by Abhishek Arunkumar Kulkarni A Thesis submitted to the Graduate Faculty of Auburn University ...

89 Pages·2016·1.05 MB·English

SELF-LOCALIZATION OF TARGET NODES USING OPPORTUNISTIC
COMMUNICATION WITH REFERENCE NODES, STATISTICAL
TIME-OF-ARRIVAL, GRID METHOD, LINEAR MATRIX INEQUALITY AND
CENTER-OF-GRAVITY
by
AbhishekArunkumarKulkarni
AThesissubmittedtotheGraduateFacultyof
AuburnUniversity
inPartialFulfillmentofthe
RequirementsfortheDegreeof
MasterofScience
Auburn,Alabama
May7,2016
Approvedby
AlvinS.Lim,Co-Chair,ProfessorofComputerScienceandSoftwareEngineering
ShiwenMao,Co-Chair,ProfessorofElectricalandComputerEngineering
ThaddeusRoppel,AssociateProfessorofElectricalandComputerEngineering
ABSTRACT
Self-localizationofmobilenodesisanimportantproblembecausemanyusefulmobileappli-
cations will become feasible once accurate position information is available. Generally, existing
methods for obtaining accurate location information require either sophisticated hardware (e.g.
Global Positioning System (GPS), Ultra-Wideband (UWB), ultrasounds transceiver) or dedicated
infrastructures (e.g. GSM, WLAN). We address this problem using a different approach that re-
quiresnospecialhardwareorinfrastructuresupport. Themainconceptadaptedtosolvethisprob-
lemis: localizationcanbeperformedandimprovedbymeansofcooperativeandopportunisticdata
exchanges among mobile nodes. Consider a GPS-denied target node with no position information
that communicates opportunistically with a number of in-range mobile peer nodes with some po-
sitioningcapabilities. Thedataexchangesbetweenthetargetnodeandthepeernodeswillthenbe
used by the target node to refine its position estimation using a combination of these algorithms:
Statistical Time-of-Arrival (TOA), Linear Matrix Inequalities (LMI), barycentric algorithms, Grid
Method and Center of Gravity (COG) techniques. Approximate ranging using Statistical TOA
method using Bhattacharyya Distance enables the LMI, Grid Method, barycentric algorithms and
Center of Gravity(COG) technique to improve position accuracy. To investigate the performance
of such an opportunistic localization algorithm, we define a simple model that describes the op-
portunistic interactions between nodes and then we run several computer simulations to analyze
the effect of the ranging error on the positioning of the target node. Along with the simulation
modelwehaveconductedtheexperimentswithreal-worlddatatomeasuretheperformanceofthe
techniquesinthereal-world. Theresultsgeneratedfrombothsimulationandreal-worlddatashow
that opportunistic interactions can actually improve self-localization accuracy, where the position
estimationerrorisabout2morlessinmanydifferentscenarios.
ii
ACKNOWLEDGEMENTS
IwouldliketoexpressmyappreciationtoDr. Limfortheguidancehehasprovidedthrough-
out my study at Auburn. I would also like to express my gratitude to the advisory committee
members,Dr. ShiwenMaoandDr. ThaddeusRoppel.
I would also like to thank several fellow students including Song Gao and Amogh Kashyap
fortheirsignificantcontributionstothisresearch.
Aboveall,IwouldliketothankmyparentswhohelpedmecometoU.S.andpursuegraduate
studyatAuburn.
iii
StylemanualorjournalusedJournalofApproximationTheory(togetherwiththestyleknown
as”aums”). BibliographyfollowsvanLeunen’sAhandbookforScholars.
ComputerSoftwareusedThedocumentpreparationpackageTEX(specificallyLATEX)together
withthedepartmentalstyle-fileaums.sty.
iv
TableofContents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
ListofFigures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
ListofTables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 MOTIVATIONS AND APPLICATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Hospitals,WarehousesandUniversityCampus . . . . . . . . . . . . . . . 6
2.2.2 IndoorTrackingofFirefighters . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 LocalizationofWirelessSensorNetwork . . . . . . . . . . . . . . . . . . 7
2.2.4 AirportsandSubwayStations . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.5 Robotlocalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.6 Museums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.7 TargetedAdvertisingandWarningAlerts . . . . . . . . . . . . . . . . . . 8
3 RELATED WORK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1 RangingTechnique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 NavigationUsingUltraWideband(UWB) . . . . . . . . . . . . . . . . . . . . . . 12
3.3 NavigationusingReceivedSignalStrengthIndicator(RSSI) . . . . . . . . . . . . 13
3.4 NavigationusingTime-of-Arrival(TOA)andTime-difference-of-Arrival(TDoA) . 14
3.5 NavigationusingAngle-of-Arrival(AoA) . . . . . . . . . . . . . . . . . . . . . . 15
4 PROBLEM STATEMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1 ProblemStatement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
v
4.2 ConceptualApproach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2.1 AssumptionsandGoals . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2.2 Time-of-ArrivalRangingTechnique . . . . . . . . . . . . . . . . . . . . . 18
4.2.3 LinearMatrixInequalityTechnique . . . . . . . . . . . . . . . . . . . . . 19
4.2.4 GridMethod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2.5 BarycentricAlgorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.6 Center-of-GravityTechnique . . . . . . . . . . . . . . . . . . . . . . . . . 20
5 DESIGN AND METHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.1 RangingModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.2 Rangingtechniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.2.1 EuclideanTime-of-ArrivalRangingMethod . . . . . . . . . . . . . . . . . 22
5.2.2 StatisticalTime-of-ArrivalRangingMethod . . . . . . . . . . . . . . . . . 23
5.3 Self-LocalizationofTargetNode . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.3.1 RawLMIEstimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.3.2 Self-LocalizationusingLinearMatrixInequalityandBarycentricAlgorithm 27
5.3.3 Self-localization using Linear Matrix Inequality and Center of Gravity
Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3.4 Self-positioningusingGridMethodTechnique . . . . . . . . . . . . . . . 29
6 IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.1 SimulationSetupandImplementation . . . . . . . . . . . . . . . . . . . . . . . . 31
6.2 ReferenceCase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.2.1 LinearMatrixInequalitywithBarycentricAlgorithm . . . . . . . . . . . . 31
6.2.2 LinearMatrixInequalitywithCenterOfGravityTechnique . . . . . . . . 32
6.2.3 VaryingTime-of-ArrivalInducedError . . . . . . . . . . . . . . . . . . . 32
6.2.4 RandomWaypointMobilityModelAnalysis . . . . . . . . . . . . . . . . 33
6.2.5 FirefighterMobilityModelAnalysis . . . . . . . . . . . . . . . . . . . . . 34
6.3 Real-WorldModelSpecifications . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
vi
7 PERFORMANCE EVALUATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.1 SimulationResult . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.2 Accuracyofreferencecase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.2.1 LinearmatrixInequalitywithBarycentricAlgorithm . . . . . . . . . . . . 36
7.2.2 LinearMatrixInequalitywithCenterOfGravityTechnique . . . . . . . . 40
7.2.3 LMIwithvaryingEuclideanTime-Of-ArrivalDistanceError . . . . . . . . 41
7.3 SimulationResultsofMobilityModels . . . . . . . . . . . . . . . . . . . . . . . . 42
7.3.1 LinearMatrixInequalitywithRandomWaypointMobilityModel . . . . . 43
7.3.2 LinearMatrixInequalitywithFirefighterMobilityModel . . . . . . . . . . 44
7.3.3 GridMethodwithCenter-of-Gravity . . . . . . . . . . . . . . . . . . . . . 49
7.3.4 Comparison of Linear Matrix Inequality Technique and Grid Method for
SimulationModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.3.5 Comparison of Linear Matrix Inequality Technique and Grid Method for
RealWorldData . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.2 FutureWork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
vii
ListofFigures
5.1 (a) Histogram of RTT samples for reference node at 10m and (b) Probability density
estimateofRTTsamplesforreferencenodeat10m . . . . . . . . . . . . . . . . . . . 24
5.2 ProbabilitydensityestimateofRTTsamplesforreferencenodeat20m . . . . . . . . 25
5.3 ProbabilitydensityestimateofRTTsamplesforreferencenodeat30m . . . . . . . . 25
5.4 RawPositionEstimationusingonlyLinearMatrixInequality . . . . . . . . . . . . . . 27
5.5 PositionEstimationusingLinearMatrixInequalityandBarycentricAlgorithm . . . . 28
5.6 Plotofallthegridpointsintheintersectionareaincludingthereferencenodeandreal
targetnodepositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.1 LMI-onlyErrorEstimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7.2 LMI+barycenterErrorEstimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.3 ErrorcomparisonBetweenLMI-onlyandLMI+barycenter . . . . . . . . . . . . . . . 39
7.4 LMI-OnlyEstimationErrorforVaryingNodeRange . . . . . . . . . . . . . . . . . . 40
7.5 LMI+BarycenterEstimationErrorforVaryingNodeRange . . . . . . . . . . . . . . . 41
7.6 LMI+BarycenterEstimationforactualdevices . . . . . . . . . . . . . . . . . . . . . 42
7.7 LMI+CenterOfGravityEstimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.8 LMI+COGwithVaryingTOAerror . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.9 LMIwithRWMMandVaryingPausetime . . . . . . . . . . . . . . . . . . . . . . . 45
7.10 LMIwithFirefighterMobilityModelandCOG . . . . . . . . . . . . . . . . . . . . . 46
7.11 FirefightermobilitywithLMIandCOGalongwithdifferentTargetnodepositions . . 47
7.12 FirefightermobilitywithLMIandCOGalongwithdifferentTargetnodepositionsand
largercommunicationrange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.13 MATLAB Map of Shelby Foyer (Auburn University) including all the peer nodes,
RealTargetnodeandEstimatedTargetnodePositions . . . . . . . . . . . . . . . . . 49
viii
7.14 The Indoor map of Shelby Foyer (Auburn University) including all the peer nodes,
Targetnodeandestimatedtargetnodepositions . . . . . . . . . . . . . . . . . . . . . 50
7.15 MatLab Map of Shelby Center Hallway (Auburn University) including all the peer
nodes,RealTargetnodeandEstimatedTargetnodePositions . . . . . . . . . . . . . . 51
7.16 TheIndoor mapofShelby CenterHallway(Auburn University)includingall thepeer
nodes,Targetnodeandestimatedtargetnodepositions . . . . . . . . . . . . . . . . . 52
7.17 Plot of reference node position, Real Target node and Target node position estimated
fromGridandLMImethod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.18 The difference between normal circle communication model and Donut circle Com-
municationmodel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.19 Plotofreferencenodepositionswithdonuts,realtargetnodeandgridpoints . . . . . 55
7.20 Plot of reference node positions, real target node position and target node position
estimatedfromGridandLMImethods . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.21 PlotofreferencenodeswithdonutsandGridpointsfordifferentintersectionareas . . 57
7.22 Plot of Reference node positions, Real target node and estimated target node position
fromGridandLMImethodforAuburnUniversityShelbyHallway . . . . . . . . . . . 58
7.23 Plot of reference node positions, real Target node position and estimated target node
positionfromGridandLMImethodsforAuburnUniversityShelbyFoyer . . . . . . . 59
7.24 Plot of reference node positions, real Target node position and estimated target node
positionfromGridandLMImethodsforAuburnUniversityHallwayandmultiplerooms 60
7.25 PlotoftargetnodelocationerrorcomparisonbetweenLMIandGridmethodwith20%
inducedTOAdistanceerror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.26 PlotoftargetnodelocationerrorcomparisonbetweenLMIandGridmethodwith50%
inducedTOAdistanceerror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.27 PlotoftargetnodelocationerrorcomparisonbetweenLMIandGridmethodwith80%
inducedTOAdistanceerror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.28 Plot of target node location error comparison between LMI and Grid method with
100%inducedTOAdistanceerror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.29 Plot of target node location error comparison between LMI and Grid method for Sta-
tisticalTOAdataforAuburnUniversityFoyHallMainFloor . . . . . . . . . . . . . . 65
7.30 Plot of target node location error comparison between LMI and Grid method for Sta-
tisticalTOAdataforAuburnUniversityShelbycenterhallwayandmultiplerooms. . . 66
ix
7.31 Plot of target node location error comparison between LMI and Grid method for Sta-
tisticalTOAdataforAuburnUniversityShelbycenterhallwayandmultiplerooms. . . 67
7.32 Plot of target node location error comparison between LMI and Grid method for Sta-
tisticalTOAdataforAuburnUniversityBrounhallSecondfloor. . . . . . . . . . . . . 68
x