are referred to the aircraft's center of gravity. ..... The airspace used for this study is the ICAO APAC region, namely Australian Airspace ... Call Sign:QFA405.
The Artificial Life and Adaptive Robotics Laboratory ALAR Technical Report Series
ATOMS: Air Traffic Operations and Management Simulator Sameer Alam, Hussein A. Abbass and Michael Barlow TR-ALAR-200611014
The Artificial Life and Adaptive Robotics Laboratory School of Information Technology and Electrical Engineering University of New South Wales Northcott Drive, Campbell, Canberra, ACT 2600 Australia Tel: +62 2 6268 8158 Fax:+61 2 6268 8581
ATOMS: Air Traffic Operations and Management Simulator Sameer Alam, Hussein A. Abbass and Michael Barlow School of Info. Tech. and Electr. Eng. Univ. College, Univ. of New South Wales ADFA, Canberra ACT 2600 Australia Email: {s.alam,h.abbass,spike}@adfa.edu.au
Abstract In this paper we introduce ATOMS (Air Traffic Operations & Management Simulator) which is an air traffic and airspace modeling and simulation system for the analysis of Free Flight concepts. This paper describes the design, architecture, functionality and applications of ATOMS. It is an intent based simulator which discretizes the airspace in equal sized hyper rectangular cells to maintain intent reference points. It can simulate end to end airspace operations and air navigation procedures for conventional air traffic as well as for Free Flight. Atmospheric and wind data modeled in ATOMS results in accurate trajectory predictions. ATOMS uses multi-agent based modeling paradigm for modular design and easy integration of various air traffic sub systems. A variety of advanced air traffic management (ATM) concepts envisioned in Free Flight are prototyped in ATOMS including airborne separation assurance (ASA), cockpit display of traffic information (CDTI), weather avoidance and decision support systems (DSS). Experimental results indicates that advanced ATM concepts makes a sound case for Free Flight, however there is a need to investigate and understand their complex interaction under non-nominal scenarios.
I. I NTRODUCTION Present day air traffic system is reaching it’s operational limits [1] and accommodating future air traffic growth will be a challenging task for aviation industry. Radical design changes in the present day air traffic control system are required to meet this challenge. Free Flight [2] is one such initiative actively pursued by government and industry alike, where pilots will have an active role in airborne separation and real time route planning. However there are several safety challenges and design issues [3] to be addressed before Free Flight can become a reality. Advanced air traffic management (ATM) concepts envisioned in Free Flight must be investigated and evaluated before they can be operationally tested [4], without which it can be a risky and expensive exercise. A wide range of ATM systems like Total Airspace and Airport Modeler (TAAM), Operational and Supportability Implementation System (OASIS), Airport and Airspace Simulation Model (SIMMOD) etc. have been developed by industry and government organizations, however they are designed around specific requirements of air traffic service providers. Further a NASA research found them to be inadequate for modeling future ATM concepts and lacking in several capabilities [5]. Based on the projected requirements for Free Flight [3], [6], we established a program to investigate future ATM concepts in Australian Airspace. An air traffic simulation environment, which can provide flexibility for rapid prototyping and evaluation of Free Flight concepts was determined as a key requirement for this purpose. Air Traffic Operations and Management Simulator (ATOMS) described here is developed to meet these objectives and to investigate future ATM concepts on safety and economic indicators. Systems with similar objectives have ´ been proposed in the literature for e.g., the Complete Air Traffic Simulator (CATS) [7] at Center dEtudes de la Navigation Adrienne (CENA), France, Future ATM Concepts Evaluation Tool (FACET) [4] at NASA Ames U.S., and the NLR Air Traffic Control Research Simulator (NARSIM) [8] at the National Aerospace Laboratory (NLR) Netherlands. However they have few limitations such as, FACET can only simulate high altitude flights and cannot model departures and arrivals, CATS do not have access to atmospheric data and NARSIM is a state based
simulator. In this paper we first provide an overview of ATOMS, followed by an outline of its design features and core capabilities and some validations results, then we discuss the implementation of the Free Flight concepts modeled in ATOMS and their evaluation. II. OVERVIEW
OF
ATOMS
The basic idea behind ATOMS is to develop a simulation and modeling environment where Free Flight concepts can be rapidly prototyped and evaluated. The main simulation engine of ATOMS models the basic air traffic and navigation features (e.g. airspace, waypoints, navigational fixes, airways, atmosphere, aircraft and trajectory generation) that are essential for evaluation of any air traffic concept. All other modules which implements present and future air traffic concepts are build around the core engine. ATOMS models enroute airspace over a contiguous region, defined by latitude and longitude pair through a graphical user interface. It uses a variety of worldwide standard databases such as: Official Airlines Guide (OAG) [9] for the flight schedules, Jeppessen Navdata for the airspace structure database, BADA [10] for the aircraft performance database and the World Area Forecast System (WAFS) [11] data structure for the atmospheric data representation. It uses a local-level right handed orthogonal earth-fixed geographic coordinate system where the x-axis points east, the y-axis points north and the z-axis points up. Since the axes of the coordinate system rotate with the earth, therefore to account for the rotation of the coordinate frame the law of Coriolis [12] is used for accurate trajectory prediction. ATOMS utilizes the Lambert conformal conical (LCC) [13] projection for displaying airspace features and air traffic movements on the graphical user interface. The LCC projection is chosen because the shapes of small areas are maintained and the amount of distortion is minimal along lines of tangency. To model four dimensional (4D) aircraft trajectory, ATOMS uses the equations of motion to calculate the velocity and position data of the aircraft over a predefined time interval in a geographic coordinate frame. Distance and track calculations are based on great circle route (GCR) [13] calculations in an elliptical earth model. The motion equations are first order integrated to compute the new aircraft position at the end of the discrete time interval. The effect of the earth’s rotation and the wind are included in the equations of motion. Aircraft performance parameters are obtained through BADA files licensed from EUROCONTROL 2 . ATOMS can be run in real time, fast time or slow time with various other options described in the subsequent sections. III. A RCHITECTURE
AND
O PERATIONAL F EATURES
OF
ATOMS
This section discusses design principals of ATOMS, followed by its operational features such as, aircraft modeling, atmosphere modeling, trajectory generation, airspace modeling and it’s graphical user interface. A. Architecture We adopted an agent-based approach to model and simulate the various entities of ATOMS. This approach helped us in the decomposition of agents (e.g. Aircraft agent) as shown in figure 1, into activities based on logical decomposition of roles (communication, navigation and surveillance) and interactions (conflict detection, weather avoidance etc.) with the environment. The interactions between the agents is implemented using a 4D discretized airspace data structure where each agent have access to state information of other agents within a predefine area. It is assumed that each aircraft broadcasts its current state and intent information via a data link and perfect information is available to all other aircraft within the broadcast range. The simulation model is discrete-event and assumes a perfect world where all the necessary data is available without any errors, whereas in real air traffic scenario the data is highly suspectable to corruption due to to electronic interference and atmospheric conditions. The software is written using the C++ c programming language and the GUI is written in Microsoft Visual C++ version 6.0. No third party softwares or libraries were used. This gave us the required flexibility for extendability and full functional control on the design of various modules. 2
Eurocontrol is the European organization for the safety of air navigation. It currently numbers 31 member states. Eurocontrol has as its primary objective the development of a seamless, pan-European air traffic management (ATM) system
Fig. 1.
A High level system flow char
ATOMS is a four dimensional simulator, i.e. it works with a latitude-longitude pair for each flight with an associated altitude and time for each of the these pairs. For Free Flight the airspace is considered in 3D which is approximated with grid cells in hyper-rectangular discrete space as shown in figure 2. The size of these 3D grid cells is based on the current set of separation standards by the ICAO 1 [14]. In Free Flight the aircraft trajectory is generated as a series of points in the airspace using these cells as reference. These cells also acts as a repository of the airspace containing information such as: weather, atmospheric properties, intent information of aircrafts etc. The fourth dimension is added to these cells by assigning them estimated time of arrival for each aircraft passing through then in the near or immediate future.
Fig. 2. Airspace modeled as hyper-rectangular discrete space. Block 1 is (1,1,1) in (i,j,k) coordinates, block 64 is (4,4,4) in (i,j,k) coordinates. On right side the labeling scheme in a 2D representation can be seen.
Looking at a 2-D version of an airspace cell block each cell can be defined by its start (latitude, longitude) and end (latitude, longitude). Using these two points, the corner co-ordinates for each cell in the whole airspace can be calculated and thus the extremes of the airspace can be defined using the geodetic earth co-ordinate system. This enables us to uniquely identify each block of the discretized airspace by (i,j,k) coordinates as shown in figure 2. Such a discretization of airspace helps maintain reference points in the space in the absence of waypoints and airways for trajectory simulation and intent information propagation, secondly it limits the airspace search volume for conflict detection, third it helps in identifying future congestions pockets in the Free Flight airspace (not discussed in this 1 International Civil Aviation Organization: ICAO was established in 1947 to develop principles and techniques of international air navigation and to support planning and development of international flight transportation
paper). ATOMS uses the flight plan (origin-waypoints-destination) when simulating the present air traffic system and uses (origin-destination) pairs when simulating the Free Flight. Flight plans are loaded for each flight from an actual flight data management system (FDMS) database provided by Air Services Australia. Flight plans are then used to generate the arrival and departure schedules for the airports in the airspace under consideration or the arrival times at the fixes on the boundary of the airspace. Then using each flight plan, a flight trajectory is developed using the aircraft performance parameters, the navigational data and the atmospheric data from the relevant database files. The simulator triggers flights based on their estimated time of departure (ETD) from airports within the designated airspace and for overflying flights, they are activated based on the estimated time of arrival (ETA) at a boundary fix. The flight profiling is done for each flight to determine the top of climb (TOC) and top of decent (TOD), these points are then inserted in the flight route generated by the flight management system (FMS). A conceptual representation of simulator modules and data flow is shown in figure 3. A simulated time clock is used to step the discrete event simulator through the sequence of time units using a fixed increment time advance. The simulation clock is advanced dT units and the aircrafts are considered to move at the end of the dT which is set at one second (real time mode) as default. However, dT can be changed for slow time and fast time modes using the GUI interface.
Fig. 3.
A conceptual representation of ATOMS’s core modules.
B. Airspace Modeling ATOMS uses airspace data (airways, waypoints, airports, navigational fixes, arrival fixes, holding points) from the Designated Airspace Handbook by Air Services Australia [15] which provide the basis for the airspace modeling. The data is pre-processed to remove unwanted fields and records. The airspace is built as a set of defined points. These points are loaded from the airspace definition file and the airspace volume is built. The waypoints that are loaded are within the airspace and those that are immediately next to the boundary of the airspace under consideration. Airports and airspace pre-boundary fixes are simulated as queues. These queues are used to simulate airport and airspace capacity constraints by imposing time restrictions on the soon to be active flights. For an airport, when a flight departs from it based on its ETD, it is put under busy status for the next 25 seconds, so if there is any other flight scheduled for departure in the next 25 seconds, then it will have to update its ETD accordingly and have its flight profile regenerated. This ensures necessary spacing between departing aircraft as well as safety from wake turbulence [16]. Similarly for an overflying flight when it becomes active at airspace boundary waypoints based on their ETA and flight level, the waypoint for that flight level is put under busy status for the next 25 seconds. If there is another flight which needs to be activated at that waypoint at the same flight level then it will search the next two lower flight levels for availability, if it finds a flight level available it gets activated else it is put on a hold pattern until the flight level is available.
C. Atmospheric Modeling Since aircraft motion, altitude and speed measurements are highly dependent on temperature and pressure, modeling of these parameters is necessary for accurate prediction of aircraft trajectory. ATOMS uses the ICAO standard atmosphere (ISA) [17] equations for atmospheric modeling, where atmosphere is considered as a perfect gas in equilibrium on a non-rotating flat earth. A temperature profile is generated using the standard temperature lapse rate equations [18]. The temperature is calculated at a series of predetermined altitudes and it is assumed that there is a linear temperature variation between these levels as shown later in the validation section. Wind data is obtained from the Bureau of Metrology which is then profiled by interpolating two known wind quantities at two different points and across altitudes to ensure no discontinuity in the wind calculation. The relative wind is assumed to apply to all points of the aircraft. Given the aircraft’s location (latitude/longitude) and altitude and interpolating between the nearest grid points and levels respectively (assuming linear characteristics of the atmosphere between these points in space) a representative value for wind aloft, temperature and density is calculated. All these atmospheric elements are set up in the airspace data structure that contains the temperature, northerly/easterly wind component and the pressure at the given altitude. D. Air Traffic Air Traffic data is provided by Air Services Australia which is obtained from the Aeronautical Fixed Telecommunications Network (AFTN) activity log. The point of activation for a flight is set to the airport of origin if it is in the airspace or it is set to the first airspace boundary waypoint if the airport of origin is outside the airspace. Similarly the point of deactivation for a flight is either the destination airport if it is within the airspace or the last waypoint point on the airspace boundary if its destination airport is outside the airspace. E. Aircraft Modeling A single point mass model is used for aircraft modeling where all the forces, moments and velocity vectors are referred to the aircraft’s center of gravity. Since our research doesn’t deal with the effects of aircraft rotation, wing down-wash, wind gradients etc., a multi point aircraft model is not necessary. The aircraft parameter values are based on the BADA database and accurately reflect the physical and operational limitations of the concerned aircraft. In addition, they provide a realistic estimate of the fuel consumed during the flight simulation.
Fig. 4.
Parametric model of aircraft and its interaction with other modules to generated flight profile.
The simulated aircraft navigates from point to point based on a predefined flight plan or free flight trajectory. The aircraft’s FMS algorithm plan 4D - trajectories (x,y,z,t), determine the respective control inputs for the aircraft to follow simulated trajectories and model the dynamic aircraft reaction to these control inputs as shown in figure 4. The aircraft state and position is computed at a sampling rate of 1 second. The subsequent integration of the aircraft state results in the desired flight profile. The aircraft class contains four key modules: • Flight Module: Contains flight plan and generic data about flight. • Flight Management System (FMS): Generates 4 dimensional flight profile from the flight plan data.
AutoPilot: Controls the navigation and steering of the aircraft based on the flight profile generated by FMS or from the navigation commands generated by Command Generator. • Navigation Command Generator: Generates appropriate navigation commands based on the flight profile and sends to autopilot for execution. Further for computation of aircraft performance parameters from the BADA database, the phase of flight is required and is defined as shown in figure 5. These flight phases are updated by the auto-pilot module based on altitude and flight profile data. For the GUI appropriate color codes (blue-take off & climb, green-cruise, grey-descent, red-approach & land) are assigned, which makes it easy to understand which phase an aircraft in at a given point of time. •
Fig. 5. Phases of flight at various altitude as modeled in ATOMS. A flight having origin-destination within the airspace undergos all the phases.
The following equations govern the motion of an aircraft in geographic reference frame [19]. Z φ = φ0 + (dφ/dt)dt λ = λ0 + h = h0 +
Z Z
(1)
(dλ/dt)dt
(2)
(dh/dt)dt
(3)
Where φ is the latitude, λ is the longitude and h is the altitude. We used first order Euler integration method which gives a good approximation for the purposes of our research. Additionally, any positional errors that are introduced are reduced to a minimum by continuously re-computing the heading. The time interval (1 second) for integration is small enough to yield approximations that are appropriate to this research and are at the same time not too coarse to allow other research issues discussed later, such as conflict detection to go unnoticed. The updated position dynamics can be obtained as follows: Z PN +1 = PN + VN (0)dt (4) PE+1 = PE +
Z
VE (0)dt
(5)
Where PE is the easterly position and PN is the northerly position respectively. VN (0) and VE (0) is the initial northerly and easterly velocity components. Therefore, knowing the variables V , PN , PE and ψ (bearing from North), PN +1 and PE+1 can be calculated. Similarly for altitude calculations, assume that the climb/descent rate is given by VF , then the new altitude can be calculated using the following formula Z HN +1 = HN + VF (0)dt (6)
Where HN +1 , HN , are the new and current altitudes(m) and VF is the vertical rate of climb/descent as calculated for the respective aircraft from the BADA database. The equation of navigation of aircraft are developed as follows, for an aircraft flying at an altitude h above the surface of the earth let the vehicle position is defined by the Latitude φ and the Longitude λ. The radius of curvature at a meridian at any latitude is given by Rλ = Ro (1 − eE 2 sin2 λ)3/2
(7)
So that the translation along a longitude can be described by dλ/dt = VN /(Rλ + h)
(8)
and the translation along a latitude can be described by dφ/dt = VE /(Rφ + h)
(9)
where Rφ = Ro (1 − eE 2 sin2 λ)1/2 . VN is the velocity component in the north direction and = Vgnd ∗ cosψ . VE is the velocity component in the east direction and = Vgnd ∗ sinψ . Vgnd is the aircraft ground speed where Vgnd = (VN 2 + VE 2 )1/2 . ψ is the aircraft’s bearing from true North. Ro is the equatorial radius of the earth (3444nmi). Rψ is the meridonial radius of curvature for north-south motion. Rλ is the normal radius of curvature for east-west motion. eE 2 = 6.7X(10)−3 (Earth eccentricity squared). F. Simulated Flight Management System (FMS) and the Autopilot The flight management system of ATOMS computes flight profiles with pre-specified arrival times at the destination airport or deactivation point in the airspace in four dimensional (4D) navigation. Given a flight plan, the FMS computes the path from one airport to another and then the autopilot flies the aircraft along that path. To achieve this, the algorithm selects the most economic speed schedules for each phase of the flight (take-off,climb, cruise, descent and approach), predicts the complex vertical and horizontal profile that the aircraft would fly and when connected with the autopilot of the aircraft, controls the aircraft along the three dimensional flight plan. For the time dimension the FMS computes the speed schedules and the flight path based on a required time of arrival at a selected point along the flight path. Top of Climb (TOC) and Top of Descent (TOD) are calculated by forward integrating from the origin and reverse integrating from the destination respectively using the aircraft equations of motion and the appropriate aircraft model. These two points are used to highlight the start and end of the cruise phase. To achieve the necessary sequencing and separation in a 4D environment, FMS provides autopilot necessary heading vectors, speed constraints, and climb/descent instructions computed by the automation algorithms. The FMS also interprets and implements the directives received from the ATCs controller via the simulated communication link. The autopilot is controlled by the simulated FMS or by the navigation command generator which simulates an auto controller for generating navigation command for aircraft maneuver. The following functions are performed by the FMS: • Flight plan processing: A flight plan is loaded by specifying the origin, destination, airways, waypoints, altitudes and estimated en route times (with wind speed and Coriolis calculations) as well as an activation airport or waypoint. The departure and arrival waypoint have pre-specified airspeed and altitude constraints as part of the navigation requirements. Crossing speeds, altitudes, and times for en route way points are then computed using the aircraft performance database. • Speed controls: The vertical trajectories are computed using the optimal rate of climb for the altitude at which the aircraft is flying which is retrieved from the aircraft performance database. Climbs are computed with a
•
transition from constant calibrated air speed (CAS) to constant mach number speed assuming maximum climb thrust. Fuel flow computation (from BADA): Assuming nominal aircraft mass, the thrust specific fuel consumption, η , in kg/min/kN is specified as a function of true airspeed, VT AS (knots) for the jet engines. The nominal R fuel flow, nom (kg/min), can then be calculated for jet engine aircraft using the thrust, T as: fnom = η × T
where η = Cf 1 (1 +
•
•
•
VT AS ) Cf 2
(10) (11)
Cf 1 (kg/min/kN )= 1st thrust specific fuel consumption coefficient, and Cf 2 (knots)= 2nd thrust specific fuel consumption obtained from aircraft performance tables. Although fuel is a major factor considered in the evaluation of the trajectory optimization, the change of the aircraft mass compared to the total mass of the aircraft is considered small. Therefore the change of the aircraft mass due to fuel burned during trajectory simulation is not considered. Aircraft acceleration/deceleration: For all aircraft, a standard acceleration model [10] is used, where the aircraft assumes a linear acceleration/deceleration profile. The maximum longitudinal acceleration is taken as 2.0f ps2 or 0.67m/s2 and the maximum normal acceleration is taken as 5.0f ps2 or 1.67m/s2 . Rate of climb and descent (ROCD) is derived from the respective aircraft performance database. Aircraft bank & turn: For all aircraft, a standard bank model [10] is used. During all turns, the aircraft maintains standard rate of turn unless the simulator requires the aircraft to perform a maximum bank angle. The nominal bank angle for civil aircraft in all phases of flight except take off and landing is 35 deg. The maximum bank angles for civil aircraft during all phases of flight except take off, landing and hold is 45 deg. The rate of turn (d/dt) is calculated as a function of bank angle and is equal to (g/VT AS ) × tan(Φ) where g is the mean earth gravity, VT AS is the aircraft true airspeed and Φ is the bank angle. Wind effect: In order to add the wind effect on aircraft speed the following computations are done by the autopilot, if WN and WE represent the northerly and easterly wind components at the location of the aircraft respectively, the equation for the ground velocity components Vg E and Vg N becomes as follows: Vg E = VT AS /E + WE
(12)
Vg N = VT AS /N + WN
(13)
where WE = Wspd × sin(Wdir )
where WN = Wspd × cos(Wdir ) VT AS /E and VT AS /N are the aircraft true airspeed (TAS) easterly and northerly components respectively. Wspd, Wdir are wind speed and direction respectively. Autopilot parameters are then initialized by setting the heading, speed and altitude as follows, if the aircraft is originating from outside the airspace boundary then current position is set as activation waypoint at airspace boundary. If the aircraft is origination from within the airspace then the departure airport coordinates are set as the current position. Based on the location of the next waypoint or TOC the heading is initialized. The phase of the flight (take-off, climb, cruise, descent, approach) is set based on its current position. Then using altitude and phase of flight, airspeed is deduced from the aircraft performance database. Altitude is initialized as departure airport altitude, if the flight is originating from within the airspace, else to the cruising level of the flight plan. Lateral navigation (LNAV) and vertical navigation (VNAV) modes are set, next altitude is initialized to the altitude of the next waypoint in the flight plan or to TOC (Free Flight). In the cruising phase as the flight get close to a waypoint the autopilot starts checking the bearing of the waypoint with respect to the aircraft. If the bearing of the waypoint from the aircraft is greater than 90 degrees and less than 270 degrees, then the waypoint is deleted from the flight plan, and the heading is updated with the next waypoint heading in the FMS.
G. Trajectory Modeling ATOMS models the 4D trajectory of the aircraft from its initial position, using either flight plan based routing or Great Circle routing (GCR) for Free Flight. In GCR the FMS extracts the origin point or the activation point and the destination or the deactivation point of the flight, then using the airspace database calculates the position coordinates of these points. Whereas in Flight plan based routing the FMS obtains the entire route which consists of waypoint/navigational fixes, and then calculates the position coordinates for each them forming an ordered pair of coordinate points representing the flight trajectory. By using Great Circle equations the course angle between these points is determined and the aircraft is navigated from one waypoint to another. Thus Free Flight can be seen as a special case of flight plan route flying, where all the middle waypoint are eliminated and the aircraft flies from origin to destination as a great circle route subject to aircraft performance parameters. While generating aircraft trajectory based on great circle routing, every cell of the airspace through which the aircraft traverse is recorded along with the estimated time of arrival in the flight plan. In case of any deviation from trajectory this cell plan is regenerated. This mechanism enabled to model intent information and propagation of future trajectory change points to neighboring aircrafts. The aircraft trajectory is generated as follows [20] : 1) The predicted position is set to the current position λ = λf , τ = τf and h = hf , where (λ,τ ,h) are current position lat, lon and alt and (λf ,τf ,hf ) are predicted position lat,lon and alt. 2) The time is set t = t0 . 3) The wind components Wn (λ,τ ,h), We (λ,τ ,h) are obtained from the wind meteorological data where Wn and We are northerly and easterly components of the wind respectively. The wind speed Vw and wind direction χw are computed using following equations respectively, Vw = (Wn2 + We2 )0.5
(14)
χw = arctan(We , Wn )
(15)
4) Then the great circle heading χg is computed by the great circle track equation χg = arctan{
sin(τf − τ ) cos λf } sin λf cos λ − sin λ cos λf cos(τf − τ )
(16)
5) Then the ground velocity is computed as: Vg = V cos{arcsin[(Vw /V ) sin(χwg )]} + Vw cos(χwg )
(17)
where track relative wind heading χwg = χw - χg and V is magnitude of the air speed. 6) By substituting these components into kinematic equation of motion given by: 1 × Vg cos χg R 1 τ= × Vg sin χg R cos λ h = Vclimb λ=
(18) (19) (20)
and integrating them using first order Euler integration for a time step of t = 1 s , the new latitude, longitude and altitude are computed. 7) This process is repeated till the entire trajectory of the aircraft is generated.
TABLE I F LIGHT PLANS REPRESENTING FOUR DIFFERENT TYPE OF Call Sign Origin Destination Activation Point Aircraft Type Speed(kn) Flight Level(00ft) Est. Time Enroute(min) Est. Time of Departure(s) ETA at Destination(s) Flight Type
SIA222 YSSY WSSS YSSY B747 480 320 466 60 26394 2
UAE412 OMDB YSSY ELATI A340 473 310 779 60 48908 1
SIA297 WSSS NZCH ATMAP B777 474 330 515 60 34535 0
FLIGHTS .
QFA405 YMML YSSY YMML A330 458 390 50 60 3641 3
TABLE II T HE ROUTE ELEMENT FOR EACH FLIGHT IN THE REPRESENTATIVE SET COMPRISING OF WAYPOINTS AND NAVAIDS Call Sign SIA222 UAE412 SIA297 QFA405
Route(Waypoints and Navaids) RIC MDG WLG TAVEV TASHA LARAB KIKEM ELATI PIPOV EMVAS EGAVI NWN RUSAD UVUPU GTH CULIN ATMAP BRM PUGUT WR UVUPU GTH CULIN SY CAWLY PLUGA SULON SBG AY MUSOP YAS BIK WELSH ODALE
IV. E XPERIMENTS & VALIDATIONS The airspace used for this study is the ICAO APAC region, namely Australian Airspace stretching in latitude from two degrees to 48 degrees south; and in longitude from 75 degrees to 163 degrees east. The simulation is set up by first loading the airspace under considerations defined by the airspace boundaries, then loading of all the waypoints, airways, navaids and airports within the airspace, finally loading of flight plans where each flight is categorized as follows: • If a flight originates and arrives from/at an airport within the area under consideration then the full flight plan is used (Type 0 Flight). • If a flight originates outside the airspace and arrives at an airport within the airspace then the flight plan is activated at the first point on the airspace boundary at the flight’s ETA at that point (Type 1 flight). • If a flight is destined for an airport outside the airspace and originating at an airport within the airspace then the flight plan is deactivated at the last point of the airspace boundary and at the flight’s ETA at that point (Type 2 flight). • For over-flying flights, the flight plan is activated at the first point on the airspace boundary at the flight’s ETA at that point and deactivated at the last point of the airspace boundary and at the flight’s ETA at that point (Type 3 Flight). A. Validation of FMS and Flight trajectory For validation of the FMS and trajectory generated by the AutoPilot, flight plans of five hundred flights comprising of four different types of actual flights were selected and fed to the FMS module of ATOMS. As an example set, four typical representative flights are shown in table I and their route element are shown in table II. The FMS processes the flight plan and outputs a flight profile for the entire route for all the flights in the flight plan. A flight profile generated for a type 3 flight is shown in table III. In the tabulated data, distance implies distance to next waypoint from the current position, heading implies heading required to turn to next waypoint, speed implies speed at that particular waypoint, latitude(Lat) and longitude(Lon) of current waypoint, altitude(Alt)
TABLE III F LIGHT PROFILE FOR A T YPE 3
WaypointID YMML SBG TOC AY MUSOP TOD YAS BIK WELSH ODALE YSSY
Distance(m) 121364.22 97456.8847 43244.83692 104168.4877 104162.2955 25099.68136 121710.3801 18794.88584 20731.35569 61784.3833 0
Heading(rad) 0.700746l 0.902369l 0.902369l 0.942513l 0.923747l 0.923747l 0.934144l 1.372062l 1.298981l 1.328189l 0.000000l
Call Sign:QFA405 Speed(m/s) Lat(rad) 115 -0.657698 236 -0.643154 235 -0.633613 236 -0.629482 236 -0.619883 235 -0.610039 237 -0.607665 191 -0.596321 183 -0.595739 146 -0.594866 74 -0.592539
TABLE IV FMS GENERATED FLIGHT PROFILE FOR A T YPE 3
WaypointID YMML TOC TOD YSSY
Distance(m) 218821.1047 312142.1223 174137.7156 0
Heading(rad) 0.930294l 0.944148l 0.948343l 0.000000l
FLIGHT
Lon(rad) 2.528109 2.543526 2.558434 2.565052 2.581342 2.597284 2.601122 2.619739 2.62323 2.627011 2.638356
FLIGHT IN
Call Sign:QFA405 Speed(m/s) Lat.(rad) 115 -0.657698 236 -0.637157 236 -0.608457 74 -0.592539
Alt(m) 21 8600 11887 11887 11887 11887 10700 4700 3800 2900 2
ETE(s) 679 412 183 441 441 106 513 98 113 423 0
ETA(s) 60 739 1151 1334 1775 2216 2322 2835 2933 3046 3469
ETE(s) 1902 1322 735 0
ETA(s) 60 1962 3284 3468
F REE F LIGHT
Lon(rad) 2.528109 2.562663 2.61147 2.638356
Alt(m) 21 11887 11887 2
at the current waypoint, ETE implies estimated time enroute from current waypoint to next waypoint, ETA implies estimated time of arrival at current waypoint. Now to validate the trajectory generated, navigation and steering by autopilot we plotted the flight profile generated by the FMS against the actual trajectory flown by the simulated aircraft and checked whether autopilot followed the flight plan, and navigated correctly. Figure 6 top left shows, for flight QFA405, waypoint generated by the FMS and the actual trajectory flown by the aircraft through autopilot control. Since the flight origin and destination are within the airspace, TOC and TOD can be seen along with the enroute waypoints and navaids, the autopilot flies the aircraft accurately as per the FMS generated flight path. Top right, shows for flight SIA222, the waypoint generated by the FMS and the actual trajectory flown by the aircraft through autopilot control. This flight has its destination outside the airspace so TOC can be seen on the flight path, but no TOD. The flight gets deactivated at waypoint KIKEM, which is the boundary waypoint on Australian airspace. Bottom left, shows flight UAE412, since its origin airport outside the airspace, the flight gets activated at ELATI, the airspace boundary waypoint on the route. TOD is present on the flight path since the destination airport is within the airspace. Bottom right shows flight SIA297, which is an overflying flight having its origin and destination outside the airspace boundary, the flight do not have TOC or TOD as it is in the cruise phase all the time it flies over the Australian airspace. It gets activated at waypoint ATMAP and deactivated at waypoint SULON. Navigation, trajectory generation and steering components of autopilot works as designed. At the turns, the autopilots maneuvers the aircraft appropriately and follow the correct waypoint heading of the flight plan. Further we examined the altitude profile of flight QFA405, for the different phases of flight. As can be seen from figure 7 that aircraft climbs off correctly, as per its flight plan, to its cruise altitude, reaches its TOC, then enters into cruise phase, upon reaching the TOD it transitions correctly into descent phase, and finally into approach phase. Further we examined the FMS and autopilot modules of ATOMS in a Free Flight airspace. In Free Flight the FMS eliminates the route data from the flight plan and performs the flight profiling as can be seen for flight QFA405 in table V, which has only origin, TOC, TOD and destination. Figure 8 shows the aircraft route flown for flight QFA405 in classical air traffic control environment and in Free Flight airspace. In Free Flight airspace the aircraft flies great circle route from origin to destination, whereas in classical model it follows the airways route, navigating
Fig. 6.
FMS generated flight profile and trajectory flown by autopilot.
Fig. 7.
Altitude log of flight QFA405, as recorded in simulator.
waypoint to waypoint.
Fig. 8. Free Flight route vs Waypoint route for flight QFA405, Free Flight route can be seen as great circle route from origin to destination.
B. Validation of Aircraft Maneuvers For validation of aircraft maneuvers we simulated the aircraft maneuvers of a MacDonnell Douglas-90 aircraft (MD90) for rate of acceleration-deceleration (ROAD), rate of climb and descent (ROCD) and rate of heading change (ROHC). Navigation commands were generated using the pilot’s interface for flight navigation for turn, climb and heading change. The experimental results in figure 9 shows that aircraft maneuvers are within its flight safety envelop. It can be seen from figure 9 top left: speed maneuver for an MD90 aircraft, which has maximum rate of accelerate-decelerate (ROAD) of 2.0f t/s2 , top right and bottom left: a climb-descent maneuver for a solution trajectory which shows the rate of climb and descend (ROCD) at 27,000 ft for MD90 aircraft, which is ±11m/s, bottom right: a heading change maneuver for a solution trajectory which shows that the rate of heading change (ROHC) MD90 aircraft, which is ±3deg/s. C. Validation of Atmospheric Model The atmospheric equations were programmed separately in C++ and the outputs were verified using the environment block of the MATLAB-Aeronautical Simulation Block Set(Aerosim) [21]. The atmospheric data for each flight level is stored in a 3D data structure, which can be accessed by other modules of simulator. The temperature, density and pressure vs altitude graphs shown in figure 10 are generated by using ICAO standard atmosphere (ISA) [17] equations for atmospheric modeling. The tropopause altitude shown in figure 10 bottom is representative of the tropopause layer where the equations of the atmosphere change. Effect of wind on the flight: For the given flight QFA405, we did two separate experiments, one with sample wind data and the other without any wind (Zero Wind). The autopilot keeps the aircraft on track by doing wind corrections, however it affects the Actual Time of Arrival(ATA) at the specified waypoints. It can be seen from figure 11 that the wind affects the speed of the aircraft and the ATA at the waypoints is more than the ATA at waypoint when there is are no winds.
Fig. 9.
Validation of climb-descent,heading and speed maneuvers for an MD90 aircraft.
Further to see whether the autopilot can keep the aircraft on track by doing wind corrections, we plotted the Cross(X) Track error, which measures the off track distance (left or right) from the original trajectory. As can be seen from figure 12 that the X-track error was within 200m either side in absence of wind and within 400m either side in presence of wind, this conforms to required navigation performance (RNP) of 1nmi envisioned in Free Flight, where aircraft are allowed to drift a maximum of 1nmi of their intended track.
Fig. 10.
Atmospheric data modeled in ATOMS
Fig. 11.
Actual time of arrival (ATA) of a flight with wind and with zero wind for flight QFA405.
Fig. 12. Cross Track error measured for flight QFA405 with wind and with zero wind. Cross Track error is within RNP-1 airspace envisioned in Free Flight
D. Fuel consumption and Economic Indicators To maintain orderly air traffic flow ATCs may allocate pilots an altitude that might not be optimum for the aircraft in terms of fuel consumption. In Free Flight pilots will be able to choose the most fuel efficient altitude depending upon the type of aircraft, thus offering immediate economic benefits to the industry in terms of lower fuel consumption. This mainly results from flying aircraft at their optimal altitude and time savings due to direct route to destination, give the airspace constraints. This will lead to reduction in flight operating time which may results in more effective use of airline fleet and passenger comfort. As shown in figure 13 for flight QFA405 when simulated in Free Flight airspace consumed 5863.12kg less fuel, saves 76.13min of flying time and covers approx. 13km less distance than in controlled airspace. This was primarily due to the direct route flown by the aircraft in Free Flight airspace leading to few maneuvers in the absence of waypoints. However, in Free Flight airspace, aircrafts may not be able to fly direct routes due to various airspace constraints but even then given the quantum of benefits, it will certainly have a significant economic impact.
Fig. 13.
Fuel and time consumption for flight QFA405 in Free Flight airspace and controlled airspace.
V. G RAPHICAL U SER I NTERFACE ATOMS is a Graphical User Interface (GUI) driven simulator. Through the GUI, the user can define the extent of the airspace to be simulated, generate a new set of flights with a varying set of traffic densities, can load and run a previously generated set of flights, and set the simulation run speed. All the desired controls, navigation and display properties of aircraft and ATCs are achieved through the GUI. The main window of the ATOMS consists of airspace display and simulator operations control tool bar. Figure 14 shows the main simulator window and the airspace modeled, waypoints, airways, airports, aircraft and their intent track. Four representative flight are also displayed with their call sign, speed and altitude information with color coded symbology for flight phase. The flights show their intended flight trajectory from current position to destination/ deactivation point. The airspace display properties are controlled through user driven features. The user can pause, resume or terminate a simulation, airspace can be zoom-in and zoom-out for any section and bad weather scenarios can also be created as described in the later section.
Fig. 14.
Main simulator window of ATOMS.
Aircraft in ATOMS can be flown using autopilot mode as well as user driven mode. On double clicking on any aircraft in the airspace, it’s control and navigation interface is activated as shown in figure 20 left. The pilot interface provides control commands (climb, descent, speed, heading etc.) to navigate the aircraft and autopilot information. Pilots can select the display of intent track, range ring around the aircraft, fuel flow computation, determine ADS-B [22] and weather radar range etc. This interface also allows user to access advance Free Flight
functions which are described later in the section. There is a simulated communication interface for Air Traffic Controllers (ATCs) shown in figure 14 left hand side, through which and ATCc can monitor the active aircraft in the airspace and can perform the basic functionalities of ATCs. The aircraft responds to ATCs directives received from the ATCs via this simulated communications interface. Once these directives are generated, they are interpreted by the FMS of the aircraft concerned, which then translates them into a series of autopilot actions depending on the type of directive, the actual aircraft state and the phase of flight. VI. F REE F LIGHT C ONCEPTS
IN
ATOMS
A variety of Free Flight concepts are implemented in ATOMS. This section discusses their modeling and results obtained. For detailed information on these concepts we refer the reader to [3], [2], [23] A. Airborne Conflict Detection and Resolution Airborne conflict detection and resolution (CD&R) to maintain aircraft separation is an essential component of air traffic management and has assumed much importance in the wake of Free Flight initiatives. Based on a survey of 68 CD&R algorithms [24] and literature available on CD&R techniques, we selected four CD&R algorithms representing three different categories and modeled them in ATOMS. The first category is known as Force Field approach [25] in which aircraft are treated as electrically charged particles, and uses modified electrostatic equations to generate resolution maneuvers. For resolution maneuvers this approach uses the repulsive forces between these charged particles (aircraft). Second is the Geometric approach [26] which utilizes current position and velocity vectors to determine miss distance and time-to-closest-approach as conflict detection mechanism, and for resolution uses combination of vertical speed and heading changes to resolve conflicts while minimizing the magnitude of the velocity change. Third is an advanced version of present day tactical collision avoidance system (TCAS) known as FAA-TCASIII, and uses range rate measurers and altitude threshold for detecting conflicts. For resolution it selects maneuvers that maximize the range vector for the time to closest point of approach or closest point of approach between two aircraft. All the three approaches uses state vectors and position velocity data to predict conflicts in the future. The idea to select the three approaches was not to elect the best but to investigate how they performed on safety and economic metrics given different scenarios. The following four CD&R algorithm were modeled in ATOMS: 1) CD&R algorithm (KB3D) by G. Dowek, C. Munoz, and A. Geser, ICASE, NASA Langley Research Center [27](Geometric). 2) CD&R algorithm (Billimoria) by K.D. Billimoria, NASA Ames Research Center [26](Geometric). 3) CD&R algorithm (Rockwell) by T.W. Rand, Rockwell Collins and M.S. Eby, Source Code Systems [28](Force Field). 4) CD&R algorithm (FAA TACS-III) by R.Y. Gazit, Department of Aeronautics and Astronautics, Stanford University [29]. The modeled algorithms were tested on the data set provided by the authors in their respective papers as well as using Free Flight scenarios. Three different air traffic scenarios in Australian airspace were simulated for the following conflict geometry: • Head-On conflict: Two aircraft flying at the same altitude, having bearing in opposite direction. • Cross-Over conflict: Two aircraft flying at the same altitude having crossing at a point in airspace at the same time. • In-Trail conflict: Two aircraft flying at the same altitude, in the same heading, where trailing aircraft have a higher speed and comes in conflict with the leading one in the future. Sample flight plans for conflict simulation have their estimated time of departure and cruise altitude adjusted accordingly to have the desired conflict scenario. For the in-trail conflict scenario, two flights departed from the same airport with a certain time difference, the flight which is ahead is assigned a slower speed than the flight which is behind it, causing in-trail conflict when the flight behind catches up. For all the CD&R algorithms lateral separation distance was set to 9200m (5nmi), separation time set to 300s (5min look ahead time) as it provide
TABLE V F LIGHT PLANS OF Scenario Head On In Trail Cross Over
Call Sign QFA405 VBB222 QFA405 ZRW444 SIA222 XCC999
CONFLICTING FLIGHTS FOR THREE SCENARIOS
Origin
Destination
YMML YSSY YMML YMML YSSY YPAD
YSSY YMML YSSY YSSY WSSS YBBN
Aircraft Type A330 A330 A330 A330 B747 B747
Speed (kn) 458 458 297 458 480 480
Flight Level(00ft) 390 390 390 390 320 320
ETE (min) 50 50 50 50 466 160
ETD (s) 60 60 60 1080 1680 60
enough time for pilots to execute a resolution maneuver [8] and vertical separation set to 305m (1000ft). A conflict search area of 80nmi was defined for an aircraft based on current ADS-B technology limitations. Each flight plan for a given geometry was run independently for each CD&R algorithm. The resolution advisory generated by the algorithms consists of a heading change maneuver, an altitude change maneuver, and speed maneuver. For each conflicting flight pair in the flight plan one is treated as own ship (aircraft which detect the conflict and initiate a resolution) and other as intruder (aircraft which does not resolve conflict and maintains course). In KB3D and FAA-TCASIII, resolution vectors were independent of each other such that implementing even one of them can resolve a conflict, whereas in the Billimoria and Rockwell-Collins cases they requires a combination of heading and speed change for a resolution maneuver. Conflicts were resolved by ownship in a non-cooperative manner. Figure 15 shows one of the conflict resolution maneuvers for a head on conflict. Note that after resolving conflict the ownship resumes back its own navigation by doing course corrections.
Fig. 15.
A conflict resolution maneuver for a head on conflict scenario.
Resolution maneuver vectors generated by the algorithms along with fuel and time cost are shown in table VI. We evaluated the algorithms on the following two performance metrics suggested by [30] • Safety: This metric records the loss of separation, if any, between the conflicting aircrafts. For all the three conflicting geometries simulated, all the four algorithm successfully resolved the conflicts and no loss of separation was recorded. • Efficiency: This metric records the operational cost of conflict resolution by each algorithm in terms of extra fuel burn due to increased drag and flight path distance traveled during a maneuver and extra time required to execute a maneuver and return back to track. Metric data for the four algorithms as shown in figure 16 suggests that the force field based algorithm is very expensive in terms of fuel consumption. For cross over scenarios when speed controls are applied as a part of the resolution vectors, proved most expensive followed by heading and altitude based maneuvers. The Rule based approach (FAA-TCASIII) in general give expensive maneuvers both in terms of fuel cost and time consumed
TABLE VI C ONFLICT RESOLUTION VECTORS GENERATED FOR DIFFERENT CONFLICT GEOMETRIES ALONG WITH EXTRA FUEL AND TIME INCURRED . Head-On Conflict Ownship:QFA405 Bearing(deg): Altitude(kft): Speed(kn):
54.84 39.00 456.80
Cross-Over Conflict Ownship:XCC999 Bearing(deg): Altitude(kft): Speed(kn):
58.33 32.00 497.62
Intrail Conflict Ownship:ZRW444 Bearing(deg): Altitude(kft): Speed(kn):
54.35 39.00 456.80
Fig. 16.
Resolution Vector Heading(deg) Altitude(kft) Speed(kn) Extra Fuel(kg) Extra Time(s)
KB3D 44.12 40.00 – 30.8 27.6
Billimoria 47.59 – 455.72 17.4 20.0
Rockwell 63.10 – 274.69 64.3 94.0
FAA-TCASIII 25.01 31.50 – 66.28 111.0
Resolution Vector Heading(deg) Altitude(kft) Speed(kn) Extra Fuel(kg) Extra Time(s)
KB3D 56.06 33.00 524.86 23.23 27
Billimoria 56.97 – 506.41 2.04 60.0
Rockwell 75.57 – 386.59 45.6 55
FAA-TCASIII 28.33 24.50 0.00 145.6 167.4
Resolution Vector Heading(deg) Altitude(kft) Speed(kn) Extra Fuel(kg) Extra Time(s)
KB3D 59.02 40.00 305.03 28.2 24.6
Billimoria 61.40 – 437.37 15.6 18.0
Rockwell 60.59 – 239.10 108 276.0
FAA-TCASIII 84.21 31.50 – 113.53 127.8
Extra fuel and extra time cost for the resolution vectors generated by the CD&R algorithms.
which is due to lack of maneuver optimization in the algorithm, however it always gives resolution vectors within the aircraft’s performance envelop. The Force Field based algorithm suggested time consuming maneuvers and recorded high on extra time consumed to execute the maneuvers. The Geometric approach based algorithm provided a good CD& R mechanism for all the three geometries examined. However the set of resolution vectors it provide is sometimes beyond the aircraft performance limits, which needs an extra function call to BADA database to
eliminate unfeasible solutions. It can be deduced from the data that speed maneuvers proved to be very costly, followed by altitude maneuvers, heading change maneuvers resulted in less fuel and extra time consumed. However the performance of these algorithms in a multi aircraft environment hasn’t examined in this paper, and results may indicate otherwise as CD&R algorithms behave differently in a multi aircraft environment. Intent based modeling used in ATOMS largely reduced the search space for detecting conflicts with intruders. Aircrafts can now search a region of six cells around it, resulting in 60nmi (alert zone) locus. This increased the efficiency of the conflict detection mechanism by doing conflict checks with only those intruders whose intended trajectory falls in the alert zone within a specified time frame. Further, it added another safety net by reducing the large number of false alarms generated by state based conflict detection algorithms. Apart from these algorithms we also developed a Neuro-Controller for CD&R in a 2D environment. Experimental results suggest that neural networks can be effectively trained using evolutionary techniques for the task of CD&R [31]. Implementation of its revised version in ATOMS is under progress. B. Cockpit Display of Traffic Information CDTI technology [32] will provide pilots with a traffic display that will be more global than that in the current TCAS situation display. It will allow pilots to determine precisely the state and intent information of aircraft in proximity to themselves and may also provide other useful and time critical information such as convective weather cells, special use airspace etc.
Fig. 17. Cockpit display of Traffic information as modeled in ATOMS, the CDTI interface besides displaying traffic, conflict and weather information also acts as decision support system by displaying the avoidance vector.
We developed a technology prototype of CDTI and implemented it in ATOMS to investigate various design issue including the dimensionality of representing the traffic around the aircraft, amount of traffic information to be displayed, level of intent information, constrained airspace information display etc. Figure 17 shows the CDTI interface in ATOMS, its features include: • AutoPilot information display: The CDTI displays key autopilot information such as ground speed, true air speed, heading and altitude. It also shows graphically the track angle, original track and current track. • ADS-B range selection: Pilots can select the range of radar display which is from 40 nmi to 120 nmi modeled on current state of art technology for Automatic dependent surveillance- broadcast(ADS-B) radar. Selecting a large range will increase the display clutter and small range will increase the risk of late collision alert from neighboring traffic.
TABLE VII C OLOR CLASSIFICATION OF THUNDER STORM CELLS BASED ON RADAR REFLECTIVITY. Color None Blue Green Yellow Red
Radar Reflectivity(dBZ) I
