Final Thesis
STCA - an aircraft conflict alert system
by
Bång Ola Norén
LiTH-IDA-EX--04/033--SE
2004-04-02
Supervisor: Børge Midtgaard Examiner: Jörgen Hansson
Division, Department Institutionen för datavetenskap 581 83 LINKÖPING Date 2004-04-02 Språk Language Rapporttyp Report category ISBN
Svenska/ Swed ish X Engelska/ English
Licentiatavhand ling
X Examensarbete ISRN LITH-IDA-EX--04/ 033--SE
C-uppsats
D-uppsats Serietitel och serienummer Title of series, numbering ISSN
Övrig rapport
____
URL för elektronisk version
http:/ / www.ep.liu.se/ exjobb/ id a/ 2004/ d t-d / 033/
Titel
Title
STCA - ett varningsystem för konflikter mellan flygplan STCA - an aircraft conflict alert system
Författare
Author
Bång Ola Norén
Sammanfattning
Abstract
The purpose of this Master’s Thesis is to prod uce a specification for the aircraft conflict alert system STCA, and implement a prototype as a mod ule in the air traffic surveillance system NOVA9000. The specification is constructed based on functional requirements from EUROCONTROL and d escribes a system using a nominal trajectory method , where the future paths of aircraft are estimated . The trajectory is created using a probabilistic approach, where future positions are d escribed with probability field s.
The prototype is implemented using the specification with some simplifications. The prototype is evaluated using record ed traffic from a heavy air traffic region surround ing an airport with parallel runways. 15 alerts were ind uced in 1,5 hour of morning traffic; this is far too much to be acceptable. Improvements are proposed and explanations to the high rate of alerts are mad e.
Nyckelord
Keyword
A
BSTRACTThe purpose of this Master’s Thesis is to produce a specification for the aircraft conflict alert system STCA, and implement a prototype as a module in the air traffic surveillance system NOVA9000.
The specification is constructed based on functional requirements from EUROCONTROL and describes a system using a nominal trajectory method, where the future paths of aircraft are estimated. The trajectory is created using a probabilistic approach, where future positions are described with probability fields.
The prototype is implemented using the specification with some
simplifications. The prototype is evaluated using recorded traffic from a heavy air traffic region surrounding an airport with parallel runways. 15 alerts were induced in 1,5 hour of morning traffic; this is far too much to be acceptable. Improvements are proposed and explanations to the high rate of alerts are made.
A
CKNOWLEDGEMENTSI would like to thank my supervisors Børge Midtgaard, Jon Arne Nysvend and Enrico Piazza for their help and guidance throughout my work. Thanks also to my examiner Jörgen Hansson. Thanks to Lars-Erik Andersson for helping calculate the trajectory of a turning aircraft. Many thanks to the employees of the NOVA division as it has been a pleasure working with them.
I would like to dedicate this thesis to my family for their support during my years at the University of Linköping. Special thanks to Astrid for her
T
ABLE OF CONTENTS 1 INTRODUCTION ________________________________________ 1 2 BACKGROUND _________________________________________ 3 2.1 STANDARDIZATION ORGANIZATIONS ______________________ 3 2.2 EXISTING METHODS ___________________________________ 3 2.2.1 Modelling approaches ________________________________ 3 2.2.2 State propagation methods ____________________________ 62.3 EXISTING SYSTEMS ____________________________________ 7
2.3.1 National Air Traffic Services __________________________ 7 2.3.2 National Aerospace Laboratory in Netherlands ____________ 8 2.3.3 SAABTech ________________________________________ 9 2.3.4 Comments ________________________________________ 11 3 PROBLEM DESCRIPTION ______________________________ 13 3.1 BACKGROUND_______________________________________ 13 3.2 PROBLEM ANALYSIS__________________________________ 13 3.3 OBJECTIVE _________________________________________ 14 4 APPROACH ___________________________________________ 15
4.1 SINGLE PATH METHODS _______________________________ 15
4.1.1 Linear prediction model _____________________________ 15 4.1.2 Preserved acceleration model _________________________ 18
4.2 PROBABILISTIC METHODS______________________________ 23
4.2.1 Portray trajectories _________________________________ 23 4.2.2 Situation connected behaviour ________________________ 24
5 SPECIFICATION _______________________________________ 26
5.1 SYSTEM OVERVIEW___________________________________ 26
5.2 FUNCTIONAL REQUIREMENTS __________________________ 26
5.3 SYSTEM-WIDE DESIGN DECISIONS________________________ 28
5.3.1 Provided data______________________________________ 28 5.3.2 Design choices_____________________________________ 29
5.4 SYSTEM ARCHITECTURAL DESIGN _______________________ 30
5.4.1 Initialization block__________________________________ 31 5.4.2 Trajectory prediction ________________________________ 31
5.4.3 Conflict search algorithm ____________________________ 32 5.5 REQUIREMENTS TRACEABILITY _________________________ 36 6 PROTOTYPE __________________________________________ 38 6.1 DESIGN DECISIONS ___________________________________ 38 6.2 DESIGN ____________________________________________ 40 6.2.1 Target info ________________________________________ 41 6.2.2 STCA class _______________________________________ 41 6.2.3 Vertical class ______________________________________ 43 6.2.4 Horizontal class ____________________________________ 44 6.2.5 Area class ________________________________________ 45 6.3 SYSTEM EVALUATION_________________________________ 46 6.3.1 Simulator description _______________________________ 46 6.3.2 System tests _______________________________________ 48 6.3.3 Real traffic situations________________________________ 51 6.3.4 Trajectory evaluation________________________________ 56 6.3.5 Testing and tuning __________________________________ 57 CONCLUSIONS ____________________________________________ 60 6.4 SYSTEM PERFORMANCE _______________________________ 60 6.5 FUTURE IMPROVEMENTS ______________________________ 61 ABBREVIATIONS __________________________________________ 62 REFERENCES _____________________________________________ 63 7 APPENDICES __________________________________________ 65
T
ABLE OF FIGURESFigure 1: Conformance system. ... 4
Figure 2: Nominal trajectory... 5
Figure 3: Escape trajectory. ... 5
Figure 4: Single path... 6
Figure 5: Worst case. ... 6
Figure 6: Probabilistic... 7
Figure 7: Vertical inaccuracy areas describing flight expected to level off. ... 8
Figure 8: Horizontal fields describing future positions for two aircraft. ... 9
Figure 9: Separation distance set out against time... 9
Figure 10: A situation predicted to be hazardous. ... 10
Figure 11: Safe situation, fields describing different timeslots are allowed to interfere. . 10
Figure 12: Two aircraft with velocity vectors... 16
Figure 13: Same situation as in Figure 12, but with relative velocity marked out. ... 17
Figure 14: Acceleration from current state to future state. ... 19
Figure 15: Mathematical trajectory and Euler trajectory coincide. ... 22
Figure 16: Euler trajectory diverges from the Mathematical trajectory. ... 22
Figure 17: Two targets with predicted uncertainties... 24
Figure 18: Predicted separation with uncertainty. ... 24
Figure 19: Block diagram of architectural design... 30
Figure 20: Block diagram of trajectory prediction process ... 31
Figure 21: Block diagram of conflict search... 33
Figure 22: Vertical conflict between two aircraft with interacting zones... 35
Figure 23: Fan shaped homogenously distributed field... 39
Figure 24: Class diagram describing the structure of the STCA module ... 41
Figure 25: An aircraft in the NOVA9000 environment... 46
Figure 26: Command window showing an alert... 47
Figure 27: Two aircraft in a head on conflict situation... 48
Figure 28: Two aircraft in a right angle conflict situation. ... 49
Figure 29: Two aircraft in a gaining conflict situation. ... 50
Figure 30: Arrival traffic at Charles de Gaulle airport in France. ... 51
Figure 31: Two aircraft in a conflict situation in the arrival area. ... 52
Figure 32: Two aircraft inducing an alert and two aircraft in a safe situation, despite horizontal conflict. ... 53
Figure 33: Conflict situation where the aircraft (AFR214P) gains on (AFR275W), which results in an alert situation. ... 54
Figure 34: A conflict situation where three aircraft are involved... 55
1 I
NTRODUCTIONAirspace is becoming more congested and automated conflict alert systems have become more important to serve as traffic monitoring aids.
Short Term Conflict Alert, STCA, is one among many different alert
systems preventing airplanes from colliding in the air. STCA communicates with air traffic controllers while there are other systems like TCAS1 that communicate directly with the pilots. STCA is a short-term alert system, there are other alert systems foreseeing dangerous situations in the longer term.
Within radar-controlled airspace STCA provides defence against either controller error or pilot error in adherence to clearance. Statistics from the United Kingdom shows that STCA reduces the probability of an air
conflict2.
STCA’s role is to warn if two planes come too close and a conflict is probable. It is important that there is enough time for the pilot to do an avoidance manoeuvre. It is also important that STCA does not cause
nuisance alerts. The number of alerts should be kept to an absolute minimum without missing any vital alerts.
When the STCA foresees, in real-time, an abnormal proximity to occur within less than two minutes, it warns the controller with a visual warning on his radar information display terminal. The STCA is a last resort backup tool, not a control support tool. In other words, the controller is not supposed to wait for an STCA alert before detecting a conflict risk and starting the resolution procedure. The controller is alerted by the STCA when he has not detected the risk of a conflict early enough.
A conflict is a violation of the separation distance minima applicable between two aircraft. In more technical terms, the STCA considers that a pair of aircraft is in conflict when a future hazardous situation puts them in danger within a time limit less than the warning time. The warning time is the time judged to be sufficient for the controller and pilot to resolve the conflict from the moment the alert is given.
1 Traffic Alert and Avoidance System
2 UK Full System Study, as reported by A C Price, C P Smith, K M Carpenter, ICAO SSR Improvements
and collision Avoidance Systems Panel SSR Mode S Airborne Collision Avoidance Systems Working Group 2, Sydney, March 1995
Chapter 2 gives a theoretical background to conflict detection and describes briefly three existing STCA systems. In chapter 3 the problem is described with a problem analysis and objectives. Chapter 4 gives a description of our approach. Three different models of how to discover conflict situations are described. Chapter 5 presents the specification to the STCA system,
functional requirements are stated and a design description is given. Chapter 6 presents the prototype that has been implemented. Moreover Design
decisions are explained and a description of the structural design is given. An evaluation of the prototype using recorded air traffic from Charles de Gaulle is also included. Our conclusions are presented in chapter 7.
2 B
ACKGROUNDIn this Chapter we give the theory of conflict detecting. First a short description of regulating recommendations followed by a description of trajectory estimation methods. Finally a survey of existing systems is given.
2.1 STANDARDIZATION ORGANIZATIONS
Air traffic often involves many countries, and it is easy to understand that international agreements are crucial for safety and efficiency reasons. There are several organizations making recommendations and regulations for air traffic on a global, continental, and a national level.
The organization we used for material regulating functionality of STCA was EUROCONTROL. EUROCONTROL has published a document with
operational requirements for Safety Nets [8]. Safety Nets is a collection of three independent safety standards of which STCA is one. The other
included systems also concerns short-term3 air traffic surveillance systems. The EUROCONTROL document recommends requirements and expected behaviour of an STCA system. These recommendations have been taken into consideration when constructing our specification.
2.2 EXISTING METHODS
In order to detect conflict situations between aircraft, future trajectories are estimated. Radar and other information sources report states of traffic, such as aircraft position, speed, and heading. Future states of traffic are estimated by a dynamic model using current traffic state information. This estimated information is processed to identify future conflicts.
2.2.1 Modelling approaches
There are many different approaches to implement an alerting system. Kuchar and Yang have studied different methods and identified several key modelling methods [13]. They could not find any single solution that stood out as being clearly the most efficient or effective model. Beneath are descriptions of three philosophies of how to model a probabilistic conflict detection system.
Conformance systems o rder ut ridor d be
he boundaries should enclose a region large enough to minimize false n
his is an intuitive way of keeping air traffic away from hazards; ensure that future
ince conformance methods rely on boundaries they are appropriate for ries
aft The idea behind this philosophy is t
alert if an aircraft crosses a boa
to a predestined area free of hazards. Typically an area can be mapped o around the normal approach cor of an airport. Then aircraft can manoeuvre within this safe area and land at the airport, but an alert is issued if they do not follow expecte behaviour. More technically it can
described: “A boundary of acceptable operating states is defined beforehand, and an alert is issued when the state of the aircraft exits this boundary.”
Figure 1: Conformance system.
T
alerts. They shall also be small enough to protect aircraft from hazards; a aircraft shall have enough time to correct a problem when it gets the alert from crossing a boundary.
T
hazards remain outside the boundaries and the air traffic inside. This approach is relatively simple in that it relies only on the current state; trajectory predictions are not required. The system delivers an alert if an aircraft crosses a boundary.
S
areas with known behaviour. For example, approach areas where bounda easily can be drawn. Free flight situations where aircraft could be located anywhere and going in any direction need more complex methods. If aircr flight plan information is available, the flight route can be checked for conflicts and safe boundaries may be defined [9].
Nominal Trajectory This philosophy predicts future trajectories of aircraft with a trajectory model. These trajectories are then compared to determine if there is a future conflict using current control
strategy. With this approach alerts are only issued if the model predicts a future dangerous situation. When a
conflict is discovered it may already be too late to prevent hazard encounter. To avoid this, alerting distances are typically determined through trial-and-error tests using fast-time simulations of aircraft encounters. Thresholds are then set to provide best performance.
Figure 2: Nominal trajectory.
The trajectory model foresees the position of the aircraft into the future. The longer time future state is predicted the worse accuracy is generally
accomplished. Therefore these methods are typically limited to a certain look-ahead time. There are many different ways of modelling these trajectories; section 2.2.2 describes different methods [9].
Escape Trajectory
rajectory Models using this design
approach ensure that there always is an available escape path. A possible escape t
is extrapolated from the current position, and if escape
possibilities are limited the model indicates an alarm. This philosophy is similar to nominal trajectory with the difference that instead of
foreseeing the most probable trajectory, an escape trajectory is predicted. Also this philosophy needs testing to adjust conditions for a safe escape [9].
Combinations
Combinations of these three philosophies are certainly possible, and in fact are probably desirable in many cases. For example an approach situation: First a conformance system detects unexpected behaviour, but before sending an alarm a nominal trajectory system checks to see if there is a hazardous situation. A nominal trajectory system alone may have induced too many nuisance alerts therefore a combination may give better results [9]. 2.2.2 State propagation methods
In order to detect conflicts between aircraft, it is necessary to project the future positions of the aircraft over time. To do so, an appropriate state propagation method is required to propagate the aircraft’s current states. Approaches to conflict analysis generally rely on one of two propagation methods, termed nominal and worst case. In between these methods is the probabilistic approach, where the likelihood of various future positions is weighed by their probability of occurrence [13].
Single path
he The single path approach assumes that the aircraft
follows a single future path. It is common to predict the path along a straight line with preserved velocity and direction. In this approach conflicts can be discovered as crossings between future paths. Uncertainties in the future trajectory are not taken into consideration and t output is either a safe or a dangerous situation.
Figure 4: Single path.
Worst case
In the worst case approach every possible future path, limited only by the aircraft aerodynamic capabilities, is considered. Here uncertainties in future trajectories are taken into consideration, but there is no possibility to evaluate the chance of conflict. The output from a worst case model is either safe or dangerous. This method is only possible to use in a short time
perspective. With a long time perspective the area of possible positions are be too big to handle.
Figure 5: Worst case.
A combination of the two methods above is the probabilistic approach. In this method uncertainties are used to develop a set of likely future trajectories, each weighed by its probability of occurrence. This provides an opportunity for a
balance between relying too heavily on the state adhering to a single path, versus relying too heavily that the states exhibit a worst case behaviour. The advantage of a probabilistic approach is that decisions can be made on the single path of
encountering a hazard – safety and false
alarm probabilities can be assessed and considered directly. The
probabilistic method is also the most general, since single path and worst-case models can be considered subsets of probabilistic trajectories.
Figure 6: Probabilistic.
2.3 EXISTING SYSTEMS
There are several STCA systems operating today. This section describes existing operating STCA systems to find out which methods and
philosophies are used.
It has been a problem to find detailed system descriptions in a competing market. We briefly describe what we have managed to learn about three different systems. The information is from commercial material describing their systems.
2.3.1 National Air Traffic Services
National Air Traffic Services (NATS) of United Kingdom has a system running that uses a single path approach [10]. It relies on a linear prediction model to calculate the closest point of arrival between a pair of aircraft. To find possible conflicts, all pairs are tested using the following three filters. Linear prediction filter
This filter predicts future horizontal and vertical positions using straight lines. It identifies aircraft that comes within given horizontal and vertical separation parameters in the same time period.
Current proximity filter
The current proximity filter measures actual horizontal and vertical separation and identifies whether aircraft violate the parameters (basic proximity conditions).
Manoeuvre hazard filter
This filter predicts ahead for turning aircraft at similar altitudes. It uses constant turn trajectories to identify aircraft, which comes within a given horizontal parameter.
Alert confirmation
Delay mechanisms exist to delay alerts when there is no immediate collision risk. The following describes the functionality of three delay mechanisms in the STCA system:
• STCA can detect when the need for an alert is marginal (low risk conditions) and delay the display of the alert until the situation gets worse.
• STCA can detect when an aircraft is turning and delay an alert if the turn takes the aircraft away from hazard.
• STCA can delay an alert if there is more than enough time for a standard manoeuvre (horizontal or vertical) to avoid a conflict. Alerts are delayed for as long as possible to allow time for resolution of a conflict situation without the need for an alert. A delay mechanism continues to delay an alert as long as there is still sufficient time for the controller to issue instructions and for these to be acted upon.
2.3.2 National Aerospace Laboratory in Netherlands National Aerospace Laboratory in N
system [2]. Below is a technical description.
etherlands has developed an STCA
oth vertical and horizontal conflict l
g-Figure 7: Vertical inaccuracy areas
B
searches consist of the comparison of future inaccuracyareas. For the vertica search, a continuous vertical speed (including deviations) is taken into account as well as a possible levellin off during the warning time. Figure 7 shows the inaccuracy areas for a
climbing flight expected to level off soon. The comparison between future vertical conflicts is done analytically.
For the horizontal search, inaccuracy limits are introduced both for the ground speed as well as the course. Using these limits, future positions are compared for iteration steps of 20 seconds. These positions can be fan-shaped whenever a turn is detected. This turn is continued for each iteration step until a certain maximum limit has been reached. The figure shows a situation of two aircraft (the first flying North, the second West) where the westbound aircraft is making a left turn. The highlighted areas are deduced from the inaccuracy limits.
Figure 8: Horizontal fields describing future positions for two aircraft.
Figure 9: Separation distance set out against time.
The lines between the fields in Figure 8 give the minimum (worst case) distance between the uncertainty areas. In Figure 9 distances are plotted against time. A separation criteria states the valid separation before an alert shall be induced. In this case an alarm would yield in the interval
[T+56…T+108]. 2.3.3 SAABTech
SAAB Technologies has developed a probabilistic methodology based on recorded flight behaviour for identifying potential conflicts between aircraft [5]. This method can handle turning aircraft in a conflict. It is designed to be operational in both En-Route and the terminal monitoring area.
The STCA assigns horizontal and vertical templates to aircraft; these
templates contain predictions for future aircraft positions at each prediction cycle (5 second intervals) for the next two minutes. Prediction templates are assigned according to an aircraft’s previous track history. This allows for
more accurate track predictions than those generated by just considering current position and velocity. Different templates are used depending on the current turn rate and vertical speed.
Ten different prediction categories are used:
Default High, Default Low, En-Route, TMA4 High, TMA-Low,
Approach High, Approach Low, Departure High, Departure Low and Stack.
Each prediction category contains 3-6 different horizontal prediction templates and 4-5 different vertical prediction templates. These depend on the number of turn rate intervals and vertical speed intervals defined.
The prediction templates are developed off-line by analyzing a large sample of historic track recordings. These templates allow for a level of uncertainty in an aircraft's future position by predicting an area of airspace where an aircraft is likely to be, rather than its precise position.
Figure 10: A situation predicted to be hazardous.
Figure 11: Safe situation, fields describing different timeslots are allowed to interfere.
The template predictions are used as inputs in probabilistic alerting logic to identify the probability that aircraft infringe horizontal and vertical
separation criteria. Here the horizontal and the vertical separation values are tuning parameters for alerting all potential conflicts. Knowledge of aircraft position and status (e.g. that the aircraft is departing) is used for selecting alerting parameters. Specific rules are used for delaying alerts in situations with a high incidence of nuisance alerts.
2.3.4 Comments
All of these systems treat the horizontal situation separately from the
vertical. This is a natural approach since rules and surrounding systems use this way of representing the situation. This strategy also makes algorithms less complex and the systems easier to grasp.
NATS is the only system that uses single path trajectories. With single path trajectories CPA5 and time to CPA can be calculated with geometric math. This way of calculating does not require a large amount of computer power. Both NLR and SAAB use probabilistic trajectories with fields describing future positions. SAAB uses distributed fields6 and NLR uses homogenous fields. These are two different methods of describing probabilistic
trajectories. The NLR system appears easier to calculate while the SAAB system seems to describe the trajectory more accurately.
SAAB uses a method with templates to predict the future trajectory. This is different from the other systems. Instead of calculating the future trajectory and tune the probable behaviour, SAAB simply records traffic and saves the information into templates. Both of these methods may accomplish a similar result. We have no information showing one method to be superior to the other.
The following theory sections describe conformance, nominal and escape trajectory systems. All of the systems above use a nominal method to predict future trajectories. NATS system is the only system using an escape
trajectory model to ensure a way out of a possible conflict situation. All of the systems are based on a design having regions with separate parameters regulating system performance. This can be compared with the conformance method described in section nr 2.2.1 where an area is free of hazards.
The biggest differences between the three systems investigated are in the horizontal plane. The vertical situation is solved in a similar and
straightforward philosophy by all three systems.
5 Described in Section 4.1.1 on page 15. 6 Described in Section 4.2.1 on page 23
Table 1: Differences in modelling philosophies and trajectory modelling among investigated systems.
System Modelling philosophy Trajectory model
NATS Conformance, Nominal, Escape trajectory
Single path
NLR Conformance, Nominal Probabilistic
3 P
ROBLEM DESCRIPTION3.1 BACKGROUND
Airspace becomes more and more congested, and the need for surveillance systems supervising the airspace increases. The methods of how the
surveillance systems operate are regulated by international standardisation organizations like EUROCONTROL. STCA is a system that alerts if there is a suspected future loss of separation, this task is about STCA.
NOVA9000 is a product by Park Air Systems, which is an international company providing surveillance solutions for the world's airspace. NOVA9000 helps air traffic controllers to supervise the airport area and surroundings. It is a system under constant improvement and STCA is an important future component of this system.
This project is about studying the design of an STCA system and start
building a partial implementation. The study shall ensure that regulations are followed and wanted performance is fulfilled.
3.2 PROBLEM ANALYSIS
Initial studies have shown that the 3D-environment usually is divided into a vertical and a horizontal view, and separate conflict detection algorithms operate vertically and horizontally to detect conflict situations. It is also common to group all aircraft into pairs. Pairs representing all possible combinations of two aircraft are formed. Every pair is then run through a process to evaluate if the constituting two aircraft are or will be in a hazardous situation.
Vertically the system has to evaluate the probability of two aircraft being at the same altitude at the same time. To do this evaluation, information about altitude and change of altitude of all aircraft are available. Also information about allotted flight level and intended flight level might be available. Horizontally the system has to evaluate if two aircraft have or will have a small horizontal separation. According to initial studies most system use a trajectory model to foresee future positions and a future conflict. To
calculate the trajectories, information about the position, speed and heading of all aircraft is available. Complementing information about turn, intent heading and future route might be available. Depending on which trajectory
method that is used, different methods of evaluating the danger in a situation can be used.
When horizontal and vertical conflicts are detected the alert decision has to be taken. The time is a crucial factor, there has to be enough time to prevent an accident, and we do not want to alert too early and induce nuisance alerts.
3.3 OBJECTIVE
Produce a specification for an STCAmodule and implement a prototype for the NOVA9000 system.
Analysis
This project includes to analyse what has been written about aircraft conflict alert handling: research reports, articles and recommendations from
international standardisation organisations. Existing solutions from other companies are also important.
Specification
The specification should be constructed taking international standards into consideration, and has to be realizable with available resources. It describes, on an algorithm level, how the STCA module works. The specification shall describe a system following recommendations from EUROCONTROL. Prototype
With prototype we mean a partial implementation of the STCAmodule. The prototype is implemented in the NOVA9000 system and follows NOVA programming standards. The purpose with the prototype is to visualize the STCA functionality and gain knowledge about the STCA problematic.
4 A
PPROACHThe following section gives a description of our approach. A description of the linear prediction model is given, followed by a mathematical model. The rest of this section describes the prediction field model.
4.1 SINGLE PATH METHODS
4.1.1 Linear prediction model
Linear prediction model is a straightforward method of discovering a conflict between two aircraft in the horizontal plane. The idea with linear prediction is to make sure that the future distance between aircraft stays greater than a minimum value. This is accomplished with the assumption that all aircraft retain their velocity and heading for a period of time ahead.
This model calculates on a single path trajectory model. The trajectory is predicted as a single line straight ahead. Turns are not taken into account and calculated future positions generate problems if one of the aircraft is turning. Statistically, aircraft fly straight most of the time; this keeps down the
number of errors. There are methods to complement this system to handle changes of speed and direction. This section describes the algorithms of the fundamental straight line method [1].
Algorithm
Two aircraft are moving in different directions with a separation between each other. Information about velocity (vector) and position in a specific moment is input data for the system. This kinematical data is assumed to be retained for a period a time ahead.
Figure 12: Two aircraft with velocity vectors.
We describe this method by showing an example with two aircraft: A1 is positioned at (x1,y1) and travels with the speed vector (velx1,vely1)
A2 is positioned at (x2,y2) and travels with the speed vector (velx2,vely2). In this example A1 is flying north and A2 is flying northwest as shown in the figure above. From this situation we calculate current separation, relative speed and CPA (Closest Point of Approach7). This information gives us with further calculations the time it takes until CPA occurs and further how close the aircraft is at this moment.
We start with calculating the separation distance and the direction angle( ) (S ) s θ :
(
) (
)
) arctan( 1 2 1 2 2 1 2 2 1 2 x x y y y y x x S s − − = − + − = θBoth aircraft flies forward with great speed, but the speed we are interested in is the closing speed (how fast they approach each other). This speed is the relative speed and is marked V in the figure below. The picture below shows that A1 is closing A2 along the broken line, with the speed V.
Figure 13: Same situation as in Figure 12, but with relative velocity marked out.
When calculating the relative velocity we use the given velocities for both aircraft, in this case given in x- and y-coordinates. The size8 (V ) and direction (θv) of the relative velocity is calculated:
) arctan( ) ( ) ( 2 1 2 1 2 2 1 2 2 1 velx velx vely vely vely vely velx velx V v − − = − + − = θ
CPA occurs where the shortest distance between A2 and the extended
relative velocity (broken line) appears. This is demonstrated as a right angle between the broken line and the separation line to A2, as shown in Figure 13.
The separation between A2 and CPA is the shortest separation that occurs in the future between A1 and A2, assumed they retain speed and direction. This value can later be compared to parameters to evaluate if this situation is hazardous. The distance is the distance to the point where the separation is at a minimum, this distance is used to calculate the time until the
minimum separation occurs. Here are the distances to CPA calculated with trigonometry: 1 D ) sin( 2 ) cos( 1 2 1 v s v s S CPA A D S CPA A D θ θ θ θ − ⋅ = → = − ⋅ = → =
Time for A1 to arrive at Closest Point of Approach (TCPA) is calculated:
V D TCPA= 1
Expressions for TCPA and D2:
(
) (
)
(
) (
)
(
) (
)
sin(arctan( ) arctan( )) )) arctan( ) ( cos(arctan 2 1 2 1 1 2 1 2 2 1 2 2 1 2 2 2 2 1 2 2 1 2 1 2 1 1 2 1 2 2 1 2 2 1 2 velx velx vely vely x x y y y y x x D vely vely velx velx velx velx vely vely x x y y y y x x TCPA − − − − − ⋅ − + − = − + − − − − − − ⋅ − + − =Now, to make sure a dangerous situation not occurs, the system has to alert when the distance is shorter than the predefined safety separation, and at the same time when TCPA is less than safety time. Safety time has to be long enough for necessary actions to be performed in order to avoid a conflict.
2
D
4.1.2 Preserved acceleration model
To extend the linear prediction model our thought was to include the acceleration in the calculations as well. In this case acceleration means change of velocity. A turn is for example represented as acceleration sideways. The linear prediction model takes the position and velocity information from the current state and assumes velocity is retained in the future. This model additionally assumes that acceleration (turn and/or speed change) is retained. Calculations can be made for air routes and closest point of approach. This gives us the time to closest point of approach and alerts can be made based on this information.
Retained acceleration in this case means that the acceleration vector stays constant in relation to the travel direction (velocity direction). As shown in the picture the acceleration vector follows the velocity as the aircraft
progresses and the velocity vector changes. This means that the acceleration vector changes direction at the same time as the aircraft direction changes.
Figure 14: Acceleration from current state to future state.
To be able to determine CPA and TCPA we want to describe the track of one aircraft with retained acceleration as a function depending on time. Algorithm Euler’s method
The track is approximated in a created simulation program. The program calculates points along the track with Euler’s (stepping) method.
This algorithm starts at the point where the aircraft is positioned. From this point it takes a small time step into the future and calculates the new position of the plane with the help of information about current velocity and
acceleration. Mathematically it can be described:
2 2 1 t a t v s sn+ = n + n⋅∆ + n⋅∆
In this new position the velocity is changed depending on the acceleration. The new velocity is calculated depending on the velocity and acceleration from the previous position:
t a v
vn+1= n + n⋅∆
The acceleration vector changes with the direction of the velocity. The acceleration in the new position is the initial acceleration vector multiplied with the current direction of velocity.
1 1 1 + + + = ⋅ n n n v v a a
These calculations give us changed values of velocity and acceleration in the new position. The next step is to calculate a new position adjusting the
values of velocity and acceleration in the same way. Iterative repetition of this procedure generates the predicted path.
Mathematical model
Euler’s method gives an approximation of points along the predicted path. However a mathematical description of the path depending on the time is desired. Complex calculations, attached as appendix, resulted in two
functions: one describing the x-coordinate part of the position depending on time and one describing the y-coordinate part.
Here follows the resulting function determining the position depending on the time (t). All variables except t in the following functions are constants and are given. This means that the exact position, assumed conditions mentioned above, are returned by these functions. Explanations to variables and special cases such as at =0 are presented in the attachment.
( )
( )
+ ⋅ ⋅ + − + + ⋅ + + + ⋅ ⋅ ⋅ + + ⋅ = 1 4 sin cos 2 1 ln sin 1 ln cos 2 1 2 1 ) ( 2 0 0 0 0 2 2 0 2 0 n t n t t t n t t n n t n t t n x a a a a a v a t a a v a t a a a a a v a t a v t r θ θ θ θ( )
( )
+ ⋅ ⋅ − − + + ⋅ − + + ⋅ ⋅ ⋅ + + ⋅ = 1 4 cos sin 2 1 ln cos 1 ln sin 2 1 2 1 ) ( 2 0 0 0 0 2 2 0 2 0 n t n t t t n t t n n t n t t n y a a a a a v a t a a v a t a a a a a v a t a v t r θ θ θ θThese expressions state one aircraft’s position depending on the time. A subtraction between the position of one aircraft and the position of the other resulted in a huge expression describing the separation between the two aircraft depending on the time (t). To determine CPA the derivate of the expression had to be found. This operation was too complex for us to handle in the common case.
Later we realised that an expression describing CPA mathematically could not be used anyway; the given data is not precise enough to make use of the calculations. In the future more precise information about acceleration speed and position may make these calculations useful.
Comparisons between this mathematical model and Euler’s method were made in a simulation program. Figure 15 describes the two trajectories calculated with the Euler method using a small time step (0.001 sec), and one using the mathematical formulas above. As one can see the two
trajectories coincide and appear to be one trajectory. This indicates that the two different calculation methods give the same result, making it believable that the correct trajectory is calculated. To visualize that the Euler method is an approximation method we raised the Euler time step to 0.5 seconds, which diverged the two trajectories as shown in Figure 16.
Figure 15: Mathematical trajectory and Euler trajectory coincide.
4.2 PROBABILISTIC METHODS
Probabilistic methods predict future trajectory similar to single trajectory methods. The difference is that the future path of the aircraft is described as regions with high probability of containing an aircraft at specific moments. The positions of an aircraft are stated with fields.
The idea with this method is to portray the future traffic picture by stating each aircraft’s probable position at specific times. By processing this picture of future air traffic hazardous situations can be found and alerted for.
4.2.1 Portray trajectories
A trajectory using this method is a sequence of fields stating the future predicted positions of an aircraft. This is needed to judge if there is a future conflict situation or not. Below are two different methods of portraying a trajectory.
Homogenous fields
One method of representing future positions is to state homogenous fields. These fields or areas represent the future position homogenously distributed over the surface. A field states the position of an aircraft at a certain related moment. One way of determining a field is to calculate the future position based on current position and speed. Around the calculated position a field of uncertainty can be placed. The shape and size of these fields have to be tuned and evaluated to fit alert criteria.
The probability of a conflict can be estimated by measuring the distance between the fields that represents future positions. If two fields, representing the same timeslot for two different aircraft, are placed close to each other the probability for a conflict in this timeslot is high.
NLR seems to use homogenous fields in their system [2], as described in section 2.3.2.
Distributed fields
Instead of using homogenous fields, the future position can be described as a distribution. This can be done with the Kalman filter method.
Target i Target j Predicted uncertainty [xi,yi] [xj,yj] Predicted uncertainty Kalman filter uses the state
vector to calculate the future position. The position is calculated as a field of predicted uncertainty. The picture to the right shows two targets moving in the
direction of the lines. The ellipses at the end of the lines state the future positions with
predicted uncertainty9. Figure 17: Two targets with predicted uncertainties To distinguish whether a
dangerous situation or not is present, the separation between these future positions is needed. A vector stating the predicted separation is calculated by subtracting the centres of the distributed positions. The uncertainty in distance is
calculated from the two predicted
uncertainties in figure 17 into a combined predicted separation uncertainty. The resulting vector and separation uncertainty is shown in figure 18.
Predicted separation
Target i - Target j
Predicted separation uncertainty dh,ij
Figure 18: Predicted separation with uncertainty.
SAAB Tech indicates they use distributed fields in their system [10]. 4.2.2 Situation connected behaviour
Another method is to record air traffic in certain situations. This method predicts the future positions of an aircraft based on previous traffic in the same situation. Air traffic is recorded and categorised into different
behaviours, for example turning or flying straight ahead. The behaviour for every category is recognised and stored in a template. The template is a
pattern for the usual behaviour in this situation. The template can be saved as homogenous fields or as distributed fields.
Region connected behaviour
Certain behaviour can be connected to certain physical areas (regions). By making rules for behaviour in these areas a more precise trajectory
prediction can be produced. If for example an aircraft is positioned in the approach area and flies in the direction of the runway a descent manoeuvre may be predicted.
Recent behaviour connected behaviour
Behaviour can also be linked to previous behaviour, for example a stacking aircraft (an aircraft circling above an airport waiting for a landing signal) is circling and has a high probability of continuing circling.
Input connected behaviour
There can be many different sources of input affecting the behaviour. Position, speed and turn rate are obvious information that can be used to predict future position. Other information like flight plan information can be linked to a certain behaviour, for example if an aircraft is flying En-Route following flight plan it is likely to continue with a straight forward, En-Route behaviour.
5 S
PECIFICATIONThis specification describes an STCA system, which is designed to work as a module in the commercial air traffic surveillance system NOVA9000.
5.1 SYSTEM OVERVIEW
STCA is a safety system constructed to alert air traffic controllers if there is a present or a future conflict within two minutes. A conflict is a loss of separation, which means a simultaneous violation of both the horizontal and vertical separation standards.
This system is designed to induce all relevant alerts with sufficient time for involved parts to act and avoid dangerous situations. It is also relevant to keep nuisance false alerts to a minimum.
STCA is part of the Safety Net (SNET) concept. 5.2 FUNCTIONAL REQUIREMENTS
These requirements consider the functionality of STCA. The requirements are constructed following recommendations from EUROCONTROL [8]. Interface (input) and HMI (presentation) are not considered in these requirements.
Alerting performance
F1 STCA shall provide the ATC (NOVA9000) with alert data for all relevant conflicts.
F2 The number of nuisance alerts produced by STCA shall be kept to a practical minimum.
F3 Where sufficient time exists, STCA should attempt to confirm that an alert is operationally necessary before notifying ATC.
Warning time
F4 The warning time shall be sufficient for all necessary steps to be taken from the controller recognizing the alert to the aircraft successfully executing an appropriate manoeuvre.
F5 Warning time shall be kept to a practical minimum given that it does not contribute to the generation of excessive nuisance alerts.
Regions
F6 STCA shall make use of regions when determining whether or not an alert is to be given for specific surveillance data.
F7 Horizontal coordinates should be defined using a reference system, which is convenient and compatible with other aspects of the ATC system.
F8 Regions should be permitted to overlap or totally enclose other regions. F9 It should be possible to assign priorities to individual regions to
determine which region is used when a position falls within more than one region.
Intention data
F10 STCA should not use controller-input Clear Flight Levels (CFL) as a basis for predicted vertical flight profile.
F11 STCA may use controller-input Clear Flight Level (CFL) to supplement surveillance data in determining whether an aircraft is in the process of levelling-off.
F12 STCA may use aircraft data from the Flight Management System (FMS) when this becomes available.
STCA Capabilities
F13 STCA shall detect conflicts based on current proximity.
F14 STCA shall detect conflicts based on predicted straight-line tracks. F15 STCA shall detect conflicts based on predicted turning tracks. F16 STCA shall inhibit alerts caused by spurious track data (e.g.
reflections10, split tracks11, coasted tracks12).
10 Radar data can give an incorrect picture of a situation due to reflections.
11 Meaning that one target is identified by the system as two separated targets with separated tracks. 12 A lost target is predicted into the future based on previous information. This is called coasting and may
F17 It shall be possible to adapt the performance of STCA dependent upon the region type.
F18 STCA shall include a mechanism to determine appropriate criteria to be used for conflicts where conflicting aircraft are in different regions 5.3 SYSTEM-WIDE DESIGN DECISIONS
5.3.1 Provided data
All input into the STCA module is distributed by NOVA9000. Information about all air targets is gained periodically (every 4th second) from the NOVA distributed database.
Following information is provided for every target: • Identifier - a number that identifies each aircraft.
• Horizontal position vector - current position of an aircraft. • Ground speed - the horizontal speed.
• Heading - the horizontal heading.
• Vertical altitude - the altitude of an aircraft.
• Vertical rate - (climb/descent), states the change in altitude with a positive or negative value.
• Additional heading information - (not for every target) 13 - information about roll angle, magnetic heading and track angle can give a better picture of future heading.
• Future altitude (not for every target) 11 - information about assigned flight level by air traffic controller, and selected altitude by aircraft, helps predicting future altitude.
5.3.2 Design choices
EUROCONTROL recommendations
We have chosen to follow recommendations stated by EUROCONTROL [8], which is an important standardisation organisation. By following their recommendations the system should perform what the customers expect of an STCA system. Following EUROCONTROL’s recommendations is often a requirement from customers.
No conflict resolution
This STCA module shall not include conflict resolution. TCAS is an alert system communicating with the pilot. This system suggests an action when a hazardous situation is discovered, whereas STCA does not.
Horizontal and vertical conflict detection
Splitting the 3-dimensional environment into vertical and horizontal conflict detection seems a natural choice for many reasons. All commercial systems investigated have chosen this way of representing the problem. The
surrounding surveillance system is built on vertical and horizontal parts. Even the information is gained in separate methods, horizontal data by radar and vertical data by air pressure. The world of flight surveillance is thinking horizontal and vertical.
Probabilistic trajectories
We chose to use a probabilistic trajectory model; this gives a natural
description of the future positions of an aircraft including uncertainty. More information about the future trajectories gives us more information to base our alert decisions on. This method also gives a visual picture to the
controller so he or she can easily understand why an alert is induced. It is also easy to test and tune the system and make it act as expected.
In order to estimate future states the dynamic model has to make some assumption about future behaviour of the aircraft. The trajectories shall be calculated based on current speed and heading taking turn and intent information into account. A turn indication shall result in a trajectory following the turn.
Distributed fields
We use distributed fields to describe future positions. The distributed fields include information about the probability of being at a certain place. This gives us a more exact positioning of an aircraft. Assuming this information is reliable, a better alert decision can be made.
Regions
The use of regions is recommended by EUROCONTROL as well as we find it practical. Certain regions are heavily trafficked and are not suitable for STCA surveillance while others are dangerous and need unique parameters. We chose to use separate parameters in different regions making the system flexible and suitable for the task.
5.4 SYSTEM ARCHITECTURAL DESIGN
STCA entry Extract all required data from TDB and update Target descriptor table Target data base Return Initialization block:
Construct for each target a descriptor record selection of targets. Flightplan data Radar data Meteorological data Calculate trajectories for all
aircrafts
Conflict search Trajectory prediction:
Extrapolate all target positions to the probed time.
Conflict search:
Group aircrafts into pairs and run through vertical and horizontal conflict detection
Target info
Alert info
STCA
5.4.1 Initialization block
Initialization block is triggered once every program invocation and it acts as an interface to the NOVA9000 target database (TDB). From this it retrieves all the data items necessary for trajectory calculation and conflict search. Initialization routine creates a target descriptor record for every target taken into consideration by STCA. A target descriptor record contains all the vital information as extracted from NOVA9000 database and data from the trajectory prediction routines.
The total set of target descriptor records is stored in the target table, which is updated every program cycle. The target table represents a snapshot of the traffic situation and acts as the interface between the trajectory prediction and the conflict search routines.
A target descriptor is created for all tracked targets; coarse filtering can be inserted if required. Coarse filtering eliminates on an early stage targets obviously out of STCA conflict, for instance an aircraft far away from other traffic.
Predict position for aircraft n at time t
t=max? t=t+timestep
t=0
Find next aircraft (n)
Processed all targets? Start calculate all
trajetories
Stop calculate all trajetories no yes no yes 5.4.2 Trajectory prediction
The trajectory prediction block calculates the trajectories for all targets given from the Target database. Figure 20 shows the process where all targets are looped through and positions are predicted iteratively for every target with a certain time step. This process results in predicted positions for every target “max time” seconds into the future with “time step” seconds between every prediction.
Each extrapolated position is described as a distributed field. A distributed field can describe probability of containing the aircraft a certain moment. Kalman filter equations are used to state this distribution as described in section 4.2.1.
Kalman equations to predict the trajectories. Information about a current turn makes the trajectory follow a turning movement. If there is intent
information available (Mode S) from the aircraft an even more precise trajectory can be predicted.
Using flight plan information
Information gained from the flight plan such as next passing point or
assigned flight level is vital information when predicting the future position. With this information a more precise direction can be estimated. Notice that passing point information is used in the horizontal conflict calculations and flight level information in the vertical conflict calculations.
Information about common behaviour
In some areas certain flight behaviour is more common. For example when aircraft approach the runway of an airport it is common that they follow an air “road”. Information about common behaviour in certain places in the air can also be used to make better estimation of future positions.
5.4.3 Conflict search algorithm
Horizontal conflict search
Vertical conflict search
Alert decision module Group aircrafts into pairs All pairs evaluated? Conflict detection start
Find next pair
Coarse filtering process
Alert
no
yes
The conflict detection routine is composed of one horizontal and one vertical conflict search r Results from these are processe the alert decision module, whic determines whether to induce an alert or not.
outine. d in h
The search routine starts with grouping all aircraft into pairs. Every aircraft shall pair together with all other aircraft. Here
follows an expression determining the number of pairs:
Number of pairs 2 ) 1 ( * − = N N w N = number of aircraft. here
After grouping, every pair goes through a process where it is evaluated if there exists a dangerous situation
between the two aircraft constituting
the pair. Figure 21: Block diagram of conflict search. This process consists of a coarse filtering process, followed by a horizontal and a vertical conflict search as shown in Figure 21. The figure indicates that all aircrafts are run through both vertical and horizontal conflict search
routines. To reduce the calculation load, pairs passing safely through the vertical conflict search can be excluded from the horizontal conflict search. A dangerous situation is only present if both vertical and horizontal conflict searches have predicted a possible conflict at the same time.
A record is created for every pair, and all records are stored in a table. These records contain information about the two aircraft included in the pair and horizontal and vertical conflict search routines stores conflict information in these records. This conflict information includes the estimated time of
conflicts.
Coarse filtering
To reduce the calculation load two coarse filters are setup to take away pairs where the included aircraft obviously are in no conflict with each other. The vertical coarse filter eliminates pairs, which obviously do not contain a vertical violation, for example one aircraft on the ground and one at an altitude of 5000 meters. The horizontal coarse filter eliminates pairs from further calculations if the two aircraft are a great horizontal distance to each other. Further reduction of the number of pairs may be possible. Testing and tuning the system decides how much reduction is necessary.
Horizontal conflict search
The horizontal conflict search is using the trajectories created in the
initialization block. The trajectories for two different aircraft are compared and the probability of a dangerous situation to occur is evaluated.
Conflict search method deals with probabilities. The probability of two aircraft being at the same place at the same time is calculated, and it is used for deciding whether a situation is dangerous or not. This is done using the distributed fields in the probabilistic trajectories stating future positions of aircraft. If the probability exceeds a threshold value the conflict search algorithm induces an STCA alert.
Kalman filter equations are used to calculate the probability of two aircraft being at the same place at the same time. In our prototype we use
homogenous fields instead to make it easy to explain and overview. If a dangerous situation is discovered, information about the time and
probability of the conflict is stored in the pair record table. The alert decision module later uses this information.
Vertical conflict search
Our approach for the vertical conflict detections is straightforward. Find the starting time and stopping time of a vertical conflict based on current
altitude information, ascend/descent information and information from the flight plan.
Aircraft are flying at allotted flight levels. If an aircraft sticks to its flight level other aircraft can fly above or below on other flight levels without having a conflict. Information about assigned flight level is included in the flight plan. This situation is easy to handle, if two aircraft both fly on their own flight level there is no vertical conflict. If they fly on the same flight level there is a vertical conflict (not necessary a horizontal conflict which is also needed for an alert).
When aircraft leave their assigned flight level a more complex situation is present. This situation arises if assigned flight levels do not match actual altitude or an ascent/descent situation is detected.
In this case an upper and a lower bound, a vertical safety zone, is calculated for every aircraft. This safety zone follows the aircraft’s
descending/ascending movement, and foresees a level off at an assigned flight level.
To determine if there is a vertical conflict, bounds are compared to see if they interact. If two aircraft have their own undisturbed safety zone there is no vertical conflict between them.
Figure 22: Vertical conflict between two aircraft with interacting zones.
Time is a crucial factor in these decisions. If for example the safety boundaries of two aircraft are crossing in 90 seconds it might not be necessary to make an alert right now. Therefore along with the conflict information from this part of the system start-time and end-time of predicted conflicts are stored in the pair record table.
Alert decision module
At the end of the evaluation process the alert decision module decides, for every pair, whether a dangerous situation exists or not. Information about the horizontal and vertical conflicts is gained from the pair record table.
To evaluate a conflict, the time of an expected situation is crucial. An alert is only induced if a conflict situation is discovered by both conflict search routines (horizontal and vertical) and the duration of conflicts coincide. For example the horizontal search routine may have found that a pair enters a dangerous situation 80 seconds from now and an alert is only induced if the vertical search routine also has a conflict detected at the same time.
The alert is not required to be induced at the first time a probable conflict is detected. If there is more than enough time to avoid the alert it may be possible to delay. The positive effect with delaying an alert is that it might not be needed to induce the alert at all, and if we delay the alert we might avoid a nuisance alert.
There are many pair situations that have a small conflict probability. To alert in all possibly dangerous situations does not make a good system. Decisions are taken and alerts balanced so that a minimal amount of nuisance alerts and all crucial alerts are induced.
The region where a future conflict is situated matters as well. This module uses different parameters for different regions to evaluate if an alert shall be induced.
This module also has functionality for inducing unwanted alerts caused by spurious track data. Here follows some spurious track data problems that have to be taken into account:
• Split tracks, when one aircraft is detected as two different aircraft close to each other, easily causes unwanted alerts.
• Reflections from objects that are not aircraft can also lead to alert problems.
• If an aircraft is lost as a target, surveillance systems often predict a track where the aircraft should have been. The predicted track
(coasted) is based on previous information about speed and position and may lead to unwanted alerts.
These problems are a result of problems with the tracking system. The multi-sensor tracker (MST) of the NOVA9000 system shall take care of these problems before STCA gets in contact with the information. If a problem situation should appear anyway, functionality to inhibit these alerts is implemented in the alert decision module.
How these decisions are made and what limits to use need to be tuned. The system has to be tested and tuned with real traffic to make sure that the system alerts in the right situations. Section 6.3.5 describes how to perform such tuning.
5.5 REQUIREMENTS TRACEABILITY
This section describes how the requirements are fulfilled. Some demands are handled by a specific part of the system while others are depending on the whole system.
The last instance delivering the alert to the ATC system is the alert decision module [F1]14. This module makes the final alert decisions and tries to keep the number of nuisance alerts to a practical minimum [F2]. The alert
decision module also attempts to confirm that an alert is operationally necessary before notifying NOVA9000 [F3].
The warning time requirements are dependent on the trajectories and the parameters set by the tuning procedure. For the warning time to be sufficient it is important that the aircraft follow the predicted trajectories and that the separation criteria is not too small [F4]. The testing and tuning procedure also affects the number of nuisance alerts and tries keeping them to a minimum [F5].
The alert decision module mostly handles the region requirements. Regions are created [F7, F8, F9] and priority stamped in other parts of the
NOVA9000 system. The alert decision module uses the region information to decide whether to induce an alert or not [F6].
The intention data is used to make the trajectory prediction. The vertical conflict detection module uses the CFL information to predict the altitude of an aircraft; this information shall be used according to the requirements [F10, F11]. Information from the FMS helps horizontal trajectory predictions with actual aircraft intention data [F12].
The requirements on the alert capabilities depend on the complete system. The alert decision module bases information from the horizontal and vertical decision modules to induce alerts. Both modules shall detect current
proximity situations [F13], straight line situations [F14] and the horizontal decision module shall take turning track situations into consideration [F15]. The alert decision module shall inhibit alerts caused by spurious track data [F16] and handle the region problematic [F17, F18].
