Generating Periodic Orbits in The Circular Restricted Three Body Problem With Applications To Lunar South Pole Coverage

GENERATING PERIODIC ORBITS IN THE CIRCULAR RESTRICTED THREE-
BODY PROBLEM WITH APPLICATIONS TO LUNAR SOUTH POLE COVERAGE
A Thesis
Submitted to the Faculty
of
Purdue University
by
Daniel J. Grebow
In Partial Fulfillment of the

Requirements for the Degree
of
Master of Science in Aeronautics and Astronautics
May 2006
Purdue University
West Lafayette, Indiana
ii
“Happy is the man that findeth wisdom; and the man that getteth
understanding. For the merchandise of it is better than the merchandise of
silver, and the gain thereof than fine gold. She is more precious than
rubies: and all the things thou canst desire are not to be compared unto
her. Length of days is in her right hand; and in her left hand riches and
honour. Her ways are ways of pleasantness, and all her paths are peace.
She is a tree of life to them that lay hold upon her: and happy is every one
that retaineth her. The LORD by wisdom hath founded the earth; by
understanding hath he established the heavens. By his knowledge the
depths are broken up, and the clouds drop down the dew.”
(Proverbs 3:13-20) KJV
iii
ACKNOWLEDGEMENTS
Socrates pursued wisdom, knowledge, and understanding. However, in Plato’s Meno,

Socrates paradoxically argues for the existence of knowledge a priori. That is, he argues
that knowledge does not come from experience but rather originates from a “spontaneous
recovery” through recollection. He proves this by posing basic geometry questions to a
young uneducated slave. The boy answers Socrates’ geometry questions correctly!
Since the boy is uneducated, his knowledge of geometry must have come from within.
Therefore, knowledge a priori must exist.
There is a flaw in the argument, of course. Socrates has failed to realize that the
nature of his own questioning has led the boy to the correct answers. That is, Socrates
has asked the questions in such a way that the uneducated boy has become educated. The
knowledge did not originate from recollection, but rather was passed from Socrates to the
boy through Socrates’ own subtle questioning.
Let me not make the same mistake. The knowledge imparted by this work did not
simply originate from within, but came from others. For this reason, I have many people
that need to be thanked.
Most definitely, Dr. Kathleen Howell has provided a strong foundation for my
knowledge in orbital mechanics. If there is value in this work, she is the creditor. She is
the Socrates and I am the uneducated young boy. It is by her subtle questioning that I
have become educated. It is my privilege to work for her.
It is also my privilege to work with Dave Folta at NASA Goddard Spaceflight Center.
In fact, Dave is a great counterexample to Socrates’ argument. Nobody possesses more
knowledge of orbits based on experience than Dave. It was his idea to investigate lunar
south pole coverage, not mine.
Much credit must also be given to Marty Ozimek. Marty and I literally worked side-
by-side at Goddard and Purdue in designing the architectures for lunar south pole
coverage. In fact, many of the ideas conveyed in Chapter 4 come directly from him. I
hope to work more with him in the future.
iv
Some of my other colleagues must also be thanked. Chris Patterson has helped me
understand so many things; there is no way I can list them all. For example, he taught me
how to use GENERATOR. Furthermore, Lucia Irrgang and I initially worked together
on computing periodic orbits near L4 and L5. I hope also to work more with Lucia in the
future.
Others also need to be thanked. Belinda Marchand initially instructed me with
modifications of the two-level corrector to include constraints. The STK® three-body
scenarios originally came from Frank Vaughn at Goddard. Analytical Graphics, Inc.
representative Paul Black has been an immediate help whenever problems were
encountered with STK.
I would also like to thank Purdue University and NASA Goddard Spaceflight Center
under Contract Number NNG04GP69G, for providing the support necessary to complete
this work.
v
TABLE OF CONTENTS
Page
LIST OF TABLES ...................................................................................................... viii
LIST OF FIGURES .................................................................................................... x
ABSTRACT ................................................................................................................ xiii
1. INTRODUCTION ................................................................................................ 1
1.1 Problem Definition ....................................................................................... 2
1.1.1 The Circular Restricted Three-Body Problem ................................. 2
1.1.2 Lunar South Pole Coverage ............................................................. 2
1.2 Previous Contributors .................................................................................. 3
1.2.1 Quasi-Periodic and Periodic Solutions ............................................ 4
1.2.2 Application to Lunar South Pole Coverage and Station-Keeping
Strategies .......................................................................................... 6
1.3 Current Work ............................................................................................... 7
2. BACKGROUND .................................................................................................. 10
2.1 The Circular Restricted Three-Body Problem ............................................. 10
2.2 The Equilibrium Points ................................................................................ 13
2.3 First-Order Analytic Approximations for Variations Relative to the
Collinear Points ............................................................................................ 15
2.3.1 Quasi-Periodic, First-Order Approximations Near the Collinear
Libration Points ................................................................................ 18
2.3.2 Periodic, First-Order Approximations Near the Collinear Libration
Points ................................................................................................ 19
2.4 Fundamental Variational Relationships ....................................................... 21
2.4.1 The State-Transition Matrix ............................................................. 21
2.4.2 Contemporaneous and Non-Contemporaneous Variations in the
State .................................................................................................. 23
vi
Page
2.4.3 An Introduction to Invariant Manifold Theory ................................ 25
3. GENERATING ORBITS IN THE CIRCULAR RESTRICTED THREE-BODY
PROBLEM ............................................................................................................ 27
3.1 The Two-Level Differential Corrections Process ........................................ 27
3.1.1 The “First Level” of the Corrections Process .................................. 28
3.1.2 The “Second Level” of the Corrections Process .............................. 30
3.1.3 Numerical Example: Lissajous Trajectories ................................... 34
3.2 A Simple Corrector ...................................................................................... 35
3.2.1 The Algorithm .................................................................................. 35
3.2.2 Numerical Results: L1, L2, and L3 Lyapunov Families of Orbits ..... 38
3.2.3 The Stability Index and Bifurcations ............................................... 45
3.3 Targeting Orbits in Three-Dimensional Space Using Symmetry ................ 48
3.3.1 Strategies to Compute Periodic Orbits .......................................... 48
3.3.2 Numerical Results: L1, L2, L3 Halo, Axial, and Vertical Families
of Orbits ........................................................................................ 53
3.4 Asymmetric Corrections Schemes ............................................................... 75
3.4.1 Strategies to Compute Periodic Orbits .......................................... 75
3.4.2 Numerical Results: L4/L5 Planar, Axial, and Vertical Families of
Orbits ............................................................................................. 83
3.5 Alternative Methods ..................................................................................... 90
3.5.1 Sequential Quadratic Programming: Completing the L3 Halo
Orbit Family .................................................................................. 90
3.5.2 Two-Level Corrections with Constraints: Completing the L4
Vertical Orbit Family .................................................................... 93
4. ORBIT SELECTION FOR LUNAR SOUTH POLE COVERAGE AND THE
TRANSITION TO A FULL EPHEMERIS MODEL ........................................... 101
4.1 Orbit Selection for Lunar South Pole Coverage .......................................... 101
4.1.1 Period and Stability Index ............................................................. 105
4.1.2 Potential Orbits for Coverage of the Lunar South Pole ................ 108
vii
Page
4.2 Transition to a Full Ephemeris Model ......................................................... 110
4.2.1 Obtaining Results with the Purdue Software Package
GENERATOR .............................................................................. 110
4.2.2 A Preliminary Coverage Analysis ................................................ 112
4.2.3 Architectures for Lunar South Pole Coverage .............................. 113
5. COVERAGE AND STATION-KEEPING ANALYSES FOR LUNAR SOUTH
POLE COVERAGE .............................................................................................. 120
5.1 Coverage Analysis ....................................................................................... 120
5.1.1 Defining the Ground Stations and Establishing the Links ............ 120
5.1.2 Results for Architectures Supporting Lunar South Pole
Coverage ....................................................................................... 127
5.2 Station-Keeping Analysis ............................................................................ 139
5.2.1 The Station-Keeping Algorithm ................................................... 139
5.2.2 Station-Keeping Results ................................................................ 141
6. RESULTS AND RECOMMENDATIONS .......................................................... 142
6.1 Summary of Results ..................................................................................... 142
6.2 Recommendations for Future Work ............................................................. 143
LIST OF REFERENCES ............................................................................................ 146
VITA ........................................................................................................................... 152
viii
LIST OF TABLES
Table Page
2.1 Locating the Collinear Libration Points .......................................................... 14
3.1 L1 Lyapunov Family Initial Conditions .......................................................... 40
3.4 Northern L1 Halo Family Initial Conditions ................................................... 56
3.7 Northeast L1 Axial Family Initial Conditions ................................................. 62
3.8 Northwest L2 Axial Family Initial Conditions ................................................ 64
3.9 Northeast L3 Axial Family Initial Conditions ................................................. 66
3.10 L1 Vertical Family Initial Conditions .............................................................. 68
3.13 Northern L2 Butterfly Family Initial Conditions ............................................. 74
3
3.14 L4 Planar Family Initial Conditions with y0 = ......................................... 85
2
3.15 Northern L4 Axial Family Initial Conditions with z0 = 0.1 ............................ 87
3.16 L4 Vertical Family Initial Conditions with y0 = 0.42545 ............................... 89

3.17 Northern L3 Family Initial Conditions Using SQP (Orange) .......................... 92
3.18 L4 Vertical Family Initial Conditions Using Two-Level Corrector with
Constraints (Green) ......................................................................................... 100
4.1 Potential Orbits for Coverage of the Lunar South Pole .................................. 109
5.1 Percent Access Times for 12-Day L1 Halo Orbits and Ground Station at the
Shackleton Crater ............................................................................................ 122
ix
Table Page
5.2 Percent Access Times for 12-Day L1 Halo Orbits and Ground Station at the
White Sands Test Facility ............................................................................... 125
5.3 Percent Access Times for 7-Day L1 Halo Orbits ............................................ 130
5.4 Percent Access Times for 8-Day L1 and L2 Halo Orbits ................................. 132
5.5 Percent Access Times for 16-Day L1 and L2 Vertical Orbits .......................... 134
5.6 Percent Access Times for 14-Day L2 Butterfly Orbits .................................... 136
5.7 Percent Access Times for 7-Day L2 Halo Orbit and 14-Day L2 Butterfly
Orbit ................................................................................................................ 138
5.8 Station-Keeping Results for One Year (~ 24 Revs) ........................................ 141
x
LIST OF FIGURES
Figure Page
2.1 Formulation of the Restricted Three-Body Problem ...................................... 11
2.2 Equilibrium or Libration Points (Magenta) in the CR3BP ............................. 14
2.3 First-Order Approximation to a Lissajous Trajectory .................................... 19
2.4 First-Order Approximation to a Lyapunov Orbit ............................................ 20
2.5 Representation of Contemporaneous and Non-Contemporaneous Variations
in the State ....................................................................................................... 24
3.1 Path Before (Red) and After (Black) First Level is Applied .......................... 29
3.2 Path Before Second Level is Applied (Black) and Final Continuous Path
(Blue) .............................................................................................................. 33
3.3 Differentially Corrected L1 Lissajous Trajectory ............................................ 34
3.4 Initial and Final Path for Simple Corrector ..................................................... 36
3.5 The L1 Lyapunov Family ................................................................................ 39
3.8 Characteristic Multipliers Associated with Center Subspace for L1
Lyapunov Orbits ............................................................................................. 47
3.9 The Northern L1 Halo Family ......................................................................... 55
3.12 The Northeast L1 Axial Family ....................................................................... 61
3.13 The Northwest L2 Axial Family ...................................................................... 63
3.14 The Northeast L3 Axial Family ....................................................................... 65
3.15 The L1 Vertical Family .................................................................................... 67
xi
Figure Page
3.18 The Northern L2 Butterfly Family ................................................................... 73
3.19 The L4 Planar Family ...................................................................................... 84
3.20 The Northern L4 Axial Family ........................................................................ 86
3.22 Path Before Second Level is Applied (Black) and Final Periodic Solution
(Blue) .............................................................................................................. 96
3.23 The L4 Vertical Orbits Using a Two-Level Corrector with Constraints ......... 99
4.1 Southern Halo Orbit Family: Earth-Moon L1 (Orange) and L2 (Blue) ........... 102
4.2 Vertical Orbit Families of Interest: Earth-Moon L1 (Red) and L2 (Cyan) ...... 103
4.3 Southern L2 Butterfly Orbit Family ................................................................ 104
4.4 Definition of Maximum xm -Distance ............................................................. 106
4.5 Period versus Maximum xm -Distance from the Moon ................................... 106
4.6 Stability Index versus Maximum xm-Distance from the Moon ....................... 107
4.7 12-Day L1 Orbit (Orange) and Patch Points (Black) CR3BP ......................... 109
4.8 Two Phased Spacecraft in 12-Day L1 Halo Orbits from GENERATOR ........ 111
4.9 zm-Displacement in the Rotating Reference Frame for Two Spacecraft in 12-
Day L1 Halo Orbits .......................................................................................... 112
4.10 Two Phased Spacecraft in 7-Day L2 Halo Orbits from GENERATOR .......... 115
4.11 Two Phased Spacecraft in 8-Day L1 and L2 Halo Orbits from
GENERATOR ................................................................................................ 116
4.12 Two Phased Spacecraft in 16-Day L1 and L2 Vertical Orbits from
GENERATOR ................................................................................................ 117
4.13 Two Phased Spacecraft in 14-Day L2 Butterfly Orbits from GENERATOR . 118
4.14 Two Phased Spacecraft in a 7-Day L2 Halo Orbit and a 14-Day L2
Butterfly Orbit ................................................................................................. 119
5.1 STK 12-Day L1 Halo Orbits ............................................................................ 121
5.2 A Facility is Placed on the Moon at the Shackleton Crater (89.9°S, 0.0°E) ... 122
5.3 Line-of-Sight Access Beams (Green) with Lunar Facility ............................. 123
xii
Figure Page
5.4 Elevation of Each Satellite Above the Horizon as Viewed from the
Shackleton Crater Facility for 12-Day L1 Halo Orbit Scenario ...................... 124
5.5 An STK Facility is Placed at the White Sands Test Facility
(32.3°N, 106.8°W) .......................................................................................... 125
5.6 Possible Communications Relay Between the White Sands Test Facility and
the Lunar Ground Station at the Shackleton Crater ........................................ 126
5.7 STK 7-Day L2 Halo Orbits .............................................................................. 129
Shackleton Crater Facility for 7-Day L2 Halo Orbit Scenario ........................ 130
5.9 STK 8-Day L1 and L2 Halo Orbits .................................................................. 131
Shackleton Crater Facility for 8-Day L1 and L2 Halo Orbit Scenario ............. 132
5.11 STK 16-Day L1 and L2 Vertical Orbits ........................................................... 133
Shackleton Crater Facility for 16-Day L1 and L2 Vertical Orbit Scenario ...... 134
5.13 STK 14-Day L2 Butterfly Orbits ..................................................................... 135
Shackleton Crater Facility for 14-Day L2 Butterfly Orbit Scenario ............... 136
5.15 STK 7-Day L2 Halo and 14-Day L2 Butterfly Orbits ...................................... 137
Shackleton Crater Facility for 7-Day L2 Halo and 14-Day L2 Butterfly Orbit
Scenario .......................................................................................................... 138
5.17 Schematic for Station-Keeping Algorithm ..................................................... 140
6.1 Earth-Moon L1 Vertical Orbit Associated Unstable (Red) and Stable (Blue)
Manifolds ........................................................................................................ 144
xiii
ABSTRACT
Grebow, Daniel J. M.S.A.A., Purdue University, May 2006. Generating Periodic Orbits
in the Circular Restricted Three-Body Problem with Applications to Lunar South Pole
Coverage. Major Professor: Dr. Kathleen C. Howell.
A potential ground station at the lunar south pole has prompted studies of orbit
architectures that ensure adequate coverage. The creation of multi-body orbit
constellations begins with a thorough investigation of periodic orbits in the Circular
Restricted Three-Body Problem (CR3BP). Strategies for the determination of periodic
solutions in the CR3BP enhance capabilities and define boundaries. Families of planar,
vertical, and axial orbits are computed that are associated with all five libration points.
Additional periodic solutions in the vicinity of the collinear libration points suggest
favorable geometries for lunar south pole coverage.
A detailed analysis of halo and vertical families, as well as other orbits near the
libration points in the vicinity of the Moon, suggests that constant communications can be
achieved with two spacecraft in combinations of Earth-Moon libration point orbits. In
particular, the investigation focuses on nine different orbits from these families with
periods ranging from 7 to 16 days. Natural solutions are generated in a full ephemeris
model including solar perturbations. Possible ground stations on the Moon and on the
Earth are established for coverage results and verification of communications
capabilities. Preliminary station-keeping costs are also computed for long term
communications with the ground stations.
1
1. INTRODUCTION
Current limitations in a number of technical areas, including mission design

capabilities, inhibit a thorough exploration of space. Thus, a demand exists for more
innovative ideas in orbital design and more sophisticated computational tools. Increased
complexity of the dynamical model, including additional gravity forces acting on the
spacecraft, offers an infinite variety of trajectories. A thorough investigation of these
trajectories aids the discovery process and facilitates space exploration.
Modeling the motion of the spacecraft by including the gravitational forces associated
with the Sun, Earth, and Moon has historically proven effective in obtaining new
scientific knowledge about the Sun. For example, NASA’s Third International Sun/Earth
Explorer (ISEE-3) mission supplied valuable information regarding solar flares and
gamma-ray bursts [1]. Furthermore, the recent Genesis spacecraft retrieved solar wind
particles over a span of two years [2] that are now being analyzed. The shapes of the
trajectory arcs used by ISEE-3 and Genesis spacecraft are advantageous for gathering
scientific data and for a communications relay with the Earth. The basic models for both
missions include the gravitational forces of the Sun, Earth, and Moon in the analysis.
Since the President’s Vision for Space Exploration [3] announcement in January
2004, preparations are underway for autonomous and manned activities on the lunar
surface. For the development of the trajectory design options for a return to the Moon,
incorporation of the gravity effects for not only the Moon, but also the Earth and Sun,
offers very useful information for spacecraft motion in the vicinity of the Moon. The
results allow flexibility for mission design and ultimately facilitate exploration of Mars.
2
1.1 Problem Definition
1.1.1 The Circular Restricted Three-Body Problem
The motion of a less massive body about an additional body with significantly larger
mass is formulated as a two-body problem in orbital mechanics. Closed form analytical
solutions exist and are conic in nature. However, the inclusion of an additional (third)
body in the gravity model, that is, the general three-body problem, is not solvable
analytically in closed form.
The general three-body model can be simplified to greatly reduce the complexity of
the problem, however. The third body, or spacecraft, is assumed to possess infinitesimal
mass in comparison with the other two bodies, or primaries. Furthermore, it is assumed
that the motion of the two primaries is Keplerian. Additional assumptions require that
the distance between the two primaries remain constant, thereby restricting the motion of
the primaries to be circular relative to the system barycenter. The motion of the third
body is described relative to a rotating reference frame determined by the motion of the
primaries. This, of course, is the classical formulation of the Circular Restricted Three-
Body Problem (CR3BP).
Even with the simplifying assumptions, the CR3BP is still not solvable in closed
form. Particular solutions do, however, exist in the three-body problem. For example, it
is known that five equilibrium solutions or libration points exist in the rotating frame.
Furthermore, approximate analytical solutions are available for motion in the vicinity of
the equilibrium points by analyzing variations with respect to the equilibrium solutions.
These solutions serve as a basis for numerical computation. Moreover, due to the
sensitivities inherent in the CR3BP, the development of numerical procedures is
challenging. Nevertheless, the solution space can be explored thereby leading to the
discovery of new types of trajectories.
1.1.2 Lunar South Pole Coverage
Current interest in autonomous and manned exploration studies is focused primarily

on the south pole of the Moon due to constant exposure to sunlight and the Shackleton
Crater site located nearby. NASA has also indicated that water ice at the lunar poles may
3
help facilitate exploration of the solar system [3]. A ground station on the Moon requires
a system of satellites that is always in view of the Earth such that constant
communications between the lunar surface and the Earth is available.
Rather than adjusting current mission objectives due to limitations inherent in the
original constellations, a libration point orbit can be considered such that current mission
design requirements are fully satisfied. However, many orbit families remain unknown
and the practical application of most known families of trajectories has not yet been
thoroughly investigated. In addition, an optimal design may involve combinations of
trajectories from various families – some yet to be exposed. Identification of the best
libration point orbit or combinations of orbits that fulfill the primary mission objectives
as the first step in the design process requires an extensive study of the dynamical theory
regarding periodic orbits in the CR3BP.
From the characteristics of the various types of periodic solutions, some useful
architectures for lunar south pole coverage exist within the context of the three-body
problem. A thorough investigation of a wide range of libration point orbits offers many
different coverage options. Of course, long term mission requirements demand that
station-keeping costs also be computed.
1.2 Previous Contributors
Much of the dynamical analysis in mechanics for the three-body problem was
developed by Sir Isaac Newton (1643-1727) in 1687 with the Principia. Newton’s
predecessor, Leonhard Euler (1707-1783), was the first to simplify the problem by
assuming an infinitesimal mass for the third body, thereby introducing the “restricted”
three-body problem. Euler was also the first to model the dynamics in a rotating
reference frame. Meanwhile, Joseph-Louis Lagrange (1736-1813), a contemporary of
Euler, identified the five equilibrium solutions, or libration points, in the CR3BP. His
solutions correctly predicted the existence of the Sun-Jupiter Trojan Asteroids. Nearly a
hundred years after Euler and Lagrange, Jules Henri Poincaré (1854-1912) demonstrated
the non-integrability of the three-body problem. His work ultimately became the
foundation of modern dynamical systems theory.
4
1.2.1 Quasi-Periodic and Periodic Solutions
One of many significant contributions, Poincaré proved that periodic orbits exist in
the three-body problem. Moreover, Poincaré claimed that there are an infinite variety of
periodic solutions [4]. In 1920, Moulton [5] published Periodic Orbits, a collection of
studies from more than five authors; the book is solely devoted to analytical methods for
approximating periodic motion near all five equilibrium points. In the fourth chapter of
this work, Buck employs a power series method to develop third-order approximations
for solutions near the triangular equilibrium points. His solutions resemble what are now
known as vertical orbits.
Due to the increasing speed and accuracy of modern computers, numerical methods
for the computation of many different types of periodic solutions have evolved in the last
fifty years. In 1965, Hénon [6] published a series of treatises on stability and bifurcations
for in-plane periodic solutions. His work established an important link between families
of solutions in the restricted three-body problem. Near this same time, Goodrich [7]
obtained numerical solutions for in-plane periodic motion in the vicinity of the triangular
equilibrium points in the Earth-Moon system. Bray and Goudas [8] obtained numerical
results for vertical orbits near the collinear libration points in 1966. Furthermore, the
results by Goodrich were confirmed a year later (1967) when Deprit, Henrard, Palmore,
and Price [9] computed similar motion in the vicinity of the triangular points. In this
same year, Szebehely [10] published Theory of Orbits: The Restricted Problem of Three
Bodies. Included in his study is an extensive compilation of both analytical and
numerical analyses in the restricted three-body problem.
In 1973, Hénon [11] expanded his analysis to include vertical stability of periodic
solutions in the restricted three-body problem. Meanwhile, Farquhar and Kamel [12]
obtained third-order approximations for quasi-periodic motion near the trans-lunar
libration point in the Earth-Moon system using a Linstedt-Poincaré method. A few years
later, Richardson and Cary [13] developed, via a method of successive approximations
truncated to the fourth-order, a model for quasi-periodic motion in the vicinity of the
interior libration point for the Sun-Earth/Moon system.
5
Markellos and Halioulias [14] created a corrections scheme in 1977 that quickly
computes asymmetric periodic solutions for the Störmer problem. The scheme proves
invaluable for obtaining solutions with no visible symmetries. Furthermore, Markellos
successfully applied the corrections scheme to the CR3BP and the stability of the
solutions was assessed using the methods developed by Hénon [15]. In 1979, Zagouras
and Kazantzis [16] numerically determined halo, axial, and vertical families near the
collinear points in the Sun-Jupiter system. Furthermore, Breakwell and Brown [17]
obtained numerical results in the Earth-Moon system for halo orbits near the collinear
points while Robin and Markellos [18] catalogued eight different families of orbits for
the Sun-Jupiter case in the vicinity of Jupiter. In 1984, using a three-dimensional method
of regularization originally developed by Kusstenheimo and Stiefel, Howell and
Breakwell [19] applied it in the restricted problem and developed approximations for
“almost rectilinear” halo orbits near the collinear points. A thorough numerical
investigation of halo orbits was also completed [20]. A year after Howell and Breakwell
developed their approximations, Zagouras [21] improved Buck’s approximations to
include fourth-order terms. He also developed a corrections scheme to obtain numerical
results for vertical and axial orbits near the triangular equilibrium points.
A method for obtaining numerical results for quasi-periodic solutions in the CR3BP
was accomplished by Howell and Pernicka [22] in 1988. The method is adaptable for
multiple flight regimes and easily modified for constrained motion [23-24]. For example,
periodic solutions are available from a more general arc simply by constraining
periodicity [25]. In general, the method remains one of the most powerful tools in
transitioning quasi-periodic and periodic solutions from the CR3BP to a full ephemeris
model with only small variations in shape.
More recently, Papadakis and Zagouras [26] studied bifurcations of families near the
triangular libration points. In 1999, Howell and Campbell [27] completed an extensive
investigation of periodic solutions and bifurcations in the Sun-Earth/Moon system for the
collinear libration points in the Earth/Moon vicinity. Dichmann, Doedel, and Paffenroth
[28] have obtained numerical results for connections between the Lyapunov, halo, axial,
and vertical orbits in the vicinity of the collinear points in the Earth-Moon system using
6
the software AUTO. In 2003, Doedel, Paffenroth, Keller, Dichmann, Galán-Vioque, and
Vanderbauwhede [29] further applied AUTO to successfully map periodic solutions near
the collinear points to those in the vicinity of the triangular points. Their study also
includes numerical results for axial and vertical orbits in the vicinity of the triangular
points.
1.2.2 Application to Lunar South Pole Coverage and Station-Keeping Strategies
In 1971, Farquhar [30] initially examined the use of ‘halo’ orbits to maintain a
communications link between the far side of the Moon and the Earth. His study includes
a communications architecture and station-keeping costs. Later, extensive work on
optimal station-keeping strategies using Floquet modes for halo orbits in the Earth-Moon
system was completed by Simό, Gómez, Libre, Martínez, and Rodríguez [31-32]. In this
approach, the unstable subspace that is available from dynamical systems theory is used
to develop a station-keeping strategy. Station-keeping analyses for the Earth-Moon halo
orbits were also completed by Howell and Pernicka [33], Howell and Keeter [34], and
Gόmez et al. [35]. Most recently, Scheeres, Han, and Hou [36] as well as Renault and
Scheeres [37] have investigated the generalized optimal placement of statistical control
maneuvers applied to orbits in the Earth-Moon restricted three-body problem. The orbits
in these studies are typical of those that might be used for lunar coverage and also
provide an additional benchmark for the station-keeping costs.
The potential use of libration point orbits for lunar south pole coverage has not yet
been fully investigated. Ely and Lieb [38-39] have investigated the placement of a
system of satellites to support a south pole station and global orbital constellations, but
their analysis is based on a two-body model with the third body effects modeled as
gravitational perturbations. They also incorporate solar radiation pressure. On a broader
scale, NASA’s Living with a Star (LWS) program was created to learn more about the
Sun-Earth system. The initiative included the possible investigation of north and south
“pole-sitters” for constant Earth atmospheric monitoring and surveillance [40]. Such
ideas may be adaptable to the Moon for feasible south pole architectures. Studies of
satellite constellations in Earth orbit are also being examined for application to the Moon.
7
1.3 Current Work
For application to the lunar south pole coverage problem, the initial design begins
with the CR3BP for the Earth-Moon system. First-order linear approximations for small
amplitude Lyapunov orbits near the collinear points are used to obtain families of
Lyapunov, halo, axial, and vertical orbits. In addition, a family of orbits that may be
described as a butterfy shape are computed. The exact bifurcation points along a family
of periodic orbits are available by applying a method of bisections. Furthermore,
strategies for computing and expanding the solution space are presented. The strategies
ultimately lead to a mapping of planar, axial, and vertical orbits in the vicinity of the
triangular points.
The southern halo families, small amplitude vertical families, and a butterfly family
in the vicinity of the cis- or trans-lunar libration points are conducive to lunar south pole
coverage. Orbits from these families are selected based on (i) feasibility for lunar south
pole coverage, (ii) predetermined altitude constraints; and, (iii) time to complete one full
period. Once a specific architecture is selected, the orbit is discretized into a series of
patch points. With modifications to a corrections scheme posed by Marchand, Howell,
and Wilson [25], the patch points are then differentially corrected to meet both the
desired time-of-flight and the orbit periodicity requirements. The solutions are
transitioned to a full ephemeris model, including solar perturbations, using the Purdue
Software Package GENERATOR [41]. Adapting models created by Folta and Vaughn
[42], the patch points that define a libration point orbit from GENERATOR are targeted
in Satellite Tool Kit®. A facility is placed at the Shackleton Crater near the south pole of
the Moon and access times between the facility and the spacecraft constellation are
analyzed. Access times with the Earth, specifically the White Sands Test Facility, are
also computed. Finally, a preliminary station-keeping analysis using invariant manifold
theory is accomplished.
8
This study is organized in the following manner:
Chapter 2:
In this chapter, the equations of motion are derived. The differential equations are
formulated within the context of the CR3BP. A method for computation of the system
equilibrium points is included. First-order analytical approximations are summarized for
quasi-periodic and periodic motion near the system equilibrium points and the state-
transition matrix is defined. Contemporaneous and non-contemporaneous variations
along the path are described and invariant manifold theory is introduced.
Chapter 3:
A two-level differential corrections scheme is derived to numerically determine quasi-
periodic motion near the system equilibrium points. Analytical approximations serve as
the initial guess. Various methods and strategies for numercially computing periodic
orbits and the associated families of solutions in the CR3BP are explored. Many
different families are represented, including the planar, axial, and vertical families in the
vicinity of all five equilibrium points.
Chapter 4:
Orbits with characteristics that support lunar south pole communications coverage are
identified for mission applications. Criteria for evaluation include feasibility of coverage,
period, and certain stability parameters. Nine different orbits from five different families
with periods ranging from 7 to 16 days are transferred to a full ephemeris model. Five
different coverage scenarios for two spacecraft are selected and preliminary coverage
analyses are summarized.
Chapter 5:
Line-of-sight access times between a lunar ground station at the Shackleton Crater
(89.9°S, 0.0°E) and dual spacecraft constellations that correspond to five coverage
scenarios are analyzed with advanced commercial software. General line-of-sight access
between the spacecraft constellation and the Earth is summarized. More specifically,
9
line-of-sight access between a spacecraft constellation and a possible transmitting site

located at the White Sands Test Facility, in New Mexico (32.3°N, 106.8°W) is computed.
Preliminary station-keeping results for the nine orbits are also developed by application
of invariant manifold theory.
Chapter 6:
The results are summarized and recommendations for future work are offered.
10
2. BACKGROUND
The differential equations describing the motion of a body of infinitesimal mass

subject to the gravity of two much more massive primaries are nonlinear and cannot be
solved analytically in closed form. However, numerical integration yields unexpected
solutions: non-planar, periodic orbits in the vicinity of the two primaries but orbiting
neither one, three-dimensional “figure-eight” trajectories, and an infinite variety of other
shapes. The differential equations describing the motion of the small body are
formulated in terms of the classical Circular Restricted Three-Body Problem (CR3BP) in
this chapter. The equilibrium points in this system are easily computed. First-order
analytical approximations are derived from the variational equations relative to the
equilibrium points and the state transition matrix is introduced. Contemporaneous and
non-contemporaneous variations in the state are defined; invariant manifold theory is
introduced.
2.1 The Circular Restricted Three-Body Problem
The position of a small body (P3 with mass m3) relative to two primaries (P1 with
mass m1 and P2 with mass m2) is depicted in Figure 2.1. Let Ri be the position of Pi with
respect to the system barycenter B. Using Newton’s Law of Gravity, the equations
describing the motion of P3 are governed by the vector differential equation,
I
d 2 R3 Gm3 m1 Gm3 m2
m3 2
=− 3
R13 − 3
R23 , (2.1)
dt R R
13 23
where G is the universal gravitational constant. Then, R13 and R23 are the relative
positions of P3 with respect to P1 and P2. The superscript I represents differentiation in
the inertial frame. The mass is restricted such that m1 > m2 >> m3. This implies that m3 is
“infinitesimal,” i.e., P3 cannot impact the motion of m1 and m2. Thus, the motion of P1
and P2 is Keplerian. Further, the motion of P2 relative to P1 is assumed to be circular.
11
ˆ
Y
ŷ
P3 ( m3 )
R23 x̂
R3
R2 P2 ( m2 )
R13
θ
R1 ˆ
X
B
P1 ( m1 )
Figure 2.1 Formulation for the Restricted Three-Body Problem
Equation (2.1) can be generalized by introducing characteristic quantities. Let the

characteristic length l* be defined such that
l * = R1 + R2 . (2.2)
Then, the characteristic mass m* is the sum of the primary masses, i.e.,
m* = m1 + m2 , (2.3)
and the characteristic time t* is computed as
l *3
t* = . (2.4)
Gm*
Since the motion of P2 about P1 is assumed to be circular, l* is constant. Using equations
(2.2)-(2.4) to non-dimensionalize equation (2.1) yields
I
d 2 r3 1− μ μ
=− r − r , (2.5)
dτ 2
r13
3 13
r23
3 23
R3 R R m
where r3 = *
, r13 = 13*
, r23 = *23 , the mass ratio μ = 2* , and the non-dimensional
l l l m
t
time τ = . Thus, the right side of equation (2.5) is fully defined.
t*
12
The left side of equation (2.5) can be expanded kinematically. For convenience, the
motion of P3 is described in terms of the barycentric, rotating frame S where x̂ is parallel
to r2 , ẑ is parallel to the angular velocity vector associated with the Keplerian primary
orbits, and ŷ completes the right-handed set. (The unit vectors associated with the S
frame are denoted as grey in Figure 2.1.) Therefore,
r3 = xxˆ + yyˆ + zzˆ, (2.6)
where x, y, and z are the components of position of P3 relative to B, defined in terms of
the rotating frame. The angular velocity of the rotating frame with respect to the inertial
frame is written Iω S = nzˆ , where the non-dimensional mean motion is unity, i.e., n = 1.
Therefore, twice differentiating equation (2.6) with respect to an inertial observer yields
the kinematic expression,
I
r3 = ( x − 2 y − x ) xˆ + ( y + 2 x − y ) yˆ + zzˆ. (2.7)
The dots in equation (2.7) represent differentiation with respect to τ. Furthermore,

expressing r13 and r23 in terms of μ, x, y, and z yields
r13 = r3 − r1 = ( x + μ ) xˆ + yyˆ + zzˆ, (2.8)
r23 = r3 − r2 = ( x − (1 − μ ) ) xˆ + yyˆ + zzˆ. (2.9)
Finally, the vector equation (2.5) can be rewritten in scalar form as
x − 2y − x = −
(1 − μ )( x + μ ) − μ ( x − (1 − μ ) ) , (2.10)
3 3
r13 r23
y + 2x − y = −
(1 − μ ) y − μ y , (2.11)
3 3
r13 r23
z=−
(1 − μ ) z − μ z . (2.12)
3 3
r13 r23
The motion of P3 relative to B in the rotating frame is determined by numerically

integrating equations (2.10)-(2.12).
13
2.2 The Equilibrium Points
The system equilibrium points, also called the Lagrange or libration points, are
determined from equations (2.10)-(2.12). The relationships that yield the constant
equilibrium solutions are the following,
(1 − μ ) ( xeq + μ ) μ ( xeq − (1 − μ ) )
− xeq = − 3
− 3
, (2.13)
r13 eq
r23 eq
(1 − μ ) yeq μ yeq
− yeq = − 3
− 3
, (2.14)
r13 eq
r23 eq
(1 − μ ) zeq μ zeq
0=− 3
− 3
, (2.15)
r13 eq
r23 eq
where rij indicates the norm of a position vector evaluated at the equilibrium point.
eq
Five equilibrium points satisfy equations (2.13)-(2.15). From equation (2.15), it is

evident that each equilibrium point possesses an out-of-plane component, zeq = 0 .
Furthermore, if r13 eq
= r23 eq
= 1, then yeq ≠ 0, as is evident from equations (2.13)-
(2.14). Using equations (2.8)-(2.9) to solve r13 eq

= r23 eq
= 1 renders the solution
1 3
xeq = − μ and yeq = ± . These equilibrium points, i.e., L4 and L5, are the triangular
2 2
or equilateral points since the points exist at the vertices of two equilateral triangles with
a common side, i.e., the straight line adjoining P1 to P2. (See Figure 2.2.) The remaining
three equilibrium points, i.e., L1, L2, and L3, are computed by defining yeq = 0 in
equations (2.13)-(2.14). Since these three equilibrium points all lie on the x-axis, they are
commonly labeled the collinear points. The form of the resulting equation is
1− μ μ
xeq + A +B = 0, (2.16)
xeq + μ xeq − 1 + μ
where the values of A and B are ±1 depending on the location of the equilibrium point.
The values are summarized in Table 2.1. An iterative method to solve equation (2.16) for
xeq is straightforward and, thus, locates the positions of L1, L2, and L3 along the x-axis.
14
Table 2.1
Locating the Collinear Libration Points
Libration Point Location A B
L1 − μ < xeq < 1 − μ −1 1
L2 xeq > 1 − μ −1 −1
L3 xeq < − μ 1 1
ŷ L4
60°
L3 L1 L2
B
x̂
P1 ( m1 ) P2 ( m2 )
60°
L5
Figure 2.2 Equilibrium or Libration Points (Magenta) in the CR3BP

15
2.3 First-Order Analytical Approximations for Variations Relative to the

Collinear Libration Points
The first-order analytical approximations for motion relative to the collinear points
are derived via linearization with respect to the equilibrium point. The first-order
variational equations can be solved to yield a useful approximation. As an initial step,
introduce the following symbols,
ξ = x − xeq , (2.17)
η = y − yeq , (2.18)
ζ = z − zeq . (2.19)
Therefore ξ, η, and ζ are the variations relative to the equilibrium point defined in terms
of its measure numbers xeq, yeq, and zeq. Then, linearizing equations (2.10)-(2.12) about
the equilibrium solution yields
ξ − 2η = U xx eq ξ + U xy eq η + U xz eq ζ , (2.20)
η + 2ξ = U yx eq ξ + U yy eq η + U yz eq ζ , (2.21)
ζ = U zx eq ξ + U zy eq η + U zz eq ζ , (2.22)
∂ 2U
where U ij is the second partial defined by . The symbol U ij indicates that the
∂j∂i eq
partials are evaluated at the equilibrium point of interest. Since zeq = 0 for all the
equilibrium points, U xz eq
= U yz = U zx eq
= U zy = 0 and U zz eq
< 0. Also, for the
eq eq
collinear points, the equilibrium point is always on the x-axis, i.e., yeq = 0, and, therefore,
U xy = U yx = 0, U xx eq
> 0, and U yy < 0. Then, for the collinear points, equations
eq eq eq
(2.20)-(2.22) are simplified to

ξ − 2η = U xx eq ξ , (2.23)
η + 2ξ = U yy eq η , (2.24)
ζ = U zz eq ζ . (2.25)
16
Notice that equation (2.25), i.e., the out-of-plane motion, is simple harmonic with
frequency v = U zz eq
and independent of the in-plane motion. Therefore, the general
solution for ζ can be written in the following form,

ζ = C1 cos vτ + C2 sin vτ , (2.26)
where C1 and C2 are constants. Correspondingly, ξ and η are coupled for the in-plane
motion. Thus, the characteristic equation associated with equations (2.23)-(2.24) is of the
form
Λ 2 + 2β1Λ − β 2 2 = 0, (2.27)
where
U xx eq
+ U yy
β1 = 2 − eq
, (2.28)
2
β 2 2 = − U xx eq U yy eq > 0, (2.29)
λ = ± Λ. (2.30)
Then, λ represents the eigenvalues associated with the coupled equations (2.23)-(2.24).
The general solutions for ξ and η are
4
ξ = ∑ Αi e λ τ , i
(2.31)
i =1
4
η = ∑ Βi e λ τ , i
(2.32)
i =1
where Ai and Bi are constants, but not independent. Direct substitution of equations
(2.31)-(3.32) into equations (2.23)-(2.24) results in the following relationship between Ai
and Bi, that is,
λi − U xx eq
Βi = Αi . (2.33)
2λi
Differentiation of equations (2.31)-(2.32) with respect to τ , substitution of equation
(2.33), and evaluation of the result at the initial time, τ = τ 0 = 0, yields
4
ξ 0 = ∑ Αi , (2.34)
i =1
17
4
ξ 0 = ∑ Αi λi , (2.35)
i =1
4 λi − U xx eq
η0 = ∑ Αi , (2.36)
i =1 2λi
4 λi − U xx eq
η0 = ∑ Αi λi . (2.37)
i =1 2λi
Expressions for Ai in terms of ξ0 , ξ 0 , η0 , η0 , λi , and U xx eq

are obtained by inversion
of equation (2.34)-(2.37). In general, as evidenced by equation (2.30), the system

possesses one unstable mode, one stable mode, and two center modes. Therefore, if A1
and A2 correspond to the stable and unstable modes, then by specifying ξ 0 and η0 , ξ 0
and η0 can be specified such that Α1 = Α 2 = 0 . Then ξ , η , and ζ can be written in the
form,
η0
ξ = ξ 0 cos sτ + sin sτ , (2.38)
β3
η = η0 cos sτ + ξ 0 β 3 sin sτ , (2.39)
ζ0
ζ = ζ 0 cos vτ + sin vτ , (2.40)
v
where
s = β1 + ( β12 + β 2 2 ) 2 ,
1
(2.41)
s 2 − U xx
β3 = eq
. (2.42)
2s
2π
Notice that the period of the motion is simply . From the linear approximation, the
s
three-dimensional motion of P3 is not periodic since the in-plane and out-of-plane
frequencies are not commensurate. However, s and v are relatively close in value for the
problem of interest. This suggests that quasi-periodic motion can be approximated.
18
2.3.1 Quasi-Periodic, First-Order Approximations Near the Collinear Libration

Points
Equations (2.38)-(2.40) can be further simplified to produce approximate conditions

for quasi-periodic, Lissajous motion in terms of discrete points along the path. The
conditions are determined simply by specifying an amplitude Aξ in the ξ direction or an
amplitude Aη in the η direction; then, an amplitude Aζ is specified in the ζ direction.
If it is desired to specify Aξ and Aζ , then equations (2.38)-(2.40) can be simplified to the
following form,
ξ = − Aξ cos ( sτ + φ ) , (2.43)
η = β3 Aξ sin ( sτ + φ ) , (2.44)
ζ = Aζ sin ( vτ + ψ ) . (2.45)
Alternatively, if it is desired that Aη and Aζ are inputs, then equations (2.38)-(2.40) can
be simplified to
Aη
ξ =− cos ( sτ + φ ) , (2.46)
β3
η = Aη sin ( sτ + φ ) , (2.47)
ζ = Aζ sin ( vτ + ψ ) . (2.48)
The phase angles φ and ψ define the starting point along the Lissajous trajectory.
Discretizing τ and differentiating equations (2.43)-(2.45) or (2.46)-(2.48) produces
approximate positions and velocities at discrete points, or patch points, along the path.
Ten revolutions of a first-order approximation to a Lissajous trajectory near L1 appear in
Figure 2.3 as plotted in the Earth-Moon system; φ = 180° , ψ = 90° , Aη = 15,000 km, and
Aζ = 20, 000 km are input with four patch points per revolution. The period
corresponding to each revolution is approximately 11.7471 days.

19
0.06
0.04
Earth-Moon System
0.02
1 unit = 385,692.5 km
0
η
μ = 0.0121505856
−0.02
−0.04
−0.06
−0.05 0 0.05
ξ
0.06 0.06
0.04 0.04
0.02 0.02
0 Earth Moon 0
ζ
−0.02 ζ −0.02
−0.04 −0.04
−0.06 −0.06 L1
−0.05 0 0.05 −0.05 0 0.05

ξ η
Figure 2.3 First-Order Approximation to a Lissajous Trajectory
2.3.2 Periodic, First-Order Approximations Near the Collinear Libration Points
According to Poincaré [4], an infinite variety of periodic solutions exist in the three-
body problem. Because equations (2.10)-(2.12) are not solvable in closed form,
computation of the periodic solution is challenging. Many of these solutions have been
studied by various researchers [7-10, 15-21, 26-29]. Most of these studies are based on
particular solutions obtained using different approaches. One way to obtain a particular
periodic solution in the nonlinear problem is to first generate a particular periodic
solution in the linear system.
20
Even planar periodic orbits in the nonlinear system are a challenge to compute
without an initial guess. A first-order, planar linear approximation is available, however.
Restricting the motion to be planar in the linear system, i.e., when Aζ = 0 in equations
(2.45) and (2.48), provides accurate approximations for small amplitude Aξ and Aη . A
single patch point is positioned on the x-axis corresponding to the angles φ = 180° ,
ψ = 90°. For the approximation in Figure 2.4, Aη = 15,000 km for a period of 11.7471
days. The resulting planar orbit is the approximation to a finite Lyapunov orbit. Other
linear approximations for particular solutions that are three-dimensional are available by
specifying the in-plane and out-of-plane frequencies to be commensurate.
0.05
Earth-Moon System
0 1 unit = 385,692.5 km
η
μ = 0.0121505856
Initial
Point
−0.05
−0.05 0 0.05
ξ
0.05 0.05
0 0
ζ
Earth Moon
L1
−0.05 −0.05
−0.05 0 0.05 −0.05 0 0.05
ξ η
Figure 2.4 First-Order Approximation to a Lyapunov Orbit

21
2.4 Fundamental Variational Relationships
The analytical approximations are only accurate for motion close to the equilibrium
point. However, other particular, numerical solutions exist for periodic motion not in the
vicinity of the equilibrium point. For example, families of large amplitude halo orbits
have been computed that originate in the vicinity of the collinear points [16, 17, 19-20,
27, 28]. Furthermore, families of solutions have been determined that ‘connect’ the
motion near the collinear points to motion in the vicinity of triangular points [7, 9, 21, 26,
29]. The stability of the solutions provides information regarding the phase space, and
the phase space is manipulated to determine other bounded motion as well as the flow to
and from periodic orbits.
2.4.1 The State-Transition Matrix
The state-transition matrix is a linear map between state variations along the path.
Since the system is conservative, a pseudo-potential function U can be defined, that is,
1− μ μ 1 2
U= + + ( x + y2 ) , (2.49)
r13 r23 2
such that
x − 2y − x = Ux, (2.50)
y + 2x − y = U y , (2.51)
z = Uz , (2.52)
∂U ∂U ∂U
where U x = , Uy = , and U z = . Let the associated six element state vector
∂x ∂y ∂z
q be defined as q = { x z} . The first-order variational equations are

T
y z x y
derived and result in the vector differential equation,

δ q = A (τ ) δ q. (2.53)
Since the matrix A (τ ) is evaluated along the reference solution, it is generally not
constant for an arbitrary trajectory as a reference solution. However, if the reference is a

22
periodic solution, then A (τ ) is periodic as well. The general form of the 6 × 6 matrix
A (τ ) is summarized in terms of four 3 × 3 submatrices,
⎡ 03×3 I 3×3 ⎤
A (τ ) = ⎢ ⎥, (2.54)
⎢⎣ U XX Ω ⎥
⎦
⎡U xx U xy U xz ⎤
⎢ ⎥
where submatrices U XX and Ω are defined as U XX = ⎢U yx U yy U yz ⎥ and
⎢U zx U zy U zz ⎥
⎣ ⎦
⎡0 2 0⎤
Ω = ⎢⎢ −2 0 0 ⎥⎥ , respectively. The components of U XX are expressed as functions of
⎣⎢ 0 0 0 ⎥⎦
∂ 2U
the states such that U ij = . The 6 × 6 state-transition matrix results in the following
∂j∂i
partials,
⎡ ∂x ∂x ∂x ∂x ∂x ∂x ⎤
⎢ ∂x ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂y ∂y ∂y ∂y ∂y ∂y ⎥
⎢ ∂x ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂z ∂z ∂z ∂z ∂z ∂z ⎥
⎢ ⎥
∂x0 ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
Φ (τ + τ 0 ,τ 0 ) = ⎢ , (2.55)
⎢ ∂x ∂x ∂x ∂x ∂x ∂x ⎥
⎢ ⎥
⎢ ∂x0 ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂y ∂y ∂y ∂y ∂y ∂y ⎥
⎢ ⎥
⎢ ∂x0 ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂z ∂z ∂z ∂z ∂z ∂z ⎥
⎢ ⎥
⎢⎣ ∂x0 ∂y0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥⎦
and can be evaluated from the governing matrix differential equation such that
Φ (τ + τ 0 ,τ 0 ) = A (τ ) Φ (τ + τ 0 ,τ 0 ) . (2.56)
The well-known form of the solution to equations (2.56) is

δ q = Φ (τ + τ 0 ,τ 0 ) δ q0 , (2.57)
23
where δ q0 is the vector of initial variations at τ 0 and the elements of the vector δ q
represent the variations at τ + τ 0 . Simultaneous numerical integration of equation (2.56)

and the equations of motion, i.e., equations (2.10)-(2.12), produces the evolving
components of Φ (τ + τ 0 ,τ 0 ) . Furthermore, from equation (2.57) it is evident that
Φ (τ 0 ,τ 0 ) = I 6×6 . The state-transition matrix supplies valuable information regarding both
the stability of the system and predictions for variations along the path.
2.4.2 Contemporaneous and Non-Contemporaneous Variations in the State
Predictions of future variations along the path are critical for the development of
targeting schemes to obtain periodic orbits. To predict variations along the path, let qi be
a six-element state vector that denotes the position and velocity states associated with the
ith point along the path as seen in Figure 2.5. If f represents a mapping via numerical
integration from point i to the next point i + 1, then
qi +1 = f ( qi , Δτ ) , (2.58)
where Δτ = τ i +1 − τ i . Then, the non-contemporaneous state variation corresponding to the
i th + 1 point, i.e., δ qi +1 , must satisfy
(
qi +1 + δ qi +1 = f qi + δ Qi ,Δτ + δτ i +1 , ) (2.59)
where δ Qi represents a contemporaneous variation in the state for the ith point along the
path. (See Figure 2.5.) Using a first-order Taylor series expansion, equation (2.58)
becomes
∂f ∂f
qi +1 + δ qi +1 = f ( qi , Δτ ) + δ Qi + δτ i +1. (2.60)
∂qi qi , Δτ
∂τ qi , Δτ
∂f ∂f
It is apparent that = Φ (τ i +1 ,τ i ) and = qi +1. Thus, in substituting equation
∂qi qi , Δτ
∂τ qi , Δτ
(2.58) into equation (2.60), the first-order approximation for the variations δ qi +1 is
rewritten as
24
⎧⎪ δ Q ⎫⎪
δ qi +1 = ⎡⎣ Φ (τ i +1 ,τ i ) qi +1 ⎤⎦ ⎨ i ⎬ . (2.61)
⎩⎪δτ i +1 ⎭⎪
Equation (2.61) and equation (2.57) together comprise an important fundamental
relationship between non-contemporaneous and contemporaneous variations in a state
vector, i.e.,
δ Qi = δ qi − qiδτ i . (2.62)
This vector relationship is also apparent in Figure 2.5. Furthermore, substituting equation
(2.61) into equation (2.60) and rearranging results in the following useful expression,
δ qi +1 − qi +1δτ i +1 = Φ (τ i +1 ,τ i ) (δ qi − qiδτ i ) . (2.63)
Equation (2.61) is the foundation of any targeting scheme when the problem is
independent of the time associated with the initial state, i.e., τi. Equation (2.63) serves as
the basis for more complex targeting schemes where there exists a more specific
dependence on τi, e.g., simultaneously predicting variations in position and time for a
series of discrete patch point states along the path.
qiδτ i
δ Qi
δ qi Nearby Path
qi
qi +1δτ i +1
Current Path
δ Qi +1 δ qi +1
qi +1
δτ i δτ i +1
τ
Figure 2.5 Representation of Contemporaneous and Non-Contemporaneous

Variations in the State
25
2.4.3 An Introduction to Invariant Manifold Theory
The phase space offers information regarding the characteristic flow near particular
solutions. The center, unstable, and stable subspaces are available from analysis of the
phase space and are useful in obtaining various types of trajectory arcs. The invariant
manifolds associated with the unstable and stable subspaces of periodic and quasi-
periodic solutions, in particular, are key factors in understanding and designing transfers
toward and away from these orbits. Perko [43] includes a thorough investigation of
invariant manifold theory as it applies to equilibrium points and periodic orbits.
For a periodic orbit, the monodromy matrix is defined as the state-transition matrix
after one full period T, i.e., Φ (T + τ i ,τ i ) . Since the orbit is periodic, any point along the
path can be represented by one six-element state vector along the path. Such a
representative point is labeled as the “fixed point.” Then, the initial time τ i corresponds
to the ith fixed point along the path of a periodic solution. Of course, information
concerning the phase space relative to this fixed point is available from the eigenstructure
of Φ (T + τ i ,τ i ) . If the eigenstructure is such that the periodic orbit possesses an unstable
and a stable mode, then the eigenvalues and eigenvectors can be used to develop linear
approximations to the unstable subspace Wu and stable subspace Ws. If Wu and Ws are
available, they are further exploited to generate the global unstable and stable manifolds.
Then, let the six-dimensional vector YîWu represent the eigenvector associated with the
unstable mode of Φ (T + τ i ,τ i ) such that it can be decomposed in terms of three-
dimensional components RiWu and ViWu as follows,
⎧⎪ RWu ⎫⎪
YîWu = ⎨ iW ⎬ . (2.64)
⎩⎪Vi ⎭⎪
u
Then, the unstable direction is represented by

YîWu
X iWu = . (2.65)
RiWu
26
A small perturbation in the unstable direction, X iWu , places the spacecraft on an unstable
manifold departing the vicinity of the reference solution. The perturbing terms are
represented by
⎧⎪δ RiWu ⎫⎪
⎨ Wu ⎬ = d ⋅ X i u ,
W
(2.66)
⎪⎩δ Vi ⎪⎭
where d is an initial displacement in the unstable direction. The unstable manifold for the
ith point along path of a periodic solution is obtained by perturbing the corresponding
{ δ ViW T } and integrating forward in time. The stable manifold is

T
states by ± δ RiWu T u
obtained using the same analysis but with the stable subspace and integrating backward
in time.
27
3. GENERATING ORBITS IN THE CIRCULAR RESTRICTED

THREE-BODY PROBLEM
In 1892, Poincaré showed that an infinite number of periodic solutions exist in the
three-body problem. Many of the solutions remain unknown and the practical application
of most of the known periodic solutions has not yet been thoroughly investigated.
However, due to the nature of the solutions, perhaps the most useful orbit architectures
can be pieced together within the context of the three-body problem. Therefore,
generating orbits in the CR3BP is a critical step in constructing a useful trajectory for the
mission design process. A thorough investigation of orbits in the CR3PB is useful in
generating a trajectory such that the current mission design requirements are fully
satisfied. Due to the sensitivities of the problem, the acquisition of many different
algorithms for convergence aids the discovery process. Furthermore, an understanding of
the stability of the solutions and their bifurcations yields knowledge about the geometry
and intersection of invariant manifolds.
The determination of quasi-periodic motion in the CR3BP is based upon a general
two-level corrections scheme. Additional periodic solutions can be computed. Five
different methods for generating periodic solutions are included in this chapter.
Strategies for predicting initial guesses for solutions and locating bifurcations along a
manifold are offered. Many different types of orbit families are generated, including
planar, axial, and vertical families of orbits near all five libration points. A general
assessment of the stability of these solutions is also summarized.
3.1 The Two-Level Differential Corrections Process
The two-level corrector was first applied to the CR3BP in 1986, by Howell and
Pernicka [22]. The corrections scheme locates a natural solution in the vicinity of an
initial guess comprised of multiple segments and is therefore a powerful tool in the
computation of quasi-periodic solutions in the CR3BP; it is also a significant aid in
28
bringing highly sensitive solutions to a full ephemeris model with only small variations in
characteristics. The quasi-periodic solutions of interest here are denoted as Lissajous
trajectories, a type of motion named after the French physicist Jules Antoine Lissajous
(1822-1880) who discovered similar motions when optically studying vibrations. The
two-level corrections scheme is also easily modified for constrained motion and
adaptable for multiple flight regimes [23-25].
3.1.1 The “First Level” of the Corrections Process
The goal of a two-level corrections scheme is the determination of a trajectory with

specified characteristics. The “first level” of the process is focused on position
continuity. Velocity discontinuities are inserted to ensure a continuous path. Figure 3.1
illustrates this first step in the process. In the figure, a path is represented as a set of
individual segments numbered 1, 2,… , n. Recall that the points that eventually connect
the segments are termed patch points. The red curves reflect initial approximations for
each segment. These “initial guesses” can be obtained analytically or numerically.
The first level of the two-level corrector allows variations in initial velocity on a
segment such that endpoint positions available from the approximations are exactly
targeted. (See the black path in Figure 3.1.) For each segment, i.e., between every ith and
i th + 1 patch point, equation (2.61) is used in an iterative process to solve for variations in
initial velocity and time that reduce the final point variations in position to zero. Then,
let Vi be the velocity vector associated with the ith patch point and τ i +1 the initial time for
integration to the i th + 1 patch point. Then, Vi and τ i +1 are iteratively updated using
⎡ ∂x ∂x ∂x ⎤
⎢ ⎥
⎢ ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂y ∂y ∂y ⎥ ⎪⎧ δ V ⎪⎫
δ Ri +1 = ⎢ Vi +1 ⎥ ⎨ i ⎬ , (3.7)
⎢ ∂x0 ∂y0 ∂z0 ⎥ ⎪⎩δτ i +1 ⎪⎭
⎢ ∂z ∂z ∂z ⎥
⎢ ⎥
⎣ ∂x0 ∂y0 ∂z0 ⎦
L
where the smallest Euclidean norm yields
29
⎧⎪ δ Vi ⎫⎪ T −1
⎬ = L ⎡⎣LL ⎤⎦ δ Ri +1 ,
T
⎨ (3.8)
⎪⎩δτ i +1 ⎪⎭
until δ Ri +1 < ε . As a result, for n patch points, there will exist new velocities for the first
n – 1 patch point states (all nodes but the final patch point n). Since time is tagged at the
endpoint along each segment, new times associated with the last n – 1 patch points will
be determined. Of course, time continuity is preserved across the entire path and each
point is fully targeted. There will also exist velocity discontinuities, or ΔV ’s, at each
internal patch points, i.e.,
ΔVi = Vi + − Vi − , (3.9)
where ‘+’ and ‘–’ denote characteristic quantities associated with outgoing and incoming
segments. In Figure 3.1, the final continuous path across all segments is black.
i
Vi − Vi +
i −1 i +1
Vi −+1 Vi −−1
V3+ Vn−−2
3 Internal Patch n−2
−
V
3
Points V+
n−2
V2+ Vn−−1
2 V2− n −1
Vn+−1
V1 Vn
1 n
Figure 3.1 Path Before (Red) and After (Black) First Level is Applied
30
3.1.2 The “Second Level” of the Corrections Process
The “second level” of the corrections process focuses on minimizing the ΔV ’s

associated with the first level, thereby locating a natural solution in the vicinity of the
initial guess. That is, the second level simultaneously shifts each patch point position and
time, i.e., δ Ri and δτ i , to minimize the ΔV ’s inserted in the first level. To accomplish
the task, it is necessary to express variations in ΔVi as functions of δ Ri and δτ i . From

equation (3.9), the velocity discontinuity identified with the ith patch point depends on
both the outgoing i th − 1 and incoming i th + 1 states. (See Figure 3.1) Therefore, the
relationship between δ Ri , δτ i , and ΔVi can be written as
⎧δ Ri −1 ⎫
⎪ ⎪
⎪ δτ i −1 ⎪
⎪⎪ δ R ⎪⎪
δΔVi = ⎡⎣ M R− M τ−i M Ri Mτi M R+i M τ+i ⎤⎦ ⎨ i ⎬ . (3.10)
⎪ δτ i ⎪
i
Mi ⎪δ R 1 ⎪
⎪ i+ ⎪
⎩⎪ δτ i +1 ⎭⎪
The state-relationship matrix Mi is identified via the partials of equation (3.9) with
respect to position and time such that
∂ΔVi ∂ΔVi ∂Vi − ∂Vi −
M R−i = = = − , (3.11)
∂Ri −1 ∂Vi − ∂Ri −1 ∂Ri −1
∂ΔVi ∂ΔVi ∂Vi − ∂Vi −

M τ−i = = = − , (3.12)
∂τ i −1 ∂Vi − ∂τ i −1 ∂τ i −1
∂ΔVi ∂ΔVi ∂Vi − ∂ΔVi ∂Vi + ∂Vi − ∂Vi +

M Ri = = + =− + , (3.13)
∂Ri ∂Vi − ∂Ri ∂Vi + ∂Ri ∂Ri ∂Ri
∂ΔVi ∂ΔVi ∂Vi − ∂ΔVi ∂Vi + ∂Vi − ∂Vi +

Mτi = = + = − + , (3.14)
∂τ i ∂Vi − ∂τ i ∂Vi + ∂τ i ∂τ i ∂τ i
∂ΔVi ∂ΔVi ∂Vi + ∂Vi +

M R+i = = = , (3.15)
∂Ri +1 ∂Vi + ∂Ri +1 ∂Ri +1
+ ∂ΔVi ∂ΔVi ∂Vi + ∂Vi +

Mτi = = = . (3.16)
∂τ i +1 ∂Vi + ∂τ i +1 ∂τ i +1
31
Each of the resulting partials in equation (3.11)-(3.16) must be further reduced for
evaluation. The partials of Vi − and Vi + with respect to the i th − 1, i th , i th + 1 positions and
times are derived from the linear variational equations. Using equation (2.63), the linear
variational equations are written,
⎧⎪δ Ri −1 − Vi −+1δτ i −1 ⎫⎪ ⎡ A i −1,i Bi −1,i ⎤ ⎧⎪ δ Ri − Vi δτ i +1 ⎫⎪
−
⎨ + ⎬=⎢ ⎨ ⎬, (3.11)
⎪⎩δ Vi −1 − Vi −1δτ i −1 ⎪⎭ ⎣ Ci −1,i
+ Di −1,i ⎥⎦ ⎪δ V − − V −δτ ⎪
⎩ i i i +1 ⎭
and
⎧⎪δ Ri +1 − Vi +−1δτ i +1 ⎫⎪ ⎡ A i +1,i Bi +1,i ⎤ ⎧⎪ δ Ri − Vi δτ i +1 ⎫⎪
+
⎨ − ⎬=⎢ ⎨ ⎬, (3.12)
⎪⎩δ Vi +1 − Vi +1δτ i +1 ⎪⎭ ⎣ Ci +1,i
− Di +1,i ⎥⎦ ⎪δ V + − V +δτ ⎪
⎩ i i i +1 ⎭
⎡ A i −1,i Bi −1,i ⎤ ⎡ A i +1,i Bi +1,i ⎤

where Φ (τ i −1 ,τ i ) = ⎢ and Φ (τ i +1 ,τ i ) = ⎢ . Notice that
⎣ Ci −1,i Di −1,i ⎥⎦ ⎣ Ci +1,i Di +1,i ⎥⎦
determination of Φ (τ i −1 ,τ i ) requires integrating backwards in time. This numerical step
can be avoided simply by recognizing that Φ (τ i −1 ,τ i ) = Φ (τ i ,τ i −1 ) . Isolating the state

−1
variations of interest in equations (3.11) and (3.12) results in the following expressions,
∂Vi −
= Bi−−11,i , (3.13)
∂Ri −1
∂Vi −
= −Bi−−11,iVi −+1 , (3.14)
∂τ i −1
∂Vi −
= −Bi−−11,i A i −1,i , (3.15)
∂Ri
∂Vi −
= Vi − + Bi−−11,i A i −1,iVi − , (3.16)
∂τ i
∂Vi +
= −Bi−+11,i A i +1,i , (3.17)
∂Ri
∂Vi +
= Vi + + Bi−+11,i A i +1,iVi + , (3.18)
∂τ i
∂Vi +
= Bi−+11,i , (3.19)
∂Ri +1
32
∂Vi +
= −Bi−+11,iVi +−1. (3.20)
∂τ i +1
Additional scalar variational relationships are added to equation (3.10) such that the
variations in the ΔV ’s are clearly expressed in terms of position and time variations
along incoming and outgoing segments, i.e.,
⎧ δ R1 ⎫
⎪ ⎪
⎪ δτ 1 ⎪
⎪ δ R2 ⎪
⎪ ⎪
⎪ δτ 2 ⎪
⎧ δΔV2 ⎫ ⎪ δR ⎪
⎪ ⎪ ⎪ 3
⎪
⎪ δΔV 3 ⎪ ⎪ δτ 3 ⎪
⎪ ⎪ ⎪ ⎪
⎨ ⎬ = [M ] ⎨ ⎬, (3.21)
⎪δΔV ⎪ ⎪δ R ⎪
⎪ n−2 ⎪ ⎪ n−2 ⎪
⎪⎩δΔVn −1 ⎭⎪ ⎪ δτ n − 2 ⎪
⎪ ⎪
⎪δ Rn −1 ⎪
⎪ δτ n −1 ⎪
⎪ ⎪
⎪ δ Rn ⎪
⎪ δτ ⎪
⎩ n ⎭
where
⎡ M R− M τ−1 M R M τ1 M R+ M τ+1 0 ⎤
⎢ 1 −
1
−
1
⎥
0 0
⎢ 3×3 3×1 M R2
M τ2 M R2
Mτ 2 ⎥
M=⎢ ⎥ . (3.22)
⎢ M R M τ n−1 M R+ M τ+n−1 03×3 03×1 ⎥
⎢ n−1 n−1
⎥
0 M R−n M τ−n M Rn M τ n M R+n M τ+n ⎥
⎣⎢ ⎦
Note that M is a banded matrix. Such a form lends tractability to the solution of equation
(3.21). Furthermore, since there are more variables than equations for the linear system
described by equation (3.21), an infinite number of solutions are available. However, the
objective is to locate a natural solution in the vicinity of the initial guess. (See Figure
3.2.) Therefore, the smallest Euclidean norm of M is used to compute the variations in
33
position and time, i.e., δ Ri and δτ i , for every patch point. A new set of patch point
positions and times are determined by subtracting these δ Ri and δτ i from the current set
of patch point positions and times. Re-targeting the new patch points again, using the
first level, results in smaller ΔV ’s at the internal points. (Recall Figure 3.1.) Then, δ Ri
and δτ i are again recomputed using the second level to locate different, new patch point
positions and times. The ΔV ’s associated with the internal points will be reduced even
further when the first level is reapplied. The process is repeated, ending with the first
level, until ΔVi < ε for i = 2,… , n − 1, thereby resulting in a continuous path. (See
Figure 3.2.)
ΔVi < ε
Vi − Vi + ΔVi +1 < ε
ΔVi −1 < ε
Vi −+1 Vi −−1
V3+ Vn−−2 ΔVn−2 < ε

ΔV3 < ε
V3− Vn+−2
V2+ Vn−−1
V2− ΔVn−1 < ε
ΔV2 < ε Vn+−1
V1 Vn
V1 Vn
Figure 3.2 Path Before Second Level is Applied (Black)

and Final Continuous Path (Blue)
34
3.1.3 Numerical Example: Lissajous Trajectories
The two-level corrections process can be used to determine particular solutions to the
nonlinear differential equations in the CR3BP. The analytical solutions to the linear
equations of motion can be used as an approximation for quasi-periodic motion, and,
thus, initiate the numerical process. The corrector is applied to the Lissajous trajectory
approximation in the Earth-Moon system near L1 (see Figure 2.3). The result appears in
n −1
Figure 3.3. In summary, ∑ ΔV
i =2
i for n = 41 patch points is reduced from an initial value
of 1.20319112 to 2.54715766 × 10−13 non-dimensional units, i.e., 13 orders of magnitude,

in just six iterations.
0.05
Earth-Moon System
1 unit = 385,692.5 km
0
y
μ = 0.0121505856
−0.05
0.8 0.82 0.84 0.86 0.88
x
0.05 0.05
0 0
z
Earth Moon
L1
−0.05 −0.05
0.8 0.82 0.84 0.86 0.88 −0.05 0 0.05
x y
Figure 3.3 Differentially Corrected L1 Lissajous Trajectory

35
Note the nonlinear effects of the natural solution depicted in Figure 3.3. For example,
the motion is not as uniform as the initial motion computed from the first-order analytical
approximation. (See Figure 2.3.) The amplitude in the y direction, Ay, and the amplitude
in the z direction, Az, are not exactly the same for each revolution. Furthermore, Ay is
now approximately 13,500 km and Az is approximately 16,000 km. However, although
the orbit is slightly smaller than the initial first-order approximation, the characteristic
shape is still maintained. Considering the sensitivities associated with the CR3BP, the
ability to compute a natural solution and maintain the characteristic shape for multiple
revolutions lends tractability to the two-level differential corrections process.
3.2 A Simple Corrector
A simple targeting scheme uses symmetry about the x-z plane to determine periodic
planar solutions in the vicinity of the collinear libration points. The resulting solutions
are termed Lyapunov orbits, named after the Russian mathematician Aleksandr
Mikhailovich Lyapunov (1857-1918) for his work on ordered differential equations,
stability theory, and non-linear analysis. First-order analytical approximations for the
solutions are available in equations (2.43)-(2.44) or (2.46)-(2.47). Small amplitude
Lyapunov orbits are expanded from these analytical approximations to obtain a manifold
family of solutions. A general assessment of the stability and the locations of bifurcation
points that signal intersections with other families of solutions are detailed below for the
L1 Lyapunov family.
3.2.1 The Algorithm
The algorithm can be used to numerically compute Lyapunov orbits in the vicinity of
the collinear points. In general, exploitation of any symmetry greatly reduces the
complexity of the three-body problem. Furthermore, restricting the motion of P3 to the x-
y plane, a required characteristic of the Lyapunov orbits, further reduces the degrees of
freedom. For convenience, the initial state vector associated with a Lyapunov orbit is
usually located on the x-axis and possesses only two non-zero components. Thus, a
36
targeting scheme for computation of Lyapunov orbits is triggered with an initial state
vector of the form
q0 = { x0 0} ,
T
0 0 0 y0 (3.23)
where initial values x0 and y0 are available for the first iteration from the analytical
approximations. Therefore, the objective is to adjust the nonzero components of the
initial state such that a perpendicular crossing is achieved at the endpoint. (See Figure
3.4.) Though not necessary for convergence, fixing either x0 or y0 allows for an
expansion of solution space once the current solution is obtained. A scheme for
constraining x0 is based on the mapping
δ x = F ( y0 ) , (3.24)
where F represents a mapping of y0 from y0 = 0 to the next crossing y = 0. This
mapping is therefore independent of the integration time, τ . (See Figure 3.4 for a
representation of F.) The variation of δ x with respect to δ y0 is available from the state-
transition matrix. Therefore, using equation (2.61), a targeter for convergence is simply
Initial
Path
Final
Path
P2
P1 L1 x
Figure 3.4 Initial and Final Path for Simple Corrector

37
∂x
δx= δ y0 . (3.25)
∂y0
Since a perpendicular crossing is targeted when F is applied to y0 , the desired value of
δ x at the second crossing is zero. Solving for δ y0 in equation (3.25), subtracting this
from y0 , and re-integrating yields a smaller δ x. The process is repeated until δ x < ε
and a perpendicular crossing is achieved.
Notice that an additional mapping can be written such that
δ y = G ( y0 ) , (3.26)
where G takes y0 to δ y when y = 0 at the next crossing. Therefore, δ y = 0. Using

equation (2.61) the targeter defined by the variational relationships then becomes
⎡ ∂y ⎤
y ⎥
⎧ 0 ⎫ ⎢⎢ ∂y0 δy
⎥ ⎧⎨ 0 ⎫⎬ .
⎨ ⎬= (3.27)
⎩δ x ⎭ ⎢ ∂x ⎥ δτ ⎭
y ⎥⎩
⎢ ∂y
⎣ 0 ⎦
The additional equation greatly increases the speed of convergence. Similar to equation
(3.25), solving for δ y0 in equation (3.27), subtracting this from y0 , and re-integrating
produces a smaller value for δ x. Notice that, given the definitions for F and G, it is
unnecessary to also subtract δτ from τ . As in the prior case, the process is repeated
until δ x < ε . Therefore, predictions in the endpoint variations are available from
equation (2.61) and are used in an iterative procedure to update the control parameter
δ y0 until the desired final point variation δ x is reduced to zero. The goal is to eliminate
the endpoint variations and, thus, these are the constraint parameters, while the initial
variations in the state are the control parameters. For example, in equation (3.27), δ y0 is
the control parameter and the constraint parameter is δ x, with a goal or target δ x = 0.
The final Lyapunov orbit is obtained by integrating the converged values in q0 (with
updated y0 ) over the period T = 2τ . A method of continuation is then employed to

expand the solution space. In applying continuation, a prediction for the neighboring nth
solution is obtained from the current result such that,
38
q0n = q0n −1 + S ⋅ Δq0n −1 , (3.28)
where q0n −1 is the initial condition corresponding to the previously converged solution.
The step size S is usually simply a fixed value Δx0 . Since it is desired to step along the
x-axis,
Δq0n −1 = {1 0 0 0 0 0} .
T
(3.29)
The process is repeated until the manifold family of solutions is obtained.
3.2.2 Numerical Results: L1, L2, and L3 Lyapunov Families of Orbits
The algorithm in the previous section is used to obtain families of Lyapunov orbits in
the vicinity of the collinear points in the Earth-Moon system. (See Figures 3.5-3.7.) The
black orbits appearing in Figures 3.5-3.7 correspond to intersections, or bifurcations,
between the Lyapunov orbit families and other families of periodic orbits. For example,
depicted in Figure 3.5 are two distinct bifurcations. Note that the smaller amplitude
bifurcation corresponds to the intersection between the L1 Lyapunov orbits and the L1
halo orbit family. Furthermore, the larger amplitude bifurcation is associated with the
intersection between the L1 Lyapunov orbits and the L1 axial orbit family. Unlike the
Lyapunov orbits, both the axial and halo orbit families are three-dimensional. (See
Section 3.3 for a more detailed discussion of the halo and axial orbit families.) Further,
since closest approach for the Lyapunov orbits near the collinear points occurs along the
x-axis, and both the initial and final point of the targeting scheme are also located on the
x-axis, the algorithm breaks down when passage by either primary is close. However,
reducing the step-size and saving the states associated with every 10 or 20 orbits yields
an appropriate representation for the shape of the solution space. The initial conditions,
period T, and stability index υ are also provided in Tables 3.1-3.3. The stability index is
further discussed in Section 3.2.3. Notice that the initial conditions, periods, and stability
indices are computed for the bifurcating orbits. Also, recall that the period T can be
dimensionalized simply by multiplying by the characteristic time t * available from
equation (2.4). For example, in the Earth-Moon system t * ≅ 4.3644 days.
39
Earth-Moon System
1 unit = 385,692.5 km
μ = 0.0121505856
0.6
L1
To Earth
0.4
0.2
Bifurcation
0 With L1 Halo Moon
y
Family
−0.2
Bifurcation
−0.4 With L1 Axial
Family
−0.6
0.2 0.4 0.6 0.8 1 1.2 1.4

x
Figure 3.5 The L1 Lyapunov Family
40
Table 3.1
L1 Lyapunov Family Initial Conditions
x0 y0 T υ
0.8234 0.1263 2.7430 1180.5771 L1 Halos
0.8189 0.1750 2.7959 1045.0317
0.8089 0.2838 3.0224 662.3978
0.7989 0.3649 3.3638 387.9805
0.7889 0.4159 3.7121 254.6194
0.7816 0.4432 3.9500 200.3471 L1 Axials
0.7689 0.4813 4.3317 144.2199
0.7589 0.5068 4.6052 117.6505
0.7489 0.5305 4.8579 99.3896
0.7389 0.5531 5.0915 86.3560
0.7289 0.5752 5.3071 76.8352
0.7189 0.5971 5.5060 69.7938
0.7089 0.6189 5.6891 64.5687
0.6989 0.6409 5.8575 60.7141
0.6889 0.6631 6.0121 57.9192
0.6789 0.6855 6.1538 55.9615
0.6689 0.7082 6.2836 54.6777
0.6589 0.7313 6.4023 53.9465
0.6489 0.7548 6.5107 53.6761
0.6389 0.7786 6.6097 53.7965
0.6289 0.8028 6.7001 54.2536
0.6189 0.8275 6.7824 55.0055
0.6089 0.8526 6.8573 56.0195
41
Earth-Moon System
1 unit = 385,692.5 km
μ = 0.0121505856
0.5
To Earth
0.4 L2
0.3
0.2
0.1
Bifurcation
0 Moon With L2 Halo
y
Family
−0.1
−0.2
−0.3
−0.4 Bifurcation
With L2 Axial
Family
−0.5
0.6 0.8 1 1.2 1.4 1.6

x
42
Table 3.2
x0 y0 T υ
1.1762 -0.1231 3.3981 651.0097
1.1809 -0.1559 3.4155 606.1130 L2 Halos
1.1843 -0.1818 3.4341 564.4278
1.1924 -0.2502 3.5128 435.7722
1.2005 -0.3239 3.6841 288.4590
1.2086 -0.3804 3.9421 189.9619
1.2167 -0.4164 4.2083 140.5503
1.2200 -0.4275 4.3105 127.8850 L2 Axials
1.2248 -0.4419 4.4584 113.2251
1.2329 -0.4621 4.6923 96.0231
1.2410 -0.4794 4.9123 84.1910
1.2491 -0.4948 5.1204 75.5738
1.2572 -0.5092 5.3180 69.0655
1.2653 -0.5227 5.5062 64.0400
1.2734 -0.5357 5.6856 60.1148
1.2815 -0.5484 5.8568 57.0416
1.2896 -0.5607 6.0203 54.6507
1.2977 -0.5729 6.1765 52.8217
1.3058 -0.5849 6.3258 51.4659
1.3139 -0.5968 6.4685 50.5160
1.3220 -0.6087 6.6049 49.9203
43
Earth-Moon System
1 unit = 385,692.5 km
μ = 0.0121505856
1.5 Bifurcation
To Moon
With L3 Halo
Family
1
0.5
0 L3 Earth
y
−0.5
−1
Bifurcation Bifurcation
With L4/L5 Planar With L3 Axial
−1.5 Family Family
−2 −1.5 −1 −0.5 0 0.5

x
44
Table 3.3
x0 y0 T υ
-1.0560 0.1017 6.2185 1.6753
-1.0884 0.1652 6.2186 1.6729
-1.1208 0.2280 6.2189 1.6694
-1.1532 0.2899 6.2192 1.6647
-1.1856 0.3512 6.2196 1.6589
-1.2180 0.4119 6.2201 1.6519
-1.2504 0.4721 6.2207 1.6437
-1.2828 0.5318 6.2213 1.6343
-1.3152 0.5911 6.2221 1.6236
-1.3476 0.6500 6.2229 1.6117
-1.3800 0.7086 6.2238 1.5985
-1.4124 0.7670 6.2248 1.5840
-1.4448 0.8252 6.2260 1.5680
-1.4772 0.8833 6.2272 1.5506
-1.5096 0.9414 6.2285 1.5316
-1.5420 0.9994 6.2300 1.5110
-1.5744 1.0576 6.2316 1.4885
-1.6068 1.1159 6.2334 1.4640
-1.6392 1.1745 6.2353 1.4372
-1.6716 1.2335 6.2374 1.4078
-1.6967 1.2795 6.2391 1.3828 L3 Halos
-1.7364 1.3531 6.2423 1.3390
-1.7688 1.4142 6.2451 1.2981
-1.8012 1.4765 6.2484 1.2511
-1.8336 1.5405 6.2522 1.1959
-1.8663 1.6072 6.2566 1.1283
-1.8963 1.6715 6.2616 1.0499 L3 Axials
-1.9118 1.70633 6.2647 1.0000 L4/L5 Planar
45
3.2.3 The Stability Index and Bifurcations
Information offered by analysis of the phase space includes stability and bifurcation
points. The eigenvalues and/or characteristic multipliers associated with the monodromy
matrix are used to evaluate the stability of the solutions, and to isolate intersections with
other families of orbits in the solution space. All of the solutions examined in this study
possess six characteristic multipliers in reciprocal pairs. Since the solution is periodic,
two of the six multipliers are equal to one. The other four include a pair associated with
the stable/unstable subspace and the final pair represents the center subspace. Since there
is one stable and one unstable mode, the average of the (reciprocal) pair of multipliers
associated with the stable subspace ( λ Ws < 1) and unstable subspace ( λ Wu > 1) , or
υ=
2
(
1 Ws
λ + λ Wu ,) (3.30)
is defined as the stability index υ. The value of the index is a single numerical quantity to
assess the stability of the system [20].
If λ Ws = λ Wu = ±1, the reference trajectory defines the exact intersection between two
different families. Thus, the orbit is a subset of both families and identifies a bifurcation
point. Furthermore, if both characteristic multipliers associated with the center subspace,
* *
i.e., λ Γ and λ Γ , are such that λ Γ = λ Γ = ±1, the solution or orbit also defines the
intersection point between two different families. When λ Ws = λ Wu = ±1 and

*
λ Γ = λ Γ = ±1 simultaneously, the trajectory solution belongs to three different families.
Thus, three families intersect resulting in a trifurcation [26]. Other types of bifurcations
also exist that are not included in this analysis. For example, the point at which two
complex conjugate multipliers leave the unit circle in the complex plane is a special type
of bifurcation known as a Krien collision [27]. In short, valuable information is gained
by analyzing the characteristic multipliers.
The exact location of the bifurcation orbits can be computed using continuation and a
method of bisections [44]. A distinct change in the eigenstructure, i.e., when the
characteristic multipliers leave the unit circle to occupy the real axis or vice versa,
triggers the location of the bifurcation point. If a distinct change in the eigenstructure
46
occurs between two successive solutions during a continuation process, then a bifurcating
orbit is clearly located between the current and previous solution. Therefore, let q0n −1 and
q0n − 2 be the initial states associated with the current and previous solutions as prescribed
by equation (3.28). Then
q0lower = q0n − 2 , (3.31)
q0upper = q0n −1. (3.32)
Also, let q0mid be the initial state associated with the prediction for a solution midway
between the previous two trajectories,
S
q0mid = q0lower + ⋅ Δq0lower . (3.33)
2
Then, as before, the location of the bifurcation is available from the eigenstructure and is
either between q0lower and q0mid or between q0mid and q0upper . If it is evident from the
eigenstructure that the bifurcating orbit is located between q0lower and q0mid , then
q0upper = q0mid . (3.34)
Alternatively, if it is evident that the bifurcation occurs between q0mid and q0upper , then
q0lower = q0mid . (3.35)
Let the initial state q0mid associated with the bisecting solution be redefined by using the
prediction,
S
q0mid = q0lower + ⋅ Δq0lower , (3.36)
4
to converge the new solution. Notice that the step size for the prediction of q0mid is
reduced by a factor of two with every new prediction of q0mid . New q0upper or q0lower are
*
again redefined and the process is continued until λ Ws = λ Wu = ±1 or λ Γ = λ Γ = ±1
thereby obtaining the exact initial state (within numerical precision) for the bifurcating
orbit.
The stability indices in Tables 3.1-3.3 are calculated using equation (3.30) and the
bifurcations are located using the method of bisections. (Recall the bifurcation
47
trajectories are plotted as black orbits in Figures 3.5-3.7.) Notice that for orbits
corresponding to the initial conditions represented in Tables 3.1-3.3, the stability index is
positive, i.e., υ > 1, with the exception of the last initial condition provided for the L3
Lyapunov orbits. Therefore, the first two bifurcations, located when expanding the
solution space for all three Lyapunov families, are associated with the condition
*
λ Γ = λ Γ = ±1. Initially, the multipliers in the center subspace for all three families are
located on the unit circle, as expected. Expanding the solutions reveals bifurcations with
the halo orbit families, where the center multipliers become real. The multipliers return
to the unit circle at the intersection of the Lyapunov and the axial orbit families. The
third bifurcation along the L3 family, the bifurcation with the L4/L5 planar orbits, is
associated with λ Ws = λ Wu = ±1, as evidenced by a stability index of one. (See Figure 3.8
for the propagation of the characterstic multipliers associated with the center subspace for
the L1 Lyapunov orbits represented in Figure 3.5 and Table 3.1.)
The strategies for locating the intersections between families are also used to obtain
the family of solutions for the intersecting families, i.e., the halo, the axial, and the L4/L5
planar families. Strategies for computing the halo, axial, and L4/L5 planar families are
further discussed in the following sections. Bifurcations within these families are located
using the same method, and the solutions are expanded to locate additional families.
1
Bifurcaton with the L1
0.5 Halo and Axial Families
Im
0
−0.5 Unit Circle
−1
−7 −6 −5 −4 −3 −2 −1 0 1
Re
Figure 3.8 Characteristic Multipliers Associated with Center Subspace

for L1 Lyapunov Orbits.
48
3.3 Targeting Orbits in Thee-Dimensional Space Using Symmetry
The simple targeter is also generalized to compute periodic orbits in three-

dimensional space. The periodic solutions bifurcating from the small amplitude
Lyapunovs, as seen in Figures 3.5-3.7, are denoted halo orbits, as first coined by
Farquhar [45] who noticed that a spacecraft in an L2 halo orbit on the far side of the
Moon appears to resemble a halo about the Moon when viewed from the Earth. A
spacecraft in such an orbit maintains continuous communications with the Earth. The
axial orbits are so named because the family appears to rotate about the x-axis from the
bifurcating Lyapunov orbit, as will be apparent later in Section 3.3. The most “vertical”
member of the axial family marks the bifurcating orbit between the axial and vertical
families. The vertical orbits resemble a “figure-8,” and are labeled vertical since the
motion of small amplitude members of the family is primarily in the z direction.
Numerical results for the halo, axial, and vertical families of orbits are available in the
literature [8, 16-17, 19-20, 27-28]. A number of algorithms can be employed
successfully to compute these families in the vicinity of the collinear libration points by
exploiting symmetry.
3.3.1 Strategies to Compute Periodic Orbits
Symmetric with respect to x-z plane. Periodic halo orbits as well as the vertical orbits are
symmetric about the x-z plane. Therefore, the initial state vector is strategically
positioned in the x-y plane such that only three nonzero components exist. Similar to the
construction of the planar Lyapunov orbits, one of these parameters is held fixed while a
differential corrections procedure is applied to the remaining two initial states until a
perpendicular crossing is achieved at the endpoint. Thus, the initial state vector takes the
form
q0 = { x0 0} ,
T
0 z0 0 y0 (3.37)
where, if it is desired to fix x0 , the following mappings can be defined
δ y = F ( z 0 , y0 ) , (3.38)
δ x = G ( z 0 , y0 ) , (3.39)
49
δ z = H ( z0 , y0 ) . (3.40)
By definition F, G, and H bring z0 and y0 to δ y, δ x, and δ z when y = 0, i.e., when
1
the solution crosses the x-y plane at τ = T . (Recall that T is the period.) Therefore, the
2
control parameters δ z0 and δ y0 are iteratively updated until the constraints,
δ x = δ z ≅ 0, are satisfied. Specifically, using equation (2.61), the targeter is defined as

⎡ ∂y ∂y ⎤
⎢ y ⎥
∂z ∂y0
⎧0⎫ ⎢ 0 ⎥ ⎧ δ z0 ⎫
⎪ ⎪ ⎢ ∂x ∂x ⎥⎪ ⎪
⎨δ x ⎬ = ⎢ x ⎥ ⎨δ y0 ⎬ . (3.41)
⎪δ z ⎪ ⎢ ∂z0 ∂y0 ⎥ ⎪ δτ ⎪
⎩ ⎭ ⎢ ⎥⎩ ⎭
∂z ∂z
⎢ z ⎥
⎣ ∂z0 ∂y0 ⎦
Solving for δ z0 and δ y0 in equation (3.41) and updating the initial state by subtracting
these from z0 and y0 , respectively, yields smaller values of δ x and δ z in the next
iteration. The process is repeated until δ x < ε and δ z < ε , thereby constraining the
motion to be perpendicular to the x-y plane at the endpoint.
Alternatively, if it is desired to fix z0 , the following mappings can be defined
δ y = F ( x0 , y0 ) , (3.42)
δ x = G ( x0 , y0 ) , (3.43)
δ z = H ( x0 , y0 ) , (3.44)
where F, G, and H bring x0 and y0 to δ y, δ x, and δ z when y = 0. As such, δ x0 and
δ y0 are the control parameters with the same constraints δ x = δ z ≅ 0. The

corresponding targeter is written as
50
⎡ ∂y ∂y ⎤
⎢ y ⎥
∂x ∂y0
⎧0⎫ ⎢ 0 ⎥ ⎧δ x0 ⎫
⎪ ⎪ ⎢ ∂x ∂x ⎥⎪ ⎪
⎨δ x ⎬ = ⎢ x ⎥ ⎨δ y0 ⎬ , (3.45)
⎪δ z ⎪ ⎢ ∂x0 ∂y0 ⎥ ⎪ δτ ⎪
⎩ ⎭ ⎢ ⎥⎩ ⎭
∂z ∂z
⎢ z ⎥
⎣ ∂x0 ∂y0 ⎦
where the controls are iteratively updated until the constraints are satisfied within the
specified tolerance.
One full revolution along the halo or vertical orbit is obtained by integrating the
converged initial state vector q0 over the period T = 2τ . To obtain families of halo
orbits, via continuation, fix z0 in regions where z0 is changing more rapidly than x0 , and
fix x0 in regions where x0 is changing more rapidly than z0 . Since the halo orbit family
originates with the bifurcating Lyapunov orbit from Section 3.2.1 (Figures 3.5-3.7), z0 is
initially fixed. Predictions for the next orbit in the family are obtained by using equation
(3.28), where the step size S is simply a fixed value Δz0 . The southern orbits result for
Δz0 < 0, whereas for the northern family of halo orbits is acquired when Δz0 > 0 . Also,
since it is desired to step through values of z0 ,
Δq0n −1 = {0 0 1 0 0 0} .
T
(3.46)
The prediction is converged to a new periodic orbit (at the specified value of z0 ), using
equation (3.45). The process is repeated for a fixed value Δz0 until
x0n − x0n −1 > z0n − z0n −1 . (3.47)
The fixed parameter is changed to x0 , and Δq0n−1 is defined from equation (3.29).
Predictions for the initial state corresponding to the next orbit in the family are computed
from equation (3.28) assuming x0 is fixed. The states converge to a periodic orbit. The
procedure steps in x0 until equation (3.47) fails to be true. The fixed parameter is then
switched back to z0 with Δq0n −1 again defined by equation (3.46) and the process is
continued. For some cases, e.g., in the L2 halo orbit family and all the vertical orbit
51
families, it is necessary to change the sign of Δz0 when the fixed parameter is switched
from x0 to z0 .
Symmetric with respect to x-axis. The axial and vertical orbits are symmetric about the x-
axis. Exploiting symmetry results in an initial state vector of the form
q0 = { x0 z0 } .
T
0 0 0 y0 (3.48)
For the axial orbits, two sets of mappings are necessary. When the motion of the orbit is
primarily horizontal, the mappings for a fixed y0 are defined as,
δ y = F ( x0 , z0 ) , (3.49)
δ z = G ( x0 , z0 ) , (3.50)
δ x = H ( x0 , z0 ) , (3.51)
where F, G, and H bring x0 and z0 to δ y, δ z , and δ x when y = 0. However, when the

motion becomes more vertical the following mappings are more useful,
δ y = P ( x0 , z0 ) , (3.52)
δ z = Q ( x0 , z0 ) , (3.53)
δ x = R ( x0 , z0 ) , (3.54)
where P, Q, and R bring x0 and z0 to δ y, δ z , and δ x when z = 0, i.e., the next

crossing with the x-y plane. Therefore, using equation (2.61), the targeter is simply
written as,
⎡ ∂y ∂y ⎤
⎢ y ⎥
∂x ∂z0
⎧ ⎫ ⎢ 0
δ y ⎥ ⎧δ x0 ⎫
⎪ ⎪ ⎢ ∂z ∂z ⎥⎪ ⎪
⎨δ z ⎬ = ⎢ z ⎥ ⎨δ z0 ⎬ . (3.55)
⎪δ x ⎪ ⎢ ∂x0 ∂z0 ⎥ ⎪ δτ ⎪
⎩ ⎭ ⎢ ⎥⎩ ⎭
∂x ∂x
⎢ x ⎥
⎣ ∂x0 ∂z0 ⎦
If the mappings F, G, and H are used, then δ y = 0 in equation (3.55), whereas if the
mappings P, Q, and R are used, then δ z = 0 in equation (3.55). The control parameters,
52
δ x0 and δ z0 , are adjusted until the final endpoint constraints δ z = δ x ≅ 0 are satisfied,
within a specified tolerance. Similarly, if it is desired to fix z0 the targeter is written
⎡ ∂y ∂y ⎤
⎢ y ⎥
∂x ∂y0
⎧δ y ⎫ ⎢ 0 ⎥ ⎧δ x0 ⎫
⎪ ⎪ ⎢ ∂z ∂z ⎥⎪ ⎪
⎨δ z ⎬ = ⎢ z ⎥ ⎨δ y0 ⎬ , (3.56)
⎪δ x ⎪ ⎢ ∂x0 ∂y0 ⎥ ⎪ δτ ⎪
⎩ ⎭ ⎢ ⎥⎩ ⎭
∂x ∂x
⎢ x ⎥
⎣ ∂x0 ∂y0 ⎦
where δ y = 0 or δ z = 0 depending on the mapping. Again, the parameters δ x0 and
δ y0 are adjusted until δ z , δ x < ε , and a perpendicular crossing with the x-axis at the
endpoint is achieved.
The complete axial or vertical orbit is obtained by integrating the converged value of
q0 over the period T = 2τ . Since the axial family originates with a bifurcating Lyapunov
orbit, it is desirable to initially fix z0 and use mappings corresponding to y = 0. Once an

out-of-plane axial orbit is established, a method of continuation can be used by applying
equation (3.28) with some fixed step size, Δy0 or Δz0 , and corresponding adjustment
Δq0n −1 determined by the fixed parameter. As the solutions become more vertical, the
mappings associated with y = 0 break down and it is necessary to switch to the mappings
associated with z = 0.
Symmetry relative to two planes. The vertical orbits are symmetric about both the x-z and
x-y planes. Therefore, any of the targeting schemes can, in theory, be employed to obtain
the vertical family of orbits once the bifurcating solution from the axial family is
obtained. However, an additional targeter can be constructed that exploits both types of
symmetry. Using the mappings defined in equations (3.49)-(3.51), the targeter uses the
control variables δ x0 and δ z0 to force the endpoint parameters δ x and δ z to zero,
thereby targeting a perpendicular crossing at the x-z plane. Using equation (2.61), this
type of targeter is written in the form
53
⎡ ∂y ∂y ⎤
⎢ y ⎥
∂x ∂z0
⎧0⎫ ⎢ 0 ⎥ ⎧δ x0 ⎫
⎪ ⎪ ⎢ ∂x ∂x ⎥⎪ ⎪
⎨δ x ⎬ = ⎢ x ⎥ ⎨δ z0 ⎬ . (3.57)
⎪δ z ⎪ ⎢ ∂x0 ∂z0 ⎥ ⎪ δτ ⎪
⎩ ⎭ ⎢ ⎥⎩ ⎭
∂z ∂z
⎢ z ⎥
⎣ ∂x0 ∂z0 ⎦
The complete vertical orbit is acquired by integrating the converged initial state vector q0
over the period T = 4τ . A method of continuation over x0 , as previously described, can

be used to obtain the vertical family of orbits.
3.3.2 Numerical Results: L1, L2, and L3 Halo, Axial, and Vertical Families of
Orbits
The algorithms summarized in Section 3.3.1 are used to numerically compute the
halo, axial, and vertical orbit families in the vicinity of the collinear points in the Earth-
Moon system. The families with bifurcating orbits (black) are provided in Figures 3.9-
3.18 with corresponding initial conditions, periods, and stability indices in Tables 3.4-
3.13. Bifurcating orbits in each of these families are located using a method of
bisections. The bifurcating orbits from the Lyapunov families are the basis to acquire the
northern families of halo orbits. The southern orbit family is obtained by reflecting the
northern orbit family across the x-y plane. For example, the northern L1 halo orbit family
is depicted in Figure 3.9. The corresponding initial conditions, periods, and stability
indices are presented in Table 3.4. The initial orbit used to compute the halo orbit family,
i.e., the Lyapunov orbit, is available from Table 3.1. A method of continuation generates
the orbit family and a method of bisections computes a bifurcation between the L1 halo
orbit family and the L4/L5 axial orbit families. (See the near-rectilinear black orbit in
Figure 3.9.) Then, as mentioned previously, the southern L1 halo orbits are computed by
reflecting the orbits represented in Figure 3.9 across the x-y plane. The algorithms fail to
converge for large amplitude L3 halo orbits since the endpoints pass very close to the
Earth, as depicted in Figure 3.11. Similarly, bifurcating orbits from the Lyapunov orbit
families are used to compute the “northeast” or “northwest” axial families (see Figures
3.12-3.14 and Tables 3.7-3.9). The remaining part of the axial family is computed by
54
reflecting the orbits across the x-z plane. The most vertical member of the axial orbit
family corresponds to a bifurcation with the vertical orbit family. The vertical orbit
families are found using the above strategies and are represented in Figures 3.15-3.17
with corresponding initial conditions, periods, and stability indices available in Tables
3.10-3.12. Since the hodographs corresponding to the L1 and L2 vertical orbit families
intersect, it is advantageous to use the final targeting scheme from the previous section
when obtaining the L2 vertical orbit family. A family of orbits that bifurcate from a near-
rectilinear L2 halo orbit and might be described as representing a “butterfly” in shape is
also generated. (See Figure 3.18 and Table 3.13.)
55
0.2
To Earth
0.1 Earth-Moon System
1 unit = 385,692.5 km
0
y
μ = 0.0121505856
−0.1
−0.2
0.8 0.9 1 1.1
x
0.3 0.3
0.2 Moon 0.2
0.1 0.1
z
0 0
L1
−0.1 −0.1
0.8 0.9 1 1.1 −0.2 0 0.2
x y
Figure 3.9 The Northern L1 Halo Family

56
Bifurcation with
L4/L5 Axial Family
x
Bifurcation with L1
Lyapunov Family
Table 3.4
Northern L1 Halo Family Initial Conditions
x0 z0 y0 T υ
0.8234 0.0000 0.1263 2.7430 1180.5771 L1 Lyapunovs
0.8234 0.0224 0.1343 2.7464 1096.9329
0.8235 0.0344 0.1439 2.7507 993.7530
0.8237 0.0464 0.1558 2.7566 865.0564
0.8242 0.0584 0.1690 2.7634 724.4702
0.8250 0.0704 0.1827 2.7707 584.3385
0.8260 0.0824 0.1964 2.7778 454.2552
0.8273 0.0944 0.2095 2.7838 340.4499
0.8289 0.1064 0.2219 2.7872 245.8832
0.8307 0.1184 0.2334 2.7864 170.8261
0.8329 0.1304 0.2437 2.7785 113.6730
0.8355 0.1424 0.2527 2.7594 71.7704
0.8389 0.1544 0.2599 2.7215 42.1098
0.8509 0.1765 0.2624 2.5402 10.2024
0.8629 0.1858 0.2504 2.3567 3.0145
0.8749 0.1914 0.2325 2.1914 1.0000
0.8869 0.1957 0.2110 2.0477 1.4145
0.8989 0.2002 0.1865 1.9289 1.9886
0.9109 0.2060 0.1590 1.8417 2.2935
0.9246 0.2180 0.1232 1.8050 2.5615
0.9305 0.2300 0.1043 1.8397 2.7481
0.9330 0.2420 0.0936 1.8959 2.8581
0.9335 0.2540 0.0874 1.9586 2.8438
0.9329 0.2660 0.0839 2.0222 2.6473
0.9315 0.2780 0.0822 2.0845 2.1802
0.9292 0.2914 0.0817 2.1509 1.0000 L4/L5 Axials
57
To Earth
0.1
Earth-Moon System
0 1 unit = 385,692.5 km
y
μ = 0.0121505856
−0.1
1 1.1 1.2
x
0.2 Moon 0.2
0.1 0.1
z
z
0 0
L2
−0.1 −0.1
1 1.1 1.2 −0.1 0 0.1
x y

58
Bifurcation with L2
Northern Butterfly
Family
x
Bifurcation with L1
Lyapunov Family
Table 3.5
x0 z0 y0 T υ
1.1809 0.0000 -0.1559 3.4155 606.1130 L2 Lyapunovs
1.1807 0.0139 -0.1570 3.4139 597.2650
1.1802 0.0259 -0.1596 3.4100 576.0367
1.1794 0.0379 -0.1637 3.4037 543.6409
1.1782 0.0499 -0.1688 3.3949 502.0654
1.1767 0.0619 -0.1747 3.3835 453.6604
1.1746 0.0739 -0.1812 3.3694 400.9064
1.1721 0.0859 -0.1879 3.3521 346.2209
1.1690 0.0979 -0.1946 3.3314 291.8115
1.1654 0.1099 -0.2011 3.3066 239.5741
1.1611 0.1219 -0.2073 3.2768 191.0295
1.1561 0.1339 -0.2130 3.2406 147.2915
1.1503 0.1459 -0.2180 3.1959 109.0641
1.1435 0.1579 -0.2220 3.1393 76.6589
1.1354 0.1699 -0.2247 3.0645 50.0298
1.1234 0.1837 -0.2253 2.9374 26.0252
1.1114 0.1934 -0.2220 2.7903 12.8967
1.0994 0.1993 -0.2153 2.6259 5.7514
1.0874 0.2020 -0.2054 2.4499 1.9040
1.0754 0.2022 -0.1926 2.2689 1.0000
1.0634 0.2003 -0.1770 2.0883 1.2995
1.0514 0.1968 -0.1589 1.9123 1.6540
1.0394 0.1919 -0.1381 1.7434 1.6636
1.0274 0.1856 -0.1146 1.5818 1.4595
1.0118 0.1739 -0.0799 1.3743 1.0000 L2 Butterflies
59
0.5 L3
Earth-Moon System
0 1 unit = 385,692.5 km
y
μ = 0.0121505856
−0.5
−1
Earth
−2 −1 0
x
2 2
To Moon
1.5 1.5
1 1
z
z
0.5 0.5
0 0
−0.5 −0.5
−2 −1 0 −1 −0.5 0 0.5 1
x y

60
Bifurcation with L3
x Lyapunov Family
Table 3.6
x0 z0 y0 T υ
-1.6967 0.0000 1.2796 6.2391 1.3828 L3 Lyapunovs
-1.6921 0.1408 1.2760 6.2391 1.3811
-1.6838 0.2358 1.2696 6.2389 1.3779
-1.6712 0.3308 1.2600 6.2387 1.3731
-1.6543 0.4258 1.2470 6.2384 1.3667
-1.6329 0.5208 1.2307 6.2381 1.3589
-1.6069 0.6158 1.2108 6.2376 1.3495
-1.5762 0.7108 1.1874 6.2370 1.3386
-1.5405 0.8058 1.1602 6.2363 1.3264
-1.4996 0.9008 1.1291 6.2353 1.3129
-1.4530 0.9958 1.0938 6.2342 1.2983
-1.4004 1.0908 1.0539 6.2328 1.2826
-1.3411 1.1858 1.0091 6.2311 1.2660
-1.2743 1.2808 0.9588 6.2289 1.2488
-1.1989 1.3758 0.9021 6.2261 1.2310
61
0.3
0.2

1 unit = 385,692.5 km
0 L1
y
μ = 0.0121505856
−0.1
Moon
−0.2
0.6 0.8 1
x
0.3 0.3
0.2 To Earth 0.2
0.1 0.1
0 0
z
z
−0.1 −0.1
−0.2 −0.2
0.6 0.8 1 −0.2 −0.1 0 0.1 0.2

x y
Figure 3.12 The Northeast L1 Axial Family

62
Bifurcation with L1
Vertical Family
Bifurcation with L1
Lyapunov Family
Table 3.7
Northeast L1 Axial Family Initial Conditions
x0 y0 z0 T υ
0.7816 0.4432 0.0000 3.9500 200.3471 L1 Lyapunovs
0.7824 0.4401 0.0500 3.9520 201.2840
0.7848 0.4307 0.1000 3.9579 204.0837
0.7888 0.4146 0.1500 3.9677 208.7091
0.7947 0.3912 0.2000 3.9814 215.0815
0.8027 0.3594 0.2500 3.9984 223.0401
0.8130 0.3176 0.3000 4.0182 232.2452
0.8262 0.2625 0.3500 4.0390 241.9622
0.8431 0.1874 0.4000 4.0574 250.5755
0.8625 0.0907 0.4434 4.0652 254.2398 L1 Verticals
63
0.2
0 L2
y
1 unit = 385,692.5 km
−0.1 μ = 0.0121505856
Moon
−0.2
0.8 1 1.2
x
0.2 To Earth 0.2

0.1 0.1
0 0
z
z
−0.1 −0.1
−0.2 −0.2
0.8 1 1.2 −0.2 0 0.2

x y
Figure 3.13 The Northwest L2 Axial Family

64
Bifurcation with L2
Lyapunov Family
Bifurcation with L2
Vertical Family
Table 3.8
Northwest L2 Axial Family Initial Conditions
x0 y0 z0 T υ
1.2200 -0.4275 0.0000 4.3105 127.8850 L2 Lyapunovs
1.2183 -0.4248 0.0500 4.3133 129.1383
1.2152 -0.4198 0.0841 4.3183 131.3556
1.2112 -0.4133 0.1139 4.3247 134.0770
1.2062 -0.4048 0.1434 4.3326 137.3638
1.1998 -0.3938 0.1740 4.3425 141.2513
1.1918 -0.3796 0.2067 4.3544 145.7239
1.1818 -0.3610 0.2419 4.3684 150.6775
1.1697 -0.3369 0.2800 4.3838 155.8679
1.1554 -0.3055 0.3214 4.3995 160.8533
1.1389 -0.2647 0.3659 4.4130 164.9638
1.1119 -0.1812 0.4358 4.4222 167.6717 L2 Verticals
65
1.5
1
0 1 unit = 385,692.5 km
y
−0.5 μ = 0.0121505856
−1
L2
−1.5 Earth
−2 −1 0
x
1.5 To Moon 1.5

1 1
0.5 0.5
0 0
z
−0.5 z −0.5
−1 −1
−1.5 −1.5
−2 −1 0 −1 0 1
x y
Figure 3.14 The Northeast L3 Axial Family

66
Bifurcation with L3
Lyapunov Family z
Bifurcation with L3
Vertical Family
Table 3.9
Northeast L3 Axial Family Initial Conditions
x0 y0 z0 T υ
-1.8963 1.6715 0.0000 6.2616 1.0499 L3 Lyapunovs
-1.8902 1.6761 0.0900 6.2617 1.0604
-1.8705 1.6893 0.1800 6.2621 1.0889
-1.8291 1.7093 0.2745 6.2628 1.1273
-1.7488 1.7269 0.3737 6.2638 1.1567
-1.6081 1.7186 0.4779 6.2651 1.1567
-1.4186 1.6683 0.5873 6.2660 1.1370
-1.2271 1.5950 0.7022 6.2664 1.1211
-1.0003 1.4900 0.8726 6.2666 1.1142 L3 Verticals
67
0.5
Earth-Moon System
0
y
1 unit = 385,692.5 km
−0.5 μ = 0.0121505856
L1
−1
−1 0 1
x
1 Earth 1
0.5 0.5
0 0
z
z
−0.5 −0.5
−1 Moon −1
−1 0 1 −1 0 1
x y
Figure 3.15 The L1 Vertical Family

68
z
Bifurcation with L1
Axial Family
Table 3.10
L1 Vertical Family Initial Conditions
x0 z0 y0 T υ
1.0118 0.1739 -0.0799 4.0652 254.2398 L1 Axials
0.9261 0.3616 -0.0544 5.0950 168.7846
0.8860 0.4766 -0.0009 5.6180 178.5819
0.8184 0.5916 0.0776 5.8965 191.1137
0.7233 0.7066 0.1817 6.0552 193.8062
0.5906 0.8216 0.3221 6.1527 184.7102
0.4756 0.8936 0.4416 6.1963 171.8511
0.3606 0.9465 0.5601 6.2235 157.0654
0.2456 0.9833 0.6780 6.2419 141.4589
0.1306 1.0057 0.7953 6.2551 125.6070
0.0156 1.0148 0.9123 6.2650 109.8492
-0.0994 1.0108 1.0291 6.2726 94.4116
-0.2144 0.9937 1.1457 6.2786 79.4644
-0.3294 0.9626 1.2621 6.2834 65.1508
-0.4444 0.9162 1.3783 6.2874 51.6049
-0.5594 0.8520 1.4945 6.2908 38.9630
-0.6744 0.7654 1.6105 6.2936 27.3749
-0.7894 0.6477 1.7265 6.2960 17.0201
-0.8756 0.5267 1.8134 6.2976 10.2053
-0.9389 0.4047 1.8772 6.2987 5.8212
-0.9809 0.2897 1.9195 6.2994 3.2509
-1.0080 0.1747 1.9468 6.2998 1.7631
-1.0214 0.0597 1.9603 6.3000 1.0856
69
1
0.5
Earth-Moon System
0
y
1 unit = 385,692.5 km
−0.5 μ = 0.0121505856
−1 L2
−1 0 1
x
1 Earth 1
0.5 0.5
0 0
z
z
−0.5 −0.5
−1 Moon −1
−1 0 1 −1 0 1
x y

70
z
Bifurcation with L2
Axial Family
Table 3.11
x0 y0 z0 T υ
1.1119 -0.1812 0.4358 4.4222 167.6717 L2 Axials
1.1003 -0.3217 0.5973 5.6753 204.3084
1.0906 -0.4317 0.7416 6.0172 240.7419
1.0842 -0.5417 0.8415 6.1305 236.0384
1.0796 -0.6517 0.9128 6.1848 217.8085
1.0762 -0.7617 0.9635 6.2162 194.9734
1.0735 -0.8717 0.9978 6.2366 170.9369
1.0713 -0.9817 1.0179 6.2508 147.2392
1.0694 -1.0917 1.0248 6.2613 124.6527
1.0679 -1.2017 1.0192 6.2692 103.5866
1.0665 -1.3117 1.0009 6.2754 84.2596
1.0654 -1.4217 0.9694 6.2804 66.7829
1.0644 -1.5317 0.9234 6.2845 51.2053
1.0636 -1.6307 0.8677 6.2876 38.8193
1.0629 -1.7198 0.8038 6.2900 28.9912
1.0624 -1.8000 0.7321 6.2919 21.2089
1.0619 -1.8721 0.6525 6.2935 15.0634
1.0616 -1.9371 0.5634 6.2947 10.2310
1.0613 -1.9891 0.4740 6.2957 6.8491
1.0611 -2.0306 0.3827 6.2964 4.4628
1.0609 -2.0639 0.2853 6.2969 2.7679
1.0608 -2.0860 0.1920 6.2973 1.7530
1.0608 -2.1007 0.0820 6.2975 1.1314
71
1
0.5
Earth-Moon System
0
y
1 unit = 385,692.5 km
−0.5 μ = 0.0121505856
−1 L3
−1 0 1
x
1 Earth 1
0.5 0.5
0 0
z
z
−0.5 −0.5
−1 Moon −1
−1 0 1 −1 0 1
x y

72
z
Bifurcation with L3
Axial Family
Bifurcation with
L4/L5 Vertical Family
Table 3.12
x0 z0 y0 T υ
-0.7916 0.6160 -0.2129 6.2505 1.5968
-0.7012 0.7160 -0.3031 6.2508 1.5604
-0.5798 0.8160 -0.4241 6.2513 1.5119
-0.4798 0.8775 -0.5239 6.2518 1.4724
-0.3798 0.9242 -0.6236 6.2523 1.4334
-0.2798 0.9583 -0.7234 6.2529 1.3947
-0.1798 0.9811 -0.8232 6.2537 1.3565
-0.0798 0.9934 -0.9229 6.2546 1.3186
0.0202 0.9956 -1.0227 6.2557 1.2812
0.1202 0.9878 -1.1225 6.2570 1.2443
0.2202 0.9697 -1.2223 6.2588 1.2078
0.3202 0.9408 -1.3221 6.2610 1.1718
0.4830 0.8679 -1.4846 6.2666 1.1142 L3 Axials
0.6202 0.7763 -1.6214 6.2752 1.0670
0.7202 0.6866 -1.7211 6.2868 1.0334
0.8213 0.5690 -1.8214 6.3111 1.0021 L4/L5 Verticals
0.8897 0.4705 -1.8880 6.3474 1.0038
0.9527 0.3705 -1.9468 6.4196 1.0071
1.0259 0.2705 -2.0078 6.5791 1.0072
1.1147 0.1705 -2.0751 6.8165 1.0000
1.1678 0.0705 -2.1154 6.9476 1.0000
73
0.1
L2
0 1 unit = 385,692.5 km
y
μ = 0.0121505856
−0.05
−0.1 To Earth
0.9 0.95 1 1.05 1.1

x
0.2 0.2
0.15 0.15
0.1 0.1
z
0.05 z 0.05
0 0
Moon
−0.05 −0.05
0.9 0.95 1 1.05 1.1 −0.1 −0.05 0 0.05 0.1
x y
Figure 3.18 The Northern L2 Butterfly Family

74
Bifurcation with L2
Northern Halo
Family
Table 3.13
Northern L2 Butterfly Family Initial Conditions
x0 z0 y0 T υ
1.0118 0.1739 -0.0799 1.3743 1.0000 L2 Halos
1.0196 0.1742 -0.0817 2.7519 1.0589
1.0249 0.1743 -0.0821 2.7582 1.1762
1.0301 0.1742 -0.0816 2.7687 1.3772
1.0354 0.1740 -0.0801 2.7845 1.6932
1.0406 0.1735 -0.0770 2.8077 2.1774
1.0459 0.1727 -0.0721 2.8409 2.9191
1.0511 0.1713 -0.0646 2.8891 4.0650
1.0564 0.1693 -0.0539 2.9596 5.8298
1.0616 0.1662 -0.0392 3.0637 8.4171
1.0669 0.1619 -0.0189 3.2255 11.5767
75
3.4 Asymmetric Correction Schemes
An asymmetric corrections scheme is a powerful tool for converging to any periodic

solution. The targeting schemes in Sections 3.2-3.3 use symmetry to adjust the control
parameters such that endpoint constraints are fully satisfied. However, there exist
asymmetric periodic solutions as well, in particular, the periodic solutions near the
triangular equilibrium points. Near L4 or L5, the planes of symmetry are either not easily
defined or are not consistently defined along the family. In such cases, equation (2.61)
cannot be used to adjust the control variables at the initial time that minimize the
endpoint variations. A more sophisticated algorithm is necessary. An algorithm
introduced by Markellos and Halioulias [14-15] successfully produces periodic solutions
of this type. A continuation method for predicting elements of future initial state vectors
by shifting along the tangent to the family is also developed. The algorithms and
continuation methods are used to locate the L4 planar, axial, and vertical families of
orbits. Similar orbits have been computed by various researchers [7, 9, 21, 26, 29].
3.4.1 Strategies to Compute Periodic Orbits
The planar case. The initial condition, associated with the planar case, that produces
asymmetric periodic orbits, e.g., planar orbits in the vicinity of L4, is of the form
q0 = { x0 0} .
T
y0 0 x0 y0 (3.58)
To later expand the solution space, one of the four nonzero components are fixed during
the integration process. Furthermore, one of the nonzero components also serves as a
reference for defining mappings. Assuming y0 is fixed, then,
x = F ( x0 , x0 , y0 ) , (3.59)
x = G ( x0 , x0 , y0 ) , (3.60)
where the mappings bring x0 , x0 , and y0 to x for F and to x for G from the reference y0
to y = y0 . The reference value remains consistent throughout the entire family.

Therefore, a reference value is determined that is common to all orbits in the family, i.e.,
76
each orbit in the family crosses the reference. Therefore, for periodic motion, the goal is
determination of the variations δ x0 and δ x0 such that x = x0 + δ x0 and x = x0 + δ x0 , i.e.,
x = F ( x0 + δ x0 , x0 + δ x0 , y0 + δ y0 ) = x0 + δ x0 , (3.61)
x = G ( x0 + δ x0 , x0 + δ x0 , y0 + δ y0 ) = x0 + δ x0 . (3.62)
From the first-order Taylor series expansion for equation (3.61)-(3.62),

∂F ∂F
F ( x, x ) = x0 + δ x0 = F ( x0 , x0 ) + δ x0 + δ x0 , (3.63)
∂x0 ∂x0
∂G ∂G
G ( x, x ) = x0 + δ x0 = G ( x0 , x0 ) + δ x0 + δ x0 . (3.64)
∂x0 ∂x0
Rearranging equations (3.63)-(3.64) results in targeter relationships,
⎡⎛ ∂F ⎞ ∂F ⎤
⎢⎜ − 1⎟ ⎥
⎧δ x ⎫ ⎢⎝ ∂x0 ⎠ ∂x0 ⎥ ⎧δ x0 ⎫
⎨ ⎬=⎢ ⎨ ⎬, (3.65)
⎩δ x ⎭ ⎢ ∂G ⎛ ∂G ⎞ ⎥ ⎩δ x0 ⎭
⎜ − 1⎟ ⎥
⎢⎣ ∂x0 ⎝ 0 ⎠ ⎥⎦
∂x
where the partials of F and G with respect to x0 , x0 , and y0 are available from the linear
variational equations. To obtain these partials, use equation (2.61) and isolate the
appropriate sensitivity partials
∂F δ x ∂x ∂y x
= = − , (3.66)
∂x0 δ x0 ∂x0 ∂x0 y
= = − , (3.67)
∂x0 δ x0 ∂x0 ∂x0 y
= = − , (3.68)
∂y0 δ y0 ∂y0 ∂y0 y
∂G δ y ∂x ∂y x
= = − , (3.69)
∂x0 δ x0 ∂x0 ∂x0 y
= = − , (3.70)
∂x0 δ x0 ∂x0 ∂x0 y
= = − . (3.71)
∂y0 δ y0 ∂y0 ∂y0 y
77
The solution for the variations in δ x0 and δ x0 is obtained from equation (3.65). Then,
δ x0 and δ x0 are subtracted from the states x0 and x0 , respectively. The process is
repeated until δ x, δ x < ε , and the periodic orbit is obtained.
Equations (3.66)-(3.71) are also employed for predictions, that is, an initial guess for
the next initial state q0n along the tangent subspace Г corresponding to the family. Let
Δq0n −1 be defined as
Δq0n −1 = {Δx0n −1 0 0 Δx0n −1 Δy0n −1 0} ,

T
(3.72)
where the following equations must be satisfied

∂F n −1 ∂F n −1 ∂F n −1
Δx0n −1 = Δx0 + Δx0 + Δy0 , (3.73)
∂x0 ∂x0 ∂y0
∂G n −1 ∂G n −1 ∂G n −1
Δx0n −1 = Δx0 + Δx0 + Δy0 , (3.74)
∂x0 ∂x0 ∂y0
such that the components of Δq0n−1 are normalized,
1 = ( Δx0n −1 ) + ( Δx0n −1 ) + ( Δy0n −1 ) .

2 2 2
(3.75)
The simultaneous solution of equations (3.73)-(3.75) yields values for the scalars Δx0n −1 ,
Δx0n −1 , and Δy0n −1. Then, equation (3.28) is used for some fixed step size S, and the
prediction q0n initiates a new iteration process to locate a periodic orbit. To compute the
entire family, it is necessary to change the sign of S when passing through the extremum
of Г. The switch time is apparent at the point when Δx0n −1 and Δx0n −1 simultaneously
change signs, i.e.,
Δx0n − 2 ⋅ Δx0n −1 < 0, (3.76)
Δx0n − 2 ⋅ Δx0n −1 < 0. (3.77)

The entire process is repeated until the family of solutions is computed.
Motion in three-dimensional space with reference y0 . To calculate the L4 vertical orbits,

the targeting algorithm must be adapted to include out-of-plane motion. Successful
78
computation of asymmetric periodic orbits in three-dimensional space depends upon the

following mappings, i.e.,
x = F ( x0 , z0 , x0 , y0 , z0 ) , (3.78)
z = G ( x0 , z0 , x0 , y0 , z0 ) , (3.79)
x = H ( x0 , z0 , x0 , y0 , z0 ) , (3.80)
z = I ( x0 , z0 , x0 , y0 , z0 ) , (3.81)
such that each one maps from the reference y0 to y = y0 , where y0 is common for all
orbits in the family. The goal is the determination of the variations
δ x0 , δ z0 , δ x0 , and δ z0 that result in this particular type of periodic orbit, i.e.,
x = F ( x0 + δ x0 , z0 + δ z0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = x0 + δ x0 , (3.82)
z = G ( x0 + δ x0 , z0 + δ z0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = z0 + δ z0 , (3.83)
x = H ( x0 + δ x0 , z0 + δ z0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = x0 + δ x0 , (3.84)
z = I ( x0 + δ x0 , z0 + δ z0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = z0 + δ z0 . (3.85)
Using a Taylor series to expand equations (3.82)-(3.85) about the reference, results in the
following first-order targeter,
⎡⎛ ∂F ⎞ ∂F ∂F ∂F ∂F ⎤
⎢⎜ − 1⎟ ⎥
⎢⎝ ∂x0 ⎠ ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ⎛ ∂G ⎞ ⎥ ⎧δ x ⎫
⎧δ x ⎫ ⎢ ∂G ∂G ∂G ∂G ⎥ ⎪ 0 ⎪
⎪δ z ⎪ ⎢ ∂x ⎜ − 1⎟ δz
⎪ ⎪ 0 ⎝ ∂z0 ⎠ ∂x0 ∂y0 ∂z0 ⎥ ⎪⎪ 0 ⎪⎪
⎨ ⎬=⎢ ⎥ ⎨δ x0 ⎬ . (3.86)
⎪δ x ⎪ ⎢ ∂H ∂H ⎛ ∂H ⎞
− 1⎟
∂H ∂H ⎥ ⎪
δ y0 ⎪
⎪⎩δ z ⎪⎭ ⎢ ∂x ⎜ ⎥ ⎪ ⎪
∂z0 ⎝ ∂x0 ⎠ ∂y0 ∂z0
⎢ 0
⎥ ⎪⎩δ z0 ⎪⎭
⎢ ∂I ∂I ∂I ∂I ⎛ ∂I ⎞⎥
⎢ ⎜ − 1⎟ ⎥
⎢⎣ ∂x0 ∂z0 ∂x0 ∂y0 ⎝ ∂z0 ⎠ ⎥⎦
To determine the partials in equation (3.86), let Q be defined as
79
⎡ ∂F ∂F ∂F ∂F ∂F ⎤
⎢ ∂x ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂G ∂G ∂G ∂G ∂G ⎥
⎢ ∂x ∂z0 ∂x0 ∂y0 ∂z0 ⎥
Q=⎢ ⎥.
0
(3.87)
⎢ ∂H ∂H ∂H ∂H ∂H ⎥
⎢ ⎥
⎢ ∂x0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂I ∂I ∂I ∂I ∂I ⎥
⎢ ⎥
⎣ ∂x0 ∂z0 ∂x0 ∂y0 ∂z0 ⎦
Then the partials in equation (3.86) are contained in Q such that the components of Q are
found using equation (2.61) to isolate the appropriate sensitivity partials. Thus,
⎡ ∂x ∂x ∂x ∂x ∂x ⎤
⎢ ∂x ∂z0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂y ∂y ∂y ∂y ∂y ⎥ ⎧x⎫
⎢ ∂x ∂z0 ∂x0 ∂y0 ∂z0 ⎥ ⎪ ⎪
Q=⎢
0 ⎥ − 1 ⎨⎪ z ⎬⎪ ⎧⎨ ∂y ∂y ∂y ∂y ∂y ⎫
⎬ . (3.88)
⎢ ∂x ∂x ∂x ∂x ∂x ⎥ y ⎪ x ⎪ ⎩ ∂x0 ∂z0 ∂x0 ∂y0 ∂z0 ⎭
⎢ ⎥
⎢ ∂x0 ∂z0 ∂x0 ∂y0 ∂z0 ⎥ ⎩⎪ z ⎭⎪
⎢ ∂y ∂y ∂y ∂y ∂y ⎥
⎢ ⎥
⎣ ∂x0 ∂z0 ∂x0 ∂y0 ∂z0 ⎦
Given the form of this targeter, a constraint to fix δ y0 is accomplished simply by
equating δ y0 to zero in equation (3.86). However, the additional equation increases the
speed of convergence of the algorithm. Furthermore, the targeter is only differentially
correcting the initial variations δ x0 , δ z0 , δ x0 , and δ z0 to compute a periodic solution, as
defined by the mappings in equations (3.82)-(3.85). Therefore it is not necessary to fix
δ y0 when computing the family. Of course, four equations in equation (3.86) implies an
infinite number of solutions. For this application, the solution with the smallest
Euclidean norm is selected. The initial state vector is updated via
δ x0 , δ z0 , δ x0 , δ y0 , and δ z0 in an iterative process until δ x, δ z , δ x, δ z < ε , thereby
obtaining a periodic orbit.
The predictions for q0n to uniformly move along the tangent subspace Г associated
with the family are available from equations (3.87) and (3.88). Let Δq0n −1 be defined as
80
Δq0n −1 = {Δx0n −1 0 Δz0n −1 Δz0n −1} ,

T
Δx0n −1 Δy0n −1 (3.89)
where the matrix equation

⎧ Δx0n −1 ⎫
⎧Δx0n −1 ⎫ ⎪ n −1 ⎪
⎪ n −1 ⎪ Δz0
⎪ Δz0 ⎪ ⎪⎪ n −1 ⎪⎪
⎨ n −1 ⎬ = [Q ] ⎨ Δx0 ⎬ , (3.90)
⎪Δx0 ⎪ ⎪Δy n −1 ⎪
⎪⎩ Δz0n −1 ⎭⎪ ⎪ 0 ⎪
n −1
⎩⎪ Δz0 ⎭⎪
must be satisfied such that for normalized Δq0n−1 ,
1 = ( Δx0n −1 ) + ( Δz0n −1 ) + ( Δx0n −1 ) + ( Δy0n −1 ) + ( Δz0n −1 ) .

2 2 2 2 2
(3.91)
Solving equations (3.90) and (3.91) simultaneously produces the components of Δq0n −1.
To obtain the entire family, the sign of S is changed when the following conditions are
simultaneously satisfied,
Δx0n − 2 ⋅ Δx0n −1 < 0, (3.92)
Δz0n − 2 ⋅ Δz0n −1 < 0, (3.93)
Δx0n − 2 ⋅ Δx0n −1 < 0, (3.94)
Δz0n − 2 ⋅ Δz0n −1 < 0. (3.95)

The process is continued until the desired family is obtained.
Motion in three-dimensional space with reference z0 . For some families of orbits, e.g.,
the L4/L5 axial orbits, no common reference y0 can be defined for every orbit in the
family. Therefore, it is necessary to change the reference to a different parameter. For
example, to obtain an L4 axial family let z0 define the reference. Then, the mappings are
defined as follows,
x = Q ( x0 , y0 , x0 , y0 , z0 ) , (3.96)
y = R ( x0 , y0 , x0 , y0 , z0 ) , (3.97)
x = T ( x0 , y0 , x0 , y0 , z0 ) , (3.98)
y = V ( x0 , y0 , x0 , y0 , z0 ) , (3.99)
81
such that each variable is mapped from the reference z0 to z = z0 . Then, for a periodic
orbit, equations (3.96)-(3.99) are rewritten in the following functional form,
x = Q ( x0 + δ x0 , y0 + δ y0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = x0 + δ x0 , (3.100)
y = R ( x0 + δ x0 , y0 + δ y0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = y0 + δ y0 , (3.101)
x = T ( x0 + δ x0 , y0 + δ y0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = x0 + δ x0 , (3.102)
y = V ( x0 + δ x0 , y0 + δ y0 , x0 + δ x0 , y0 + δ y0 , z0 + δ z0 ) = y0 + δ y0 . (3.103)
A Taylor series expansion about the reference solution to equations (3.100)-(3.103)

results in the following variational relationships
⎡⎛ ∂Q ⎞ ∂Q ∂Q ∂Q ∂Q ⎤
⎢⎜ − 1⎟ ⎥
⎢⎝ ∂x0 ⎠ ∂y0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ⎛ ∂R ⎞ ⎥ ⎧δ x ⎫
⎧δ x ⎫ ⎢ ∂R ∂R ∂R ∂R ⎥ ⎪ 0 ⎪
⎪δ y ⎪ ⎢ ∂x ⎜ − 1⎟ δy
⎪ ⎪ 0 ⎝ ∂y0 ⎠ ∂x0 ∂y0 ∂z0 ⎥ ⎪⎪ 0 ⎪⎪
⎨ ⎬=⎢ ⎥ ⎨δ x0 ⎬ . (3.104)
⎪δ x ⎪ ⎢ ∂T ∂T ⎛ ∂T
− 1
⎞ ∂T ∂T ⎥ ⎪
δ y0 ⎪
⎪⎩δ y ⎪⎭ ⎢ ⎜ ⎟ ⎥ ⎪ ⎪
∂x0 ∂y0 ⎝ ∂x0 ⎠ ∂y0 ∂z0
⎢ ⎥ ⎩⎪δ z0 ⎭⎪
⎢ ∂V ∂V ∂V ⎛ ∂V ⎞ ∂V ⎥
⎢ ⎜ − 1⎟ ⎥
⎣⎢ ∂x0 ∂y0 ∂x0 ⎝ ∂y0 ⎠ ∂z0 ⎦⎥
In order to obtain the partials in equation (3.104), let P be defined as

⎡ ∂Q ∂Q ∂Q ∂Q ∂Q ⎤
⎢ ∂x ∂x0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂R ∂R ∂R ∂R ∂R ⎥
⎢ ∂x ∂x0 ∂x0 ∂y0 ∂z0 ⎥
P=⎢ ⎥,
0
(3.105)
⎢ ∂T ∂T ∂T ∂T ∂T ⎥
⎢ ⎥
⎢ ∂x0 ∂x0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂V ∂V ∂V ∂V ∂V ⎥
⎢ ⎥
⎣ ∂x0 ∂x0 ∂x0 ∂y0 ∂z0 ⎦
such that P is computed using equation (2.61) and isolating the appropriate sensitivity
partials, i.e.,
82
⎡ ∂x ∂x ∂x ∂x ∂x ⎤
⎢ ∂x ∂y0 ∂x0 ∂y0 ∂z0 ⎥
⎢ 0 ⎥
⎢ ∂y ∂y ∂y ∂y ∂y ⎥ ⎧x⎫
⎢ ∂x ∂y0 ∂x0 ∂y0 ∂z0 ⎥ ⎪ ⎪
P=⎢
0 ⎥ − 1 ⎨⎪ y ⎬⎪ ⎧⎨ ∂z ∂z ∂z ∂z ∂z ⎫
⎬ . (3.106)
⎢ ∂x ∂x ∂x ∂x ∂x ⎥ z ⎪ x ⎪ ⎩ ∂x0 ∂y0 ∂x0 ∂y0 ∂z0 ⎭
⎢ ⎥ ⎪⎩ y ⎪⎭
⎢ ∂x0 ∂y0 ∂x0 ∂y0 ∂z0 ⎥
⎢ ∂y ∂y ∂y ∂y ∂y ⎥
⎢ ⎥
⎣ ∂x0 ∂y0 ∂x0 ∂y0 ∂z0 ⎦
The solution for the variations δ x0 , δ y0 , δ x0 , δ y0 , and δ z0 in equations (3.104) is

selected as that with the smallest Euclidean norm. The initial state vector is updated in an
iterative process until δ x, δ y, δ x, δ y < ε , thereby obtaining a periodic orbit.
For the predictions q0n to uniformly move along the tangent subspace Г, let Δq0n−1 be
defined as
Δq0n −1 = {Δx0n −1 Δy0n −1 0 Δx0n −1 Δy0n −1 Δz0n −1} ,

T
(3.107)
where
⎧ Δx0n −1 ⎫
⎧ Δx0n −1 ⎫ ⎪ n −1 ⎪
⎪ n −1 ⎪ Δy0
⎪Δy0 ⎪ ⎪⎪ n −1 ⎪⎪
⎨ n −1 ⎬ = [ P ] ⎨ Δx0 ⎬ . (3.108)
⎪ Δx0 ⎪ ⎪Δy n −1 ⎪
⎪⎩Δy0n −1 ⎭⎪ ⎪ 0 ⎪
n −1
⎩⎪ Δz0 ⎭⎪
Of course,
1 = ( Δx0n −1 ) + ( Δy0n −1 ) + ( Δx0n −1 ) + ( Δy0n −1 ) + ( Δz0n −1 ) ,

2 2 2 2 2
(3.109)
must also be satisfied. Solving equations (3.108) and (3.109) simultaneously produces
the components of Δq0n −1. The sign of S is switched when the following conditions are all
satisfied,
Δx0n − 2 ⋅ Δx0n −1 < 0, (3.110)
Δy0n − 2 ⋅ Δy0n −1 < 0, (3.111)
Δx0n − 2 ⋅ Δx0n −1 < 0, (3.112)

83
Δy0n − 2 ⋅ Δy0n −1 < 0. (3.113)

The process is repeated until the family of orbits is obtained.
3.4.2 Numerical Results: L4/L5 Planar, Axial, and Vertical Families of Orbits
The strategies and the corresponding algorithms from Section 3.4.1 are used to
compute asymmetric periodic orbits in the vicinity of the triangular points in the Earth-
Moon system. The orbit families with their bifurcating orbits are computed using a
method of bisections and appear in Figures 3.19-3.21. The initial conditions, periods, and
stability indices are provided in Tables 3.14-3.16. The largest amplitude L3 Lyapunov
orbit from Figure 3.7 corresponds to a bifurcation orbit intersecting the L4/L5 planar orbit
family. The L4 planar orbit family, as depicted in Figure 3.19, is computed using the
3
reference y = y0 = . The reference is selected such that all orbits in the family cross
2
the reference. (See Table 3.14 for corresponding initial conditions, periods, and stability
indices.) Furthermore, the bifurcation associated with the near-rectilinear L1 halo orbits
is a basis for computing the northern L4 axial orbit family. (Recall Figure 3.9.) The
reference z = z0 = 0.1 is selected such that it is common to all orbits in the family but,
also, not significantly close to the Moon. The northern L4 axial orbit family is
represented in Figure 3.20 with corresponding parameters presented in Table 3.15. The
southern orbit family is determined by reflecting the northern orbit family across the x-y
plane. The L4 axial orbit families terminate with an L4 vertical orbit. This orbit, with
reference y = y0 = 0.42545, is a basis to compute the L4 vertical orbit family. (See
Figures 3.21 and Table 3.16.) Then, the corresponding L5 vertical and axial orbit families
can be obtained by reflecting the L4 orbit families across the x-z plane. Thus, a mapping
between periodic orbits near L1 to orbits in the vicinity of L3 is available via the L4/L5
axial and vertical orbit families.
84
Earth-Moon System
1 unit = 385,692.5 km
μ = 0.0121505856
Bifurcation With L3
L4
1.5 Lyapunov Family
0.5
y
0 Earth
−0.5
Moon
−1
L3
−1.5
−2 −1.5 −1 −0.5 0 0.5 1
x
Figure 3.19 The L4 Planar Family

85
Table 3.14
3
L4 Planar Family Initial Conditions with y0 =
2
x0 x0 y0 T υ
0.4750 0.0697 -1.0915 6.2647 1.0000 L3 Lyapunovs
0.5837 0.0606 -1.0896 6.2657 1.0000
0.6853 0.0637 -1.0828 6.2691 1.0000
0.7656 0.0726 -1.0703 6.2747 1.0000
0.8312 0.0839 -1.0509 6.2829 1.0000
0.8826 0.0959 -1.0245 6.2937 1.0000
0.9203 0.1077 -0.9931 6.3064 1.0000
0.9459 0.1183 -0.9595 6.3201 1.0000
0.9621 0.1277 -0.9261 6.3338 1.0000
0.9717 0.1356 -0.8943 6.3470 1.0000
0.9775 0.1438 -0.8581 6.3621 1.0000
0.9788 0.1518 -0.8174 6.3791 1.0000
0.9753 0.1594 -0.7724 6.3977 1.0000
0.9665 0.1661 -0.7229 6.4178 1.0000
0.9520 0.1715 -0.6689 6.4391 1.0000
0.9314 0.1748 -0.6102 6.4612 1.0000
0.9043 0.1754 -0.5467 6.4836 1.0000
0.8736 0.1728 -0.4850 6.5035 1.0000
0.8401 0.1669 -0.4249 6.5210 1.0000
0.8043 0.1579 -0.3667 6.5361 1.0000
0.7667 0.1457 -0.3104 6.5488 1.0000
0.7274 0.1306 -0.2562 6.5592 1.0000
0.6867 0.1126 -0.2040 6.5675 1.0000
0.6447 0.0918 -0.1542 6.5738 1.0000
0.6015 0.0685 -0.1069 6.5783 1.0000
0.5572 0.0428 -0.0622 6.5812 1.0000
86
1.5
L4
1
Earth-Moon System
0.5 1 unit = 385,692.5 km
y
μ = 0.0121505856
0
L1
−0.5
−0.5 0 0.5 1 1.5
x
1 1
Moon
0.5 0.5
0 0
z
z
−0.5 Earth −0.5
−1 −1
−0.5 0 0.5 1 1.5 −0.5 0 0.5 1 1.5
x y
Figure 3.20 The Northern L4 Axial Family

87
z
Bifurcation with L4
Vertical Family
Bifurcation with L1
Halo Family
x
Table 3.15
Northern L4 Axial Family Initial Conditions with z0 = 0.1
x0 y0 x0 y0 z0 T υ
0.9767 -0.0334 -0.0737 -0.0626 0.4671 2.1509 1.0000 L1 Halos
0.9522 -0.0146 -0.1089 0.0426 0.4599 2.1684 1.2665
0.9353 -0.0019 -0.1254 0.1104 0.4375 2.2359 1.4014
0.9167 0.0122 -0.1328 0.1753 0.4026 2.3433 1.9466
0.8958 0.0292 -0.1297 0.2354 0.3601 2.4992 2.7963
0.8580 0.0661 -0.1077 0.3173 0.2917 2.8625 4.3600
0.8342 0.0949 -0.0924 0.3566 0.2589 3.1374 5.0376
0.8067 0.1326 -0.0815 0.3939 0.2321 3.4831 5.3890
0.7717 0.1819 -0.0838 0.4337 0.2144 3.9253 5.3009
0.7080 0.2548 -0.1289 0.4969 0.2175 4.6214 4.5915
0.6695 0.2848 -0.1739 0.5331 0.2351 4.9622 4.1337
0.6301 0.3060 -0.2311 0.5692 0.2645 5.2632 3.6779
0.5946 0.3174 -0.2918 0.5997 0.3028 5.4992 3.2707
0.5620 0.3221 -0.3562 0.6244 0.3522 5.6920 2.8854
0.5326 0.3217 -0.4220 0.6401 0.4148 5.8483 2.5153
0.5074 0.3183 -0.4850 0.6421 0.4922 5.9719 2.1603
0.4874 0.3140 -0.5394 0.6237 0.5849 6.0671 1.8235
0.4740 0.3117 -0.5768 0.5776 0.6899 6.1379 1.5161
0.4713 0.3236 -0.5690 0.3981 0.8913 6.2216 1.1156
0.4848 0.3462 -0.5045 0.2472 0.9807 6.2495 1.0403
0.5014 0.3731 -0.4279 0.1236 1.0239 6.2641 1.0202
0.5208 0.4071 -0.3372 0.0078 1.0440 6.2743 1.0113
0.5409 0.4472 -0.2392 -0.0943 1.0455 6.2816 1.0067
0.5621 0.4981 -0.1267 -0.1894 1.0313 6.2873 1.0038
0.5823 0.5617 -0.0005 -0.2722 1.0021 6.2917 1.0019
0.5982 0.6402 0.1380 -0.3369 0.9591 6.2948 1.0007
0.6035 0.7821 0.3538 -0.3843 0.8750 6.2966 1.0000 L4 Verticals
88
1 L4
0 1 unit = 385,692.5 km
y
μ = 0.0121505856
−0.5
L
−1 3
−1 0 1
x
1 Moon 1
0.5 0.5
0 0
z
z
−0.5 −0.5
−1 Earth −1
−1 0 1 −1 0 1
x y

89
z
Bifurcation with L4
Axial Family
y x
Bifurcation with L3
Vertical Family
Table 3.16
L4 Vertical Family Initial Conditions with y0 = 0.42545
x0 z0 x0 y0 z0 T υ
0.6032 0.6678 -0.3544 -0.3857 0.5728 6.2966 1.0000 L4 Axials
0.6390 0.6333 -0.3106 -0.4867 0.6482 6.2977 1.0012
0.6617 0.6093 -0.2749 -0.5583 0.6967 6.2985 1.0020
0.6825 0.5857 -0.2350 -0.6310 0.7410 6.2995 1.0028
0.7013 0.5631 -0.1915 -0.7050 0.7803 6.3005 1.0035
0.7180 0.5416 -0.1444 -0.7806 0.8138 6.3015 1.0040
0.7325 0.5219 -0.0942 -0.8577 0.8407 6.3025 1.0044
0.7450 0.5042 -0.0412 -0.9364 0.8600 6.3035 1.0047
0.7553 0.4889 0.0143 -1.0162 0.8708 6.3045 1.0048
0.7636 0.4763 0.0718 -1.0966 0.8722 6.3055 1.0048
0.7700 0.4666 0.1311 -1.1767 0.8636 6.3064 1.0047
0.7745 0.4598 0.1918 -1.2554 0.8443 6.3072 1.0045
0.7773 0.4560 0.2537 -1.3314 0.8142 6.3080 1.0042
0.7782 0.4552 0.3167 -1.4031 0.7731 6.3086 1.0039
0.7775 0.4573 0.3806 -1.4691 0.7210 6.3092 1.0036
0.7753 0.4620 0.4411 -1.5237 0.6631 6.3097 1.0032
0.7718 0.4686 0.4981 -1.5674 0.6015 6.3101 1.0030
0.7668 0.4773 0.5553 -1.6025 0.5335 6.3104 1.0027
0.7604 0.4881 0.6127 -1.6278 0.4603 6.3106 1.0025
0.7525 0.5009 0.6701 -1.6424 0.3832 6.3108 1.0023
0.7428 0.5156 0.7271 -1.6453 0.3038 6.3110 1.0022
0.7314 0.5321 0.7836 -1.6361 0.2238 6.3110 1.0021
0.7167 0.5521 0.8442 -1.6120 0.1380 6.3111 1.0021 L3 Verticals
90
3.5 Alternative Methods
Other approaches are also available for generating periodic orbits in the CR3BP. Due
to the sensitivity of the problem, some computational methods converge better than
others depending on the geometry of the solution. Ultimately, the most useful tool is one
that employs an automated procedure to map all solutions of interest. Two alternatives
offer different perspectives in approaching the problem of computing periodic orbits in
the CR3BP. One approach formulates the search for periodic orbits in terms of an
optimization problem. With parameter optimization, specifically Sequential Quadratic
Programming (SQP), additional orbits in the L3 halo family are determined. (See Figure
3.11.) Returning to the two-level corrector, the final convergence method to be
implemented involves discretizing the path into a series of patch points and adding
constraints in the two-level corrections algorithm to converge to a periodic orbit. For
both methods it is not necessary to exploit symmetry. All approaches produce families
that are, ultimately, entwined.
3.5.1 Sequential Quadratic Programming: Completing the L3 Halo Orbit Family
The Sequential Quadratic Programming (SQP) algorithm employed here uses a

method posed by Broydon, Fletcher, Shanno, and Goldfarb to approximate the
Lagrangian and Hessian [46]. Bounding the problem effectively reduces the design space
to within ±0.2 non-dimensional units of the initial guess. This increases the speed of
convergence and maintains a directed search that will produce orbits with the desired
characteristics. For the collinear halo orbits, a possible form for the initial state vector for
integration is,
q0 = { x0 0} ,
T
0 z0 0 y0 (3.114)
where q0 corresponds to the apse point along the orbit and perpendicular to the x-z plane.
Then, for the endpoint state q , the problem is simply formulated as least-squares
minimization, i.e.,
min f ( v ) = q − q0 ,
2
(3.115)
with design variables v . If it is desired to fix x0 , the design variables are

91
v = { z0 y0 τ } ,
T
(3.116)
subject to the bounds

z0i − 0.2 ≤ z0 < z0i + 0.2, (3.117)
y0i − 0.2 ≤ y0 < y0i + 0.2, (3.118)
τ 0 − 0.2 ≤ τ 0 < τ 0 + 0.2,

i i
(3.119)
where z0i = z0 , y0i = y0 , and τ 0i = τ 0 before the minimization of f. Similarly, if it is
desired to fix z0 , the design variables are
v = { x0 y0 τ } ,
T
(3.120)
subject to the bounds

x0i − 0.2 ≤ x0 < x0i + 0.2, (3.121)
y0i − 0.2 ≤ y0 < y0i + 0.2, (3.122)
τ 0 − 0.2 ≤ τ 0 < τ 0 + 0.2,

i i
(3.123)
where x0i = x0 , y0i = y0 , and τ 0i = τ 0 before the minimization of f. Notice that the
formulation of equation (3.115) does not necessarily converge to a perpendicular

crossing. If it is desired to achieve a perpendicular crossing at the endpoint, equation
(3.115) is simply redefined to be
min f ( v ) = ( x − x0 ) + ( z − z0 ) + ( y − y0 ) .
2 2 2
(3.124)
If either equation (3.115) or (3.124) is used for the minimization, predictions for the next
orbit along the family are detailed in Section 3.2.1, for motion perpendicular to the x-y
plane.
The entire process is also easily adapted to obtain asymmetric periodic solutions.
However, improvements to be made to the current design are recommended. For
example, the current design uses numerical gradients to approximate the Lagrangian and
Hessian. Modifying the process to include analytical gradients greatly increases the
speed of convergence. Furthermore, better predictions for neighboring periodic solutions
while expanding the solution space are available by exploiting the tangent Г space along
the family as presented in section 3.4.1. However, even without such modifications, the
92
algorithm successfully determines periodic solutions and steps along x0 or z0 to produce

an entire family of orbits. In fact, the L3 halo family in the Earth-Moon system is further
expanded using SQP in regions where the previous targeter scheme failed to converge.
(See Table 3.17 and corresponding orbits in orange.) Another way to compute these
orbits is to employ regularization [19].
x Bifurcation with L3
Lyapunov Family
Table 3.17
Northern L3 Halo Family Initial Conditions Using SQP (Orange)
x0 z0 y0 T υ
-1.1131 1.4708 0.8376 6.2224 1.2129
-1.0145 1.5658 0.7636 6.2173 1.1941
-0.9195 1.6449 0.6923 6.2114 1.1776
-0.8245 1.7130 0.6208 6.2042 1.1618
-0.7295 1.7713 0.5491 6.1952 1.1461
-0.6345 1.8204 0.4772 6.1838 1.1297
-0.5395 1.8606 0.4051 6.1686 1.1119
-0.4445 1.8920 0.3329 6.1472 1.0921
-0.3495 1.9138 0.2607 6.1148 1.0688
-0.2545 1.9237 0.1887 6.0590 1.0385
-0.1595 1.9131 0.1169 5.9394 1.0000
93
3.5.2 Two-Level Corrections with Constraints: Completing the L4 Vertical Orbit

Family
The second level of the two-level differential corrector is easily adapted to

accommodate constraints [25]. Periodicity can be modeled as a special type of
constraint. This is accomplished by the addition of requirements on the patch points. To
enforce periodicity, it is necessary to ensure that the patch point associated with the first
state is exactly the same as the patch point, i.e., the elements of the state vector,
associated with the last state. The convergence process is completely independent of any
symmetries associated with the solutions, and therefore asymmetric periodic solutions
can also be determined. Isolating a specific patch point in the orbit and constraining the
point to be an extremum enables a method of continuation to expand the solution space
and compute the family of orbits.
The periodicity constraint. For a periodic solution, in addition to minimizing the ΔV ’s at

the internal patch points (recall Figure 3.1), it is also necessary to minimize the values
ΔR1,n = R1 − Rn , (3.125)
ΔV1,n = V1 − Vn , (3.126)
where ΔR1,n and ΔV1,n are the discontinuities in position and velocity, respectively,
between the first and last point, i.e., the nth point. (See Figure 3.22.) Therefore, it is
necessary to determine the variations in δ Ri and δτ i that also minimize the endpoint
discontinuities. The endpoint discontinuities in position depend only on the first and nth
points. Therefore, using equation (3.125), the partial of ΔR1,n with respect to position and
time can be evaluated as follows,

∂ΔR1, n
= −I 3×3 , (3.127)
∂Rn
∂ΔR1, n
= 03×1 , (3.128)
∂τ n
94
∂ΔR1, n
= I 3×3 , (3.129)
∂R1
∂ΔR1, n
= 03×1. (3.130)
∂τ 1
However, the velocity discontinuity also depends on the outgoing state associated with
the nth − 1 patch point and the incoming state associated with the second patch point.
Therefore, via equation (3.126) the partials of ΔV1,n with respect to position and time are
computed from the following relationships,

∂ΔV1, n ∂ΔV1,n ∂Vn− ∂Vn−
= = − , (3.131)
∂Rn −1 ∂Vn− ∂Rn −1 ∂Rn −1
∂ΔV1,n ∂ΔV1,n ∂Vn− ∂Vn−

= = − , (3.132)
∂τ n −1 ∂Vn− ∂τ n −1 ∂τ n −1
∂ΔV1, n ∂ΔV1,n ∂Vn− ∂Vn−

= = − , (3.133)
∂Rn ∂Vn− ∂Rn ∂Rn
∂ΔV1, n ∂ΔV1,n ∂Vi − ∂Vn−

= = − , (3.134)
∂τ n ∂Vn− ∂τ n ∂τ n
∂ΔV1,n ∂ΔV1,n ∂V1+ ∂V1+

= = , (3.135)
∂R1 ∂V1+ ∂R1 ∂R1
∂ΔV1,n ∂ΔV1,n ∂V1+ ∂V1+

= = , (3.136)
∂τ 1 ∂V1+ ∂τ 1 ∂τ 1
∂ΔV1,n ∂ΔV1,n ∂V1+ ∂V1+

= = , (3.137)
∂R2 ∂V1+ ∂R2 ∂R2
∂ΔV1,n ∂ΔV1,n ∂V1+ ∂V1+

= = , (3.138)
∂τ 2 ∂V1+ ∂τ 2 ∂τ 2
where the partials of Vn− and V1+ with respect to the nth − 1, nth , 1st, and 2nd point
positions and times are available from equations (3.13)-(3.20). Therefore, the new
targeter that incorporates periodicity constraints takes the form
95
⎧ δ R1 ⎫
⎪ ⎪
⎪ δτ 1 ⎪
⎧ δΔV2 ⎫ ⎪ δ R2 ⎪
⎪ ⎪ ⎪ ⎪
⎪ δΔV 3 ⎪ ⎪ δτ 2 ⎪
⎪ ⎪ ⎪ δR ⎪
⎪ ⎪ * ⎪
⎪ 3 ⎪⎪
= ⎡
⎨δΔVn − 2 ⎬ ⎣ ⎦ ⎨
M ⎤ ⎬, (3.139)
⎪δΔV ⎪ ⎪δτ ⎪
⎪ n −1
⎪ ⎪ n−2 ⎪
⎪ δΔR1,n ⎪ ⎪δ Rn −1 ⎪
⎪ ⎪ ⎪ ⎪
⎩ δΔV1,n ⎭ ⎪ δτ n −1 ⎪
⎪ δ Rn ⎪
⎪ ⎪
⎩⎪ δτ n ⎭⎪
where
⎡ ⎤
⎢ M ⎥
⎢ ⎥
⎢ ⎥
M* = ⎢ ⎥ , (3.140)
⎢ I 3×3 03×1 03×3 03×1 03×3 03×1 03×3 03×1 −I 3×3 03×1 ⎥
⎢ + + + +
∂Vn− ∂Vn− ∂Vn− ∂Vn− ⎥⎥
⎢ ∂V1 ∂V1 ∂V1 ∂V1 03×3 03×1 − − − −
⎢⎣ ∂R1 ∂τ 1 ∂R2 ∂τ 2 ∂Rn −1 ∂τ n −1 ∂Rn ∂τ n ⎥⎦
with M evaluated from the expressions in equation (3.22). Notice that the form of
equation (3.140) requires at least four patch points be specified. Use the smallest
Euclidian norm of M as the solution to equation (3.139) for the variations in position and
time. Subtract these variations from the existing patch point positions and times to
update the patch point states to reduce the internal ΔV ’s and endpoint discontinuities in
the next iteration. The process is continued until ΔVi , ΔR1,n , ΔV1,n < ε , thereby
obtaining a periodic solution. (See Figure 3.22.)

96
ΔVi < ε
Vi − Vi + ΔVi +1 < ε
ΔVi −1 < ε
Vi −+1 Vi −−1
V3+ Vn−−2
ΔV3 < ε V3− Vn+−2 ΔVn−2 < ε
V2+ Vn−−1
V2−
Vn+−1
V1 Vn ΔVn−1 < ε
ΔV2 < ε
ΔR1,n , ΔV1,n < ε
Figure 3.22 Path Before Second Level is Applied (Black)

and Final Periodic Solution (Blue)
Extrema constraints for continuation. To obtain a neighboring solution, it is necessary to

control the extrema points of the converged solution. For example, a patch point to be
associated with a neighboring solution can be predicted or estimated by fixing the step
size Δz0 and applying equations (3.28) and (3.46) to a point along the current solution.
However, unless additional constraints are added, the corrections procedure will simply
return to the same known periodic solution. Constraints are added to ensure that the
patch point is associated with zmax and that variations δ zmax are not applied during the
new convergence process.
In general, adding constraints will eventually produce an over-constrained problem.
However, this problem is easily avoided by specifying additional patch points along the
97
path as well, thereby increasing the number of degrees of freedom. However, for only
four patch points, there exist sixteen degrees of freedom. Reducing the ΔV ’s associated
with the internal state requires three equations of constraint. Adding the periodicity
constraint and the extremum constraint with δ zmax = 0 implies eleven total equations of
constraint. Since there are still more degrees of freedom than constraints, the problem
possesses a solution.
Furthermore, since there are already six equations of constraint directly associated
with first and last patch points, additional patch points are carefully selected when
applying the extremum constraint. Therefore, to constrain the second patch point such
that z0 = zmax , a constraint is also required to force z0 = 0. Since the incoming velocity
associated with the second patch point is already being used to constrain periodicity, the
outgoing velocity is employed to aid in satisfying the constraint z0 = 0. Notice that
equations (3.17)-(3.20) can be rewritten for the second patch point in the form
⎡ ∂x2+ ⎤ ⎡ ∂x2+ ⎤ ⎡ ∂x2+ ⎤ ⎡ ∂x2+ ⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ∂R2 ⎥ ⎢ ∂τ 2 ⎥ ⎢ ∂R3 ⎥ ⎢ ∂τ 3 ⎥
∂V2+ ⎢ ∂y2+ ⎥ ∂V2+ ⎢ ∂y2+ ⎥ ∂V2+ ⎢ ∂y2+ ⎥ ∂V2+ ⎢ ∂y2+ ⎥
=⎢ ⎥, =⎢ ⎥, =⎢ ⎥, =⎢ ⎥ . (3.141) -(3.144)
∂R2 ⎢ ∂R2 ⎥ ∂τ 2 ⎢ ∂τ 2 ⎥ ∂R3 ⎢ ∂R3 ⎥ ∂τ 3 ⎢ ∂τ 3 ⎥
⎢ +⎥ ⎢ ∂z + ⎥ ⎢ +⎥ ⎢ ∂z + ⎥
⎢ ∂z2 ⎥ ⎢ 2⎥ ⎢ ∂z2 ⎥ ⎢ 2⎥
⎢⎣ ∂R2 ⎥⎦ ⎢⎣ ∂τ 2 ⎥⎦ ⎢⎣ ∂R3 ⎥⎦ ⎢⎣ ∂τ 3 ⎥⎦
Then the targeter for the second level of the corrections process, including the constraint
z0 = 0 is,
⎧δ R1 ⎫
⎧ δΔV2 ⎫ ⎪ ⎪
⎪ ⎪ ⎪ δτ 1 ⎪
⎪ δΔV3 ⎪ ⎪δ R2 ⎪
⎪ ⎪ ⎪ ⎪
⎪ ⎪ ⎪ δτ 2 ⎪
⎪δΔVn − 2 ⎪ ** ⎪ ⎪
⎨ ⎬ = ⎡⎣ M ⎤⎦ ⎨δ R3 ⎬ , (3.145)
⎪ δΔVn −1 ⎪ ⎪ δτ ⎪
⎪ δΔR1,n ⎪ ⎪ 3⎪
⎪ ⎪ ⎪δ R4 ⎪
⎪ δΔV1,n ⎪ ⎪ ⎪
⎪ δz ⎪ ⎪ ⎪
⎩ 2 ⎭
⎪⎩ δτ n ⎭⎪
98
for
⎡ ⎤
⎢ M * ⎥
⎢ ⎥
M** = ⎢ ⎥. (3.146)
⎢ + + + + ⎥
⎢0 ∂z ∂z ∂z ∂z
03×1 2 2 2 2
03×3 03×1 ⎥
⎢ 3×3 ∂R2 ∂τ 2 ∂R3 ∂τ 3 ⎥
⎣ ⎦
Notationally, z2 is the z component of velocity for the second patch point. Also, the
variation δ z2 is defined as zero in equation (3.145). The next solution in the family is
determined using a fixed step Δz0 and equations (3.28) and (3.46) are applied to the
second patch point. The process is repeated until a family of solutions is obtained.
This process is used to complete the L4 vertical orbit family originally seen in Figure
3.21 and Table 3.16. The initial conditions associated with the first orbit are listed in
Table 3.16. These were numerically simulated over the period of motion and discretized
into five patch points. Convergence is obtained using equation (3.145). The remaining
portion of the family is computed using continuation and a fixed step Δz0 where Δz0 < 0.
The converged orbits appear in Figure 3.23 and the initial conditions are listed in Table
3.18. Note that the family is determined by stepping through values of z for one patch
point where the corresponding velocity z is forced to zero. This approach produces
favorable results but, in general, is very dependent on the geometry of the solution.
Furthermore, for some periodic solutions, a step through other parameters associated with
different patch points may be superior. However, the constraints corresponding to the
various other parameters must be derived and then applied to the patch point of interest.
When considering all possible cases, the problem becomes very complicated, perhaps
intractable. Alternatively, the continuation process can be enhanced by simultaneously
predicting initial states for all the patch points. This is accomplished by shifting along
the tangent Г as discussed in Section 3.4.1. Such a scheme also aids in ‘following’ the
solution through unpredictable twisting or curving paths as in the configuration space
seen in the L4 axial orbits.
99
Constrained Patch Points (Red)

δz = 0
0.8
0.6
Earth-Moon System
1 unit = 385,692.5 km
0.4
μ = 0.0121505856
0.2
0 L4
z
−0.2
−0.4
−0.6
−0.8
0.8
0.6
0.4
0.2 0.4 0.6
y x
Figure 3.23 The L4 Vertical Orbits Using a

Two-Level Corrector with Constraints
100
z
Bifurcation with L4
Axial Family
x
y
Bifurcation with L3
Vertical Family
Table 3.18

Using Two-Level Corrector with Constraints (Green)
x0 y0 z0 x0 y0 T υ
0.3006 0.4431 0.8411 -0.3762 0.2559 6.2951 1.0000
0.3308 0.4921 0.8011 -0.3299 0.2221 6.2938 1.0000
0.3558 0.5355 0.7611 -0.2897 0.1927 6.2926 1.0000
0.3769 0.5745 0.7211 -0.2542 0.1669 6.2915 1.0000
0.3947 0.6097 0.6811 -0.2224 0.1441 6.2905 1.0000
0.4099 0.6418 0.6411 -0.1939 0.1238 6.2896 1.0000
0.4229 0.6711 0.6011 -0.1680 0.1059 6.2888 1.0000
0.4340 0.6978 0.5611 -0.1446 0.0900 6.2880 1.0000
0.4435 0.7222 0.5211 -0.1234 0.0758 6.2873 1.0000
0.4517 0.7444 0.4811 -0.1042 0.0632 6.2866 1.0000
101
4. ORBIT SELECTION FOR LUNAR SOUTH POLE COVERAGE AND

THE TRANSITION TO A FULL EPHEMERIS MODEL
The various families of orbits and their connections to each other, as described in
Chapter 3, can be key components for actual mission design. Many of the orbits
computed possess geometries favorable for lunar south pole coverage. For example,
spacecraft in L1 and L2 southern halo orbits are in direct view of the lunar south pole for
nearly their entire period of motion, a favorable characteristic for lunar south pole
coverage. Thus, the initial design begins in the CR3BP. The known families of orbits
with their computed characteristics form a solid basis for selecting orbits to support lunar
south pole coverage. The periods and stability indices associated with the orbits offer
valuable information for initial design. Nine different orbits with periods ranging from 7
to 16 days are selected and transferred to a full ephemeris model, including solar
perturbations. A preliminary coverage analysis is discussed for potential coverage
architectures.
4.1 Orbit Selection for Lunar South Pole Coverage
Specific members of the known families of orbits in the Earth-Moon CR3BP are
isolated based on communications instruments and subsurface constraints, bounding the
useful range of orbits within a family to those with lunar altitudes between 50 km and
100,000 km. (See Figures 4.1–4.3.) Furthermore, it is useful to redefine the orbits in a
Moon centered, rotating frame such that
x m = x − (1 − μ ) , (4.1)
y m = y, (4.2)
z m = z, (4.3)
where xm, ym, and zm define the position of the spacecraft with respect to the Moon in the
rotating frame and μ = 0.0121505856, the associated mass ratio.
102
4
x 10
5
To Earth
km)
104 km)
4
zmw( (xx10
−5
Moon
−10
−8 −6 −4 −2 0 2 x 4 6 8
m 4 4 4
ux ( x( 10
x 10km)km) x 10
4
x 10
5
0
km)
104 km)
4
z ( (xx10
−5
m
w
−10
5
0 5
4 0
x 10 −5 −5 4
m
xx 10
44 44
yv ( (xx10
10 km)
km) xum( (xx10
10 km)
km)
Figure 4.1 Southern Halo Orbit Families:

Earth-Moon L1 (Orange) and L2 (Blue).
103
5
x 10
1
0.8 To Earth
0.6
0.4
zw ( x 10 km)
0.2
5
0
−0.2
m
−0.4
−0.6 Moon
−0.8
−1
−1 −0.5 0 0.5 1
m
x u ( x 105 km) x 10
5
5
x 10
1
0.5
zw ( x 105 km)
0
m
−0.5
−1 1
1
0.5 0
0 5
5 −0.5 x 10
x 10 −1 −1
m m 5
y v ( x 105 km) xu (( x 10 km)
Figure 4.2 Vertical Orbit Families of Interest:

Earth-Moon L1 (Red) and L2 (Cyan)
104
4
x 10
1 To Earth
−1
zw ( x 104 km)
−2
4
−3
m
−4
Moon
−5
−6
−7
−4 −2 0 x 2 4
m 44 4
xu ( (xx10
10 km)
km) x 10
4
x 10
1
0
−1
4 km)
km)
−2
4
z ( (xx1010
−3
m
−4
w
−5
−6
−7
4
2 4
0 2
4 −2 0
x 10 −4 −2
−4 xx 104
m 4
104 km)
yv ((xx10 km) m 4 4
ux( x( 10
x 10km)
km)
Figure 4.3 Southern L2 Butterfly Orbit Family

105
A number of L1 and L2 southern halo orbits satisfying the altitude constraints appear
in Figure 4.1. The L1 and L2 southern halo orbits are particularly effective in this problem
since the motion is almost always within line-of-sight to the Earth. Many of the orbits
pass very close to the Moon and are near-rectilinear in terms of the out-of-plane motion.
The near-rectilinear orbits possess a line-of-sight to the lunar south pole for almost the
entire period of the motion. Alternatively, the L1 and L2 vertical orbits with altitudes less
than 100,000 km are also feasible for coverage of the lunar south pole. (See Figure 4.2.)
These orbits bend toward both the north and south poles of the Moon, a favorable
characteristic for polar coverage. Finally, the southern butterfly orbits are also effective
for lunar south pole coverage. As evidenced in Figure 4.3, the butterfly orbits wrap
around both the near and far side of the Moon such that a direct line-of-sight to the lunar
south pole exists for nearly the entire orbital period. Of course, complete coverage is
only achieved when at least two vehicles are placed in one or more of these orbits.
4.1.1 Period and Stability Index
The time to complete one full period is a useful selection parameter in the initial
design phase. Let the maximum excursion distance, as it appears in the example halo
orbit in Figure 4.4, be defined as the maximum xm-distance for each orbit in the Moon
centered, rotating frame. Orbital periods are plotted against maximum excursion distance
during initial design selection, as indicated in Figure 4.5. Suitable regions for the
production of feasible architectures occur when the orbital periods are commensurate.
One such region consists of orbits in L1 and L2 halo families sharing periods between 7.9
and 12.2 days. An example that exhibits feasible south pole coverage consists of a 12-
day L1 and 12-day L2 halo orbit combination, illustrated by the black dashed line in
Figure 4.5. Another region with commensurate combinations consists of orbits with a
ratio of periods that is 2:1, i.e., one period is exactly twice that of the other. An example
from this region consists of a 14-day L2 butterfly orbit and a 7-day L2 halo orbit
combination, as noted by the grey dashed lines in Figure 4.5. The information in Figure
4.5 serves as a basis for the determination of many other commensurate orbit
106
combinations that lead to complete south pole coverage. Furthermore, orbits not
investigated here may be added to Figure 4.5 without affecting the analysis.
zm
xm
Max x m
Figure 4.4 Definition of Maximum xm-Distance
20 L Halo Family
1
L2 Halo Family
L Vertical Family
1
17.5 14-Day L2 Butterfly and 7-Day L2 L2 Vertical Family
Halo Orbit Combination L Butterfly Family
2
15
Period (days)
12.5
10
12-Day L1 Halo and 12-Day L2

7.5 Halo Orbit Combination
0 1 2 33 44 55 6 7 8
m 44 4
Max |u|
|x (| x( x1010km)
km) x 10
Figure 4.5 Period versus Maximum xm-Distance from the Moon

107
Also useful for design purposes is the stability index, υ, as defined in Section 3.2.1.
A stability index of one indicates a stable orbit, whereas stability indices greater than one
reflect instability. Of course, a large stability index indicates a divergent mode that
departs from the vicinity of the orbit very quickly. Generally, the stability index is
directly correlated to the station-keeping costs and is inversely related to transfer costs.
The stability indices for orbits from the various families appear in Figure 4.6 as functions
of maximum excursion distance from the Moon. In general, the stability index increases
with distance from the Moon.
1600 L Halo Family

1
L Halo Family
2
1400 L Vertical Family
1
L2 Vertical Family
1200 L2 Butterfly Family
Stability Index, ν
1000
800
600
400
200
0
0 1 2 33 44 55 6 7 8
4 4
Max |u|m(
Max |x | x( x1010km)km)
4
x 10
Figure 4.6 Stability Index versus Maximum xm-Distance from the Moon
108
4.1.2 Potential Orbits for Coverage of the Lunar South Pole
Using periodicity and stability criteria, orbits from Figures 4.1–4.3 are selected for
use in a coverage scenario. Either a single orbit is selected and two spacecraft are placed
in the same orbit (but, out-of-phase), or a unique orbit is selected for each spacecraft such
that their periods are commensurate. When an orbit with a precisely defined period is
desired, the modified two-level differential corrections scheme accepts the patch points of
a previously generated, neighboring orbit as an initial guess and generates a new orbit
with the specified period. When this process is complete for all desired orbits, patch
points are obtained; patch points for multiple revolutions are added to develop a baseline
180-day mission. Once the final trajectories are obtained, the initial time corresponding
to one of the two spacecraft is phase shifted by a half-period, thus allowing the greatest
chance for complete coverage of the lunar south pole.
For this investigation, nine different orbits from the families represented in Figures
4.1–4.3 are selected. Coverage can be adequately ensured with two spacecraft in just one
of these nine orbits by phasing the vehicles appropriately. However, as already noted,
selecting orbits with commensurate periods allows for architectures with combinations of
two different orbits for complete lunar south pole coverage. For example, rather than
placing two spacecraft in the same 7-day L2 near-rectilinear halo orbit, one spacecraft is
placed in a 7-day near-rectilinear halo and the other vehicle in a 14-day L2 butterfly orbit.
The combinations may provide more complete coverage. Possible orbits for use in
combination to ensure lunar coverage are presented in Table 4.1.
For example, one possible solution to the coverage problem employs two spacecraft
in the same 12-day L1 halo orbit. The lunar periapsis for this orbit is approximately
36,500 km with corresponding apoapsis at 81,200 km. One spacecraft is phase shifted by
a half-period, thus allowing the greatest chance for lunar south pole coverage, and the
relative spacecraft positions appear in Figure 4.7. Fifteen revolutions are added for a
baseline 180-day mission.
109
Table 4.1
Potential Orbits for Coverage of the Lunar South Pole
Orbit Type Libration Point Period (days) Stability Index
Near-Rectilinear Halo L2 7.0 1.00
Halo L1 12.0 60
Halo L2 14.0 115
Vertical L1 14.0 690
Butterfly L2 14.0 11.3
4
x 10
4
3
2
1
km)
0
104 km)
4
−1
Lunar South
zw (( xx 10
−2 Pole
m
−3
−4
−5
−6
5 2
0
−2
0 −4
4 −6 4
x 10 −5 −8 x 10
x
4
4 4
vym( x( x1010 km)
km) u (xxm 10 4
km)
( x 10 km)
Figure 4.7 12-Day L1 Orbit (Orange) and Patch Points (Black) from CR3BP
110
4.2 Transition to a Full Ephemeris Model
All orbits are initially designed under the assumption that a spacecraft is subject only
to the gravitational force of the Earth and Moon in the CR3BP. The Purdue software
package GENERATOR transitions trajectories computed in the CR3BP to a full
ephemeris model, including solar perturbations, with only small variations in shape.
Potential architectures for lunar south pole coverage are established and preliminary
coverage analyses are completed.
4.2.1 Obtaining Results with the Purdue Software Package GENERATOR
The Purdue software package GENERATOR [41] is a mission design tool that is
based on multi-body equations of motion including solar perturbations. Preliminary
baseline trajectories can quickly be determined within the context of the two-, three- or
four-body problem. Any number of bodies and the corresponding ephemeris information
can be incorporated as desired (other forces, control schemes, and design components are
available but not employed here). For this application, the 180-day baseline orbits
acquired by using the modified two-level differential corrector in the CR3BP are
transferred to the GENERATOR full ephemeris model. The patch point velocity
discontinuities are minimized in a two-level corrections procedure within GENERATOR
and a modified orbit emerges.
Recall that one possible scenario places two spacecraft in the same 12-day L1 halo
orbit. The orbits for the two spacecraft that result from GENERATOR appear in Figure
4.8 in both a Moon centered, rotating system of coordinates and the inertial Mean J2000
frame. Note that the trajectory of one spacecraft is plotted in blue; the motion of the
other spacecraft appears in orange. The quasi-periodic motion of the spacecraft is most
apparent in the Earth-Moon rotating frame where both orbit trajectories follow nearly the
same path. Note that the patch point positions (black) have been adjusted for continuous
motion in the full ephemeris model. In the Inertial Mean J2000 frame, the motion
appears to “umbrella” around the Moon.
111
4
4 x 10
x 10
3 To Earth
2
2
1
km)
104 km)
km)
( x 104 km)
0 0
4
4
zw ((xx10
−1
−2
m
−2
m
zw
−3
−4
−4 Moon
−6 0 −5
4 −2
2 −4
0 −2 −6
4 −4 −8 x 1x −6
x 10 4
−8 −6 −4 x−2 0
m 4
yv ( x 10 km) xum (( xx 10
104 km)
km) m ( x 1044 km)
xu ( x 10 km) x 10
4
4
x 10
5
km)
104km)
4
Wm((xx10
−0
Inertial Z
−
−5
−
5
5
4
x 10 0 0 4
x 10
−5 −5
m 44 km)
Inertial
InertialYV ( x 10 km) InertialXUm((xx10
4
Inertial 104km)
km)
Figure 4.8 Two Phased Spacecraft in 12-Day L1 Halo Orbits from GENERATOR;
Moon Centered, Rotating Reference Frame (Top Left), xm-zm Projection (Top Right);
Inertial Mean J2000 Reference Frame (Bottom)
112
4.2.2 A Preliminary Coverage Analysis
The resulting coverage schemes are initially analyzed by examination of the zm-
displacement of each of the spacecraft at the same instant of time. The zm-displacement
reflects the out-of-plane component of the position vector. The potential to maintain line-
of-sight to the lunar south pole exists only if at least one of the spacecraft is below the
Earth-Moon fundamental plane z m = 0 at all times. Consider two spacecraft in a single
L1 halo orbit. A typical two-spacecraft coverage scheme is achieved by displacing the
motion of each spacecraft by a half period. Thus, the two spacecraft are then phase
shifted in the L1 halo orbit and the zm-displacement of each spacecraft as a function of
time appears in Figure 4.9. The dashed line highlights the zm-value at which the two
spacecraft possess a common zm-component but are moving in opposite directions along
the orbit. The dashed line in Figure 4.9 also demonstrates that the zm-crossing occurs
16,500 km below the fundamental plane, ensuring that at least one spacecraft is always
within direct line-of-sight to the south pole. The first goal is to maximize the distance
between the dashed line and the fundamental plane ( z m = 0). This initial step does not
fully account for the actual position of the lunar south pole due to tilt and nutation of the
rotation axes, but provides an estimate for the line-of-sight coverage behavior over time.
Additionally, analyzing the zm-displacement in this way also offers a visual confirmation
that the proper periodicity constraint is implemented correctly, i.e., both spacecraft
maintain the prescribed phasing over the entire mission time.
4
x 10
4
km)
104km)
2
0
4
zm( (xx10
−2
−4
w
−6
0 20 40 60 80 100 120 140 160 180
time (days) m
Initial Time Phase Shift Line of Equal z Alignment
Figure 4.9 zm-Displacement in the Rotating Reference Frame

for Two Spacecraft in 12-Day L1 Halo Orbits
113
4.2.3 Architectures for Lunar South Pole Coverage
Besides two spacecraft in the same 12-day L1 halo orbit, other combinations of
various orbits from Table 4.1 also yield viable coverage options. For example, the L2 7-
day halo orbit combination, with the orbits plotted in Figure 4.10, also utilizes only one
halo orbit for both spacecraft. This near-rectilinear orbit passes approximately 2,750 km
from the lunar surface at periapsis, 72,100 km at apoapsis altitude, and possesses a
stability index of 1.00. Analyzing the GENERATOR zm-displacement components, the
scenario results in a distance between the dashed line for equal zm alignment and the
fundamental plane of 55,000 km. (See Figure 4.10.)
The 8-day L1 and L2 halo orbit scenario depicts an orbit combination that utilizes halo
orbits in the vicinity of two distinct libration points (Figures 4.11). Like the 7-day
scenario, the 8-day L1 and L2 halo orbits offer low minimum altitudes at approximately
1,400 km and 6,000 km, with apoapsis altitudes of 90,200 km and 76,900 km, and
stability indices of 1.25 and 1.00, respectively. The GENERATOR analysis in Figure
4.11 demonstrates that the minimum distance between the dashed line for equal zm
alignment below the fundamental plane is 60,000 km.
The 16-day L1 and L2 vertical orbit scenario also utilizes multiple libration point
orbits (Figure 4.12). The periapsis altitudes are much larger than the near-rectilinear halo
orbits at 48,600 km and 55,800 km, respectively. The corresponding apoapsis altitudes
are approximately 82,100 km for the L1 vertical orbit and 69,600 km for the L2 vertical
orbit. The corresponding stability indices are 370 and 515 for the L1 orbit and L2 orbit,
respectively. Due to the nature of the “figure-8” shape, the dashed line for equal zm
alignment occurs at the fundamental plane, suggesting brief intervals of time when both
spacecraft are not within line-of-sight of the south pole. (See Figure 4.12.) An alternative
scenario that employs the use of a vertical orbit for complete coverage is to place just one
spacecraft in a 14-day L1 vertical orbit and the other in a 14-day L2 halo orbit.
The orbits for two spacecraft in a single 14-day L2 butterfly orbit appear in Figure
4.13. The associated stability index of 11.3 is slightly higher than the near-rectilinear
halo orbits. As evidenced by Figure 4.13, the GENERATOR minimum zm-displacement
crossing occurs 36,000 km below the fundamental plane. The orbit possesses a low
114
periapsis altitude of 8,800 km. The corresponding apoapsis altitude for this orbit is
approximately 67,900 km.
Finally, Figure 4.14 depicts a unique 14-day L2 butterfly orbit and 7-day L2 halo orbit
scenario utilizing two different orbits from two different families. The scenario also
exploits a 2:1 ratio between the periods. For this specific scenario, the minimum zm-
displacement crossing occurs 45,000 km below the fundamental plane (Figure 4.14).
115
4
4 x 10
x 10
0
0
−1
−1
−2 −2 Moon
km)
km)
104 km)
1044km)
−3 −3
4
zw (( xx 10
zwm((xx10
−4
−4
m
−5
−5
−6
−6
To Earth
−7
4 −7
2 2 4
0 0
−2 −2
x 10
4 −4 −4 x 4 −4 −2 0 x2 4
m 4 x 10
104 km)
y v (( xx 10 km) m
x 410km)
ux( x(10
4
km) ux(mx(10
4 4
km)km)
x 10
4
x 10
4
x 10
0
km)
1044 km)
−2
InertialZW ( xx 10
−
m
−4
−
Inertial
−
−6
−
4
4 2 2
x 10 0 4
0 −2 x 10
m
Inertial Y V( (xx 10 km)
Inertial 1044km) 4 4
Inertial Xm
Inertial U ( x( 10
x 10km)
km)
4
x 10
km)
104 4km)
0
−2
zm( (xx10
−4
−6
w
−8
0 20 40 60 80 100 120 140 160 180
time (days)
Figure 4.10 Two Phased Spacecraft in 7-Day L2 Halo Orbits from GENERATOR;
Inertial Mean J2000 Reference Frame (Center); zm-Displacement from
Rotating Reference Frame (Bottom)
116
4
x 10
4
x 10 0 To Earth
−1
0
−2
Moon
−2 −3
km)
104 km)
km)
104 km)
−4
4
−4
4
w (( xx 10
z w (( xx 10
−5
−6
zm
m
−6
−8 −7
5 −8
5
0 0 4
−9
4 x 10
x 10 −5 −5 x −5 0 x 5
4
yvm((xx10
4
104km)
km) xum( (x x10104km)
km) u (xm
x 10
4 4
km)km)
( x 10
4
x 10
4
x 10
0
km)
104 km)
−2
4
Zm((xx10
−
−4
Inertial W
−
Inertial
−6
−
−8
−
6
4 5
4 2
x 10 0
0 x 10
4
−2
Inertial m
InertialYV ( x 10
4
104 km)
km) −5 4
Inertial XUm((xx10
Inertial 104km)
km)
4
x 10
km)
104km)
0
4
wzm( (xx10
−5
−10
0 20 40 60 80 100 120 140 160 180
time (days)
Figure 4.11 Two Phased Spacecraft in 8-Day L1 and L2 Halo Orbits from GENERATOR;
117
4
x 10
8
4
x 10
6
4
5 To Earth
2
km)
104 km)
km)
104 km)
4
0
zmw((xx10
4
zm((xx10
0
−2
w
−5 −4
Moon
5
−6
0 4
5 x 10
0 −5 x −8
4 −5 −5 0 x 5
x 10 m
m 44
ux ( x( x1010 km)
km)
yv ((xx10 km)
km) 1044 xum((xx10
4
104km)
km) x 10
4
4
x 10
km)
10 km)
5
44
Z ( (xx10
−
0
m
Inertial W
−
Inertial
−
−5
5
5
4 0
x 10 0 4
−5 −5 x 10
m
InertialYV ( x 104 km)
Inertial InertialXUm (( xx 10
4
Inertial 104 km)
km)
5
x 10
1
( x 105 km)
0
m
zw
−1
0 20 40 60 80 100 120 140 160 180
time (days)
Figure 4.12 Two Phased Spacecraft in 16-Day L1 and L2 Vertical Orbits from
GENERATOR; Moon Centered, Rotating Reference Frame (Top Left), xm-zm Projection
(Top Right); Inertial Mean J2000 Reference Frame (Center); zm-Displacement from
118
4
x 10
4
x 10 1
0
0
km)
−1
104km)
4 km)
10 km)
−2
4
−2
zmw((xx10
Moon
4
z w(( xx 10
−4 −3
m
−6 −4
4
4 −5 To Earth
2
2
4 0 0 −6
x 10 4
−2 −2 x 10
−4 −4 x −4 −2 0 x2 4
4
yvm (( xx 10
104 km)
km) m
ux( x( 104 4
x 10km)
km) ux(mx(10 4 4
km)
x 10 km) 4
x 10
4
x 10
2
km)
10 km)
0
4 4
Wm(( xx 10
−
−2
Inertial Z
−
−4
−
−6
−
−8
6
4 5
4 2
x 10 0 0 4
x 10
−2
Inertial m
1044 km) −5
InertialYV (( xx 10 km) Um( (xx10
Inertial X
Inertial
4
104km)
km)
4
x 10
2
km)
104km)
0
4
−2
wzm( (xx10
−4
−6
0 20 40 60 80 100 120 140 160 180
time (days)
Figure 4.13 Two Phased Spacecraft in 14-Day L2 Butterfly Orbits from GENERATOR;
119
4
x 10 x 10
4
0 0
−1
km)
104km)
−2
km)
104km)
−2
4
Moon
z w((xx10
4
zmw( (xx10
−4 −3
m
−4
−6
−5
4 −6 To Earth
2 4
0 2
4 −2 0 −7
x 10 −2 4
−4 −4 xx 10 −4 −2 0 x 2 4
m 44 44
yv ( (xx10
10 km)
km) uxm
( x( x1010km)
km) xum( (xx10
4
104km)
km) x 10
4
4
x 10
0
km)
104 km)
4
−2
Zm((xx10
−
−4
Inertial W
−
Inertial
−6
−
−
−8
6
4 5
4 2
x 10 0
0 x 10
4
m 44
−2
Inertial
InertialYV (( xx 10
10 km)
km) −5 4
Inertial Um((xx10
Inertial X 104 km)
km)
4
x 10
2
km)
1044 km)
0
−2
zwm((xx 10
−4
−6
−8
0 20 40 60 80 100 120 140 160 180
time (days)
Figure 4.14 Two Phased Spacecraft in a 7-Day L2 Halo Orbit and a 14-Day L2 Butterfly
Orbit; Moon Centered, Rotating Reference Frame (Top Left), xm-zm Projection (Top
Right); Inertial Mean J2000 Reference Frame (Center); zm-Displacement from
120
5. COVERAGE AND STATION-KEEPING ANALYSES

FOR LUNAR SOUTH POLE COVERAGE
Mission scenarios for lunar south pole coverage have been developed in support of
activities at NASA Goddard Spaceflight Center. Thus, trajectories designed in the
Purdue software package GENERATOR are further evaluated in the NASA Goddard
version of the Analytical Graphics, Inc. software package Satellite Tool Kit (STK®). To
accomplish such a task, the solutions obtained from GENERATOR are confirmed by
targeting GENERATOR patch points in STK using STK’s Astrogator. The propagator in
the STK targeting sequence is an eighth-order, full Runge-Kutta-Verner integrator with
ninth-order error control including solar perturbations. In general, slight differences in
modeling and numerical precision require only small corrections at each point. For
example, transitioning from GENERATOR to STK using Astrogator Connect requires
less than 1 m/s per year in corrections, an insignificant size in comparison with the costs
necessary for station-keeping from other “real” errors. Obtaining the orbits in STK
provides a complete coverage analysis for potential ground stations. Using invariant
manifold theory, preliminary station-keeping analyses for the orbits in Table 4.1 are also
available.
5.1 Coverage Analysis
Assuming an omni-directional communications link, the satellites are only accessible

to the ground stations if they are within direct line-of-sight. Intervals when the ground
stations are unable to access either satellite are recorded.
5.1.1 Defining the Ground Stations and Establishing the Links
Consider two satellites in a 12-day L1 halo orbit as depicted in Figure 4.8. Using
Astrogator Connect to target the converged GENERATOR patch point positions results
in Figures 5.1. Due to constant exposure to sunlight and possibly the existence of frozen
121
volatiles, the lunar south pole is a likely location for a ground station on the Moon. In
addition, current interest is directed at exploration options expanding out from the south
pole. One such location that is of scientific interest is the site of the Shackleton Crater
[47]. Therefore, a facility is placed on the Moon with coordinates 89.9°S, 0.0°E
corresponding to the Shackleton Crater. (See Figure 5.2.)
Figure 5.1 STK 12-Day L1 Halo Orbits:

Moon Centered, Rotating Frame (Top);
Moon Centered, Inertial Mean J2000 Frame (Bottom)
122
Shackleton Crater (89.9°S, 0.0°E)
Figure 5.2 A Facility is Placed on Moon at the Shackleton Crater (89.9°S, 0.0°E)
Both satellites are placed in a constellation and the times when either satellite or both
satellites have line-of-sight access with the facility is computed via an access chain.
Furthermore, the line-of-sight access times between the lunar facility and each individual
satellite is available. (See yellow access beams in Figure 5.3.) The results for the 12-day
L1 halo orbit scenario are available in Table 5.1 for a simulation time of 173.99 days. At
least one satellite is always in direct line-of-sight of the Shackleton Crater.
Table 5.1
Percent Access Times for 12-Day L1 Halo Orbits and
Ground Station at the Shackleton Crater
Facility at Shackleton Crater
Only Satellite 1 63.51%
Both Satellites 27.42%
Either Satellite 100.00%
Simulation Time (days) 173.99
123
Figure 5.3 Line of Sight Access Beams (Yellow) with Lunar Facility
Information from the ground station regarding each satellite during access times is
also available. For example, a plot of the elevation angle of each satellite relative to the
facility as a function of time appears in Figure 5.4. As in Table 5.1, the plot demonstrates
that the desired coverage is achieved. In fact, from Figure 5.4 it is apparent that at least
one satellite is always 15° above the horizon as viewed from the facility (see black
dashed line in Figure 5.4). That is, constant communications is achieved anywhere inside
the Shackleton Crater only if the walls of the crater are inclined less than 15°.
Furthermore, there is a satellite between the elevation angles 0° and 15°, 27.42% of the
simulation time. Incorporated in Table 5.1 and Figure 5.4 is direct access with the true,
exact location of the ground facility (89.9°S, 0.0°E) including lunar librations from
wobble and nutation of the spin axis.
Alternatively, three fictional ground stations are placed in a new configuration along
the Earth equator separated by 120°. The line-of-sight access between the ground-based
Earth stations and the satellite constellation is computed via another access chain. The
results are representative of access between the Earth and the satellites, i.e., times when
Earth can communicate with either satellite. For two satellites in a 12-day L1 halo orbit,
each satellite always maintains line-of-sight with at least one ground station 100% of the
simulation time. Of course, this is also evident from Figures 5.1 and 5.3 where each
124
satellite is visibly always in direct line-of-sight with the Earth. Such a result is expected
for satellites in Earth-Moon halo orbits [45]. It might be more accurate for actual mission
analysis to specify a real transmitting site on the Earth for ground station communications
with the lunar-based satellites. A probable location for this ground station is the White
Sands Test Facility (WSTF) located in New Mexico (32.3°N, 106.8°W). (See Figure 5.5.)
A chain is created for access between the satellite constellation and a facility located at
WSTF. The times when each satellite, either satellite, or both satellites possess line-of-
sight with the facility are computed. As evidenced from Table 5.2, each satellite
maintains line-of-sight for nearly half the simulation time. Given Earth rotation, this is
an excellent result.
Region of
Dual
Coverage
Figure 5.4 Elevation of Each Satellite Above the Horizon as Viewed from
the Shackleton Crater Facility for 12-Day L1 Halo Orbit Scenario
125
White Sands Test Facility

(32.3°N, 106.8°W)
Figure 5.5 An STK Facility is Placed at the White Sands Test Facility
(32.3°N, 106.8°W)
Table 5.2
Percent Access Times for 12-Day L1 Halo Orbits and
Ground Station at the White Sands Test Facility
Facility at WSTF
Both Satellites 44.61%
Either Satellite 51.82%
126
An additional communications link between the satellites also provides valuable

information. For example, a satellite in direct line-of-sight with the lunar facility may
not, at the same time, possess line-of-sight with WSTF. However, if the other satellite
possesses line-of-sight in direct communications with WSTF, then a communications link
is established between the lunar facility and WSTF via a relay between the two satellites.
(See Figure 5.6.) For some orbits, line-of-sight does not exist when the Moon is between
the satellites, i.e., when one satellite is at periapsis and the other is at apoapsis. However,
for the configuration of two satellites in 12-day L1 halo orbits, there is line-of-sight
between both satellites 100% of the simulation time.
Satellite to WSTF
Ground Station Satellite to
(White) Shackleton Ground
Station (Yellow)
Satellite to Satellite
Link (Green)
Figure 5.6 Possible Communications Relay Between the White Sands Test Facility
and the Lunar Ground Station at the Shackleton Crater
127
5.1.2 Results for Architectures Supporting Lunar South Pole Coverage
Besides the 12-day L2 halo orbit combination, the entire coverage analysis is easily
repeated for the remaining architectures corresponding to Figures 4.10-4.14. The access
and elevation results for all the remaining constellations in this study appear in Figures
5.7-5.16 and Tables 5.3-5.7. Recall the plots of the out-of-plane phasing between
satellites in each figure. In general, smaller distances between the dashed line for equal
zm alignment and the fundamental plane, z m = 0, as computed in GENERATOR,
translate to smaller ranges over which the elevation angles of the two satellites allow
simultaneous coverage scenarios in STK. A smaller range over the elevation angles
reflects smaller time intervals over which the lunar facility has access with both satellites.
Furthermore, nearly all the coverage scenarios possess nearly 50% access with WSTF.
(See Tables 5.3-5.7.)
For the 7-day L2 halo orbit combination (Figure 5.7), complete coverage of the
facility occurs at all times, as observed from Figure 5.8. Both satellites simultaneously
achieve access with the lunar facility for elevation angles from 0° to 65° for 92.78% of
the simulation time. Therefore, if the walls of the Shackleton Crater are less than a 65°
incline, constant communications can be maintained anywhere inside the crater. Also
notable in Figure 5.7, is a time when both satellites are accessed by the Shackleton
facility, i.e., both yellow access beams are active. Since the 7-day L2 halo orbit is near-
rectilinear, the motion is similar to a highly elliptical two-body orbit. Therefore, during
the majority of the orbital period both satellites are within direct line-of-sight of the lunar
facility, leading to large periods of redundant coverage. In fact, one satellite possesses
line-of-sight with the Shackleton facility over 95% of the simulation time. Furthermore,
the satellites are fully accessible from the Earth 100% of the time with only small
intervals when one satellite cannot communicate with the other due to interference with
the Moon. That is, for 99.21% of the simulation time, the satellites possess direct line-of-
sight with each other.
The 8-day L1 and L2 halo orbit scenario shares many similarities with the 7-day L2
halo orbit scenario (Figure 5.9). Simultaneous coverage of the lunar facility is achieved
128
at approximately 0° to 60° in elevation angle, as apparent in Figure 5.10. That is, at least
one satellite is always 60° above the horizon. The satellite configured in the 8-day L1
halo orbit is in direct line-of-sight with the Shackleton facility nearly 99% of the
simulation time. As with the previous scenarios utilizing halo orbits, satellites placed in
the 8-day L1 and L2 halo orbits are always in direct access with the Earth. The satellites
possess line-of-sight with one another 99.73% of the simulation time.
Due to the nature of the “figure-eight” shape, no two-satellite vertical orbit
combination exhibits complete coverage of the lunar south pole, as evidenced with the
16-day L1 and L2 vertical orbit scenario. Neither satellite is within direct line-of-sight of
the Shackleton facility when both satellites cross the fundamental plane at the same time.
The elevation plot in Figure 5.12 confirms that there are, in fact, small intervals in time,
i.e., 1.39% percent of the simulation time, when the lunar facility cannot access either
satellite. At this same instant, the Moon passes directly between the satellites, and
therefore the satellites are also not able to communicate with each other. In fact, the STK
results in Figure 5.11 reveal an instant in time when no access beams (yellow) exist
between the lunar facility and either satellite. At this same instant, an access beam
(green) does not exist between the satellites either. Satellite-to-satellite line-of-sight
occurs only 94.33% of the simulation time. Furthermore, due to passage behind the
Moon, L2 vertical orbits do not possesses 100% line-of-sight access with the Earth.
However, since the out-of-plane motion for the vertical orbit is larger than the halo orbits,
the verticals are able to communicate with WSTF nearly 52% of the time. (See Table
5.5.)
Two satellites in 14-day L2 butterfly orbits, as depicted in Figure 5.13, exhibit the
same properties as the near-rectilinear halo orbit scenarios. Figure 5.14 demonstrates that
at least one satellite is always 45° above the horizon. That is, between 0° and 45° there is
simultaneous coverage, and from Table 5.6 this occurs 78.25% of the time. For 0.39% of
the simulation time, the Moon interferes with communication between the satellites.
Similar to a satellite in an L2 vertical orbit, a satellite in a 14-day L2 butterfly orbit passes
behind the Moon and therefore does not posses 100% line-of-sight with the Earth.
129
Finally, a constellation with one satellite in a 14-day L2 butterfly orbit and another
satellite placed in a 7-day L2 halo orbits appears in Figure 5.15. Simultaneous coverage
occurs for 85.57% of the time between elevations of 0° and 50° degrees. (See Figure 5.16
and Table 5.7.) Furthermore, the satellites are in direct line-of-sight with one another for
99.76% of the simulation time.
Figure 5.7 STK 7-Day L2 Halo Orbits:

Moon Centered, Inertial Frame (Bottom)
130
the Shackleton Crater Facility for 7-Day L2 Halo Orbit Scenario
Table 5.3
Percent Access Times for 7-Day L2 Halo Orbits
Facility at Facility at
Earth
Shackleton Crater WSTF
Only Satellite 1 96.41% 100.00% 47.26%
Only Satellite 2 96.37% 100.00% 47.21%
Both Satellites 92.78% 100.00% 45.72%
Either Satellite 100.00% 100.00% 48.75%
131
Figure 5.9 STK 8-Day L1 and L2 Halo Orbits:

132
the Shackleton Crater Facility for 8-Day L1 and L2 Halo Orbit Scenario
Table 5.4
Percent Access Times for 8-Day L1 and L2 Halo Orbits
Earth
Only L1 Satellite 98.67% 100.00% 46.13%
Only L2 Satellite 93.44% 100.00% 47.43%
Both Satellites 92.11% 100.00% 45.46%
133
Figure 5.11 STK 16-Day L1 and L2 Vertical Orbits:

134
the Shackleton Crater Facility for 16-Day L1 and L2 Vertical Orbit Scenario
Table 5.5
Percent Access Times for 16-Day L1 and L2 Vertical Orbits
Earth
Only L2 Satellite 51.47% 98.27% 49.29%
Only L1 Satellite 47.02% 100.00% 48.04%
Both Satellites 0.08% 98.27% 45.19%
135
Figure 5.13 STK 14-Day L2 Butterfly Orbits:

136
the Shackleton Crater Facility for 14-Day L2 Butterfly Orbit Scenario
Table 5.6
Percent Access Times for 14-Day L2 Butterfly Orbits
Earth
Only Satellite 1 89.10% 99.45% 46.90%
Only Satellite 2 89.15% 99.39% 46.61%
Both Satellites 78.25% 98.89% 45.29%
137
Figure 5.15 STK 7-Day L2 Halo and 14-Day L2 Butterfly Orbits:

138
the Shackleton Crater Facility for 7-Day L2 Halo and 14-Day L2 Butterfly Orbit Scenario
Table 5.7
Percent Access Times for 7-Day L2 Halo and 14-Day L2 Butterfly Orbits
Earth
Only 14-Day Satellite 89.22% 99.53% 47.18%
Only 7-Day Satellite 96.35% 100.00% 46.86%
Both Satellites 85.85% 99.53% 45.60%
139
5.2 Station-Keeping Analysis
Another important factor in determining architectures for lunar south pole coverage is
the station-keeping cost associated with each satellite. Once a baseline coverage scenario
is constructed, potential station-keeping costs are investigated. The unstable subspace
that is available from dynamical systems theory is used to develop a station-keeping
strategy. Results are provided for the nine orbits that are defined in Table 4.1.
5.2.1 The Station-Keeping Algorithm
A preliminary station-keeping analysis is constructed by modifying the targeting

sequence in Astrogator Connect. A schematic representation the station-keeping
algorithm for four patch points appears in Figure 5.17. Note that the green path
represents the reference orbit as computed directly from GENERATOR. Along the green
path, the GENERATOR patch points are numbered #1– #4 and are defined as the target
points in Astrogator. Note that the six-dimensional state vectors are decomposed into
{ }
T
three-dimensional position and velocity vectors, i.e., qi = RiT Vi T . However, the
vehicle is not actually associated with the reference state. Perturbing the satellite into the
unstable subspace, and determining the maneuvers necessary to offset the error and
maintain the orbit, ultimately yields a good approximation of the maximum station-
keeping cost [31-37]. So, the perturbations along the unstable direction are computed
{ δ V1T } . Recall that position and

T
from equation (2.66) and appear in magenta as δ R1T
velocity perturbations are represented by the displacement d along the unstable direction
in equation (2.66). These perturbations are analogous to navigation errors and the impact
of such errors is most significant if they are in the unstable direction. Therefore, the
magnitude of d is sized to be consistent with the magnitude of some average navigation
3σ position errors; that is, the value might typically cover a range from 2 to 3 km in
position. To account for slight differences between the unstable directions from the
CR3BP and the full model, a baseline 3σ position error of 5 km is used in this analysis.
{ δ V1T } are added to the initial state and propagation in

T
These perturbations δ R1T
140
Astrogator over the time interval necessary to reach point #2 moves the vehicle along a
new actual path (in blue). Of course, the state associated with the satellite at the end of
{ }.
T
this propagation is actually δ R2T δ V2T A correction maneuver, Δv2 , is computed
via a targeting sequence in Astrogator to actually reach the next target point #3. This
{ δ V1T } . The Δv2 from the

T
Δv2 is necessary to offset the error introduced by δ R1T
targeting sequence is adjusted by ± 2% to include hot (+) or cold (–) burn errors, i.e.,
δΔv2 . However, to incorporate possible errors, perturbations along the unstable direction
are again added. After the adjusted Δv2 + δΔv2 is implemented, the satellite is perturbed
{δ R δ V2T } in the unstable direction and again propagated forward. The process is
T
T
2
repeated for a specified time interval, i.e., a pre-determined number of target points. The
n −1
total station-keeping cost is Δv = ∑ Δvi for n target points.
i=2
Actual Path
Reference Path
Target Vector
Perturbing Terms (Invariant Manifold Theory) ⎧⎪ R ⎫⎪ ⎪⎧δ R ⎪⎫ ⎪⎧ 0 ⎪⎫ ⎪⎧ 0 ⎪⎫
⎨ ⎬+⎨
3
Burn Errors ⎬+⎨ ⎬+⎨
3
⎧⎪ R ⎫⎪ ⎬
δV Δv δΔv3 ⎭⎪
⎧⎪ R ⎫⎪ ⎧⎪δ R ⎫⎪ ⎧⎪ 0 ⎫⎪ ⎧⎪ 0 ⎫⎪ ⎩⎪V3 ⎭⎪ ⎪⎩ 3 ⎪⎭ ⎪⎩ 3 ⎪⎭ ⎩⎪
3
⎨ ⎬
⎨ ⎬+⎨ ⎬+⎨ ⎬+⎨
2 2
⎬ ⎩⎪V3 ⎭⎪
⎧⎪ R ⎫⎪ ⎪⎩V2 ⎪⎭ ⎩⎪ δ V2 ⎭⎪ ⎩⎪Δv2 ⎭⎪ ⎪⎩δΔv2 ⎭⎪
2
⎨ ⎬
⎪⎩V2 ⎪⎭
⎧⎪ R ⎫⎪ ⎪⎧ 0 ⎫⎪
⎧⎪ R1 ⎫⎪ ⎧⎪δ R1 ⎫⎪ ⎧⎪ R ⎫⎪ ⎪⎧ 0 ⎪⎫ ⎨ ⎬+⎨ ⎬
3
⎨ ⎬+⎨ ⎬ Δv
⎨ ⎬+ ⎨ ⎩⎪V3 ⎭⎪ ⎪⎩ 3 ⎭⎪
2
⎬
⎩⎪V1 ⎭⎪ ⎩⎪δ V1 ⎭⎪ ⎪⎩V2 ⎭⎪ ⎩⎪ Δv2 ⎭⎪
#1 #2 #3 #4
⎪⎧ R1 ⎪⎫ ⎪⎧ R2 ⎪⎫ ⎪⎧ R3 ⎪⎫ ⎪⎧ R4 ⎪⎫
⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎪⎩V1 ⎪⎭ ⎪⎩V2 ⎪⎭ ⎩⎪V3 ⎭⎪ ⎩⎪V4 ⎭⎪
Figure 5.17 Schematic for Station-Keeping Algorithm

141
5.2.2 Station-Keeping Results
The station-keeping costs are obtained in Astrogator Connect using the algorithm
from the previous section. (See Figure 5.17.) Results for each orbit in Table 4.1 are
specified in Table 5.8 for one complete year. In general, station-keeping cost increases
with stability index. The lowest costs correspond to the 7- and 8-day L2 near-rectilinear
halo orbits while the highest costs are associated with the 14-day L1 vertical orbit. The
algorithm depicted in Figure 5.17 does not compute optimal cost, therefore these costs
are not optimized. However, experience suggests that these results are consistent with an
impulsive control scheme [33].
Table 5.8
Station-Keeping Results for One Year (~ 24 Revs)
Avg. Time
Libration Period Avg. 3σ No. of Between Avg. Δv Total Δv
Orbit Type
Point (days) Δv (cm/s) Maneuvers Maneuvers (m/s) (m/s)
(days)
Near-Rectilinear Halo L2 7.0 2.06 86 4.20 0.057 4.82
Halo L1 12.0 3.82 60 6.00 1.106 66.33
Halo L2 14.0 2.77 156 2.33 0.183 28.47
Vertical L1 14.0 3.13 68 5.25 2.527 171.82
Butterfly L2 14.0 9.78 78 4.67 0.409 31.86
Vertical L1 16.0 2.81 91 4.00 0.347 31.55
Vertical L2 16.0 2.75 60 6.00 1.472 88.32
142
6. RESULTS AND RECOMMENDATIONS
6.1 Summary of Results
Particular solutions in the CR3BP are available using analytical approximations in the
vicinity of the libration points. The solutions are expanded with five different corrections
schemes for numerically computing periodic orbits in the CR3BP. Strategies for
numerical computation of families of periodic solutions are offered. A method of
bisections establishes a means for locating bifurcations within families and mappings to
different families.
The numerical techniques establish a basis for the computation of planar, axial, and
vertical families emanating from five equilibrium points in the Earth-Moon system. In
addition, halo orbit families in the vicinity of the collinear points and a butterfly family in
the vicinity of the trans-lunar libration point are computed. The analyses are easily
modified for numerical computation of similar solutions in other systems of interest.
Orbits from the halo, vertical, and butterfly families in the vicinity of the Earth-Moon
libration points are selected for potential lunar coverage scenarios. Orbits with altitudes
between 50 km and 100,000 km are identified as feasible. Nine different orbits with
periods ranging from 7 to 16 days are studied in detail. Two phased spacecraft in a single
orbit or two spacecraft in a combination of two different orbits, with periods that are
commensurate, ensure adequate coverage of the lunar south pole. The orbit with the
desired period is determined in the CR3BP from a corrections process with a periodicity
constraint. Multiple revolutions are added and the orbit from the CR3BP is transitioned
to a full ephemeris model, including solar perturbations.
A complete coverage analysis is available using advanced commercial software. A
ground station is positioned at the Shackleton Crater near the lunar south pole and a
transmitting station is located at the White Sands Test Facility. Line-of-sight percent
access times with the both stations are computed for five potential coverage scenarios. A
preliminary station-keeping analysis is available by targeting points along the baseline
143
orbit and perturbing the spacecraft in a direction consistent with the unstable subspace.
Complete coverage of the lunar south pole is achieved with two spacecraft in
combinations of Earth-Moon libration point orbits.
This work contributes to the continuing evolution of a general baseline strategy for
mission design:
1. Computing periodic solutions in the CR3BP
2. Identifying feasible solutions for mission design requirements
3. Transitioning to a full ephemeris model
4. Verifying mission design requirements using advanced software
The general method is applicable to a broad range of mission applications for libration
point orbit design.
6.2 Recommendations for Future Work
Analysis continues to obtain transfers from the Earth to such orbits. An example of
the stable and unstable manifolds corresponding to a vertical orbit appears in Figure 6.1.
In general, the stability index offers information regarding the relative transfer costs to
reach these orbits. Moreover, current designs for lunar south pole coverage include
“frozen,” polar orbits computed using a two-body model, and with third body effects
modeled as gravitational perturbations [37-38]. Establishing a connection between these
orbits and the orbits computed in this work may identify even more favorable orbits for
lunar south pole coverage. In addition, libration point orbits from other families may also
render coverage options for the lunar south pole. A further investigation of periodic
orbits and the connections between families in the CR3BP may lead to the discovery new
families of orbits with smaller lunar altitudes to aid communications with the lunar
ground station. Furthermore, a more complete mapping of the solution space in the
CR3BP uncovers even more solutions. The nature of the solutions may aid in return to
the Moon and ultimately facilitate exploration to Mars.
144
Earth
Moon
Figure 6.1 Earth-Moon L1 Vertical Orbit with Associated

Unstable (Red) and Stable (Blue) Manifolds
145
LIST OF REFERENCES
146
LIST OF REFERENCES
[1] R. Farquhar, “The Flight of ISEE-3/ICE: Origins, Mission History, and a

Legacy.” Paper No. AIAA 1998-4464, AIAA/AAS Astrodynamics Specialist
Conference and Exhibit, Boston, Massachusetts, August 10-12, 1998.
[2] M. Lo, B. Williams, W. Bollman, D. Han, Y. Hahn, J. Bell, E. Hirst, R. Corwin,

P. Hong, K. Howell, B. Barden, and R. Wilson, “Genesis Mission Design.” Paper
No. AIAA 1998-4468, AIAA/AAS Astrodynamics Specialist Conference and
Exhibit, Boston, Massachusetts, August 10-12, August 1998.
[3] “The Vision for Space Exploration,” National Aeronautics and Space
Administration Publication, NP-2004-01-334-HQ, February 2004.
[4] H. Poincaré, Les Méthodes Nouvelles de la Mécanique Céleste. Vol. 1, 1892; New
Methods of Celestial Mechanics (English Translation). Vol. 13, History of
Modern Physics and Astronomy, American Institute of Physics, 1993.
[5] F. Moulton (in collaboration with D. Buchanan, T. Buck, F. Griffin, W. Longley,

and W. MacMillan), “Periodic Orbits.” Carnegie Institute of Washington,
Washington D.C., 1920.
[6] M. Hénon, “Exploration Numérique du Problème Restreint. II. Masses égales,

stabilité des orbites périodiques.” Annales d’Astrophysique, Vol. 28, February
1965, pp. 992-1007.
[7] E. Goodrich, “Numerical Determination of Short-Period Trojan Orbits in the

Restricted Three-Body Problem.” The Astronomical Journal, Vol. 71, No. 2,
March 1966, pp. 88-93.
[8] T. Bray and L. Goudas, “Doubly Symmetric Orbits about the Collinear
Lagrangian Points.” The Astronomical Journal, Vol. 72, No. 2, March 1967, pp.
202-213.
[9] A. Deprit, J. Henrard, J. Palmore, and J. Price, “The Trojan Manifold in the
System Earth-Moon.” Royal Astronomical Society, Monthly Notices, Vol. 132,
April 3, 1967, pp. 311-335.
[10] V. Szebehely, Theory of Orbits: The Restricted Problem of Three Bodies. New
York: Academic Press, 1967.
147
[11] M. Hénon, “Vertical Stability of Periodic Orbits in the Restricted Problem.”

Celestial Mechanics, Vol. 8, 1973, pp. 269-272.
[12] R. Farquhar and A. Kamel, “Quasi-Periodic Orbits About the Translunar

Libration Point.” Celestial Mechanics, Vol. 7, 1973, pp. 458-473.
[13] D. Richardson and N. Cary, “A Uniformly Valid Solution for Motion about the
Interior Libration Point of the Perturbed Elliptic-Restricted Problem.” Paper No.
AAS 75-021, AAIA/AAS Astrodynamics Conference, Nassau, Bahamas, July 28-
30, 1975.
[14] V. Markellos and A. Halioulias, “Numerical Determination of Asymmetric

Periodic Solutions.” Astrophysics and Space Science, Vol. 46, 1977, pp. 183-193.
[15] V. Markellos, “Asymmetric periodic orbits in three dimensions.” Royal

Astronomical Society, Monthly Notices, Vol. 184, 1978, pp. 273-281.
[16] C. Zagouras and P. Kazantzis, “Three-dimensional periodic oscillations

generating from plane periodic ones around the collinear Lagrangian points.”
Astrophysics and Space Science, Vol. 61, No. 2, April 1979, pp. 389-409.
[17] J. Breakwell and J. Brown, “The ‘Halo’ Family of 3-Dimensional Periodic Orbits
in the Earth-Moon Restricted 3-Body Problem.” Celestial Mechanics, Vol. 20,
1979, pp. 389-404.
[18] I. Robin and V. Markellos, “Numerical Determination of Three-Dimensional

Periodic Orbits Generated From Vertical Self-Resonant Satellite Orbits.”
Celestial Mechanics, Vol. 21, 1980, pp. 395-434.
[19] K. Howell and J. Breakwell, “Almost Rectilinear Halo Orbits.” Celestial

Mechanics, Vol. 32, No. 1, January 1984, pp. 29-52.
[20] K. Howell, “Three-Dimensional, Periodic, ‘Halo’ Orbits.” Celestial Mechanics,

Vol. 32, No. 1, January 1984, pp. 53-71.
[21] C. Zagouras, “Three-Dimensional Periodic Orbits about the Triangular

Equilibrium Points of the Restricted Problem of Three Bodies.” Celestial
Mechanics, Vol. 37, June 6, 1984, pp. 28-46.
[22] K. Howell and H. Pernicka, “Numerical Determination of Lissajous Trajectories

in the Restricted Three-Body Problem.” Celestial Mechanics, Vol. 41, 1988, pp.
107-124.
148
[23] S. Hughes, D. Cooley, and J. Guzmán, “A Direct Method for Fuel Optimal
Maneuvers of Distributed Spacecraft in Multiple Flight Regimes.” Paper No.
AAS 05-158, Space Flight Mechanics Meeting, Copper Mountain, Colorado,
January 23-27, 2005.
[24] B. Marchand and K. Howell, “Aspherical Formations Near the Libration Points in
the Sun-Earth/Moon Ephemeris System.” Paper No. AAS 04-157, 14th
AAS/AIAA Space Flight Mechanics Conference, Maui, Hawaii, February 8-12,
2004.
[25] B. Marchand, K. Howell, and R. Wilson, “A Trajectory Design Strategy in the

Multi-Body Regime Including Constraints.” (in preparation)
[26] K. Papadakis and C. Zagouras, “Bifurcation Points and Intersections of Families

of Periodic Orbits in the Three-Dimensional Restricted Three-Body Problem.”
Astrophysics and Space Science, Vol. 199, No. 2, January 1992, pp. 241-256.
[27] K. Howell and E. Campbell, “Three-Dimensional Periodic Solutions That

Bifurcate From Halo Families in the Circular Restricted Three-Body Problem.”
AAS/AIAA Paper No. 1999-161, AAS/AIAA Space Flight Mechanics Meeting,
Breckenridge, Colorado, February 7-10, 1999.
[28] D. Dichmann, E. Doedel, and R. Paffenroth, “The Computation of Periodic

Solutions of the 3-Body Problem Using the Numerical Continuation Sofware
AUTO.” Libration Point Orbits and Applications, G. Gómez, M. Lo, J.
Masdemont (Editors), Hong Kong, World Scientific, 2003.
[29] E. Doedel, R. Paffenroth, H. Keller, D. Dichmann, J. Galán-Vioque, and A.

Vanderbauwhede, “Computation of Periodic Solutions of Conservative Systems
with Application to the 3-Body Problem.” International Journal of Bifurcation
and Chaos, Vol. 13, No. 6, 2003, pp. 1353-1381.
[30] R. Farquhar, “The Utilization of Halo Orbits in Advanced Lunar Operations.”

NASA TND-365, Goddard Spaceflight Center, Greenbelt, Maryland, 1971.
[31] C. Simó, G. Gómez, J. Libre, and R. Martínez, “Station Keeping of a

Quasiperiodic Halo Orbit Using Invariant Manifolds.” Proceedings of the Second
International Symposium on Spaceflight Dynamics, Darmstadt, Germany, October
1986, pp. 65-70.
[32] C. Simó, G. Gómez, J. Libre, R. Martínez, and J. Rodríguez, “On the Optimal
Station Keeping Control of Halo Orbits.” Acta Astonautica, Vol. 15, No. 6/7,
1987, pp. 391-197.
149
[33] K. Howell and H. Pernicka, “Stationkeeping Method for Libration Point

Trajectories.” Journal of Guidance and Control, Vol. 16, No. 1, 1993, pp. 151–
159.
[34] K. Howell and T. Keeter, “Station-Keeping Strategies for Libration Point Orbits:
Target Point and Floquet Mode Approaches,” Proceeding of the AAS/AIAA
Spaceflight Mechanics Meeting 1995, Advances in the Astronautical Sciences,
Vol. 89, R. Proulx, J. Liu, P. Seidelmann, and S. Alfano (editors), 1995, pp. 1377-
1396.
[35] G. Gómez, K. Howell, J. Masdemont, and C. Simó, “Station-Keeping Strategies

for Translunar Libration Point Orbits.” Paper No. AAS 98-168, AAS/AIAA
Spaceflight Mechanics Meeting, Monterey, California, February 9-11, 1998.
[36] D. Scheeres, D. Han, and Y. Hou, “The Influence of Unstable Manifolds on Orbit
Uncertainty.” Journal of Guidance, Control, and Dynamics, Vol. 24, No. 3. May-
June 2001, pp. 573-585.
[37] C. Renault and D. Scheeres, “Statistical Analysis of Control Maneuvers in

Unstable Orbital Environments.” Journal of Guidance, Control, and Dynamics,
Vol. 26, No. 5, September-October 2003, pp.758-769.
[38] T. Ely, “Stable Constellations of Frozen Elliptical Inclined Orbits.” Journal of the
Astronautical Sciences, Vol. 53, No. 3, July-September 2005.
[39] T. Ely and E. Lieb, “Constellations of Elliptical Inclined Lunar Orbits Providing
Polar and Global Coverage.” Paper No. AAS 05-158, AAS/AIAA Spaceflight
Mechanics Meeting, South Lake Tahoe, California, August 7-11, 2005.
[40] NASA Space Science Enterprise, Strategic Plan.

http://spacescience.nasa.gov/admin/pubs/strategy/2000/, Sections 2-3, November
2000.
[41] K. Howell and J. Anderson, User’s Guide: Purdue Software GENERATOR, July
2001.
[42] D. Folta and F. Vaughn, “A Survey of Earth-Moon Libration Orbits: Station-

Keeping Strategies and Intra-Orbit Transfers.” Paper No. AIAA 04-4741,
Proceedings of AAS/AIAA Astrodynamics Specialist Conference, Providence,
Rhode Island, 2004.
[43] L. Perko, Differential Equations and Dynamical Systems. Springer-Verlag, 3rd ed.,
2001.
150
[44] W. Gautschi, Numerical Analysis: An Introduction. Birkhäuser, Boston, 1997, pp.

217-220.
[45] R. Farquhar, “Lunar Communications with Libration-Point Satellites.” Journal of

Spacecraft and Rockets, Vol. 4, 1967, pp. 1383-1384.
[46] G. Vanderplaats, Numerical Optimization Techniques for Engineering Design.

Vanderplaats Research & Development, Inc., Colorado Springs, 2001, pp. 250-
256.
[47] Information regarding current interest in lunar south pole sites and activities
conveyed by José Guzmán, Applied Physics Laboratory, private communication.
151
VITA
152
VITA
Daniel J. Grebow received his undergraduate degree in Aeronautical and Astronautical

Engineering from Purdue University in May of 2004. He spent the summer of 2005 at
NASA Goddard Spaceflight Center. At NASA, under the guidance of David Folta, he
participated in a research study concerning lunar coverage using libration point orbits. In
collaboration with Martin Ozimek, the results of this work were presented to the Goddard
Flight Dynamics branch. He also recently presented the results of the investigation in a
paper at the 16th AIAA/AAS Space Flight Mechanics Conference in January, 2006.
Additional results have been obtained and a paper, entitled “Multi-Body Orbit
Architectures for Lunar South Pole Coverage,” is currently in preparation to be submitted
for publication in an archival journal. Mr. Grebow anticipates working at NASA
Goddard again during summer 2006 and continuing his studies at Purdue University for a
Doctoral degree under Dr. Kathleen Howell.

Generating Periodic Orbits in The Circular Restricted Three Body Problem With Applications To Lunar South Pole Coverage

Uploaded by

Copyright:

Available Formats

Generating Periodic Orbits in The Circular Restricted Three Body Problem With Applications To Lunar South Pole Coverage

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Generating Periodic Orbits in The Circular Restricted Three Body Problem With Applications To Lunar South Pole Coverage

Uploaded by

Copyright:

Available Formats

GENERATING PERIODIC ORBITS IN THE CIRCULAR RESTRICTED THREE-

BODY PROBLEM WITH APPLICATIONS TO LUNAR SOUTH POLE COVERAGE

In Partial Fulfillment of the

Master of Science in Aeronautics and Astronautics

Socrates pursued wisdom, knowledge, and understanding. However, in Plato’s Meno,

3.16 L4 Vertical Family Initial Conditions with y0 = 0.42545 ............................... 89

Current limitations in a number of technical areas, including mission design

1.1 Problem Definition

1.1.1 The Circular Restricted Three-Body Problem

1.1.2 Lunar South Pole Coverage

Current interest in autonomous and manned exploration studies is focused primarily

1.2 Previous Contributors

1.2.1 Quasi-Periodic and Periodic Solutions

1.2.2 Application to Lunar South Pole Coverage and Station-Keeping Strategies

1.3 Current Work

This study is organized in the following manner:

line-of-sight access between a spacecraft constellation and a possible transmitting site

The differential equations describing the motion of a body of infinitesimal mass

2.1 The Circular Restricted Three-Body Problem

Figure 2.1 Formulation for the Restricted Three-Body Problem

Equation (2.1) can be generalized by introducing characteristic quantities. Let the

The dots in equation (2.7) represent differentiation with respect to τ. Furthermore,

r13 = r3 − r1 = ( x + μ ) xˆ + yyˆ + zzˆ, (2.8)

r23 = r3 − r2 = ( x − (1 − μ ) ) xˆ + yyˆ + zzˆ. (2.9)

Finally, the vector equation (2.5) can be rewritten in scalar form as

The motion of P3 relative to B in the rotating frame is determined by numerically

2.2 The Equilibrium Points

Five equilibrium points satisfy equations (2.13)-(2.15). From equation (2.15), it is

(2.14). Using equations (2.8)-(2.9) to solve r13 eq

L1 − μ < xeq < 1 − μ −1 1

Figure 2.2 Equilibrium or Libration Points (Magenta) in the CR3BP

2.3 First-Order Analytical Approximations for Variations Relative to the

(2.20)-(2.22) are simplified to

solution for ζ can be written in the following form,

Expressions for Ai in terms of ξ0 , ξ 0 , η0 , η0 , λi , and U xx eq

of equation (2.34)-(2.37). In general, as evidenced by equation (2.30), the system

2.3.1 Quasi-Periodic, First-Order Approximations Near the Collinear Libration

Equations (2.38)-(2.40) can be further simplified to produce approximate conditions

amplitude Aη in the η direction; then, an amplitude Aζ is specified in the ζ direction.

If it is desired to specify Aξ and Aζ , then equations (2.38)-(2.40) can be simplified to the

corresponding to each revolution is approximately 11.7471 days.

−0.05 0 0.05 −0.05 0 0.05

Figure 2.3 First-Order Approximation to a Lissajous Trajectory

2.3.2 Periodic, First-Order Approximations Near the Collinear Libration Points

Figure 2.4 First-Order Approximation to a Lyapunov Orbit

2.4 Fundamental Variational Relationships

2.4.1 The State-Transition Matrix

q be defined as q = { x z} . The first-order variational equations are

derived and result in the vector differential equation,

constant for an arbitrary trajectory as a reference solution. However, if the reference is a

A (τ ) is summarized in terms of four 3 × 3 submatrices,

The well-known form of the solution to equations (2.56) is

represent the variations at τ + τ 0 . Simultaneous numerical integration of equation (2.56)

Φ (τ 0 ,τ 0 ) = I 6×6 . The state-transition matrix supplies valuable information regarding both

2.4.2 Contemporaneous and Non-Contemporaneous Variations in the State

where Δτ = τ i +1 − τ i . Then, the non-contemporaneous state variation corresponding to the

i th + 1 point, i.e., δ qi +1 , must satisfy

δ qi +1 − qi +1δτ i +1 = Φ (τ i +1 ,τ i ) (δ qi − qiδτ i ) . (2.63)

Figure 2.5 Representation of Contemporaneous and Non-Contemporaneous