced by:
• The airspace concept used and the routes that it proposes
• The take-off/landing zones
• The scenario where traffic demand appears
For the impact of the crash, the following factors play a role:
• Consequence area in terms of aircraft size expressed in maximum (take-off) weight. This is
the area in which the people on ground could receive fatal injury should an aircraft accident
occur.
• Density of the population
• Shielding of population by buildings [21]
• Commuters/traffic on the street
• The time of day in relation of the number of people in the area.
The model should include the lethality of a crash, which is determined as the ratio of the number of
people killed in the crash area and the number of people present in that area. A separate method
can be used to determine the number of people in the crash area at a certain time of day.
A summation of all traffic shall be provided, but it is expected that the TPR of the smallest aircraft
type (the microdrone) will have a negligent impact on the resulting, total TPR, because of the limited
number of microdrone movements. As such, the inclusion of microdrone in the model is still crucial
for the overall concept and for the determination of the level of risk acceptable for the population in
the city. The same is true of aircraft movements over the lower density areas of the city (outer ring),
that will have a relative small contribution to the total risk.
5.3.3 Model restrictions
The third party risk model makes use of some assumed values, based on existing aircraft accident
data. This is not representative for operations of PAVs/UAVs in a dense city as there is not enough
data available for accidents of this kind. For this reason, the data should only be compared between
the different scenarios. For this reason, the full-mix concept is chosen as the reference concept for
Third party risk. The third party risk of all other three concepts will be related to the safety of this
scenario. For instance, the layers concept can score a certain percentage better or worse than the
full-mix concept.
The data can also be used to determine target level of safety for future PAV/UAV operations, but this
is not part of the goal of the Metropolis project.
Metropolis
39
The TPR model implementation can be computational intensive due to the number of tracks that are
being processed, the length of the simulation, and/or the chosen model parameters. For this reason,
further simplifications to the metrics calculation may need to be chosen, so the metrics can still be
calculated. Examples of simplification can be:
• Reduction of interpolation of the tracks or grid size or density of the simulation area
• Reduction of number of tracks or simulation time that is considered for this metrics
• Reduction of aircraft types, so only aircraft types that have a significant influence on the TPR
are taken into account.
5.3.4 Input from simulation
• Aircraft tracks for each aircraft in the simulation.
• Population density for the city.
5.3.5 Output from metrics
• Percentage of Third Party risk compared to the reference concept (full-mix).
5.4 Third party risk model design
The model shall make use aircraft tracks and iterate over the given tracks. For each of the aircraft
tracks, the risk is calculated by the combination of:
1. The location and the chance of an accident of the aircraft.
2. The location of the crash area, in case of an accident.
3. The lethality chance based upon the aircraft and the location.
5.4.1 Probability and location of an accident
To calculate the probability of an accident of one flight, this flight will be split up in a take-off phase,
a cruise phase, and a landing phase. For each of these phases, prior research exists on accidents in
these phases, and a model can be made that calculate the risk on the ground for the take-off, cruise,
and landing phase. After that, these risks can be summed up for the total risk of the whole flight on
the ground.
Take-off and landing phases
For the take-off and landing phases, the probability of an accident can be calculated based upon
prior research of similar aircraft types. There is extensive data available that includes both the total
movements, and the kind and number of accidents during start or landing. However, such approach
in data research is not feasible for the given aircraft types in METROPOLIS, as it is impossible to find
any historic data at this moment. Therefore, similar aircraft types must be used, such as other rotor-
based aircraft. The probability that a landing or a take-off results in an accident will be a parameter
in the model. A Weibull distribution [22], as used in the NLR TPR model [16] to describe the aircraft
accident locations, will be used to determine where the accident will take place. The specific
parameters of the Weibull distribution shall be based on prior data research and engineering
judgment.
Metropolis
40
Specific take-off and landing failure rates for rotor-craft aircraft can be found [23], but they explicitly
exclude data from tilt-rotor aircraft. However, a Congressional study indicates that a typical tilt-rotor
aircraft such as the V-22 Osprey has similar failure rates [24]. Therefore, for the aircraft in
METROPOLIS, either the single engine or dual engine turbine failure rates are taken. As there is no
data available for the microdrone UAV, the worst-case values are.
Table 5.1: METROPOLIS Aircraft types matched on helicopter types from Third Party Risk studies
Assumed type
PAL-V One Piston/single turbine
Terrafugia TF-X Twin turbine engine
AgustaWestland 609 Twin turbine engine
Microdrone MD4-3000 Piston/single turbine
To determine potential impact area of an accident during take-off or landing, the flight track needs
to be available from take-off to a certain distance where the Weibull distribution can be considered
1. For landing, the track should be reversed from touch-down back to the point where the Weibull
distribution is valid for this phase of flight. For these calculations, iteration along the route should
take place based on fixed distance steps.
Two modifications are proposed compared to the NLR TPR model. First, in the NLR TPR model, an
aircraft is considered to be in cruise over 500 feet [23]. For METROPOLIS is can be expected that
some aircraft, in particular UAVs, will have a more vertical than horizontal flight path. For this
reason, instead of using the height of 500 feet as cut-off, the total flown distance of 1000 metres is
taken as limit what part of the flight path is considered part of the start or landing. Second, it could
be that there is still some risk left on the Weibull curve during the take-off or landing phase at the
moment of cut-off. For an airport-based TPR model, this risk is negligible, as it concerns areas further
away from the airport. In Metropolis, this residual risk will be evenly distributed over the trajectory
in which the aircraft within the 1000 metres.
Cruise phase
Table 5.2: METROPOLIS Aircraft cruising speed
Cruise speed [kts] Cruise speed [m/s]
PAL-V One 75 39
Terrafugia TF-X 160 82
AgustaWestland 609 200 103
Microdrone MD4-3000 31 16
For the cruise-phase, the probability of an accident will be calculated based upon flight-hours. In
[25], an accident rate of 5.6x10-5 per flight hour is given based upon data from the National
Transportation Safety Board (NTSB) that leads to an accident resulting in a fatality on the ground to
1.48x10-7. For METROPOLIS, as no extensive data is available on PAV/UAV accident rates, the
decision to compare the concepts in relation with each other can be justified. For this reason, this
number is only used to balance the relation between the accident risk in cruise phase and the
accident risk in start/landing phase.
Metropolis
41
As the risk is based upon flight-hours, the speed of the aircraft shall be taken into account. To
determine the risk for the cruise-phase, the model needs to iterate through the given route, and
calculate the risk for each iteration step. The time step (in seconds) of this iteration speed will be a
parameter in the model.
If the aircraft route is given as a list of waypoints, an approximation must be made for the iteration
step between the waypoints. It is expected that this list of waypoints is not too far away and the
iteration can be considered to be a straight line. Alternative methods are the use of interpolation
techniques like splines3, but these methods are more computational intensive and are not needed
for a simple model set-up described in this context. It is also assumed that the aircraft speed is
known for each of the given waypoints, and therefore, the average speed between the previous and
next waypoint is used to calculate the aircraft speed during the interpolation.
Mid-air collision risk
The risk on mid-air collision for a specific number of aircraft within a volume of airspace can be
translated into certain chance on a mid-air collision per flight hour [26]. For this reason, an
additional factor can be added to the risk that is found in cruise-phase.
5.4.2 Location of the crash area
Figure 5.2: Accident chance and impact area of a single track
A grid will be defined that will contain the operating area of the aircraft. Each grid cell will receive a
risk value based upon the offered number of tracks and the model that calculates the risk of this cell.
For obvious reasons, simplicity and consistency, the size of the grid cells are considered to be square.
The number of grid cells and the size of the grid cell will be a parameter in the model. The smaller