Академический Документы
Профессиональный Документы
Культура Документы
Theneedforincreasingsafetyandefficiencylevelsintheairtransportsystemrequiresmoderncontroland
traffic management (ATM) systems for aircraft in air and in ground operations, as well as for service
vehicles on the airport surface. The related Communications, Navigation and Surveillance (CNS)
infrastructures call for enhanced positioning and identification techniques such as Multilateration (MLAT)
andWideAreaMLAT(WAM),automaticdependentsurveillancebroadcast(ADSB),automaticlocationof
vehiclesandmanagement(AVMS).Thistypeofenhancedsurveillanceinfrastructureisspatiallydistributed
(i.e. with many receiving or transmitting/receiving stations) and logically distributed (i.e. with local and
central processing and with fusion of different information sources, including the traditional primary and
secondaryradar).
In this frame, new system architectures and new algorithms for integrity monitoring and for multisensor
datafusionarerequired.
Securityanddefencesystemsusesimilaralgorithmsforpassivelocatingoftargetsbasedonmeasurements
of Time of Arrival (TOA) and its differences (TDOA) as well as of Doppler frequency and its differences
(FDOA),possiblycombinedwithangular/directionmeasurements(AOA/DOA).
The emerging dual use concept for surveillance and identification systems and the intrinsic
commonalities of location algorithms in the different applications did suggest to include the pertaining
itemsintheWorkshopstopics.
Theseenhancedsurveillancesystemsareconsideredintheirmanyfacets:
x ADSBimplementation,services,equipageandapplications
x ADSBatairports
x WideAreaMultilateration
x MultilaterationatAirports
x Radartechnologyforairportsurveillance
x PSRandSSRtechnologyforAirTrafficSurveillance
x GNSSapplicationstoAirTrafficManagement
x Safetyissues,solutionsandstandards
x Interoperabilitybetweencommercial,militaryandGeneralAviation
x Integrationofunmannedaircraftsystems(UAS)
x Dualuseapplications(security,defence)
x SystemsandSubsystems:Architectures,Newconcepts
x Passivelocationbasedontime,Doppler,angleMeasurements
x Sensordatafusion
x Technologies(hardware,firmware,software)
x Environmentalaspects(includingradiopropagation)
x TestingandFieldAnalysis,IntegrityMonitoring
x ImplementationplansandOperationalresults
Thisworkshopisaimedtocoverallofthesefacets,atvariouslevelsofdetail.Wetrustthatitwillprovidea
true flavour of the current developments and trends, and therefore serve the best interests of the
operationalandscientificcommunities.
Gaspare Galati Piet van Genderen
iv
Members of the Technical Program Committee
SimonAtkinson,RokeManorResearchLtdU.K.
JuanVicenteBalbastreTejedor,TechnicalUniversityValencia
CristianoBaldoni,ENAVItaly
JuanBesada,TechnicalUniversityMadrid
JochenBredemeyer,FlightCalibrationServicesGermany
RonaldBruno,ITTUSA
StevenCampbell,MITLincolnLab
EnzoDalleMese,UniversityofPisaItaly
PatrizioDeMarco,SelexS.I.Italy
SergiodiGirolamo,ThalesAleniaSpaceItaly
PietroFinocchio,AFCEAItaly
MicheleFiorini,IETItaly
PeterForm,TechnicalUnivewrsityBraunschweigh
GaspareGalati,TorVergataUniversityRomeItaly
FulvioGini,UniversityofPisaItaly
RalfHeidger,DFSGermany
AdamKawalec,MilitaryUniversityofTechnologyWarsawPoland
DirkKugler,DLRGermany
MauroLeonardi,TorVergataUniversityRomeItaly
IanLevitt,FAAUSA
KonstantinLukin,IREKharkovUkraine
PravasMahapatra,IndianInstituteofScienceBangalore
MichelMoruzzis,ThalesFrance
DanielMuller,ThalesAirSystemsFrance
BenitoPalumbo,PrivateExpertConsultantItaly
GabrielePavan,TorVergataUniversityRomeItaly
GiorgioPerrotta,SpaceSystemsItaly
NicolasPetrochilos,UniversityofReimsFrance
DanielaPistoia,ElettronicaS.p.A.Italy
ChristosRekkas,Eurocontrol
HermannRohling,Techn.UniversityHamburgHarburg
JohnScardina,FAAUSA
FaustoSimoni,ENAVItaly
RobertoSorrentino,PerugiaUniversityandEUMA
DanielStamm,Skyguide
LorenzPeterSchmidt,UniversitaetErlangenNuernbergGermany
VojtechStejskal,ERABeyondRadarPardubiceCzeckRepublic
FilippoTomasello,EASACologne
PietvanGenderen,TechnicalUniversityDelft
CarloVertua,ThalesItalia
FelixYanovski,UniversityofKiev
AlexYarovoy,TechnicalUniversityDelftIRCTR,TheNetherlands
MaurizioZacchei,ENAVItaly
v
TYRRHENIAN INTERNATIONAL WORKSHOP ON DIGITAL COMMUNICATIONS
ENHANCED SURVEILLANCE OF AIRCRAFT AND VEHICLES
LIST OF SESSIONS
1 - SESAR 1
Chair: C. Baldoni
Co-chair: R. Heidger
2 - REGULATORY AND OPERATIONAL ASPECTS 33
Chair: P. Finocchio
Co-chair: C. Rekkas
3 - HIGH RESOLUTION RADAR TECHNIQUES 59
Chair: J. Besada
Co-chair: D. Kgler
4 - NON-COOPERATIVE LOCATION 87
Chair: D. Pistoia
Co-chair: A. Kawalec
5 - NON-COOPERATIVE PASSIVE COHERENT LOCATION AND
MULTISTATIC PRIMARY SURVEILLANCE RADARLOCATION 117
Chair: D. Muller
Co-chair: V. Stejskal
6 - NEWCONCEPTS AND TOOLS 139
Chair: G. Pavan
Co-chair: A. Mahapatra
7 - MULTILATERATION AND ADS-B (1) 165
Chair: S. Di Girolamo
Co-chair: S. Atkinson
8 - MULTILATERATION AND ADS-B (2) 195
Chair: S. Atkinson
Co-chair: S. Di Girolamo
INTERACTIVE SESSION 223
Chair: P. van Genderen
Co-chair: M. Leonardi
vi
vii
Focused Session 1
SESAR
Chair:C.BaldoniCochair:R.Heidger
Theframeofthissessionistheprovisionofsurveillance,navigationandseparationinahighly
integratedgroundbasedandairbornefutureATMSystemaimedatgreener,moreefficient,more
costeffectiveandallsafeairtransportoperations.
Fromidentificationofrequirementstotheoperationalvalidation
ofanintegratedsolution 2
Claudio Vaccaro, Gabriella Duca - SICTA, Italy
AnewconceptforATMServiceSupervisionforthenewSESAR
EnterpriseArchitecture 9
Gabriella Carrozza - SESM s.c.a.r.l., Italy
Stephen Straub - Deutsche Flugsicherung GmbH (DFS), Germany
Hakim Souami - THALES Air Systems, France
ADSBIntegrationintheSESARsurfacesurveillancearchitecture 13
Andrs Soto, Pedro Merino, Jorge Valle - Indra Sistemas S.A., Spain
Futuremobilesatellitecommunication 19
Pierpaolo Tavernise - THALES Alenia Space, Italy
OptimisingRunwayThroughputthroughWakeVortexDetection,
PredictionandDecisionSupportTools 27
Frdric Barbaresco, Philippe Juge, Mathieu Klein, Yves Ricci,
Jean-Yves Schneider, Jean-Franois Moneuse - THALES Air Systems, France
Session 2
Regulatory and Operational Aspects
Chair:P.FinocchioCochair:C.Rekkas
ADSBandWAMdeploymentinEurope 35
Christos Rekkas - Eurocontrol, Belgium
U.S.ActivitiesinADSBSystemsImplementation 41
Paul Douglas Arbuckle - US Federal Aviation Administration, USA
DetectandavoidforUnmannedAircraftSystemsin
thetotalsystemapproach 47
Filippo Tomasello - EASA, Italy
David Haddon - EASA, Germany
viii
NorthSeaHelicopterADSB/MLatPilotProjectFindings 53
Paul Thomas - Bristow Helicopters, United Kingdom
Session 3
High Resolution Radar Techniques
Chair:J.BesadaCochair:D.Kgler
LAOTSE,anApproachforForeignObjectDetection
bymultimodalnetted2D/3DSensors 61
Sebastian Hantscher, Helmut Essen, Paul Warok Ruediger Zimmermann,
Martin Schrder, Rainer Sommer, Stefan Lang - Fraunhofer FHR, Germany
Marek Schikora, Klaus Wild, Wolfgang Koch - Fraunhofer FKIE, Germany
MillimeterwaveRadarforRunwayDebrisDetection 65
Helmut Essen, Frank P. Lorenz, Sebastian Hantscher, Paul Warok,
Ruediger Zimmermann, Martin Schrder - Fraunhofer FHR, Germany
Wolfgang Koch, Marek Schikora, G. Luedtke,
Klaus Wild - Fraunhofer FKIE, Germany
OFDMWaveformsforaFullyPolarimetricWeatherRadar 69
Oleg A. Krasnov, Zongbo Wang, R. Firat Tigrek,
Piet van Genderen - Delft University of Technology, The Netherlands
Polarimetryappliedtoavionicweatherradar:improvementon
meteorologicalphenomenadetectionandclassification 73
Alberto Lupidi, Christian Moscardini, Fabrizio Berizzi - University of Pisa, Italy
Andrea Garzelli - University of Siena, Italy
Fabrizio Cuccoli - CNIT, Italy
Marcello Bernab - SELEX Galileo S.p.A., Italy
PrinciplesofUtilizationofPolarizationInvariantParametersfor
ClassificationandRecognitionofDistributedRadarObjects
PartI.Simplestmodelofadistributedobjectpaper 79
Victor N. Tatarinov, Sergey V. Tatarinov
Tomsk University of Control Systems and Radioelectronics, Russian Federation
Piet van Genderen - Delft University of Technology, The Netherlands
PartII.Multipointmodelandcorrelationtheory 83
Victor N. Tatarinov, Sergey V. Tatarinov
Tomsk University of Control Systems and Radioelectronics, Russian Federation
Piet van Genderen - Delft University of Technology, The Netherlands
ix
Focused Session 4
Non-cooperative location
Chair:D.PistoiaCochair:A.Kawalec
Scopeofthesessionistocollectstateoftheartresultsrelatedtotheproblemoflocationofnon
cooperativetargets.Particularemphasiswillbegiventoallthosepapersaddressingpassive
techniques,bothstandaloneandintegratinginformationfromactivesensors.
Processingofsignalsrelatedtothemeasurementofradiofrequencyand/orIR/VIS/UVsignatureare
ofinterest.Papersshowingresultsfromtrialswillhavepriority.
Enhancingsensitivityforemittergeolocation 89
Gran Tengstrand, Viktor Andersson,
Peter Hultman - SAAB Electronic Defence Systems, Sweden
Dario Benvenuti - ELETTRONICA, Italy
Jean-Franois Grandin, Luc Bosser - THALES Systmes Aroports, France
Brje Andersson, Anders Johansson - Swedish Defence Research Agency, Sweden
NewsolutiontoenhancethesecurityinAirTrafficControl 95
Enrico Anniballi, Roberta Cardinali - SESM s.c.a.r.l., Italy
SolvingtheDataLinkbottleneckforMPEGLocation 101
Jean-Franois Grandin, Luc Bosser - THALES Systmes Aroports, France
Gran Tengstrand, Viktor Andersson,
Peter Hultman - SAAB Electronic Defence Systems, Sweden
Dario Benvenuti - ELETTRONICA, Italy
Brje Andersson, Anders Johansson - Swedish Defence Research Agency, Sweden
ParasiticDopplereffectinpassivelocation 107
Dario Benvenuti - ELETTRONICA, Italy
Aninairpassiveacousticsurveillancesystemforairtrafficcontrol111
Domenico Donisi, Marco Bonamente - D'Appolonia S.p.A., Italy
Vincenzo Quaranta, Salvatore Ameduri - CIRA, Italy
Focused Session 5
Non-cooperative Passive Coherent Location and
Multistatic Primary Surveillance Radarlocation
Chair:D.MullerCochair:V.Stejskal
ThesessionisaimedtodiscusstheprogressofPassiveCoherentLocationandotherindependent
noncooperativesurveillancetechniquesbasedonradar,likemultistaticprimaryradar.
OnboardPCLsystemsforairborneplatformprotection119
Krzysztof S. Kulpa, Mateusz Malanowski,
Piotr Samczyski, Jacek Misiurewicz - Warsaw University of Technology, Poland
Maciej Smolarczyk - Telecommunications Research Institute, Poland
x
FMBasedPassiveCoherentRadar.
Fromdetectionstotracks 123
Radek Plek, Vojtch Stejskal, Martin Pelant,
Martin Vojek - ERA Beyond Radar, Czech Republic
HighRangeResolutionMultichannelDVBTPassiveRadar:
AerialTargetDetections 129
Dario Petri, Amerigo Capria - CNIT, Italy
Michele Conti, Fabrizio Berizzi,
Marco Martorella, Enzo Dalle Mese - University of Pisa, Italy
RangeOnlyMultistaticTrackinginClutter 133
Darko Muicki, Taek Lyul Song - Hanyang University, Republic of Korea
Session 6
New Concepts and Tools
Chair:G.PavanCochair:A.Mahapatra
TrackerQualityMonitoringbyNonDedicatedCalibrationFlights 141
Matthias Hess, Ralf Heidger - Deutsche Flugsicherung GmbH (DFS), Germany
Jochen Bredemeyer - FCS Flight Calibration Service GmbH, Germany
TheTransponderDataRecorder:firstimplementationandapplications 147
Gaspare Galati, Mauro Leonardi, Emilio G. Piracci - Tor Vergata University, Italy
Soumem Samanta - National Institute of Technology, India
ADSB/MLATsurveillancesystemfromHighAltitudePlatformSystems 153
Mauro Leonardi, Silvio Spinelli, Gaspare Galati - Tor Vergata University, Italy
SpacebasedADSBAsmallstepfortechnologyagiantleapforATM? 159
Adam Parkinson - Helios, United Kingdom
Focused Session 7
Multilateration and ADS-B (1)
Chair:S.DiGirolamoCochair:S.Atkinson
AimofthesessionisatechnicaldiscussionontheadvantagesandlimitationsofseparateWAMand
ADSBsystemsandthebenefitsandtechnicalchallengesofdatafusionincombinedsystems.
StrategiestoDesignandDeployModeSMultilaterationSystems 167
Ivan A. Mantilla-G, Juan V. Balbastre-T,
Elas de los Reyes - Universidad Politcnica de Valencia, Spain
Mauro Leonardi, Gaspare Galati - Tor Vergata University, Italy
xi
CorrectionofsystematicerrorsinWideAreaMultilateration 173
Jorge Abbud, Gonzalo de Miguel,
Juan Besada - Universidad Politcnica de Madrid, Spain
Multilaterationsystemtimesynchronizationvia
overdeterminationofTDOAmeasurements 179
Martin Pelant, Vojtch Stejskal - ERA Beyond Radar, Czech Republic
ImprovementofMultilateration(MLAT)AccuracyandConvergence
forAirportSurveillance 185
Ivan A. Mantilla-G, Juan V. Balbastre-T,
Elas de los Reyes - Universidad Politcnica de Valencia, Spain
Mauro Leonardi, Gaspare Galati - Tor Vergata University, Italy
AssessingthesafetyofWAMoveranonradarsurveillancearea 191
James Hanson, Ben Stanley - Helios, United Kingdom
Focused Session 8
Multilateration and ADS-B (2)
Chair:S.AtkinsonCochair:S.DiGirolamo
AimofthesessionisatechnicaldiscussionontheadvantagesandlimitationsofseparateWAMand
ADSBsystemsandthebenefitsandtechnicalchallengesofdatafusionincombinedsystems.
ImplementationofADSBSystems
BenefitsandConsiderations 197
Abraham Barsheshat - Sensis Corporation, USA
InvestigationofMeasurementCharacteristicsof
MLAT/WAMandADSB 203
Klaus Pourvoyeur, Adolf Mathias,
Ralf Heidger - Deutsche Flugsicherung GmbH (DFS), Germany
RealTimePerformanceMonitoringandNoise
AnalysisinanoperationalWAMSystem 207
Alexander Pawlitzki, Holger Neufeldt - THALES Air Systems GmbH, Germany
ADSBviaIridiumNEXTsatellites 213
Paolo Noschese, Silvia Porfili,
Sergio Di Girolamo - THALES Alenia Space, Italy
xii
IndependentSurveillanceBroadcastADSBReceivers
withDOAEstimation 219
Christoph Reck, Lorenz-P. Schmidt
Max S. Reuther - University of Erlangen-Nuremberg, Germany
Alexander Jasch - TU Braunschweig, Germany
INTERACTIVE
SESSION
Chair:P.vanGenderenCochair:M.Leonardi
SBASAvailabilityImprovementBasedonthe
ModifiedRadarTechniques 225
Boriana Vassileva
Institute for Information and Communication Technologies, Bulgaria
Boris Vassilev - Technical University of Sofia, Bulgaria
SmartconcatenationofCorrelativeDirectionFindingand
Superresolutiontechniques 231
Libero Dinoi, Marco Guerriero, Gianpiero Panci - Elettronica SpA, Italy
DevelopmentofHighPerformanceWAMSystem 237
Hiromi Miyazaki, Tadashi Koga, Eisuke Ueda,
Yasuyuki Kakubari, Shirou Nihei - Electronic Navigation Research Institute, Japan
MachineReadableEncodingStandardSpecificationsinATC 241
Adolf Mathias, Matthias Hess - Deutsche Flugsicherung GmbH (DFS), Germany
ExperimentalADSBbasedsurveillance 247
Juan A. Besada, Gonzalo de Miguel
Ana M. Bernardos, Jos R. Casar - Universidad Politcnica de Madrid, Spain
ADSBHILSTestforCollisionAvoidanceofSmartUAV 253
Changsun Yoo, Am Cho, Bumjin Park
Youngshin Kang - Korea Aerospace Research Institute, Republic of Korea
Sangwook Shim, Ilhyung Lee - Korea Advanced Institute of Science and
Technology, Republic of Korea
xiii
Ageneticalgorithmandlocalsearchfortheautomated
calculationofdilutionofprecisionofmodeS
multilaterationsystemsatairports 259
Ruy F.Ruiz-Mojica, Ivan A. Mantilla-Gaviria,
Emilio J. Gmez-Prez, Juan V. Balbastre-Tejedor,
Elias de Los Reyes-Dav - Universidad Politcnica de Valencia, Spain
MultiApproachStrategyforMultiSensorData
FusionEnhancement 265
Carlo A. Vertua, Luca Saini - THALES, Italy
Olivier Baud, Nicolas Honor, Peter E. Lawrence - THALES, France
AutomaticIdentificationofProcessStepsin
MultilaterationData 271
Stefanie Helm - German Aerospace Center (DLR), Germany
LISTOFAUTHORS 277
SESAR
Fromidentificationofrequirementstotheoperational
validationofanintegratedsolution
Session1.1page3
AnewconceptforATMServiceSupervision
forthenewSESAREnterpriseArchitecture
Session1.2page9
ADSBIntegrationintheSESARsurface
surveillancearchitecture
Session1.3page13
Futuremobilesatellitecommunication
Session1.4page19
OptimisingRunwayThroughputthroughWake
VortexDetection,PredictionandDecisionSupportTools
Session1.5page27
):
R
: 5m
: 0.15
In this simulation, two trajectories (TR1 and TR2) are
presented. TR1 represents an aircraft which departs from a
stand in Terminal 3 and takes off from runway 15R and TR2
represents an aircraft which departs from a stand in Terminal
4S and takes off from runway 36R.
For the sake of simplicity, some of the already presented
effects are not taken into account in order to focus on the
simulated ones and not being disturbed by the joint effects of
the rest of them. Thus, the following assumptions are
considered:
The system receives ADS-B data from an ADS-B
server, so the problem with duplicates is not
present.
The ADS-B time bias and the coordinate
transformation effect are negligible.
ADS-B data is based on GPS receiver and never in
the INS system of the aircraft.
The biases of SMR systems have been cancelled.
Thus, ADS-B is the primary surveillance source of
information which is supported by an SMR sensor (SMR1). So,
there are two independent monosensor tracks. The main one is
the one based on ADS-B and the one provided by the SMR is
the backup one. The following figure represents the RMS error
of both monosensor tracks for TR1 after performing a Monte
Carlo simulation of 100 iterations.
Figure 3. RMS error for the two sources of information (SMR and ADS-B)
Assuming that the results for TR1 are representative
enough, the integration of ADS-B will considerably improve
the quality of the output of the Surveillance Function provided
that the already known issues are conveniently addressed. In
order to achieve A-SMGCS level II, which implies the 7.5 m at
95% percentile of accuracy, the integration of ADS-B is almost
mandatory.
Besides, the output of the SMR sensor guarantees that its
performances also comply with the A-SMGCS level II
accuracy requirement, so its output can perfectly take over the
role of back up surveillance source. For those cases in which a
better performance is required, the output is improved by
including an IMM filter which exploits the airport layout at the
expense of increasing the complexity of the MSDF function.
REFERENCES
[1] SESAR Definition Phase, Deliverable 3. The ATM Target Concept,
September 2007.
[2] EUROCONTROL, Definition of A-SMGCS Implementation Levels,
November 2005.
[3] SESAR Development Phase, 12.03.01.D02. Phase1-Architecture
design, Ed. 00.01.00, January 2011.
[4] ICAO, Doc 9871 - Technical provisions for Mode S services and
extended squitter., 2008.
[5] G. de Vela et al., Integration of ADS-B surveillance data in operative
multiradar tracking processors, Information Fusion, 2008 11th
International Conference on, 2008, 1-8, 0.1109/ICIF.2008.4632309.
[6] RTCA, DO-242A, Minimum Aviation System Performance Standards
for Automatic Dependent Surveillance Broadcast (ADS-B), June 2002.
[7] SC 186 WG-6, Proposed Revisions to ADS-B MASPS: Integrity and
Accuracy Monitoring, August 2001
[8] J.A. Besada et al., On-line sensor calibration for airport data fusion,
Radar Conference, Proceedings of the IEEE, 2004.
[9] Luca Saini et al., 1090ES ADS-B Surveillance for vehicle tracking-
Performance Results, ESAVS 2007, Bonn, Germany, March 2007.
[10] Paul Askew, NATS, Evaluation of ADS-B at Heathrow for
EUROCONTROL ADS Programme Report, June 2002.
[11] A. Smith et al., System-Wide ADS-B Back-Up and Validation,
presented at the Integrated Communications, Navigation, & Surveillance
Conference, Hyatt Regency Baltimore, Baltimore, Maryland, 2006.
[12] ICAO, Guidance Material on Issues to be Considered in ATC Multi-
sensor Fusion Processing Including the Integration of ADS-B Data,
September 2008.
[13] J.D. Powell, C. Jennings, y W. Holforty, Use of ADS-B and
perspective displays to enhance airport capacity, in Digital Avionics
Systems Conference, 2005. DASC 2005. The 24th, vol. 1, 2005, 4.D.4-
4.1-9 Vol. 1, 10.1109/DASC.2005.
[14] Campbell, S.D.; Grappel, R.D.; Flavin, J.M., Multi-sensor processing
for aircraft surveillance in mixed radar/ADS-B environments, in Digital
Communications - Enhanced Surveillance of Aircraft and Vehicles,
2008. TIWDC/ESAV 2008. Tyrrhenian International Workshop on,
September 2008, 10.1109/TIWDC.2008.4649022.
18 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Future mobile satellite communication
Project 15.2.6: The feeder link to the ESA Iris Programme
Pierpaolo Tavernise
Thales Alenia Space Italia
Via Saccomuro 24 00131 Rome - Italy
pierpaolo.tavernise@thalesaleniaspace.com
Abstract SESAR Concept of Operations set out in the SESAR
Definition Phase (2006-2008) addressed the issue of the
inadequacy of the current ATM communication link identifying
new communication technologies that will enable improved voice
and data exchanges due to the estimated increase of the flight
traffic in the future. Currently SESAR WP15 addresses the non-
avionic CNS technologies development and validation comprising
the definition of the future mobile datalink systems to serve
communication services. Within WP15, the overall objective of
the Project 15.2.6 Future mobile satellite communication is to
define the requirements for the future Satellite Communication
System developed in the ESA Iris Programme, to perform the
complementary activities of validation and to support the
standardization of the new satellite link. The present paper
describes the status of P15.2.6 on going activities.
Keywords-SESAR; satellite; ATM; communication; SATCOM;
ESA; Iris
I. INTRODUCTION
SESAR Concept of Operations [1] and the ATM Master
Plan [4], endorsed by the EU Transport Council, indicates the
following solutions to meet the new long-term communication
requirements:
increase the use of digital technology and protocols. Data
link in the future will become the primary means of
communications [2] and voice while will not be able to be
used as back back-up will remain available for emergency
communications [6],
implement terrestrial and satellite based new data links to
complement VDL2 over high-density continental areas
[4],
provide satellite communication as the primary mean over
oceanic, remote and polar areas [5],
fully integrate the terrestrial and satellite networks
realizing a dual link system [3].
The new ATM satellite data link will moreover provide the
advantage to offer complementarities to the terrestrial links in
terms of coverage and infrastructure and radio spectrum
diversity.
II. THE MULTI LINK OPERATIONAL CONCEPT
Among the SESAR WP15 (Non avionic CNS Aspects)
projects, the project P15.2.4 (Future Mobile Data Link System
Definition) is a key project aiming to lead the definition of the
future communication system meeting the future requirements
for all phases of flight.
In particular SESAR Project 15.2.4 (P15.2.4) is in charge of
the definition of the overall system aspects of the Future
Communications Infrastructure (FCI) to be developed under
the SESAR Programme.
A new specific functionality has been identified as one of
the main concepts of the FCI with the objective to provide a
robust and high availability connectivity adapted to safety
critical application services. This specific functionality is
identified as the Multilink concept. The multilink concept
should enable the seamless use of different technologies
supporting the future ATM concept.
The SESAR ATM concept introduces new ATM services
that are demanding in data exchanges (latency, capacity,
availability, ) as
o 4D Trajectories Management, CDM
o Meteo info, SWIM,
SESAR introduced initially the Dual link concept [3] as a
way to limit the impact of events such as Loss of Service. The
contemporary availability and usage of multiple new data links
in a transparent way is the object of the multilink operational
concept (ML OC) defined in the P15.2.4 [10].
The Multilink Concept is involving three new
communications links:
AeroMACS - a ground-based, high-capacity,
airport surface datalink system for the aeronautical
mobile airport communications
LDACS - a ground-based datalink system for
continental airspace in L-band for digital
aeronautical communications
SATCOM - a satellite-based datalink system for
the oceanic, remote and continental environments
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 19
SESAR P15.2.4 is focused on the overall FCI and in
particular on the LDACS. SESAR Project 15.2.6 (P15.2.6) is
focused on the activities related to the new satellite datalink
(SATCOM). Following the SESAR vision, the Future
Communication Infrastructure (FCI) will be a system of
systems integrating existing communication sub networks
(VDL2) as well as the new communications sub networks
Figure 1: SESAR FCI Multilink Concept
Currently, there are two satellite communications system for
ATM services: the INMARSAT3 and IRIDIUM systems.
However, the performance requirements in the current ICAO
satellite standards are insufficient to cover the quality of
service (QoS) requirements of the applications supporting the
future operating concept. There is therefore a need to update
the AMS(R)S SARPs with more stringent performance-based
requirements in line with the requirements of the future ATM
concept.
In the European context, it is the European Space Agency
(ESA) with the Iris Programme that is leading the satellite
system design for ATM communication in collaboration with
the Sesar Joint Undertaking and in particular with the project
P15.2.6 .
In particular the Iris programme is complemented by the
SESAR P15.2.6 project (Future mobile satellite
communication) that is defined as the direct interface of
SESAR Programme to the ESA Iris Programme.
Figure 2: SESAR P15.2.6 and ESA Iris Programme
III. ESAIRIS PROGRAMME
The ARTES Element 10, Iris Programme, is the
European Space Agencys programme in support of the
implementation of safety-of-life communications via satellite
for the future European Air Traffic Management System
(EATMS) .
On the basis of the communications needs and the
technological solutions identified through SESAR definition
phase [1][2], ESA started the Phase II.1 of the Iris programme
which aim at designing, developing, verifying and validating
the new satellite communication system for ATM services
within the framework of the SESAR Programme.
Indeed the Iris Programme is financed by ESA Member
States but its implemented in close collaboration with SESAR
Joint Undertaking.
The collaboration is highlighted by the Iris Mission
Requirements definition and the validation of the satellite
system that are in charge to SESAR Joint Undertaking (SJU).
Furthermore SJU has been involved in the formal review
process of the ESA Iris projects contributing to the System
Requirements Review board with Eurocontrol, the European
Aviation Safety Agency (EASA), Airbus and an independent
ATM Safety Board.
The design and development of the satellite
communication system for ATM implies two parallel
developments:
o the development of a new satellite communication
standard to be adopted at worldwide level, designed ad hoc
for the future ATM services being also an open standard.
o the development of a European satellite infrastructure,
which will enable the provision of the future ATM
services service in the defined coverage area.
From a technical perspective, the ESA role with the Iris
Programme covers the design of the new Communication
Standard, the design and development of the Satellite
Communication System, the procurement of the subset and the
System Verification.
The Iris Phase II.1 has started in 2009, studying two
alternative approaches for the system design:
o the design of a purpose-built system (called ANTARES)
supported by three parallel studies preparing for future
service provision (HERMES, OPERA and SIRIO studies),
o the analysis of the feasibility of adapting Inmarsats
SwiftBroadband system for provision of safety services
(THAUMAS study). This activity includes modifications
of the SBB communication system, the analysis of
required modifications to the satellite constellation, user
terminals and the ground segment.
Based on the result of the Iris Phase II.1 studies, ESA will
submit a proposal for a Satcom system solution to SJU at the
end of the 2012. The ESA proposal will include different
technical options (based on ANTARES and THAUMAS
solutions) in terms of geographical service coverage and
system capacity.
20 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
The technical options will be supported by Business Cases,
Safety Cases and Security Case.
Based on all these elements, EC and SJU will reach a
consensus on the final design option of the Satcom solution for
SESAR.
A decision by ESA Member States is forecast in 2012 to
give the go-ahead for the Phase II.2 and complete the Iris Phase
II in full alignment with the SESAR Development Phase.
At the end of Phase II, all transferable ESA Iris Programme
assets will be transferred to the future owner of the system.
A Phase III of the ESA Iris programme foresees for ESA
the only role to provide technical support to the system owner.
.
Figure 3: ESA Iris Programme logo
IV. SESAR P15.2.6 AND THE FUTURE MOBILE SATELLITE
COMMUNICATION LINK
P15.2.6 is carried out by the following SESAR Members.
Airbus, AENA, Alenia, Eurocontrol, Frequentis, Indra,
NORACON and Thales.
Thales Alenia Space Italia (one of the 4 Thales Entities
involved in the project) is in charge of the project coordination.
P15.2.6 is part of the SESAR WP 15 the Non Avionic
CNS System addressing CNS technologies development and
validation also considering their compatibility with the Military
and General Aviation user needs and constitutes, as previously
indicated, the interface of SESAR Programme to the ESA Iris
Programme.
P15.2.6 is started in April 2010 and its completion is
foreseen in 2016 (inline with the ESA Iris and P15.2.4
schedule)
P15.2.6 aims to perform the following main activities:
1. Define the Iris mission requirements identifying the
satellite link operational concepts and define the
SATCOM system interfaces with the overall SESAR
infrastructure
2. Validate the ESA Iris development from an end-to-end
point of view in collaboration with P15.2.4 for the FCI
Operational Concept
3. Supporting the standardizations process related with
satellite communication system and promoting this
aviation technical standard to regional and international
standardizations bodies.
The definition of the new datalink requirements, the data
link validation and standardization is an activity implying a
coordination not only with Iris Programme and P15.2.4 but in
addition with the other SESAR WPs working on tasks
complementary to P15.2.6 tasks. In particular links have been
identified with the P15.1.6 for the satellite spectrum
assessment, WP 9 for the Aeronautical Flexible User Terminal
definition, WP 3 for the Validation and Verification exercises
coordination, WP 10 for the ground network, WP 14 for the
SWIM application and WP 16 for the Cost Benefit Analysis.
Figure 4: The 16 WPs of the SESAR Programme
Next figure summarizes the links between the Iris Programme ,
P15.2.6 and the other SESAR Projects and WPs:
Figure 5: Links between the Iris Programme and the other
SESAR WPs.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 21
Figure 6: Links between P15.2.6, the Iris Programme and
the other SESAR Projects
V. SESAR P15.2.6 IRIS MISSION REQUIREMENT
DEFINITION
Following the E-OCVM [12], the SESAR reference for the
ATM Concept Validation, the level of maturity of a new
concept as the usage of a new satellite datalink for the ATM
services implies an acceptance and a commitment from all
involved stakeholders. Moreover the expectations of all the
involved stakeholders have been discovered and collected in
the Mission requirement Document.
Once the MRD is issued, ESA will translate the Mission
Requirements generated by the stakeholders into requirements
applicable to the design and development of the Satellite
Communication System performed in the Iris Programme.
These requirements will be collected in the Iris System
Requirement Document (SRD).
The satellite system design is carried out in an iterative
process between Iris design studies and SESAR P15.2.6. Based
on inputs from P15.2.6, Iris Programme will consolidate
technical requirements in an iterative process consisting of the
following steps:
1. Mission (User) Requirements capture from the SJU
and the Aviation Community,
2. Translation of Mission Requirements into System
Requirements by ESA,
3. Assessment of impact of System Requirements on the
design of the Satellite Communication System
4. feed-back on such impacts and their high-level
consequence to SJU and the Aviation Community, leading to a
possible update of the Mission Requirements and a new
iteration cycle.
The following stakeholders have been identified for the
Satellite Communication System:
the SESAR JU is responsible for the definition of the
users requirements for the overall European ATM System, for
the apportionment of technical budgets requirements among its
different communication systems (Terrestrial System, Satellite
System, Airport network) and for the definition of the
interfaces between these systems.
The EC is responsible for establishing the future
regulation for the deployment of SESAR and for the provision
of services within the SES legislation.
ESA will be responsible for the management and
coordination of the design, development and verification
activities of the satellite communication system until the
validation phase. At the end of the development phase, ESA
will transfer the assets to the system owner.
The European Aviation Safety Agency (EASA) will
be responsible for future safety regulations for ATM, according
to the SES legislation.
Satellite and Aeronautical Manufacturing Industry
will develop the satellite system and the user terminals
Satellite Operators and Satellite Service Providers
(Satellite Communication Service Providers and
Communication Service Providers) will be in charge to
operate the satellite and provide the service to the final users of
the system (ANSPs, Airlines, General Aviation).
The Future System Owner is foreseen to be identified
at the end of the next phase of Iris Programme.
The International Civil Aviation Organisation (ICAO)
will be involved to approve and endorse the SARPs and
required updates to the AMS(R)S Technical Manual (a new
Part)
the Aeronautical Community (ANSPs, Airspace users
and industry associations such as IATA) as final users of the
system.
Representatives of all the stakeholders have been involved
in the MRD definition.
In particular SESAR P15.2.6 MRD definition task is
carried out with the involvement of the Satellite and
Aeronautical Manufacturing Industry (Thales, Indra,
Frequentis, Airbus), a Satellite Operator and Satellite Service
Provider (Telespazio is the affiliate of Alenia Member involved
in the project), 2 ANSPs (AENA and Avinor - member of
Noracon) and Eurocontrol.
Airspace Users contribution to the P15.2.6 MRD definition
task and SESAR projects more in general is also planned to be
obtained by SJU contracts with Air France, EBAA, KLM,
Iberia, the Lufthansa Group, Novair, SAS Scandinavian
Airlines, TAP Portugal, IATA, IAOPA and ELFAA [13].
In this way SJU has reinforced the user-driven approach to
technologies and procedures development including the
Airspace Users on the analysis of the outcomes of SESAR
projects [13].
In particular EBAA and SAS will be directly involved in
the MRD review.
22 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
The involvement of EASA is obtained by Iris Programme
where EASA is involved for the Expert technical safety advice
to ESA and for the chair of the ATM SATCOM (Iris) Safety
Board (ASSB) [14].
Satellite Operators and Satellite Service Providers are
involved in the Service Operators Studies (SIRIO, OPERA,
HERMES) of the Iris Programme.
Currently the mission requirements are under definition
within P15.2.6 and the Aviation Community. The main
requirements having strong impact on the Satellite system
design are reported in the following:
o the coverage area of interest for the provision of the
SATCOM services
o the list of ATC and AOC services to be included and the
reference performances (reference are COCR [15] and the
outcomes of P15.2.4 and other SJU external studies as
AOC Study and OPTIMI)
o the validation assumptions as the foreseen infrastructure
and the timeframe
o security features and countermeasure to the impact of
interferences on the mobile link.
Figure 7: SES Airspace and European FABs are
investigated by P15.2.6 as possible coverage areas
With the Iris MRD, P15.2.6 will deliver to ESA the Iris ICD
that will define the interface between the future European
ATM Network (EATMN) with the Iris SATCOM system.
VI. SESAR P15.2.6 ESAIRIS DEVELOPMENT VALIDATION
The reference in SESAR for the validation and verification
methodology is E-OCVM [12].
Validation is intended moreover as an iterative process
providing evidence that a new system or operational concept
fits for purpose or, in other words, validation answers the
question: Are we building the right system?
Verification provides proof that technology components are
feasible and can be safely and economically implemented or, in
other words, verification answers the question: Are we
building the system right?
Verification shall be conducted in parallel with validation
in order to discover problems early and to resolve them before
costly deployment. Indeed Validation and Verification (V&V)
are not considered in isolation by the E-OCVM but are
combined.
P15.2.6 is a System project and for this reason validation is
not part on the scope of the project that will be focused on the
System verification. However P15.2.6 will support the
operational validation exercises within WP3 and/or individual
Operational projects to be identified in the near future.
The current WP3 objective is to build a comprehensive
and integrated V&V Infrastructure to fulfil the overall
validation needs for the development of the SESAR ATM
system. The resulting V&V Infrastructure should be capable of
validating from a single ATM Service up to the whole SESAR
ATM Target Concept ....[16]
WP3 will provide support to the Operational projects for
the setting up of validation activities plus the specification and
development of specific validation tools.
Main tools are the IBPs representing the Industry
Based/Pre-Operational V&V Platforms.
Main tasks of the V&V activity performed in P15.2.6 are :
o Definition of the SATCOM system test procedures and
plan
o Specification of the Test bed(s) requirements
o Definition of the test bed architecture to integrate the Iris
SATCOM component inside the SESAR IBPs validation
platform
o Support to the integration of the SATCOM test bed
platform(s) developed in the Iris Programme in the FCI
Verification Test Bed for the end to end V&V activities
o Execution of the tests and collection of the results.
The boundaries SATCOM datalink defines the boundaries
of the P15.2.6 V&V activity respect the FCI V&V. The
assumption is that the V&V task of the satellite
communication system will be carried out between the
interface between the Satellite Communication Service
Provider and the Communication Service Provider on the one
end and the interface between the Satellite Data Unit and the
on-board router of the aircraft on the other end.
This is depicted in the following scheme (from [18]):
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 23
Figure 8: Boundaries of the Satellite Communication
System V&V
A possible architecture of the FCI Test bed platform for
V&V purposes is reported in the next figure (from [17]):
Verification platform and test controler
ATN/IPS ground infrastructure
mockup (including A/G router,
G/G router, etc)
LDACS
AGDL
emulator
ATN/IPS airborne infrastructure
mockup
SATCOM
AGDL
emulator
AeroMACS
AGDL
emulator
ATC/AOC
Traffic
Generator
Figure 9: A possible FCI Test bed platform architecture to
support V&V of the SATCOM datalink from an end to
end perspective.
P15.2.4 FCI test bed will be connected to Air Ground Data
Link AGDL emulators or prototypes developed in P15.2.4
(LDACS), P15.2.6/Iris (SATCOM) and P15.2.7 (AeroMACS)
P15.2.4 FCI test bed will support V&V of the SATCOM
datalink from an end to end perspective and in particular:
Verification exercises where the FCI architecture (and
multilink management) impacts the SATCOM datalink
performances ;
Validation exercises performed with the context of
WP3 or associated operational projects, where the performance
of the FCI impacts the satcom datalink and where the FCI and
the SATCOM in particular provides the communication
channel to be connected to the SESAR IBPs.
The SATCOM test bed could be used in addition (if agreed
in the test bed requirement document) for:
Demonstrations,
Certification exercises.
VII. SUPPORTING THE STANDARDIZATIONS PROCESS
Because aviation operates on a global basis, any new ATM
solution must be supported and coordinated on a worldwide
basis. This requires ICAO standardisation and coordination
with the international stakeholders leading to support for the
new standard.
For this reason SJU under the US/EC Memorandum Of
Cooperation agreed to establish a dedicated activity
(Coordination PLAN 4.4 Data-link technology) to coordinate
FAA the development of future communication technologies,
including the supporting avionics architecture and in particular
including developments of LDACS, Aeromacs and future
SATCOM technologies.
The activities in this Coordination Plan are very much linked to
providing input to aviation standardisation groups such as
ICAO (ACP), EUROCAE, RTCA and (maybe) SAE.
Furthermore P15.2.6 supports the standardization activities
and will drive the development of the required global ICAO
standards by the involvement of its Members in a dedicated
Working Group (WG). The international aspect of satellite
system standardization are dealt in the EUROCONTROL
NEXUS WG with contributions from all interested parties in
Europe and other interested Countries such as USA and Japan:.
NEXUS WG is a subgroup of Nexsat WG and is based on
voluntary contributions. EUROCONTROL is the facilitator
(rapporteur ) of the discussions in the group [19].
NEXUS is currently working in developing a proposal for
ICAO for an update of the AMS(R)S SARPs with stringent
performance requirements and in the future will be involved in
developing a proposal for an update of the AMS(R)S Technical
Manual with the specifications of a new SATCOM technology
meeting the updated requirements.
This imply a strict coordination between the work carried out
in the Nexus WG and the Mission Requirement definition.
VIII. CONCLUSION
The main objectives of the Project 15.2.6 Future mobile
satellite communication is to define the requirements for the
future Satellite Communication System that will be developed
by ESA in the Iris Programme, to perform the technical
validation of the Iris SATCOM system, support the
standardization of the new satellite link with aviation
standardization groups such as ICAO and EUROCAE.
A new concept as the usage of a new satellite datalink for
the ATM services as part of a more general multilink
24 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
operational concept implies an acceptance and a commitment
from all the users and service providers.
P15.2.6 aims moreover to identify, define and collect the
expectations of all the involved aeronautical stakeholders in
order to feed the ESA Iris Programme with requirements that
will bring ESA to develop a system corresponding to the
Airspace Users and ANSPs expectations for the future.
Operational validation of the SATCOM system is in charge
to SESAR ANSPs Operational projects by the methodologies
and tools made available by WP3. P15.2.6 will be focused
mainly on the system verification and will support the
operational validation process. Verification activities will
consist on defining the SATCOM system test procedures and
plan, the specification of the SATCOM Test bed architecture,
the support to the integration of the SATCOM test bed in the
FCI test bed and the execution and analysis of the related tests.
The standardization activities for the new SATCOM
datalink are supported by P15.2.6 with the involvement of
P15.2.6 Members in the Nexus Group aimed to deliver to
ICAO an agreed amendment to the ICAO SATCOM SARPs
(Annex 10 Vol. 3).
First issue of Iris MRD will be provided to ESA within
2011 and will be updated in a iterative way as soon as the FCI
concepts will be consolidated.
REFERENCES
[1] SESAR The ATM Target Concept (D3), DLM-0612-001-02-00a
(approved), September 2007.(Page 11)
[2] "Technology Assessment", SESAR Definition Phase Task 2.5.x -
Milestone 3. SESAR Consortium, 2007. DLT-0612-25x-00-05. (Sec.
1.5.1).
[3] SESAR The ATM Target Concept (D3), DLM-0612-001-02-00a
(approved), September 2007.(Page 52)
[4] European Air Traffic Management Master Plan Edition 1 - 30 March
2009 (Pag. 101)
[5] SESAR Deployment sequence (D4), DLM-0706-001-02-00 (approved)
January 2008 (pag. 77)
[6] Action Plan 17: Future Communications Study Final Conclusions and
Recommendations Report, EUROCONTROL/FAA, Version 1.1,
November 2007
[7] http://telecom.esa.int/iris
[8] ESA Iris Programme Status and proposed way forward Document
produced by ESA, EC and SESAR JU Version 20 September 2010
[9] Inter-Regional Satcom Voice Ad Hoc Task Force (SATCOM ad hoc TF)
First Meeting (Paris, France, 25 to 27 January 2011) Agenda Item 4:
Review of the status of implementation and available documentation
Status Of ESA Iris Programme
[10] Project P15.2.4 DEL EWA02-T1-D1 Multilink Concept: An Operational
Perspective edition. 00.00.08 (14/02/2011)
[11] The EUROCONTROL Skyway magazine - Number 54 - Winter 2010
[12] E-OCVM European Operational Concept Validation Methodology V.3.0
2010
[13] SESAR Magazine Issue #6 June 2011
[14] EASA Requirements for Service Providers of aeronautical satellite
mobile (en-route) Communications (CSP) Issue 1.0 10.12.2010 Iris-SB-
CP-TNO-0402-ESA-C2
[15] EUROCONTROL/FAA Communications Operating Concept and
Requirements for the Future Radio System Version 2.0
[16] SESAR Revision Framework for WP03 2nd June 2010 Edition 01.01.00
[17] SESAR P15.2.4 PIR 29/04/2011 Edition 00.01.00
[18] ESA Iris Programme Validation Assumptions draft 0.1
[19] Eurocontrol Nexus Terms of Reference v1.0
(www.eurocontrol.int/nexsat/public/standard_page/NEXUS.html)
[20] ICAO Document 9869, Manual on Required Communication
Performance (RCP) First Edition -2008
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 25
\
|
I
=
|
.
|
\
|
B
r
f
e
r
r v 1
2
) (
0
t
u
(1)
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 27
where
0
ADSBandWAMdeploymentinEurope
Session2.1page35
U.S.ActivitiesinADSBSystemsImplementation
Session2.2page41
DetectandavoidforUnmannedAircraftSystemsin
thetotalsystemapproach
Session2.3page47
NorthSeaHelicopterADSB/MLatPilotProjectFindings
Session2.4page53
LAOTSE,anApproachforForeignObjectDetection
bymultimodalnetted2D/3DSensors
Session3.1page61
MillimeterwaveRadarforRunwayDebrisDetection
Session3.2page65
OFDMWaveformsforaFullyPolarimetricWeatherRadar
Session3.3page69
Polarimetryappliedtoavionicweatherradar:
improvementonmeteorologicalphenomenadetectionandclassification
Session3.4page73
PrinciplesofUtilizationofPolarizationInvariantParametersfor
ClassificationandRecognitionofDistributedRadarObjects
PartI.Simplestmodelofadistributedobjectpaper
Session3.5page79
PartII.Multipointmodelandcorrelationtheory
Session3.6page83
(2)
Then, the range is calculated by (1). If the modulation
frequency is stable enough, the integration time can be chosen
comparably high to improve the signal-to-noise-ratio.
However, the choice of the modulation frequency f is crucial
as it directly constraints the maximum range. Only phase
differences of 2 0 < within one period can be used for
range calculation without loss of the unambiguity. Hence, the
unambiguity range is given by
TABLE I. TECHNICAL SPECIFICATIONS OF THE PMD CAMERA [3]
Parameter Value
Standard measurement range 0.3 m to 7 m
Repeatability (1) 3 mm
Frame rate
40 fps @ 200 x 200 pixels
60 fps @ 176 x 144 pixels
80 fps @ 160 x 120 pixels
Field of view 40 x 40
Illumination wavelength 870 nm
Power supply 12 V 10%
Interface USB 2.0
Operating temperature 0 to 50C
Storage temperature -20C to 85C
Figure 3. Optical image of the test objects (from top to bottom: wooden ruler,
spoon, nail file, screw driver, rubber, scissors, push pin, screws)
f
c
R
unamb
2
= (3)
Commercial ToF cameras often operate with a modulation
frequency of 20 MHz, yielding a 7.5 m long measurement area.
In order to increase this, signal processing techniques as
known from image processing for radar interferometry can be
applied to evaluate the phase information of adjacent pixels.
Such 2D phase unwrapping methods allow an unambiguous
phase determination over multiple periods [9]. Another option
would be to use multitone techniques as they are known from
automotive applications [10]. An overview about the technical
specifications is given in Table 1. The camera lens is
changeable and thus customised for different applications,
such as for automotive applications, for long range
surveillance or for medical applications (short range
surveillance).
Figure 4. Measurement of empty space (intensity plot)
Figure 5. Result (intensity plot) of various items located at the ground. The
data of the empty space had been removed.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 63
Figure 6. 3-D image of scissors shown in the figures above
B. Measurements
The combination of radar technology and optical systems
facilitate a reliable surveillance of runways. To demonstrate
this, a ToF camera from PMD Technologies GmbH was used
to image different objects and materials on a smooth plane
(Fig. 3). To demonstrate the potential of such a ToF system, a
870 nm PMD camera comprising of about 200 x 200 pixels
was used to image different objects and materials (see Fig. 5).
To obtain detailed and undistorted images a free-space
measurement (shown in Fig. 4) has to be carried out with the
camera system, which will be subtracted from the following
measurements. This is a similar idea as for the surveillance
radar which detects objects in the change detection sense. On
the one hand, only deviations of the surface of the runway are
imaged (change detection), on the other hand, the range
increase in dependency of the perspective as well as the
intensity distribution of the illumination spot on the ground can
be removed. The result (intensity image) is depicted in Fig. 6.
Different objects made of different materials have been
imaged. The objects can be identified easily by their shape.
This allows a reliable evaluation of the threat on the runways
by the airport personnel which can decide, whether a removal
is mandatory or not as in the case of a non-dangerous item like
a piece of paper. As described in the section above, ToF
cameras are able to measure objects three-dimensional. To
illustrate this, the scissors from the test set-up in Fig. 3 have
been selected exemplarily. Fig. 6 shows the result. The height
information is coded by the red colour. The bars as well as
blades are depicted in a clear way, such that the item can be
identified as scissors. The range deviation could be
compensated successfully as the object is imaged on a constant
ground height. It should be mentioned that both measurements
can also be carried out under foggy or rainy conditions making
the system very attractive for a non-stop operation.
V. CONCLUSION
An innovative solution for runway surveillance and debris
reconnaissance was proposed. Radar sensors as well as optical
three dimensional sensors were fused to get as much
information as possible and to achieve a high and reliable
detection rate. The system supports the airport personnel
responsible for the safety of aircrafts on runways during take-
off and landing by giving an alarm in the case of a dangerous
item on the runway. The radar system monitors the runway in
change detection mode and gives the fused information to the
optical system that images the item in more detail. Test
measurements showed the feasibility of this technique by
using a ToF camera. Both, the intensity plot as well as the 3D
range image enabled a clear identification of the objects.
REFERENCES
[1] A. Tessmann, A. Leuther, M. Kuri, H. Massler, M. Riessle, H. Essen, S.
Stanko, R. Sommer, M. Zink, R. Stibal, W. Reinert, M. Schlechtweg,
220 GHz Low-Noise Amplifier Modules for Radiometric Imaging
Applications, Proceedings of the 1st European Microwave Integrated
Circuits Conference, pp. 137-140, Sept. 2006.
[2] H. Essen, A. Wahlen, R. Sommer, G. Konrad, M. Schlechtweg, A.
Tessmann, Very high bandwidthmillimetre-wave radar, Electronics
Letters. Vol. 41, No. 22, Oct. 2005, pp. 1247 1249
[3] P.D.L. Beasley, G. Binns, R.D. Hodges,R.J. BAdley, Tarsier, a
millimetre wave radar for airport runway debris detection, Radar
Conference, 2004. EURAD. First European Radar Conference, 2004,
Amsterdam, pp. 261 264
[4] H. Essen, F. Lorenz, S. Hantscher, P. Warok, R. Zimmermann, M.
Schrder, W. Koch, M. Schikora, K. Wild, Millimeterwave Radar for
Runway Debris Detection; Enhanced Surveillance of Aircraft and
Vehicles, Capri, Italien, Sept: 2011
[5] C. Harris and M. Stephens, "A combined corner and edge detector".
Proceedings of the 4th Alvey Vision Conference 1998: pages 147-151.
[6] D.G. Lowe, Distinctive Image Features from Scale-Invariant
Keypoints, International Journal of Computer Vision, 60, 2, pp. 91-110,
2004. Millimeter Wave Radar for Runway Debris Detection, to be
published, ESAV 2011, Capri, Sep. 2011
[7] www.pmdtec.com
[8] T. Ringbeck, C. Schaller, M. Profittlich, Kameras fr die dritte
Dimension, Optik & Photonik, no. 3, October 2009 [in German
language]
[9] T. Ringbeck, B. Hagebeuker, Dreidimensionale Objekterfassung in
Echtzeit, AVN, no. 7, July 2007 [in German language]
[10] N. Fatihi, S. Hantscher, J. Rubart, C. Krebs, D. Nler, H. Essen
Imaging Permittivity Measurements for Automated Material
Inspection, Progress in Electromagnetics Research Symposium, March
2011
64 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Millimeterwave Radar for Runway Debris Detection
H. Essen, F. Lorenz, S. Hantscher, P. Warok, R.
Zimmermann, M. Schrder
Fraunhofer - FHR
D- 53343 Wachtberg, Germany
helmut.essen@fhr.fraunhofer.de
W. Koch, M. Schikora, G. Luedtke, K. Wild
Fraunhofer - FKIE
D-53343 Wachtberg, Germany
marek.schikora@fkie.fraunhofer.de
AbstractFor foreign object detection on runways highly
sensitive radar sensors give the opportunity to detect even very
small objects, metallic and non-metallic, also under adverse
weather conditions. As it is desirable for airport applications to
install only small but robust installations along the traffic areas,
millimeter wave radars offer the opportunity of small antenna
apertures and miniaturized system hardware. A 220-GHz radar
was developed, which is capable to serve this application if
several of these are netted to cover the whole traffic area.
Although under fortunate conditions the radar allows a
classification or even an identification of the debris, the complete
system design incorporates 3-D PMD-cameras for assistance in
the identification process, if the environmental conditions allow
for this. The latter part of the system is described in a separate
paper [3].
Keywords- FOD; mmW-radar; high resolution; netted sensors
I. INTRODUCTION
The detection and classification of runway debris is a major
concern for airport operation, as related flight delays and
accident claims as well as aircraft repairs are very costly. In
forthcoming years traffic on airports will even increase and
thus the demand on maintaining an adequate safety level. For
safe airport operation it is necessary, that the runways and
taxiways are permanently in an undisturbed condition.
Following the recommendations of the ICAO, the
inspection of runways is done at a period of 6 hours, visually
by security personnel at the time being. This procedure is
costly and may be subject to human failure and is also highly
dependent on weather and daylight conditions. Especially at
adverse weather the visual inspection cannot guarantee the
detection of small metallic particles like screws. Moreover the
traffic density on airports has been considerably increasing,
and a secure inspection between consecutive take-offs and
landings by human visual inspection is not possible.
A few automatic or semi-automatic systems have been
proposed or are in use at some airports to serve this purpose
[1]. Video surveillance is one of the proposals, but suffers
from obscuration by bad weather. Radar sensors, which are
capable to be operated under all weather conditions [2] are
also taken under consideration.
The radar proposed in this contribution is operating at
millimeterwaves with high bandwidth to achieve a very good
range resolution. The 220 GHz band is used, which in
comparison to lower frequencies allows considerably smaller
antennas with high directivity and thus a miniaturization of the
complete set-up. Further a wider bandwidth and thus a higher
ranger resolution can be accommodated at higher frequencies.
As output power at this frequency band is limited, a netted
approach, using several radars along the traffic area, is
proposed.
II. 220-GHZ LAOTSE RADAR CONCEPT
A. System Concept
The LAOTSE [3] concept is based upon a net of
miniaturized 220-GHz radar modules, which are positioned
along the traffic areas. This netted approach is necessary as the
output power of each radar is not sufficient to cover the whole
traffic area, and additionally the probability of false alarms is
considerably reduced for areas, which are monitored from
different aspects. The total LAOTSE system is additionally
using distributed PMD cameras, which are focussed to objects,
which have been detected by the radar [4] and which are also
part of the sensor net.
B. Radar Geometry
Tests of an experimental LAOTSE system, with a limited
number of sensors are conducted on the Cologne-Bonn
Konrad-Adenauer Airport.
Figure 1. Photo of Cologne-Bonn Airport with LAOTSE Site, old Runway
(below) and Cross-Wind Runway
The first LAOTSE radar module is mounted at a slightly
elevated position near to the crossing of the old runway and
the cross-wind runway to be able to overlook a certain area.
The project is financed by the Ministry for Innovation, Science and Tech-
nology of the German Federal State of North-Rhine Westphalia under the
ZIEL2 Programme of the European Fund for Regional Development (EFRE)
LAOTSE
Test site
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 65
Fig. 1 shows the situation, Fig. 2 demonstrates the geometry.
Fig. 3 shows the illumination by the antenna beam.
Figure 2. Typical Geometry for Radar above Runway
Figure 3. Illumination of Traffic Area by one Radar Module
To miniaturize the system, the 220 GHz frequency band is
used, which allows to reduce the antenna diameter by more
than a factor of two. This is essential for operation in an
airport environment, where obstacles like the netted sensors
along the runway have to be avoided. This approach, however,
demands for a wider net of sensors to cover the complete
runway area. For complete demonstration of LAOTSE
performance a second sensor will be installed along the old
runway to maintain an overlapping survey region.
In the following, the approach for a single radar module is
described.
C. 220-GHz Sensor Front-End
The system concept is based on a design, which was
already realized at 94 GHz [5]. The radar FM-CW front-end
uses miniaturized, monolithically integrated components in
GaAs technique [6]. To be capable to cover a bigger range,
like 200 m, coherency of the signal is most essential.
The classical FM-CW approach uses a reference signal,
which is derived from a voltage controlled oscillator (VCO),
which generally is not stable enough to maintain the necessary
coherency over a range > 200 m. This is due to the fact, that
for tuning of the oscillator the Q-factor of the resonator must
not be too high, which may lead to an uncontrolled frequency
drift. The drift can be so high, that during the transmit-receive
cycle the phase of the transmit signal has been changing so
much, that the mixing between reference- and receive- signal
does not result in a sharp frequency response but to a widening
of the receive signal. This widening increases with range and
at a certain range limit no resolution can be achieved. As this
concept does not allow, to achieve a sufficient range, a
concept as sketched in the block diagram of Fig. 3 was
developed. The output frequency is shifted to 105 GHz instead
of 94 GHz and a subsequent doubling results in an operating
frequency around 210 GHz.
Figure 4. Block Diagram of 220-GHz Front-End
All system frequencies are generated or derived from a
stable crystal oscillator. A dielectric, resonator stabilized
oscillator (DRO) generates the reference frequency of 17.2
GHz upon which the radar waveform, a linear FM-chirp with a
bandwidth of 220 MHz 380 MHz is mixed. A multiplication
by a factor of 6 leads to a basic millimetre wave frequency
band of 104.52 GHz 105.48 GHz. After amplification this
signal is split into the transmit branch and the local oscillator
branch. The latter is used for downconversion of the receive
signal using a subharmonic mixer. The transmit signal is
derived by doubling to result in a frequency band of 209.04
GHz 210.96 GHz, a total bandwidth of 1.92 GHz, equivalent
to a range resolution of about 8 cm.
The RF front-end is quite compact and avoids long
waveguide connections. Fig. 5 shows a photo.
Figure 5. Photo of LAOTSE 220-GHz Front-End
The complete radar RF part is mounted within the primary
focus of the Antenna. IF and DC connections to the data
RADAR
66 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
acquisition and control box are maintained by rotary joints
leading through the centre of the antenna pedestal.
For the first experiments a self contained steering and data
acquision box was developed, which is capable to operate one
single LAOTSE sensor, to store the measured radar and
distribute all relevant data into the sensor network.
For the operational system only a miniaturized data-
interface box will be assigned to each radar sensor. An FPGA
based processing electronics will deduce high resolution range
profiles by applying an FFT algorithm , which contain the
information on RCS and position of debris objects, and which
serve as a basis for further netted processing and the change
detection algorithm.
D. Antenna and Pedestal
As the FM-CW principle is used for the LAOTSE radar,
special care has to be taken for a good isolation between
receiver and transmitter, which is essential for a good dynamic
range of the system. For an optimum performance, separated
antennas were used. To allow a wide operation bandwidth,
which is necessary for a high range resolution, offset feed
antennas were used. To avoid long waveguides serving the two
antenna feeds, a design as shown in Fig. 6 was used.
Figure 6. Sketch of Antenna Assembly and Photo of Radar upon Pedestal
Each radar module is mounted upon an antenna pedestal
capable to survey 360 angular region in azimuth and + 20 in
elevation. During operation the elevation is fixed according to
the geometrical conditions. Fig. 6 shows also a photo of a radar
module on its pedestal.
III. TEST MEASUREMENTS
A. ISAR Measurements on small Objects
To test the sensitivity of the system and to evaluate the
range under which small items can be detected, measurements
with an assembly of 25 different small objects on a turntable at
a range of about 200 m were conducted. The objects were
turned over 360, and the measured data were evaluated using
an Inverse Synthetic Aperture (ISAR) algorithm, which, in
contrary to the final LAOTSE algorithm gives a synthetic radar
image of each object. Tab. 1 gives a catalogue of the measured
objects.
TAB. 1
CATALOGUE OF MEASURED ITEMS
Objects
Tools Metal Pieces dielectric Pieces
pliers Wire bundle plastic 100mm x
40mm x35 mm
screw driver screw M8x60 plastic tray 2 shelf
folding ruler metal bar 30mm
x30mm x 30mm
plastic tray 3 shelf
file wheel d = 75mm wood pieces
100mm x 40mm x
20mm
saw antenna fixing stone
paint brush metal plate 200
mm x 15 mm
cable channel
(plastic)
Fig. 7 shows a photo of the assembly of parts on the
turntable.
Figure 7. Photo of Sample Pieces on Turntable
Before the measurements, the system was calibrated by
means of a trihedral corner reflector, which was positioned in
the middle of the turntable. This allows a thorough phase
calibration, which is necessary for the ISAR process and at the
end allows to assign a distinct radar cross section to each item
on the turntable.
Figure 8. Accumulated ISAR Image for 360 Aspect Angle Range
The evaluation was done using different radar bandwidths
ranging from 2 GHz to 8 GHz, which corresponds to a
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 67
resolution of about 2 cm to 8 cm. Fig. 8 shows an ISAR image
accumulated over the full 360 aspect angle range, which
means, that each objects was measured from all sides, and
which is more related to the approach of netted sensors, which
also allow to see an object from considerably different aspect
angles.
More realistic for the result for a single scanning radar
sensor is an evaluation, where only a limited fraction of aspect
angles is evaluated. Fig. 9 shows an image related to that of
Fig. 8 with an evaluation aspect range of only 10.
Figure 9. ISAR Image for 10 Aspect Angle Range
The evaluation showed, that all small items could be
detected and imaged independent of aspect angle. Only for dry
wooden blocks this did not apply, as they could be seen only
at those angles, where they were facing the radar. If they were
only a little bit humid, they were also visible for all aspect
angles. Resolution does not play an important role, the
detectability is not influenced for a resolution range between 2
and 8 cm. It can be concluded, that for an operational system a
resolution of about 10 cm is fully sufficient.
B. Test Measurements with the LAOTSE Radar
While for the test measurements, described above, the radar
was fixed and the objects were turning to be visible for all
aspect angles, further tests were conducted with the
operational geometry of a scanning radar. By chance the
weather was very bad with low visibility during the
measurements. Fig. 10 shows a photo of the measurement
scene. Fig. 11 shows the result of a sector scan. Clearly the
fence surrounding the terrain can be identified by its very
strong radar echo (red) and the returns of some small items in
front of the fence.
Figure 10. Photo of Test Scenario
Figure 11. Radar Image for Sector Scan over Scene shown in Figure 7
IV. CONCLUSION
Radar modules at a frequency of 220 GHz were developed
as elements for a netted set-up along traffic ways on airports. It
was demonstrated, that also under adverse weather conditions
small objects, metallic and non-metallic could be detected.
Further research is done on sensor data fusion among several
radar modules and additional 3-D IR cameras [4] using
photonic mixer devices technology. This topic is highlighted in
a separate paper.
REFERENCES
[1] P.D.L. Beasley, G. Binns, R.D. Hodges,R.J. BAdley, Tarsier, a
millimetre wave radar for airport runway debris detection, Radar
Conference, 2004. EURAD. First European Radar Conference, 2004,
Amsterdam, pp. 261 - 264
[2] I. J. Patterson, Foreign Object Debris (FOD) Research, International
Airport Revue, Issue 2, 2008
[3] H. Essen, G. Luedtke, P. Warok, W. Koch, M. Schikora, K. Wild,
Millimeterwave Radar Network for Foreign Object Detection, 2
nd
Intl.
Workshop on Cognitive Information Processing (CIP), Elba, June 14
16, 2010
[4] T. Ringbeck, A 3D Time of Flight Camera for Object Detection,
Optical 3-D Measurement Techniques 09-12. 07. 2007, ETH Zurich
[5] H. Essen, A. Wahlen, R. Sommer, G. Konrad, M. Schlechtweg, A.
Tessmann, Very high bandwidthmillimetre-wave radar, Electronics
Letters. Vol. 41, No. 22, Oct. 2005, pp. 1247 1249
[6] A. Tessmann et al. 220 GHz Low-Noise Amplifier Modules for
Radiometric Imaging Applications, Proceedings of the 1st European
Microwave Integrated Circuits Conference, pp. 137-140, Sept. 2006.
68 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
OFDM Waveforms for a Fully Polarimetric Weather
Radar
O.A. Krasnov, Z. Wang, R.F. Tigrek, P. van Genderen
Delft University of Technology
International Research Centre for Telecommunications and Radar (IRCTR)/ MTSRadar
Delft, The Netherlands
AbstractRetrieval of cloud parameters in weather radar
benefits from polarimetric measurements. Most polarimetric
radars measure the full back scatter matrix (BSM) using a few
alternating polarized sounding signals. Using specially encoded
orthogonal OFDM signals however, the BSM can be measured in
a single simultaneous transmission of two orthogonally polarized
signals. Based on a set of parameters for weather radar, the
properties of such a signal are explored and its merit as a useful
capability is shown.
Keywords; weather radar; polarimetry; OFDM; coded
waveforms.
I. INTRODUCTION
Weather radar is an important source of information for air
traffic control and, in general, for meteorological services and
research of the environment. For retrieval of parameters on the
water content of clouds, it is important the measure the full
backscatter matrix (BSM) of clouds, with a highly detailed
analysis of the Doppler shifts. Currently, various radars are
capable of measuring the BSM, but the waveforms used consist
of transmitting sounding signals in two orthogonal basis
directions alternatingly. So in order to measure the full BSM
two consecutive transmissions are needed. By consequence,
some time has passed in between the two transmissions and
due to the decorrelation of the echoes of the hydrometeors the
quality of the parameter retrieval is somewhat compromised.
The research addressed in this paper explores orthogonal
signals, such that the sounding signals in the two orthogonal
basis directions can be transmitted at the same time, so that the
BSM can be measured using a single though complicated -
transmission. Also in [1] the measurement of the BSM using
orthogonal signals has been addressed.
IRCTR has developed a radar platform called PARSAX
(Polarimetric Agile Radar in S- and X-band; currently the radar
is being set to work in S-band) supporting to do such analyses
with a variety of waveforms [2]. In [3] and [4] the problem of
simultaneous transmission of orthogonal signals has been
addressed using various modes of Frequency Modulated
Continuous Wave (FMCW) modulation. The current paper
explores a particular pulse coded waveform, Orthogonal
Frequency Division Multiplexing (OFDM). It will be
demonstrated that such waveforms can be used, albeit with a
number of constraints.
The paper is structured in the following way. Chapter 2 will
detail the OFDM sounding signal both in a generic way and in
the specific coding that is used for the benefit of the isolation of
the two orthogonal components. Chapter 3 discusses a set of
parameters applicable to this type of weather radar. Chapter 4
addresses the effects of Doppler shift. Doppler shift is a
parameter that is of vital relevance for the retrieval of the cloud
parameters. However, in OFDM is the major cause of loss of
orthogonality. Chapter 5 concludes the paper.
II. SIGNAL DESCRIPTION
OFDM is a waveform that is widely used in communication
[5]. It is being considered for application in radar only recently
[6], inspired by the availability of signal generators due to the
application in communication. The signal is composed of a
number of carriers at a mutual, constant spacing that is the
inverse of the signals duration. The complex baseband
description is:
( ) ( )
( )
1
0
2 0
0
N
k
k
s t a exp j tk / f t T
s t
=
= <
=
, elsewhere
(1)
where N is the number of carriers,
k
a
is the complex
amplitude of carrier k (also called the code of the carrier) and
f is the frequency spacing between the carriers. T is the
duration of the signal, 1 T f = .
All carriers composing this signal are mutual orthogonal, as
can also be seen from Fig.1 representing a detail in the power
spectral density of an OFDM signal.
Usually the duration T of the signal is quite long and
therefore some type of pulse compression is applied in order to
achieve an appropriate range resolution. Given the nature of the
OFDM signal, the most widely known procedure is to
transform the received signal to the frequency domain and
multiply the carriers with the complex conjugate of the Fourier
transformed transmitted signal. Then after this multiplication
the time domain range profile is generated by the inverse
Fourier transform. The level of the sidelobes in this range
profile can be managed by any weighting technique.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 69
The carrier codes
k
a
can be chosen independent from each
other. They have an important impact on the signal envelope in
the time domain to the extent that the ratio between the peaks
and the average of the power is affected. Certainly, the
envelope is not constant and the transmitter cannot be driven
into saturation as it usual in radar, reason for radar engineers to
be reluctant to apply this type of waveform. Special coding
optimizes the peak to average power ratio [6, 7].
One of the comments on the Doppler sensitivity of the
OFDM signal is that as soon as the Doppler shift is no longer
very small compared to the carrier frequency spacing, the
sidelobes of the compressed signal come up rapidly and the
orthogonality is lost. When the carrier codes are random
variables, the sidelobes will ultimately be proportional to 1 N .
When on the contrary, the codes are chosen to be equal, the
loss of the sidelobes will be negligible, but the peak to average
power ratio will be equal to N , so the signal envelope will be
very spiky.
The concept of the fully polarimetric radar is based on an
approach as shown in the block schematic diagram of Fig. 2. It
is shown there that one of the two signals is exciting one of the
two polarization directions, while the other signal is exciting
the orthogonal polarization direction. Then on receive, the two
signals are separated in the pulse compression process by using
the two complex conjugated transmitted waveforms as a
reference. If no special precautions are taken, the cross
compressed signal, i.e. the signal after application of the code
of the other signal for the purpose of the compression, will be
noise-like. The resulting average signal power is called here the
isolation of the codes. An important case for achieving high
code isolation can be appreciated by comparing Figs 1a and 1b.
a. two orthogonal sets of carriers, all
having a different random code
b. two orthogonal sets of carriers,
one with odd, the other with even
carriers. All carriers have a different,
random code.
Figure 1. Power spectral density of OFDM waveforms. Subplot a shows a
regular set of OFDM carriers. Subplot b shows a set of two special cases,
having either only the odd frequencies or the even frequencies. Their code
isolation when the Dopplershift is zero is perfect.
The underlying assumptions are that in Fig1a the two
orthogonal signals show a full spectral occupancy, i.e. all
values of
k
a are non-zero, and additionally by assuming that
they are all have the same amplitude but a random phase in
[ ) 0 2 ,
. In Fig.1b the sounding signals are different to the
extent that transmitter 1 is using all carriers with even values of
k and transmitter 2 is using all carriers with odd values of k .
The output of the compression process of a signal received in
one of the two channels using the code of the other channel
consists of the code isolation. It will be illustrated later in this
paper that it is proportional to
1 N
in the first case (two fully
populated sets of carriers) and close to perfect isolation in the
second case (two disjoint sets of even and odd carriers
respectively).
According to this line of thinking, the signal used for the
experimentation for the fully polarimetric radar is defined as
( ) ( )
( ) ( ) ( )
( )
2 1
2
0
2 1
2 1
0
2 2 0
2 2 1 0
0
N /
H k
k
N /
V k
k
s t a exp j t k / f t T
s t a exp j t k / f t T
s t
+
=
= <
= + <
=
elsewhere
,
(2)
Here
( )
V
s t represents the sounding signal exciting the
vertical-polarization and
( )
H
s t the horizontal polarization.
Figure 2. Simplified block schematic diagram of the radar.
(the symbols refers to coding/decoding)
OFDM1 OFDM2
R
H
H
V
V
H
V
s
VH
s
HH
s
VV
s
HV
Transmitter
Receiver
70 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
An important side effect of this coding however is that the
unambiguous range is half the value of the full set of carriers.
Also, the method for reducing the peak to average power ratio
fails in this coding.
III. A SET OF PARAMETERS FOR WEATHER RADAR
In the application of weather radar as described in the
introduction, a high resolution of the speed is needed, a typical
value being
0 1
r
v . m / s ec =
. For an X-band radar, this
corresponds to a Doppler resolution of 6.6Hz and thus on a
time interval of evaluation of the phase of 0.15sec. In order to
appreciate the significance of the Doppler resolution, let us
develop a benchmark set of radar parameters compliant with
this requirement.
For the application of detailed weather measurements using
the X-band radar, the maximum range should be not less than
10km. Since in OFDM the maximum range corresponds to
2
u,max
R c f =
, and given the effect of the odd/even coding, this
translates into a frequency spacing
3
7 5 10 f . Hz < . Because
the duration of this carrier spacing is directly linked to the time
duration of the signal by
4
1 1 33 10 T f . s
= = , also the
pulse repetition frequency is an immediate consequence:
3
7 5 10 PRF . Hz = . Hence, using a Doppler filterbank, the
number of pulses to be integrated in order to arrive at the
required Doppler resolution is
3
0 15 7 5 10 1125
p
N . . = =
. The
number of carriers affects two parameters at the same time: the
bandwidth,
f
B N f = , and the sidelobes of the pulse
compression in case of too high a Doppler shift, so in case of
loss of orthogonality, SLL=1/N
f
. The range resolution of
weather radar is a compromise between sensitivity (given the
number of rain droplets per resolution volume) and detail in
finding the structure of clouds. A fair number is to assume that
it should not be lower than 3m, hence
( ) 2 50 B c R MHz = = and the corresponding number of
carriers is N
f
=6667. The correlation noise floor, which is the
mean value of the sidelobes of the pulse compression in the
case of total loss of orthogonality, then is 38 SLL dB w.r.t.
the peak value in case of full orthogonality.
IV. THE EFFECTS OF DOPPLER SHIFT
The concept of orthogonality of the various carriers in the
OFDM waveform is compromised if the carriers are shifted
due to Doppler. In communications this effect is called Inter
Carrier Interference (ICI). Effectively the impact of Doppler
a. Correlation noise b. Code Leakage
Figure 3. Comparison of the correlation noise and the code leakage of the two
full sets of carriers (solid lines) and the odd/even carriers (dashed lines).
shift depends on the ratio of the Doppler shift to the frequency
spacing between the carriers. In case of randomly coded
carriers, the pulse compression in the receiver will fully fail as
soon as the Doppler shift equals the carrier spacing.
Because of the same reason of loss of orthogonality, also
the code isolation is affected by a Doppler shift and it is getting
more reduced the higher the ratio between Doppler shift and
carrier spacing is.
In the case of fully random carrier coding, the correlation
noise floor is described by [8]:
( )
( ) ( )
( )
1 1 1 1
2
2
0 0 0 0
2
1
2
f f f f
N N N N
f
k l m n f
f
m k n l
exp j
N
N
n l
exp j
N
= = = =
=
(3)
Here
d
f f = is the ratio between the Doppler shift and the
carrier spacing. This expression is valid for the fully occupied
set of carriers. A graphical representation is in Fig.3a, while in
Fig.3b the code leakage is presented as a function of the
relative Doppler shift . In these graphs also the
corresponding graphs are shows for the odd/even carriers
waveform.
These graphs demonstrate two major effects, the correlation
noise and the code leakage.
A. Effects on the correlation noise floor
The correlation noise, consisting of the range sidelobes in
the co-polar channel due to the Doppler shift, is very low for
low Doppler shift. More precisely, if the Doppler shift would
be exactly zero, the profile of the range sidelobes would
correspond to the weighting function used in the pulse
compression. Just off the zero Doppler values, the waveform
with all carriers used has a slightly lower correlation noise floor
than the odd/even carriers waveform, as a consequence of the
higher number of carriers. In all other cases, composing the
greater part of the span of Doppler values, the odd/even
waveform is superior.
B. The code leakage
Regarding the code leakage, consisting of the signal having
code 1 leaking into the other channel and there compressed
using the code 2, the odd/even waveform is superior to the
waveform with all carriers used for the Doppler shifts of
interest. Obviously, this was the reason why this waveform was
devised. Such a property is extremely important. It supports to
effectively filter away the strong echoes from stationary ground
objects, that otherwise might clutter the meteorological echoes.
It should be reminded though that at system level the total
leaked power from one channel into the other channel is not
just due to code leakage, but also due to other system
components, e.g. the antenna. A typical number for co/cross
polar antenna isolation is 25-30 dB for parabolic antennas.
Calling
A
I the antenna co/cross polar isolation and
C
I the
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 71
code isolation, preferably the inequality
max p A C t
S G I I N + < should hold (all values in dB). Here
max
S is the maximum signal that is within the linear dynamic
range of the receiver,
p
G
is the gain due to the processing, and
t
N is the level of the thermal noise. Using for instance
70
max
S dB =
with respect to thermal noise,
( ) 10 1125 30 5
p
G log . dB = =
,
30
A
I dB =
then it follows
70 5
C
I . dB < . Such a value is achieved for Doppler shifts close
to zero, but not for
0 02 150 2 25
d
f . f Hz . m / sec > =
. Thus
stationary discrete objects, also if some internal motion is
present, dont pose a problem. Hydrometeor echoes themselves
will be much weaker than the discretes and dont pose a
problem either.
V. DISCUSSION AND CONCLUSION
Using OFDM-coded waveforms tailored to minimize
leakage of signals from one channel out of two orthogonally
polarized channels into the other one is considered to be very
promising. Such coding outperforms any other type of coding
identified so far on the figure of merit for isolation of the
codes. Apart from the straightforward coding as shown in this
paper, also other codes can be devised, always based on the
orthogonality of the carriers in the waveform. A recent
experimental result obtained using the PARSAX radar for the
simultaneous transmission of an odd/even coded pair of
OFDM-signals as in (2) is shown in Fig.4. It shows the range
profiles of each of the four elements of the BSM for a single
OFDM-sweep. The number of carriers is N=25000 and the
total bandwidth is B=50MHz. The carriers are random QPSK-
coded. The antenna is pointing at the horizon while the weather
was clear. The reflections in this example therefore only
concern ground objects. The vertical scale is adjusted to have
the highest value (actually the object concerned is a tall
chimney) at 0dB. The floor of the signal level is close to the
dynamic range of the receiver, approx. 70dB. More details are
in [8].
The limit of applicability of the coding comes from the loss
of orthogonality, as it can be caused for instance by the
Doppler effect. Waveforms like Linear Frequency Modulation
are far more tolerant to Doppler, but cannot provide the
isolation between the codes as discussed here unless the
orthogonal signals are timed in a partially overlapping order,
like proposed in [4].
An effect that has not been addressed in this paper, but that
is highly important for keeping the carriers orthogonal while
receiving echoes with an unknown delay, is the necessity to
introduce a cyclic prefix as an extension to the duration T of a
single pulse. This effect has been discussed in detail in [9]. It
doesnt invalidate the conclusions.
0 5 10 15
-80
-60
-40
-20
0
HH
range, KM
a
m
p
lit
u
d
e
,
d
B
0 5 10 15
-80
-60
-40
-20
0
VH
range, K
a
m
p
li
t
u
d
e
,
d
B
0 5 10 15
-80
-60
-40
-20
0
HV
range, KM
a
m
p
l
it
u
d
e
,
d
B
0 5 10 15
-80
-60
-40
-20
0
VV
range, KM
a
m
p
l
it
u
d
e
,
d
B
Figure 4. Rangeprofiles of the four elements of the BSM. The polarization
basis is linear, V/H
In this paper the orthogonality of codes was exploited for
the benefit of measurement of the BSM in fully polarimetric
radar. The concept of orthogonality might also be used in
different applications, for instance for the benefit of retrieving
the echoes from multiple beams generated by phased array
antennas.
REFERENCES
[1] D. Giuli, M. Fossi, L. Facheris, Radar target scattering matrix
measurement through orthogonal signals, Vol.F of the Proc. of IEE,
Vol.140, issue 4, 1993, pp233-242
[2] O.A. Krasnov, L.P. Ligthart, Z. Li, P. Lys, F. van der Zwan, The
PARSAX Full Polarimetric FMCW Radar with Dual-Orthogonal
Signals, Proc. EuRAD 2008, Oct. 2008, Amsterdam, The Netherlands,
pp. 8487.
[3] G. Babur, Processing of Dual-Orthogonal CW Polarimetric
RadarSignals, PhD thesis, TU Delft, 2009.
[4] G. P. Babur, O.A. Krasnov, L. P. Ligthart, Quasi-Simultaneous
Measurements of Scattering Matrix Elements in Polarimetric Radar with
Continuous Waveforms Providing High-Level Isolation in Radar
Channels Proc. EuRAD2006, 30 Sep-2 Oct 2009, Rome, Italy, pp 1-4
[5] R. Prasad, OFDM for Wireless Communication, Artech House, 2004
[6] N. Levanon, Radar Signals, Hoboken, NJ, John Wiley&Sons, 2004
[7] M.R. Schroeder, Synthesis of Low-peak-factor Signals and Binary
Sequences with Low Autocorrelation, IEEE Tr. Information Theory,
Vol.16, No.1, January 1970, pp 85-89
[8] Z. Wang, R.F. Tigrek, O.A. Krasnov, F. van der Zwan, P. van Genderen,
A.Yarovoy, Application of I-OFDM Signals for Simultaneous
Polarimetric Measurement, Proc. EuRAD2011, 12-14 Oct 2011,
Manchester, UK
[9] R.F. Tigrek, A processing Technique for OFDM-Modulated Wideband
Radar Signals, PhD thesis, TUDelft, 2010
72 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Polarimetry applied to avionic weather radar:
improvement on meteorological phenomena detection
and classification
Alberto Lupidi
1
, Christian Moscardini
2
, Andrea
Garzelli
3
, Fabrizio Berizzi
4
, Fabrizio Cuccoli
5
CNIT-RaSS (National Interuniversitary Consortium for
Telecommunication-national laboratory of radar and
Surveillance Systems)
Italy
{
1
a.lupidi,
2
c.moscardini,
4
f.berizzi}@iet.unipi.it
3
andrea.garzelli @dii.unisi.it
5
fabrizio.cuccoli @cnit.it
Marcello Bernab
SELEXGalileo S.p.A.
Campi Bisenzio-Italy
marcello.bernabo @selexgalileo.com
AbstractAvionic Weather Radar is an essential equipment in
aircraft. Polarimetry can improve the detection and the
classification of hydrometeors and thus the safety and the
efficiency of the flight. Here a 3D polarimetric radar simulator
for the feasibility study on avionic weather polarimetric radar is
presented.
I. INTRODUCTION
In current avionic systems is impossible to distinguish the type
of precipitation, water, snow, hail. Of course, assumptions can
be done, i.e., high reflectivity in a zone where temperature is
15-20 degrees below zero is likely to indicate an hailstorm, but
we can have no precise information on type of precipitation
near and below the melting height (which also depend on
season and geographic region). About 70% of the high-
reflectivity echoes that pilots see on their radar is non-
hazardous (other than causing a decrease in visibility and
making runways wet). To determine whether or not a particu-
lar red echo is hazardous in terms of turbulence and hail and
other dangers, the pilot must first know if the atmosphere in
which he is flying is conducive to of hail and high turbulence.
It is worth noting to recall that heavy rain without turbulence is
not an issue for the safety of the flight. But even with
atmospheric knowledge, a pilot cannot say whether a particular
high-reflectivity area is hazardous. Usually, the pilot evades
that area, with an increase of costs, time and polluting
emissions due to the detour. The use of polarimetry can help
giving us more precise details on hydrometeor types [1].
For example, rain tends to have an elliptical form with
minor axis oriented vertically, resulting in HH signal to be
higher than VV signal thus having a positive high Differential
Reflectivity. On the contrary, hail, due to its tumbling motion,
appears as spherical, thus having a nearly zero Differential
Reflectivity, even at higher reflectivity (and higher hazard)
level. Classification algorithms which utilize the polarimetric
information on the three channels (HH, VV, HV/VH) can be
developed with the knowledge of Total (Z) and Differential
Reflectivity (Z
DR
) and Linear Depolarization Ratio LDR
defined as
DR HH dBZ VV dBZ
Z Z Z = (1)
VH dBZ HH dBZ
LDR Z Z = (2)
In this work we assume X-Band based system (around 9-
10 GHz) that are preferred because they have an antenna
whose dimensions are compact and compatible also for
business aircrafts. Polarimetric classification algorithms for
ground based S-C bands systems already exist and in general
there is no great difference between ground based and airborne
operation in the application of these algorithms [2], [3].
Differences arise from the technical limitations of the airborne
system, like antenna size, transmitted power and scanning
speed. Main issues for avionic weather radars in conjunction
with the use of X-band are:
1. heavy beam path attenuation and Mie scattering effects
2. ground clutter
3. wider beam width
4. data availability
In this paper we did not deal with path attenuation and ground
clutter. These problems will be addressed in future works. We
solved the problem of data availability simulating real radar
data with a physical based approach described later. Section II
describes the scenario and the mentioned approach, while in
section III and IV we show some results and conclusion
respectively.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 73
II. ATMOSPHERIC SCENARIO AND RADAR MODEL
One of the problems in weather radar engineering is the
availability of data. Moreover, in radar meteorology, data are
available mainly in S-band (around 3 GHz) because this is the
band chosen for ground based weather radar. To simulate
realistic polarimetric radar data in X-Band, the two most
important things we need to know to compute the radar
reflectivity are:
1. the Drop Diameter Distribution (DSD) of hydrometeors
N(D) measured in m
-4
2. their polarimetric Radar Cross Section (RCS) o
H,V
(D)
measured in m
2
.
Polarimetric reflectivity is finally computed as
4
, , 5
0
( ) ( )
0.93
H V H V
Z D N D dD
o
t
=
}
, (3)
Total reflectivity is the result from summing the contributes of
hail and rain calculated separately.
For DSD calculation we adopted the Weather Research and
Forecast Model (WRF), a state-of the art NWP developed by a
consortium of research institutes including NOAA and NCAR
[4]. The WRF can also provide the temporal evolution of
parameters based on a real scenario. This NWP gives us
important parameters needed for the definition of an analytical,
physical based Drop Size Distribution (DSD):
1) Hydrometeor mixing ratio [Kg/Kg]
2) Pressure [Pa]
3) Potential temperature [K]
4) Particle Number Concentration [particles/m
3
].
Additionally WRF provides the wind field used to compute
Doppler shifts.
The DSD that we used in our computation is a Gamma
probability density function.
To compute the polarimetric RCS, we utilized a T-Matrix
method. The T-Matrix method is the fastest exact technique for
the computation of non-spherical scattering based on a direct
solution of Maxwell equations [5],[6]. Dielectric constants,
particle orientation, diameter and the relationship between
diameter and axial ratio are set as parameters to calculate the
electromagnetic scattering. Details on the generation of the 3D
reflectivity maps for every polarimetric channel can be found
in our previous work [7].
The received radar signal is then generated using a custo-
mized version of Airborne Windshear Doppler Radar
Simulation (ADWRS), extensively used by NASA in various
campaigns [8].
The simulation input values include the radar systems
parameters, the cinematic characteristics of the airborne
platform, the antenna parameters and the scanning angle
strategy. Other inputs specify the phenomenon characteristics
in term of wind field and radar reflectivity. Last two variable is
represented by a 3D data cube, described before. From both the
initial aircraft position and the initial antenna scan direction,
the simulation consists of the generation of the instantaneous
received signal. For each range bin, the amplitude and phase of
the received signal can be seen as the coherent sum of a
number of contributions that came from volumetric scattering
mechanism.
III. 3DSIMULATION RESULTS
A. Description of the simulated scenario
Simulations were performed with the transponder
characteristics summarized in Table 1. It is worth nothing that
the radar simulator can perform a full 360 scanning, but for
avionic uses we can reduce this range to 180 or less. The
aircraft is positioned in the center (0,0), heading south at 150
knots. The relatively low transmitted power is meant to
simulate the latest state-of-the-art solid-state GaAs radar
transmitters equipping civil avionic weather radars, designed to
work with such low peak power.
TABLE I. TRANSCEIVER CHARACTERISTICS
Transmitted frequency 9.353 GHz
Pulse length 1 s
PRF 6.5 kHz
Range resolution 150 m
Beamwidth 3
Transmitted power 195 Watt
Antenna Gain 33 dB
Noise figure 4 dB
An area of about 1800 km
2
in the Mediterranean Sea, close
to Barcelona, Spain, was selected, with a maximum height of
8000 m. Figure 1 shows the profiles of hydrometeor mixing
ratios obtained from WRF at altitudes of 450 m, 1000 m and
2000 m with a RGB mapping. Red indicates hail/graupel, blue
indicates rain and mixed precipitation zones are in purple.
B. Results
Figures 2 to 4 show some simulation results regarding Z
HH
,
Z
DR
and LDR, which accounts for the more or less pronounced
oscillations of hydrometeors. All these parameters are useful
for classification between liquid and solid dangerous particles.
Figure 2 shows results for the lowest altitude level, well
under the melting layer, dominated by rain. We can notice the
presence of a heavy storm characterized by strong reflectivity
74 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
echoes up to 53 dBZ, however we cannot distinguish if these
echoes are due to hail or rain. Analyzing Z
DR
, the radar
retrieves values from 1 up to 3.6 dB in the storm core. As
explained in section I, this behavior indicates a rain dominated
zone, as. Moreover, LDR level do not surpass -25 dB level,
indicating small oscillations of particles during fall, which is
another characteristics of rain. Over -15 dB values appear
usually where both Zhh and Zvh are very low, so the ratio is
similar.
Figure 3 represents an intermediate altitude where rain and
hail are heavily mixed. As expected, total reflectivity level
remains the same as before, but we can appreciate variations in
the values of Z
DR
and LDR. Z
DR
begins to decrease steadily
reaching his top at 2.8 dB, while LDR rises up to a value of -
21.2 dB. This behavior is typical of a mixed precipitation zone,
but we can still detect rain presence in near the borders of the
scanned area at (0,-20) and (-15,10) coordinates.
Where polarimetry shows its potential in detecting
dangerous area is well shown in Figure 4. Once more, total
reflectivity level remains in the 55 dBZ range, but observation
of Z
DR
and LDR supports the evidence of a hail dominated
zone. Maximum value of Z
DR
do not exceed 0.3 dB, and it even
has negative value, -0.1 dB, which can be caused, other than
the tumbling motion of hail, by the Mie scattering effects for
larger stones. LDR rise up to a value of -18/-17 dB, indicating
a very high signal power in the VH channel (see Eq. 2).
IV. CONCLUSIONS
It is clear that even in an uniform reflectivity phenomenon,
in both its horizontal and vertical structure, polarimetric data
processing can provide useful information for feature
discrimination and thus to reduce risk due to solid particles
impact. Even if the beamwidth is three degrees, combining the
signal received from partially overlapping azimuthal sectors it
is possible to have information which permit to make a good
discrimination and resolve different scattering behaviour.
Further studies will be conducted to evaluate returns from very
long distances. Long ranges suffer also from heavy attenuation
which can be compensated using an additional polarimetric
variable, the Specific Differential Phase (KDP), that is also a
good estimator for rainfall rate. This accurate risk assessment
is not possible with single-polarization avionic radar, so the
only action that is taken is making long detours, even if the
phenomenon would pose no threats.
REFERENCES
[1] F. J. Yanovsky, Evolution and Prospects of Airborne Weather Radar
Functionality and Technology, 18th International Conference on
Applied Electromagnetics and Communications, 2005.
[2] V.N. Bringi, and V. Chandrasekar, Polarimetric Doppler Weather
Radar, Cambridge University Press, 2004.
[3] J.M. Straka, D.S. Zrnic, and A.V. Ryzhkov, 'Bulk Hydrometeor
Classification and Quantification Using Polarimetric Radar Data:
Synthesis of Relations, J. Appl. Meteor. 39, 2000, pp. 13411372.
[4] S.E. Koch, The Use of Simulated Radar Reflectivity Fields in the
Diagnosis of Mesoscale Phenomena from High-Resolution WRF Model
Forecasts, 32nd Conference on Radar Meteorology, 2005.
[5] P.C. Waterman, Scattering by Dielectric Obstacles, Alta Frequenza
(Speciale), 1969, pp. 348352., 1969.
[6] M. Mishchenko, L.D. Travis, and A.A. Lacis, Scattering, Absorption
and Emission of Light by Small Particles, Cambridge University Press,
2
nd
ed., 2005.
[7] A. Lupidi, C. Moscardini, F. Berizzi, M. Martorella, "Simulation of X-
Band Polarimetric Weather Radar Returns based on the Weather
Research and Forecast Model", 2011 IEEE Radar Conference, Kansas
City, 2011.
[8] Britt, C., L., Kelly, C., W., Users Guide for an Airborne Doppler
Weather Radar simulation (ADWRS), Center for Aerospace
Technology, Tech. Rep. 7473/029-05S NASA, 2002.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 75
1 (a)
1 (b)
1 (c)
Figure 1: Mixing Ratio: (a) 450 m (b) 1000 m (c) 2000 m altitude
2 (a)
2 (b)
2 (c)
Figure 2: 450 m altitude: (a) Total Reflectivity, (b) ZDR, (c) LDR
76 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
3 (a)
3 (b)
3 (c)
Figure 3: 1000 m altitude: (a) Total Reflectivity, (b) ZDR, (c) LDR
4 (a)
4 (b)
4 (c)
Figure 4: 2000 m altitude: (a) Total Reflectivity, (b) ZDR, (c) LDR
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 77
G G
(1)
where
m
X is distance between the center of gravity of the
object and the m
th
bright point,
0
R is the distance between the
radar and the center of gravity of the object, is the aspect
angle of the object and
0
E
G
,
1
2
2
0
0
b
S
b
=
(2)
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 79
Fig.1. The scattering geometry for the two-point radar object
We will consider coherent scattering. The geometry is shown
in the Fig. 1. Here the distances
1 2
, R R
between the scatterers
and an arbitrary point Q in the far field can be written as
2,1 0 0
0, 5 sin 0, 5 R R l R l under the condition
0
0, 5l R << . Using these expressions, we can find the Jones
vector of the scattered field for the case of the radiated signal
having a linear polarization at an inclination angle of 45 . It is
should be mentioned here that we are using the Cartesian
(linear) polarization basis both for the scattered matrices and
for the Jones vector of the scattered field.
( )
( ) ( )
( ) ( )
1 1
2 2
exp exp 2
2 exp exp
S
a j b j
E
a j b j
+
=
+
(3)
where kl = . The angular dependence of the polarization-
energetic response functions in the form of the Stokes
parameters
0 3
, S S is
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
0
3
;
[ ].
X X Y Y
X Y Y X
S E E E E
S i E E E E
= +
=
The expanded form of the energetic response function ( )
0
S
can be found as
( )
0 0 0
0, 5
a b
S S S = + +
( )
2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2 1
cos 2 a b a b a a b b a a b b
+ + + + +
(4)
where
( ) ( ) { } 1 1 1 2 2 1 1 2 2
Im / Re arctg a b a b a b a b
= + +
and
2 2
0 1 2
a
S a a = + ,
2 2
0 1 2
b
S b b = + . The values
0 0
,
a a
S S are the
Stokes zero-parameters of the elementary scatterers a and
b.
The polarization-angular response function ( )
3
S can be
found as
( )
3 3 3
0, 5
a b
S S S = +
+
( )
2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2 2
2 ( ) sin 2 a b a b a a b b a a b b
+ + + +
(5)
where
( ) ( ) { } 2 1 2 2 1 1 2 2 1
Im / Re arctg a b a b a b a b
=
and
( )
3 1 2 1 2
0.5
a
S j a a a a
= ,
( ) 3 1 2 1 2
0.5
b
S j b b b b
=
are the 3-rd
Stokes parameters of the elementary scatterers a and b.
The angular harmonic functions [ ] cos ... , and [ ] sin ... in the
expressions (4) and (5) represent the influence of the spatial
separation l on the distribution of the polarization-energetic
parameters of scattered field in the far zone. The derivative of
the full phase ( ) 2
k
kl = + ( 1, 2 k = ) of the angular
harmonic functions along the angular variable is the space
frequency:
[ ]
1 2
2
2
SP k
d l
f kl
d
= + =
(6)
Thus, the space frequency in the distributed radar object
theory equals twice the distance between the elementary
scatterers constituting the radar object, normalized to the wave
length.
Next we will analyze the amplitudes of the angular harmonic
functions [ ]
1
cos 2kl + , [ ]
2
sin 2kl + in order to assess the
impact of the polarization properties of the elementary
scatterers on the polarization-energetic parameters of the field
scattered by the complex radar object. Let us write the
polarization ratios
2 1
/
A
P a a =
and
2 1
/
B
P b b =
which are
characterizing the point radar objects A and B on the complex
plane of radar objects [3]. Using the method of the
stereographic projections, we can find the spherical distance
between the points ,
A B
S S , laying on the surface of the
Riemann sphere having unit diameter, which are connected
with the points ,
A B
P P
on the radar objects complex plane.
The coordinates of the points ,
A B
S S on the surface of the
sphere are
2 2 2 2
1 2 3
Re /(1 ); Im /(1 ); /(1 ) X P P X P P X P P = + = + = +
and the spherical distance between these points can be found
to be:
2 2
( , )
1 1
A B
S A B
A B
P P
S S
P P
= =
+ +
( )
2 2
2 2
1 1
A B A B A B
A B
P P P P P P
P P
+ +
=
+ +
(6)
where
A B
P P
is the Euclidian metric on the complex plane
of radar objects. After substitution of the polarization ratios
2 1
/
A
P a a =
and
2 1
/
B
P b b =
into the expressions (6) we can
write
( )( )
2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2
2 2 2 2
1 2 1 2
( )
( , )
S A B
a b a b a a b b a a b b
S S
a a b b
+ +
=
+ +
(7)
The value
( )( )
2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2
2 2 2 2
1 2 1 2
( ) a b a b a a b b a a b b
D
a a b b
+ +
=
+ +
(8)
0,5l 0,5l
1
R
2
R
0
R
1
S
2
S
80 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
is the so-called polarization distance between two waves (or
radar objects), having different polarizations [5]. When the
waves having coincident polarizations, (
A B
P P =
) can be
characterized by the polarization distance value 0 D = and the
waves having orthogonal polarizations ( 1/
B A
P P
=
) can be
characterized by the polarization distance value 1 D = . Thus, it
follows from (7) and (8) that
( )( )
2 2 2 2 2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2 1 2 1 2
( ) a b a b a a b b a a b b D a a b b
+ + = + +
.
We can also use the so-called polarization proximity value N
that can be defined as 1 N D = . Then
( ) ( )
2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2
2 2 2 2
1 2 1 2
1
a b a b a a b b a a b b
N D
a a b b
+ + +
= =
+ +
(9)
The waves with coinciding polarizations (
A B
P P =
) can be
characterized by the polarization proximity value 1 N = and
the waves having orthogonal polarizations ( 1/
B A
P P
=
) can
be characterized by the polarization proximity value 0 N = .
Then we can write
( )( )
2 2 2 2 2 2 2 2
1 1 2 2 1 2 1 2 1 2 1 2 1 2 1 2
a b a b a a b b a a b b N a a b b
+ + + = + +
.
If we compare the amplitudes of the space harmonic
oscillations with the expressions (7) and (9), we can see that
the expressions (4) and (5) can be rewritten as
( ) ( )
0 0 0 0 0 1
0, 5 2 cos 2
a b a b
S S S S S N
= + + +
(10)
( ) ( )
3 3 3 0 0 2
0, 5 2 sin 2
a b a b
S S S S S D
= + + +
(11)
We can consider these expressions as generalized interference
laws [4]. It follows from the expression (10) that the
orthogonally polarized waves cannot give an interference
picture in case the polarization proximity value 0 N = .
However, the expression (11) demonstrates that in this case
the third Stokes parameter will have the maximal value of this
interference picture visibility.
It follows from expressions (10), (11) that every Stokes
parameter has some constant component, which is defined by
the respective Stokes parameters of both objects (a and b),
and space harmonics function [ ]
1
cos 2kl + , [ ]
2
sin 2kl + ,
having amplitudes
0 0
a b
S S N ,
0 0
a b
S S D and space initial
phase
k
. So, the polarization-energetic properties of complex
radar object cannot be found using only the properties of its
individual elements. The properties of the integral system
appear by taking the relations between the individual elements
into account. These relations in our case are the polarization
distance and the polarization proximity. The use of these
values leads to the emergence of new properties which did
not exist for every element separately.
We define an instantaneous visibility of the generalized
interference law as
( ) ( )
( ) ( )
0 0 0 0
0 0 0 0
2 .
A B MAX MIN
MAX MIN A B
S S S S
W N
S S S S
= =
+ +
(12)
It can be seen that the equation (12) is coinciding with the well
known expression for the Fresnel-Arago interference law
( ) ( )
( ) ( )
1 2
12
1 2
2 ,
MAX MIN
MAX MIN
I I I I
W
I I I I
= =
+ +
where
1 2
, I I are the integrated powers (energies) of the waves
and
12
is a degree of coherence. If
1 2
I I = then the visibility
of the interference law is defined by the degree of coherence
of second order.
So we can state that from a physical point of view the
parameter N can be considered as a polarization coherence
parameter, which defines the proximity of the polarization
states of elementary scatterers, and in the same way a degree
of coherence of stochastic waves is summarized. In this case
we have an instantaneous value of the polarization
coherence, while at the same time the coherence degree
12
is
the correlation value.
IV. EXPERIMENTAL INVESTIGATIONS OF THE
POLARIZATION-ENERGETIC PROPERTIES OF
ELECTROMAGNETIC WAVES SCATTERED BY A TWO-POINT
RADAR OBJECT
A measurement campaign to investigate jointly the
generalized FresnelArago interference laws and the
polarization-energetic properties of the electromagnetic field
scattered by reflecting interferometers (man-made radar
objects consisting of two elements) was realized by the
International Research Centre for Telecommunications and
Radar of TU Delft [5]. In this paper a small part of the results
is presented and interpreted from the point of view of the
generalized FresnelArago interference laws and the
emergence principle with respect to power and polarization
harmonics ( ) ( )
0 3
, S S , corresponding to the space
frequency caused by the distributed radar object used in the
campaign.
A collection of twoelements manmade distributed radar
objects with known polarization properties of their elements
was used in the campaign. The difference between the
properties of the various elements constituting the distributed
radar objects leads to different values of the polarization
proximity or polarization distance of these elements. The
following combinations of two elements manmade
distributed radar objects were used:
1).Two empty trihedral corner reflectors ( 1; 0 N D = = );
2).Two trihedrals, where the first one was empty and the
second one was fitted with a linear polarizer consisting of a
special polarizing grid. ( 0.5; 0.5 N D = = );
3). Two trihedrals, where the first one was empty and the
second one was fitted with an elliptic polarizer consisting of a
special polarizing grid. The transmission coefficients along the
OX and OY axes are 0.5
Y X
b b = and the mutual phase shift
between the polarizers eigen axes is / 2
XY
=
( 1; 0.5;
A B
P P j = =
).
The phase centers of the trihedrals were separated by 100 cm,
while the wavelength of the radar was 3 cm. For these
parameters the space frequency and space period are
1
2 / ( ) ,
SP
f l Rad
= 0, 015
SP
T Rad = (or 0.855 ). The
mechanical construction, on which the trihedrals were
mounted, rotated with an angular step 0, 25 .
When the object includes the trihedral with the elliptic
polarizer and the empty trihedral (combination 3 ), we can find
the theoretical estimation of the polarization proximity and
distance as 0.5. N D = = In the Fig.3a,b the experimental
angular harmonics functions (generalized interference
pictures)
( ) ( )
0 3
, S S are shown. It follows from these figures
that the visibility for interference picture ( )
0
S is
0
0, 3 W
corresponding to a polarization proximity
0
0.54 N = (note
that the theoretical estimation is N=0.5). The visibility
for ( )
3
S is
3
1 W = , corresponding to a polarization distance
0.5 D = . For the system including the trihedral arranged by
the linear polarizer and empty trihedral (object N2), we can
0
0,1
0,2
0,3
1 3 5 7 9 11 13 15 17
Fig.3a. Generalized interference law
for the Stokes parameter ( )
0
S (object N3)
0
0,1
0,2
0,3
1 3 5 7 9 11 13 15 17
Fig. 3b. Generalized interference law for the Stokes parameter
( )
3
S (object N3)
0
0,1
0,2
0,3
1 3 5 7 9 11 13 15 17
Fig.4a Generalized interference law
for the Stokes parameter ( )
0
S (object N2)
0
0,1
0,2
0,3
1 3 5 7 9 11 13 15 17
Fig. 4b. Generalized interference law
for the Stokes parameter ( )
3
S (object N2)
find the theoretical estimation visibility values
0
0, 66; W =
3
1 W = that correspond to polarization proximity
values
0 0 3 3
0.82; 1 N W N W = = = = . In the Fig.4a,b the
angular harmonics functions ( ) ( )
0 3
, S S for this situation
are shown.
The experimental estimation based on Fig.4a,b gives us
0 3
0.85; 1 N N = what is the closely coinciding with the
theoretical estimation.
V. CONCLUSIONS
The results presented demonstrated the phenomenons which
occur when electromagnetic waves are coherently scattered by
two-point radar objects. These results can be used for
preliminary classification of complex radar objects in the
flight safety problem
REFERENCES
[1] J.A.Stratton, L.J.Chu. Diffraction theory of electromagnetic waves.
Phys. Rev, v.56, pp 308- 316.
[2] F.I.Peregudov, F.P.Tarasenko. The principles of systems analysis.
Tomsk, 2001, 350 p. (In Russian)
[3] V.N. Tatarinov, S.V. Tatarinov, L.P. Ligthart. An introduction to radar
signals polarization modern theory. Tomsk State University Publ
House, 2006, vol.1 380 pp.
[4] V.N. Tatarinov, S.V. Tatarinov. A Generalization of Fresnel-Arago
Interference Laws. Proc. of Int. Conf. SoftCOM 09, Split-Hwar,
Croatia, Sept. 2009
[5] V.N. Tatarinov, S.V. Tatarinov, P. van Genderen, D.Tran, P.Usov,
J.Zijderveld. The Reports IRCTR-S-028-03 and IRCTR-S-029-04.
82 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Principles of Utilization of Polarization Invariant
Parameters for Classification and Recognition of
Distributed Radar Objects
Part II. Multipoint model and correlation theory
Victor N. Tatarinov, Sergey V. Tatarinov
Radiodesign Department
Tomsk University of Control Systems and Radioelectronics
Tomsk, Russian Federation
v-n-tatarinov@yandex.ru
Piet van Genderen
IRCTR/MTSRadar
Delft Technology University
Delft, The Netherlands
p.vangenderen@tudelft.nl
Abstract The paper is the second step towards the utilization of
polarization invariant parameters for classification and
recognition of distributed radar objects. A the theory on the
statistical properties of the polarization parameters of the
scattered field is established and confirmed by the experimental
data
Keywords- Stokes parameters; angular distribution;
autocorrelation function; space spectra of distributed objects
I. INTRODUCTION
This paper demonstrates that the interference process of the
field scattered from multi-point random complex radar objects
(RCRO) leads to polarization-energetic speckles. The
polarization-energetic response function of an RCRO can be
considered to be a collection of space harmonics. Every space
harmonic of this collection is initiated by one pair out of many
pairs, which can be formed by the multi-points constituting the
scattering RCRO. Every space harmonic will have an
amplitude, which will be determined by the value of the
proximity (or distance) of the polarization states of the points
involved in the respective pair. The positions of the elementary
scatterers composing the RCRO are stochastic and we have a
random number of interfering pairs. The polarization proximity
of each pair also is a stochastic parameter, and thus, even when
the spatial separation between points in a pair is the same, we
will have a classical stochastic process at each change of the
aspect angle. Our approach to the problem is novel and it is
formulated like this for the first time.
II. THE POWER ANGULAR DISTRIBUTION OF THE
ELECTROMAGNETIC FIELD SCATTERED BY A DISTRIBUTED
OBJECT
Let us consider first the geometry at the scattering by one of
the scatterers of the multi-point (complex) radar object (fig.1).
For the case of coinciding linear polarizations both for
transmission and reception we can write the field scattered by
a point scatterer
I
(where
I
is the Radar Cross Section
(RCS) of this scatterer) , observed in some point Q in the far
field as
( )
0
exp 2
4
I
S I
I
j kR
E E
R
=
(1)
where
I
R is the distance between the scatterer at
I
X and the
point Q at X (ref. Fig.1) ;
0
E
and
S
E
=
(3a)
When all scatterers are characterized by a scattering matrix
( ) ; , 1, 2
il
I
S i l =
(3b)
Let us now consider the formation of the distribution of the
polarization-energetic parameters of the electromagnetic field
resulting from the interference process at the scattering by the
RCRO. To this end we will analyze the scattering process by
the multi-point RCRO. Without loss of generality we will
assume that the point scatterers constituting the RCRO, are
located on a line (see Fig.1).
For the example of Fig.1 we will find that the electrical vector
of the field, scattered by the 4-points complex object, observed
in a point Q located in the far field for the case of coinciding
linear polarization both for transmission and reception is:
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 83
( )
( )
( )
0 0
1 1
0
exp 2
{ exp 2
4
S
j kR E
E j kX
R
= +
( )
2 2
exp 2 j kX + +
( ) ( )
3 3 4 4
exp 2 exp 2 }. j kX j kX + +
(4)
Then we can find the instantaneous distribution of the power
of the scattered field in space as a function of the positional
angle as:
( ) ( ) ( ) ( )
1 1
{ exp 2
S S
P E E A j kX
= = +
( ) ( )
2 2 3 3
exp 2 exp 2 j kX j kX + + +
( ) ( )
4 4 1 1
exp 2 } { exp 2 j kX j kX + +
( ) ( )
2 2 3 3
exp 2 exp 2 j kX j kX + + +
( )
4 4
1 2 3 4
exp 2 } j kX +
= + + + +
( ) ( )
1 2 12 1 3 13
2 cos 2 2 cos 2 kd kd + + +
( ) ( )
1 4 14 2 3 23
2 cos 2 2 cos 2 kd kd + + +
( ) ( )
2 4 24 3 4 34
2 cos 2 2 cos 2 kd kd + +
(5)
So, the instantaneous distribution of the power of the scattered
field in space as a function of the positional angle is formed
by the sum of the radar cross section of the elementary
scatterers (4 terms) plus 6 cosine oscillations. It can be seen
that all of these cosine terms are caused by the interference
effect between the fields scattered by all pairs of elementary
scatterers constituting the RCRO. The number of these pairs
can be found by the binomial coefficient
( )
!
! !
N
M
M
C
N M N
=
= =
, the derivative of which is
1 2
2
SP
d d
f
d
= = . It represents the space frequency with
dimension
1
Rad
. The period 1/ / 2
SP SP
T f d = = having
the dimension [ ] Rad corresponds to this frequency.
So, the full power distribution of the field scattered by a
complex radar object, is the sum of the interference patterns
formed by a collection of elementary two-points interfering
scatterers.
Thus, we can write the random angular representation of the
scattered power, depending on the positional angle as
( ) ( )
2
1 1
2 cos 2
M C
m i l il
m
P kd
=
= +
(6)
where
2
M
C C = is number of combinations, and M is the total
number of elementary scatterers constituting the RCRO.
III. THE ANGULAR DISTRIBUTION OF THE STOKES
PARAMETERS OF THE ELECTROMAGNETIC FIELD SCATTERED
BY A DISTRIBUTED OBJECT
It was demonstrated in [1] that the angular distribution of the
Stokes parameter
0 3
and S S of the electromagnetic field
scattered by a two-point distributed object has the form
( ) ( )
( ) ( )
0 0 0 0 0
3 3 3 0 0
2 cos 2 0.5 ;
2 cos 2 0.5 .
a b a b
ab
a b a b
ab
S S S S S N kl
S S S S S D kl
= + + +
= + +
(7)
It follows from expression (7) that the space harmonic
functions ( ) cos 2kl have amplitudes
0 0
a b
ab
S S N
and
0 0
a b
ab
S S D . Here the values ,
ab ab
N D are respectively
the proximity and distance of the polarization states of the
elementary scatterers of the distributed object [1, 2].
Taking this into account, we can write the angular distribution
of the Stokes parameters of the field scattered by a random
complex radar object as the sum of the generalized
interference patterns, which are formed by a collection of
elementary two-points interfering scatterers (see Fig.1):
2
X
3
X
4
X
1
X
4
R
3
R
2
R 1
R
0
R
0
Q
4
R
X
Fig. 1. The scattering geometry for multi-point radar object
84 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
( ) ( )
0 0 0 0
1 1
2 cos 2
M C
m
i l il il il
m
S S S S N kd
=
= + +
(9a)
( ) ( )
3 3 0 0
1 1
2 cos 2
M C
m
i l il il il
m
S S S S D kd
=
= + +
(9b)
where
2
M
C C = is the total number of combinations. The
amplitudes of the space harmonics and the initial space phases
of these harmonics will be stochastic values. Thus the further
analysis must be based on statistics.
IV. ATHEORETICAL DEFINITION OF THE
AUTOCORRELATION FUNCTION OF THE ANGULAR
DISTRIBUTION OF THE STOKES PARAMETERS OF THE
SCATTERED FIELD. SPACE SPECTRA
Now we will develop a theoretical form of the autocorrelation
function of the angular distribution of the Stokes parameter
( )
3
S of the scattered field.
Since we would like to find the autocorrelation function (not
the covariance function!), we must eliminate a random
constant term
3
1
M
m
m
S
=
(10)
Lets suppose that the function (10) is a stationary stochastic
function. Then its autocorrelation function can be found as
( ) ( ) ( ) { }
1 2 S
B M S S = =
[ ]
1 1
cos 2
C C
N L N
N L
M D D kd
= =
= +
[ ]} cos 2 ( )
L
kd + +
(11)
Here the space harmonics amplitudes D and the space
initial phase are random values, which can be characterized
by a two-dimensional probability distribution density
( ) 2
, W D , M is the averaging operator and
1 2
= .
We will suppose that the random amplitudes and phases are
independent variables. Then the two-dimensional probability
distribution can be presented as the product of two one-
dimensional distributions:
( ) ( ) ( )
2 1 1
, W D W D W =
Using the orthogonality condition
( ) ( )
1
cos cos
0
N L
SP SP
forN L
d
forN L
=
=
[ ] [ ]} cos 2 cos 2 ( )
N N
kd kd + + +
(12)
Considering that the initial stochastic realization
( )
3
S
is a
function of the random variables and D , we will use for
the definition of the autocorrelation function of this realization
the expression for the mean value of a function of two random
variables:
( ) { } ( ) ( )
1 2 1 2 2 1 2 1 2
, , , M y x x y x x W x x dx dx
=
Using this expression we can write the autocorrelation
function (12):
( ) ( ) [ ]
2
1
cos 2
C
S N N
N
B D kd
=
= +
[ ] ( ) ( ) 2
cos 2 ( ) ,
N
kd W D d D d + +
(13)
For the calculation of the double integral in (13) we will use
the condition
( ) ( ) ( )
2 1 1
, W D W D W = mentioned before.
Lets suppose also that the random phase has a uniform
probability distribution density on the interval
( ) , , i.e.
( ) 1/ 2 W = . A probability distribution density for the random
amplitude D can be preassigned, however for all cases it has
to be one-sided. Now we will transform the product of the
cosine functions and substitute it into (13) together with the
probability distributions of the random variables and D :
( ) ( )
1 2 3
1
C
S
N
B I I I
=
= + +
(14)
Lets now evaluate the integrals
1 2 3
, , I I I . The first integral
gives
( ) ( ) ( ) ( )
2
1 1
0
0.5
cos 2
2
N N
I D kd W D d D d
= =
( ) 0, 5 cos 2
N N
D kd = < >
(15)
where
N
D < > is the mean value of the polarization distance,
which was found by averaging over the statistical ensemble of
random values
N
D for all space harmonics having the space
frequency 2 /
N
SP N
f d = .
The second integral
2
I can be evaluated using the condition
cos
0
sin
d
+
( ) ( ) ( ) 1
cos 2 0.
N
kd W D d D d =
It can be shown that the third integral, containing the sinus
functions, is zero as well:
3
0 I = . Thus, we can write the
theoretical form of the autocorrelation function of the angular
distribution of the Stokes parameters of the scattered field as:
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 85
( ) ( )
1
cos 2
2
C
N
S N
N
D
B kd
=
< >
=
(16)
Taking into account that every term of the sum in (16) is the
autocorrelation function for an isolated space harmonic
oscillation
( ) ( ) cos 2
N N N N
S D kd = + having random
amplitude
N
D and random initial space phase
N
, i.e.
( ) ( ) cos 2
2
N
SN N
D
B kd
< >
=
(17)
it can be seen that the autocorrelation function of the
stochastic realization of the Stokes parameters is the sum of
the individual autocorrelation functions of all space harmonics
which are contributing to this stochastic realization:
( ) ( )
1
C
S SN
N
B B
=
=
(18)
Lets now develop complex radar objects averaged space
spectra using expression (18) for the autocorrelation function
of the polarization-angular response. The power spectrum in
the case of isolated space harmonics can be found as the
Fourier transform of the above the autocorrelation function
(17)
( ) ( ) ( ) ( ) exp
SP SN SP
P B j d
= =
( ) ( )
0.5 [ ]
N N
N SP SP SP SP
D = < > + +
(19)
where
2
2 2
SP SP
d
f
(20)
So, the power spectra of the polarization angular response
function have a discrete form. It is caused by the discrete
structure of the RCRO. Besides, man-made distributed radar
objects have a finite extension. In this context we have to
emphasize that the power spectra of radar objects have a
limited character.
V. EXPERIMENTAL INVESTIGATIONS OF THE ACF AND
SPACE SPECTRA
The experimental autocorrelation functions (ACF) and space
spectra of the stochastic polarization-angular response of a
rotated complex radar object (it is a Caterpillar vehicle, a
heavy construction machine) are shown in Fig.2 and Fig.3.
Fig.2 shows the autocorrelation function in the angular
interval 20 w.r.t. the objects board (dashed line) and the
autocorrelation function over the same interval w.r.t. the stern
of the object (solid line). The measurements in these directions
allow us to consider the difference in the radar objects space
spectral bands when it is observed in areas perpendicular to
the board (dashed line) and to the stern of the object (solid
line). It can be seen from Fig. 3 that the RCROs mean power
spectra have a two-mode form. It shows that the so-called
equivalence principle can be used on the average in order to
describe a model of a random complex radar object [3] as two
distributed scatterers in space.
-1
-0,5
0
0,5
1
Fig. 2. Autocorrelation functions of RCRO
stochastic polarization-angular response
-0,5
0
0,5
1
Fig.3. Mean power space spectra of RCRO
VI. CONCLUSION
The results of this paper confirm that we can consider the
formation of the polarizationpower parameters of the
electromagnetic field scattered by an RCRO as an interference
process. This fact allows us to find both the autocorrelation
function and the RCRO space power spectra of the
polarization-angular stochastic response. It gives us the
possibility for recognition and classification of distributed
radar objects.
REFERENCES
[1] V. N. Tatarinov, P. van Genderen, S. V. Tatarinov. Principles of
Utilization of Polarization Invariant Parameters for Classification and
Recognition of Distributed Radar Objects. Part 1. Simplest Model of a
Distributed Object. Proc. of the Int. Conf. ESAV2011.
[2] V.N. Tatarinov, S.V. Tatarinov, L.P. Ligthart. An introduction to radar
signals polarization modern theory. Tomsk State Univ.. Publ. House,
vol.1 2006, 380 pp. (In Russian)
[3] S.V.Tatarinov, L.P.Ligthart, V.N.Tatarinov. The use of an equivalence
principle on the average for a statement of definition of an random
complex radar object model. Proc. Of MIKON2000, Wrozlaw, Poland,
Vol 2, pp. 12-17
86 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Noncoopeiative location
Enhancingsensitivityforemittergeolocation
Session4.1page89
NewsolutiontoenhancethesecurityinAirTrafficControl
Session4.2page95
SolvingtheDataLinkbottleneckforMPEGLocation
Session4.3page101
ParasiticDopplereffectinpassivelocation
Session4.4page107
Aninairpassiveacousticsurveillancesystemforairtrafficcontrol
Session4.5page111
1
n
L
1
n
h
1
n
L
+
=
=
N
Lh
N
Lh
s L
Figure 2. EEP caracteristics
In this equation N is the number of equivalent uncorrelated
measurements. N is also called the number of effective
The authors acknowledge the European Defence Agency,
the French DGA, the Swedish FMV and the Italian MoD,
which have contributed to the funding of this study under
EDA contract nB-0055-IAP2-ERG.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 101
R
R
For time limited illumination (critical target)
VT
R R
=
N R
R
12
If the radar target is at 50 km distance, and a quite
advanced interferometer is used (0.3 accuracy) the best
location accuracy can only be obtained flying a long baseline
(e.g more than 150 seconds for an aircraft at 300m/s speed) as
a good angular aperture between the measurements is required.
B. M- AOA Location Basics
The MP-AOA aims at delivering an instantaneous and
accurate target location by using several Lines of Bearing
coming from N different sensors located at several positions
remote from each other. Theoretically, MP-AOA can achieve
accurate location as soon as the targets beam has illuminated
all the sensors during one scanning excursion.
Multiple platforms AOA offer a direct improvement in time-
line, and the location accuracy will then depend on the
measurement accuracy of each of the contributors and their
geometrical positions. The angular aperture and distances of
the platforms in the network will provide a result without
needing a long integration time. This is the major advantage of
MPEG over SPEG.
Figure 3. Advantage of MPEG over SPEG
If the target manage its illumination time, this will defeat SPEG
method.
III. DATA LINK REDUCTION PRINCIPLE
Sensors are geographically dispersed, and to combine the
information that they acquire requires the information to be
communicated to some location(s) where the data integration
process occurs. There are 2 possible ways to proceed:
First possibility consists in communicating
exhaustively each AOA measurement that has been
carried out.
Figure 4. Measurements Fusion
Second possibility consists in making an angular
tracking of each target, performed by each ESM
sensor. In this case, the state of the track is regularly
communicated.
Figure 5. Tracks or Information Fusion
Due to the constraints in the communications bandwidth,
in a practical multi-sensor tracking network, the sensor data
processors communicate only a subset of the data available
from each sensor, usually in the form of tracks but alternatively
in the form of information summaries.
This can also be applied to MP-TDOA location according
the following diagram:
Figure 6. AOA (Angle Of Arrival) and HOA (Hyperbole Of Arrival)
102 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Local Angular Tracking
(2D)
3D Tracks initiation
(Deghosting)
3D Tracking
2D tracks non assigned
3D tracks 3D Tracks
Selected
2D tracks
New 3D
Tracks
Local ESM
perimeter
IV. ARCHITECTURE CHOICE
Architectures and Algorithms for Network Sensor Fusion
are presented in famous tutorial books like[1][2]. Paper [3]
gives an overview of different architectures from centralized,
hierarchical to distributed.
A central approach non only presents a saturation risks with
all the measurements transmitted but turns out " inextricable
from the point of view of the association " while a local
tracking establishes a reasonable approach reducing
considerably the difficulties of association with a loss of almost
nonexistant intrinsic performance. The general Hybrid-
architecture is the following:
Figure 7. Hybrid Architecture for ESMs passive tracking
In this architecture the 2D or 3D tracks are initialized at the
level of the passive sensors.
The association of the passive data (DOA or 2D tracks) is
treated(handled) at the level of the passive sensor, whether it is
the association of the DOA in the 2D angular tracks or the
formation of the 3D tracks from a treatment of association of
the passive 2D tracks.
The central tracker (3D tracking) has for objective to
maintain the estimation and the prediction of the 3D tracks by
taking into account the knowledge of the dynamics of targets.
This central tracker realizes no treatment of association. In the
case of fixed target this estimation and prediction is the most
simple. The updated 3D tracks must be sent back to the local
ESM to increase the quality of the local ESM association.
These 3D tracks must be also feedback towards the process of
initialization of the 3D tracks to avoid the creation of redundant
3D tracks. The process of initialisation of the 3D tracks realizes
the deghosting.
This architecture is distributed on all the observers.
V. TRACK AND INFORMATION FUSION COMPARED
A. Track Fusion
Track fusion is the best popular approach and its limitations
when used with feedback have been explained and solved in
different papers [4][5][6][7]. Above all, Track fusion is not
well suited for passive sensors where the local unobservability
problem occurs.
In track fusion sensor-level tracks are combined to form
global-level tracks that are based on data from all sensors.
If ( )
K k
P X , is the information (state, covariance) given by
sensor k. The track fusion without feedback and with
uncorrelated measurements is given by:
=
=
N
k
k k
N
k
k F
X P P X
1
1
1
1
1
and
1
1
1
=
N
k
k F
P P
(Equ.1)
In this equation
K
P has to be invertible which is not the
case in a passive system above all at the beginning of the
measurements. Even if
K
P is invertible, if it is ill-conditioned
numerical errors appear and are accumulated during successive
integration of measurements. When using track fusion, there is
the implicit requirement that the tracks and their covariance
matrices exist and are invertible. This is always the case in
radar and the track fusion has been designed at the beginning
for this type of sensor.
As integrating all the measurements from all sensors global
tracks are more accurate that sensor-level tracks. Thats why it
is relevant to feed back global tracks to the sensor-level
trackers to reduce the data association errors.
The global tracks, however, are cross correlated with the
sensor-level tracks. This track to track correlation should be
taken into account in algorithm design. The paper [4] presents
algorithm architectures and methods for dealing with the cross-
correlations of the tracks in track fusion for feeding back the
global level tracks. All these algorithms use a decorrelator in
order to remove the cross-correlation between the sensor level
tracks and the global level tracks.
But there is an alternative to this scheme which consists to
exchange Information and not tracks.
B. Information Fusion
In Information fusion sensor-level information summaries
are combined to form global-level tracks that are based on data
from all sensors.
In linear or pseudo linear cases measurements and state are
related by the following type of linear equation:
X H Z
n n
=
The optimal solution that minimizes the mean quadratic
error is given by:
( )
( )
n
T
n n
T
n n
Z H H H X
1
=
An iterative calculation can be used by applying the
recursive process on the coefficients of matrix
( ) n
T
n n
H H Inf = and
n
T
n
Z H .
Its interesting to notice that
( ) n
T
n n
H H Inf = is the
expression of the Information matrix, whose inversion can
provide the Cramer Rao Boundary in the Gaussian case.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 103
If ( )
K K
J I , is the information given by sensor k.
k
T
k k
H H I =
k
T
k k
Z H J =
( )
F F F
J I X
1
=
with
=
=
N
k
k F
I I
1
and
=
=
N
k
k F
J J
1
In the general case, in passive systems information matrices
exist but are often ill-conditioned and even non invertible.
Fortunately the combined information matrix becomes
invertible with good conditioning until the passive systems are
well separated in space.
To avoid the decorrelation phase the purpose is to not
communicate tracks but statistical summaries which are
independent each other. The processing is according the
following diagrams:
Figure 8. Information Fusion Processing
In the processing the global level tracks are feedback but just
for association.
For non-linear least squares systems similar arguments
shows that the solution should be generalized as in VI.
C. Comparison
As said previously, there are two opposite ways to proceed
when we deal with multi-platform case:
First one is sharing exhaustively every line of bearing
calculated after each target beam interception,
Second one is sharing summarized relevant
information about each target: average line of bearing
an associated covariance, for instance.
However, each solution presents serious drawbacks:
In the first case, we communicate a huge amount of
redundant information, without any relevant filtering
or compression. As a consequence, this method
requires high data rate communication capabilities.
In the other case, some interesting information can be
lost.
The optimal solution consists in communicating some
compressed but relevant information that offers the best trade
off between the following competing requirements:
Optimal geo-location performances (obtained when all
the measurements are available).
Minimal communication rate between platforms.
The question now is what summary to transmit?. This
question is fundamental either for track fusion or for
information fusion. It has been addressed for information
fusion in [8] and for track fusion in [9].
How to provide an exhaustive statistics for practical
measurements system in the case of information fusion will be
presented in paragraph xx. It will be done either on Information
Matrix form or based on a Taylor series expansion of the
measurement in time or space.
VI. DETAILS OF SUMMARIZATION METHOD
A. Estimators decomposition - Principle
The optimal estimation is obtained by processing the entire
amount of available measurements collected by every platform.
We can make a decomposition of this optimal estimation by
separate processing of measurements collected by each
observer, before sharing results coming from these different
observers. The overall amount of calculation is parallelized
between observers, which finally share a limited amount of
data. This data is composed of intermediate results that
participate to the final geo-location estimation.
1) Linear case
In that case, the calculation does not require any iteration
and an additive decomposition can be easily obtained. Each
observer calculates partial estimations using his own data,
before sharing these intermediate results with the other
observers.
Only a limited number of coefficients have to be exchanged
via the communication network.
2) Non Linear case
The issue of measurements compression consists in
representing accurately numerous observations by using a
limited number of coefficients, without any significant loss of
information.
Generally speaking, Taylor expansion can offer an
interesting solution to this problem in some case, as well as
Fourier Transform, for instance.
We can assess compression efficiency by comparing:
the actual observations
the observations that are re-estimated from the limited
number of coefficients that have been calculated
according to the chosen compression modeling.
104 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
d
vr
v
ESM 1
ESM 0
d
vr
v
ESM 1
ESM 0
In order to choose a relevant modeling for compressing
angular measurements, several parameters must be taken into
account (geometrical configurations of acquisitions, sensors
characterization, ).
The method is based on Taylor series Expansion of the
measurement equation:
( ) ( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
1
1
2
! 1 !
....
! 2 ! 1
+
+
+
+ + +
+
+ =
n
n
n
n
a x
n
f
a x
n
a f
a x
a f
a x
a f
a f x f
For a function of several variables (3 here) it becomes:
( ) ( )
( )
( )
( )
( )
( )
( ) .......
, , , , , ,
, , , , +
+ =
=
=
=
=
=
=
=
=
=
c z
z
z y x f
b y
y
z y x f
a x
x
z y x f
c b a f z y x f
c z
b y
a x
c z
b y
a x
c z
b y
a x
The validation of regression validity considers values of
higher order terms.
If the relative motion is MRU Expansion in time is
sufficient.
3) Application to TDOA and AOA
According to the following diagram:
The measurement equation is:
( ) ( ) ( ) ( )
( )
c
t r
t r t r
c
t
= =
0 1
1
= <<1
( ) ( ) ( ) ( ) ( ) ( )
( ) ( )
+ + =
8 2
1
1 1
1
2
2 / 1
1
t t
t r
c
t t r
c
t
This gives
( )
( ) ( ) ( ) ( )
( )
( ) ( )
( ) ( ) t cr
d
t cr
t d
t cr
t d
c
t d
t
3
4
2
3 2 2
1
8 2
cos
2
sin cos
+ =
=
t
t
t r
v
( )
0 =
=
t
t r
t
vr
( ) ( )
( )
4
4
3
3
2
2
0 1
8
3
cos 2 sin
2
cos 2 sin
2
sin sin
cr
d
vr v
cr
d
v vr
cr
d
r
v
c
d
t
t
t
+ +
+ =
+ =
=
( ) ( ) ( ) ( )
( )
( ) ( )
=
=
+ =
cos 2 sin
2
sin sin
,
2
sin
cos ,
where
. , ,
2
1
2
1
1 1 1
v vr
r
d
r
v
r
r
d
r
t r r
c
d
t
AOA case is when d/r <<<< 1. This leads to:
( ) ( ) ( ) ( )
( )
( )
=
=
+ =
r
v
r
r
t r r
c
d
t
sin
,
cos ,
where
. , ,
1
1
1 1 1
4) Implementation
A possible implementation of the method is the following:
Robust regression on Measurements.
Analysis of Regressor variances to validate regression
o If not validated: Divise the data and iterate.
o Once validated: Transmit regression
coefficients:
First Bearing (denoised)
Last Bearing (denoised)
Number of estimated independant
bearings
Bearing rates (not fully required for
fixed emitters)
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 105
The full MP-AOA location distribution is represented
according the following figure:
Figure 9. Location Distribution Principle
VII. ILLUSTRATIVE SIMULATION
The following example considers a patrol of two aircrafts
targeting an hostile aircraft which is emitted.
In the centralized case Instantaneous triangulation are
filtered.
Figure 10. Location with centralized method.
Figure 11. Location with information fusion method
Due to the information fusion method the DATA LINK rate is
reduced.
VIII. CONCLUSION
By 2010-2020 the infosphere will be very different of what
we know today. There will be manned and unmanned
platforms in air/sea/land all connected via data links. The
requirements on amount of data to exchange will increase by
several orders of magnitude. It will be of first importance to
minimize data transmission while preserving important data.
The method presented here offers a very attractive alternative
to solve the bottleneck of data transmission.
ACKNOWLEDGMENT
The authors acknowledge the European Defence Agency,
the French DGA, the Swedish FMV and the Italian MoD,
which have contributed to the funding of this study under EDA
contract nB-0055-IAP2-ERG.
REFERENCES
[1] Y. Bar-Shalom, Multitarget-Multisensor tracking : Principles and
techniques , 3
ime
dition, 1995.
[2] S. Blackmann, R. Popoli, Design and analysis of modern tracking
systems , Artech House, 1999.
[3] C.Y. Chong, S. Mori, W.H. Barker and K.C. Chang, Architectures and
Algorithms for Track Association and Fusion , in IEEE AES Systems
Magazine, January 2000
[4] Olivier E.Drummond Track Fusion with Feedback SPIE vol 2759
[5] Y. Bar-Shalom, On the Track-to-Track Correlation Problem, in IEEE
Transactions on Automatic Control, vol. ac-26, n2, April 1981
[6] X. Rong Li, Y. Zhu and C. Han, Unified Optimal Linear Estimation
Fusion Part I : Unified Models and Fusion Rules, in Proceedings of
2000 International Conference on Information Fusion, Paris, France,
July 2000
[7] O. Drummond, A Hybrid Sensor Fusion Algorithm Architecture and
Tracklet, in Proceedings of SPIE, vol. 3163, 1997
[8] JF Grandin, L Ratton Procd et systme de pistage et de suivi
dmetteurs patent number : N 08 01715. date 28.03.2008 Extended
to European level : Date 12.03.2009 N09154963.4
[9] Xuezhi Wang, Rob Evans, Jonathan Legg Distributed Sensor Fusion
with Network Constraints SPIE vol 5429 2004
106 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Parasitic Doppler effect in passive location
Dario Benvenuti
Research and Advanced System Design
Elettronica SpA
Via Tiburtina Valeria, km 13.7, Rome, Italy
dario.benvenuti@elt.it
Abstract Large base interferometry is a fast and accurate
technique for passive emitter location. Experiments both in direct
and inverse configuration have been executed and in this paper
some results are presented in which the main sources of errors in
Doppler measurements are analyzed.
The observer attitude measurements need high accuracy, because
it can lead to large estimation errors. Vibrations can be a
problem in case of wings mounted antennas. The Doppler
induced by the scanning antenna phase center displacement has
been analysed as a possible source of error also in the inverse
configuration where much of the disturbances are eliminated.
Keywords--passive location; large base interferometry; parasitic
doppler
Passive location of surface emitters by an airborne observer
through the large base interferometry is much faster and more
accurate than standard triangulation and bearing-only tracking.
The principle is that the frequency difference of arrival
(FDOA) is related to the apparent angular velocity of the
emitter with respect to the observer,
( )
sin B
FDOA =
(1)
Moreover the time difference of arrival (TDOA) is related
to the direction of arrival (DOA), :
=
B
c TDOA
1
cos (2)
so with two differential measurements, FDOA and TDOA
location can be calculated:
( )
= sin
V
R (3)
where is the angle between the velocity vector and the
interferometer [1, 2].
The error analysis shows a very good behaviour of this
approach for emitter that are not pointed by the observer, and
that are not aligned with the interferometer antennas. In fact the
emitter on the observer velocity direction show a null angular
rate which in turn does not allow the range estimation, while
the angular measurement is bad conditioned around the
direction of the interferometer antennas [2].
I. EXPERIMENTAL TRIALS
A series of field test have been conducted by ELT in the
framework of a research contract with Italian MoD, with the
collaboration of Aeronautica Militare Italiana (AMI). The first
campaign was performed in 2006 with a Passive Location
Demonstrator (PLD) installed on a helicopter [2]. The
constrained baseline along with the low velocity of the
helicopter allows a reduced performance location.
A second field test has been recently carried on with the
collaboration of an aircraft. These trials were in the inverse
configuration, that is the PLD was standing on the ground and
the aircraft was emitting with his own radar. That is not a
problem because the relative motion is the physical entity that
is to be measured, provided the aircraft route and velocity is
known.
Figure 1. Localization result of a trial from [2].
The first campaign results have already been published [2]
showing the average result of about 10%, in this paper a bad
trial with a larger error will be analysed.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 107
The aircraft campaign yielded better results as it can be
expected given the higher velocity, a larger base and better
conditions due to inverse configuration. Some trials yield a
final location accuracy on the order of 2-3%, however some
trials have still an error higher than 5% and an hypothesis about
the error will be presented.
-16 -14 -12 -10 -8 -6 -4 -2 0 2
-6
-4
-2
0
2
4
6
X[km]
Y
[
k
m
]
True trajectory
Measurements
Filtering
Emitting positions
Figure 2. Localization result of a trial of the aircraft campaign.
II. SOURCE OF ERRORS IN HELICOPTER TRIALS
The main source of error in some of these trials has been
due to the helicopter attitude: in fact though the trajectory was
rectilinear and uniform some perturbations during the flight are
unavoidable and the attitude sensor may be not as accurate as
needed. Equation (3) with the contribution of own platform
rotation
P
becomes:
( )
= sin
P
V
R
(4)
where the contribution of
P
to the error is quite important [2].
Figure 3. Trial 1: platform own motion compared to LOS variation.
In figure 3 the measured yaw rate and the total estimated
are reported along with their difference and the expected line of
sight (LOS) variation. It is evident that the not perfect
cancellation of the own platform rotation is comparable to the
quantity to be measured, and in fact in this trial the localization
error has been very large (see figure 4)
0 5 10 15 20 25 30 35
0
10
20
30
40
50
60
70
80
90
100
Relative Time [s]
[
%
]
Localization Trial 17/07/2006, 09:31
Percentage Range Error
Figure 4. Trial 1: localization error.
Another possible source of error in operative conditions is
the antenna vibrations, especially for wing mounted antennas.
In these trials that was not the case because the antenna
structure was quite robust and rigid. Simulations have been
performed with different amount of vibrations and the results
are presented.
In Figure 5 a typical vibration spectrum is reported: the
z-component displacement is plotted as a function of frequency
and the corresponding velocity is evaluated. The main mode is
about 5 Hz with maximum displacement 30 cm, this should be
conservative for a combat aircraft; the higher modes present
less displacement but, due to higher frequency, still have a
considerable effect on velocity.
0 5 10 15 20 25 30 35 40 45 50
0
0.1
0.2
0.3
0.4
Z
[
m
]
0 5 10 15 20 25 30 35 40 45 50
0
5
10
15
20
f [Hz]
V
[
m
/
s
]
Figure 5. Vibrations spectrum: displacement (in blu) and velocity (in red)
The relative localization error due to vibrations is reported
in figure 6. In the left hand vibrations have been imposed as a
108 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
common mode on the two wings, while on the right they are
partially uncorrelated.
0 20 40 60
0
0.2
0.4
0.6
0.8
1
f [Hz]
R
e
l
a
t
i
v
e
e
r
r
o
r
Common mode vibrations
0 20 40 60
0
0.2
0.4
0.6
0.8
1
f [Hz]
R
e
l
a
t
i
v
e
e
r
r
o
r
Partially uncorrelated vibrations
Figure 6. Error induced by wings vibrations: (left) common mode vibration;
(right) partially uncorrelated vibrations
The effect of vibrations as a common mode on the two
antennas is just detectable and, being short time correlated, can
be averaged away. However as soon as a mismatched vibration
is set on the two wings a much higher effect arises which
severely penalizes the wing mount configuration; thus it is a
much better choice, also for the DOA estimation constraints, to
install the interferometer antennas on the aircraft fuselage than
on the wings.
III. ERRORS IN INVERSE CONFIGURATION TRIALS
In the inverse configuration the effect of platform attitude
variations is much less important and vibrations are not
present, by the way the antenna configuration is longitudinal
(fuselage mount) thus vibrations should be negligible also in
real conditions. Let's consider the following model (see figure
7) for the FDOA calculation, in which the origin of the
coordinate system coincides with the center of PLD, B is the
antennas distance along the x-axis, P(x
0
, y
0
) is the position of
the aircraft (i.e. the emitting antenna) at time t
0
, and V is the
aircraft velocity, assumed parallel to the x-axis.
Figure 7. Geometrical model for FDOA evaluation.
The FDOA is the time derivative of the following phase
difference:
( ) ( ) ( )
( ) ( )
+ +
+ =
2
0
2
0
2
0
2
0
2 /
2 / 2
t r y Vt B x
t r y Vt B x t
(5)
where the attitude variations are summarized in the (yaw
rate) and r is the distance of the emitting antenna from the
aircraft center of mass.
In a rectilinear trajectory heading variations of as much as
1/s have been observed, causing an error in the FDOA less
than 2 %, with the geometry depicted, where x
0
= 20 km, y
0
=
1 km.
In some trials a much larger error has been observed, like in
figure 8, where the true trajectory is reported in black solid line
along with the estimated positions in red circles. Here each
measurement is achieved on a burst of 128 pulses.
-25 -20 -15 -10 -5 0 5 10
-10
-5
0
5
10
X[km]
Y
[
k
m
]
True trajectory
Measurements
Figure 8. Trial a/c 14: localization result.
A possible cause of error is due to the antenna phase center
displacement during the scansion: if the phase center is not
coincident with the rotation center a parasitic Doppler effect
arises. The phase center displacement can exist also for phased
array antennas, if the scan phase differences are not symmetric
with respect to the geometric center of the array.
Figure 9. Antenna phase center displacement.
In figure 9 a) a simple 2-element array is depicted with
broadside configuration: the phase center is the geometric
center. When the array is scanning, if the phases are symmetric
the phase center remains fixed. Often phases of the array
elements are all the same sign, giving rise to a displacement of
the phase center, this is particularly evident with true time
delay phased array. This effect is meaningless for the radar
itself, and can be however corrected for, while it has an effect
on the FDOA estimation.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 109
To evaluate this effect (5) can be used again where now r is
the antenna half width and is the scan rate; for reasonable
values of the parameters (r = 0.5 m, = 2 rad/s) this yields an
error of 8% with the same geometrical setup as above.
This value certainly does not explain the large errors of
figure 8, but it must be noted that a phased array antenna can
point the beam in a pulse to pulse way, giving rise to a much
higher . That is the case of the radar used in the
experimentation, as it can be seen from the amplitude pattern
reported in figure 10: it shows the main beam and a replica
about 4 dB lower which is probably due to a second beam
position.
Unfortunately the exact working of the radar and the details
of its construction are not accessible to our analysis, but the
performed evaluation can be useful to interpret some
unexplainable results and in general to give a warning on the
subtleties that lay behind these kind of techniques.
0 0.01 0.02 0.03 0.04 0.05 0.06
-42
-41
-40
-39
-38
-37
-36
-35
-34
-33
-32
Time [s]
A
m
p
[
d
B
]
Figure 10. Pulse to pulse beam pointing of radar used in a/c trials.
IV. CONCLUSIONS
In this paper results of experimental trials of passive
localization through phase difference rate estimation are
presented. Several sources of error have been analyzed and
taken into account.
The first experimental campaign was performed by the use
of a helicopter and thus presented critical features like slow
platform velocity, perturbed straight flight and not perfect
attitude compensation.
The second campaign was performed in more controlled
conditions (inverse configuration) and with the use of a combat
aircraft; the results were surely better than the first campaign,
however still in some trials the error was surprisingly high.
An hypothesis on the aircraft radar has been done: the
phase center is conjectured to move with beam scanning, and
the effect has been evaluated with reasonable parameters
values showing a not negligible effect on localization.
Though we are not sure about the real working of the
aircraft radar the performed analysis can be useful to have an
insight on the localization techniques based on phase
estimation.
ACKNOWLEDGMENT
I wish to thank my colleagues involved in the passive
localization topic, in particular Antonio Zaccaron for the clarity
of his suggestions, and my head Daniela Pistoia for the support.
REFERENCES
[1] X.-P. Deng, Z. Liu, W. -L. Jiang, Y. -Y. Zhou and Y. -W. Xu, Passive
location method and accuracy analysis with phase difference rate
measurements, IEE Proc. Radar, Sonar Navig., Vol. 148, No. 5,
October 2001
[2] G. Severino, A. Zaccaron and R. Ardoino, Performance of a Doppler
based direct passive location technique, ESAV 2008, Capri, 3-5
September 2008
[3] K. Becker, An efficient method of passive emitter location, IEEE
Trans. on AES, vol. 28, No. 4, October 1992.
110 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
An in-air passive acoustic surveillance system for air
traffic control
GUARDIAN project
Vincenzo Quaranta and Salvatore Ameduri
CIRA, Italian Aerospace Research Centre
Vibration & Acoustics Lab.
Via Maiorise, Capua (CE), Italy
E-mail: v.quaranta@cira.it
Domenico Donisi and Marco Bonamente
DAppolonia S.p.A.
Largo Carlo Salinari, 18/19 00142 Rome Italy
Headquarters: Via San Nazaro 19 16145 Genova Italy
E-mail: domenico.donisi@dappolonia.it
AbstractThe air traffic control inside ATZ (Aerodrome
Traffic Zone) is a key activity for airport management services
to meet increased security and a low environmental impact on
air transport systems. The GUARDIAN system aims to the
development of an acoustic system as support for airplane
traffic control in aerodrome zone.
Keywords: traffic management systems, aircraft detection
and location, acoustic devices, beam-forming algorithms
I. INTRODUCTION
In recent years an increasing demand has interested
airports and the related air traffic management authorities.
The risk of collision between taking off and landing aircrafts
and between aircraft and ground vehicles due to increased air
traffic are issues that have highlighted the need to improve
airborne surveillance systems by means of real time aircraft
identification and tracking procedures in aerodrome zone.
State-of-the-art ATM (Air Traffic Management) systems for
aerodrome surveillance include especially radar technology.
On the other hand, there is currently a significant increase of
interest in the international scientific community, in the
definition of alternative acoustic systems for locating,
tracking and identification of moving acoustic sources with
particular attention to aspects of intruder aircrafts and for
monitoring the movement of ground forces. This interest
stems primarily from the fact that unlike radar detection,
acoustic detection can be performed with totally passive
sensors only by listening to the noise of the target. This
represent an obvious advantage from the environmental point
of view (no emissions of any kind), safety (no possibility of
locating the sensor in the absence of emission, inability to
drastically reduce the noise of the target), and costs (reduced
energy consumption due to lack of transmission power, lack
of critical components working at microwave, etc.). In this
paper, the preliminary results of MIUR (Italian Ministry of
University and Research) - funded GUARDIAN project are
reported. The aim of this research project is the design and
development of a novel acoustic system for the improvement
of co-operative management of ATZ (Aerodrome Traffic
Zone) control. ATZ is the air space of defined dimensions
interested by taking off and landing of aircrafts. The
proposed novel cognitive sensor offers flexibility and
adaptivity to the airport scenario while ensuring a neat
capability improvement in aircraft and ground vehicles
detection, location and tracking.
II. DESCRIPTION OF THE ACOUSTIC SYSTEM
The GUARDIAN sensor is a multi-modal, in-air passive
acoustic device working in arrayed/sparse configuration by
means of an innovative ensemble of digital processing
stages. This system allows the detection and tracking of in
flight aircraft and in take-off or landing by means of a
passive detection and spectral analysis of airplane acoustic
emission.
Figure 1. Physical architecture of the GUARDIAN system
The system prototype consists of two planar acoustic
sensors, one rotating (Master) and one fix antenna (Slave),
based on a passive phased array of randomly distributed
microphones and a command & control unit to allow the user
to manage and visualize tracking data. Figure 1 shows the
sketch of the system.
III. PHASED ARRAY DESIGN PARAMETERS
The performance of an acoustic array is determined by its
geometry (shape, number and spatial position of
microphones), which defines the response of the array, called
Array Pattern. After fixing the geometry, the frequency and
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 111
the antenna pointing direction, the Array Pattern shows the
attenuation performed by the antenna on the acoustic waves
incident from directions other than the pointing direction.
The MSL (Maximum Side lobe Level) value, which is the
ability of the array in reducing false noise sources in
directions other than the antenna pointing one, and the
resolution, which defines the minimum angular distance at
which the array is able to separate two nearby sources in
space, can be deduced from the Array Pattern. Thus, acoustic
array performance figures are: angular resolution (-3 dB
main lobe width), MSL, maximum range and working
frequency.
IV. GUARDIAN PROCESSING CHAIN
In present section, the Data Processor architecture is
reported. Scope is to highlight the different functionalities
that can be identified as sub-blocks of the processor and to
describe their logical interconnections. Architecture
describes the logic chain of the Processor and operates first
important design choices that highly impact on resulting
performances, complexity and computational load of the
Processor. Architecture design is thus a critical phase and is
best tackled keeping resulting schema as modular and
scalable as possible. Figure 2 shows the GUARDIAN
Processor architecture which is organized in the following
sub-blocks: Data acquisition (i), Beam-forming (ii), Detector
(iii), Triangulator (iv), Tracker (v) and Classifier (vi).
Figure 2. Logical architecture of the GUARDIAN processing chain
A. Data Acquisition
Data acquisition is the first step in Data Processing. In
this phase, acoustic samples are collected in the time-domain
at each array microphone and they are made available for
further processing. The parameters relevant for this
processing step are the time acquisition window, i.e. the time
interval in which data are acquired, and the sampling
frequency. In GUARDIAN system, the time acquisition
window is equal to T = 0.2 seconds, while data sampling
frequency is set to f
s
= 48 kHz. The data acquired by the
acoustic antenna are processed with the FFT operation in
order to translate them in the frequency-domain. The data
collected in time-domain, which are in form of a [Array
Elements x Time Samples] matrix, are then converted in a
[Array Elements x Frequency Bins] matrix.
B. Beam-forming
The operation of beam-forming has the scope of
electronically forming pointing beams for listening,
searching and/or tracking purposes. The SOI (Signal Of
Interest) for this kind of application is a broadband signal.
Most of the energy contribution of aircrafts spectral signature
is in the range 500 2500 Hz. For this reason, a frequency-
domain multiple beam-forming approach is used in order to
implement the acoustic data processing. Indeed a low
computational cost is achieved by splitting the signal into its
frequency components through an FFT and applying the
beam-forming algorithm to each component. Given the
planar localization of microphones on antenna array, spatial
beams are synthesized along azimuth () and elevation ()
direction by means of beam-forming operation. Acoustic
data are processed in parallel through two different filters:
the MBMF (Matched Beam-forming) and the MVDR
(Minimum Variance Distortion less Response). This choice
is due to the fact that MBMF and MVDR [1] exhibits
different benefits. MBMF presents: (i) robustness to non
idealities, such as steering mis-matches or the mis-
positioning of array elements, (ii) low computational cost
and an high-fidelity response i.e., the acoustic shapes are
preserved for spectral analysis. On the other hand, it provides
low accuracy, low resolution and it is sensible to the
presence of interferences. Moreover acoustic classification
issues may be faced by extracting spectral information from
detected acoustic targets by implementing MBMF beam-
forming in the angle/beam of detected target by means of
MVDR algorithm. For all the mentioned reasons, to
overcome the MBMF limitations, also the MVDR algorithm
is implemented. It is an adaptive beam former which has
high accuracy and high resolution. In addition, according to
its adaptively, it is robust to scenarios which include
interferences. However, it is computational expensive and it
does not preserve the signal phases, i.e. the acoustic shape of
signals is corrupted. Moreover, it requires the knowledge of
the correlation matrix with interferences and noise; more in
practice, since this knowledge is not given, correlation
matrix must be estimated on the basis of the received signal.
Due to their different properties, MBMF and MVDR are
both performed with different aims:
- The MBMF is used for detection (its performance
depends on circumstances) and classification purposes and it
is performed on the entire band of interest (from 300 Hz to 3
kHz);
- The MVDR is used for detection purpose and it is
performed only on a reduced band in which it is expected to
find power for the aircraft (from 1 kHz to 1.5 kHz).
C. Delay and Sum Matched Beamformer
In the matched beam-forming all microphone signals are
matched and summed in phase (coherent summation). For
each looking (steering) direction time delays have to be
applied corresponding to the arrival of a plane wave on each
microphone. In the frequency domain a time delay
corresponds to phase shift and therefore a frequency domain
implementation is the simplest implementation. The DSBF
112 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
(Delay and Sum Beam-forming) in the frequency domain is
formulated as:
=
N
n
n n n f
r k
c
f
j f X w f B
1
) , ( 2 exp ) ( ) , , ( (1)
where:
B
f
is the beamformer output in the frequency
domain;
f is the frequency and k is the wave vector;
, is the bearing to look at;
n is the microphone number and N is the number of
microphones;
r
n
contains the coordinates of the n-th microphone;
w
n
the MBMF weight (tapering coefficient) of
microphone n;
tapering coefficients are based on a Hamming
(raised cosine) window;
X
n
(f) is the Fourier component of microphone n at
frequency f. X
n
(f) is derived from the microphone
time-series x
n
(t) by FFT with N
FFT
points
corresponding to integration time of t
int
= N
FFT
/Fs;
c the sound speed value;
beam-forming coefficients C
n
can be computed in a
fast way by iteration making use of the equidistant
frequency grid. In this way no complex
exponentials have to be computed: C
n
+1 = C
n
* C
since:
+
r k
c
f
j r k
c
f
j r k
c
f
j
n n
) , ( 2 exp ) , ( 2 exp ) , ( 2 exp
1
(2)
D. Minimum Variance Distortionless Response
Beamformer
The MBMF described in the previous paragraph is the
solution that maximizes the gain if the noise is uncorrelated
(and no interferences are present). The adaptive beam-
forming bases its weights on the data collected by the
microphones (thats why they are called adaptive) and
consequently are scenario dependent. It can maximize the
gain in case of presence of correlated unwanted disturbances
impinging on the array. In GUARDIAN system, the adaptive
beam-forming is implemented as an MVDR (minimum
variance distortion less response) type of beam former in the
frequency domain:
1
1
) , ( 2 exp ) ( ) , ( 2 exp ) , , (
=
T
f
r k
c
f
j f r k
c
f
j f B R
(3)
where:
r is a vector of length N with the positions of the
microphones;
T
f x f x f R ) ( ) ( ) ( = is the N*N covariance
matrix of the microphones signals around frequency
f , where the underline denotes an average.
In case of uncorrelated noise on the array, R = I and the
adaptive beam former becomes a matched beam former.
E. Detection, Triangulation and Tracking
The detector will run on the high resolution adaptive
beam-forming data and in parallel on the conventional beam
former for a more robust detection. Indeed MVDR suffers
low SNR targets but has low false alarm rate, while MBMF
may have an high false alarm rate due to side lobes.
Normally, when a target is detected in an adaptive beam, the
nearest beam of the conventional beam-forming is used for
acoustic classification. This approach reduces the false alarm
rate. This detection algorithm consists of the following steps:
1. The beam energy
2
f
B is integrated over a
selectable frequency band; typically for the GUARDIAN
system f =1000-1500 Hz. The selected band depends on the
expected signal-to-noise ratios in the band, which will
depend on aircrafts spectra, acoustic propagation, processing
again, etc.
2. The resulting data in all beams are normalized by
removing their median value or more in a more robust way
by a moving mean.
3. In latter case, the outer beams should be removed
from the further process due to edge effects of the filter.
4. Across all remaining beams, local maxima are
identified. These local maxima are defined as those beams
that have a higher energy than their two neighboring beams.
5. The local maxima are compared against a user-
defined detection threshold (DT).
6. Local maxima that exceed DT are identified as
detections.
The above 6 steps are carried out for both receivers at
each temporal snapshot, resulting in two sets of DOAs
(Direction Of Arrival) where aircrafts are detected. These
two sets of bearings are then employed to find the actual
position in Cartesian coordinates of the aircraft by means of
a standard triangulation procedure. These plots feed the
tracker which gives the aircraft track as output [2]. The track
is the ultimate product of the Processor and can be
complemented with the aircraft classification made by the
Classifier. Finally, they are sent to the User interface and
duly presented on the screen.
V. MICROPHONES DISTRIBUTION OPTIMIZATION
Due to the wide amount of optimisation parameters, (i.e.
the microphones coordinates), an heuristic genetic approach
was adopted (see flow chart in Figure 3), [3]. According to
the natural selection principle, an initial population
constituted by a certain amount of individuals, each one
characterised by a specific genetic (chromosomes) is
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 113
subjected to environmental aggressions of different type
(predators, hostile weather, diseases). The higher the
individual adaptability level, the higher the surviving chance
and the possibility of transmitting own genetic to the next
generation. This transmission occurs through the cross-over
process: each individual, on the bases of its transmission
chance, will interchange a part of the chromosomes with
another individual, thus generating a new one, potentially
better. By repeating this process (selection and cross-over)
for an adequate number of times, the average performance
(adaptability or fitness) of the population is increased.
Anyway, since the first population contains a finite number
of information and the next ones result just from a
recombination of it, the improvement possibility is in
practice confined to the specific initial chromosomes values.
Hence, to hinder the premature convergence to a local
maximum and to enrich the genetic content of a population,
another phenomenon acts together selection and cross-over:
the mutation, i.e. the random and spontaneous variation of a
small part of the genetic code of some individual. Thus,
individuals slightly different enrich the current population,
guaranteeing a wider genetic content. Aforementioned logic
was implemented for the antenna optimisation problem. Each
population is constituted by a certain amount of antennas
(individuals), differing each other for microphones abscissas
and ordinates (chromosomes). The Maximum Side Lobe
Level (MSL) in dB was assumed as fitness function: the
lower the MSL, the better the performance. Constraints were
taken into account both for the optimisation parameter
domain (microphone locations are onto the plane of the
antenna) and for the global performance of the antenna. In
practice the locations of the microphones belong to an
uniform square grid, while a cut-off level for the angular
resolution was set, thus discarding antennas with
unacceptable resolution, even exhibiting a good MSL. In
order to fully investigate the dependence of the array
performance on antenna shape and dimensions, two different
sizes of the array have been investigated, both made of 256
microphones. The smaller was a 4 x 4 square meters array,
while the larger was a 6 x 6 square meters one. First of all, a
regular grid array fulfilling the array performance was
designed. In particular, the fixed distance between adjacent
microphones has been derived by the following formula
(assuming f
max
=1250 Hz), [4] [5] [6]:
max
min
2 2 f
c
d d
(4)
while the side length D, linked to the desired resolution at
the minimum frequency of interest, has been derived from:
( )
min
3
cos
1
f
c
R
z
D
=
(5)
At low frequency, if a good resolution is required, the
combination of a large antenna diameter and a small
microphones distance causes the need for a large number of
sensors. In order to reduce the number of microphones, only
n=256 microphones positions are randomly chosen from all
possible N locations (N=3600 for the larger array, N=1600
for the smaller array, thus n << N) of the regular mesh. An
optimization process was then performed to determine a
sparse array configuration with the best possible
combination of antenna performance parameters.
In TABLE I. adopted simulation parameters in the
optimisation procedure are summarised.
TABLE I.
Algorithm settings
Parameter Smaller array (4 x 4 m) Bigger array (6 x 6 m)
Number of iterations
(populations)
217 46
Number of individuals 4000 4000
Number of chromosomes
per individual (x,y,
coordinates of 256
microphones)
512 512
Cross-over %
(chromosomes
interchange percentage)
50 50
Mutation % (percentage
of mutated individual per
population)
10 10
Optimisation constraints
Microphone on a
uniform square grid (m)
grid step: 0.1
grid dimension: 4 x 4
grid step: 0.1
grid dimension: 6 x 6
Resolution (deg): max
(elevation, azimuth)
6 4
Figure 3. Genetic algorithm flow chart.
114 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Several optimisation processes were carried out, moving
from different initial populations. Figure 4 gives an idea of
the convergence ability of the algorithm for some of these
processes. At the end, different optimal results were
achieved, each one characterised by the minimum MSL
compatible with assumed constraints.
The optimal individuals were located onto a 2D graph
(see Figure 5 for the bigger antennas), whose axis refer to
MSL and angular resolution. In the same pictures, also
results achieved from a wide amount of random no optimised
configurations are reported, to highlight the quality of the
results obtained through the optimisation process.
0 20 40 60 80 100 120 140 160 180 200 220
0
2
4
6
8
10
12
14
16
18
Fitness function vs iteration for the 6x6 m Grid Random Array
Population number
F
i
t
n
e
s
s
f
u
n
c
t
i
o
n
(
d
B
)
Max
Mean
Figure 4. Fitness function vs iteration for the 6 m square antenna.
Figure 5. 100.000 randomly generated arrays (GREEN cross) Vs.
optimized configuration (RED star) Size 6x6 m.
VI. ARRAY PERFORMANCES
Figure 6 and Figure 7 summarise optimized antennas
performance parameters (angular resolution and MSL) vs.
elevation angle of antenna pointing direction in the case of
MBMF approach. According to the theory, the angular
resolution strongly decreases with the elevation steering
angle, while MSL, mainly depending on sensors number,
does not exhibit large variations.
Figure 6. Larger antenna performance parameters
Figure 7. Smaller antenna performance parameters
The corresponding optimal microphone distribution is
depicted in Figure 8 and Figure 9.
Figure 8. Larger antenna optimal microphones distribution
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 115
Figure 9. Smaller antenna optimal microphones distribution
TABLE II. and TABLE III. summarise antennas
performance parameters for a MBMF approach. The
elevation linear resolution has been evaluated at a distance of
7 km. The range of the antenna, i.e. the maximum distance in
the steering direction at which the array can yet detect the
sound source, has been determined by considering a tonal
(with different frequencies) sound source with an SPL of 140
dB at 10 m and a tonal (at the same frequency) background
noise with an SPL of 60 dB. Only the sound attenuations due
to air absorption (frequency dependent) and spherical
divergences were considered. As reported in TABLE IV. and
in, a range of 7 km can be theoretically achieved for a sound
source with a tonal frequency of 1250 Hz and an array with
MSL of -16 dB (i.e. with 256 microphones).
TABLE II.
Steering elevation Elevation angular Elevation linear MSL
angle (deg) resolution (deg) resolution (m) @ 7km (dB)
30 2.77 338.68 -16.07
60 4.80 587.81 -16.36
GRID RANDOM ARRAY (6x6 m)
TABLE III.
Steering elevation Elevation angular Elevation linear MSL
angle (deg) resolution (deg) resolution (m) @ 7km (dB)
30 3.96 484.58 -16.79
60 7.07 868.17 -16.72
GRID RANDOM ARRAY (4x4 m)
TABLE IV.
Sound source-array Spherical divergence Air absorbing Gain over Background
distance (m) attenuation (dB) attenuation (dB) Noise (dB)
10 0.00 0.05 95.95
20 6.02 0.10 89.88
40 12.04 0.20 83.76
80 18.06 0.40 77.53
100 20.00 0.51 75.50
200 26.02 1.01 68.97
400 32.04 2.02 61.94
500 33.98 2.53 59.50
1000 40.00 5.05 50.95
2000 46.02 10.10 39.88
3000 49.54 15.15 31.31
4000 52.04 20.20 23.76
5000 53.98 25.25 16.77
6000 55.56 30.30 10.14
7000 56.90 35.35 3.75
8000 58.06 40.40 -2.46
9000 59.08 45.45 -8.53
10000 60.00 50.50 -14.50
Gain over Background Noise @ array - Tonal noise=140 dB @10 m
Array gain=16 dB - Background noise=60 dB - Air: 15 C & 70%U.R.
-10.00
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Sound source-array distance (m)
G
a
i
n
V
s
.
B
a
c
k
g
r
o
u
n
d
n
o
i
s
e
(
d
B
)
500 Hz
1000 Hz
1250 Hz
2000 Hz
4000 Hz
5000 Hz
Figure 10. Array gain over background noise Vs. frequencies
VII. CONCLUSIONS
The GUARDIAN prototype system based on planar
acoustic antenna has been described, showing the results of
beam-forming technique for aircrafts detection.
The results herein illustrated proved the possibility of
using microphone arrays for air traffic control inside ATZ.
However, in order to achieve performance parameters
comparable to those of conventional radar, the use of
antennas with great diameter (tens of meters) and with a
large number of microphones (1000) is required. The use of
MEMS (Micro Electro-Mechanical Systems) digital
microphones will make antennas of this typology
economically and technically feasible.
For what concern the improvement of the array resistance
to harsh environmental operating conditions usually found at
airports (high humidity, rain, wind, low and high
temperature,), the development of innovative antenna
employing FBG (Fiber Bragg Grating) sensors will allow the
implementation of permanent outdoor stations.
ACKNOWLEDGMENT
The product was developed within the MIUR-funded
GUARDIAN project.
REFERENCES
[1] H.L. Van Trees; Detection, Estimation, and Modulation Theory, John
Wiley & Sons, 1968
[2] Y. Bar-Shalom et al., Estimation with Applications to Tracking and
Navigation, Johon Wiley & Sons, 2001
[3] Genetic Algorithms in Search, Optimisation & Machine Learning,
David E. Goldberg, 1989, Addison-Wesley Publishing Company,
Inc., pp. 60 88.
[4] J.J. Christensen and J. Hald, Beam-forming, Technical Review N. 1
2004, Bruel & Kjaer Sound & Vibration Measurement A/S, DK-2850
Naerum, Denmark.
[5] C.A.Balanis Antenna Theory Analysis and Design, Third Edition, a
John Wiley & Sons, Inc., Publication.
[6] M. Brandstein and D. Ward, Microphone Arrays, Springer, Berlin
2001, pp. 157-177.
116 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
OnboardPCLsystemsforairborneplatformprotection
Session5.1page119
FMBasedPassiveCoherentRadar.
Fromdetectionstotracks
Session5.2page123
HighRangeResolutionMultichannelDVBTPassiveRadar:
AerialTargetDetections
Session5.3page129
RangeOnlyMultistaticTrackinginClutter
Session5.4page133
i
t
t
c
Fv
j
T R
dt e
c
r
t X t X t w v r y
0
2
*
) ( ) ( ) ( ) , (
S
(2)
where
*
T
X is the reference signal, received by the beam
directed towards transmitter, ) (t X
R
is the signal from
surveillancebeam,
i
t is theintegration time, F is the carrier
frequency and ) (t w is weighting function. Thetarget bistatic
range r and velocity v areestimatedas thecoordinates of the
maximumof function given by (2). TheCartesian coordinates
of detect targets are calculated by finding the cross-section
points of at least 3 ellipsoids of bistatic distances r and focal
points placed in the transmitter and radar positions. Thus the
own platform position (e.g. from GPS) and position of all
transmitters of opportunitymust be known.
Theconcept of airborne PCL system is presented in Fig. 2.
The illuminators of opportunity are the DVB-T transmitters,
while they have sufficient transmitting power, good, time-
invariant cross-ambiguity properties and good coverage.
120 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
The required detection range is almost one order of
magnitude smaller than theoretical detection range, predicted
by formula (1) and presented in Fig. 1, so theproblemof low
elevation coverageof DVB-T is not of high importance, while
it is possible to exploit vertical side-lobes of transmitter
antenna for illumination. The single-frequency network is a
benefit, while it is possible to tunethe radar receivers only to
one frequency and obtain multi-transmitter tracking system.
The typical integration time is 0.1..1 s so tracking can reach
required accuracy within 1-5 s, and thus reaction time is sped
up by 3 times. In such case the detection range can be
shortened to c.a. 6 km. The targets location is estimated
calculating ellipsoids crossing, as stated in Fig. 2.
One of the basic problems in passive airborne radar is
cancelation of ground clutter. In ground-based PCL system
ground clutter has no Doppler spread and can be easily
canceled using adaptivelattice filter [14]. In PCL radar placed
on moving platform clutter signal has significant Doppler
spread caused by theplatformmotion. Theclutter-free region
is limited in rangeby theplatformelevation (when only direct
signal compete with the target echoes) and in Doppler by
platform velocity. Using classical Doppler spread clutter
cancelation method described in [17], not only clutter but also
thetargets of interest can be canceled.
Figure2. Theairborne PCL scenario.
The other possibility is to apply adaptive space-time
processing (STAP) to canceled theground clutter. The classical
STAP algorithms are derived for active radar, where all
sounding pulses arethesame. In PCL system theilluminating
signal is continuous and due to changes of information
contents, thetime shifted parts arenot identical. As theresults,
STAP algorithms for airborne PCL systems have to be
modified to obtain required properties.
III. EXPERIMENTA RESULTS
To gain deeper knowledge of thePCL signal properties the
set of experiments wereperformed at Radar Technology Lab at
Warsaw University of Technology(WUT). Thefirst part of the
experiments were performed using Raw Radar Signal
Simulator, simulating thescenario using FM radio and DVB-T
signals for scene illumination and 6 elements antenna array.
Thesignals fromall antennas wereamplified by selective FM
band COTS amplifiers (88-108 MHz) and directly digitized.
The digital signals were down converted to the baseband where
IQ digital decoding was perform. The IQ signals from 2
selected FM channels (100 kHz each) were recorded on the
hard drive, and processed offline.
The experiments proved that the direct use of classical
STAP do not ensure good results, while combining the
CLEAN technique [17] for direct signal cancelation with
modified STAP approach givessignificant improvement. Up to
now only FM based experiments on moving platforms, namely
on car platform[11] and on BRYZA airplane [12] werecarried
out at WUT Radar Technology Lab(Fig. 3).
Figure3. Bryza airplane (left) and theFM antenna mounted temoprarly at
theplane window (right)
The result of signal processing using combination of both
methods is presented in Fig 4. It can be seen that theresidual
clutter returns arepresent in thepicture, but also thetarget is
clearly seen.
Figure4. Detection results using CLEAN method combined with simplified
STAP amplituse (in dB scale) color coded
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 121
Theresidual clutter can mask someof theslow targets, so
further development of efficient signal processing algorithms is
required.
The experiments with DBV-T illuminators are scheduled
for summer 2011.
IV. CONCLUSIONS
The paper presents the idea of Airborne Passive radar
dedicated for aircraft and UAVs platformfor air surveillance
for thepurpose to increase air safety and avoid collisions in air-
space. It can be used for men-operating platforms, but main
benefits will be to connect thePCL radar output to autopilot to
provide safety for unmanned platforms. The first trials on
passive airborne radars performed in Poland [12, 16] and
England [13] shows that this technology can be used in near
future.
ACKNOWLEDGMENT
The authors would like to thanks Polish ministry of Science
and High Education for support under scientific grant
5525/B/T00/2010/39and Polish Cost Guard for cooperation.
This work was partially supported by the European Union in
the framework of European Social Fund through theWarsaw
University of Technology Development Programme, realized
by Center for Advanced Studies.
REFERENCES
[1] P. E. Howland, D. Maksimiuk, and G. Reitsma, 'FM radio based
bistaticradar, IEE Proc. Radar, Sonar and Navigation, vol. 152, no. 3,
pp. 107115, J une2005.
[2] B. D. Nordwall, ' `Silent Sentry` a new type oI radar, Aviation
Weekand SpaceTechnology, no. 30, pp. 7071, 1998.
[3] A. Guner, M. A. Temple, R.L. Claypoole Jr. 'Direct-path filtering of
DAB waveIorm Irom PCL receiver target channel, Electronics Letters,
vol. 39, Issue: 1, 9 J an. 2003, pp:118 119.
[4] D. Gould, R. Pollard, C. Sarno, and P. Tittenso, 'A multiband
passiveradar demonstrator, in Proc. International Radar Symposium
2006, Cracow, Poland, 2426 May 2006, pp. 657660.
[5] A. D. Lallo, A. Farina, R. Fulcoli, P. Genovesi, R. Lalli, andR.
Mancinelli, 'Design, development and test on real data oI an FM based
prototypical passive radar, in Proc. RadarCon, Rome, Italy, 2630 May
2008, p. CD.
[6] D. Poullin and M. Flecheux, 'Recent progress in Passive
CoherentLocation (PCL) concepts and techniques in France, in Proc.
RadarCon,Rome, Italy, 2630 May 2008, p. CD.
[7] P. E. Howland, 'Target tracking using television-based bistatic radar,
IEE Proc.-Radar, Sonar & Navig., Vol. 146, No. 3, J une1999.
[8] H. Kuschel, M. Glende, J . Heckenbach, S. Mller, J . Schell, and C.
Schumacher, 'Experimental passive radar systems using digital
illuminators (DAB/DVB), in Proc. International Radar Symposium
2007, Cologne, Germany, 57 September 2007, pp. 411417.
[9] M. Malanowski, K. Kulpa, and J . Misiurewicz, 'PaRaDe Passive
Radar Demonstrator family development at Warsaw University of
Technology, in Proc. Microwaves, Radar and Remote Sensing, Kiev,
Ukraine, 2224 September 2008, p. CDC.
[10] J. Baker, H. D. GriIIiths, and I. Papoutsis, 'Passive Coherent
Locationradar systems. Part 2:WaveIorm properties, IEE Proc. Radar,
Sonar and Navigation, vol. 152, no. 3, pp. 160168, J une2005.
[11] B. Dawidowicz and K. Kulpa, 'Experimental results Irom PCL radar on
moving platIorm, in Proc. International Radar Symposium 2008,
Wroclaw, Poland, 2123 May 2008, pp. 305308.
[12] K. Kulpa, M. Malanowski, J . Misiurewicz, M. Mordzonek, P.
Samczyski, and M. Smolarczyk, 'Airborne PCL radar: the concept and
primary results, in Proc. Military Radar 2008, Amsterdam, The
Netherlands, 2729 October 2008, p. CD.
[13] J. Brown, K. Woodbridge, A. Stove, and S. Watts, 'Air target
detectionusing airborne passive bistatic radar, ELECTRONICS
LETTERS, vol. 46, no. 20, September 2010.
[14] K. Kulpa, 'Ground clutter suppression in noise radar, in Proc.
International Conference on Radar, Tuluse, France, 1822 October
2004, p. CD.
[15] M. Malanowski, G. Mazurek, K. Kulpa, and J. Misiurewicz, 'FM
basedPCL radar demonstrator, in Proc. International Radar Symposium
2007, Cologne, Germany, 57 September 2007, pp. 431435.
[16] M. Malanowski, K. Kulpa, M. Mordzonek, and P. Samczynski,
'PaRaDe reconIigurable soItware deIined passive radar, in NATO
Specialist Meeting SET-136, Lisbon, Portugal, 2325 J une2009, p. CD.
[17] K. Kulpa, 'Simple sea clutter canceller Ior noise radar, in Proc.
International Radar Symposium 2006, Cracow, Poland, 2426 May
2006, pp. 299302.
[18] U. Nickel, 'Passive radar operation with GSM base stations features
and concepts, 2nd Passive Radar FHR-PCL-focus day(s), 17-18
November 2009, Wachtberg, Germany, pp. CD
[19] P. Samczynski, K. Kulpa, M. Malanowski, P. Krysik, L. Malikowski,
,A Concept of GSM-based Passive Radar
for Vehicle Traffic Monitoring, to be published in proceedings of
MRRS-2011, 25-27 August 2011, Kiev, Ukraine.
[20] F. Colone, P. Falcone, C. Bongioanni, P. Lombardo, 'WiFi Based
Passive Bistatic Radar Ior Local Area Surveillance, 2nd Passive Radar
FHR-PCL-focus day(s), 17-18 November 2009, Wachtberg, Germany,
pp. CD
122 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
FM Based Passive Coherent Radar
From detections to tracks
Radek Plsek, Vojtch Stejskal, Martin Pelant, Martin Vojacek
Advanced programs
ERA a.s.
Prumyslova 387, Pardubice, Czech Republic
rplsek@erabeyondradar.com
Abstract This paper provides description of individual steps
leading to successful targets tracking in multichannel passive
coherent system. Emphasis is given on targets association
principle in bistatic space prior to Cartesian tracking. The angle
of arrival from antenna array as a mean of targets association
and its properties is discussed.
Keywords - Passive Coherent Location; multistatic radar;
tracking; Angel of arrival
I. INTRODUCTION
A long term experience with military Electronic Support
Measurement (ESM) systems like Tamara and recent Vera-N
was motivation for company to follow with research dedicated
to passive radars technology usually called Passive Coherent
Location (PCL). In contrary to ESM system which exploits
different kind of signal emissions generated by target, the PCL
utilizes reflected signal and therefore it is able to detect non
cooperative and quiet targets by means of bistatic Radar Cross
Section (RCS). Bistatic configuration is also very suitable for
detection of stealth targets because their primary focus is on
monostatic RCS minimalization and property of non
electromagnetic emissions which makes them undetectable.
Moreover, the PCL system can utilize wide range of
commercial used illuminators of opportunity within VHF and
UHF bands.
Figure 1 Passive Multistatic Coherent Radar (one receiver with multiple
transmitters configuration).
II. SYSTEM CONCEPT
The PCL system being under development in ERA consists
of one receiver site and it is proposed to process signal from
four transmitters of opportunity, hence the system falls in the
group of multistatic configurations (see figure 1). It works with
signals intended for commercial terrestrial radio broadcasting
in band from 88 MHz to 108 MHz, where speech or music is
frequency modulated with ca. 100 kHz bandwidth. Conception
oI so called direct sampling`, i.e. sampling without any
frequency down-conversion, is being used. Therefore, only a
very simple analogue front-end is needed which consequently
yields in maximum cross channels` coherency.
A. Analog signal processing
Figure 2 Individual elements pattern (left) and combined patterns after
applied beam forming (right).
The proposed system consists of circular antenna array with
eight horizontally placed elements. Each element is connected
into signal divider which gives arise of required number of
output signals. It is possible to slightly reduce required high
dynamic range with analogue beam forming technique.
Therefore, each adjacent elements` pair (i.e. signals coming
from adjacent dividers) is connected to passive beam-former
consisting of delay links and attenuators network. With proper
beam-former settings, the minimum of radiation pattern
derived from combined antenna elements (pair) is set to
direction of the transmitter whose signal is being processed
(see figure 2). After that, all signals are passed through filter
and ampliIier. The receiver`s analogue signal processing chain
ends with an Analogue-to-Digital Converter (ADC). All
subsequent operations on signals (radio channel selection,
filtration, etc.) are made digitally. Since the output signal
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 123
bandwidth is ca. 150 kHz, signals are decimated, and therefore,
the native sampling ADC speed is reduced from 125 MSps to
approx. 240 kSps and complex signal is generated.
B. Digital signal processing
Digital signals in complex form are moved forward to
sophisticated digital signal processing chain which is
appropriately divided into two basic blocks: Multichannel
coherent processing and Air picture acquisition (see figure 3
for more details).
Figure 3 Signal processing chain block diagram
Signal from each antenna pair is processed separately on
Graphic Processing/Central Processing Units starting with
ground clutter rejection based on adaptive filtering within using
the auto-ambiguity function of reference signal. At this stage,
the static targets (clutter) are removed from the signal and only
moving targets remain. Subsequently the matched filter is
applied i.e. Cross Ambiguity Function (CAF) is computed for
each processed signal whereas peak in CAF represent elliptical
distance and elliptical Doppler frequency shift of appropriate
target. The longer is the integration interval, the lower (hence
better) the sensitivity of the system is, however its maximal
length is limited by required refreshing rate of the system
(bellow one second typically) as well as target`s
maneuverability possibilities. It is reasonable to notice that
different kind of modulation signals (speech vs. music) lead to
different power of ambiguity function side lobes and for that
reason Sequential Target Elimination algorithm is applied
instead of widely used simple Constant False Alarm Rate
detector (CFAR) to discover possibly hidden peaks in CAF
surface.
Every single detected peak is equipped with a number of
parameters and such detections, or plots, are further processed
by Air picture acquisition block.
C. Coverage prediction
The proposed PCL system`s parameters, especially
coverage, can be roughly predicted considering
transmitter/receiver geographic location and other parameters.
Taking into account omnidirectinal radiation pattern for both
transmitting and receiving antenna together with their real
location, it is possible to obtain signal to interference (reflected
to reference signal) ratio Ior constant target`s RCS 10dBsm. It
can be seen impossibility to detect any target with such system.
On the other hand, the situation will change significantly by
addition the processing gain` concerning adaptive clutter
rejection (ca. 40 dB depends on modulation type), CAF (ca.
48 dB depends on number of samples entering each
integration) and rejection by antenna pattern`s null steering (ca.
10 dB depends on particular pair of antenna elements).
Contours of the minimum required RCS to let the target be
detectable by system with consideration of all mentioned
effects are depicted in figure 4 for two transmitters. The
decreasing system`s detection capability behind transmitter is
easily observable. Note both subplots in the figure are results of
real radiation pattern measurements and their adjustment by
beam-former.
Figure 4 Covergade predictions: minimal detectable RCS [dBsm] contours
for current antenna patterns and beam-former settings for Krasne (left) and
Cerna Hora (right) transmitters
III. DETECTIONS
The system is still under development, and therefore its
efficiency is measured by means of only one FM radio
processing at the moment. Detections across all individual
processed channels (i.e. signals from individual elements)
during ten minute record are depicted in figure 5 as an example
oI system`s capabilities. There is significantly lower number of
detected targets in channel number 8. This is given by fact, that
this signal has the same contents as reference one. The figure 6
than, contains detections from all channels joined in single
Range vs. Doppler plane. Note several detections in bistatic
range up to 260km observed.
124 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Figure 5 Detections in individual channels (top to bottom channel 1 to
channel 8)
Figure 6 Detections from all channels in single range vs. Doppler plane
Since exact target 3D positions cannot be determined by
such system, the reIerence 'ground truth system output with
exact target positions in Cartesian space was involved into
testing.
The reIerence is taken Irom ERA`s Multilateration
Surveillance System (MSS) which is situated in area of
Pardubice in the Czech Republic. Its coverage is angularly
dependent, typically 400 km in range. Comparison between
PCL and MSS is performed in range-Doppler plane, when a
real time data from MSS system are converted from Cartesian
into this coordinate system. Each detected PCL target is then
associated with its nearest MSS target if they meets condition
of minimal range-Doppler distance. This threshold is formed
by an ellipse which center is situated at location of any MSS
target, and its semimajor axis is 3 km and semiminor axis is 10
Hz. The PCL system efficiency is than evaluated in such
manner. Target association also provides overview of detected
targets in Cartesian space (see example in figure 7).
Figure 7 PCL detections associated with MSS ground truth (transmitter
'Cerna Hora)
IV. ANGLE OF ARRIVAL
The Angle of Arrival (AoA) as a mean of targets
association from different FM channels could be derived based
on signal`s phase and amplitude conditions among individual
antenna array elements and their radiation pattern knowledge.
Let us mark vectors representing complex samples of
signals from each channel (i.e. antenna element or beam former
output) as
defined as
where
=
=
(1)
where N
c
is the number of channels, f
m
is the carrier
frequency for the m-th channel and ( )
m
s t is the complex
envelope of the m-th channel. Under the assumption that the
N
c
channels are equally spaced, it is possible to write f
m
as
0 f
f m + where
f
represents the channel bandwidth. If
( )
ref
s t is downconverted respect to f
0
, it is possible to write
the complex envelope of the signal as:
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 129
( ) ( )
1
2
0
c
f
N
j m t
ref m
m
s t s t e
=
=
The DVB-T multichannel ambiguity func
signal ( )
ref
s t can be written as:
2 *
1 1
2 2 ( *
0 0
1 1
2 2 *
0 0
( , ) ( ) ( )
( ) ( )
( ) ( )
d
c c
f f
c c
f f
j f t
d ref ref
N N
j m t j p t
m p
m p
N N
j p j m t j
m p
m p
f s t s t e dt
s t e s t e
e s t e s t e
+
= =
+
= =
= =
=
=
Under the following assumptions:
( )
m
s t is a bandwidth-limited
bandwidth equal to 2B)
the signal bandwidth is always
channel bandwidth, 2
f
B (i.
do not overlap)
the Doppler frequency is n
respect to the signal bandwidth, f
it is possible to rewrite eq. (3) as:
( )
1
2 *
0
1
2
0
( , ) ( ) ( )
,
c
f
c
f
N
j p
d p p
p
N
j p
p d
p
f e s t s t e
e AF f
=
=
=
where ( ) ,
p d
AF f is the ambiguity funct
DVB-T channel. Under the realistic assum
auto-ambiguity function of a generic single
exhibits the same main characteristics,
simplified to:
( )
( )
(
(
1
2
0
| ( , ) | | ,
sinc
,
sinc
c
f
N
j p
d d
p
c
d c
f
f AF f e
N
AF f N
=
(4)
tion for a single
mption that the
DVB-T channel
eq.(4) can be
)
)
|
f
f
(5)
nge resolution is
e single DVB-T
nction relative to
he multichannel
els
c
N , and the
and the sidelobe
DVB-T channels
oftware Defined
to obtain the AF
m. The central
sed signal shows
Fig. 1 DVB-T mu
In this case, the ambiguit
and compared with the one
channel. Plots of the ambig
(range) and Doppler frequenc
worth noting that the range
times with respect to the sin
Doppler profile maintained th
Fig. 2 Multichannel AF from
3-D view, Range p
The ambiguity function prese
peaks due to the known st
ultichannel spectrum
ty function have been computed
e obtained for a single DVB-T
guity function along time delay
cy are represented in Fig. 2. It is
e resolution is improved by N
c
ngle DVB-T channel, while the
he same behaviour.
real data (starting from the top):
rofile, Doppler profile
ents unwanted deterministic side
tructure of the DVB-T signal,
130 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
which includes pilots, guard intervals, and the guard band
between adjacent channels. In this work a novel signal pre-
processing technique based on DVB-T reference signal
power spectral density (PSD) [7] is used to reduce these
peaks as shown in Fig. 3.
Fig. 3 Multichannel AF from real data 3-D view after reference
signal pre-processing
III. EXPERIMENTAL SET UP
The equipment that has been used in this experiment is
composed by commercial off-the-shelf low cost TV
antennas, two synchronized Ettus USRP2 board equipped
with a RF front-end tunable from 800 MHz to 2400 MHz.
The main technical specifications of the USRP2 are:
FPGA Xilinx Spartan 3-2000 EP1C12 Q240C8
Cyclone
2 High-Speed Analog to Digital Converters (ADCs)
operating at 14 bits with a sampling rate of 100
Mega-samples per seconds (100 MS/s)
2 High-Speed Digital to Analog Converters (DACs)
operating at 16 bit with a sampling rate of 400 MS/s
Gigabit Ethernet interface
The antenna used during preliminary measurements and
the experiment for the target channel is a Yagi-Uda antenna
with a receiving gain equal to 18 dB and a Half Power Beam
Width of 20 degrees in the horizontal plane. On reference
channel, a Yagi-Uda antenna with a gain of 15 dB has been
used.
IV. EXPERIMENTAL RESULTS
The experiment scenario geometry is shown in Fig. 4.
Specifically, the receiver was located at the Department of
Information Engineering in Pisa and the DVB-T transmitter
was 14 km away from the receiver at 36 North-East as
indicated by the red arrow in Fig. 4. Moreover, the
surveillance antenna was pointed at 15 of azimuth and 30
of elevation. The targets of interest were airplanes taking off
from the nearby Pisa airport. Fig. 4 shows the trajectory of
the considered target. The expected Doppler frequencies for
the target in the surveillance area are shown in Fig. 5.
Fig. 4 Experiment Scenario geometry and target trajectory
Fig. 5 Expected Doppler Frequencies for a target that is
moving between the two red dots of Fig. 4 trajectory
Fig. 6 presents a photo of the detected target and
summarizes the target main technical information.
British Airways Boeing 737-400
Technical information Values (metres)
Length 36.5
Wingspan 28.9
Height 11.1
Fig. 6 Main technical information of the detected target
The reference and surveillance channels have been
simultaneously acquired with the equipment presented in
section III. Then the pre-processing technique has been
applied and finally the Cross-Ambiguity Function (CAF)
relative to three adjacent DVB-T channels has been
evaluated (Fig. 7). The peak due to the target echo is clearly
visible at the 87
th
range bin (i.e.: around 1700 m for the
geometry considered in Fig. 4).
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 131
Fig. 7 CAF of the surveillance area
Moreover, a Doppler frequency value of -169 Hz is in
accordance with the expected velocity (more than 450
km/h).
Fig. 8 Target range profile for a single DVB-T channel (blue
line on the top) and for three adjacent DVB-T channels (red
line on the bottom)
In order to better evaluate the range resolution
improvement, the CAF has been calculated and compared
for one and three DVB-T channels. Particularly, Fig. 8
presents the range profile along the Doppler frequency of
the target echo (i.e.: -169 Hz). Considering the geometry in
Fig. 4, the bistatic range resolution achievable by using a
single channel is around 57 m, whereas exploiting three
adjacent channels it is around 18 m. It is worth noting as the
range resolution relative exploiting three DVB-T channels
(red line) is improved respect to the single DVB-T channel
case (blue line). As a matter of fact, the blue line range
profile shows only one main peak while two peaks are
clearly visible on the red line one. This result is actually
consistent with the target size reported in Fig. 6.
V. CONCLUSIONS
In this paper, the exploitation of multiple DVB-T channels
for a passive radar system has been considered in order to
enhance the radar range resolution. A preliminary study has
been focused on the ambiguity function analysis in order to
verify the suitability of a multichannel DVB-T signal for
radar applications. The theoretical study has been supported
by preliminary measurements relative to three adjacent
DVB-T channels. Experimental results in aerial scenario
have been carried out and discussed. As a matter of fact,
high resolution DVB-T passive radar can be the first step to
perform passive radar imaging and target classification.
REFERENCES
[1] H D. Griffiths and C. J. Baker, Passive Coherent Location radar
systems. Part 1: Performance prediction. Radar, Sonar and
Navigation, IEE Proceedings, vol. 152, no. 3, pp. 124 132, 2005.
[2] .C J. Baker, H. D. Griffiths, and I. Papoutsis, Passive coherent
location radar systems. part 2: waveform properties, Radar, Sonar
and Navigation, IEE Proceedings -, vol. 152, no. 3, pp. 160168,
2005.
[3] P. E. Howland, Target tracking using television-based bistatic
radar,Radar Sonar and Navigation IEE Proceedings, vol. 146, no. 3,
pp. 166174, june 1999.
[4] Olsen, K.E.; Woodbridge, K.; , "Analysis of the performance of a
multiband passive bistatic radar processing scheme," Waveform
Diversity and Design Conference (WDD), 2010 International , vol.,
no., pp.000142-000149, 8-13 Aug. 2010
[5] Bongioanni, C.; Colone, F.; Lombardo, P.; , "Performance analysis of
a multi-frequency FM based Passive Bistatic Radar," Radar
Conference, 2008. RADAR '08. IEEE , vol., no., pp.1-6, 26-30 May
2008
[6] Conti, M.; Berizzi, F.; Petri, D.; Capria, A.; Martorella, M.; , "High
range resolution DVB-T Passive Radar," Radar Conference
(EuRAD), 2010 European , vol., no., pp.109-112, Sept. 30 2010-Oct.
1 2010
[7] M. Conti, D. Petri, A. Capria, M. Martorella, F. Berizzi, E. Dalle
Mese, "Ambiguity Function Sidelobes Mitigation in Multichannel
DVB-T Passive Bistatic Radar", accepted to International Radar
Symposium (IRS) 2011, September 7-9, 2011, Leipzig, Germany
132 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Range Only Multistatic Tracking in Clutter
Darko Mu sicki and Taek Lyul Song
Department of Electronic Systems Engineering,
Hanyang University, Republic of Korea
Darko.Musicki@gmail.com and tsong@hanyang.ac.kr
AbstractThis paper presents an algorithm for multistatic
target tracking in clutter, using only range difference information
(neither bearing nor Doppler information are assumed available).
Presence of false tracks, Data association issues as well as the non-
linear measurement equation makes this a challenging problem.
This paper proposes a solution to this problem by using the
Gaussian Mixture Measurement likelihood - Integrated Track
Splitting algorithm.
I. INTRODUCTION
1
Multistatic tracking involves using non-collocated transmit-
ters and receivers to track the targets. One transmitter can
be paired with one (bistatic) or many receivers (multistatic
conguration). The receiver measures the time difference
between the signal propagated directly, and the signal reected
from the target. This setup has many practical benets, as well
as numerous challenges, and is an active area of research, a
nice overview can be found in [1].
In this paper we also assume no angle information, which
introduces a severe measurement non-linearity and also ne-
cessitates using more than one stationary receiver (sensor)
to ensure system observability. The situation is made more
complex by uncertain detections and the presence of spurious
measurements (clutter). The uncertain presence of target(s)
also increases the problem dimensionality.
The signal to clutter ratio is usually low enough so that a sig-
nicant number of clutter detections are present in each scan.
The tracks are initialized and updated using measurements,
thus both true tracks (which follow targets) and false tracks
(which do not) are initialized and updated. Furthermore, true
tracks may become false, either because the target disappears,
or the track may lose a target due to random detections and
measurement noise phenomena.
A false track discrimination procedure must be used to
recognize and eliminate (a vast majority of) false tracks, and
recognize and conrm (a vast majority of) true tracks.
One approach to track information involves using the specu-
lar phenomena [2]. At certain target to receiver geometries, the
power of received signal increases sharply. Thus, the detection
threshold is kept high until the specular detection occurs and
the track is initialized. Then the detection threshold is lowered,
but only in the vicinity of established track(s). This approach
is innovative, and robust in some environments. However it
1
This work was supported by Defense Acquisition Program Administration
and Agency for Defense Development (Republic of Korea) through Project
ADD-09-70-01-03, as collaboration with the University of Melbourne, Aus-
tralia under the contract UD090002DD
usually results in signicant delays before track is initializied
and, subsequently, conrmed. Additional measures have to
be taken to detect the event of track becoming a false track
(losing its target).
Here we propose using the Gaussian Mixture Measurement
likelihood - Integrated Track Splitting algorithm (GMM-ITS)
algorithm to handle both the false track discrimination and
target trajectory estimation, in a natural (integrated) fashion.
The GMM-ITS was introduced in [3] as a general algo-
rithm for target tracking in clutter, when (severely) non-linear
measurements are available. It approximates the trajectory
state pdf by a dynamic Gaussian mixture (in the state space).
The measurement nonlinearity is handled by projecting the
measurement likelihood from the measurement space into the
surveillance space, where it is then approximated by a Gaus-
sian Mixture (summing up to a proportionality coefcient).
Trajectory state update is then handled by using the standard
Kalman lter update operation between every pair of trajectory
state and measurement likelihood Gaussians. To facilitate the
false track discrimination, the probability of target existence
is recursively calculated in the standard manner [4], [5] using
the measurement likelihood ratio.
The track conrmation and termination operations use the
calculated probability of target existence as the track quality
measure.
This approach offers signicant benets. A standard lter
is being used; the only module that is changed between appli-
cation is the measurement likelihood projection and approxi-
mation [6][9]. The atomic operation is the standard Kalman
Filter update, which has been studied in every conceivable
detail during the last decades. On the negative side, the number
of operations (and computational requirements) is signicantly
higher than that of a single Kalman lter, but nevertheless it
is usually an order of magnitude more frugal than a Particle
Filter implementation [10].
This paper is a natural extension of [11], which limits
use of GMM-ITS to a trajectory estimation in a multistatic
environment (no data association environment and associated
false track discrimination requirements).
Models and assumptions are presented in Section II, fol-
lowed by the description of the measurement likelihood trans-
formation in Section III. The GMM-ITS tracker is briey
presented in Section IV. Simulation results in Section V
vindicate this approach, followed by the concluding remarks
in Section VI.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 133
II. MODELS AND ASSUMPTIONS
In this paper we consider two dimensional multistatic surveil-
lance scenarios. A single emitter broadcasts signal, which is
received by multiple sensors. The sensors process the signal,
perform the detection and provide detection measurements.
All measurements are processed at a fusion center. Here we
assume that each sensor receives (takes) measurements simul-
taneously, although the algorithm provides a straightforward
extension for asynchronous scanning.
A single target may or may not be present (although the
lter may be extended to the multitarget version).
The sensors also report clutter measurements, which are
assumed to follow Poisson distribution with known mea-
surement density. Thus, at each scan the tracking algorithm
receives multiple measurements from a number of sensors,
without prior knowledge on which measurement is the target
measurement.
The usual simplifying assumptions of point targets and
the innite sensor resolution are used here. The point target
(if it exists and follows a trajectory whose only dynamic
model is known) creates up to one detection per measurement
time per receiver, with probability of detection P
D
1.
The innite resolution sensor assumption translates into each
measurement has only one source; i.e. it is either the target
measurement or a clutter measurement, but not both.
The target existence event
k
at time k is modeled as a
Markov Chain [4], [5], where the propagation equals
P{
k
} = P{
k1
}, (1)
where denotes the Markov transition probability. Eq. (1) is
conditioned on the data set used so far.
Standard target trajectory propagation is assumed here
x
k
= F x
k1
+
k
(2)
where F is (known) propagation matrix, and
k
is zero mean
white Gaussian sequence not correlated with any other random
sequence, with (known) correlation matrix Q.
Target trajectory is usually expressed in Cartesian coordi-
nates, almost always with position and velocity as elements
of the trajectory state. Here we assume that exists linear pro-
jection H from the trajectory state space into the surveillance
(position) space, i.e. that Hx
k
equals the position of the target.
Each sensor receives both the direct signal from emitter, and
the signal reected by the target. The time difference between
these signals is a measurement of the difference in two lengths
the signal had to travel. Each sensor corrupts measurement
with additive measurement noise, and the target measurement
from sensor (s) at time k equals
y
(s)
k
= h
_
x
k
; x
(s)
_
+
(s)
k
(3)
with x
k
denoting the trajectory state, x
(s)
denoting sensor
trajectory state, and
(s)
k
denoting a sample of zero mean,
white Gaussian process with standard deviation
(s)
. The
measurement function is
h
_
x
k
; x
(s)
_
= a
(s)
k
b
(s)
(4)
where a
(s)
k
denotes the total distance from the emitter to
target to sensor (s), and b
(s)
denotes the direct distance from
sensor (s) to the emitter. Both the (stationary) emitter and the
(stationary) sensor positions are assumed known.
Disregarding the measurement noise and given the mea-
surement value of h(x
k
; x
(s)
), and therefore the value a
(s)
k
,
possible target positions form an ellipse where the emitter and
sensor (s) form the focal points, and with the major axis of
a
(s)
k
, and the minor axis of
_
_
a
(s)
k
_
2
_
b
(s)
_
2
. (5)
Due to the measurement noise, the line of this target position
uncertainty ellipse becomes wide. This measurement equation
is non-linear.
In addition to the signal reected by the target (and de-
tected with the probability of detection P
D
, each sensor
also receives clutter (plus amplication noise) signal, which
results in spurious detections termed the clutter measurements.
The clutter measurements are assumed to follow a Poisson
distribution parameterized by the clutter measurement density
in the measurement space. The clutter measurement density
is assumed a priori known, it can be estimated otherwise [12].
Thus, at measurement time k, receiver i receives a set
of measurements Y
k
(i) with random cardinality m
i
k
; where
Y
k,j
(i) denotes the jth measurement received by receiver i
at time k. The fusion center receives measurement sets from
all receivers, Y
k
= {Y
k
(1), . . . Y
k
(R)} where R denotes
the number of receivers. Denote by Y
k
the sequence of all
measurement sets up to and including time k.
III. MEASUREMENT LIKELIHOOD TRANSFORMATION
The likelihoods of measurement Y
k,j
(i) is approximated by a
Gaussian Mixture in the surveillance space:
p(Y
k,j
(i)|x
k
) c
j
k,i
G
g=1
j,g
k,i
N
_
z
j,g
k,i
; Hx
k
, R
j,g
k,i
_
(6)
where N(x; m, P) denotes the Gaussian pdf of variable x
with mean m and covariance P. The value of constant c
j
k,i
is
important for data association and equals
c
j
k,i
=
A
(s)
2
(s)
, (7)
where A
g=1
j,g
k,i
= 1,
j,g
k,i
> 0 (8)
constraint.
To obtain the Gaussian Mixture presentation of measure-
ment likelihood in the surveillance space, the area between
(s)
ellipses is divided into component ellipses, where each
component ellipse is a footprint of one measurement Gaussian
134 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
0 2000 4000 6000 8000 10000
5000
0
5000
receiver 2
receiver 3
receiver 1
target
emitter
Figure 1. Simulation scenario.
Mixture component, as presented on Figure 1. The process of
obtaining (6) are presented in [11].
The likelihood of each measurement received by each
receiver is converted into a (separate) Gaussian Mixture.
IV. GMM-ITS
Track state at time k is hybrid, as it consists of a binary random
variable (target existence) denoted by
k
and a continuous
random variable (trajectory state) x
k
:
P{
k
, x
k
} = P{
k
}p(x
k
|
k
), (9)
where trajectory state x
k
is only dened given the target exis-
tence event
k
. Trajectory state pdf is initialized and updated
by nonlinear measurement(s). Therefore, the trajectory state
pdf is non-Gaussian, and is approximated here by a Gaussian
Mixture; for example prior pdf at time k approximation is
p
_
x
k
|
k
, Y
k1
_
= (10)
Ck
c=1
c
N
_
x
k
; x
(c)
k|k1
, P
(c)
k|k1
_
where c denotes track component, or index of the track
component, depending on the context, and
Ck
c=1
c
= 1,
c
> 0. (11)
The advantage of using Gaussian Mixtures for both mea-
surement likelihood presentation and for the track state pdf
presentation, is that we may replace non-linear operations by
a number of simple Kalman lter updates.
The GMM-ITS recursion cycle at time k starts with
the posterior state at time k 1, P{
k1
|Y
k1
} and
p(x
k1
|
k1
, Y
k1
) and consists of
track prediction,
measurement component selection and likelihood calcu-
lation,
track update,
track component management, and
track output.
The measurement component selection and likelihood calcu-
lation, the track update and track component management
are performed sequentially on each measurement set Y
k
(i),
where the updated track state using the measurement sets
Y
k
(1) . . . Y
k
(i1) serves as the prior track state with respect
to the measurement set Y
k
(i) update. Thus, for the rest of this
section, due care has to be exercised when using P{
k
|Y
k1
},
C
k
,
c
, x
(c)
k|k1
and P
(c)
k|k1
, as they are the propagated values
only when using Y
k
(1), otherwise they are the track state after
Y
k
(1) . . . Y
k
(i 1) have been applied.
The probability of target existence serves as the track quality
measure and is used for false track discrimination. When (and
if) the probability of target existence rises above a predeter-
mined conrmation threshold, the track is assumed true, is
conrmed, and the information is forwarded to the operators
or the higher level of information fusion. When the probability
of target existence falls below a predetermined termination
threshold, the track is declared false and terminated.
A. Track Prediction
Track prediction propagates track state pdf from time k 1 to
time k,
p
_
x
k
|Y
k1
_
=
Ck
c=1
c
N
_
x
k
; x
(c)
k|k1
, P
(c)
k|k1
_
(12)
Relative probability
c
of each component c does not change
when propagating and each track component propagates indi-
vidually as per standard Kalman Filter prediction,and
_
x
(c)
k|k1
, P
(c)
k|k1
_
= (13)
KF
P
_
x
(c)
k1|k1
, P
(c)
k1|k1
, F, Q
_
.
The probability of target existence propagates by
P{
k
|Y
k1
} = P{
k1
|Y
k1
} (14)
B. Selection and Likelihood Calculation
Each track component c at time k selects a subset of mea-
surement components for update. The selection procedure will
not be detailed here, sufce it to say that it is described in
many references [13][15], and will not be detailed here. Each
component selects measurement components with signicant
mutual likelihoods.
The likelihood of a selected measurement Y
k,j
(i) compo-
nent g, with respect to the track component c is
p
j,g,c
k,i
=
c
j
k,i
P
G
N
_
z
j,g
k,i
; H x
(c)
k|k1
, S
j,g,c
k,i
_
(15)
where P
G
is the gating probability and
S
j,g,c
k,i
= HP
(c)
k|k1
H
T
+R
j,g
k,i
. (16)
The likelihood of a measurement Y
k,j
(i) component g which
is not selected by the track component c, and with respect
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 135
to the track component c is p
j,g,c
k,i
= 0. The likelihood of
measurement Y
k,j
(i) with respect to track component c is
p
j,c
k,i
=
G
g=1
j,g
k,i
p
j,g,c
k,i
, (17)
and the likelihood of measurement Y
k,j
(i) with respect to track
is
p
j
k,i
=
Ck
c=1
c
p
j,c
k,i
. (18)
C. Track Update
Track measurement likelihood ratio of measurement set Y
k
(i)
is
k,i
= 1 P
D
P
G
+ P
D
P
G
m
i
k
j=1
p
j
k,i
. (19)
Updated probability of target existence equals
P{
k
|Y
k
} =
k,i
P{
k
|Y
k1
}
1 (1
k,i
)P{
k
|Y
k1
}
. (20)
During the track update, each existing track component
is replaced by a number of new track components. Each
pair of track component c and a measurement component g
(of measurement Y
k,j
(i) forms a new component, given that
track component c selects measurement component g. Denote
by c
+
new component formed by track component c and
measurement component g, then the updated track state pdf is
approximated by
p
_
x
k
|Y
k
_
=
Ck+1
c
+
=1
c
+N
_
x
k
; x
(c
+
)
k|k
, P
(c
+
)
k|k
_
(21)
where relative component probability
c
+ is
c
+ = P
D
P
G
k,i
p
j,g,c
k,i
, (22)
or, when the component is associated with the null measure-
ment
c
+ =
c
1 P
D
P
G
k,i
. (23)
which satisfy constraint
Ck+1
c
+
c
+ = 1. (24)
Mean x
(c
+
)
k|k
and covariance P
(c
+
)
k|k
of new component c
+
are
obtained by simple Kalman lter update of track component
c by measurement component g
_
x
(c
+
)
k|k
, P
(c
+
)
k|k
_
= (25)
KF
U
_
x
(c)
k|k1
, P
(c)
k|k1
, z
j,g
k,i
, R
j,g
k,i
, H
_
,
where KF
U
denotes the Kalman lter update operation. For
the association with the null measurement,
_
x
(c
+
)
k|k
, P
(c
+
)
k|k
_
=
_
x
(c)
k|k1
, P
(c)
k|k1
_
. (26)
D. Track Component Management
As presented in this section, the number of track compo-
nents grows exponentially in time. To prevent the saturation
of available computational resources, some track component
management has to be implemented [16]. Proposed algorithm
is an instance of track splitting algorithm [3], [5], [17], [18],
and all track splitting track component management methods
can also be used here.
These techniques include track component pruning, where
components with small relative probabilities are removed,
track component subtree pruning, where whole subtrees of
components are removed based on track component relative
probabilities [13]. Finally, one can merge close track com-
ponents, where denition of close differs between various
proposals [19][21].
E. Track Output
As the track trajectory estimate, the authors use the trajectory
state mean x
k|k
and covariance P
k|k
, dened as the mean and
covariance of a posteriori track state pdf at time k
x
k|k
=
c
+
c
+ x
(c
+
)
k|k
(27)
P
k|k
=
c
+
c
+
_
P
(c
+
)
k|k
+ x
(c
+
)
k|k
_
x
(c
+
)
k|k
_
T
_
(28)
x
k|k
x
T
k|k
V. SIMULATIONS
Simulations have been used to verify proposed algorithm. A
simple scenario is shown in Figure 1. Three sensors generate
multistatic range only target measurements with the probabil-
ity of detection P
D
= 0.9. At each scan each sensor generates
a random number of clutter measurements, on the average each
sensor generates 3 clutter measurements per scan.
Multistatic measurements are taken once per 10 seconds,
and total simulated time is 1000 s. Each simulation scenario
is repeated 500 times with independent measurements.
0 200 400 600 800 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
time
c
o
n
f
i
r
m
e
d
t
r
u
e
Figure 2. True track conrmation rate
136 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Tracks are initialized using all measurements, and the
tracks are propagated as proposed in this paper. False track
discrimination has been applied using the probability of target
existence as the track quality measure. No false tracks were
conrmed, and the true track success rate is depicted on
Figure 2.
Tracking results depend on the geometry of the situation
which is the current research area. Another area is the adaptive
choice of sensors to be used, the resource allocation problem.
VI. CONCLUSIONS
This paper illustrates use of the GMM-ITS algorithm when
tracking a single target in clutter, using the highly-nonlinear
multistatic measurements without bearings information.
The initial results are highly encouraging, and the research
continues in various practical aspects of the application.
REFERENCES
[1] S. Coraluppi, Multistatic sonar localization, IEEE Journ. Oceanic
Engineering, vol. 31, no. 4, October 2006.
[2] D. Grimmett, Multistatic target tracking using specular cue initiation
and directed data re-trieval, in 11th International Conference on Infor-
mation Fusion, Fusion 2008, Cologne, Germany, July 2008.
[3] D. Mu sicki and R. Evans, Measurement Gaussian sum mixture tar-
get tracking, in 9th International Conference on Information Fusion,
Fusion 2006, Florence, Italy, July 2006.
[4] D. Mu sicki, R. Evans, and S. Stankovi c, Integrated Probabilistic Data
Association (IPDA), IEEE Trans. Automatic Control, vol. 39, no. 6, pp.
12371241, Jun 1994.
[5] D. Mu sicki, B. La Scala, and R. Evans, The Integrated Track Splitting
lter - efcient multi-scan single target tracking in clutter, IEEE Trans.
Aerospace Electronic Systems, vol. 43, no. 4, pp. 14091425, October
2007.
[6] D. Mu sicki, Bearings only multi-sensor maneuvering target tracking,
Systems Control Letters, vol. 57, no. 3, pp. 216221, March 2008.
[7] , Multi-target tracking using multiple passive bearings-only asyn-
chronous sensors, IEEE Trans. Aerospace and Electronic Systems,
vol. 44, no. 3, pp. 11511160, July 2008.
[8] , Bearings only single-sensor target tracking using Gaussian mix-
tures, Automatica, vol. 45, no. 9, pp. 20882092, September 2009.
[9] D. Mu sicki, R. Kaune, and W. Koch, Mobile emitter geolocation
using TDOA and FDOA measurements, IEEE Trans. Signal Processing,
vol. 58, no. 3, pp. 18631874, Mar 2010.
[10] B. Ristic, S. Arulampalam, and N. Gordon, Beyond the Kalman Filter.
Artech House, 2004.
[11] T. L. Song and D. Mu sicki, Range only multistatic tracking, in
13th International Conference on Information Fusion, Fusion 2010,
Edinburgh, UK, July 26-29 2010.
[12] , Adaptive clutter measurement density estimation for improved
target tracking, IEEE Trans. Aerospace Electronic Systems, vol. 47,
no. 2, pp. 14571466, April 2011.
[13] S. Blackman and R. Popoli, Design and Analysis of Modern Tracking
Systems. Artech House, 1999.
[14] Y. Bar-Shalom and E. Tse, Tracking in a cluttered environment with
Probabilistic Data Association, Automatica, vol. 11, pp. 451460, Sep
1975.
[15] Y. Bar-Shalom, K. Chang, and H. Blom, Multitarget Multisensor Track-
ing. Artech House, 1990, vol. 1, ch. Automatic track formation in
clutter with a recursive algorithm, pp. 2542.
[16] S. Blackman, Multiple-target tracking with radar applications. Artech
House, 1986.
[17] T. Kurien, Multitarget Multisensor Tracking, Y. Bar-Shalom, Ed. Artech
House, 1990, vol. 1.
[18] D. B. Reid, An algorithm for tracking multiple targets, IEEE Trans.
Automatic Control, vol. 24, no. 6, pp. 843854, Jun 1979.
[19] R. A. Singer, R. G. Sea, and K. Housewright, Derivation and evaluation
of improved tracking lters for use in dense multi-target environments,
IEEE Trans. Information Theory, vol. 20, no. 4, pp. 423432, Jul 1974.
[20] D. J. Salmond, Mixture reduction algorithms for target tracking in
clutter, in SPIE: Signal and Data Processing of Small Targets, vol.
1305, Orlando, Florida, April 1990, pp. 434445.
[21] J. L. Williams and P. S. Mayback, Cost-function-based gaussian mix-
ture reduction for target tracking, in 6th International Conference on
Information Fusion, Fusion 2003, Cairns, Queensland, Australia, July
2003, pp. 10471054.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 137
TrackerQualityMonitoringbyNonDedicatedCalibrationFlights
Session6.1page141
TheTransponderDataRecorder:firstimplementationandapplications
Session6.2page147
ADSB/MLATsurveillancesystemfromHighAltitudePlatformSystems
Session6.3page153
SpacebasedADSBAsmallstepfortechnologyagiantleapforATM?
Session6.4page159
where
and
and
,
respectively. Trajectories can be modeled as mappings
that is
spanned by position , velocity and acceleration .
Here we neglect the acceleration and consider the
velocity to be equal to the first time derivative of the position
. So we can reduce a trajectory to a mapping from time to
position:
and we can set
So we obtain for
and
the
distances across and along two trajectories
and
the
following expressions:
and
and
and
m
, the maximum distance D
M
and imposing
m
equal to
62 is possible nd all the other parameters (considering also
the Earth radius R
E
= 6378 km and Satellite altitude of H =
780 km). In particular, dening L = R
E
+ H, the following
equations can be written:
L = D
M
cos(
m
) +R
E
cos(
m
) (1)
D
2
M
= L
2
+R
2
E
2R
E
Lcos(
m
) (2)
and is possible to compute
m
and
m
:
m
= arcos
L D
M
cos(
m
)
R
E
(3)
Figure 2. System geometry for a Satellite ADS-B receiver.
m
=
2
m
m
(4)
The coverage area S and the coverage radius for each satellite
can be calculated by the following equations:
S = 2R
2
E
(1 cos(
m
)) (5)
D = 2R
E
m
(6)
All the resulting geometry information are summarized in
Table I .
H[km] D
M
[km] m m S [km
2
] D [km]
780 2503.31 7.72 20.27 15838703.965 4514.2283
Table I
COVERAGE PARAMETERS FOR AN IRIDIUM SATELLITE WITH AN ANTENNA
BEAMWIDTH OF 124.
For this coverage it is necessary to verify if the transmitted
signal from the airplanes arrive at the satellite receiver with
enough power to be decoded (i.e. greater then the receiver
sensibility) and also optimize the radiation pattern of the
antenna. It is possible to calculate the satellite antenna pattern
from the Friis equation[6]:
P
rx
= P
tx
+G
tx
L
tx
L
at
A
fs
(
m
)+G
rx
(
m
)L
rx
[dB]
(7)
where P
rx
is the receiver sensitivity, P
tx
is the transmitted
power of the transponder (Class A3 transponder, i.e. 21 dBW),
L
tx
e L
rx
, are the transmitting and receiving losses, L
at
is the
propagation loss and A
fs
is the free-space attenuation.
Ptx Prx Ltx Gtx Lat Lrx
21 dBW -120 dBW 3 dB 0 dBi 3 dB 3 dB
Table II
PARAMETERS FOR LINK-BUDGET COMPUTATION.
154 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Given the parameters in Table II, considering a target at
the maximum distance D
M
(
m
) is possible to compute the
free-space attenuation:
A
fs
= 10log
4D
M
(
m
)
2
[dB] (8)
and the the Antenna gain directly from the equation 7.
In Figure 3 a vertical (azimuth) cut of the pattern is
represented.
Figure 3. Ideal vertical section of the radiation pattern of a full coverage
Iridium ADS-B antenna.
This kind of radiation pattern cannot be generated with a
small and simple antenna. This means that the ADS-B payload
must be more complex (multichannel receiver and multibeam
antenna), at least as complex as in the primary application of
the satellite (for example, the antenna should be very near,
in dimensions, to the Iridium antenna, i.e. composed by 3
panels of about 180 cm X 90 cm). It may be concluded that
global coverage by ADS-B on satellites may bring to too
large and too complicated systems, with an unfavorable cost
effectiveness.
III. HAP PAYLOAD DESCRIPTION
Considering the architectures of the system given in Figure
1 is possible to dene the HAPs payload for the 1090ES
channel as described in Figure 4. The payload receives the
Mode S signals and ADS-B report, decodes it and sends the
information decoded to the CPS. The payload must add infor-
mation to every Mode S reply received about Time Of Arrival
(TOA) of the message and the platform precise position. So,
the platform has a GPS/GNSS receiver. It is important to verify
if this payload respects the requirements to y on an HAP (in
term of weight, volume and power consumption). Therefore a
brief analysis of the needed hardware with COTS components
was done; the results are reported in Table III.
Considering that a typical Unmanned Aircraft System
(UAS) or a typical airship for stratospheric y can carry a
payload of 50 kg and can supply a power of kW, the ADS-
B/Multilateration receiver, with its weight of less than 5 Kg
Component Weigh Size Power
[kg] [cm
3
] [W]
Antenna 0.1086 8.20 10 3 = 246 0
1090 MHz
Mode S 0.7 3.8 17.4 12.4 = 841.46 2.8
Receiver
Process and 0.13 9 9.6 2 = 172.8 6
Control unit
GPS antenna 0.111 8.73 5.59 3.2 = 156.16 1
GPS Receiver 0.54 9.5 4.2 16.8 = 670.32 1.7
Estimate Weight, 1.59 2087 11.5
size and power
Table III
POWER CONSUMPTION, WEIGHT AND VOLUME OF A COTS ADS-B
RECEIVING SYSTEM.
Figure 4. HAP ADS-B Payload diagram
and a power consummation smaller then 3 W can be a sec-
ondary payload for an HAP mission (i.e. Telecommunication
or Satellite Navigation augmentation [7]).
IV. COVEREGE OF THE HAP
The real coverage of the HAP system must be also eval-
uated. This is essentially limited from: (a) geometry; (b)
capacity of the channel and (c) the power, i.e the link budget.
Concerning the geometry and the coverage, it is possible to
use the formulas for the satellite application, section II, while
considering an altitude of 20 km for the HAP. The results are
reported in Table IV.
H D
M
m m m S D
[km] [km] [] [] [] [km
2
] [km]
20 505.48 0 4.53 85.8203 798038 1008.02
20 195.54 5 1.74 83.3 118494 388.43
Table IV
PARAMETERS FOR THE MAXIMUM GEOMETRICAL COVERAGE
ACHIEVABLE WITH THE USE OF AN HAP (m = 0 AND m = 5.
Considering that the airplane can be at 10 km of altitude
the real coverage of an HAP can be also larger: for an airplane
ying at 10 km of altitude the maximum distance from the
HAP in line of sight is 862.25 km, this means that only one
HAP can cover the European core.
The capacity of the channel depends essentially on the
trafc scenario and on the ability of the receiver to decode
superimposed replies. Here, the trafc scenarios proposed
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 155
in the CASCADE Program from EUROCONTROL [8] is
considered to exploit the problem of system capacity. The
CASCADE program considers the following scenario: a Mode
S receiver in Bruxelles at an altitude of 3300 feet and a
coverage area with a radius of R
max
= 300NM with three
scenarios for the fruit densities:
1) high interference (
max,1
= 105000 fruits per second)
2) reduced Radar infrastructures (
max,2
= 55000 fruits
per second)
3) medium interference (
max,3
= 50000 fruits per second)
4) low interference (
max,4
= 27500 fruits per second).
If an uniformly distributed trafc in the coverage area and an
arrival process of the fruits Poisson distributed are considered
[9], it is possible to write:
p(n) =
(T)
n
n!
e
T
(9)
in which p(n) is the probability of receiving n replies in the
time interval T (length of the Mode S reply), for a xed
number of fruits per second . When n is larger than one
it means that one or more fruits arrive before the end of the
reply, i. e. interference condition. Not all the tranponders in
the coverage area will interfere with a reply but only the reply
that have enough power to produce a Signal to Interference
Ratio (SIR) smaller then a given value. For this reason only the
message coming from airplanes which have a distance from
the receiver between 0 and R+r are considered, where R is
the distance of the interfered airplane and the receiver and r
is such that the received power of the interfering reply is no
less than 3dB below the received power of the interfered reply.
This means that it is assumed that a signal to interference ratio
greater than 3 dB can be managed. In Figure 5 the probability
of receiving a non interfered signal from a transponder at a
given distance R using a standard receiver is reported; we have
assumed the previous hypothesis and that the number of fruits
() is computed by using the effective radius of the interfering
area:
i
=
(R + r)
2
(R
max
)
2
max,i
(10)
It is possible also to consider an enhanced receiver that
has the capability to decode the message also in the case
of interference conditions, as described in [9]. In particular,
calling t
p
the Mode S preamble duration and t
ES
the Mode
S reply duration the following events can be considered:
A:(0 interfering signals in [0 t
ES
]);
B:(0 interfering signals in [t
p
t
ES
]);
C:(1 interfering signals in [t
p
t
ES
]);
F:(0 interfering signals in [0 t
p
]);
and for this enhanced receiver we can say that we do not have
interference if:
P
free
= P(B C|F) =
P((B C)F)
P(F)
=
P(BF) +P(CF)
P(F)
=
P(A) +P(C)P(F)
P(F)
=
P(A)
P(F)
+P(C)
(11)
Table V
COVERAGE LIMIT DUE TO THE DIFFERENT RECEIVER CAPABILITY
(STANDARD (*), ENHANCED RECEIVER FOR ONE (), TWO (), THREE (),
INTERFERING SIGNALS).
this means that if only the preamble is free of interference
and the data-block has zero or one superimposed signals, the
message can be considered free of interference because the
interference can be managed and solved [9]. In the same
manner also the interference with 2 or 3 replies can be
modeled. Table V shows the resulting coverage area (due to
the capacity of the channel) for different scenarios.
Figure 5. Probability of receiving one long squitter without interference.
In the Table V it is possible to see that the capacity of
the channel (i.e. the number of aircraft in the coverage of the
sensor) is the most important factor that limits the systems
performance, only using enhanced receiver is possible to
156 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
achieve a coverage area range of 136 NM for a single reply
(P
D
= 0.9) or 232 NM for 9 replies (at least 1 reply every 10
seconds).
The last system parameter to be investigated is the link
budget. Staring from the Friis equation, it is possible to change
the coverage area choosing the appropriate antenna for the
receiving station.
In gure 6 the ideal vertical section of the radiation pattern
of an HAP ADS-B antenna (from 20 Km of altitude) is shown,
it is computed with the same formulas used for computing the
satellite ideal radiation pattern.
Different types of antenna have been analyzed from a 20
Km high HAPS, in particular:
Dipole antenna;
Vertical array of 4 dipoles with about 8 dBi of maximum
gain that allows 60 NM of coverage;
Vertical array of 6 dipoles with about 9 dBi of maximum
gain that allows about 120 NM of coverage.
Figure 6. Ideal vertical section of the radiation pattern of a ideal HAP ADS-B
antenna.
In table VI the coverage due to the antenna conguration
are reported and the considered capacity scenario are also
compared with the antenna coverage. It is possible to see that
existing congurations for HAP application in high trafc or
low trafc condition match also with the proposed antennas.
V. INNSBRUCK WAM - SIMULATION TRIALS
To understand the capability of an HAP ADS-B receiver
some simulations were done. In particular the real Innsbruck
WAM system was supposed to be enhanced with an HAP
platform. The actual Innsbruck WAM system consists of 9
stations (Figure 7): 3 transmitting and receiving station, 5
receiving -only, and one reference transponder (Patscherkofel
station) [10]. Considering the position of all the sensors (Table
VII) we add an HAP (204 km of altitude) at coordinates:
(4716N, 1123E).
It is possible to understand the performance of these systems
(with or without the HAP receiver) using the Cramer Rao
Table VI
PROPOSED ANTENNA FOR THE CONSIDERED TRAFFIC SCENARIOS.
A:{DIPOLE, AVIONIC ANTENNA}, B:{4 OR 6 ARRAY OF DIPOLES)}.
Position Latitude Longitude Altitude [m]
Patscherkofel 4712
31.4
1127
36.7
2245
Flughafen 4715
28.2
1121
9.8
616
Hafelekar 4718
46.4
1123
10.3
2336
Rangger K op 4714
37.4
1110
51.9
1910
Tulferer Berg 4715
11.6
1133
20.9
1360
Hecher 4719
15.7
1144
30.8
1895
Kanzelkehre 4724
34.6
1147
15.1
1006
Telfs 4718
11.9
1104
23.1
605
Gschwandtkopf 4718
49.3
1110
39.5
1459
Table VII
WAM INNSBRUCK- STATIONS COORDINATES [10].
Figure 7. Innsbruck WAM real layout and taking off path.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 157
Figure 8. Comparison of CRLB for a taking off airplane in the Innsbruck
scenario with or without one HAP over the airport.
Lower Bound (CRLB) method described in [11], considering
a TOA accuracy of:
TDOA,i
() = max
1
B
2SNR
, 1
[meters] (12)
Where B is the receiver bandwidth supposed equal to 20 MHz
and SNR the signal-to-noise ratio.
The CRLB was computed along the taking off path reported
in Figure 7 and the lower bound for Vertical (
2
z
) and 3D
error (
2
x
+
2
y
+
2
z
) are reported in Figure 8.
Figure 8 clearly shows the benets introduced by using
the HAP over the airport. The height of the HAP allows
improvement of the Dilution Of Precision (DOP) and therefore
of the system accuracy (e.g. 3D position error below 5 meters
until 40 Km of coverage). This kind improvement cannot be
obtained with any type of ground stations.
VI. CONCLUSION
The paper shows that it is possible to develop a low cost
ADS-B Mode S receiver to be installed in a HAP with a very
simple antenna and receiver. The problem of the coverage of
this kind of station was discussed and different solution have
been proposed for high density and medium density trafc. A
coverage radius greater than 140 NM (limited by the capacity
of the channel) can be achieved with all the advantages due to
the fact that we have a ying receiver station. The advantages
of having a ying station is very clear when the station is a part
of Multilateration system (Local or Wide area), in this case
the performance increases a lot with respect to the classical
deployment due to the high decrease in the value of the vertical
DOP, as shown with simulation for the Innsbruck WAM
system. The aim of the paper was to study the possibility
to develop a small, not heavy and cheap system to be used
as a piggy-back system over the primary application of the
HAP (telecommunication or satellite navigation applications)
but the results call for a study for the development of a more
complex system with a multibeam antenna that can manage
more than one independent coverage area (this is possible
because the maximum geometrical coverage is about 500 NM)
with an ad hoc designed platform with a big array antenna and
a multichannel receiver.
REFERENCES
[1] M. Leonardi, G. Galati, P. Magaro, and V. Paciucci, Wide area surveil-
lance using ssr mode s multilateration: advantages and limitations, in
European Radar Conference, 2005, Parigi, 6-7 october 2005, p. 225.
[2] T. Delovski, L.-C. Hauer, and J. Behrens, Ads-b high altitude measure-
ments in non radar airspace. ESAVS 2010 proceedings, 16-17 march,
2010, berlin, pp. 1-5.
[3] A. Arag` on-Zavala, J. L. Cuevas-Ru`z, and J. A. Delgado-Pen`n, High-
Altitude Platforms for Wireless Communications. John Wiley and Sons,
2008, pp. 1, 16-33, 155-157.
[4] R. Miura and M. Suzuki, Preliminary Flight Test Program on Telecom
and Broadcasting Using High Altitude Platform Stations. Wireless
Personal Communication 24, 2003, pp.341-361.
[5] D. Grace, K. Katzis, D. Pearce, and P. Mitchell, Low-latency mac-layer
handoff for a high-altitude platform delivering broadband communica-
tions. The Radio Science Bulletin No 332, March 2010.
[6] H.T.Fris, A note on a simbol trasmission formula, May 1946, vol.34,
pp.254-256.
[7] F. Dovis, L. L. Presti, and P. Mulassano, Support infrastructures based
on high altitude platforms for navigation satellite systems, Wireless
Communications, IEEE, vol. 12, no. 5, p. 106, october 2005.
[8] 1090 MHz Capacity Study-Final Report, Cascade program Std., July
2006, edition 2.6, pp.13-25, 32-34, 46-54.
[9] E.G.Piracci, N.Petrochilos, and G.Galati, 1090 es receiving capacity
improvement using ads-b ground receivers with signals discrimination
capability. ESAV08 Proceedings, 3-5 September 2008, pp. 17.
[10] Wieser, Wolfmayr, Langhans, Cernin, and Scheiinger, Wide area
multilateration at terminal area innsbruck, in Wide Area Multilateration
Workshop, Eurocontrol, Bruxelles, June 2007.
[11] G. Galati, M. Leonardi, and M. Tosti, Multilateration (local and
wide area) as a distributed sensor system: Lower bounds of accuracy.
Amsterdam: EuRad 2008 Conference, 30-31 October 2008.
158 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Space-based ADS-B
A small step for technology a giant leap for ATM?
Adam Parkinson
Senior Consultant
Helios
Farnborough, UK
adam.parkinson@askhelios.com
Abstract - This paper investigates the feasibility and merits of a
space-based ADS-B system. The primary concept that will be
investigated is the reception of ADS-B transmissions from
aircraft in oceanic airspace using satellites and the subsequent
relay of the data to end-users on the ground.
Keywords: ADS-B, space systems, new concepts, applications,
oceanic, requirements)
I. THE CHALLENGES FACING AIR TRAFFIC MANAGEMENT
There are several significant challenges facing Air Traffic
Management (ATM) today. These can be summarized as the
need for increased capacity to cope with traffic growth whilst
at the same time reducing delays. Across Europe these
challenges are being strategically addressed through two
initiatives. The Single European Sky (SES) is concerned with
providing the necessary legislative framework while the SES
ATM Research (SESAR) Programme is concerned with the
modernization and interoperability of ATM infrastructure
across Europe. In particular SESAR has the following
performance targets:
Accommodate a 3 times increase in movements whilst
reducing delay.
Improve safety by a factor of 10.
Enable a 10% reduction in environmental effects per
flight.
Reduce ATM unit cost to airspace users by at least
50%.
II. THE ROLE OF ADS-B
For SESAR and NextGen (the US equivalent), Automatic
Dependent Surveillance Broadcast (ADS-B) is one of the most
important underlying technologies in the plan to transform
ATM from the current radar-based surveillance to Global
Navigation Satellite Systems (GNSS) surveillance. ADS-B is
defined by the International Civil Aviation Organization
(ICAO) as a surveillance application transmitting parameters,
such as position, track and ground speed, via a broadcast mode
data link, and at specified intervals, for utilization by any air
and/or ground users requiring it. The ADS-B reports are sent
periodically by the aircraft with no intervention from the
ground function and ADS-B reports may be received by any
suitable receiving equipment in range of the transmitting
aircraft. The data transmitted is derived from the aircraft
systems themselves and in this sense ADS-B is known as a
dependent surveillance technology. The transmitting aircraft
does not know which, if any, recipients are receiving and
processing the position reports as they are not acknowledged.
The concept with ADS-B is that position reports are
transmitted so frequently that the loss of a small number of
position reports is not operationally significant.
ADS-B does not require a specific data link however
throughout the rest of this paper when referring to ADS-B we
mean the transmission of Extended Squitter messages
transmitted over the 1090MHz channel 1090ES. The 1090MHz
channel is the downlink channel for Secondary Surveillance
Radar (SSR) replies from aircraft and the Extended Squitter
messages are compatible with the Mode S radar data link
formats. Amongst other data items Extended Squitter messages
contain aircraft position information derived from GNSS.
The potential benefits of ADS-B are [1]:
Faster data update rate than that typically available
with radar.
Lower cost ground infrastructure compared to radar.
Surveillance data can be directly received in the
aircraft cockpit increasing situational awareness.
Position accuracy is potentially higher than radar and is
not range dependent.
It can display both airborne and ground traffic.
It can potentially enable new applications and
operational procedures resulting in more efficient flight
profiles and reduced emissions e.g. allows 5 NM of
separation in Non-Radar Airspace (NRA) compared to
current procedural separation.
However, in order to realize the full benefits from ADS-B
aircraft must be equipped and plans for mandates in both
Europe and the US are illustrating the difficulty in achieving
universal equipage and the need to ensure that once fitted the
avionics are fully utilized. The mandates are currently being
put in place in the US and Europe for 1090ES ADS-B (dates
2015-2020) and 1090ES ADS-B is already widely implemented
on commercial traffic. In addition ADS-B is being deployed
world-wide in places like Australia, Canada, Thailand,
Jamaica, United Arab Emirates and South Korea.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 159
Other potential issues related to ADS-B include [2]:
Incorrectly coded 24-bit (technical) aircraft address
used to uniquely identify aircraft.
Dependent on GNSS for surveillance and navigation,
and GNSS position potentially leads to oscillating
position quality.
May need addition surveillance cover provide by
independent surveillance technology to provide a
separation service in current radar airspace.
III. SURVEILLANCE OF REMOTER REGIONS
A. The oceancic problem
Traditionally surveillance has been performed by
monitoring on the ground signals received from aircraft,
requiring ground infrastructure to do so. However, in oceanic
or remote regions it may not be practically possible to install
ground infrastructure even with ADS-B.
For long periods of time in oceanic airspace, aircraft are
unable to communicate directly with controllers and are
therefore issued with strategic clearances between exit and
entry points to what is know as procedural airspace. Because of
the lack of communication and surveillance information,
aircraft in oceanic airspace are required to maintain large
separation distances/times, typically 10 minutes longitudinally
and 60NM laterally, in order to maintain safety standards [3].
This procedural separation has the undesired effect of limiting
airspace capacity and also the flexibility of aircraft to fly
efficient routes.
The question therefore arises how can we make use of the
ADS-B surveillance data already being transmitted by aircraft
to improve oceanic operations?
B. Current oceanic operations and technical solutions
Aircraft with different speeds on the same track in oceanic
airspace will gradually get closer or further apart. It is
imperative to monitor this change of spacing closely for loss of
separation. Pilots are therefore required to report their position
verbally at regular intervals along the route, for example at
each waypoint or every 45 minutes which ever is shorter [3].
Typically this is using High Frequency (HF) or satellite
communications.
However, the advent of data link communications has
already removed the need for voice reporting and enabled
higher rates of position reporting in oceanic regions for suitably
equipped aircraft. This type of position reporting is known as
Automatic Dependent Surveillance Contract (ADS-C). It is
based on setting up a point-to-point communications contract
between the aircraft and the ground where position reports are
acknowledged and are either made at an agreed regular rate, are
event driven or are made on demand.
The only operational implementation of ADS-C in oceanic
airspace is the Future Air Navigation System (FANS) 1/A
equipment on Boeing and Airbus aircraft. Nearly all long haul
aircraft are now equipped with FANS 1/A. FANS 1/A makes
use of the Aircraft Communication Addressing and Reporting
System (ACARS) and sub-networks used by many aircraft to
communicate aircraft information to the Airline Operations
Centre (AOC).
The increased monitoring provided by FANS 1/A ADS-C
services potentially enable the following benefits in oceanic
airspace [4]:
reduced separation (typically 30NM longitudinally and
laterally);
more direct routes;
more optimal climb and descend profiles;
increased access to cruise altitudes or closer to optimal;
reduced controller and pilot workload;
increased level of safety.
However, it is unlikely that ADS-C data will ever be used
operationally to provide a radar-like separation service in
oceanic airspace. The application of ADS-C based separations
would require extensive evaluations and agreements with
adjacent Area Control Centers (ACC) [4].
C. Future oceanic operations and technical solutions
A potential future ADS-B enabled application for oceanic
airspace is the In-Trail Procedure (ITP). It was originally
envisaged that this could be provided by ADS-C. However, it
was decided that this was an impractical solution, and that any
airborne surveillance application used in the oceanic airspace
should be feasible without ground surveillance [5]. Therefore,
the application is now either Airborne Traffic Situational
Awareness (ATSA) with similar procedural limits to today or
Airborne Separation (ASEP) with new airborne separation
standards (yet to be defined).
ATSA-ITP has some of the most noticeable benefits for a
relatively small investment and is therefore likely to be one of
the first airborne applications to be implemented. Although
originally a spacing application, ATSA-ITP has now been re-
classified as situational awareness. The ATSA-ITP allows
pilots to identify the relative position of other aircraft, and pass
this information to the controller to clear the aircraft for a
procedural climb. Due to the higher accuracy of surveillance
information available to the controller (and flight crew) via the
ADS-B reports, lower procedural limits can be applied during
the duration of the climb, assuming some geometric
constraints. Since the separation limits are lower, there is
effectively more airspace within which the aircraft can climb,
thus maximizing the fuel efficiency for the given traffic density
[5].
ATSA or ASEP applications may also enable additional
procedures which are not possible with just ADS-C, such as
passing maneuvers, to be implemented in procedural oceanic
airspace.
160 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
IV. SATELLITE ADS-B
A. Why do satelite ADS-B?
One of the aims of SESAR is to develop a global
interoperable ATM system. ADS-B is a significant contributor
to a future ATM system capable of providing high accuracy,
high update rate position reports with a low cost ground
infrastructure. Satellites also play a key role in enabling global
communications services and high accuracy positioning and
navigation through GNSS. Given that aircraft will be equipped
for ADS-B operation over continental regions it is only natural
to investigate the feasibility of receiving the ADS-B position
reports via satellite to provide cost-effective surveillance
coverage in remote regions without current surveillance
infrastructure. The potential benefits are the improved
monitoring of aircraft in remote areas to increase safety and
enabling subsequent changes to current procedures to make
more efficient use of the airspace.
B. Assessment of a potential implementation
A key question to answer is whether satellite ADS-B is
feasible, or under what conditions is it feasible? To help us
answer this question we postulate a possible satellite ADS-B
implementation and assess its potential performance
characteristics.
The analysis consists of an ADS-B receiver installed on a
satellite in a sun synchronous orbit at an altitude of 670km
receiving ADS-B reports at 10 locations in the North Atlantic
oceanic region. The assumptions used in the analysis are
summarized in Table I. A visual representation of the scenario
is presented in Fig. 1 where the red crosses represent the 10
analysis points, the blue triangles are aircraft and the purple
circles are radar locations [6].
C. Assessment results and analysis
The 1090MHz interference received by the satellite ADS-B
receiver at each of the test points was analyzed using a
1090MHz interference model developed by Helios on behalf of
Eurocontrol. From the analysis the following observations were
noted:
The highest interference levels are recorded at test
points that either have the highest traffic densities or
are within range of radar interrogations.
The level of TCAS interference is generally low as the
test point locations are primarily in areas of low or
medium traffic density and the probability of an
aircraft being close enough to generate TCAS
transmissions is low.
The level of Mode A/C SSR interference is low as
most test points are out of range of SSR radars.
TABLE I. ANALYSIS SCENARIO ASSUMPTIONS
Scenario element Description
Aircraft
Predicted 2015 oceanic and core Europe traffic
levels based on actual Eurocontrol CFMU data
and STATFOR growth predocitions
Mode S equipage 100%
Extended Squitter
equipage
100% of Mode S
equipped aircraft
TCAS equipage 80%
Aircraft Mode S
transponder
Short Squitter
trasnmission rte
1Hz
Extended Squitter
transmissions rate
6.2Hz
Transmission power 57dBm
Antenna pattern
Omni-directional
with 0dB gain
Radar
Predicted Mode S and SSR radar installations
in Europe in 2015 based on information
gathered from Eurocotnrol and European
ANSPs
Number of civil SSR 20
Number of military
SSR
340
Number of civil
Mode S
160
Number of military
Mode S
45
Satellite
ADS-B receiver installed on a satellite in a sun
synchronous orbit at an altitude of 670km
Antenna horn size 5mm
Antenna peak gain 9.6dB
Antenna beamwidth -3dB
Antenna pattern
shape
Elliptical with semi-
major aperture of 25
degrees and semi-
minor aperture of
22.5 degrees
Cable losses 0dB
Minimum signal
level for detection
-92dBm
Figure 1. Satellite ADS-B analysis scenario
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 161
The figures below show the detailed analysis of a high
interference test point in the mid-Atlantic which contains 32
aircraft within the satellite spot beam. Fig. 2 shows the
frequency and type of messages received on 1090MHz split
into four 90 degree sectors while Fig. 3 shows the cumulative
total of messages against the received power [6]. Fig. 4
presents results from a similar analysis to predict interference
levels on the ground and in the air in mainland Europe [7].
The interference at the satellite ADS-B receiver is much
lower than that expected at an ADS-B ground station in
mainland Europe. An ADS-B ground station needs to be able
to decode wanted Extended Squitters from aircraft in its
operational range in the presence of unwanted transmissions or
transmissions from aircraft outside of its operational range.
However, the majority of signals received at the satellite will
be Extended Squitters which we want to decode. The Extended
Squitters also come from aircraft which are at similar distances
from the satellite and are therefore likely to have similar
received signal powers. Therefore it is highly likely that any
overlap between the received transmissions will lead to
garbling of the Extended Squitter.
Using the assumption that any overlap between received
messages would lead to garbling, and an interference model
developed as part of the Eurocontrol ADS-B Coverage
Analysis and Planning Tool (CAPT), we further analyzed the
potential Update Probability (UP) for receiving ADS-B
position updates [8]. Assuming a 2Hz transmission rate for
Extended Squitters containing position data the satellite ADS-
B receiver could support up to 160 aircraft in the spot beam in
an oceanic region whilst still achieving a 95% probability of
update within 5 seconds. However, it should be noted that this
figure is likely to decrease dramatically as aircraft within the
spot beam fall within radar cover and interference levels
increase.
Mode S inteference in each of four 90 degree sectors
0
20
40
60
80
100
120
1 2 3 4
Subsector
F
R
U
I
T
(
H
z
)
Mode S TCAS replies
extended squitters
Short squitters
All-call ModeS (own)
All-call ModeS (other)
Roll-call ModeS (other)
Figure 2. Detailed analusis of interference in mid-Atlantic
Cumulative 1090MHz interference
0
50
100
150
200
250
300
-97 -92 -87 -82 -77
Received Signal Level (dBm)
N
u
m
b
e
r
o
f
M
e
s
s
a
g
e
s
p
e
r
s
e
c
o
n
d
Mode A/C
Short Mode S
Extended Squitter
Total Mode S
Figure 3. Cumulative interference against receieved signal level
Results from study predicting interference levels at
Brussels in 2015
0
20000
40000
60000
80000
100000
120000
Ground test point (0km) Airborne test point (10km)
F
R
U
I
T
(
H
z
)
Mode A/C TCAS
Mode A/C
Short Squitters
Mode S All Call
Mode S TCAS
Extended Squitter
Roll Call Mode S
Figure 4. Predicted levels of 1090MHz interference in mainland Europe
Considering the onward transmission of the ADS-B data
received by the satellite there are two methods:
Bent-pipe where the data received by the satellite is
forwarded on to the ground with amplification and a
shift to the downlink channel frequency.
Regenerative where the data is decoded by the satellite
and then and re-encoded onto the downlink signal.
Table II provides a comparison of these two methods for
the onward transmission of the received ADS-B data against
the likely ADS-B data rate requirements. Considering this
analysis and the analysis above the initial conclusion is that
from a technical point of view it is feasible to receive ADS-B
data using a satellite.
162 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
TABLE II. COMPARISON OF TWO METHODS FOR ONWARD
TRANSMISSION OF ADS-B DATA
1090Mhz ES
Bent-pipe Regenerative
Channel
bandwidth
2.6Mhz (-3dB
point)
Typcially 10s
of MHz for
Ka/Ku band [9]
Typcially 10s of
MHz for Ka/Ku
band [9]
Data rate
22/111 Kpbs
required for
decode of
Extended
Squitters from
32/160 aircraft
Typically 1000s
Kbps per sub-
channel [9]
Typically 1000s
Kbps per sub-
channel [9]
Latency ~0.5 seconds
Up to 2 seconds
[10]
Pros
Shorter latencies
Offers ability ro
post process
signal on ground
for better decode
performance
More efficient
use of downlink
bandwidth
Time of
applicability of
ADS-B report
can be adjusted
for latency of
satellite link
Cons
Time of
applicability of
ADS-B report
ignores latency
of satellite link
Less efficient
use of downlink
bandwidth
Longer latencies
D. Other satellite ADS-B initiatives
The feasibility and potential benefits of tracking aircraft
(and ships) via satellites are now being more and more widely
recognized. There are several other recent initiatives
investigating the feasibility of providing satellite tracking
services and these are summarized below identifying any
further lessons that can be learned.
The SESAR project OPTIMI is currently investigating
technical solutions including the potential for satellite services
to improve monitoring and position tracking of aircraft while in
remote or oceanic areas. In particular this is to support Search
and Rescue operations and accident investigation in these
remote areas. The project aims to deliver recommendations that
can be implemented in 2011. ESA has also recently launched
an invitation to tender to procure an ADS-B payload for the
reception and processing of ADS-B signals on a satellite in-
orbit demonstration mission.
This year the satellite service provider Globalstar has
signed an agreement with ADS-B Technologies in order to
develop a system allowing ADS-B equipped aircraft in remote
and oceanic regions to relay information to the ground and
other aircraft using the Globalstar network of satellites and
ground stations. The proposed Globalstar service is based on a
constellation of low-earth-orbit satellites using a bent-pipe
architecture which they claim will provide near-real-time data
relay.
The Satellite service provider Iridium also announced this
year that they plan on monitoring ADS-B transmissions using
their next generation of 66 communication satellites. Their
current generation of satellites is currently being certified to
provide aviation safety services requiring the satellites meet
defined levels of robustness, reliability and latency. The next
generation of satellites will be fully operational by 2017
providing global coverage. The main technical challenge being
investigated is the need to blank out ADS-B reports received
from high density areas as these could saturate the ADS-B
receiver.
Similar initiatives exist, and in many cases are further
advanced, in the maritime domain for tracking ships using
satellites and AIS - the maritime equivalent of ADS-B.
Originally designed as a terrestrial system, there are now AIS
receivers installed on satellites decoding AIS transmissions
from ships. The Canadian company ExactEarth is already
offering a commercial space-based AIS service using two
microsatellites. It also has plans to launch two more
microsatellites this year that will enable post-processing of the
received AIS data on the ground to improve probability of
detection performance. However, the requirements for the use
of the AIS data may be different from the potential use of
ADS-B data. The dynamics of a ship are very different to
aircraft and AIS data is not used for providing separation
services or collision avoidance. Satellite AIS data is often fused
with other remote imaging data and the primary applications
under investigation are homeland security, search and rescue
and environmental monitoring. Current performance targets are
of the order 80% probability detection and update rates in the
order of minutes or even hours rather than seconds. Experience
from current test or operational AIS satellites also indicate that
probability of detection performance can be variable [11].
V. CONCLUSIONS
The assessment of a potential implementation of satellite
ADS-B demonstrates the feasibility of decoding ADS-B
messages with a high update probability via an ADS-B receiver
installed on a satellite. The assessment also provides evidence
that satellite ADS-B can successfully provide aircraft position
updates to controllers at update rates similar to radar, and much
higher than ADS-C, even considering oceanic traffic growth
well beyond 2015. However, it is noted that this conclusion
only holds true when the aircraft are out of the range of SSR
radars. When aircraft are within radar coverage (e.g. close to
shore) the interference environment experienced by the satellite
will increase dramatically and detection performance will drop.
It is unlikely satellite ADS-B will ever enable surveillance
separations in oceanic airspace that are currently achievable in
Radar (RAD) airspace because for safety reasons a second
layer of surveillance cover would be required that is
sufficiently different from ADS-B. However, satellite ADS-B,
particular when combined with ADS-B ATSA or ASEP
applications may:
Provide a cost-effective means of monitoring from the
ground ATSA and ASEP enabled maneuvers in
oceanic airspace (such as passing) providing increased
safety.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 163
Enable more efficient flight procedures such as those
enabled by ADS-B ground stations installed in Non-
Radar Airspace (NRA).
The ADS-B-NRA application is designed to enhance the
following ICAO air traffic services:
1) ATC service and flight information service principally
for:
ATC separation services, including the possible
reduction of separation minima;
transfer of responsibility for control;
ATC clearances;
flight information services;
flight crew guidance for flight operations in ADS-B
only surveillance airspace.
2) Alerting services, principally for:
notification of rescue coordination centers;
plotting of aircraft in a state of emergency or when
deviating from intended track (e.g. because of bad
weather).
However, the following critical differences between the
application of satellite ADS-B in the oceanic region and ADS-
B-NRA should be noted:
the increased latency of ADS-B reports received by a
satellite compared to a ground station;
the lack of real-time voice (or data) communications in
oceanic airspace compared to NRA.
In terms of implementation timescales it is likely that initial
ADS-B satellite services may be available around the time of
the European ADS-B mandate in 2015. The maritime domain
is leading the way in that it already has commercially available
satellite tracking services and satellite service providers have
identified that a similar business opportunity may exist in
aviation.
However, a word of caution: initial maritime applications
appear to be focused on homeland security and environmental
monitoring applications. If satellite ADS-B is to be used to
improve the efficiency of oceanic operations through new
procedures and reduced separations its actual performance in
terms of probability detection, update rates, latency etc. must
be demonstrated through live trials. Decisions on optimum spot
sizes and the total number of satellites required to give
sufficient satellite coverage also need to be made. Furthermore,
if the satellite ADS-B concept is to be developed within the
aviation community a rigorous cost-benefit analysis is required
to ensure that additional benefits enabled by satellite ADS-B
are cost-effective compared to other technical solutions.
REFERENCES
[1] W. Richards, K. OBrien, D. Miller, New air traffic surveillance
technology, Boeing aero quarterly, quarter 2 2010, unpublished.
[2] B. Stanley, S. Kelly, Helios new surveillance technologies, Helios 2009,
unpublished.
[3] Joint North Atlantic airspace standard operating procedures, August 25
2006, unpublished.
[4] Guidance material on SISTAL FANS-1/A implementation on SAL
airspace, unpublished
[5] B. Stanley, S. Kelly, Helios ADS-B training course, 2008, unpublished.
[6] A. Parkinson, P1175 ADS-B satellite 1090MHz scenario simulation
technical note, version 0.2, September 2009, unpublished.
[7] R. McDonald, CASCADE programme 1090MHz interference study
final report, version 2.3, July 2006, unpublished
[8] A. Parkinson, P895 CAPTv2 software requirements specification
(algorithms), version 0.k, December 2010, unpublished.
[9] M. Bever, J. Freitag, S. Linsky, J. Myers, R. Nuber, J. Prieto Jr., E.
Wiswell, Fast-packet vs. circuit switch and bent pipe satellite network
architectures, Fourth Ka-Band Utilization Conference Venice, Italy,
November 1998
[10] O. Gupta, Global augmentation of ADS-B using Iridium NEXT hosted
payloads, February 2011, unpublished.
[11] F. te Hennepe, R Rinaldo, A. Ginesi, C. Tobehn, M. Wieser,
. Helleren, Feasibility of an European constellation for space-based
detection of AIS signals, European symposium on satellite-AIS,
December 2010.
164 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
StrategiestoDesignandDeployModeSMultilaterationSystems
Session7.1page167
CorrectionofsystematicerrorsinWideAreaMultilateration
Session7.2page173
Multilaterationsystemtimesynchronizationviaoverdetermination
ofTDOAmeasurements
Session7.3page179
ImprovementofMultilateration(MLAT)AccuracyandConvergence
forAirportSurveillance
Session7.4page185
AssessingthesafetyofWAMoveranonradarsurveillancearea
Session7.5page191
stations
(with
(1)
Equation (1) provides the number of possible combinations
given the size of the discrete search space P and the number of
stations to be deployed
(2)
where PoD is the probability of detection of one station and it
should be provided by the manufacturer and,
is the number
of stations that cover the jth point. In (2) it is assumed that at
least four stations are needed to calculate the position. By (2) it
can be estimated the minimum number of stations that make
(3)
where cond is the total number of requirements and
restrictions,
on
the design. The corresponding values of
-array of integer
numbers, where the value of the ith array position represents
the index of the selected site for the ith station. Instead, in [6]
each individual is composed by the set of (x,y) coordinates of
the stations. Moreover, it is worth to say that the information
contained in a specific individual position can change and
depends on the parameters to be optimized in the design. This
particularity is commented in the next section.
III. SIMULATION AND RESULTS
To validate the strategies proposed in this work three
different simulations have been carried out over the layout of
Barcelona (Spain) Airport. The common objective for all the
simulations is to obtain a MLAT system which cover the three
runways, the taxiways and the apron centrelines, given a set of
requirements and restrictions. The first simulation consists in
the design of a MLAT system with a fixed number of Time
Difference of Arrival (TDOA) stations. The second one
consists in the design of a MLAT system with a variable
number of stations. In this simulation, the objective is to find a
design that satisfies all the requirements and restrictions by
using the possible minimum number of TDOA stations. The
last simulation consists in the design of a MLAT system with a
fixed number of TDOA and AOA stations. Fig. 2 shows the
Barcelona airport layout and the P-set of available sites for the
simulations. For these simulations P=41.
For all the simulations, the antenna station height (mast
length) has been assumed to be equal to 2 m and the
calculations for LoS and CRLB are performed for a spatial grid
of 5m 5m. This spatial grid is also in concordance with the
Digital Terrain Model (DTM) used to calculate the LoS. The
Genetic Algorithms (GA) parameters for all the simulations are
those described in [6].
A. MLAT System with a Fixed Number of TDOA Stations
The first scenario shows the first and the standard strategy
proposed herein. It consists in the design of a MLAT system
for a given set of requirements and restrictions. The
requirements for this particular simulation are based on those
described in [1], which are basically: Horizontal accuracy must
be within 3.75 m and the System Probability of Detection must
be better than 99.9%. The number of stations to use in this
design is twelve and they measure only the TDOA parameter.
The restriction of LoS redundancy, using a station probability
of detection of PoD=97%, provided by a quick evaluation of
(2) is 7 and the minimum spatial separation is
= 400 m.
For this scenario, an individual is an array of 121 size,
where the ith position represents the index of the possible
position for the ith station and it can be written as
, where
and
(4)
where
(5)
and
(6)
Finally, the value of the weight factors depends on the
importance given to each requirement or restriction on the
design; they can be chosen by the designer. Here, we have used
=0.95 and
(7)
where
.
Fig. 3 shows the horizontal accuracy for this scenario and
how the interested airport areas are covered with the assumed
requirements. From the theory [2], [4-5] it is well known that a
correct system geometry, to obtain high accuracy levels, is to
set the stations in a polygon enclosing the interest area. In Fig.
3 it can be observed that the proposed procedure provides a
solution that is in the line of this theoretical aspect. Finally,
Fig. 4 shows the procedure convergence. In this scenario, the
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 169
Figure 3. Horizontal accuracy for the design with a fixed number of
TDOA stations.
Figure 4. GA convergence for the design with a fixed number of TDOA
stations.
number of possible combinations, provided by (1), is
7.898710
9
and a relative good solution is obtained within 50
iterations, which means only 500 problem evaluations.
However, it is advisable to expend more iterations (up to 200)
because the random component of the GA allows to the
procedure the exploration of new values in the search space. In
any case the total number of problem evaluations is much
smaller than that value provided by (1).
B. MLAT System with a Variable Number of Stations
The second scenario consists in the design of a MLAT
system with a variable number of TDOA stations. In this kind
of scenario, the objective is not only to calculate the stations
sites but it is also to calculate a relative minimum number of
stations that satisfy all of the assumed requirements and
restrictions. All requirements and restrictions for this problem
are those described for the first problem. Moreover, for this
problem it is necessary to stipulate a range for the number of
stations. For this work, we have used a range of
.
For this scenario, an individual is an array of variable
length, where the first position sets the length of this. It can be
written as
, where
is the number of
stations calculated at time t. The fitness function for this
scenario takes the following form,
(8)
where
and
min
max
min
(9)
Finally, the weight factors values used for this problem are
=0.85,
=0.05 and
=0.1.
Fig. 5 shows the results for the horizontal accuracy. Also in
this scenario, all the areas of interest are covered satisfying all
requirements and restrictions. The important aspect in this
scenario is that the minimum number of stations calculated is
11, it is, one less station than in the first scenario. This kind of
simulation is useful to know an approximate minimum number
of stations that meets the requirements and restrictions.
However, due to the random component of the GA it is
advisable to run the procedure, for this scenario, once or twice
more, just to validate the calculated minimum number. Finally,
Fig. 6 shows the procedure convergence for this scenario, for
this scenario a good solution is found after 150 iterations. It can
be understood because the complexity of this problem (number
of possible combinations) is much greater than that of the first
scenario.
C. MLAT System with a Fixed Number of TDOA/AOA
stations
This scenario consists in the design of an improved MLAT
system with a fixed number of TDOA/AOA stations.
Normally, the AOA measurement capabilities are added to
improve the horizontal accuracy in surface movement
applications [2]. For this scenario the requirements and
restrictions are those described for the first problem and the
AOA measurements capabilities are added only to the station
number 1 (the AOA measurements error is assumed to be 10
-3
rad).
For this scenario, an individual is represented as in the first
scenario, i.e., as an array of 121 size
. The
difference lies in that, for this scenario, the pertaining LoS
coverage of the station number 1 is relatively more important
than those of the remaining stations. This particular aspect is
introduced in the fitness function as follows,
(10)
where
and
(11)
Finally, the weight factors values used for this problem are
=0.9,
=0.05 and
=0.05.
Fig. 7 shows the horizontal accuracy for this scenario. The
complexity of this problem is basically of the same order than
that of the first one but, here the CRLB calculation has been
carried out by taking into account the accuracy improvement
provided by the TDOA/AOA station [2]. The final site for this
station is shown in Fig. 7 as the diamond. Also for this kind of
scenario it is advisable to run the procedure once or twice
more. Similarly to the first problem, here a good solution is
found after 50 iterations (see Fig. 8).
IV. CONCLUSION
In this work, a set of practical and useful strategies to
design and deploy Mode-S Multilateration systems has been
presented. These strategies are based on the use of genetic
algorithms along with the well-known CRLB analysis. A
general procedure to use these strategies is also proposed and it
is useful to design new MLAT systems but also to validate
whether a previous system design could be the optimum
solution regarding to a set of available resources.
Three kinds of scenarios have been presented. The first one
is able to design new MLAT systems with a fixed number of
TDOA stations but also to validate whether a final design
(clearly before the implementation) can be improved by
feasible but not obvious sites changes. The second one
provides a strategy to obtain a minimum number of stations
which satisfy all the stipulated requirements and restrictions.
The third scenario is proposed to design improved MLAT
systems, i.e., by using other type of measurements like AOA or
RTD. For this third scenario, an example with a MLAT system
using TDOA/AOA stations has been presented but, the use
with other measurements combinations is straightforward.
Finally, it is worth to say that also these strategies can be used
together in order to obtain more reliable results, e.g., firstly the
second scenario can be used to obtain a possible minimum
number of stations that meets all the requirements and
restrictions and then, by means of the first scenario, obtain the
optimum sites or just to validate that set obtained with the
second scenario.
The use of new requirements or restrictions is also possible
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 171
only by modifying the corresponding cost function and their
weight factors.
REFERENCES
[1] "Minimum Operational Performance Specification for Mode S
Multilateration Systems for Use in Advanced Surface Movement
Guidance and Control Systems (A-SMGCS)," EUROCAE, 2003.
[2] G. Galati, M. Leonardi, and M. Tosti, "Multilateration (Local and Wide
area) as a distributed sensor system: Lower bounds of accuracy," in
European Radar Conference, EuRAD, Amsterdam, 30-31 Oct. 2008, pp.
196-199.
[3] H. B. Lee, "Accuracy Limitations of Hyperbolic Multilateration
System," IEEE Transactions on Aerospace and Electronic System, vol.
AES-11, January 1975.
[4] N. Levanon, "Lowest GDOP in 2-D scenarios," IEE Proc. Radar, Sonar
Navig., vol. 147, pp. 149-155, June 2000.
[5] D. J. Torrieri, "Statistical Theory of Passive Location Systems," IEEE
Transactions on Aerospace and Electronic System, vol. AES-20, pp.
183-198, March 1984.
[6] I. A. Mantilla-G, R. F. Ruiz, J. V. Balbastre-T, and E. d. l. Reyes,
"Application of Metaheuristic Optimization Techniques to
Multilateration System Deployment," in Enhanced Solutions for Aircraft
and Vehicle Surveillance Applications, ESAVS 2010, German Institute of
Navigation (DGON), Berlin, Germany, 16-17 March 2010, p. Session
2B/3.
172 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Correction of systematic errors in
Wide Area Multilateration
Jorge M. Abbud, Gonzalo de Miguel, Juan Besada
GPDS-SSR
Universidad Politcnica de Madrid
Madrid, Spain
jorge.jose.abbud.momma@gmail.com, gonzalo@grpss.ssr.upm.es, besada@grpss.ssr.upm.es
AbstractThis work presents a method to estimate and correct
slow time-dependent position errors due to non perfect ground
station synchronization and tropospheric propagation. It uses
opportunity traffic emissions, i.e. signals transmitted from the
aircrafts within the coverage zone. This method is used to
overcome the difficulty of installing reference beacons
simultaneously visible by all the base stations in a given Wide
Area Multilateration (WAM) system.
Keywords- Wide Area Multilateration; Air Traffic Control;
ADS-B estimation
I. INTRODUCTION
Due to the performance improvements of multilateration
systems, their application range has been extended from short-
range applications (airport surveillance) to medium-range
surveillance, such as surveillance in Terminal Maneuvering
Areas (TMA) [1]. This system has been called Wide Area
Multilateration system (WAM). Under this positive
performance evolution, WAM becomes a firm candidate to
replace secondary radars in the surveillance network for Air
Traffic Control (ATC) [2].
Multilateration determines the aircrafts position by
using the time of arrival (TOA) of the signal travelling from
the aircraft itself to a network of fixed receivers (base
stations). If the signal is properly coded, it is easy to associate
the TOAs relative to one single transmission in the different
base stations. This is the typical case in ATC, where RF
emissions (ADS-B or TCAS) are used. This way, when the
system has the complete set of TOAs in all ground stations,
the aircraft position can be determined. The trend in the future
ATC surveillance is to use ADS-B as the main source of
aircraft positioning. But it is still necessary to have a
collaborative backup system in order to enhance surveillance
integrity [2]. A promising solution is the use of ADS-B ground
stations as WAM base stations. Each base station will send the
measured TOA together with the ADS-B information to the
ATC control center. Multilateration is performed by
processing the TOAs [2].
The accuracy of the multilateration position is determined
by the errors in the TOA estimates. From a data processing
perspective, these errors can be grouped into three main
categories [2]: white noise, synchronization issue among
ground stations and propagation effects. The first two are
present in any multilateration scenario, although white noise
effects in the position determination are not critical for the S/N
values usually managed in these systems. On the other hand,
although propagation error has not been taken into account for
short-range applications, given the distance between base
stations in WAM scenarios (up to 20 or 30NM [1]) this source
of error has to be considered. This is required in order to
preserve the accuracy from suffering degradation along the
coverage area (i.e. a low Dilution of Precision).
This characteristic rules out the calibration philosophy to
reduce both synchronization and propagation errors, since
installing fixed beacons in Line-Of-Sight with all base stations
is costly, if not impossible, for large baseline separations.
Also, the propagation error has a hard dependence with
aircraft altitude. This way, calibrations for on-ground targets
are not valid for flying aircrafts.
In order to solve the synchronization issue between
stations, GPS-based methods could be used, but these would
not reduce propagation errors. Furthermore, a backup
synchronization subsystem would be necessary in order to
mitigate hypothetical failures of GPS.
Therefore, one solution to this problem is to add a
processing subsystem which corrects synchronization and
propagation errors simultaneously, by analyzing the signals
transmitted by all aircrafts currently present within the WAM
coverage (opportunity traffic).
In a recent paper, authors have studied the possibility of
performing calibration using opportunity traffic for WAM
systems [5]. Figure 1 presents the block diagram of the
proposed calibration mechanism. TOAs measured in each
station are associated and sent to the central processor which
computes the position. This can be either a master station in
the multilateration system or a remote station fusing the
information of many sensors (this can be the case for ADS-B
technology [3]). The first operation is to apply calibration
corrections to pseudoranges (for synchronization issues and
propagation). Then, target coordinates are determined as if the
calibration was perfect, modeling the propagation error as a
polynomial depending on distance. The output of this block is
delivered as a position determined by WAM system. In order
to compensate for slow time variations in the real propagation
and calibration conditions, the system has an open-loop
The work has been financed by Spanish Science and Technology Office
under projects TEC-2008-06732 y TIN-2008-06742.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 173
control system that modifies the estimated propagation
constants, maintaining the system calibrated.
The system selects the targets located at a determined
flight level and spread around the coverage area. Then, the
system determines their position, as well as the calibration
constants, triggering the iterative algorithm from the target
coordinates delivered by WAM system and the calibration
constants at the output of the averaging filter. Calibration
constants are averaged in order to reduce their variance. Once
filtered, they are used for the correction of future
pseudoranges.
The position determination is done using the proposed
mechanism in [6]. It uses an iterative algorithm with the
linearized multilateration equation [7]. This algorithm needs an
initial position which is determined using the closed form
algorithm of [8]. The system can be implemented in TOA form
(determining time of emission, cT
e
) or in TDOA form
(eliminating cT
e
). If cT
e
is not necessary, the TDOA form is
more accurate due to the complete ignorance about the time of
emission in WAM. This is the approach assumed in this work.
The same method applies to the determination of calibration
constants plus position.
Averaging
Computation of
calibration
constants
Selection of
calibration
targets
cTi
cT
im
, K
m
(x,y,z, cTe) TOD Calibration
K
Position
determination
(DTOA or TOA)
TODs
Figure 1. Block diagram of calibration mechanism using opportunity traffic.
This approach uses a linear model for propagation error
and a constant error to represent the synchronization error of
each base station. Linear propagation models are appropriate
for medium distances (around 75-100 km). For longer
distances, calibrations based on linear models do not have the
required performance. This paper will extend the method
proposed in [5] by including second-order propagation models.
The inclusion of a polynomial model to correct the
synchronization error in each base station will be studied as
well.
The paper is structured as follows: section 2 states the
problem of position determination with WAM including
propagation and clock synchronization errors. Section 3
compares the accuracy of the system using opportunity traffic
when propagation effects are modelled through first or second-
order model (with respect to distance). Finally, section 4
analyzes the accuracy degradation due to the clock drifts,
followed by a mitigation method proposal.
II. CHARACTERISATION OF SLOWLY-VARIANT ERRORS IN
WAM SYSTEMS
The aircraft position is determined by means of the time
difference of arrival (TDOA) of the signal at the different base
stations. As a first step, a method based on hyperbolic location
as described in [9] can be used. Additionally, in the presence of
error, a gradient method starting from the solution of the
previous treatment will be used in order to refine the location
[7].
Each base station in the scenario measures the TOA of the
signal received from the target aircraft. The TOA of the signal
traveling from the j-th aircraft located in (x
j
,y
j
,z
j
) to the i-th
base station located in (x
i
,y
i
,z
i
) can be represented by the
following expression:
( ) ( ) ( )
i i e
j
i
j
i
j
i
j
i i
j
i
n T T P
c
z z y y x x
c
TOA
+ + + +
+ + = =
1
1 2 2 2
(1)
Equation (1) can be rewritten as:
i i e
j
i ij i
j
i
n T T P
c
R
c
TOA + + + + = =
1 1
(2)
where R
ij
stands for the Euclidean distance between the i-th
station and j-th aircraft, P
i
j
represents the propagation error,
T
e
represents the signal emission time, T
i
is the synchronism
error, and n
i
the white noise random error.
In order to eliminate the signal emission time uncertainty,
the aircraft position will be assessed based on the Time
Difference of Arrival (TDOA). This means that all available
TOAs for a single emission are referenced to the TOA on one
of the base stations. Thus, the TOA equation system is now
replaced by a TDOA equation-system, with one less unknown,
as well as one less equation:
( )
( ) ( ) ( )
m i m i
j
m
j
i
mj ij
j
i
j
m
j
i
j
m i
n n T T P P
c
R R
c
TDOA
+ + +
= = =
1
1
,
(3)
where the reference base station is the m-th station.
Let us now characterize the propagation error P
i
j
. It is the
uniform vertical gradient of atmospheric refractive index that
bends the signal propagation trajectory and changes the
velocity of light, delaying its arrival to the base station. Figure
2 represents this propagation error with respect to the slant
range for aircrafts flying at altitudes between 6000m and
14000m AMSL, using the expressions defined in [10]. The
base station is considered to be at sea level.
174 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Figure 2. Systematic error (range bias) due to radio wave propagation for a
standard atmosphere.
Two relevant characteristics can be observed concerning
this propagation error. First, a second degree polynomial seems
to be a good fit for modeling the range bias with respect to the
distance. Actually, even a linear approximation can prove to be
sufficient for short distances [5]. Second, the coefficients of the
polynomial depend on the aircraft altitude. The last observation
forces to estimate different propagation models as a function of
aircraft height. The system divides the height in different layers
(with 1-2 km of thickness). For each altitude layer, the system
performs an independent propagation calibration based on the
aircrafts inside it.
Taking into account the previous observations, equation
(2) can be approximated substituting propagation error by a
first-order or second-order model versus Euclidean range:
( )
i i e ij i
j
i
n T T R K
c
TOA + + + + = 1
1
(4)
( ) [ ]
i i e ij ij i
j
i
n T T R K R K
c
TOA + + + + + =
2
2 1
1
1
(5)
The method to estimate parameters for both models will be
described in section 3.
Now we shall characterize the clock drift occurring in the
base stations. In this study, we consider that since the signal
emissions are quasi-periodic, a Time Interval Error (TIE)
model of a local clock will be used based on the philosophy of
[11]. It consists of a polynomial model projecting ahead on a
horizon of N points from the starting point with the k-th
degree Taylor expansion:
( ) ( ) T n w
p
T n
n T
k
p
p p
p i
,
!
0
1 _
=
+ = (6)
where n is the sample number, T is the time step,
( ) [ ] k p
p p
, 0 , 0
1 1
+ +
the initial states of the clock and
w
1
(n,T) is a clock noise with known properties. For large
values of n, the polynomial component dominates over
w
1
(n,T). Thus, in this study, we shall characterize this local
clock with a second-order polynomial without considering
w
1
(n,T). Expression (6) now becomes:
( ) ( ) ( ) ( )
2 2
2 , 1 , 0 ,
0
2
1
0 0 T n nT nT T
i i i i
+ + = (7)
This way, TDOA between stations i-th and m-th for a
signal transmitted at time t=nT for the j-th aircraft under
coverage can now be written in the following ways:
( ) ( ) [ ( )( )
( ) ] ( ) ( ) ( ) ( )
m i m i mj ij
mj ij
j
i
j
m i
n n nT T nT T R R K
R R K
c
nT TDOA nT
+ + +
+ = =
2 2
2
1 ,
1
1
(8)
On the right side, we are now in a position to set a system
of TDOA equations in order to determine the aircraft position.
On the other hand, the number of unknowns has increased due
to the characterization of propagation effects and relative
synchronization errors. Therefore, the solution of the system
shall not only contain the aircraft coordinates, but also the
constants relative to both propagation error and clock drift.
In order to avoid an indeterminate system of non-linear
equations, a set of new independent equations must be
obtained. For this purpose, the opportunity traffic method will
be used.
III. CORRECTION OF THE SYSTEMATIC ERROR DUE TO
PROPAGATION
This section focuses solely on the propagation effects and
the technique used to solve the equation system without
considering clock drifts (synchronization error is considered
constant with time).
Since the calibration constants must be determined together
with coordinates (three spatial coordinates plus emission time
for each aircraft, two propagation constants and one
synchronization constant for each base station minus one for
the reference station), there is a need to process jointly the
TDOA of M (N/(N-4)) aircrafts for the linear propagation
error model to obtain the sufficient number of equations. It is
necessary that the number of base stations, N, is greater than 4,
the minimum number of pseudo-ranges required to determine
spatial coordinates and emission time. The pseudo-ranges of
the extra stations are used as equation to determine the
additional unknowns of the calibration models. When using the
parabolic propagation error model, the minimum number of
aircrafts under coverage shall be M (N+1)/(N-4).
A larger amount of jointly-processed aircrafts yield a
stronger degree of over-determination, providing therefore
higher stability in the estimates. The drawback is that larger
sparse matrices and larger amount of data shall be handled.
The non-linear equation system allowing the simultaneous
determination of calibration constants as well as the location of
the aircrafts is as follows (noise terms are not included for the
sake of clarity):
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 175
( ) ( )
( )( ) ( ) [ ] ( )
( ) ( )
( )( ) ( ) [ ] ( )
( ) ( )
( )( ) ( ) [ ] ( )
( ) ( )
( )( ) ( ) [ ] ( ) 0 1
1
0 0
...
0 1
1
0 0
...
0 1
1
0 0
...
0 1
1
0 0
1
2
1
2
2 1 1
1 ,
21
2
1
2
2 2 1 2 1
2 1 , 2
1
2
11
2
1 2 11 1 1
1 1
1 ,
21
2
11
2
21 2 11 21 1
1
2
1
1 , 2
N j Nj j Nj
j
N
j
N
j j j j
j j
N N N
N N
R R K R R K
c
TDOA
R R K R R K
c
TDOA
R R K R R K
c
TDOA
R R K R R K
c
TDOA
+ + +
=
+ + +
=
+ + +
=
+ + +
=
(9)
where
21
(0)=
2
(0)-
1
(0), denoted
21
within the remaining
part of this paper represent the difference between
synchronization errors in two base stations. Note that it is
sufficient to determine the difference between synchronization
errors since the measured magnitude is the TDOA. So, in a
scenario involving N stations and M aircrafts, the amount of
TDOA equations is M(N-1).
One way to solve this system is by using a gradient method
[7][8], setting the initial value around the intersection of the
hyperboloids. The initial condition is determined as indicated
in figure 1. Iteration using the linearized system shall be
performed until the convergence criteria based on the accuracy
requirements have been met.
The vector composed by the unknowns is as follows:
[
]
T
N
M M M
K K c c
z y x z y x z y x
2 1 1 , 1 , 2
2 2 2 1 1 1
, , ,...,
, , , ,..., , , , , ,
= x
(10)
(x
j
,y
j
,z
j
) being the position of the j-th aircraft, c
i,1
being the
synchronization errors between the clocks of stations 1 and i,
and K
1
, K
2
the coefficients relative to degrees 1 and 2 modeling
the propagation error effect.
The system in (9) can be solved using the following
iteration:
k k k
x x x + =
1
(11)
The initial value
0
x must be set around the true solution.
This is done using previous estimations of calibration constants
and the position determined for each aircraft by the WAM
system before calibration process.
For each iteration process, an estimation of the TDOAs is
performed using the estimates of vector
1 k
x . The error
between the measured values and its estimates is defined as
follows:
( )
1
=
k k
x f (12)
where f is the set of TDOA estimates. As an example,
) (
1 ,
k j
i
f x is defined in expression (13) below:
( ) ( )( ) [ ( )]
k k
j
k
j i
k k
j
k
j i
k k j
i
i
R R K R R K
c
f
1 ,
2
, 1
2
, 2 , 1 , 1 1 ,
1
1
+ + + = x
(13)
The differential increment is obtained by solving the first
derivative terms of the linearized version of system (9). Thus,
expression (12) can also be written as:
k k k
x A = (14)
where
k
A is the gradient matrix of system (9) at the k-th
iteration. Matrix A is obtained using the following expressions:
(
(
(
(
(
(
(
=
2
,1
s
N ,1
s
N ,1
s
N
2
4,1 4,1 4,1
2
3,1 3,1 3,1
2
2,1 2,1 2,1
R R 1 0 0
0
R R 0 0 0
R R 0 1 0
R R 0 0 1
A
"
# # " # # #
"
"
"
(15)
where:
(
(
(
(
(
=
A/C
N
i,1 i,1 i,1
i,1
2
i,1 i,1
i,1 i,1
1
i,1
i,1
0 0
0 0
0 0
"
# " # #
"
"
[ ] 0 0 0 =
i,1
0
[ ] [ ]
T T
0 0 0 ; 1 1 1 " " = = 0 1
[ ]
T
i j j i i i
R R R R R R R R
A/C A/C
N N
i,1 R
, 1 , , 1 , 2 , 1 2 , 1 , 1 1 ,
... , , , = "
[ ]
T
i j j i i i
R R R R R R R R
2
, 1
2
,
2
, 1
2
,
2
2 , 1
2
2 ,
2
1 , 1
2
1 ,
... , ... , ,
A/C A/C
N N
2
i,1 R =
( ) ( )
( ) ( )
( ) ( )
T
b
i
b
j
b
j
ij
b
i j
b
i
b
j
b
j
ij
b
i j
b
i
b
j
b
j
ij
b
i j
j
i
z z K
R
z z
R
z z
K
y y K
R
y y
R
y y
K
x x K
R
x x
R
x x
f
K
|
|
|
|
|
|
|
|
|
.
|
\
|
+
|
|
.
|
\
|
+
+
|
|
.
|
\
|
+
+
|
|
.
|
\
|
+
= =
1 2
1
1
1
1 2
1
1
1
1 2
1
1
1
1 ,
2 1
2 1
2 1
j
i,1
Since system (14) is usually over-determined, it must be
solved using the minimum mean square error (MMSE)
expression:
176 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
( ) ( ) ( ) ( ) ( )
k k
T
k k k
T
k k
S A A S A x =
1
1
1
(16)
where S
k
is the TDOA covariance matrix and can be estimated
from the S/N in each receiver.
In order to assess the performance of the opportunity traffic
method, a hypothetical WAM scenario is simulated. The
system here considered is composed by six stations, located in
the corners of a square (side: 100 Km) and two in the middle of
two vertical sides. The station altitudes are arbitrary, but near
sea level (this implies a poor performance in aircraft altitude
determination). Six aircrafts are considered for the calibration
process, all of them outside the square delimited by base
stations 1 to 4 with a height of 10 Km ((x Km,y Km): (-150,60)
(100,90) (80,-40) (-120,-70) (20,80) (-30,-110)).
Figures 3 to 6 display the mean and standard deviation of
the WAM location error for pseudo-range error standard
deviation values ranging between 1 and 10m. The results have
been obtained using Monte Carlo experiments, averaging
sufficient independent solutions to turn the simulation variance
negligible. Results displayed in figures 3 and 4 have been
obtained using the linear model for propagation error, whereas
figures 5 and 6 show the results obtained using the parabolic
model.
1 2 3 4 5 6 7 8 9 10
40
60
80
100
120
140
160
180
200
220
240
Standard deviation of pseudorange error (m)
M
e
a
n
v
a
l
u
e
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
i
n
X
Y
(
m
)
A/C 1
A/C 2
A/C 3
A/C 4
A/C 5
A/C 6
Figure 3. Mean of WAM position error using the order 1 estimation of the
propagation error.
1 2 3 4 5 6 7 8 9 10
0
20
40
60
80
100
120
140
160
180
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
i
n
X
Y
(
m
)
Standard deviation of pseudorange error (m)
A/C 1
A/C 2
A/C 3
A/C 4
A/C 5
A/C 6
Figure 4. Standard deviation of WAM position error using the order 1
estimation of the propagation error.
1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
M
e
a
n
v
a
l
u
e
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
i
n
X
Y
(
m
)
Standard deviation of pseudorange error (m)
A/C 1
A/C 2
A/C 3
A/C 4
A/C 5
A/C 6
Figure 5. Mean of WAM position error using the order 2 estimation of the
propagation error.
1 2 3 4 5 6 7 8 9 10
0
50
100
150
200
250
300
350
400
450
500
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
i
n
X
Y
(
m
)
Standard deviation of pseudorange error (m)
A/C 1
A/C 2
A/C 3
A/C 4
A/C 5
A/C 6
Figure 6. Standard deviation of WAM position error using the order 2
estimation of the propagation error.
Concerning the performance in the X-Y plane, results show
that the parabolic model sacrifices the variance of the location
error on behalf of its mean value.
IV. CLOCK DRIFT EFFECT ON POSITION ACCURACY
This section covers the assessment of the dual correction
opportunity traffic method, since it considers the propagation
effects, as well as the drifts suffered by the clocks placed on
each base station. Expression (6) suggests that the TOA
measurement induced by this drift varies over time. Figure 7
shows how the errors increase rapidly in a biased way.
Therefore, the TDOA estimates generated by the algorithm
must consider the behavior of each local clock (more precisely,
the difference between drifts).
Consequently, the iterative algorithm presented in section 3
is to be used once again, in order to refine the target location.
Since two new unknowns shall be taken into account for each
equation of the TDOA system, necessary expansions are to be
made in expressions (9) to (16) in order to accommodate the
TDOA measurements gathered at different instants.
Besides, the scenario described in section 3 has been
upgraded, as local oscillators are modeled with a second-order
polynomial along the time axis. The experiments have been
carried out considering clock parameters typically used in base
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 177
stations (e.g. atomic clocks, with the following coefficients for
(6): 10
-10
, 310
-11
, 10
-13
).
Figure 8 show how the two-step approach based on the
opportunity traffic method is capable of bounding the mean
error with a small increase in standard deviation.
0 200 400 600 800 1000 1200
-800
-600
-400
-200
0
200
400
600
800
M
e
a
n
v
a
l
u
e
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
(
m
)
Time elapsed (s)
1
st
A/C - Xaxis
1
st
A/C - Yaxis
2
nd
A/C - Xaxis
2
nd
A/C - Yaxis
3
rd
A/C - X axis
3
rd
A/C - Yaxis
4
th
A/C - X axis
4
th
A/C - Yaxis
Figure 7. Mean of WAM position error obtained without considering clock
drift.
0 200 400 600 800 1000 1200
0
2
4
6
8
10
12
Time (s)
M
e
a
n
o
f
W
A
M
l
o
c
a
t
i
o
n
e
r
r
o
r
(
m
)
A/C 1
A/C 2
A/C 3
A/C 4
A/C 5
A/C 6
Figure 8. Mean value of horizontal position error with the two-step approach.
V. CONCLUSION
This paper presents a collaborative backup system that
enhances ATC surveillance integrity. Its cost is relatively low
since this system reuses the ADS-B ground stations.
This system is able to mitigate the impact of the
propagation effects, as well as the impact of the clock drift
effects for a limited period of time without using calibration
stations. Cases in point are scenarios leading to temporary GPS
unavailability, such as spoofing, insufficient number of
acquired satellites or even a failure in the GPS receiver aboard
the aircraft.
REFERENCES
[1] W.H.L.Neven, T.J. Quilter, R. Weedon, R. A. Hogendoorn, Wide Area
Multilateration, Report on EATMP TRS 131/04. Version 1.1 - National
Lucht en Ruimtevaartlaboratorium, August 2005.
[2] SESAR Definition Phase Deliverable 3: the ATM target concept,
SESAR Consortium, September, 2007.
[3] EUROCONTROL standard document for surveillance data exchange.
Part 12: Category 21, ADS-B messages, EUROCONTROL, ed 1.1
September 2008.
[4] M. J. Leeson, Error Analysis for a Wide Area Multilateration
System,QinetiQ/C&IS/ADC/520896/7/19, 2006.
[5] G. de Miguel, J. Besada, J. Garca, Correction of propagation errors in
Wide Area Multilateration systems, European Radar Conference 2009
(EuRad 2009), Rome (Italy), September 2009.
[6] G. Galati, M. Leonardi, P. Magar, V. Paciucci, Wide Area
Surveillance using SSR Mode S Multilateration: advantages and
limitations, European Radar Conference 2005, EURAD 2005.
[7] W. H. Foy, Position-Location Solutions by Taylor-Series Estimation,
IEEE Trans. on Aer. and Elec. Syst., Vol. AES-12, No. 2, pp. 187-194,
March 1976.
[8] G. Strang, K. Borre, Linear Algebra, Geodesy, and GPS, Wellesley-
Cambridge Press, 1997.
[9] Y.T.Chan, and K.C.Ho, A simple and efficient estimator for hyperbolic
location, IEEE Transactions on Signal Processing, Vol. 42, No. 8, pp.
1905-1915, August 1994.
[10] D. K. Barton, Radar System Analysis and Modeling, Artech House,
2005.
[11] Y. S. Shmaliy, An unbiased FIR filter for TIE model of a local clock in
applications to GPS-based timekeeping, IEEE Transactions on
Ultrasonics, Ferroelectrics, and Frequency Control, UFFC-53, pp. 862-
870, 2006.
178 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Multilateration system time synchronization via over-
determination of TDOA measurements
Martin Pelant
Advanced Programs
ERA a.s.
Pardubice, Czech Republic
mpelant@erabeyondradar.com
Vojtch Stejskal
Advanced Programs
ERA a.s.
Pardubice, Czech Republic
vstejskal@erabeyondradar.com
Abstract Sensors of any multilateration system have to be
precisely synchronized to achieve accurate position estimation
IURP WDUJHWV 7DOA measurements. The GPS common view
principle is frequently implemented in Distributed Time systems
and on the other hand additional delays with one time reference
in Central Time system. Such systems, however, may not operate
effectively if the reference time is not available. This paper
provides description together with test results of clock
synchronization estimated using position over-determination
principles applied to TDOA measurements of well-tracked
targets in WAM system (Brno, Ostrava and simulated joint
system operated by Czech ANS) with a height of about 5 km and
higher. Clock correction factors or delays can then be applied to
sensor clock signals, providing accurate tracking of targets such
as aircraft, even in the event of GPS failure or jamming.
Keywords - Multilateration, TDOA measurements, time
synchronization, over-determination principle.
I. INTRODUCTION
Nowadays Wide Area Multilateration (WAM) systems use for
their sensors` synchronization a Distributed Time (DT) or
Central Time (CT) principle. This means implementation of a
time reference into side stations when using Distributed Time
which is usually in form of GPS Common View equipment.
Such a system requires only minor levels of systematic time
error corrections on a Central Processing Station (CPS).
Incoming pulse trains are already time stamped (TOA) on each
sensor. On the other hand, Central Time uses only one time
reference implemented at the CPS. TOA stamping is performed
at the CPS to which are all trains forwarded.
Both time synchronization principles are more or less
affected by systematic errors of Time Difference of Arrival
(TDOA) measurements. The main challenge in WAM system
design and deployment is its systematic error reduction (e.g.
systematic errors of additional time delays in CT system)
which can be called calibration. The principle of
synchronization/calibration is based on the assumption that all
targets are tracked in real 3D space, and the fact that the vector
of primary, linearly independent, TOA measurements of the
signal belonging to the target may have dimension greater than
three (rather than a real 3D space). The theoretical maximum
dimension of TDOA measurements vector is equal to the
number of stations reduced by one.
A special case occurs when any time reference is missing
(i.e. GPS in DT). It this case, the system can be
synchronized/calibrated only from one source respectively
from real traffic TDOA measurements with targets of unknown
position. There exist two different means of calibration:
- Time calibration from over determined TDOA
measurements with known behavior in height.
Minimal operational requirements are 4 stations
and one independent source with known height
(this can be also ground beacon).
- Time calibration from over determined TDOA
measurements of targets with unknown position.
Minimal system requirement is 5 stations.
This paper will show results of the second method of time
synchronization/calibration. The systematic time errors (of
additional delays) together with position accuracy (covariance
matrix) and current estimate of clock deviations (TDOA
deviations) can be then estimated.
II. OVER-DETERMINATION PRINCIPLE
In operation, the method uses the previously reported known
position, for example, the position prior to the GPS failure,
and/or the ADS-B signal, Mode-S or Barometric height
reported by mode-C transponder. This position data may be
only an approximate location of the aircraft, since the data may
not be accurate, or the aircraft has moved since reporting the
last position. The position of the aircraft is then calculated a
number of times using the TDOA measured from a number of
stations forming target`s track, preferably five or more, even
though the clocks of the different stations may be slightly out
of sync.
To calculate an initial position in three dimensions, at least
four reporting stations are required. Or, if altitude is known
(barometric reporting from a Mode-C transponder, for
example) only three reporting stations may be needed to
determine position. In either event, the over-determination
technique calculates aircraft position based on different groups
of receivers within the overall set. Thus, in the simplest
example for a group of four sensors, S1, S2, S3, and S4, the
position of an aircraft or other object may be determined by
using time-stamp signals from different groups of three
sensors. For example, a first position may be determined from
S1, S2, S3 sensors, a second position from S1, S2, S4 sensors,
a third position from S1, S3, S4 sensors, a fourth position from
S2, S3, S4 sensors. Thus, a number of different position values
(and thus track values) can be readily calculated.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 179
By merging this position data together, discrepancies in
reporting station clock signal values can then be calculated, by
determining, iteratively, the clock corrections necessary be
applied to make all the position values corresponding one to
another. The same technique can be used in three dimensions
using five or more sensors, gathered into groups of four. In the
test data presented further, more than 5 sensors were used, with
results plotted relatively to the one sensor S1 (whose plot
would show a straight line).
Figure 1 is a block diagram illustrating the system of the
present invention for calibration of TOA/TDOA measurements
of dimension 4 or higher. The block diagram illustrates the
iterative process (X axis) over time (Y axis). A number of
primary measurements are made, using different groups of
sensors, as previously discussed. These positions are then
converted to Cartesian 3-D coordinates, and when compared to
other such measurements, a clock deviation estimate is made.
From these clock deviation estimates, additional delays or
corrections to clocks are made, using a Kalman filter, and the
process is repeated. With each iteration, the positioning data
becomes more coherent, and the clock values are brought into
sync using the correction factors.
Figure 1. Block diagram illustrating calibration via over-determination of
TOA/TDOA measurements of dimension 4 or higher.
The Primary Measurements are computed as follows: Any
TDOA measurements with dimensions equal or higher than
four may be used in this example. The following processing
can be speed up if the primary measurements are supplemented
by corresponding approximate target`s position in 3D which
can be taken from the prior target tracking process. This will
omit analytical conversion from the three selected TDOA
measurements into Cartesian 3D. Analytical conversion
provides only an approximate position estimate which serves as
initial condition for iterative computation of real 3D position,
exploiting a whole vector of primary measurements.
Conversion from hyperbolic to Cartesian 3D space is made
as follows. An approximate or initial target`s position in 3D
can be computed from primary TDOA measurements via
analytical conversion or can be updated through primary
measurement (dimension 4 or more) associated with an
existing tracked target. This block computes the most probable
target position (from TDOA) in 3D space according to whole
vector of primary measurements and its regular covariance
matrix defining position accuracy. Afterwards, it computes the
difference between the most probable target position in 3D
(estimated in previous steps and converted back to hyperbolic
space) and entering primary TDOA measurements (from whole
set of primary measurement vectors). This difference is
estimate of inferior clock deviations (additional delays) with
singular covariance matrix. The difference, including the
probability matrix, can be computed with an assumption that
dimensions of primary TDOA measurements (in hyperbolic
space) are higher than dimensions of 3D (Cartesian), i.e., 3D
space is subspace of primary measurement (hyperbolic) space.
The system is thus over-determined (redundant).
The Clock Deviation Estimate is calculated as follows.
Coalescence of one-minute consequent differences, mentioned
above, between measured and the most probable target`s
position in hyperbolic space are made, i.e., the corrections of
the clocks` additional delays are estimated. Every measurement
may vary in dimension and it is necessary to have one unified
measurement, to which all measurements are transformed. All
differences are accumulated into one multidimensional value
with Gaussian distribution Additional delay errors (reference
clocks` deviations at stations) are assumed to be constant
during this one-minute period. Note that coalescent estimates
of differences may comprise the final calibration output in
some applications, depending on the desired accuracy level and
the amount of initial discrepancy in the clock values among
stations.
Correction of Additional Delays and the Kalman filter
operate as follows. Update (or correction) of additional delays
(reference clock errors) uses Kalman filter feature, which
allows the system to model the error trend of an additional
delay (error of clock synchronization). Clock errors are then
filtered.
These delay values may be applied to the sensor clock
values (time stamps) as received at a central station, where
time-stamped data values are received from a number of
stations to perform TDOA calculations to determine aircraft
position. Alternatively, these time delay values may be
transmitted to sensor stations to correct or update sensor clock
signals. In either event, the position signals may be continually
over-determined in order to verify that the clock signals remain
accurate over time and to update or revise the clock correction
factors.
In this manner, a WAM system can be kept in operation
even without a centrally synchronized clock signal, though the
means of over-determination techniques. Such a system and
method is important, as modern aircraft tracking systems are
switching over to an ADS-B type system, where GPS position
signals are self-reported by aircraft to determine aircraft
positions in a controlled airspace. If GPS signals go offline or
180 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
are jammed, position of aircraft would be difficult to
determine, unless legacy radar systems were still employed.
Wide Area Multilateration offers a backup solution to ADS-B,
but may also be dependent on a GPS clock signal for
synchronization purposes. The presented principle allows an
WAM system to operate in the absence of a central
synchronized clock signal, thus providing an effective
redundant backup to ADS-B systems.
III. MEASUREMENTS
The real traffic from two WAM systems, namely Brno and
Ostrava, were taken for the purpose of synchronization
algorithm design and evaluation. The Brno as well as Ostrava
system includes 7 stations. The synchronization method was
then evaluated on basis of a Kalman filter which estimates
clock deviations from over-determined TDOA measurements
at stations 2-7 from the reference clock at station 1. For
evaluation purposes, a random number generator generates
'real mutual deviations oI clocks on each station which tends
to slowly diverge without any synchronization, which
corresponds to the real clock behavior. The Kalman filter
contains a state vector (current estimate of clock deviations),
including its accuracy (covariance matrix) computed from
measurements. The difference between "real" clock deviation
(from the clock at station 1) and estimated deviations produced
by synchronization measurements, which can be found in the
status vector, is the key result description. That difference
corresponds to systematic errors from primary WAM TDOA
measurements and affects all tracked targets.
Note that stations that are close in position to other stations
are synchronized much more accurately than other pairs. This
effect is caused by the small weight of TDOA coordinates from
the base between stations in the calculation of the target
position, when a relatively large change in 3D position of the
target causes only a small change in TDOA coordinates.
Fig. 2 is a series of graphs illustrating systematic errors of
TDOA measurements on sensors 2-7, relative to sensor 1 (not
shown) of WAM Brno during the worst time period for
synchronization (e.g., least amount of traffic) estimated by the
over-determination method, where the Y-axis represents clock
deviation in nanoseconds, and the X-axis represents time in
minutes. The graphs of fig. 2 illustrate the systematic errors of
primary WAM TDOA measurements where clock
synchronization on each station is based only on real traffic
conditions. The graph elements represent progress of
systematic errors and the continuous solid portion "lock up"
corresponds to the theoretical limit of 99% percentile. These
errors can be interpreted as error for a fixed station after
application of the over-determination method. Thus, output
data from a fixed station will be affected by this error during
given time interval.
Figure 2. Systematic errors of over-determined TDOA measurements on
each of WAM Brno stations (related to 1. reference station which isn`t
shown).
Referring to fig. 2, one can see the TDOA measurement
error dependency on traffic density at any given time. The
weak synchronization during night traffic from a worst time
period for synchronization by the over-determination method is
between 600th and 660th minute (from 08/30/2010, 1 AM until
08/30/2010, 2AM). Fig. 3 shows this particular traffic interval.
The dark trajectories represent lower altitude targets (aircraft)
while the lighter trajectories represent higher altitude targets.
Only the lighter (higher altitude) trajectories were used for
sync, as accuracy in obtaining sync from lower level targets is
far more difficult, if not impossible, in some situations.
Figure 3. Traffic during the worst time period for synchronization by over-
determination method (from 30.8.2010, 1am until 30.8.2010, 2am). Only red
trajectories were used for sync.
Fig. 4 is a graph illustrating the target altitudes showed in
fig. 3, over time, during the worst hour of system operation
during the test, in terms of synchronization. Again, the dark
lines represent lower altitude targets (aircraft) while the lighter
lines represent higher altitude targets. As illustrated in fig. 3,
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 181
only targets with an altitude of 4800 meters and above were
used for generating sync using over-determination.
Figure 4. Targets` altitudes during the worst time period Ior synchronization
by over-determination method (from 30.8.2010, 1am until 30.8.2010, 2am).
Only red trajectories above 4800m were used for sync.
Above described synchronization by over-determination for
Brno WAM system was performed also for related Ostrava
WAM system. Traffic situation was recorded on both systems
simultaneously thus multiple targets appeared in both systems
during their over flight. The Ostrava system test was chosen
due to different siting and traffic conditions. It provides also
possibility to perform test on artificial joint system Morava
(partly overlapping Brno with Ostrava system) which results
will be showed further.
Fig. 5 presents systematic errors in ns of over-determined
TDOA measurements for Ostrava system in similar manner as
fig.2 for Brno system. The increasing error trend can be seen
again during night hours` week traIIic. The 99% error 'lock
up curve is not higher than 10 ns which comply with MLAT
requirements.
Figure 5. Systematic errors of over-determined TDOA measurements on
each of WAM Ostrava stations (related to 1. reference station which isn`t
shown).
As previosly mentioned, the test data recorded for both
systems can be used to simulate traffic in joint, wider, WAM
system called Morava. Morava name is chosen on base of
geographical relation between these two regions and their
location in Moravian part of Czech Republic, see fig.6.
Figure 6. Situation map of WAM systems Brno and Ostrava.
Fig. 7 represents graphs illustrating systematic errors of
TDOA measurements on all Brno and Ostrava stations 2-14,
relative to station 1 of WAM Brno. The simulation length was
shortened to 7 hours according to high amount of input data
which are representing again the worst time period for
synchronization (night traffic). The progress of systematic
errors increases with decreasing air traffic and vice versa
decreases with increasing traffic. The main difference in
comparison with stand alone systems is in maximal value of
99 percentile which doesn`t exceed 2 ns even during the
weakest traffic conditions. This result implies previously
mentioned dependency between synchronization error and
number of sparsely separated stations.
Figure 7. Systematic errors of over-determined TDOA measurements on
each of WAM Morava stations, i.e. all Brno and Ostrava stations (related to 1.
reference station which isn`t shown).
182 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Figure 8. Traffic interval with common targets for Ostrava and Brno system.
Only red trajectories were used for sync.
IV. CONCLUSIONS
It was presented and proved on real data that any Wide
Area Multilateration system can by synchronized by over-
determination method in situation when any kind of clock
reference is lost (GPS, beacon, Mode-C). It is required that
such a system will satisfy basic prerequisites of sparse sensor
distribution and amount of at least five sensor. Test on two
WAM systems, namely Brno and Ostrava proved capability of
synchronization from targets when 99% of all synchronization
error, deviations from reference station, didn`t exceed 10 ns for
system with 7 stations. This property is moreover valid for low
air traffic during night. Once these two systems are artificially
merged into one system, the 99 % percentile margin is even
lower and doesn`t exceed 2 ns.
Any multilateration system can provide by above presented
synchronization method from flying targets sufficient
performance in case of time reference lost for necessary
amount of time during which all airplanes can land safely and
any additional surveillance system can be involved.
REFERENCES
[1] M. Pelant, 'Estimate of Systematic Errors of A Passive Surveillance
System (PSS) According to Emitters with Unknown Position, at
German Radar Symposium GRS 2002, Bonn, Germany, 03-04
September 2002.
[2] M. Pelant et al, Provisional U.S. Patent Ser. No. 61/474,350 for: "Time
Synchronization via Over-determined Measurements".
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 183
(1)
where c is the velocity of light, (x,y,z) the unknown target
position (aircraft position) and (x
i
,y
i
,z
i
) is the known position of
the ith station (i=1 denotes the reference station). Linearizing
(1) by Taylor series expansion [11-12] is the most accepted
strategy to solve an inverse problem with the hyperbolic
equations, in order to estimate the target position. In the current
literature, the solution of this inverse problem has been
presented as an iterative procedure in the sense of the Least-
Squares (LS) [11-12]. Denoting the unknown target position as
ground stations) in a
vector
(2)
where ; is the
Jacobian matrix of
the
and
is a
. Finally,
because this method is based on an iterative procedure, K is the
number of refinement iterations.
The solution provided by (2) is the minimum residual norm
solution and the matrices product
is known as the
pseudoinverse matrix
(3)
where is the exact coefficient matrix for the inverse problem,
is called regularization parameter and is called
regularization matrix. The regularization parameter controls
the importance given to the regularization term
.
Using the Tikhonov regularization concept, the likelihood
function [12] for the Mode-S location problem can be
expressed as follows,
(4)
where is the covariance matrix of the TDOA
measurements noise and det denotes the determinant operator.
The maximum likelihood solution of (4) is that
which
minimizes the following function,
(5)
Solving (5) by Taylor-series expansion, the estimation for
the unknown target position in the Tikhonov sense takes the
following form,
(6)
where
(7)
It is worth to say that, due to the fact that the covariance
matrix , for real applications, is often not known because
it depends on the true target position, in practice it is common
remove it from (7), assuming an identity matrix.
The choice of regularization parameter and regularization
matrix is the most critical aspect to make a correct use of the
procedure described above. Firstly, the choice of the
regularization matrix is directly connected with the statistics of
the target position vector . If the components of are
assumed to be non-random and uncorrelated, a standard choice
of the regularization matrix is
, where
is a
identity matrix.
On the other hand, the choice of the regularization
parameter value is not as straightforward as the choice of
regularization matrix. In the literature there exist a considerable
number of methods and procedures to calculate/estimate an
approximated regularization parameter value. These methods
provide good results for a variety of applications (e.g. image
processing, biologic computer, remote sensing, electromagnetic
scattering, etc.) and they are basically based on the solution of
an optimization problem, i.e., find a parameter that satisfies
some equalities [15] or find a parameter that minimizes some
special functions [16-18]. However, it is worth to say that, due
to nature of these methods, they introduce a significant
computational load and therefore the computation time can be
not acceptable for real time location in Mode-S Multilateration.
In this work, we evaluate the problem for several
regularization parameters values (no more than three) and then
we choose as true solution the one which corresponds with the
minimum residual error. This option is feasible for this
application because the typical size of the coefficients matrices
(Jacobian matrix) is normally smaller than .
In general, the residual error for an inverse problem is
given by,
186 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Figure 1. Linate airport system layout.
Figure 2. Horizontal accuracy for Linate airport. Each point in the
abscissa corresponds to a point in the simulated path.
(8)
Remembering that for Taylor-series expansion method, the
matrix is an approximation of an exact coefficient matrix,
then (8) could not be a correct value for the residual error
regarding to the true target position . Therefore, in this work,
we propose to calculate the residual error by replacing the
regularized solution
(9)
where the vector
is given by,
(10)
IV. RESULTS
Preliminary results are shown to validate the improvement
of the system accuracy and its convergence by applying the
Tikhonov method in the iterative procedure of Taylor-series
expansion. Two scenarios have been simulated; the first one is
the operating system of Linate Airport (Milan, Italy) and the
second one is a multilateration system which is well described
and studied by Cramr-Rao Lower Bound -CRLB- analysis in
[6].
For each scenario, the horizontal (2D) R.M.S error
(obtained via Monte-Carlo simulation with 100 trials), the
theoretical accuracy provided by the CRLB [6], the bias of the
estimator and the spatial convergence are calculated.
A. Linate Airport System
The Linate airport system is composed of eight ground
stations. For this scenario we have simulated a path of surface
movement around the airport. The system layout and simulated
path are shown in Fig. 1.
For this scenario, the starting point for the Taylor-series
expansion method has been assumed to be a fixed point over
the airport and it is shown as the star in Fig. 1. For this scenario
it has been found that using only one regularization parameter
value ( ) is enough to obtain satisfactory results.
Fig. 2 shows the horizontal R.M.S error for the horizontal
projection of Taylor-series expansion method and the non-
projected (3D) version solved by the pseudoinverse matrix. It
also shows the non-projected (full version) Taylor-series
expansion method solved by Tikhonov regularization and the
corresponding CRLB analysis.
Initially, the CRLB analysis predicts a good accuracy over
the entire path, presenting only a few peaks around the points
40 and 50, where the horizontal accuracy is slightly larger than
7 meters. However, for the non-projected Taylor (circles), it
can be seen how the ill-conditioned problem avoids the
convergence of the method solved by the pseudoinverse
matrix, i.e., the R.M.S error tends to infinity in the most of the
points. On the other hand, the horizontal projected version
obtains acceptable accuracy levels but the effect of the spatial
bias is present, for this scenario, in most of the points (more for
those points within the 30 and 120). Finally, it is evident how
the solution obtained by applying Tikhonov regularization
improves both the ill-conditioned problem, which is directly
related with the system accuracy and convergence and the
spatial bias added for the projected version.
Fig. 3 shows the bias of the estimator for the projected
version of Taylor as solved by the pseudoinverse and that one
corresponding to the full version of Taylor as solved by the
Tikhonov regularization. In this figure it can be noted the
improvement, regarding to the spatial bias of the horizontal
projection of Taylor-series method, added by using the
Tikhonov regularization. This aspect is very important when
using tracking algorithms (which are present in all the Air
Traffic Control -ATC- systems) because they can improve the
R.M.S error of the location algorithm but not the bias. In this
way, it is clear to see how Tikhonov method also helps to the
tracking algorithms to reach more accurate tracks.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 187
Figure 3. Horizontal bias of the estimator for Linate airport. Each point
in the abscissa corresponds to a point in the simulated path.
Figure 4. Spatial convergence for one trial.
Figure 5. Layout of the MLAT system for a takeoff line. Figure 6. Horizontal accuracy over the takeoff line.
Finally, Fig. 4 shows the spatial convergence for a specific
Monte-Carlo trial. In this figure it can be observed how the
solution by Tikhonov regularization allows the Taylor-series
expansion to ensure the convergence to the true point.
B. MLAT System for a Takeoff Line
This system is composed for four stations and it is well
analyzed in [6]. The layout of the simulated scenario is shown
in Fig. 5.
For this scenario, the starting point for the Taylor-series
expansion method has been obtained by means of the closed
form algorithm described in [19]. This algorithm is based on
spherical intersections and it does not need a starting point but,
as it is shown in the results, it is also affected for the ill-
conditioning of the problem due to the system geometry. The
horizontal coordinates of the starting point (x,y) are taken from
the closed form algorithm and the vertical coordinate (z) is
simulated as the barometric altitude, i.e., with a bias of 40m
regarding to the real target height. Also for this scenario it has
been found that only using one regularization parameter value
( ) is enough to obtain satisfactory results.
The amount of ill-conditioning of this scenario is
significantly greater than that of the first scenario. It is because
the number of stations here (four) is much smaller than the first
one (eight). This effect can be noted in the CRLB analysis
shown in Fig. 6 since the theoretical accuracy diverges for
points within 0 and 5 km and for those around 20 km. On the
other hand, due to the fact that for this scenario, the target
height is increasing with the distance, the vertical separation of
this with the plane of the ground stations considerably affect
the accuracy provided by the horizontal projection of Taylor-
series method (crosses) and the spatial bias added by this is
considerably large for points beyond 15 km.
Due to the ill-conditioning, it can be observed that, for this
scenario, the accuracy levels provided by the full version of
Taylor, using the pseudoinverse matrix, diverges far from the
theoretical accuracy values (CRLB) for points within 0-5 km
and 15-20 km. In contrast, the closed form algorithm presents a
more stable accuracy but it is also affected by system geometry
(Dilution of Precision -DOP-). Finally, it is evident the
significant improvement, of the system accuracy, obtained by
applying Tikhonov regularization. The accuracy for this option
is stable for the whole of takeoff line and it is not larger than 25
188 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Figure 7. Horizontal bias of the estimator over the takeoff line. Figure 8. Spatial convergence for one trial over the takeoff line.
meters. It is worth to say that this solution is below the CRLB
values because also the CRLB is affected by the ill-
conditioning of the problem, specifically that part due to the
system geometry.
Fig. 7 shows the bias estimator for the solutions obtained
by the closed form algorithm and by the full version of Taylor-
series using both pseudoinverse matrix and Tikhonov
regularization. Firstly, it can be noted that for a few points
close to 10 km and 15 km, the bias of the solution obtained by
pseudoinverse is smaller (nor more than 1 m) than that of
Tikhonov method. It can be explained because in the case of
well-conditioned problems the pseudoinverse matrix is the
solution with minimum norm [3] and in contrast Tikhonov
always adds certain amount of bias [13]. The important aspect
is that, if the correct regularization parameter value is chosen,
this amount of bias can be neglected regarding to the rest of the
options to improve the problem (i.e., the horizontal projection
of Taylor-series method). Moreover, due to the ill-conditioned
problem, for the rest of the points, the bias added by
pseudoinverse matrix solution is infinity and for most of the
points the bias added by the closed form algorithm has been
found greater than that of Tikhonov regularization.
Finally, Fig. 8 shows the spatial convergence for a specific
Monte-Carlo trial; in this figure it can be noted how the
regularization of the location problem ensures the convergence
also for this scenario.
V. CONCLUSIONS
The implementation of Tikhonov regularization to solve the
inverse problem of Taylor-series expansion, for location in
Multilateration systems, has been described and evaluated. The
theoretical aspects of the method with a practical strategy to
calculate the regularization parameter have been described.
For the scenarios simulated here, significant improvements,
for the system accuracy and convergence, have been found
with the implementation of Tikhonov regularization. For both
scenarios, it was found that the regularization of the location
problem significantly mitigates the ill-conditioning due to the
system geometry, i.e., those points where the CRLB analysis
predicts poor accuracy levels; to the measurements noise, i.e.,
those points where the CRLB predicts good accuracy levels but
the solution obtained by the pseudoinverse matrix diverges;
and also due to the quality of the starting point for Taylor-
series expansion method.
For both scenarios it was found that a regularization
parameter value of was enough to obtain satisfactory
results but, it is worth to say that in the situations where the
problem is better conditioned, it is necessary to use, at least,
one or two more values smaller than , i.e., the smaller
the amount of ill-conditioning the smaller should be .
The regularization of the location problem is more useful
for those situations where the vertical separation between the
ground stations and the target is quite small or for those
situations with a small number of stations.
ACKNOWLEDGMENT
The authors are grateful to Thales Italia S.p.A. (Eng. R.
Scaroni) who supplied the geometry of the Multilateration
system in Linate (Milan, Italy) airport.
REFERENCES
[1] "The ATM Surveillance Strategy for ECAC," Eurocontrol,2008.
[2] "Minimum Operational Performance Specification for Mode S
Multilateration Systems for Use in Advanced Surface Movement
Guidance and Control Systems (A-SMGCS)," EUROCAE, 2003.
[3] G. H. Golub and C. F. V. Loan, Matrix Computations, Third ed.
Baltimore: the Johns Hopkins University Press, 1996.
[4] J. Hadamard, Lectures on Cauchy's Problem in Linear Partial
Differential Equations. New Haven: Yale University Press, 1923.
[5] M. Leonardi, A. Mathias, and G. Galati, "Two Efficient Localization
Algorithms for Multilateration," International Journal of Microwave
and Wireless Technologies, vol. 1, pp. 223-229, 2009.
[6] G. Galati, M. Leonardi, and M. Tosti, "Multilateration (Local and Wide
area) as a distributed sensor system: Lower bounds of accuracy," in
European Radar Conference, EuRAD, Amsterdam, 30-31 Oct. 2008, pp.
196-199.
[7] M. Bertero, P. Boccacci, G. J. Brakenhoff, F. Malfanti, and H. T. M. v.
d. Voort, "Three-dimensional image restoration and super-resolution in
flourescence confocal microscopy," Journal of Microscopy, vol. 157, pp.
3-20, 1990.
[8] R. F. Harrington, Field computations by moment methods. New York:
Macmillan, 1993.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 189
[9] W. Menke, Geophysical data analysis: discrete inverse theory. San
Diego: Academic Press, 1989.
[10] A. Kawalec, M. A. Kojdecki, and B. Wajszczyk, "Regularised iterative
estimation of emitter position for passive localisation systems,"
presented at the International Conference on Microwaves, Radar &
Wireless Communications, MIKON 2006, 22-24 May, 2006.
[11] W. H. Foy, "Position-Location Solution by Taylor-Series Estimation,"
IEEE Transactions on Aerospace and Electronic System, vol. AES-12,
pp. 187-194, March 1976.
[12] D. J. Torrieri, "Statistical Theory of Passive Location Systems," IEEE
Transactions on Aerospace and Electronic System, vol. AES-20, pp.
183-198, March 1984.
[13] A. N. Tikhonov, "Solution of incorrectly formulated problems and the
regularization method," Sovieth Math. Dokl., vol. 4, pp. 1035-1038,
1963.
[14] D. L. Phillips, "A technique for the numerical solution of certain integral
equations of the first kind," Journal of the ACM, vol. 9, pp. 84-97, 1962.
[15] V. A. Morozov, "On the solution of functional equations by method of
regularization," Sovieth Math. Dokl., vol. 7, pp. 414-417, 1966.
[16] H. Gfrerer, "An a posteriori parameter choice for ordinary and iterated
Tikhonov regularization of ill-posed problems leading to optimal
convergences rates," Math. Comp., vol. 49, pp. 507-522, 1987.
[17] M. Hanke and T. Raus, "A general heuristic for choosing the
regularization parameter in ill-posed problems," SIAM J. Sci. Comput.,
vol. 17, pp. 956-972, 1996.
[18] G. H. Golub, M. T. Heath, and G. Wahba, "Generalized cross-validation
as a method for choosing a good ridge parameter," Technometrics, vol.
21, pp. 215-223, 1979.
[19] H. C. Schau and A. Z. Robinson, "Passive Source Localization
Employing Intersecting Spherical Surfaces from Time-of-Arrival
Differences," IEEE Transactions on Acoustics, Speech, And Signal
Processing, vol. ASSP-35, pp. 1223-1225, August 1987.
190 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Assessing the safety of WAM over
a non-radar surveillance area
James Hanson
Helios
James.Hanson@askhelios.com
Ben Stanley
Helios
Ben.Stanley@askhelios.com
This paper discusses potential issues and solutions for developing
an argument that Wide Area Multilateration (WAM) is acceptably
safe as a sole means of surveillance in normal and failure-based
operations. Using a practical example it discusses techniques to
analyse and characterise the behaviour of a distributed receiver and
interrogator system. It will also investigate the challenges to gather
appropriate evidence, and how they can potentially be overcome.
Keywords: Safety; WAM; multilateration; distributed system;
remote sites
I. INTRODUCTION
Under the standards and guidelines currently promulgated
by ICAO, EASA and EUROCONTROL, a safety case must
accompany the introduction of a new surveillance system such
as WAM (Wide Area Multilateration). This requires a
proactive approach to understanding risk and applying
appropriate controls - i.e. the system being placed into
operation must be proven in theory, and subsequently in
practice, to meet an acceptable level of safety.
It is important that attention is paid to all aspects of the
behaviour of the new WAM system; acceptable safety should
be shown when WAM is working as designed, when it has a
critical failure (e.g. CPS failure), and when it has no critical
failure but, through gradual system degradation, increases risk
(e.g. small corruption of position plotted).
The defining case from a risk-based perspective tends to be
that of undetected corruption; the gradual degradation of
position away from the true value or undetected loss (or failure
lack of initiation) of a track. In other words, it is this system
failure which has the greatest risk, with the combination of
probability of occurrence and severity of credible effects,
taking account of all potential mitigations. Whichever way this
is caused at an individual component (or function) level, it
needs to be identified and mitigated.
A further focus of the analysis can be found in the
calculation of expected failure rates for the provision of
surveillance information based on individual components or
functions. With the distributed nature of WAM, this is not as
straightforward a solution as for other technologies, as it needs
to take account of local geometries (i.e. the level of redundancy
across the airspace at all altitudes under surveillance-based
control) and the availability of communications and power to
each remote site.
This paper therefore discusses potential methods for
carrying out a safety case for the installation and operation of a
Wide Area Multilateration (WAM) system across an area in
which no surveillance has previously been available i.e. a
non-radar airspace (NRA) environment. It is not intended to be
comprehensive, but sets out some thoughts on which it is
hoped others will build.
II. KEY PROPERTIES OF WAM
Wide Area Multilateration is a surveillance technique for
monitoring the position of aircraft with ground-based sensors.
As with Secondary Surveillance Radars (SSR) it is a
cooperative independent technique. The benefits of WAM over
radar are typically the lower cost of the receivers and improved
accuracy.
A series of ground stations receive the same signal from an
aircraft and use the time difference of arrival (TDOA) from
each sensor to determine hyperboloids. When plotted from a
sufficient number of receivers, these hyperboloids intersect to
determine the precise 3D position of the aircraft. This is
transmitted as an ASTERIX message (using Category 20 [2]
for the target reports and Category 19 [1] for status messages).
The position can be further augmented by ranging techniques
in which an approximation is made of the time taken for a
signal to reach an individual ground station. Integrity of the
system can be checked through reference transponders. The
reference transponders transmit from a fixed location known to
the central processing system and can potentially be used to
alert to system inaccuracies.
III. MAKING A SAFETY CLAIM
Before putting into service a WAM system, it is necessary
to demonstrate that the system will be acceptably safe by
design and in operation. This entails the development of a
safety argument that shows how a target level of safety will be
achieved by the new system in operation. In practice this
involves analyzing the system in two conditions:
a) When everything is working as designed; and
b) When failures have occurred.
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 191
The safety claim is made only once it can be shown that the
normal and failed performances of the system are both
within tolerable limits.
In terms of performance, a set of reference values have
been derived by EUROCONTROL, ostensibly to be independent
of the surveillance source (eg the draft Surveillance
Performance and Interoperability Implementing Rule). These
references values are able to relate certain probabilities of
position error (e.g. Y metres at 95% confidence) to the
separation minima through the use of a mixture of
mathematical analysis and experiential evidence. The latter
uses the argument that if radars have been able to provide
position accuracy to a certain level, and can support e.g. 5NM
separation, then that level of position accuracy is acceptably
safe for nominal operations.
The parameters can be simulated and analysed in
operational implementation using tools such as
EUROCONTROLs SASS-C (Surveillance Analysis Support
System for ATC Centre) [3] and its derivative CAPT
(Coverage Analysis and Planning Tool) [4]. Both tools are
evolving to handle radar, ADS-B and Multilateration data.
One of the concerns with this approach is that the
probability of error values for WAM at other points on the
curve (e.g. 98%, 99.5%) may not exactly correspond to the
radar error distribution model. At the time of writing,
EUROCONTROL and others in Europe are currently responding
to this issue through the derivation of surveillance separation
error curves, allowing any value to be determined theoretically
then tested (or monitored) in practice, and giving a greater
degree of confidence in the ability of the WAM system to
support a given separation minima (see ED-161 for an example
of this approach for ADS-B [6]).
It should be noted that in the absence of CAPT (or other
tools), the manufacturers models of coverage should be
extensively validated through use of flight trials and targets of
opportunity. Note the need to look at different flight levels (i.e.
the base of coverage) in this process, and also to consider
carefully the level of redundancy in the distributed sensors. If a
certain performance is required at 2000ft across a wide
coverage area (met by 4 receivers and 2 interrogators viewing
the airspace), typically greater redundancy will be seen at
higher flight levels and therefore the effect of an error in any
one of these remote units is minimised.
The driving hazards in a WAM surveillance system
typically relate to undetected loss or corruption of aircraft
position (assuming the system detects all aircraft within the
service volume when working as designed) which, as with any
surveillance safety case, would be determined as part of a
hazard identification workshop with controllers. Using current
safety methodology [5], the next step in the safety case is to
examine the failures of the end-to-end surveillance system that
could lead to the hazard. This is important in the design and
deployment of WAM, as the failure rates of multiple
components need to be correlated to understand the failure rate
at different points in the service volume. This is not a trivial
task in a non-radar environment where no radars exist to
interrogate aircraft and the WAM system must actively
interrogate to ensure sufficient replies are received to plot the
aircraft position. Over a large area this then means a complex
distribution and geometrical spread of installed receivers and
interrogators, each one having a specified:
failure rate (Mean Time Between Failure);
repair rate (Mean Time To Repair, dependent on not
only the maintenance contracts but also the location in
the country e.g. a WAM receiver station mounted on
an oil rig may take longer to repair in the event of
failure than one nearby a busy aerodrome);
communications availability (from the remote site to
the central processor for WAM);
power, particularly for environments where a link to a
National Grid is not available;
From these parameters, it can be seen that an understanding
of the failure rate for different parts of the airspace (e.g. low
altitude, high altitude) can assist in determining the
communications and power performance to be required, which
can be especially important when these are delivered by third
party suppliers and service level agreements need to be signed.
The communications in particular has the potential to add
unnecessary cost if a broad-brush stringent assumption is made
instead of an analysis based on operational need carried out.
In addition to the varying reliability of ground components
of the system, there is also the aircraft to take into account. For
example the reliability of the transponder including the
information it is transmitting (with associated human error
rates) as well as its trigger level, and overall mean time
between failures.
IV. AN APPROACH TO DETERMINING EXPECTED FAILURE
RATES FOR A DISTRIBUTED SYSTEM
As well as taking into account the different failure rates of
individual sites it is also necessary to consider the distributed
network of possible sites that could potentially be used in any
position derivation. One approach is to consider the airspace as
a patchwork of different probabilities where the likelihood of
position determination in each patch is determined by:
The number of interrogators in view
The number of receivers in view
The reliability of each receiver/interrogator
Analysing the failure rate of each receiver/interrogator
should take into account the factors identified in the previous
section to yield the different failure rates within each patch.
Mathematical analysis of the combinations of failures within
each patch can then provide a reasonable estimate of the failure
rate of an interrogator and receiver within the patch. Coverage
maps (which, as mentioned earlier, would require validation by
eg flight tests) can then be used to identify the number of
receivers and interrogators available to provide a position (ie
avoid the hazard from occurring).
In the following simplified example, the distribution of
receivers (Rx) and transmitters (Tx) is indicated by the orange
(Rx) and blue (Tx) markers. The distribution is not intended to
192 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
represent a recommended siting of receivers and interrogators,
but solely acts as an illustration of the approach:
Rx
Tx
Figure 1
The simplified receiver coverage pattern might therefore
look as follows, where a deeper colour indicates a greater
overlap between receivers:
Figure 2
The transmitter pattern may conversely look as follows:
Figure 3
Combining these two patterns together provides an overall
coverage within the surveillance volume as follows:
Figure 4
Where the different colours indicate the number of
receivers and transmitters in view:
>3 Rx 2 Tx
>3 Rx 1 Tx
2 Rx 1 Tx
1 Rx 1 Tx
3 Rx 0 Tx
2 Rx 0 Tx
1 Rx 0 Tx
>3 Rx 2 Tx
>3 Rx 1 Tx
2 Rx 1 Tx
1 Rx 1 Tx
3 Rx 0 Tx
2 Rx 0 Tx
1 Rx 0 Tx
Table 1
This can then be used to calculate the probability of
position being lost, corrupted or un-initiated in a particular area
of airspace due there being an insufficient number of receivers
or transmitters. The defined surveillance volume should ensure
that the necessary number of receivers and transmitters are
always in view to support the safety objective. This will
depend upon the availability of each receiver and transmitter
but as the combinations used to derive position are so
numerous it may be better to take either a conservative
assumption and assume that each receivers or transmitters has
the reliability of the most unreliable receiver/transmitter or to
mathematically analyse the various combinations to arrive at
more accurate reliability for receivers/transmitters in that
patch of airspace.
V. VALIDATING THE FAILURE RATE AND PERFORMANCE
Having assessed the expected performance and safety
values of the implemented system from a theoretical
perspective, the next stage is to validate the operational
practice and ensure the safety case remains valid. In particular
this should include flight trials and where possible
confirmation, through monitoring, of the contributing sensors
to position reports.
In addition, integrity monitoring, can be used to provide
confidence during operation as a technical mitigation to any tail
errors in the position distribution. The distribution of WAM
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 193
errors is still being defined but may well be different to the
normal distribution assumed for radar (see below)
99.9% 99.9%
Figure 5
Integrity monitoring allows detection of the tails of this
type of distribution using reference transponders or additional
receivers (when in view), with the resultant mitigation that the
hazard is detected and leads to a lower severity effect. Four
receivers are needed to detect position and additional receivers
will provide further integrity checks, that when combined with
a means of monitoring the contributing sensors can ensure
detection and potential elimination of erroneous receivers.
Monitoring of the ASTERIX Category 19 status messages
provide further insight and can usefully supplement integrity
monitoring techniques.
VI. OPERATIONAL AND BUSINESS IMPLICATIONS
From both an operational and business perspective it is
important to explore the safety case fully. An overly redundant
system means extra cost with little added benefit. Insufficient
redundancy on the other hand may lead to a reduced
surveillance volume due to issues at low level or at the
boundaries of the volume. The usual response to the latter issue
is a need to re-procure additional receivers or interrogators,
with an obvious impact on the investment case.
In the event that receivers/interrogators do fail, the
continuity of operations is often an important consideration,
particularly for a WAM system placed over a wide area with
centralised processing. One mitigation could be to sectorise the
airspace taking into account the coverage of the receivers. This
may then limit the impact of that failure to a smaller area. For
example the base altitude of a sector might be defined at a
particular height beyond which there is increased redundancy.
Airspace below that sector could revert to uncontrolled
airspace in the event of a failure and the controller might only
use surveillance below that sector to detect airspace
infringements, whereas there would still be sufficient
redundancy in the sector above to support a surveillance
service. It is also worth considering the traffic volumes in each
sector as lower volume sectors could be allocated a different
proportion of the overall safety objective (depending on
whether or not traffic volumes have been taken into account in
determining hazard severity, in which case they mustnt be
double-counted as a mitigation)
VII. CONCLUSION
A safety case must accompany the introduction of WAM to
prove in theory, and subsequently in practice, that it meets an
acceptable level of safety. As shown above, it also has a role in
enabling the best possible trade-off between the operational
requirements (particularly integrity and continuity) and the cost
(redundancy). The defining hazard tends to be that of
undetected corruption or undetected loss (or lack of initiation)
and it needs to be identified and mitigated.
With the distributed nature of WAM, determining expected
failure rates is not straightforward and the number of
interrogators/receivers in view as well as their reliability should
be considered. This will be helped by tools being developed by
EUROCONTROL, along with the manufacturer models. However,
these do not currently include a full mathematical treatment of
the expected failure rates taking account of all aspects.
This paper is not intended to be comprehensive, but
nevertheless sets out some thoughts on potential methods for
carrying out a safety case for the installation and operation of
WAM in a NRA environment. Particular consideration is given
to airspace as a patchwork of failure rates that should be
validated and monitored. The implications of such analyses on
operational and business aspects are also considered.
VIII. REFERENCES
[1] EUROCONTROL STANDARD DOCUMENT FOR SURVEILLANCE
DATA EXCHANGE, Part 18 : Category 019, Multilateration System
Status Messages. Edition 1.3, December 2010
[2] EUROCONTROL STANDARD DOCUMENT FOR SURVEILLANCE
DATA EXCHANGE Part 14 : Category 020, Multilateration Target
Reports, Edition 1.8, December 2010
[3] http://www.eurocontrol.int/sass/public/subsite_homepage/homepage.htm
l
[4] http://www.eurocontrol.int/sass/public/standard_page/CAPT.html
[5] http://www.eurocontrol.int/safety/public/standard_page/samtf.html
[6] ED-161 Safety, Performance and Interoperability Requirements
Document for ADS-B-RAD Application, September 2009
194 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
ImplementationofADSBSystems
BenefitsandConsiderations
Session8.1page197
InvestigationofMeasurementCharacteristics
ofMLAT/WAMandADSB
Session8.2page203
RealTimePerformanceMonitoringandNoise
AnalysisinanoperationalWAMSystem
Session8.3page207
ADSBviaIridiumNEXTsatellites
Session8.4page213
IndependentSurveillanceBroadcast
ADSBReceiverswithDOAEstimation
Session8.5page219
180
45
90
135
Nuremberg
LHFT Radar
Ingolstadt
Augsburg
Munich
Regensburg
Pilsen
Latitude
L
o
n
g
i
t
u
d
e
Reference pos.
Sensor pos.
Sensor normal
47.5 48 48.5 49 49.5 50 50.5 51
8
9
10
11
12
13
14
(a) ADS-B reference
10
10
20
20
30
30
40
40
50
50 km
45
90
135
Radar
Eltersh.
W urzburg
Latitude
L
o
n
g
i
t
u
d
e
Reference pos.
Sensor pos.
Sensor norm.
49 49.1 49.2 49.3 49.4 49.5 49.6 49.7 49.8
8.8
9
9.2
9.4
9.6
9.8
10
10.2
10.4
10.6
10.8
(b) Cessna F172 track
Fig. 4. Reference positions of the two measurement campaigns
distance of up to 200 km were received.
A small selected part of 33 telegrams (Airborne Position Code
TYPE 13) is singled out and applied to calibrate the
DOA sensor in the angular range from 45
to 135
based
on Eigenstructure analysis.
The second measurement campaign was carried out at Adolf
W urth Airport, Schw abisch Hall, Germany. To evaluate the
accuracy of the constructed DOA sensor, a dedicated reference
aircraft (see Fig. 5) is used. This measurement aircraft is
owned and operated by the Institute of Flight Guidance, TU
Braunschweig. It is equipped with a Novatel OEM IV receiver
and an iMAR iVRU-FC Inertial Measurement Unit (IMU).
In the rst step of post processing the recorded GPS data is
improved by differential corrections. In a following step the
GPS data and the data provided by the inertial measurement
unit are merged, taking the lever arms between GPS antenna
220 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
Fig. 5. Measurement aircraft (TU Braunschweig): Cessna F172 Skyhawk
TYPE
R
M
S
E
[
10 12 14 16 18
10
1
10
0
10
1
Fig. 6. DOA estimation RMSE versus TYPE-number
and IMU into account. The resulting reference track provides
an accuracy in the sub-meter range.
The gathered reference positions shown in Fig. 4(b) are not
ideally distributed over the angular range and the covered
distance is smaller than with the ADS-B campaign. The ight
track consists mainly of several aerodrome circlings aroundthe
airport and then part of the aircraft track back to Braunschweig
until connection is lost. The comparably small range is due to
the low transponder power (70 W).
Nevertheless, 1170 Mode S telegrams were recorded and
related to the aircraft position based on its time stamps. Again
33 telegrams were used for Eigenstructure calibration in the
angular range from 45
to 135
.
B. Results
The database of recorded ADS-B airborne position tele-
grams were transmitted by a large mixture of different
transponders. Not all of those transponders provide reliable
positioning information. The standard for ADS-B [7] denes
a number of reliability classes encoded in the airborne position
TYPE code. With each of those classes a dedicated contain-
ment radius around the transmitted position is dened in which
the real position needs to be located with a percentage of 95%.
The database of recorded ADS-B telegrams contains
TYPE [10, 11, 12, 13, 14, 17, 18]. The database is now
sorted for TYPE-numbers and the RMSE of the differences
between reference angles and DOA estimates is plotted versus
the according TYPE in Fig. 6. The RMSE remains about
equal and below 1
. As a
correlation of low accuracy class and DOA RMSE is evident,
those telegrams should not be trusted.
Distance [km]
R
M
S
E
[
All TYPE
Only TYPE 13
50 100 150
10
1
10
0
10
1
Fig. 7. DOA estimation RMSE versus distance with and without TYPE-
number ltering
DOA [
]
E
r
r
o
r
[
|Error|
RMSE
10 45 90 135 170
10
1
10
0
10
1
(a) Error vs. DOA
Error [
]
N
u
m
b
e
r
o
f
e
s
t
i
m
a
t
e
s
-10 -5 0 5 10
0
100
200
300
400
500
600
(b) Histogram
Fig. 8. DOA difference of ADS-B reference and DOA estimation sensor
As DOA error according to horizontal positioning error de-
creases with increasing distance, the resulting RMSE should
decrease over distance too. Thus, the database is sorted for
distance by bins of 30 km and the resulting RMSE of each
bin is plotted versus the according distance in Fig. 7. As an-
ticipated a clear decrease of RMSEis observed with increasing
distance when the whole database is regarded. If all telegrams
with TYPE > 13 (a total of 16000 telegrams remain) are
excluded, the decrease in RMSE with increasing distance is
much lower. At a distance of 135 km both curves meet, at an
RMSE of 0.63
.
Using this selected database with TYPE > 13, an error
plot versus impinging angle can be computed like shown in
Fig. 8(a). By restricting the angular range by 45
to 135
, the
histogram shown in Fig. 8(b) visualizes the according error
distribution. The RMSE in this angular range computes to
0.9
to 135
by evaluating
616 reference positions in this range.
V. CONCLUSION
To show the potential of ADS-B receivers with DOA esti-
mation function, two measurement campaigns were conducted.
By evaluating a large number of ADS-B airborne positions
and comparing the resulting reference DOAs with estimated
DOAs, a clear correlation between accuracy class and DOA
error was found. Depending on the aircraft distance, a cross
check between DOA estimate and ADS-B position can thus
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 221
DOA [
]
E
r
r
o
r
[
|Error|
RMSE
10 45 90 135 170
10
1
10
0
10
1
(a) Error vs. DOA
Error [
]
N
u
m
b
e
r
o
f
e
s
t
i
m
a
t
e
s
-10 -5 0 5 10
0
10
20
30
40
50
60
(b) Histogram
Fig. 9. DOA difference of differential GPS reference and DOA estimation
sensor
gather valuable information regarding ADS-B reliability.
By a dedicated reference aircraft the potential of the con-
structed DOA sensor to resolve DOA down to a RMSE of
0.66
BalbastreTejedorV.Juan
BarbarescoFrdric
BarsheshatAbraham
BaudOlivier
BenvenutiDario
BerizziFabrizio
BernabMarcello
BernardosM.Ana
BesadaA.Juan
BonamenteMarco
BosserLuc
BredemeyerJochen
UniversidadPolitcnicadeValencia,Spain
THALESAirSystems,France
SENSISCorporation,USA
THALES,France
ELETTRONICA,Italy
UniversityofPisa,Italy
SELEXGalileoS.p.A.,Italy
UniversidadPolitcnicadeMadrid,Spain
GPDS,UniversidadPolitcnicadeMadrid,Spain
D'AppoloniaS.p.A.,Italy
THALESSystmesAroports,France
FCSFlightCalibrationServiceGmbH,Germany
167, 185, 259
27
197
265
89, 107
73, 129
73
247
173, 247
111
89, 101
141
UniversidadPolitcnicadeMadrid,Spain
KoreaAerospaceResearchInstit.,RepublicofKorea
UniversityofPisa,Italy
CNIT,Italy
247
253
129
73
DalleMeseEnzo
delosReyesDavElas
deMiguelGonzalo
DiGirolamoSergio
DinoiLibero
DonisiDomenico
DucaGabriella
UniversityofPisa,Italy
UniversidadPolitcnicadeValencia,Spain
UniversidadPolitcnicadeMadrid,Spain
THALESAleniaSpaceItaly
ELETTRONICA,Italy
D'AppoloniaS.p.A.,Italy
SICTA,Italy
129
167, 185, 259
173, 247
213
231
111
2
G
GalatiGaspare TorVergataUniversity,Italy 147, 153, 167, 185
GarzelliAndrea
GmezPrezJ.Emilio
GrandinJeanFranois
GuerrieroMarco
UniversityofSiena,Italy
Univers.PolitcnicadeValencia,Spain
THALESSystmesAroports,France
ELETTRONICA,Italy
73
259
89, 101
231
HaddonDavid
HansonJames
HantscherSebastian
HeidgerRalf
HelmStefanie
HessMatthias
HonorNicolas
HultmanPeter
EASA,Germany
Helios,UnitedKingdom
FraunhoferFHR,Germany
DFSDeutscheFlugsicherungGmbH,Germany
GermanAerospaceCenter(DLR),Germany
DeutscheFlugsicherungGmbH(DFS),Germany
THALES,France
SAABElectronicDefenceSystems,Sweden
47
191
65
141, 203
271
141, 241
265
89
JaschAlexander
JohanssonAnders
JugePhilippe
TUBraunschweig,Germany
SwedishDefenceResearchAgency,Sweden
THALESAirSystems,France
219
89, 101
27
LangStefan
LawrenceE.Peter
LeeIlhyung
LeonardiMauro
LorenzP.Frank
LuedtkeG.
LupidiAlberto
LyulSongTaek
FraunhoferFHR,Germany
THALES,France
AdvancedInst.ofScienceandTech.,Korea
TorVergataUniversity,Italy
FraunhoferFHR,Germany
FraunhoferFKIE,Germany
UniversityofPisa,Italy
HanyangUniversity,RepublicofKorea
61
265
253
147, 153, 167, 185
65
65
73
133
MalanowskiMateusz
MantillaGaviriaA.Ivan
MartorellaMarco
MathiasAdolf
MerinoPedro
MisiurewiczJacek
MiyazakiHiromi
MoneuseJeanFranois
MoscardiniChristian
MuickiDarko
WarsawUniversityofTechnology,Poland
UniversidadPolitcnicadeValencia,Spain
UniversityofPisa,Italy
DeutscheFlugsicherungGmbH(DFS),Germany
INDRASistemasS.A.,Spain
WarsawUniversityofTechnology,Poland
ElectronicNavigationResearchInstitute,Japan
THALESAirSystems,France
UniversityofPisa,Italy
HanyangUniversity,RepublicofKorea
119
167, 185, 259
129
203, 241
13
119
237
27
73
133
NeufeldtHolger
NiheiShirou
NoschesePaolo
THALESAirSystemsGmbH,Germany
ElectronicNavigationResearchInstitute,Japan
THALESAleniaSpaceItaly
207
237
213
CIRA,Italy
111
R
ReckChristoph
RekkasChristos
ReutherS.Max
RicciYves
RuizMojicaF.Ruy
UniversityofErlangenNuremberg,Germany
EUROCONTROLBelgium
UniversityofErlangenNuremberg,Germany
THALESAirSystems,France
UniversidadPolitcnicadeValencia,Spain
219
35
219
27
259
S
SainiLuca
SamantaSoumem
SamczyskiPiotr
SchikoraMarek
SchmidtLorenzP.
SchneiderJeanYves
SchrderMartin
ShimSangwook
SmolarczykMaciej
SommerRainer
SotoAndrs
SouamiHakim
SpinelliSilvio
StanleyBen
StejskalVojtch
StraubStephen
THALES, Italy
NationalInstituteofTechnology,India
WarsawUniversityofTechnology,Poland
FraunhoferFKIE,Germany
UniversityofErlangenNuremberg,Germany
THALESAirSystems,France
FraunhoferFHR,Germany
AdvancedInst.ofScienceandTech.,Korea
TelecommunicationsResearchInstitute,Poland
FraunhoferFHR,Germany
INDRASistemasS.A.,Spain
THALESAirSystems,France
TorVergataUniversity,Italy
Helios,UnitedKingdom
ERABeyondRadar,CzechRepublic
DeutscheFlugsicherungGmbH(DFS),Germany
265
147
119
65
219
27
61, 65
253
119
61
13
153
191
123, 179
9
280 Proceedings of ESAV'11 - September 12 - 14 Capri, Italy
T
TatarinovN.Victor
TatarinovV.Sergey
TavernisePierpaolo
TengstrandGran
ThomasPaul
TigrekR.Firat
TomaselloFilippo
TomskUniv.ofControlSyst.andRad.Russian
TomskUniv.ofControlSyst.andRad.Russian
THALESAleniaSpace, Italy
SAABElectronicDefenceSystems,Sweden
BristowHelicopters,UnitedKingdom
DelftUniversityofTechnology,TheNetherlands
EASA,Italy
79, 83
79, 83
19
89, 101
53
69
47
U
UedaEisuke ElectronicNavigationResearchInstitute,Japan 237
V
VaccaroClaudio SICTA,Italy 2
ValleJorge
vanGenderenPiet
VassilevBoris
VassilevaBoriana
VertuaCarlo
VojekMartin
INDRASistemasS.A.,Spain
DelftUniversityofTechnology,TheNetherlands
TechnicalUniversityofSofia,Bulgaria
Inst.forInformationandCom.Tech,Bulgaria
THALES, Italy
ERABeyondRadar,CzechRepublic
13
69, 79, 83
225
225
265
123
W
WangZongbo
WarokPaul
WildKlaus
DelftUniversityofTechnology,TheNetherlands
FraunhoferFHR,Germany
FraunhoferFKIE,Germany
69
61, 65
61, 65
Y
YooChangsun KoreaAerospaceResearchInstitute,Rep.Korea 253
Z
ZimmermannRuediger FraunhoferFHR,Germany 61, 65
Proceedings of ESAV'11 - September 12 - 14 Capri, Italy 281