THÈSE
THÈSE
En vue de l’obtention du
DOCTORAT DE L’UNIVERSITÉ DE TOULOUSE
Délivré par : l’Université Toulouse 3 Paul Sabatier (UT3 Paul Sabatier)
Présentée et soutenue le 29/10/2013 par :
Georgia Iuliana DEACONU
On the trajectory design, guidance and control for spacecraft rendezvous
and proximity operations.
JURY
M. Ali ZOLGHADRI
Professeur, Université Bordeaux I
Président du Jury
M. Denis ARZELIER
Directeur de Recherche, CNRS
Examinateur
M. Jean-Claude BERGES
Ingénieur, CNES
Membre Invité
M. Thomas CARTER
Em. Prof., East. Connecticut SU
Examinateur
M. Patrick DANES
Professeur, UT3
Examinateur
M. Alexandre FALCOZ
Ingénieur, Astrium EADS
Membre Invité
M. Christophe LOUEMBET
Maître de conférences, UT3
Examinateur
M. Sorin OLARU
Maître de conférences, SUPELEC
Rapporteur
M. Andrew SINCLAIR
Associate Prof., Auburn Univ.
Rapporteur
M. Alain THERON
Chercheur affilié, LAAS-CNRS
Examinateur
École doctorale et spécialité :
EDSYS : Automatique 4200046
Unité de Recherche :
Laboratoire d’Analyse et d’Architecture des Systèmes du CNRS
Directeur(s) de Thèse :
M. Christophe LOUEMBET et M. Alain THERON
Rapporteurs :
There are many people that I would like to thank for their direct or indirect contribution to the
completion ofthis thesis. Iwould like tothank rst myadvisors,Christophe LouembetandAlain
Théron,fortheirguidanceandforthefreedomtheygrantedmeinchoosingtheresearchdirections.
ThisthesishasbeensupportedbytheFrenchNationalCenterforSpaceStudiesandbyAstrium
EADSandIwouldlike tousethisoccasionto thankJean-Claude BergesandAlexandreFalcoz for
their help andfor their participation.
IwouldliketothankDenisArzelier,thehead oftheMethods andAlgorithmsinControl which
hosted me at LAAS-CNRS for the opportunity he oered me a while ago for the internship and
for supporting mycandidature for this subject. Iwouldalso like to thank all themembersof the
MACgroupfor making theworkenvironment soenjoyable.
Special thanksto Eric Kerrigan and to Paola Falugi and to theother members of theControl
and Powergroupat theImperialCollege ofLondon,fortheinterestingand challenging discussions
and for thegood timesduring mystayinLondon.
A³vrea deasemeneasãlemulµumesc pãrinµilormei,Nicoleta³iMarian, precum³isuroriimele
Teodora pentrutotajutorul oferit peperioada studiilor.
IwouldalsoliketothankFrancescoforhiscareandforhissupportandmyoldandnewfriends
that have shared good and bad days with me over the years: Rãzvan, Dinu, Oana, Magda and
Recentspacemissionsrelymore andmoreonthecooperationbetween dierentspacecraftinorder
to achieve a desiredobjective. Among thespacecraft proximity operations, theorbitalrendezvous
is a classical example that has generated a large amount of studies since the beginning of the
space exploration. However, the motivations and objectives for the proximity operations have
considerably changed. The need for higher autonomy, better security and lower costs prompts
for the development of new guidance and control algorithms. The presence of dierent types of
constraints and physical limitations also contributes to the increased complexity of the problem.
In this challenging context, this dissertation represents a contribution to the development of new
spacecraft guidance andcontrol algorithms.
The works presented in this dissertation are based on a structural analysis of the spacecraft
relative dynamics. Using a simpliedmodel, a new set of parametric expressions is developed for
the relative motion. This parametrization is very well suited for the analysis of the geometric
properties of periodicrelative trajectories and for handling dierent types of state constraints. A
formal connection is evidenced between the set of parameters that dene constrained trajectories
and theconeof positive semi-denite matrices. Thisresult isexploited inthedesign of spacecraft
relative trajectories for proximity operations, in the impulsive control framework. The resulting
guidance algorithms enable the guaranteed continuous constraints satisfaction, while still relying
on semi-denite programming tools. The problem of the robustness of the computed maneuvers
La réalisation des missions spatiales repose de plus en plus souvent sur la coopération entre
dif-férentsengins spatiaux. Parmi les opérations de proximité, lerendez-vousorbital estune pratique
aussianciennequelaconquêtespatiale,quicontinuedegénérer denombreuxtravauxderecherche.
Cependant, les motivations etles objectifs desrécentes missions de rendez-vous orbital ont
large-ment évolués. En eet, les besoins d'une autonomie accrue, d'une sécurité améliorée, d'une plus
grande exibilité etd'une réduction descoûts, constituent autant d'incitations au développement
de nouvelles méthodes de guidage et contrôle. La satisfaction de contraintes très variées, dues à
des considérations de sécurité ouà deslimitations technologiques incontournables desactionneurs
ou descapteurs, contribuent à la richesse du problème posé. Dans ce contexte, le développement
de nouveauxalgorithmes de commande constitue unvrai déscientique quecette thèse tente de
relever.
Les travaux de cette thèse sont basées sur l'analyse structurelle des expressions décrivant le
mouvement relatifentredeuxvéhiculesenorbite. Surlabasedesmodèlesdetransitiondisponibles
dans la littérature, une nouvelle paramétrisation du mouvement relatif est proposée. Celle-ci,
particulièrement adaptée àlacaractérisation destrajectoires périodiques, orelapossibilitéd'une
prise en compte de contraintes d'état très variées. Un lien formel est mis en évidence entre les
paramètres dénissant les trajectoires contraintes et le cône des matrices semi dénies positives.
Ces résultatssont exploités dans ledéveloppement desalgorithmes de design de trajectoires pour
desopérationsde proximité,sous hypothèse de poussée impulsionnelle. Ces algorithmesont,entre
autre,lapropriétédepermettrelasatisfactiondescontraintessurlatrajectoiredemanièrecontinue
dans letemps, tout en utilisant lesoutils numériques de laprogrammation convexe. Le problème
spéciquedelarobustessedesman÷uvresauxincertitudesdelachaînedemesureestaussiabordé
danscemanuscrit. Desapprochesdetypecommandeprédictivesontmisesenplaceandegarantir
Acknowledgements i
Abstract iii
Résumé v
Nomenclature ix
Introduction 1
1 Spacecraft relative motion 11
1.1 Introduction . . . 11
1.2 Dynamics ofa spacecraftorbiting theEarth . . . 12
1.3 Spacecraft relative motion . . . 16
1.3.1 LocalCartesian dynamics . . . 16
1.3.2 Orbital elements dierencesdynamics . . . 19
1.4 Linearized Cartesianrelative motion . . . 20
1.4.1 State-space representation . . . 20
1.4.2 The state transitionmatrix . . . 21
1.5 Properties ofrelative trajectories . . . 24
1.5.1 Periodicity conditions . . . 24
1.5.2 Inter-satellite distance . . . 27
1.5.3 Geometry oftheperiodicspacecraftrelative motion. . . 28
1.6 Conclusions . . . 30
2 Parametric expressions for the spacecraft relative trajectory 33 2.1 Denition ofthe parameters . . . 34
2.2 Properties ofspacecraft relative trajectories . . . 35
2.2.1 Dynamics ofthevector of parameters . . . 36
2.2.2 Properties ofperiodictrajectories . . . 38
2.3 Numericalanalysis oftheperiodic relative motion. . . 40
2.3.1 The eectsof theeccentricity of theleaderorbit . . . 40
2.3.2 The eectsof thevalues of theparameters . . . 40
2.4 Conclusion. . . 43
3 Constrained spacecraft relative trajectories 45 3.1 Denition ofadmissible trajectories . . . 46
3.2 Finite descriptionof admissible trajectories . . . 47
3.2.1 Finite descriptionusing constraints discretization . . . 48
3.2.2 Finite descriptionusing non-negative polynomials . . . 48
3.3 Description ofconstrained trajectories usingnon negative polynomials . . . 49
3.3.1 Rational expressionsfor thespacecraft relative motion . . . 49
3.3.2 Constrained nonperiodic trajectories . . . 52
4 Trajectory design for spacecraft rendezvous 57
4.1 Fixed-time linearizedimpulsivespacecraft rendezvous. . . 58
4.1.1 General formulation oftheguidance problem . . . 58
4.1.2 Consumption criteria . . . 59
4.1.3 Saturation constraints . . . 60
4.1.4 Using directshooting methods for theguidance problem . . . 61
4.2 Fixed-time rendezvouswithtrajectoryconstraints . . . 63
4.2.1 Guidance towardsa constrained periodicrelative motion . . . 63
4.2.2 Passively safe trajectories for spacecraftrendezvous . . . 66
4.2.3 Spacecraft rendezvous withvisibilityconstraints . . . 68
4.3 Numericalexamples . . . 70
4.3.1 Reaching aconstrained periodicrelative trajectory . . . 70
4.3.2 Passively safe rendezvoustrajectories . . . 74
4.3.3 Constrained nonperiodic relative trajectories . . . 78
4.4 Conclusion. . . 79
5 Spacecraft rendezvous robust to navigation uncertainties 81 5.1 ModelPredictive Control and spacecrafttrajectorydesign . . . 82
5.2 The robusttrajectoryplanning problem . . . 84
5.2.1 The spacecraftrelative dynamics . . . 85
5.2.2 The eectsof navigation uncertainties . . . 86
5.2.3 The nominaltrajectory . . . 87
5.2.4 General formulation oftheguidance problem . . . 87
5.3 Ane state-feedbackMPC. . . 89
5.3.1 Computation ofthefeedbackgains . . . 89
5.3.2 Computation ofthenominal control . . . 92
5.4 Ane disturbance feedbackMPC . . . 93
5.5 Numericalevaluation oftherobust control techniques. . . 97
5.5.1 Description ofthe simulation procedure . . . 97
5.5.2 The PRISMAmission . . . 99
5.5.3 The Simbol-X mission . . . 103
5.6 Conclusion. . . 108
6 Analytical bi-impulsive controlaround a desired periodic trajectory 109 6.1 Stabilityarounda periodicrelative trajectory . . . 110
6.2 Analytical bi-impulsive stabilizing control for theperiodicmotion . . . 111
6.2.1 Computation ofthecontrol . . . 112
6.2.2 Domain ofvalidity . . . 113
6.2.3 Performances inpresenceof navigation uncertainties . . . 115
6.3 Robust guidancetowards a spacecraftperiodicrelative motion . . . 117
6.4 Numericalexamples . . . 119
6.4.1 Inuenceof theeccentricityof thereferenceorbit . . . 120
6.4.2 Inuenceof theintervalbetween controls . . . 122
6.4.3 Inuenceof thenavigation uncertainties . . . 123
6.5 Conclusion. . . 124
B Properties of non negative polynomials 137
B.1 Checking polynomials non negativityon anite interval . . . 137
B.2 Checking polynomials non negativityon aninnite interval . . . 138
C Ellipsoidal sets 141
C.1 Representations ofellipsoidal sets . . . 141
C.2 Operations withellipsoids . . . 142
C.3 The S-procedure . . . 142
µ
Earth'sgravitational constantν
true anomalyΩ
longitudeof theascendingnodeω
argument ofperigeea
semi-major axisB
0
Earthcentered inertial baseB
l
Spacecraft centered localCartesian baseE
eccentric anomalye
eccentricityi
orbit inclinationM
mean anomalyn
mean motionLMI Linear MatrixInequality
LP Linear Program
LTI Linear TimeInvariant
LTV Linear TimeVarying
LVLH Local Vertical LocalHorizontal
MPC ModelPredictive Control
Résumé: Le succès des missions spatiales repose de plus en plus souvent sur la coopération entre
plusieurs véhicules en orbite. L'approvisionnement de la Station Spatiale Internationale par
ex-emple est assuré par des opérations de rendez-vous orbital, tandis que des nombreuses missions
scientiques utilisent des formations de satellites pour relever des mesures. Ce type d'opérations
ont des besoins spéciques en termes d'algorithmes de contrôle, vue la distance réduite entre les
véhicules, les contraintes d'autonomieetde sécuritédes missionspatiales etles ressources limitées.
Les travaux decettethèse portent sur ledéveloppement des algorithmes deguidage pour des op
éra-tionsdeproximitéentreles satellites,oùla distance réduitepermetla navigationrelative. L'objectif
est de fournir des plans de man÷uvres optimisés du point de vue de la consommation de
com-bustible, qui prennentencompteles contraintes opérationnellesde la missionetqui soientrobustes
à des incertitudes. Le cadre detravail choisiest celui des méthodes dîtesdirectes, qui permettentla
formulation duproblème deguidage comme unproblème d'optimisationparamétrique.
Background and motivations
Spacecraft rendezvousand docking capabilities arerequired for alarge arrayofspace applications
thatinvolvemorethanonespacecraft. Itisakeytechnologyforthein-orbitassemblyoflargeunits,
such as the space stations (Mir, Skylab, ISS). The space stations further rely on rendezvous and
dockingmissionsinorderto receivesupplies orto exchange thecrew. Forinstance, theunmanned
Automated Transfer Vehicle (ATV) from the European Space Agency periodically supplies the
International SpaceStation(ISS)withpropellant,water, air, payloadsand experiments. Recently,
theDragon spacecraft became therstcommercial spacecraft to successfullydockwiththeISS.
Space rendezvous has also been used for a variety of other purposes, including the service
missions to the Hubble Space Telescope and the EURECA spacecraft retrieval. Other on-orbit
servicing missions areunder study for existing spacecraft [7,85]. The increasing number of space debris in the Low Earth Orbit originating from mutual collisions, motivated the study of active
debris removalmissions[13].
In the recent years, a lot of interest hasbeen shown for space scientic missions that rely on
dierent instruments distributedovera eet of spacecraft. This congurationcan provide several
can usemultiple "massproduction" vehicles to assemble theeet. The robustness of themission
is alsoincreased bythis congurationsince thepayloadsaredistributedamongthespacecraft and
can eventually be replaced in case of failure [98]. Formation ying oers more exibility because the formation can be recongured in order to follow new mission requirements. This approach
has been considered for scientic missionswith very diverse objectives, such asEarth observation
(A-train),interferometryforEarth-likeplanetsdetection(DARWIN),measurementofgravitational
waves fromsupermassiveblackholebinaries (LISA)or X-raysspace telescope (Simbol-X).
The success of spacecraft rendezvous and formation ying missions depends on the precise
control of the spacecraft relative state, often-times in the context of relatively small spacecraft
separations. Inordertoensurethesecurityofthemission,ahighdegreeofautonomyandrobustness
is desiredfor therelative motioncontrol procedure. For missionssuchastheMarsSample Return
[86], for which the communication delay between the ground station and the spacecraft is very large,anautonomousguidancealgorithmwhichguaranteesthatnocollisionwilloccurbetweenthe
spacecraft isof vitalimportance.
The fuel-cost of the spacecraft maneuvers is also a matter of concern. The propulsion system
canaccountforupto50%ofthespacecraftmassatthelaunchtime,reducingtheavailablepayload
massandinuencingthecostofthelaunch. Thecontrolalgorithmsmustensurethatthecomputed
maneuversarefuel-optimal,suchthatthedesiredlifetimeforthemissioncan be achieved withthe
smallest amount of propellant.
The spacecraft rendezvous
Theorbitalrendezvousprocessconsistsinexecutingaseriesoforbitalmaneuverswiththepurpose
of bringing two spacecraft in close vicinity of each other. Usually one of the spacecraft, called
thetarget, isconsideredto be inert,while thesecondspacecraft, calledthefollower or thechaser,
executesthemaneuvers. Whentheobjectiveisto physicallyjointhetwospacecraftinvolvedinthe
rendezvous, we speakabout docking or berthing.
W. Fehse identied in[29]several phases ofa rendezvous mission, each one withits own chal-lenges. For the launch stage, the purpose is to bring the two spacecraft in vicinity by placing
them in the same orbital plane. The phasing stage aims at reducing the phase angle between
the target and the follower (see the illustration inFigure 1). During the phasing maneuvers, the
follower spacecraft is controlled from the ground station and the navigation is based on absolute
Figure 1: Viewofthetarget's orbital planeat thebeginning of thephasing stage
a stablepositioninproximityof thetarget,using relative navigation measurements. For theATV
rendezvous scenario with the ISS for instance, this stage starts at a range of few tens of
kilome-tres and ends at a range of few kilometres from the target spacecraft. The following close range
rendezvous phase is usually divided in two stages: the closing maneuvers guiding the spacecraft
towards thenalapproachcorridor andthenal approach stage leading to matingconditions.
The dierent stages of an orbital rendezvous mission aresummarized inFigure 2. The works
presentedinthisdissertationarerelatedtothehomingandtheclosingphases,duringwhichthetwo
spacecraftrelyonrelativenavigationmeasurementsinordertoachievethedesirednalconditions.
Some of the presented examples also refer to the nal approach phase, leading to the spacecraft
docking.
Mission constraints and technical challenges
The spacecraft rendezvous guidance is a complex process due to thedierent types of conditions
andconstraintsthatmustberespectedduringeachphaseofthemission. Forthephasesconsidered
in this dissertation, the far range and close range rendezvous stages, theapproach trajectory can
be required for instance to pass through specied hold points where the follower vehicle must
waitfor thepermissionto proceed,either fromtheground control station or fromthe crew ofthe
targetspacecraft[29]. Securityconsiderationsmightimposethechoiceofapproachtrajectoriesthat are inherently safe, meaning that they are guaranteed to avoid any collision with target vehicle,
Figure 2: Thedierent phases ofa rendezvousmission
visibilityconeofthetargetspacecraftforcontinuousvisualcontact. Thesespecicationscorrespond
to constraints that the rendezvous trajectory must respect in order to certify that the mission
requirementsaremet.
Another factor that must be taken into considerationwhen designing the approach trajectory
is the fuel-cost of the maneuvers. Thrust maneuvers can be approximated with impulses, i.e.
instantaneous changes of velocity at the time of maneuver. This simplies the computation and
the analysis of a fuel-optimal maneuvers plan for the spacecraft rendezvous [29]. The impulsive approximation is especially well adapted for the liquid propellant engines which are used for a
wide spanofspacecraftmaneuvers, rangingfromorbitaltransfermaneuversto stationkeepingand
Orbitaldisturbances,navigationerrorsandcontrolexecutionerrorscanaltertheoutcomeofthe
computed maneuvers. Thepresenceandtheeectsofthesedisturbancesneedto beintegratedinto
therendezvoustrajectorydesignphase. Navigationerrorsaredenedasthedierencebetweenthe
state perceivedbytheonboardsystemandtherealstateofthevehicle. Theycanbecausedbythe
sensors measurement performancelimitations, byerrorsin thealignment between thesensors and
spacecraft axes, by the onboard information processing and ltering, etc. The control execution
errors refer to deviations in magnitude, direction or application time from the desired impulsive
thrusts. They can be due to mounting errors, to misalignments with the mechanical axes, to the
engine performances, etc. The decision autonomy of the spacecraft cannot be increased without
providing a priori guarantees for its behaviour inperturbed conditions. Thismust be done while
usingcontrolalgorithmsofreducedcomplexitysincethecomputationalresourcesavailableonboard
thespacecraft arelimitedwithrespectto thoseavailableon ground.
Some of these challenges are addressed in this dissertation. The main objective is to provide
algorithms for thecomputation of robust fuel-optimal maneuversplans leading to rendezvous
tra-jectories that respecttheconstraintsimposedby themission'srequirements, even inpresenceof a
certain classofuncertainties. A study ofthespacecraftconstrained naturally periodictrajectories
is carriedout inrelation tothesecurityspecicationsfor theapproachtrajectory.
Thespacecrafttrajectorydesignisachallenging problemduetothepresenceoftrajectoryand
control constraints, totherobustness considerationsandtothelargenumberofdesignparameters.
In the most general case, only the initial time of the mission is xed and the trajectory design
procedure must provide a choice for the nal time, the number and the distribution of thrusting
instants, the amplitude and the direction of the thrusts. If the design algorithm is intended for
use onboard the spacecraft, then restrictions are added on its computational complexity. A brief
presentationofthemaintrajectorydesignapproachesisgiveninwhatfollows,withafocusontheir
abilityto handlethedierentmission requirements.
Spacecraft relative trajectory design approaches
Thetrajectorydesignforspacecraftrendezvousandproximityoperationsreferstothecomputation
ofaseriesofmaneuversthatsteerthespacecraftfromsomeknowninitialrelativeconditionstosome
nal desired relative conditions. The design procedure generally consists insolving an open-loop
optimal control problemwhose solution corresponds tothe best approach trajectory thatrespects
respected and can increase thelifetime of the spacecraft. The techniques for solving this type of
constrainedopen-loopoptimalcontrolproblemsareusuallydividedintodirectmethodsandindirect
methods [24].
Indirect methods are based on analytical necessary optimality conditions derived using the
calculus of variations and thePontryagin maximumprinciple. The optimal solution can be found
by solving the two-point-boundary-value problem (TPBVP) resulting from these conditions [24]. When using the indirect methods, the optimal spacecraft trajectory for the rendezvous problem
is computed indirectly, based on the evolution of the adjoint state vector or the so-called primer
vector [58,62]. For impulsive trajectories, the primer vector indicates the times and thepositions of the thrust impulses that minimize the total fuel cost. However, the resolution of the problem
is complicated in thegeneral case, especially when constraints are added to the problem. It also
requires a good guess for the initial value of the primer vector. Recent works on the spacecraft
rendezvous problem have focused on transforming the necessary conditions for optimality into
constructive conditions forthe optimalsolution [3,4].
Direct methods rely on the transformation of the optimal control problem into a parameter
optimization problem. This isusually achieved through control parametrization and through
dis-cretization [45]. Theobtained nite-dimensional optimal control problemcan be eciently solved using the existing algorithms [11]. There are dierent types of direct methods depending on the choiceforthedecisionvariablesandontheusedintegrationmethod. Amongthem,thedirect
shoot-ing methods are used in the cases where the parametrisation concerns only the control variables.
The system'sdynamics areusuallylinear andare integrated analyticallyor numerically [49].
Theindirectresolutionmethodscertifytheglobaloptimalityofacomputedsolutionbychecking
a set of necessary and sucient (ifavailable) conditions. Howeverthey lead to problems thatare
hard to solve numerically, especially when constraints are considered. Direct methods are ableto
deal withstate andcontrol constraints more eectively and to integraterobustness elementswith
respect to dierent types of disturbances. Even if the obtained solution can only be certied as
optimal for the particular parametrization and/or discretization that has been considered, they
provide an attractive alternative for therendezvousguidance problem.
The algorithms developed in this dissertation for the design of spacecraft rendezvous
trajec-tories fall into the category of direct shooting methods. Other than theadvantage related to the
reduced complexity of theresulting optimization problem, this approach also oersthepossibility
willbeonreducingtheeectsofrelativenavigationuncertainties onthenalrendezvousprecision.
Spacecraft trajectory control: closing the loop
The direct and indirect approaches for spacecraft relative trajectory design provide a series of
fuel-optimal maneuvers that need to be executed at the specied instants in order to reach the
desired nal objective. The maneuvers plan is obtained based on open-loop predictions of the
evolution of the spacecraft relative trajectory. As previously discussed, the presence of orbital
perturbations, navigation uncertainties or control execution errorsmight alter theoutcome of the
computed maneuvers. Inorder to limit their undesiredeects andto reach aspecied rendezvous
precision, thetrajectory control needsto be implementedina closed-loop manner.
Theresolutionofaconstrainedopen-loopoptimalcontrolproblemcanbeintegratedina
closed-loop setting by using the Model Predictive Control (MPC) methodology [84]. Model Predictive ControlorRecedingHorizon Controlisacontroltechnique forwhichthecontrol actionisobtained
bysolving at each sampling instant anite-horizon open-loop optimal control problem, using the
current state ofthesystemasinitial state. The optimizationdeliversevery timeaseries ofcontrol
actions out of which only therst one is applied to thesystem. Therest of the planis discarded
because a new solution, based on new measurement information, will be computed at the next
sampling time[68].
ModelPredictiveControlisapopularcontroltechniqueforspacecraftrendezvousandproximity
operations [16,18,26,32,41,43,86]. Itspopularity isdue totheabilityto integrateconstraints and uncertainties directly into the trajectorydesign problem. Dierent othercontrol approaches have
been proposed for spacecraft proximity operations and formation ying, spanning over a large
range of techniques. A non exhaustive list includes adaptive control [2,95], non-linear quadratic regulator [6], feedback impulsive control [89], Lyapunov-based nonlinear output feedback control [104], time-delayed feedbackcontrol [12]andseveralothers [87,88]. Butvery fewofthem consider thepresenceofconstraintsorthefuelcostofthemaneuvers,andfocusonlyonreachingthespecied
nal conditions.
Insteadofdeterminingo-lineafeedbackpolicythatprovidestheoptimalcontrolforallsystem
states, MPC solves an open-loop optimal control problem on-line which takes into consideration
the current state of the system. The periodic recomputation of the solution creates an implicit
closed-loop. Therobustnesspropertiesofthisimplicit closed-loopwithrespecttodierenttypesof
The presence of uncertainties raises questions related to the changes induced in the control
performances. Inthecaseofspacecrafttrajectorycontrol, theperformances aredened inrelation
to the fuel consumption and to the precision with respect to the desired nal objective of the
maneuvers. The Model Predictive Control possesses some inherent robustness properties, dened
as the robustness of the closed-loop for the control that has been computed without explicitly
considering theuncertainties [36,67]. But for problems thatinclude control and state constraints, the computed control actions must guarantee that no transgressions of theconstraints will occur
for allthepossiblerealizationsoftheuncertainties. Inthiscase, theinherent robustnessproperties
areno longersucientand thepresenceofuncertainties needsto be included inthewriting ofthe
optimizationproblem[27,59,69,79]. Anotherkeyaspectisthepropertyofrecursivefeasibilityofthe control problem inpresenceof uncertainties. The optimal control is recomputed at each sampling
instant anditisimportantto providetheoreticalguarantees that,iftherst optimizationproblem
is feasible, then all the subsequent optimization problems will also be feasible. These important
properties areinvestigated for theguidance algorithmsproposedin thisdissertation.
Objectives and organization of the dissertation
The works presented inthis dissertation are oriented following two main axes: theanalysisof the
spacecraftrelativemotionandthedesignandcontrolofthespacecraftrelativetrajectory. Thestudy
of the relative motion concentrates on spacecraft naturally periodic relative trajectories. These
periodictrajectories,intheabsenceofperturbations,require nocontrolinorderto bemaintained.
Thispropertycouldmakethemgoodcandidatesfor parkingorbitsinbetween dierent phasesofa
rendezvousmission,forautonomousinspectiontrajectoriesforon-orbitservicingmissionsorforfail
trajectories in case of systemmalfunction. Chapter 1 summarizes the most common-used models
for representing the spacecraft relative motion. It also provides an overview of the properties
of the spacecraft relative trajectories that are of interest for the rendezvous guidance problem,
suchasperiodicityconditions, inter-satellite distanceand geometric properties ofperiodicrelative
trajectories.
Thedierentperiodicmotioninitialisationtechniquespresentedintheliteraturedonotgiveany
information about the geometric properties of the resulting trajectory. To address this problem,
a new parametrization for the spacecraft relative trajectories is developed in Chapter 2. This
parametrizationprovidesagoodframeworkforanalysingtheirpropertiesanditisinusedinChapter
The spacecraft relative trajectory control concentrates around the problem of designing
fuel-optimal maneuvers plans leading the spacecraft from an arbitrary initial relative state towards a
desirednalrelativestate,followingtrajectorieswhichrespectsdierentmissionconstraints.
Chap-ter 4 details the writing of the spacecraft rendezvous guidance problem as an impulsive optimal
control problem using direct shooting methods. It illustrates the contribution of the results
pre-sented in Chapter 3 in obtaining approach trajectories that respect visibility constraints or that
are guaranteed to be safe for a large rangeof system errors. The robustness aspects with respect
to navigation uncertainties aretreatedinChapter5. Theguidanceproblemismodiedinorderto
provideasolutionwhichguaranteesapriori constraintssatisfactionforalladmissiblevaluesforthe
uncertainties, without modifyingthecomplexityof thecontrol algorithm. Moreover, theproposed
control strategyalsominimizestheeectsofthesensingnoise ontheprecisionwithwhichthenal
objective isachieved.
Thepresence ofperturbationsalso aectsthespacecraft naturally periodic motion. Chapter 6
presentsalow-complexitystabilizingcontrolstrategyforthespacecraftperiodicmotioninpresence
of sensingnoise. Thedeveloped methodisbasedontheparametrizationfor thespacecraftrelative
trajectorypresentedinChapter 2.
Spacecraft relative motion
Contents
1.1 Introduction . . . 11
1.2 Dynamicsof a spacecraft orbitingthe Earth . . . 12
1.3 Spacecraft relative motion . . . 16
1.3.1 LocalCartesiandynamics . . . 16
1.3.2 Orbitalelementsdierencesdynamics . . . 19
1.4 Linearized Cartesianrelative motion. . . 20
1.4.1 State-spacerepresentation . . . 20
1.4.2 Thestatetransitionmatrix . . . 21
1.5 Properties of relative trajectories. . . 24
1.5.1 Periodicityconditions . . . 24
1.5.2 Inter-satellitedistance . . . 27
1.5.3 Geometryoftheperiodicspacecraftrelativemotion . . . 28
1.6 Conclusions . . . 30
Résumé: L'étude du mouvement relatif des satellites consiste à analyser la dynamique d'un
satelliteappelélechasseurparrapportàunautre satellite,appelélacible. Diérenteschoixexistent
pour la représentation de l'état relatif, chacune avec ses avantages. Plusieurs représentations sont
passéesenrevueence chapitre,notammentdans lecadre des orbitesképlériennes. L'accent estmis
sur ladescriptionbasée surles positionsetlesvitessesrelatives,exprimées dansunrepère cartésien
local attaché au satellite cible. Les propriétés des trajectoires relatives sont également étudiées,
comme les distances minimale et maximale entre les satellites, l'existence des trajectoires relative
périodiques et leur propriétés géométriques.
1.1 Introduction
Thespacecraftrelativemotionreferstothestudyofthedynamicsofaspacecraft,calledthefollower,
with respect to the dynamics of another spacecraft, called the leader or the target. The motion
of an individual satellite orbiting the Earth can be expressed using dierent representations for
most common descriptions will be presented in this chapter. The nal choice is usually driven
by the purpose of the study. Historically, models based on orbital elements and orbital elements
dierenceshavebeenusedforformationyingapplications[16,34,60],whileCartesianmodelshave been preferredfor spacecraftrendezvous andcollision avoidance problems [17,31,41,57].
Regardless of the representation chosen for the spacecraft relative motion, a distinction can
be made between Keplerian models and non Keplerian models. Under Keplerian assumptions,
the Earth is represented as an homogeneous sphere and the spacecraft motion is aected only
byNewtonian accelerations. The non Keplerian models take into account theEarth's oblateness,
usually through the spherical harmonic model for the Earth's potential, the atmospheric dragor
thesolar radiation pressure, amongother orbitaldisturbances.
TheKeplerian framework leadsto lessaccurate butsimplieddynamicalmodels for the
space-craftrelative motion. Thesesimpliedmodels arewellsuitedfor controlsynthesispurposes,likein
thecase of maneuversplans design for spacecraft rendezvousmissions for instance. Therelatively
small distances between the spacecraft when compared to thedistance with respect to the center
of the Earth and the short time horizons associated with rendezvous missions justify the usage
of simplied relative motion models. For this reason we will focus mainly on Keplerian models
throughout thisdissertation, while referringthe interested readerto publicationstreating some of
theother representations.
Inwhat follows, aparticular interestwill be paidto theperiodicsolutions of theequations
de-scribing thespacecraft relative motion. Thesesolutions enable thesatellitesto maintaina desired
conguration without external intervention and without any fuelexpenditure. This property has
beenextensivelyusedintheformationightliterature[1,5,46,55,92]andhasrecentlygained atten-tion for orbital rendezvous and collision avoidance applications [25,41,43]. Dierent initialization methods forperiodicmotionwillbepresentedalongwithsomeofthegeometricalpropertiesofthe
resulting trajectories.
1.2 Dynamics of a spacecraft orbiting the Earth
The Keplerian dynamics of a spacecraft with respect to the Earth can be derived from Newton's
equations of motionbetween two massparticles. In this case, themotion of a spacecraft orbiting
theEarth isdescribed bythefollowing dierential equation[8]:
d
2
R
~
dt
2
!
B0
=
−
µ
k ~
R
k
3
~
R
(1.1)where
R
~
representsthevectorfromthecenteroftheEarthtothespacecraftcenterofmassandµ
is theEarth's gravitational constant. The dynamics areexpressedwithrespectto an Earthcenteredinertial frame
R
0
= (0, ~
X, ~
Y , ~
Z)
illustrated in Figure 1.1. The fundamental plane forR
0
is the Earth's equatorial plane, theZ
~
axis coincides with the rotation axisof theEarth and is oriented towards theNorth Pole, theX
~
axis points thevernal equinoxand theY
~
axisis orthogonalto the~
X ~
Z
plane.Figure 1.1: The EarthCentered Inertialframe and thesatellite trajectory
Even thoughthedierential equation (1.1)governing therelative motionof two bodies is
non-linear,theequationadmitsageneralanalyticalsolution[8]. Theconstantsofintegrationassociated to thesolution arecalled theorbital elements of thesatellite motion and they playan important
role inthestudy ofthe properties of thespacecraft trajectory.
Let the orbital plane be the plane which contains the trajectory of the orbiting spacecraft
(see Figure1.1). Theequation ofthespacecraft trajectoryexpressedusing polar coordinateswith
respectto this planeisgiven by [8]:
R =
k ~
R
k =
a(1
− e
2
)
1 + e cos ν
(1.2)where
a
is called the semi-major axis of the spacecraft orbit,e
is called the eccentricity andν
is calledthetrueanomaly. Thesatellite'sorbitisboundedife < 1
andunboundedife
≥ 1
. Fore = 0
the spacecraft trajectory is acircle of radiusa
and for0 < e < 1
the trajectory is an ellipse. The true anomalyν
representstheangle between thespacecraft's current positionand thedirection of theperigee(Figure 1.1).orientation of the orbital plane are required in order to completely characterize the spacecraft
trajectory. A common choice is represented bythe angles
i
,Ω
andω
dened withrespect to the Earth's equatorialplane, asindicatedinFigure 1.2.Figure 1.2: The denitionof theclassicalorbital elements
Theline of nodes denotes theline of intersection between thespacecraftorbital planeand the
equatorial plane. The ascending node refers to thepoint where the satellite iscrossing the line of
nodesinanorthbound direction. Thelongitude of the ascending node,
Ω
,istheangle between theX
axis of theR
0
frame and theascending node, the argument of perigee,ω
,is the angle between theascending node and theperigeewhilethe inclination,i
,is theanglebetween theorbital plane and theequatorialplane.Theset oforbital elementsis dened by:
oe =
h
a e i Ω ω ν
i
T
(1.3)
anditcompletelydescribesthestateofasatelliteorbitingtheEarth. UnderKeplerianassumptions,
therst ve parameters areconstant andonly thetrueanomaly changeswithtime[8]:
˙ν =
r
µ
a
3
(1
− e
2
)
3
(1 + e cos ν)
2
(1.4)
Sometimes, the eccentric anomaly,
E
, or the mean anomaly,M
, are used instead ofν
as the varyingstate. Theeccentric anomalyandthetrueanomalyarerelatedthrough geometricaltrans-formations (Figure 1.3):
tan
ν
2
=
r
1 + e
1
− e
tan
E
2
(1.5)while eccentric anomaly andthemean anomalyare relatedthroughKepler'sequation:
M = E
− e sin E = M
0
+ n(t
− t
0
)
(1.6)As shown in (1.6), the mean anomaly can also be dened as a linear function of time, where
n =
p
µ/a
3
is the mean motion of the satellite,
t
0
is the reference time andM
0
is the mean anomaly att
0
.Figure1.3: Thedenition oftheeccentric anomaly
When the orbit is circular or near circular (
e
≈ 0
) or when theorbit is planar or near planar (i
≈ 0
), some of the classical orbital elementsoe
are not dened. Inthose cases, the state of the spacecraft can be represented using dierent functionsof the classical orbital elements thatavoidthis problem. Among the solutions proposed in the literature, we can mention the nonsingular
orbital elements, the equinoctial elements or the Delaunay canonical elements, used for studying
thesatellite motionina Hamiltonianframework[90].
Thechoiceofusingtheinertialpositionandvelocityorthevarioussetsoforbitalparametersin
ordertodescribethestateofaspacecraftorbitingtheEarthismadedependingontheapplication.
Throughout this dissertation, theclassical orbital elements
oe
arepreferred for therepresentation oftheleader'sstate. Thischoiceismotivatedbythefactthat, intheKepleriancontext consideredhere, the resulting dynamics have a very simple form (only one state that changes over time).
To complete the description of the spacecraft relative motion, the state of the follower satellite
1.3 Spacecraft relative motion
The spacecraft relative motion refers to the study of the dynamics of the leader spacecraft
com-bined with the study of the dynamics of the follower spacecraft. As previously stated, there are
dierent possible state denitions which can be used inthe description of the motion of a single
spacecraft (Cartesian positionand velocity, dierent sets of orbital parameters). In a similarway,
dierent representations can be considered for the spacecraft relative state, each one bearing its
ownadvantages.
1.3.1 Local Cartesian dynamics
The spacecraft relative motionrepresented using local Cartesiandynamics is dened with respect
to a localrotatingCartesianframe centered on theleader satellite. A commonlyusedframeisthe
Local Vertical Local Horizontal (LVLH) frame
R
l
= (S
l
, ~x, ~y, ~z)
illustrated in Figure (1.4). The~z
axis is radially oriented from the leader satellite towards the center of the Earth, the~
y
axis is orthogonal to the orbital plane, inthe oppositedirection with respect to theangular momentumvector, andthe
~x
axislays intheleader's orbitalplane inthedirectionof thesatellite's velocity.PSfragreplacements
~x
~z
ν
a
O
P
~
~
Q
S
l
S
2
~r
Figure 1.4: The spacecraftrelative positionand theleader's LVLHframe
The relative position between theleader spacecraft
S
l
and the follower spacecraftS
f
is repre-sentedby~r =
−−→
S
l
S
f
inFigure 1.4. Considering that theKeplerian dynamics of each satellite with respectto the Earthcan be described using (1.1), therelative inertial acceleration can bewrittenas:
d
2
~r
dt
2
B0
=
−
µ
k ~
R + ~r
k
3
( ~
R + ~r) +
µ
k ~
R
k
3
~
R
3
(1.7)where
~
R =
−−→
OS
f
representstheinertial positionof theleader spacecraft. Theterm onthelefthand side of (1.7) can be furtherdeveloped usingthederivation rulewithrespectto arotating frame:d
2
~r
dt
2
B0
=
d
2
~r
dt
2
B
l
+ 2 ~
Ω
Bl/B0
×
d ~r
dt
B
l
+
d ~
Ω
Bl/B0
dt
!
Bl
× ~r + ~Ω
Bl/B0
×
~
Ω
Bl/B0
× ~r
(1.8)Theterms inthesumcorrespond to thespacecraftrelative acceleration inthelocalframe,the
Euleracceleration, theCoriolisaccelerationandthecentrifugalaccelerationrespectively. Theterm
~
Ω
B
l
/B0
representstherotationvelocity ofthelocal basisB
l
withrespectto theinertial basisB
0
.Assuming that the dynamics of the leader spacecraft areexpressed using the orbital elements
dened in (1.3) and that the spacecraft relative state is given by the local relative position and
velocity
X =
h
x y z v
x
v
y
v
z
i
T
, the dierent terms in (1.8) can be computed individually.
In thecaseof Keplerian motion,wehave:
~
Ω
Bl/B0
=
0
− ˙ν
0
Bl
~
R =
0
0
R
Bl
~r =
x
y
z
Bl
(1.9)Afterintroducing theelements from(1.9),equation (1.8) becomes:
d
2
~r
dt
2
B0
=
¨
x
− 2 ˙ν ˙z − ¨ν z − ˙ν
2
x
¨
y
¨
z + 2 ˙ν ˙x + ¨
ν x
− ˙ν
2
z
B
l
(1.10)Developing therighthand sideof (1.7)leadsto thefollowingnonlinearequations forthe
space-craft relative dynamics:
¨
x
− 2 ˙ν ˙z − ¨ν z − ˙ν
2
x =
−
p
µ x
(x
2
+ y
2
+ (R
− z)
2
)
3
¨
y =
−
p
µ y
(x
2
+ y
2
+ (R
− z)
2
)
3
¨
z + 2 ˙ν ˙x + ¨
ν x
− ˙ν
2
z =
−
p
µ(R
− z)
(x
2
+ y
2
+ (R
− z)
2
)
3
+
µ
R
2
(1.11)In the case where the distance between the two satellites is a lot smaller than the distance
equations can beusedto describe thespacecraft relative motion[101]:
¨
x = 2 ˙ν ˙z + ¨
ν z + ˙ν
2
x
−
µ
R
3
x
¨
y =
−
µ
R
3
y
¨
z =
−2 ˙ν ˙x − ¨ν x + ˙ν
2
z + 2
µ
R
3
z
(1.12)It can be noticed thatfor thelinearizedequations,thedynamics on the
y
axisaredecoupled from thedynamics inthexz
plane and dene aharmonical oscillator.Inthecasewheretheorbitoftheleader spacecraftiscircular,asimpliedformcanbeobtained
for theaboveequations. If
e = 0
thenR = a =
const,˙ν = n =
const andν = 0
¨
. After introducing these values in(1.12),thewell knownHill-Clohessy-Wiltshire equations for thespacecraftrelativemotionwith respectto a circular referenceorbit canbe deduced[23,42]:
¨
x = 2 n ˙z
¨
y =
−n
2
y
¨
z =
−2 n ˙x + 3 n
2
z
(1.13)It can be noticed that in this case the spacecraft relative dynamics correspond to a Linear Time
Invariant system.
The non Keplerian relative dynamics
Long term predictions of thespacecraft relative trajectory arenecessaryfor formation ying
mis-sions. In this case, maintaining the assumption that there are no external perturbing forces or
nonlinear terms introduces unacceptable prediction errors. Therefore, dierent models of
space-craftrelativemotionaccountingforsomeoftheeectsoforbitaldisturbanceshavebeendeveloped.
For circular reference orbits, Schweighart and Sedwick presented in [91] a set of constant-coecient lineardierential equationsthatinclude theperturbationdue totheEarth'soblateness,
representedthroughthe
J
2
potential. HamelanddeLafontainedevelopedin[39]asetoflinearized equations ofrelativemotionaboutaJ
2
perturbed ellipticalreferenceorbit. Kechichiangavein[50] the expression of the rotation velocityΩ
~
Bl/B0
for the case where disturbances due to air drag and Earth oblateness are considered. The resultis very general but itleads to complex nonlinearexpressions for therelative motionthatarenot easy to useinpractice.
Even ifthe dynamics modelled by the Tschauner-Hempel equations (1.12) do not include the
model which is well suited for control synthesis and has been widely used for spacecraft relative
trajectorydesign [6,41,47,86,93,99].
1.3.2 Orbital elements dierences dynamics
The dierential orbital elements are dened asthe dierence between the orbital elements of the
leader spacecraft
oe
l
andtheorbitalelementsof follower spacecraftoe
f
:X
oe
= oe
l
− oe
f
=
h
δa δe δi δΩ δω δν (
orδM
orδE)
i
T
(1.14)
UnderKeplerian assumptions, ve of thesixorbital elementsdening thestate ofa spacecraft
are constant. In this case, the relative dynamics expressed using thedierential orbital elements
exhibitsimilarproperties. Thesimplestformfortherelativedynamicsisobtainedwhenthevarying
term intheorbital elementsis chosen to be themeananomaly
M
:˙
X
oe
=
δ ˙a
δ ˙e
δ˙i
δ ˙
Ω
δ ˙ω
δ ˙
M
=
0
0
0
0
0
−
3
2
r
µ
a
5
δa
(1.15)Variationalmethodscan beusedto analysetheeectofperturbingaccelerationsontheorbital
elementsdescribingthespacecraftmotion,inthenonKepleriancase[90]. Theperturbing accelera-tionscanmodelforinstancetheeectsoftheEarthoblatenessand/ortheeectsoftheatmospheric
drag. The well known GaussVariational Equations (GVE)represent a specicformulationof the
orbitalelements variation problem,written fordisturbancesexpressedintheleader'sLVLHframe.
The spacecraft relative dynamics represented using the orbital elements dierences have been
successfully usedinformation ight applications, especiallyfor congurations that require a large
separation between the spacecraft [1,16]. In the case of the spacecraft rendezvous, the mission's objectivesareusuallyspeciedusingtherelativeCartesianlocalcoordinates,intermsofnal
rela-tive positionandvelocity,givensome position/velocityconstraints. For thisreason thedescription
1.4 Linearized Cartesian relative motion
StartingfromtheTschauner-Hempelequations(1.12)forthelinearizedCartesianrelativedynamics,
a state space representation of the spacecraft relative dynamics can be obtained. Based on this
formulation, closed form solutions for the relative trajectories can be computed. These solutions
enablethepropagationofthespacecraftrelativestatewithoutmakinguseofnumericalintegration,
which makesthemvery valuable for spaceapplications where computationalpowerislimited.
1.4.1 State-space representation
Letthespacecraftrelativestatevectorbedenedbytherelativepositionandvelocityprojectedon
each axis ofthe leader's LVLHframe:
X =
h
x y z v
x
v
y
v
z
i
T
. Ifin (1.12) theindependent
variable time is replaced by the true anomaly of the leader spacecraft, a simplied form can be
obtained for the equations describing the relative dynamics between the leader and the follower
spacecraft. The derivatives withrespectto timearereplaced by:
d(
·)
dt
=
d(
·)
dν
dν
dt
= (
·)
0
˙ν
d
2
()
dt
2
=
d
2
()
dν
2
˙ν
2
+
d()
dν
ν
¨
(1.16)and thefollowing variable change isused:
˜
X(ν) =
(1 + e cos ν)I
3
0
3
−e sin νI
3
(1 + e cos ν)
˙ν
I
3
X(t)
(1.17)where
I
3
∈ R
3×3
is theidentity matrix and
0
3
∈ R
3×3
is the zeromatrix. Thisoperationleads to
a periodicstate-space modelfor thespacecraftrelative dynamics:
˜
X
0
(ν) = ˜
A(ν) ˜
X(ν) + ˜
B ˜
u
(1.18)where thedynamical matrix
˜
A(ν)
isgiven by:˜
A(ν) =
0
0
0
1
0 0
0
0
0
0
1 0
0
0
0
0
0 1
0
0
0
0
0 2
0
−1
0
0
0 0
0
0
3
1 + e cos ν
−2 0 0
(1.19)the control matrix
B
˜
is dened byB = [0
˜
3
I
3
]
T
and
u = [˜
˜
u
x
u
˜
y
u
˜
z
]
T
represents the acceleration
generated bythespacecraft thrusters.
Closedform solutions can becomputed for theperiodicsystem(1.18) and thegeneral method
for obtaining themis summarizednext.
1.4.2 The state transition matrix
The statetransitionmatrixprovidesaconvenientwayto representthesolutionoftheautonomous
dynamics of a linear system. For the spacecraft relative motion, computing the state transition
matrix would enable the propagation of the relative state starting from any initial conditions,
without relying onnumerical integration:
˜
X(ν) = Φ(ν, ν
0
) ˜
X(ν
0
)
(1.20)Fromthedynamics ofthesystem(1.18) ,itcanbededucedthatthestate transitionmatrixveries
thefollowing dierential equation:
Φ
0
(ν, ν
0
) = A(ν)Φ(ν, ν
0
),
Φ(ν, ν) = I
∀ν
(1.21)For Linear Time Varyingsystems such as(1.18), there is no general analytical expression for the
state transition matrix. Numerical methods developed for computing
Φ
are usually based on the resolution of the dierential equation (1.21). In the case of the spacecraft relative motion, thespecialstructure of thedynamicalmatrix
A(ν)
enables thecomputation of ananalytical solution. Forthelinearizedspacecraftrelativemotion,thedynamicsonthey
axisarenotaectedbythe motion in thexz
plane and aredescribed bythe following homogeneous second order dierential equation (see(1.18)):˜
y
00
=
−˜y
(1.22)The solutionof (1.22) can be directlyexpressedasa function oftheinitial conditions:
˜
X
y
(ν) = Φ
y
(ν, ν
0
) ˜
X
y
(ν
0
)
(1.23)where
ν
0
istheinitial true anomalyfor theuncontrolled motionand:˜
X
y
(ν) =
y(ν)
˜
˜
v
y
(ν)
Φ
y
(ν, ν
0
) =
cos(ν
− ν
0
)
sin(ν
− ν
0
)
− sin(ν − ν
0
) cos(ν
− ν
0
)
(1.24)From(1.18),thehomogeneous dierential equations for the
xz
planearegiven by:˜
x
00
= 2˜
z
0
(1.25)˜
z
00
=
3
1 + e cos ν
z
˜
− 2˜x
0
(1.26)Integrating (1.25) onceleads to:
˜
x
0
= 2˜
z + K
(1.27)where Kis aconstant ofintegration. After introducing(1.27)in(1.26), asecond orderdierential
equation only in
˜
z
is obtained:˜
z
00
+
4
−
3
1 + e cos ν
˜
z = K
(1.28)Asrecalled byCarter in[20], themethod for solving this type of dierential equationconsists in nding a family of particular solutions
ϕ
1
,ϕ
2
for the homogeneous dierential equation such that :ϕ
1
ϕ
0
2
− ϕ
2
ϕ
0
1
=
constant (1.29)and then applying the technique of variation of parameters [80]. The choice of the particular solutions
ϕ
1
,ϕ
2
determines thenalformof thetransitionmatrix.A transition matrix for the periodic system (1.18) has been proposed by Carter in [20]. A slightly dierent solution hasbeen given by Yamanaka and Ankersen in [103], which presents the advantage of having a simpler form. The Yamanaka-Ankersen transition matrix will be used for
some of thedevelopments inthisdissertation and itis reproduced herefor completeness.
Taking
˜
X
xz
(ν) =
h
˜
x(ν) ˜
z(ν) ˜
v
x
(ν) ˜
v
y
(ν)
i
T
,thepropagationoftherelative stateisgivenby:
˜
X
xz
(ν) = Φ
xz
(ν, ν
0
) ˜
X
xz
(ν
0
)
(1.30)where thetransition matrix
Φ
xz
(ν, ν
0
)
can bewritten as:The matrix
φ
xz
(ν)
isdened by[103]:φ
xz
(ν) =
1
− cos ν(2 + e cos ν)
sin ν(2 + e cos ν)
3(1 + e cos ν)
2
J
0
sin ν(1 + e cos ν)
cos ν(1 + e cos ν)
2
− 3e sin ν(1 + e cos ν)J
0
2 sin ν(1 + e cos ν)
2 cos ν(1 + e cos ν)
− e
3
− 6e sin ν(1 + e cos ν)J
0
cos ν + e cos 2ν
− sin ν − e sin 2ν
−3e
(cos ν + e cos 2ν)J +
sin ν
1 + e cos ν
(1.32)Theterm
J
isrelatedtothechoiceoftheparticularsolutionϕ
2
andinthecaseofthe Yamanka-Ankersentransition matrix isgivenby:J(ν) =
Z
ν
ν0
dτ
(1 + e cos τ )
2
=
n(t
− t
0
)
(1
− e
2
)
3/2
(1.33)From(1.33)it follows thatfor theinitial trueanomaly
ν
0
wehaveJ(ν
0
) = 0
. Thisenables the analytical computation of theinverse of theφ
xz
(ν)
matrix atν
0
:φ
−1
xz
(ν
0
) =
1
e
2
−1
e
2
−1 −
3e sin ν
0
(2 + e cos ν
0
)
1 + e cos ν
0
e sin ν
0
(2 + e cos ν
0
)
2
−e cos ν
0
(1+e cos ν
0
)
0
3 sin ν
0
(e cos ν
0
+1+e
2
)
1 + e cos ν
0
− sin ν
0
(2 + e cos ν
0
)
−(cos ν
0
+e cos
2
ν
0
−2e)
0
3(e + cos ν
0
)
−(2 cos ν
0
+e cos
2
ν
0
+e)
sin ν
0
(1 + e cos ν
0
)
0
−(3e cos ν
0
+ e
2
+ 2)
(1 + e cos ν
0
)
2
−e sin ν
0
(1 + e cos ν
0
)
(1.34)The complete transition matrix
Φ
corresponding to the state vectorX(ν)
˜
can be obtained by combining theblocks fromtheΦ
y
andΦ
xz
matrices inthe appropriateorder.Overview of closed form solutions
Several works have been dedicated to thecomputation of the transition matrix for the spacecraft
relativemotion,inthecasewheretheleadersatelliteevolvesonanarbitraryellipticalorbit. Melton
provides in[70]a solution that uses directlythe timeas theindependent variable, obtained using seriesexpansionsoftheeccentricity. However,thisisanapproximatesolutionanditlosesaccuracy
for higher values of the eccentricity. Recently, a transition matrix obtained starting from the
Tschauner-Hempelequationsthatalsoincludestheeectsofthe
J
2
perturbationhasbeenproposed byYamadaandKimurain[102]. Thegiven solutioniscumbersomeandnot easyto usefor control design purposes. Moreover, the obtained transition matrix is shown to be accurate only for shortprediction horizons.
motionhavebeenpresentedin[33,39]. Theyarebasedontheconnectionbetween thelocal Carte-sian relative state and thedierential orbital elementsand no longer require theresolution of the
dierential equations ofmotion. Gim and Alfriend consider in[33]both the short-period and the long-periodeectsofthe
J
2
perturbation,leadingtoaveryaccuratebutcomplexsolutionthatstill requirestheknowledgeoftheevolutionoftheorbitalparametersfortheleadersatellite. Hamelandde Lafontaine simplify theproblemin[39]byneglecting theshort-term eects of
J
2
. Theyobtain a solution thatguarantees abounded prediction erroreven for longhorizonsbut thatrequirestheknowledgeof therelativesecular driftof themeanorbital elements.
Closed form solutions of the spacecraft relative dynamics are sought for the computational
advantage obtained from removing the integration process from thetrajectory design algorithms.
Moreover, they can also provide some insight into thegeometrical properties of theresulting
tra-jectories. Some examples of trajectory parametrizations that have been derived from such closed
form solutions will be presented inthenextsection.
1.5 Properties of relative trajectories
The spacecraft ability to maintain a naturally periodic relative motion has been thoroughly
in-vestigated, especially in the context of formation ight applications. Some of the initialisation
techniques for obtaining periodic solutions to the equations of spacecraft relative motion will be
presentednext,along withsome ofthegeometrical propertiesof theresulting trajectories.
The connection between the initial conditions of the periodic motion and the dimensions of
the obtained trajectory bears a lot of importance in the mission design process. The estimation
of theminimal distance between the spacecraft is essential for collision avoidance purposes while
theevaluationof themaximaldistanceplaysan important role inthechoice ofthesensorsfor the
relative navigation. However, sucient understandingofthisconnectionhasnot yetbeen reached.
The next sections summarize some interesting results found in the literature in relation to this
topic.
1.5.1 Periodicity conditions
The distance between two spacecraft on Keplerian orbits cannot grow unboundedly [37]. This observationisbasedonthefactthatintheKeplerian casethespacecraftevolveontrajectoriesthat
arebounded anddonotchangeovertime. However,unlesssome particularconditionsaremet,the
motionbetween spacecraft evolving onorbitsthat verifythefollowing condition:
p T
l
= q T
f
, p, q
∈ N
(1.35)where
T
l
andT
f
aretheorbitalperiodsoftheleaderandthefollowerspacecraftrespectively. Since:T = 2π
s
a
3
µ
,
(1.36)thecondition(1.35)canbeeasilytransformed into aconditiononthesemi-majoraxisoftheorbits
correspondingto thetwo spacecraft:
a
f
=
3
s
p
2
q
2
a
l
(1.37)or in a condition between the energy of theorbits. The restriction in (1.37) induces a restriction
ontherelativetrajectory. Figure1.5illustratesthetrajectoryobtainedbypropagatingtherelative
motion over 10 orbital periods for dierent ratios between the orbital periods of two spacecraft.
The relative trajectoryappears to layon a closed surface whose shape anddimensions depend on
theratio chosenbetween theorbital periods.
−12
−10
−8
−6
−4
−2
x 10
6
−5
0
5
x 10
6
−2
−1
0
1
2
x 10
6
y [m]
x [m]
z [m]
p=2, q=3
−12
−10
−8
−6
−4
−2
x 10
6
−6
−4
−2
0
2
4
6
x 10
6
−2
−1
0
1
2
x 10
6
y [m]
x [m]
z [m]
p=5, q=7
Figure1.5: Relative trajectories obtained for dierent ratiosbetween theorbital periods
In thecasewhere
p = q = 1
,constraint (1.37) becomes:a
f
= a
l
(1.38)In thiscase, therelative trajectorybetween thetwo spacecraftis periodic(see Figure1.6).