Table Of ContentLink¨opingStudiesinScienceandTechnology
ThesisNo. 1476
Positioning Algorithms for
Surveillance Using Unmanned
Aerial Vehicles
by
Per-Magnus Olsson
SubmittedtoLink¨opingInstituteofTechnologyatLink¨opingUniversityinpartial
fulfilmentoftherequirementsfordegreeofLicentiateofEngineering
DepartmentofComputerandInformationScience
Link¨opinguniversitet
SE-58183Link¨oping,Sweden
Link¨oping2011
ISBN 978-91-7393-200-4, ISSN 0280–7971
Printed by LiU-Tryck, 2011
Copyright ' Per-Magnus Olsson 2011
Electronic version available at :
http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-66060
Positioning Algorithms for Surveillance Using
Unmanned Aerial Vehicles
by
Per-MagnusOlsson
April2011
ISBN978-91-7393-200-4
Link¨opingStudiesinScienceandTechnology
ThesisNo. 1476
ISSN0280–7971
LiU–Tek–Lic–2011:15
ABSTRACT
Surveillanceisanimportantapplicationforunmannedaerialvehicles(UAVs). Thesensed
informationoftenhashighpriorityanditmustbemadeavailabletohumanoperatorsas
quickly as possible. Due to obstacles and limited communication range, it is not always
possibletotransmittheinformationdirectlytothebasestation. Inthiscase,otherUAVs
canformarelaychainbetweenthesurveillanceUAVandthebasestation. Determining
suitablepositionsforsuchUAVsisacomplexoptimizationprobleminandofitself,and
ismadeevenmoredifficultbycommunicationandsurveillanceconstraints.
TosolvedifferentvariationsoffindingpositionsforUAVsforsurveillanceofonetarget,
two new algorithms have been developed. One of the algorithms is developed especially
forfindingasetofrelaychainsofferingdifferenttrade-offsbetweenthenumberofUAVs
andthequalityofthechain. Theotheralgorithmistailoredtowardsfindingthehighest
qualitychainpossible,givenalimitednumberofavailableUAVs.
Finding the optimal positions for surveillance of several targets is more difficult. A
study has been performed, in order to determine how the problems of interest can be
solved. Itturnsoutthatveryfewoftheexistingalgorithmscanbeusedduetothechar-
acteristics of our specific problem. For this reason, an algorithm for quickly calculating
positionsforsurveillanceofmultipletargetshasbeendeveloped. Thisenablescalculation
of an initial chain that is immediately made available to the user, and the chain is then
incrementallyoptimizedaccordingtotheuser’sdesire.
ThisworkhasbeensupportedbyCUGS(theSwedishNationalGraduateSchoolinCom-
puter Science), LinkLab (www.linklab.se), the Swedish National Aeronautics Research
ProgramNFFP04–S4203andNFFP05–01263,theELLIITExcellenceCenteratLinko¨ping-
Lund for Information Technology, the Swedish Foundation for Strategic Research (SSF)
Strategic Research Center MOVIII, the Center for Industrial Information Technology
CENIIT(grantnumber06.09)andtheLinnaeusCenterforControl,Autonomy,Decision-
making in Complex Systems (CADICS), funded by the Swedish Research Council (VR).
DepartmentofComputerandInformationScience
Link¨opinguniversitet
SE-58183Link¨oping,Sweden
Acknowledgements
This thesis would not have been possible without the support of co-workers
and friends. Especially I would like to thank:
Mysupervisors, professorPatrickDohertyandassociateprofessorJonas
Kvarnstr¨om for guidance in research as well as in writing of this thesis.
My co-researchers Kaj Holmberg and Oleg Burdakov for many valuable
discussions.
ThecolleaguesatAIICSandespeciallyOlovAndersson, FredrikHeintz,
David Land´en, Martin Magnusson and Piotr Rudol for spending time read-
ing this thesis as well as earlier papers.
Anne Moe for guiding me through the bureaucratic maze of graduate
studies.
Otherfriends,toonumeroustomentionbyname,forencouragementand
inspiration.
Linda for being the world’s best girlfriend.
Contents
1 Introduction 1
1.1 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Related Work 7
2.1 UAVs As Relays . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Single Target . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Multiple Targets . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Area Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Exploration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 Ad-hoc Networks and Wireless Sensor Networks. . . . . . . . 12
3 The Relay Positioning Problems 15
3.1 Problem Setup . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Definitions of the Single Target Relay Problems . . . . . . . . 16
3.3 Cost Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3.1 Transmission Quality. . . . . . . . . . . . . . . . . . . 19
3.3.2 Position Visibility . . . . . . . . . . . . . . . . . . . . 20
3.3.3 Minimum Free Angle Between Positions . . . . . . . . 21
3.3.4 Minimum Distance to Obstacles . . . . . . . . . . . . 23
3.3.5 Surveillance Cost Functions . . . . . . . . . . . . . . . 23
3.4 Reachability Functions . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Problem Properties . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 Continuous Solution Methods . . . . . . . . . . . . . . . . . . 27
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4 Environment Representation and Discretization 31
4.1 Discretization and Graph Creation . . . . . . . . . . . . . . . 31
4.2 Fixed-Size Grids . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Octrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
ix
4.4 Expanded Geometry Graphs . . . . . . . . . . . . . . . . . . 35
4.5 Voronoi Diagrams . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6 Discretization Methods Used in Motion Planning . . . . . . . 38
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5 Relay Positioning Algorithms for Single Target Problems 41
5.1 Existing Algorithms for the STR-ParetoLimited Problem . 42
5.2 A New Label-Correcting Algorithm . . . . . . . . . . . . . . . 44
5.2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . 44
5.2.2 Algorithm Details . . . . . . . . . . . . . . . . . . . . 46
5.2.3 Correctness Proof . . . . . . . . . . . . . . . . . . . . 50
5.2.4 Time Complexity . . . . . . . . . . . . . . . . . . . . . 52
5.2.5 Improved Preprocessing . . . . . . . . . . . . . . . . . 52
5.2.6 Example . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.3 A New Dual Ascent Algorithm . . . . . . . . . . . . . . . . . 54
5.3.1 Algorithm Details . . . . . . . . . . . . . . . . . . . . 55
5.3.2 Theoretical Properties . . . . . . . . . . . . . . . . . . 57
5.3.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.4 Performance Improvements . . . . . . . . . . . . . . . 60
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6 Relay Positioning for Multiple Targets 65
6.1 Definition of the Multiple Target Relay Problems . . . . . . . 65
6.2 Relation to Steiner Tree Problems . . . . . . . . . . . . . . . 67
6.2.1 Continuous Steiner Trees . . . . . . . . . . . . . . . . 67
6.2.2 Discrete Steiner Trees . . . . . . . . . . . . . . . . . . 68
6.3 Adapting the Cheapest Path Heuristic . . . . . . . . . . . . . 78
6.3.1 Theoretical Properties . . . . . . . . . . . . . . . . . . 80
6.3.2 Extensions . . . . . . . . . . . . . . . . . . . . . . . . 81
6.4 Calculating Pareto-optimal Relay Trees For Two Targets. . . 83
6.4.1 Determining the Set of Pareto-optimal Relay Trees . . 85
6.4.2 Duplicate Edges in the Relay Tree . . . . . . . . . . . 86
6.5 Improving Relay Trees . . . . . . . . . . . . . . . . . . . . . . 87
6.5.1 Reduced Trees . . . . . . . . . . . . . . . . . . . . . . 89
6.5.2 Choosing Subtrees for Optimization . . . . . . . . . . 89
6.5.3 Different Tree Structures . . . . . . . . . . . . . . . . 94
6.5.4 Collisions Between Trees . . . . . . . . . . . . . . . . . 95
6.5.5 Algorithm Details . . . . . . . . . . . . . . . . . . . . 97
6.5.6 Time Complexity . . . . . . . . . . . . . . . . . . . . . 99
6.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7 Implementation and Experimental Results 103
7.1 Software Architecture . . . . . . . . . . . . . . . . . . . . . . 103
7.2 Problem Setup for Empirical Testing . . . . . . . . . . . . . . 107
7.3 Pareto-Optimal Relay Chains . . . . . . . . . . . . . . . . . . 110
7.4 Optimal Chains Using At Most M UAVs . . . . . . . . . . . 114
7.5 Relay Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
8 Discussion 127
8.1 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . 128
Description:suitable positions for such UAVs is a complex optimization problem in and of itself, and .. to and from such an altitude might require significant transmission power, Algorithms for finding trajectories are not part of this thesis for example be flight time, fuel consumption or visibility A grid c