Mathematical

Problems in
Engineering
Theory, Methods, and Applications
Editor-in-Chief: Jose Manoel Balthazar

Special Issue
Space Dynamics

Guest Editors: Antonio F. Bertachini A. Prado, Maria Cecilia Zanardi,

Tadashi Yokoyama, and Silvia Maria Giuliatti Winter

Hindawi Publishing Corporation

http://www.hindawi.com Volume 2009
Space Dynamics
Mathematical Problems in Engineering

Space Dynamics
Guest Editors: Antonio F. Bertachini A. Prado, Maria Cecilia Zanardi, Tadashi
Yokoyama, and Silvia Maria Giuliatti Winter
Copyright q 2009 Hindawi Publishing Corporation. All rights reserved.

This is an issue published in volume 2009 of Mathematical Problems in Engineering. All articles are open access
articles distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Editor-in-Chief
Jose Manoel Balthazar, Universidade Estadual Paulista, Brazil

Associate Editors
John Burns, USA Shi Jian Liao, China J. Rodellar, Spain
Carlo Cattani, Italy Panos Liatsis, UK Ilmar Ferreira Santos, Denmark
Miguel Cerrolaza, Venezuela Bin Liu, Australia Nickolas S. Sapidis, Greece
David Chelidze, USA Angelo Luongo, Italy Massimo Scalia, Italy
Jyh Horng Chou, Taiwan Mehrdad Massoudi, USA Alexander P. Seyranian, Russia
Horst Ecker, Austria Yuri V. Mikhlin, Ukraine Christos H. Skiadas, Greece
Oleg V. Gendelman, Israel G. V. Milovanovic, Serbia Alois Steindl, Austria
Paulo Batista Goncalves, Brazil Ben T. Nohara, Japan Jitao Sun, China
Oded Gottlieb, Israel Ekaterina Pavlovskaia, UK Cristian Toma, Romania
K. Stevanovic Hedrih, Serbia Francesco Pellicano, Italy Irina N. Trendalova, UK
Wei-Chiang Hong, Taiwan F. Lobo Pereira, Portugal Kuppalapalle Vajravelu, USA
J. Horacek, Czech Republic Sergio Preidikman, USA Victoria Vampa, Argentina
J. Jiang, China Dane Quinn, USA Jerzy Warminski, Poland
Joaquim J. Judice, Portugal Saad A. Ragab, USA Mohammad I. Younis, USA
Tamas Kalmar-Nagy, USA K. R. Rajagopal, USA
Ming Li, China Giuseppe Rega, Italy
Contents
Space Dynamics, Antonio F. Bertachini A. Prado, Maria Cecilia Zanardi,
Silvia Maria Giuliatti Winter, and Tadashi Yokoyama
Volume 2009, Article ID 732758, 7 pages

Optimal On-O Attitude Control for the Brazilian Multimission Platform Satellite,
Gilberto Arantes Jr., Luiz S. Martins-Filho, and Adrielle C. Santana
Volume 2009, Article ID 750945, 17 pages

Highly Ecient Sigma Point Filter for Spacecraft Attitude and Rate Estimation,
Chunshi Fan and Zheng You
Volume 2009, Article ID 507370, 23 pages

Spin-Stabilized Spacecrafts: Analytical Attitude Propagation Using Magnetic Torques,

Roberta Veloso Garcia, Maria Ceclia F. P. S. Zanardi, and Helio Koiti Kuga
Volume 2009, Article ID 753653, 18 pages

Using of H-Innity Control Method in Attitude Control System of Rigid-Flexible Satellite,

Ximena Celia Mendez Cubillos and Luiz Carlos Gadelha de Souza
Volume 2009, Article ID 173145, 9 pages

Hill Problem Analytical Theory to the Order Four: Application to the Computation of Frozen
Orbits around Planetary Satellites, Martin Lara and Jesus F. Palacian
Volume 2009, Article ID 753653, 18 pages

Collision and Stable Regions around Bodies with Simple Geometric Shape, A. A. Silva,
O. C. Winter, and A. F. B. A. Prado
Volume 2009, Article ID 396267, 14 pages

Dynamical Aspects of an Equilateral Restricted Four-Body Problem, Martha Alvarez-Ramrez

and Claudio Vidal
Volume 2009, Article ID 181360, 23 pages

Nonsphericity of the Moon and Near Sun-Synchronous Polar Lunar Orbits,

Jean Paulo dos Santos Carvalho, Rodolpho Vilhena de Moraes, and Antonio Fernando Bertachini
de Almeida Prado
Volume 2009, Article ID 740460, 24 pages

GPS Satellites Orbits: Resonance, Luiz Danilo Damasceno Ferreira and

Rodolpho Vilhena de Moraes
Volume 2009, Article ID 347835, 12 pages

Some Initial Conditions for Disposed Satellites of the Systems GPS and Galileo
Constellations, Diogo Merguizo Sanchez, Tadashi Yokoyama, Pedro Ivo de Oliveira Brasil,
and Ricardo Reis Cordeiro
Volume 2009, Article ID 510759, 22 pages
Quality of TEC Estimated with Mod Ion Using GPS and GLONASS Data,
Paulo de Oliveira Camargo
Volume 2009, Article ID 794578, 16 pages

The Impact on Geographic Location Accuracy due to Dierent Satellite Orbit Ephemerides,
Claudia C. Celestino, Cristina T. Sousa, Wilson Yamaguti, and Helio Koiti Kuga
Volume 2009, Article ID 856138, 9 pages

Simulations under Ideal and Nonideal Conditions for Characterization of a Passive Doppler
Geographical Location System Using Extension of Data Reception Network,
Cristina Tobler de Sousa, Rodolpho Vilhena de Moraes, and Helio Koiti Kuga
Volume 2009, Article ID 147326, 19 pages

A Discussion Related to Orbit Determination Using Nonlinear Sigma Point Kalman Filter,
Paula Cristiane Pinto Mesquita Pardal, Helio Koiti Kuga, and Rodolpho Vilhena de Moraes
Volume 2009, Article ID 140963, 12 pages

Orbital Dynamics of a Simple Solar Photon Thruster, Anna D. Guerman, Georgi V. Smirnov,
and Maria Cecilia Pereira
Volume 2009, Article ID 537256, 11 pages

Alternative Transfers to the NEOs 99942 Apophis, 1994 WR12, and 2007 UW1 via Derived
Trajectories from Periodic Orbits of Family G, C. F. de Melo, E. E. N. Macau, and O. C. Winter
Volume 2009, Article ID 303604, 12 pages

Controlling the Eccentricity of Polar Lunar Orbits with Low-Thrust Propulsion,

O. C. Winter, D. C. Mourao, C. F. Melo, E. N. Macau, J. L. Ferreira, and J. P. S. Carvalho
Volume 2009, Article ID 159287, 10 pages

Internal Loading Distribution in Statically Loaded Ball Bearings Subjected to an Eccentric

Thrust Load, Mario Cesar Ricci
Volume 2009, Article ID 471804, 36 pages

The Determination of the Velocities after Impact for the Constrained Bar Problem,
Andre Fenili, Luiz Carlos Gadelha de Souza, and Bernd Schafer
Volume 2009, Article ID 384071, 16 pages

Gravitational Capture of Asteroids by Gas Drag, E. Vieira Neto and O. C. Winter

Volume 2009, Article ID 897570, 11 pages

Atmosphentry Dynamics of Conic Objects, J. P. Saldia, A. Cimino, W. Schulz, S. Elaskar,

and A. Costa
Volume 2009, Article ID 859678, 14 pages
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 732758, 7 pages
doi:10.1155/2009/732758

Editorial
Space Dynamics

Antonio F. Bertachini A. Prado,1 Maria Cecilia Zanardi,2

Tadashi Yokoyama,3 and Silvia Maria Giuliatti Winter2
1
INPE-DMC, Brazil
2
FEG-UNESP, Guaratingueta, Brazil
3
UNESP, Campus de Rio Claro, Brazil

Correspondence should be addressed to Antonio F. Bertachini A. Prado, [email protected]

Received 31 December 2009; Accepted 31 December 2009

Copyright q 2009 Antonio F. Bertachini A. Prado et al. This is an open access article distributed
under the Creative Commons Attribution License, which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work is properly cited.

The space activity in the world is one of the most important achievements of mankind.
It makes possible live communications, exploration of Earth resources, weather forecast,
accurate positioning and several other tasks that are part of our lives today.
The space dynamics plays a very important rule in these developments, since its study
allows us to plan how to launch and control a space vehicle in order to obtain the results we
need.
This eld considers the study of Celestial Mechanics and Control applied to spacecraft
and natural objects. The main tasks are to determine the orbit and the attitude of the
spacecraft based in some observations, to obtain its position and attitude in space in a given
time from some initial conditions, to nd the best way to change their orbits and attitude, to
analyze how to use the information of the satellites to nd the position and the velocity of a
given point e.g., a personal receptor, a satellite or a car, etc.
This eld of study comes from Astronomy. The main contributors from the past have
important names like Johannes Kepler 15711630 and Isaac Newton 16421727. Based
on the observations of the motion of the planets realized by Tycho Brahe 15461601, Kepler
formulated the three basic laws, which govern the motion of the planets around the Sun. From
these laws, Newton formulated the universal Law of Gravitation. According to this Law,
mass attracts mass in a ratio that is proportional to the product of the two masses involved
and inversely proportional to the square of the distance between them. Those laws are the
scientic bases of the space exploration age that ocially begin with the launch of the satellite
Sputnik in 1957 by the former Soviet Union. Since then, a strong battle between the United
States of America USA and the Soviet Union took place leading to many achievements in
Space. One of the most important results was the landing of the man on the Moon, achieved
2 Mathematical Problems in Engineering

by the USA in 1969. From this point, several dierent applications of the space research were
developed, changing for better the human life on Earth.
In that scope, this special issue of Mathematical Problems in Engineering is focused on
the recent advances in space dynamics techniques. It has a total of 21 papers that are briey
described below.
Four of them are concerned with the attitude motion, control and determination.
Optimal On-O Attitude Control for the Brazilian Multimission Platform Satellite by G.
Arantes Jr. et al. is the rst one. This work deals with the analysis and design of the reaction
thruster attitude control for the Brazilian Multi-Mission Platform satellite. The aim of this
work is to provide smoother control for improved pointing requirements with less thruster
activation or propellant consumption. The fuel is a deciding factor of the lifetime of the
spacecraft and reduced propellant consumption is highly required, specially, regarding a
multi-mission spacecraft wherein dierent payloads are being considering. The three-axis
attitude control is considered and it is activated in pulse mode. Consequently a modulation
of the torque command is compelling in order to avoid high non-linear control action.
The paper considers the Pulse-Width Pulse-Frequency PWPF modulator, composed of
a Schmidt trigger, a rst order lter, and a feedback loop. This modulator holds several
advantages over classical bang-bang controllers such as close to linear operations, high
accuracy, and reduced propellant consumption. The Linear Gaussian Quadratic LQG
technique is used to synthesize the control law during stabilization mode and the modulator
is used to modulate the continuous control signal to discrete one. The results of the
numerical simulations show that the obtained on-o thruster reaction attitude control system,
based on the LQG/PWPF modulation, is optimal with respect to the minimization of
the quadratic cost function of the states and control signals and propellant consumption.
The paper presents a set of optimal parameter for the PWPF modulator by considering
static and dynamics analysis. The obtained results demonstrate the feasibility of combining
LQG/PWPF modulator in a unique controller for on-o thruster reaction attitude control
system. Stability remains by adding the PWPF modulator and reasonable accuracy in attitude
is achieved. Practical aspects are included in this study as ltering and presence of external
impulsive perturbations. The advantages of less spent propellant shall contribute to the
Brazilian Multi-Mission Platform project, specially, a satellite conceived to be used on a large
number and dierent types of missions, in the context of an ever-advancing Brazilian space
program.
The second paper on this subject is Highly Ecient Sigma Point Filter for Spacecraft
Attitude and Rate Estimation by C. Fan and Z. You. In this paper, for spacecraft attitude
determination problem, the multiplicative extended Kalman lter MEKF and other similar
algorithms, have been good solutions for most nominal space missions. However, nowadays,
due to their overload computational complexity, they are prohibitive for actual on board
implementation. In this paper, the authors present a new and quite competitive algorithm,
with signicant lower computational complexity even when compared to the reduced sigma
point algorithms. The precision is the same as the traditional unscented Kalman lters. In
terms of eciency, the proposed algorithm rivals MEKF, even in severe situations.
The next one is Spin-Stabilized Spacecraft: Analytical Attitude Propagation Using
Magnetic Torques by R. V. Garcia et al.. This paper considers the problem in obtaining the
attitude of a satellite in a given time based on information from a previous time. It analyzes
the rotational motion of a spin stabilized Earth articial satellite. It makes derivation of
an analytical attitude prediction. Particular attention is given to torques, which come from
residual magnetic and eddy currents perturbations, as well as their inuences on the satellite
Mathematical Problems in Engineering 3

angular velocity and space orientation. A spherical coordinated system, xed in the satellite,
is used to locate the spin axis of the satellite in relation to the terrestrial equatorial system.
The last paper of this topic is Using of H-Innity Control Method in Attitude Control
System of Rigid-Flexible Satellite by X. C. M. Cubillos and L. C. G. Souza. This paper
considers the attitude control systems of satellites with rigid and exible components. In the
current space missions, this problem is demanding a better performance, which implies in
the development of several methods to approach this problem. For this reason, the methods
available today need more investigation in order to know their capability and limitations.
Therefore, in this paper, the HInnity method is studied in terms of the performance of the
Attitude Control System of a Rigid-Flexible Satellite.
There were four papers studying the problem of nding space trajectories. The rst
one is Hill Problem Analytical Theory to the Order Four: Application to the Computation
of Frozen Orbits around Planetary Satellites by M. Lara and J. F. Palacian. In this paper,
applications to the computation of frozen orbits around planetary satellites are made. The
Hill problem, a simplied model of the restricted three-body problem, also gives a very
good approximation for the dynamics involving the motion of natural and articial satellites,
moons, asteroids and comets. Frozen orbits in the Hill problem are determined through
the double averaged problem. The developed method provides the explicit equations of the
transformation connecting averaged and non averaged models, making the computation of
the frozen orbits straightforward.
The second one covering this topic is Collision and Stable Regions around Bodies with
Simple Geometric Shape by A. A. Silva et al.. Collision and stable regions around bodies
with simple geometric shape are studied. The gravitational potential of two simple geometric
shapes, square and triangular plates, were obtained in order to study the orbital motion of a
particle around them. Collision and stable regions were also derived from the well known
Poincare surface of section. These results can be applied to a particle in orbit around an
irregular body, such as an asteroid or a comet.
The next paper is Dynamical Aspects of an Equilateral Restricted Four-Body Problem
by M. Alvarez-Ramrez and C. Vidal. It is an immediate extension of the classical restricted
three body problem ERFBP: a particle is under the attraction of three nonzero masses
m1 , m2 , m3  which move on circular orbits around their center of mass, xed at the origin of
the coordinate system in a such way that their conguration is always an equilateral triangle.
In particular, it is assumed m2  m3 . In a synodical system, a rst integral of the problem
is obtained. Using Hamiltonian formalism the authors dene Hills regions. Equilibrium
solutions are obtained for dierent cases and the number of them depends on the values
of the masses. The Lyapunov stability of these solutions is studied in the symmetrical case
assuming m1  m2  m3  . Under certain conditions and for very small , circular and
elliptic keplerian periodic solutions can be continued to ERFBP. For  1/2, Lyapunov
Central theorem can provide a one-parameter family of periodic orbits. Some numerical
applications are also shown.
The last one in this category is Nonsphericity of the Moon and Near Sun-Synchronous
Polar Lunar Orbits by J. P. S. Carvalho et al.. Here, the dynamics of a lunar articial satellite
perturbed by the nonuniform distribution of mass of the Moon taking into account the
oblateness J2  and the equatorial ellipticity sectorial term C22  is presented. A canonical
perturbation method based on Lie-Hori algorithm is used to obtain the second order
solutions. A study is performed for the critical inclination and the eect of the coupling terms
J2 and C22 are presented. A new second order formula is obtained for the critical inclination as
a function of the argument of the pericenter and of the longitude of the ascending node. In the
4 Mathematical Problems in Engineering

same way, for Lunar Sun-synchronous and Near-Polar Orbits, a new formula is obtained to
provide the value of the inclination. This formula depends on the semi-major axis, eccentricity
and the longitude of the ascending node. For Lunar low altitude satellites, the authors call
the attention for the importance of the additional harmonics J3 , J5 , and C31 , besides J2 and
C22 . In particular they mention that, for small inclinations, some contributions of the second
order terms can become as large as the rst order terms. Several numerical simulations are
presented to illustrate the time variation of the eccentricity and inclination.
After that, there are ve papers considering the problem of localization with
information obtained from space, in particular using GPS and/or GLONASS constellations.
The rst paper of this topic is GPS Satellites Orbits: Resonance by L. D. D. Ferreira and
R. V. Moraes. In this paper, the eects of the perturbations due to resonant geopotential
harmonics on the semi major axis of GPS satellites are analyzed. The results show that it is
possible to obtain secular perturbations of about 4m/day using numerical integration of the
Lagrange planetary equations and considering, in the disturbing potential, the main secular
resonant coecients. The paper also shows the amplitudes for the long period terms due
to the resonant coecients for some hypothetical satellites orbiting in the neighborhood of
the GPS satellites orbits. The results can be used to perform orbital maneuvers of the GPS
satellites to keep them in their nominal orbits.
The second paper is Some Initial Conditions for Disposed Satellites of the Systems GPS
and Galileo Constellations by D. M. Sanchez et al.. In this paper the stability of the disposed
objects of the GPS and Galileo systems can be aected by the increasing in their eccentricities
due to strong resonances. A search for initial conditions where the disposed objects remain
at least 250 years, without crossing the orbits of the operational satellites, was performed. As
a result, regions where the values of the eccentricity prevent possible risk of collisions have
been identied in the phase space. The results also show that the initial inclination of the
Moon plays an important role in searching these initial conditions.
Then, we have Quality of TEC Estimated with Mod Ion Using GPS and GLONASS
Data by P. O. Camargo. The largest source of error in positioning and navigation with the
Global Navigation Satellite System GNSS is the ionosphere, which depends on the Total
Electron Content TEC. The quality of the TEC was analyzed taking into account the ModIon
model developed in UNESP-Brazil the more appropriate model to be used in the South
America region.
After that, we have the paper The Impact on Geographic Location Accuracy due to
Dierent Satellite Orbit Ephemeredes by C. C. Celestino et al.. Here, it is assumed that there
are several satellites, hundreds of Data Collection Platforms DCPs deployed on ground
xed or mobile of a large country e.g. Brazil, and also some ground reception stations. It
considers the question of obtaining the geographic location of these DCPs. In this work, the
impact on the geographic location accuracy, when using orbit ephemeris obtained through
several sources, is assessed. First, by this evaluation is performed by computer simulation
of the Doppler data, corresponding to real existing satellite passes. Then, real Doppler data
are used to assess the performance of the location system. The results indicate that the use of
precise ephemeris can improve the performance of the calculations involved in this process
by reducing the location errors. This conclusion can then be extended to similar location
systems.
There is also the paper Simulations under Ideal and Non ideal Conditions for
Characterization of a Passive Doppler Geographical Location System Using Extension of
Data Reception Network by C. T. Sousa et al.. It presents a Data Reception Network DRN
software investigation to characterize the passive Doppler Geographical Location GEOLOC
Mathematical Problems in Engineering 5

software. The test scenario is composed by Brazilian Data Collection Satellite SCD2 and
the National Oceanic Atmospheric Administration satellite NOAA-17 passes, a single Data
Collecting Platform DPC and ve ground received stations. The Doppler measurements
data of a single satellite pass over a DCP, considering a network of ground reception stations,
is the rule of the DNR. The DNR uses an ordering selection method that merges the collected
Doppler shift measurements through the stations network in a single le. The pre-processed
and analyzed measurement encompasses the DCP signal transmission time and the Doppler
shifted signal frequency received on board of the satellite. Thus, the assembly to a single le of
the measurements collected, considering a given satellite pass, will contain more information
about the full Doppler eect behavior while decreasing the amount of measurement losses,
as a consequence, an extended visibility between the relay satellite and the reception stations.
The results and analyses were rstly obtained considering the ground stations separately, to
characterize their eects in the geographical location result. Six conditions were investigated:
ideal simulated conditions, random and bias errors in the Doppler measurements, errors in
the satellites ephemeris and errors in the time stamp. To investigate the DNR importance
to get more accurate locations, an analysis was performed considering the random errors
of 1 HZ in the Doppler measurements. The results show that the developed GEOLOC is
operating appropriately under the ideal conditions. The inclusion of biased errors degrades
the location results more than the random errors. The random errors are ltered out by the
least squares algorithm and they produce mean locations results that tend to zero error,
mainly for high sampling rate. The simulations results, considering biased errors, yield
errors that degrade the location for high and low sampling rates. The simulation results
for ephemeredes error shows that it is fundamental to minimize them, because the location
system cannot compensate these errors. The satellites ephemeredes errors are approximately
similar in magnitude to their resulting transmitter location errors. The simulations results,
using the DRN algorithm, show that to improve the locations results quality it would be
necessary to have more Reception Stations spread over the Brazilian territory, to obtain
additional amount of data. Then, on the other hand, it improves the geometrical coverage
between satellite and DCPs, and better recovers the full Doppler curves, yielding, as a
consequence, more valid and improved locations.
A similar problem, but concerned with the determination of an orbit of a satellite, is
considered in A Discussion Related to Orbit Determination Using Nonlinear Sigma Point
Kalman Filter by P. C. P. M. Pardal et al.. The goal of this work is to present a Kalman
lter based on the sigma point unscented transformation, aiming at real-time satellite
orbit determination using GPS measurements. Firstly, some underlying material is briey
presented before introducing SPKF sigma point Kalman lter and the basic idea of the
unscented transformation in which this lter is based. Through the paper, the formulation
about orbit determination via GPS, dynamic and observation models and unmodeled
acceleration estimation are presented. The SPKF is investigated in many dierent applications
and the results are discussed. The advantages indicate that SPKF can be used as an emerging
estimation algorithm to nonlinear system.
Orbital maneuvers for space vehicles are also considered in three papers, as in Orbital
Dynamics of a Simple Solar Photon Thruster by A. D. Guerman et al.. This paper studies
the orbital dynamics and control for two systems of solar propulsion, a at solar sail FSS
and a simple solar photon thruster SPT. The use of solar pressure to create propulsion can
minimize the spacecraft on-board energy consumption during the mission. Modern materials
and technologies made this propulsion scheme feasible, and many projects of solar sail are
now under development, making the solar sail dynamics the subject of numerous studies.
6 Mathematical Problems in Engineering

To perform the analysis presented in this paper, the equations of the sailcrafts motion are
deduced. Comparisons for the performance of two schemes of solar propulsion Simple Solar
Photon ThrusterSSPT and Dual Reection Solar Photon ThrusterDRSPT are shown
for two test time-optimal control problems of trajectory transfer Earth-Mars transfer and
Earth-Venus transfer. The mathematical model for the force acting on SSPT due to the solar
radiation pressure takes into account multiple reections of the light ux on the sailcraft
elements. In this analysis it is assumed that the solar radiation pressure follows inverse-
square variation law, the only gravitational eld is the one from the Sun central Newtonian,
and the sails are assumed to be ideal reectors. For a planar motion of an almost at sail
with negligible attitude control errors, the SSPT equations of motion are similar to those for
a DRSPT. The analysis showed a better performance of SPT in terms of response time and
the results are more pronounced for Earth-Venus transfer. It can be explained by the greater
values of the transversal component of the acceleration developed by SSPT compared to FSS.
Then, we have the paper Alternative Transfers to the NEOs 99942 Apophis, 1994
WR12, and 2007 UW1 via Derived Trajectories from Periodic Orbits of Family G by C. F.
Melo et al.. This paper explores the existence of a natural and direct link between low Earth
orbits and the lunar sphere of the inuence to get low-energy transfer trajectory to the three
Near Earth Objects through swing-bys with the Moon. The existence of this link is related to a
family of retrograde periodic orbits around the Lagrangian equilibrium point L1 predicted by
the circular, planar, restricted three-body Earth-Moon-particle problem. Such orbits belong to
the so-called Family G. The trajectories in this link are sensitive to small disturbances. This
enables them to be conveniently diverted, reducing the cost of a swing-by maneuver. These
maneuvers allow a gain in energy enough for the trajectories to escape from the Earth-Moon
system and to be stabilized in heliocentric orbits between Earth and Venus or Earth and Mars.
The result shows that the required increment of velocity by escape trajectories G is, in general,
fewer than the ones required by conventional transfer Patched-conic, between 2% up to
4%. Besides, the spacecraft velocities relative to the asteroids are also, in general, less than
that value obtained by the conventional methods. In terms of the transfer time, the results
show that in the Apophis and 1994WR12 it is possible to nd Closest Point Approaches. The
longest time always corresponds to the smallest relative velocity in Closest Point Approaches
for trajectories G. Therefore, the trajectories G can intercept the Near Earth Objects orbits and,
they can be a good alternative to design future missions destined to the Near Earth Objects.
After that, we have the paper Controlling the Eccentricity of Polar Lunar Orbits with
Low-Thrust Propulsion by O. C. Winter et al.. This paper approaches the problem that lunar
satellites in polar orbits suer a high increase on the eccentricity, due to the gravitational
perturbation of the Earth leading them to a collision with the Moon. Then, the control of the
orbital eccentricity leads to the control of the satellites lifetime. This paper introduces an
approach in order to keep the orbital eccentricity of the satellite at low values. The method
presented in the paper considers two systems: the 3-body problem, Moon-Earth-satellite and
the 4-body problem, Moon-Earth-Sun-satellite. A system considering a satellite with initial
eccentricity equals to 0.0001 and a range of initial altitudes, between 100 km and 5000 km,
is considered. An empirical expression for the length of time needed to occur the collision
with the Moon as a function of the initial altitude is derived. The results found for the 3-
body model were not signicantly dierent from those found for the 4-body model. After
that, using low thrust propulsion, it is introduced a correction of the eccentricity every time
it reaches the value 0.05.
Mechanical aspects of spacecrafts are considered in two papers. The rst one is Internal
Loading Distribution in Statically Loaded Ball Bearings Subjected to an Eccentric Thrust Load
Mathematical Problems in Engineering 7

by M. C. Ricci. In this paper an iterative method is introduced to calculate internal normal

ball loads in statically loaded single-row, angular-contact ball bearings, subjected to a known
thrust load which is applied to a variable distance from the geometric bearing center line.
Numerical examples are shown and compared with the literature. Fifty gures are presented
and the results are discussed.
The other paper is The Determination of the Velocities after Impact for the Constrained
Bar Problem by A. Fenili et al.. In this paper, a mathematical model for a constrained
manipulator is studied. Despite the fact that the model is simple, it has all the important
features of the system. A fully plastic impact is considered. Analytical expressions for the
velocities of the bodies involved after the collision are derived and used for the numerical
integrations of the equations of motion. The theory presented in the paper can be used to
problems where the robots have to follow some prescribed patterns or trajectories when in
contact with the environment.
One paper deals with the astronomical side of the space dynamics: Gravitational
Capture of Asteroids by Gas Drag by E. Vieira-Neto and O. C. Winter. The orbital
conguration of the irregular satellites, present in the giant planets system, suggests that
these bodies were asteroids in heliocentric orbits that have been captured by the planets.
Since this capture is temporary, it has been necessary a dissipative eect in order to turn this
temporary capture into a permanent one. This paper deals with this problem by analyzing the
eects of the gas drag, from the Solar Nebula, in the orbital evolution of these asteroids after
they have being captured by the planets. The results show that, although this dissipative
eect is important, it is not the only mechanism responsible for keeping the asteroids in a
permanent orbit about the planet.
Then, we also have one paper studying the motion of a spacecraft when traveling in
the atmospheric region of the space: Atmospheric Reentry Dynamics of Conic Objects by J. P.
Saldia et al.. In this paper, the accurate determination of the aerodynamics coecients is an
important issue in the calculation of the reentry trajectories of an object inside the terrestrial
atmosphere. The methodology to calculate these coecients and how to include them in a
code, in order to compute the reentry trajectories, is considered. As a result, a sample of
trajectories of conical objects for dierent initial ight conditions is presented.

Acknowledgments
The guest editors would like to thank all the authors, the reviews, the Editor of the journal,
and all the sta involved in the preparation of this issue for the opportunity to publish the
articles related to this important subject.

Antonio F. Bertachini A. Prado

Maria Cecilia Zanardi
Tadashi Yokoyama
Silvia Maria Giuliatti Winter
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 750945, 17 pages
doi:10.1155/2009/750945

Research Article
Optimal On-Off Attitude Control for the Brazilian
Multimission Platform Satellite

Gilberto Arantes Jr.,1 Luiz S. Martins-Filho,2

and Adrielle C. Santana3
1
Center for Applied Space Technology and Microgravity, (ZARM) Am Fallturm, University of Bremen,
28359 Bremen, Germany
2
Center of Engineering, Modelling and Applied Social Sciences, Federal University of ABC,
Rua Catequese 242, 09090-400 Santo Andre, SP, Brazil
3
Graduate Program on Information Engineering, Federal University of ABC, Rua Catequese 242,
09090-400 Santo Andre, SP, Brazil

Correspondence should be addressed to Luiz S. Martins-Filho, [email protected]

Received 24 June 2009; Accepted 13 August 2009

Recommended by Maria Zanardi

This work deals with the analysis and design of a reaction thruster attitude control for the
Brazilian Multimission platform satellite. The three-axis attitude control systems are activated in
pulse mode. Consequently, a modulation of the torque command is compelling in order to avoid
high nonlinear control action. This work considers the Pulse-Width Pulse-Frequency PWPF
modulator which is composed of a Schmidt trigger, a rst-order lter, and a feedback loop.
PWPF modulator holds several advantages over classical bang-bang controllers such as close
to linear operation, high accuracy, and reduced propellant consumption. The Linear Gaussian
Quadratic LQG technique is used to synthesize the control law during stabilization mode and
the modulator is used to modulate the continuous control signal to discrete one. Numerical
simulations are used to analyze the performance of the attitude control. The LQG/PWPF
approach achieves good stabilization-mode requirements as disturbances rejection and regulation
performance.

Copyright q 2009 Gilberto Arantes Jr et al. This is an open access article distributed under the
Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
One of the intentions of this work is to support the ongoing Brazilian multimission platform
MMP satellite project 1. The project takes into consideration a special platform satellite
which can supply multimissions capabilities supporting dierent payloads to lift up on the
platform. Applications including Earth observation, communication, scientic experiments,
and surveillance are few examples of suitable use of the MMP satellite. The MMP adopted
pulse or on-o reaction thruster for attitude maneuvers, therefore, modulating continuous
command signal to an on-o signal is a required task. Selecting the properly method to
2 Mathematical Problems in Engineering

modulate the control command signal is a key assignment. The aim of this work is to
provide smoother control for improved pointing requirements with less thruster activation
or propellant consumption. The fuel is a deciding factor of the lifetime of the spacecraft
and reduced propellant consumption is highly required, specially, regarding a multimission
spacecraft wherein dierent payloads are being considered.
In this paper a pulse-width pulse-frequency PWPF modulator is considered as a
feasible option for the MMP reaction thruster modulation due to advantages over other types
of pulse modulators as bang-bang controllers which has excessive thruster actuation 2, 3.
The PWPF modulator translates the continuous commanded control/torque signal to an on-
o signal. Its behavior is a quasilinear mode which is possible by modulating the width of
the activated reaction pulse proportionally to the level of the torque command input pulse-
width and also the distance between the pulses pulse-frequency. A PWPF modulator
is composed of a Schmidt trigger, a lag network lter, and a feedback loop. The PWPF
design requires iterative tuning of lag lter and Schmidt trigger. The optimal parameters
achievement is based on the static test signals and dynamic feedback signals simulation
results. The optimality is in respect to either the number of rings or spent fuel. The work in
35 provides good guidelines for the PWPF tuning task.
The PWPF is synthesized with a Linear Quadratic Gaussian LQG controller which
is designed for the MIMO attitude system. The LQG controller, refered to as H2 , allows a
tradeo between regulation performance and control eort. In order to reduce the control
eort or fuel consumption, an iteratively searching of the trade-o can be carried out.
Nevertheless the controller has to attempt all the involved requirements and specications.
A previous study of the LQG approach applied to the MMP satellite is presented in 6. The
reaction attitude control system is applied to the stabilization mode of the MMP. The paper
is divided into 5 sections. Section 2 presents the nonlinear model of the satellite, assumed
a rigid body, its linearization around the operation point, and the developed virtual reality
model of the satellite for visualization purposes. Section 3 presents a brief description of the
PWPF modulator and design of the LQG controller, which includes the description of the
LQG controller and provides the tuning parameters range for the PWPF modulator. Section 4
presents the numerical simulation for the reaction thruster attitude control system during
the stabilization mode. Regulation, ltering, and disturbance rejection are investigated and
discussed. Conclusions are presented in Section 5 based on the obtained results.

2. Problem Formulation
In this section we describe the mathematical model of the attitude motion, including
kinematics, dynamics, and the linerization of the satellite model around LHLV reference
frame. Based on that linear model the LQG controller is designed for the stabilization mode.

2.1. Satellite Attitude Model

The attitude of the satellite will be dened in this work by the orientation of the body
frame x, y, z coincident with the three principal axes of inertia with respect to the orbital
reference frame xr , yr , zr , also known as Local-Vertical-Local-Horizontal LVLH 7. The
origin of the orbit reference frame moves with the center of mass of the satellite in orbit.
The zr axis points toward the center of mass of the Earth, xr axis is in the plane of the orbit,
perpendicular to zr , in the direction of the velocity of the spacecraft. The yr axis is normal to
Mathematical Problems in Engineering 3

Figure 1: LVLH axis representation.

the local plane of the orbit, and completes a three-axis right-hand orthogonal system. Figure 1
illustrated the LVLH reference frame.
The attitude is represented by the direction cosine matrix R between body frame and
reference frame. During the stabilization mode only small angular variations are considered,
in this case the Euler angles parametrization is an appropriate choice due to the guarantee of
nonsingularity. Thus, by using Euler angles , ,  in an asymmetric sequence 3-2-1 z-y-x
for describing a rotation matrix, one nds 7, 8

cc sc s

Rzyx 
cs ssc cc sss sc.
2.1
ss csc sc css cc

For a rotating body the elements of the direction cosine matrix change with time, this change
relative to any reference frame xed in inertial space can be written as follows 9:


Rt  S bib Rt, 2.2

where bib  x , y , z T is the angular velocity of the body frame relative to the inertial
frame, expressed in the body frame, S is the skew-symmetric operator given by

0 z y

S bib 
z 0 x
. 2.3
y x 0

According to 10 the angular velocity can be expressed as function of the mean orbital
motion 0  and the derivatives , , , thereafter the kinematics of the rigid body is
4 Mathematical Problems in Engineering

expressed by

1 0 s cs

bib 
0 c sc 0 sss cc 2.4
0 s cc css sc

since large slewing maneuvers of the satellite are not considered, it is save to approximate
c 1, s , 0. According to 2.4 for small Euler angles, the kinematics can be
approximated as



bib 
0 1 . 2.5

The dynamics of a satellite attitude, equipped with six one-sides thrusters is modelled by
using the Euler equations. Furthermore, the attitude dynamic is written in the body frame, it
yields


dh
ext  bib hb , 2.6
dt b

where hb  Jbib is the momentum of the rigid body, J is the satellite inertia matrix, and ext
are the external torques acting in the system including perturbation and thruster actuation.
Using dh/dtb  J bib , 2.6 becomes

J bib S bib Jbib  bd bc , 2.7

where bd represents all the disturbance torques, for example, atmosphere drag, gravity
gradient, and so on, and bc represents the control torques used for controlling the attitude
motion. The control torques about the body axes, x, y, and z provide by the thrusters are
x , y z , respectively. The thruster reaction system is discussed in detail in the following
section. The torque eect caused by the gravity gradient is taken into account and it is
included in the linearization process. An asymmetric body subject to a gravitational eld
experience a torque tending to align the axis of the least moment of inertia with the
eld direction 8. For small angle maneuvers, the model of the gravity gradient torque is
approximated as 8, 9

Jz Jy

bg  302
Jx Jz  . 2.8
0

Substituting 2.4 into 2.6 and adding the control and gradient gravity torque, we linearize
the satellite attitude model. Moreover, the linearization is performed around the LHLV orbital
Mathematical Problems in Engineering 5

frame, it is thus adopted for the stabilization mode. Afterwards the attitude model can be
represented in the state space form 6, 10

x  Ax Bu,
2.9
y  Cx Du,

T
with states x  , , , , ,  , and inputs u  x , y , z T .A is the state matrix, B is the
input matrix, C is the output matrix, and D is the direct transmission matrix. In the particular
problem they are given by

0 0 0 1 0 0

0 0 0 0 1 0


0 0 0 0 0 1

2 
4 Jz Jy 0 Jx Jy Jz
A

0
0 0 0 0 ,

Jx Jx
302 Jx Jz 

0 0 0 0 0
Jy
 
0 Jx Jy 0 Jy Jx Jz
2
0 0 0 0
Jz Jz

0 0 0

0 0 0


0 0 0


l
B 0 0 , C  I66 , D  063 .
J
x

0 l 0

J y
l
0 0
Jz
2.10

It is worth to note that x row and z yaw axes belong to a multi-input and multi-output
MIMO system 4 2 and the y pitch axis could be dealt as a single input and single output
system SISO by assuming a tachometry feedback control. Although the controller is project
over the linear model, the nonlinear model is used in the simulations.

2.2. Virtual Reality Model of the Spacecraft

In this work a Virtual Reality VR model are developed as a visualization tool. The purpose is
to visualize the simulations giving a fast and a visual feedback of the simulation models over
time. The model is produced by using the virtual reality model language VRML format
which includes a description of 3-dimensional scenes, sounds, internal actions, and WWW
anchors. It enables us to view moving three-dimensional scenes driven by signals from the
6 Mathematical Problems in Engineering

Figure 2: A graphical interface in VRML for visualization.

dynamic model, that is, attitude dynamics. The VR model was created with the use of V-
Realm builder tool, a more detailed description can be found in 11. Figure 2 shows the
basic structure representation of the spacecrafts bus. The payload is not illustrated.

3. Thruster Attitude Control System

In this section the controller design based on the Linear Quadratic Gaussian LQG technique
is briey described, afterwards the PWPF modulator is presented in details.

3.1. LQG Controller Design

The Linear Quadratic Gaussian LQG or H2 control consist of a technique for designing
optimal controllers. The approach is based on the search of the tradeo between regulation
performance of the states and control eort 12. The referred optimality is expressed by a
quadratic cost function and allows the designer to shape the principal gains of the return
ration, at either the input or the output of the plant, to achieve required performance or
robustness specications. Moreover the method is easily designed for Multi-Input Multi-
Output MIMO systems. The controller design takes into account disturbances in the plant
and measurement noise from the sensors. Formally, the LQG approach addresses the problem
where we consider a linear system model perturbed by disturbances w, and measurements
of the sensor corrupted by noise which includes also the eects of the disturbances by
measurement environment. The state-space model representation of the linear or linearized
system with the addition of the disturbance eects can be mathematically expressed by

x  Ax Bu Gw,
3.1
y  Cx Du ,

in our problem A, B, C, and D are given by 2.10. The matrix G is the disturbance balance
matrix. The disturbance and measurements noises are assumed both white noises. The
principle of the LQG is combine the linear quadratic regulator LQR and the linear-quadratic
estimator LQE, that is, a steady-state Kalman lter. The separation principle guarantees that
those can be design and computed independently 13.
Mathematical Problems in Engineering 7

3.1.1. LQR Problem

The solution for the optimal state feedback controller is obtained by solving the LQR problem.
Namely the LQR optimal controller automatically ensures a stable close-loop system, and
achieves guarantee levels of stability and robustness for minimal phase systems, for example,
multivariable margins of phase and gain. The LQR approach gives the optimal controller
gain, denoted by K, with linear control law:

u  Kx, 3.2

which minimizes the quadratic cost function, given by

 
JLQR  xT Qx uT Ru dt, 3.3
0

where Q is positive denite, and R is semipositive denite, these are weighting or tuning
matrices that dene the trade-o between regulation performance and control eorts. The
rst term in 3.3 corresponds to the energy of the controlled output y  x and the second
term corresponds to the energy of the control signal. The gain matrix K for the optimization
problem is obtained by solving the algebraic matrix Riccati equation:

AT P P A P BR1 BT P Q  0. 3.4

The optimal control gain is then obtained by

K  R1 BT P. 3.5

The close-loop dynamics model is obtained by substituting 3.5 into 3.1, and taking w 
v  0, as follows

x  A BKx, 3.6

which corresponds to an asymptotically stable system.

In order to adopt the LQR formulation the whole state x of the process has to be
measurable. In this case it is necessary to estimate the absent states, so the estimated states
are denoted by x. Notice that the output matrix in our case is C  I66 , it means that the
whole state is measurable. Physically, the angular rates are obtained from the gyros and the
attitude/orientation from the solar sensor. Nevertheless, because of the presence of noise, an
estimation is advice in order to produce better and reliable information about the real states.
The estimation is performed by employing the steady-state Kalman lter.
8 Mathematical Problems in Engineering

3.1.2. Kalman Filter Design

The Kalman lter is used to obtain the estimated state x. The lter equation in view of the
attitude model is given by

x  Ax Bu Ly Cx, 3.7

where L is the Kalman lter gain. The optimal gain L minimizes the covariance of the error E
between real x and estimated x states, by dening the state estimation error as e : x x, the
cost function is given by 13
 
JLQE  lim E e eT . 3.8
t

We assume that the disturbances aecting the process w and v are zero-mean Gaussian white-
noise process with covariances Qe  EwwT  and Re  EvvT , respectively. The process and
measurement noises are uncorrelated from each other. The gain L is obtained solving the
algebraic matrix Riccati equation:

AT S SA SCR1
e C S Qe  0.
T
3.9

The optimal estimator gain is then obtained by

L  R1 T
e C S, 3.10

and the error dynamics is given by

e  A LCe, 3.11

where A LC is asymptotically stable. From 3.6 and 3.11 the open-loop transfer function
for the LQG controller is found as follows:

Klqg sGs  KsI A BK LC1 Ls, 3.12

where Gs  s  CsI A1 B is the transfer function of the attitude model, in this case
Gs is a matrix of transfer functions.

3.2. Pulse-Width Pulse-Frequency Modulator

The control signals from the LQG controller are of continuous type. However, pulse thruster
devices can provide only on-o signals generating nonlinear control action. Nonetheless,
those can be used in a quasilinear mode by modulating the width of the activate reaction
pulse proportionally to the level of the torque command input. This is known as pulse-width
modulation PW. In the pulse-width pulse-frequency PWPF modulation the distance
between the pulses is also modulated. Its basic structure is shown in Figure 3.
Mathematical Problems in Engineering 9

To thruster
Demanded
torque Um
rt rp t et km ft UonUo Um t
kpm
m s 1 Uo Uon
Um
Filter

Schmitt trigger

PWPF modulator

Figure 3: Pulse-width pulse-frequency PWPF modulator.

The modulator includes a Schmitt trigger which is a relay with dead zone and
hysteresis, it includes also a rst-order-lter, lag network type, and a negative feedback loop.
When a positive input to the Schmitt trigger is greater than Uon , the trigger input is Um . If
the input falls below Uo the trigger output is 0. This response is also reected for negative
inputs in case of two side-thrusters or those thruster that produce negative torques clockwise
direction. The error signal et is the dierence between the Schmitt trigger output Uon and
the system input rt. The error is fed into the lter whose output signal ft and it feeds the
Schmitt trigger. The parameters of interest for designing the PWPF are: the lter coecients
km and m , the Schmitt trigger parameters Uon , Uo , it denes the hysteresis as h  Uon Uo ,
and the maximal/minimal Um . The PWPF modulator can incorporate an additional gain
kpm which will be considered separately from the control gain.
In the case of a constant input, the PWPF modulator drives the thruster valve with
on-o pulse sequence having a nearly linear duty cycle with input amplitude. It is worth to
note that the modulator has a behavior independent of the system in which it is used 3. The
static characteristics of the continuous time modulator for a constant input C are presented
as follows:

i on-time
 
h
Ton  P W  m ln 1 , 3.13
km C Um  Uon

ii o-time
 
h
To  m ln 1 , 3.14
km C Uon h

iii modulator frequency

1
f , 3.15
Ton To
10 Mathematical Problems in Engineering

Table 1: Recommended range for the PWPF parameters.

Static analysis Dynamic analysis Recommended

km 2 < km < 7 N/A 2 < km < 7
m 0.1 < m < 1 0.1 < m < 0.5 0.1 < m < .5
Uon Uon > 0.3 N/A Uon > 0.3
h h > 0.2Uon N/A h > 0.2Uon
kpm N/A kpm 20 kpm 20

iv duty cycle

1
ln1 a/1 x
DC  , 3.16
1 ln1 a/x

v minimum pulse-width PW


h
 m ln 1 , 3.17
km Um

where the following internal parameters are also dened: dead zone Cd  Uon /km , saturation
level Cs  Um Uon h/km , normalized hysteresis width a  h/km Cs Cd , and normalized
input x  C Cd /Cs Cd .
In order to determine the range of parameters for the PWPF modulator, static
and dynamic analyses are carried out. The static analysis involves test input signals, for
example, step, ramp, and sinusoidal signals. The dynamic analysis uses plant and controller.
Afterwards the choice is based upon the number of rings and level of fuel consumption
results. The number of rings gives an indication of the life-time of the thrusters. Table 1
presents the obtained results for the particular problem.

3.3. Specications and Tuning Schemes

The specication of the requirements for the attitude control system are determined by the
capabilities of the MMP satellite to attempt some desired nominal performance for the linked
payload. Considering the stabilization mode the following specications are given in terms
of time and frequency domain:
i steady state error less than 0.5 degrees for each axis;
ii overshoot less than 40%;
iii short rise time or fast response against disturbances;
iv stability margins gain GM 6 db and PM 60 for each channel.
For the control design, it is necessary to check the limitations and constraints imposed by
the plant. In this sense the optimality of the LQG only holds for the following assumptions:
the matrix A B must be stabilizable and A C must be detectable. In the case of
the attitude model, both conditions are satised. The next step is to design a controller
which achieves the required system performance. During the stabilization mode, it is desired
Mathematical Problems in Engineering 11

attenuation of the eects of disturbances acting on the satellite and accomplishment of

regulation to maintain the satellite in the required attitude. Moreover the output has to be
insensitive to measurements errors. Unfortunately there is an unavoidable tradeo between
attenuation disturbances and ltering out measurement error. This tradeo has to be kept
in mind during the design of the controller. In the case of attitude model, the disturbances
acting in the system belong to the spectrum at low frequencies, note that the regulation
signals belongs also to spectrum at low frequencies. On the other hand, the measurement
noises and unmodeled system terms are concentrated at high frequencies. In order to fulll
the specications, tuning of LQG gains and PWPF gains have to be careful performed. The
nature of the tuning is an iterative process which turns out less arduous with the use of a
computational tool, in this work the Matlab package is used. In the following, the obtained
weights for LQG controller and PWPF modulator are presented.

3.3.1. LQR Tuning

The rst choice for the tuning matrices Q and R is taken from the Brysons rule, selecting Q
and R diagonal matrices with the form

1
Qii  i {1, 2, . . . , n},
maximum acceptable value of xi2
3.18
1
Rii  j {1, 2, . . . , m},
maximum acceptable value of u2j

where xi and uj are the states input signals boundaries, respectively. The rule is used to keep
the states and inputs below some boundaries. It is advised to avoid large control signals
which from the engineering point of view are unacceptable. On the other hand, the controller
has to fulll all the system specications and the LQR formulation does not directly allow
one to achieve standard control system specications. Nevertheless those can be achieved
by iteration over the values of the weights of Q and R in the cost function till it arrives
at satisfactory controller. For the proposed reaction attitude control system the boundaries
for the states are kept 5 in attitude , , , and 1 degree per second for the rates. The
boundary for the input signals are 1 Newton meter. The result weighting matrices for the
controller which achieved satisfactory controller are

Q  Qii , R  1 101 Rii . 3.19

The control tuning matrices R and Q were obtained through iterative process following
expectable requirements, for example, allowed non-saturation control eort and reasonable
stabilization time.

3.3.2. Filter Tuning

The tuning weight matrices Re and Qe for the Kalman lter are obtained considering Re
large compared to Qe . It corresponds to weighting the measurements less than the dynamics
model. This also leads to a reduction of the poles values for A LC. The relative magnitude
12 Mathematical Problems in Engineering

Table 2: PWPF parameters used to compose the ACS.

km m Uon h Kp
1 0.1 0.45 0.3 20

Table 3: Simulation parameters.

Parameters Values
Jx  305.89126
Principal momentum of inertia without payload kgm2  Jy  314.06488
Jz  167.33919
Torque arm m l  1.0
Mean orbital motion rad/s 0  0.001
Mass kg 578.05239
Orbit altitude km 750
Maximum force N 5
Eccentricity
0
Initial attitude degrees slew maneuver ,,10,10,10
Initial Angular Rate degrees/s bib  1, 1, 1T

of Re and Qe is determined iteratively till achieves satisfactory gain L in terms of ltering and
smoothing of the measurement vector signal yv . The matrices values are given by


Qe  diag 0, 0, 0, qe , qe , qe ,
3.20
Re  diagre , re , re , ve , ve , ve ,

where qe  5 103 , re  1 101 , and ve  1 102 . Note that the precision for the rate
measurements is bigger than for the attitude measurements, and the tuning values for the
dynamic noise in the attitude are selected as zeros.

3.3.3. Selected PWPF Parameters

In order to compose the entire reaction thruster attitude control system and to achieve the
desire performance the parameters for the PWPF are selected from the optimal range. Table 2
presents those PWPF parameters.
Next section presents the performance of the reaction thruster attitude controller
during the stabilization mode. Filtering noise, rejection of impulse disturbances, and
regulation performance are analyzed.

4. Numerical Simulation and Results

The reaction thruster attitude control is tested through numerical simulations. The tuning
matrices schemes presented in Section 3 are used to obtain the controller and observer gains.
They are able to attempt pointing requirements <0.5  and reasonable margins of stability
for the attitude control system during the stabilization mode. Table 3 includes the values of
Mathematical Problems in Engineering 13

1 0.1

Error deg
0.5
deg

0 0
0.5
1 0.1
0 50 100 150 200 0 50 100 150 200
Time s Time s

a b

1 0.1

Error deg
0.5
deg

0 0
0.5
1 0.1
0 50 100 150 200 0 50 100 150 200
Time s Time s

c d

1 0.1
Error deg

0.5
deg

0 0

0.5
1 0.1
0 50 100 150 200 0 50 100 150 200
Time s Time s

yv measurement Error
x estimated 3 sigma

e f

Figure 4: Filtering of the measurement data.

principal momentum of inertia without payload 1. Although several simulations over a
wide range of initial condition for attitude were performed, just one case is shown.

4.1. Noise Filtering

In order to lter the noise in the measurement yv the steady-state Kalman lter is applied. The
estimated and measurement attitude is shown in Figure 4, on the left side. The errors and the
respective 3 sigma boundary results are shown on the right side of Figure 4. The steady-state
Kalman achieves good estimation of the real attitude with a standard deviation  0.027. The
estimation results present smoother prole compare to the measurements which is favorable
wish for the control system.

4.2. Short Slew Maneuver during Stabilization Mode

Although a set of dierent initial conditions are simulated and analyzed, we present only the
case when the satellite has a displacement of 10 degrees for each axis in attitude with respect
to LHLV orbital frame. To regulate or stabilize the satellite a short maneuver is needed. The
14 Mathematical Problems in Engineering

Slew maneuver
12

Attitude deg
10
8
6
4
2
0
2
0 200 400 600 800 1000 1200 1400 1600
Time s

a

Impulse disturbances
0.4
Attitude deg

0.2
0
0.2
0.4
0 200 400 600 800 1000 1200 1400 1600
Time s

b

Figure 5: Attitude prole during slew maneuver and disturbance eect.

attitude prole is shown in Figure 5. The simulation time corresponds to quarter of the orbital
period  104 minutes. The satellite executes the maneuver in approximately 100 seconds.
The duty cycle generated by the PWPF modulation is shown in Figure 6. The duty
cycle for row, pitch, and yaw angles are the same order of magnitude. The maximal spent time
to complete a close path is quasi 800 seconds and it occurs in row direction. The specication
of pointing accuracy is achieved, less than 0.5 degrees. In fact reading out Figure 6 the
maximal errors in row, pitch, and yaw are 0.3, 0.3, and 0.25 degrees, respectively. It shows
a high accurate performance of the reaction thruster which is possible by modulating the
control signal using the PWPF modulator.
Figure 7 shows the control command, executed by LQG controller, and the modulation
during the slew maneuver. The thrusters prole present small pulse-width modulation ton 
which leads small impulses and hence less fuel consumption. Positive torques are executed
by 3 of thrusters and negative ones by another 3 thrusters.

4.3. Disturbance Rejection During Stabilization Mode

In order to test the action of the controller against disturbance eects e.g., atmospheric drag
or uncertainties in the system e.g., sloshing, simulations considering impulse disturbance
signals acting in the system are carried out. Figure 5 presents the results obtained for the atti-
tude. It shows a maximal error in attitude of 0.2 degrees which fulll the desired precision
specication. The attitude control system is capable of respond fast to the disturbance eects,
less than 60 seconds for an exogenous pulse of 1 Newton. The control signal command and
the PWPF modulation results are presented in Figure 8. The results are satisfactory in terms
of accuracy and fuel since the modulation of the pulse-width is kept small.
Mathematical Problems in Engineering 15

Duty cycle Roll

0.2 0.01
0

v
0.1 0.01
0.02
0.4 0.2 0 0.2 0.4 0.6
0
Pitch
0.02
Rates deg/s

0.1

v
0

0.2
0.02
0.4 0.2 0 0.2 0.4 0.6

0.3 Yaw
0.02

0.4

v
0

0.5 0.02
2 0 2 4 6 8 10 0.4 0.2 0 0.2 0.4 0.6
Euler angles deg

Figure 6: Duty cycle response for stabilization during a short slew maneuver.
Control command

0.1
0
Nm

0.1
0.2
0.3
0 20 40 60 80 100 120 140 160 180 200
Time s

u
u
u

a

5 5 5
T N

T N

0 0 0

5 5 5
0 50 100 150 200 0 50 100 150 200 0 50 100 150 200
Time s Time s Time s

b c d

Figure 7: Control command and PWPF modulation during the slew maneuver.
16 Mathematical Problems in Engineering

Control command
0.2
0

Nm
0.2
0.4
0.6
0.8
0 200 400 600 800 1000 1200 1400 1600
Time s

u
u
u

a

5 5 5

T N
T N

T N

0 0 0

5 5 5
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
Time s Time s Time s

b c d

Figure 8: Control command and PWPF modulation for disturbance rejection.

It is worth to note that the use of magnetic coils actuators or reaction wheel devices can
zero the error residue, for example, duty cycle, by a damped actuation. This actuation will be
very small because of the level of accuracy in attitude achieved by the thruster actuation.

5. Conclusions
The obtained on-o thruster reaction attitude control system based on the LQG/PWPF
modulation is optimal with respect to regulation, i.e., minimizing the quadratic cost function
of states and control signals, and propellant consumption. The optimality for fuel is obtained
through o-line simulations varying the parameters of the PWPF modulator till less fuel
consumption is achieved. This work presents the set of optimal parameters for the PWPF
modulator by considering static and dynamic analysis.
The LQG design is an ecient way to achieve exponentially stability, moreover it
allows to weight the magnitude of input signal u, restricting the torque commands till
acceptable performance is achieved. The weighting matrices for tuning the optimal LQG
controller are presented and discussed in this work. The previous work, see 6, using the
LQG design, demonstrated successfully, the applicability and suitability of the controller for
the stabilization mode. However, in the foregoing work the required on-o modulation was
not taken into consideration. It is worth to note that the LQG controller is able to stabilize the
system even for large initial attitude displacements within nonlinear dynamics. It shows how
resistent the controller is in face of internal changes in behavior.
The obtained results demonstrate the feasibility of combining LQG/PWPF modulator
in an unique controller for on-o thruster reaction attitude control system. Stability remains
Mathematical Problems in Engineering 17

by adding the PWPF modulator and reasonable accuracy in attitude is achieved, that is,
magnitude of the duty cycle. Practical aspects are included in this study as ltering and
presence of external impulsive perturbations. The advantages of less spent propellant shall
contribute to the MMP project, specially, a satellite conceived to be used on a large number
and dierent types of missions, in the context of an ever-advancing Brazilian space program.

Acknowledgments
The authors acknowledge the support of Conselho Nacional de Desenvolvimento Cientico e
Tecnologico-CNPq, Coordenacao de Aperfeicoamento de Pessoal de Nivel Superior CAPES Brazil,
and Fundacao de Amparo a Pesquisa do Estado de S. Paulo-FAPESP Sao Paulo, Brazil.

References
1 AEBAgencia Espacial Brasileira, National space activities program PNAE 19982007, Report
Brazilian Space Agency, AEB, Brasilia, Brazil, 1998.
2 Q. Hu and Y. Liu, A hybrid scheme of feed-forward/feedback control for vibration suppression of
exible spacecraft with on-o actuators during attitude maneuver, International Journal of Information
Technology, vol. 11, no. 12, pp. 95107, 2005.
3 T. D. Krovel, Optimal tunning of PWPF modulator for attitude control, M.S. thesis, Department of
Engineering Cybernetics, Norwegian University of Science and Technology, 2005.
4 T. D. Krovel, PWPF modulation of thrusters on the micro-satellite SSETI/ESMO, Tech. Rep.,
Department of Engineering Cybernetics, Norwegian University of Science and Technology, 2005.
5 R. S. McClelland, Spacecraft attitude control system performance using pulse-width pulse-frequency
modulated thrusters, M.S. thesis, Naval Postgraduate School, Monterey, Calif, USA, 1994.
6 A. C. Santana, L. S. Martins-Filho, and G. Arantes Jr., Attitude stabilization of the PMM satellite
using a LQG-based control strategy, Tendencias em Matematica Aplicada e Computacional, vol. 9, no. 2,
pp. 321330, 2008.
7 J. R. Wertz, Ed., Spacecraft Attitude Determination and Control, Reidel, Dordrecht, The Netherlands,
1978.
8 M. H. Kaplan, Modern Spacecraft Dynamics and Control, John Wiley & Sons, New York, NY, USA, 1976.
9 B. Wie, H. Weiss, and A. Arapostathis, Quaternion feedback regulator for spacecraft eigenaxis
rotations, Journal of Guidance, Control, and Dynamics, vol. 12, no. 3, pp. 375380, 1989.
10 G. Arantes Jr., Estudo comparativo de tecnicas de controle de atitude em tres eixos para satelites articias, M.S.
thesis, INPE, S. Jose dos Campos, Brazil, 2005.
11 LIGOS, V-Relam Builder: Users Guide and Reference, Ligos Corporation, 1997.
12 J. B. Burl, Linear Optimal Control, Prentice Hall, Upper Saddle River, NJ, USA, 1998.
13 B. D. O. Anderson and J. B. Moore, Optimal Control: Linear Quadratic Methods, Dover, Mineola, Tex,
USA, 2007.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 507370, 23 pages
doi:10.1155/2009/507370

Research Article
Highly Efcient Sigma Point Filter for Spacecraft
Attitude and Rate Estimation

Chunshi Fan1 and Zheng You2

1
Department of Precision Instruments and Mechanology, Tsinghua University, Beijing 100084, China
2
State Key Laboratory of Precision Measurement Technology and Instruments, Tsinghua University,
Beijing 100084, China

Correspondence should be addressed to Chunshi Fan, [email protected]

Received 9 July 2009; Accepted 30 September 2009

Recommended by Tadashi Yokoyama

Nonlinearities in spacecraft attitude determination problem has been studied intensively during
the past decades. Traditionally, multiplicative extended Kalman lter MEKF algorithm has been a
good solution for most nominal space missions. But in recent years, advances in space missions
deserve a revisit of the issue. Though there exist a variety of advanced nonlinear ltering
algorithms, most of them are prohibited for actual onboard implementation because of their
overload computational complexity. In this paper, we address this diculty by developing a new
algorithm framework based on the marginal ltering principle, which requires only 4 sigma points
to give a complete 6-state attitude and angular rate estimation. Moreover, a new strategy for sigma
point construction is also developed to further increase the eciency and numerical accuracy.
Incorporating the presented framework and novel sigma points, we proposed a new, nonlinear
attitude and rate estimator, namely, the Marginal Geometric Sigma Point Filter. The new algorithm
is of the same precision as traditional unscented Kalman lters, while keeping a signicantly lower
computational complexity, even when compared to the reduced sigma point algorithms. In fact, it
has truly rivaled the eciency of MEKF, even when simple closed-form solutions are involved in
the latter.

Copyright q 2009 C. Fan and Z. You. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
Nonlinearities in spacecraft attitude determination problem have been studied intensively
during the past decades. Since the early 1980s, multiplicative extended Kalman ltering
MEKF algorithm 1 has proven to be a successful solution for engineering application. The
MEKF algorithm has a very low computing cost and performs quite well in most nominal
space missions where the spacecrafts angular rate is slow and the nonlinearities are not so
impactive.
2 Mathematical Problems in Engineering

In recent years, advances in space missions, such as the greater agility demand and
the application of lower cost sensors, deserve a revisit of the nonlinearity issue. Although
a variety of advanced nonlinear ltering algorithms exist, only few of them are close to the
restrict numerical expense requirements of actual onboard implementations. In the existing
methods, the well-known sigma point Kalman lters SPKFs 2, 3 have approven to
be among the most ecient ones. SPKF bases on a Gaussian distribution approximation
technique, namely, the unscented transformation UT, where a deterministic set of weighted
points known as the sigma points are used to make probabilistic inference 4. Eliminating
the complex Jacobian matrix derivations, SPKF algorithms are much easier to implement and
have better performance than traditional widely used EKF algorithms; so they have found
widespread application in a variety of elds. In recent years, spacecraft attitude estimation
problems have also been addressed by SPKF approaches in literature and engineering
practice 57.
In spite of being ecient among nonlinear lters, baseline SPKF still seems
computational costly for engineering implementation. If we denote m as the number of
sigma points required, then for an n-dimension random state vector x, standard unscented
transformation needs a set of m  2n 1 points to capture the states statistical distribution
properties. More seriously, when we develop a complete SPKF estimator for the n-
dimensional state model, the actual m needed is no longer 2n 1it easily becomes
4n or even larger, because standard SPKF algorithm requires a state augmentation to
include all the propagation and measurement noise terms, hence leading to an unacceptable
increase in computational burden. For avoidance of state augment, iterated and trapezoidal
approximation approaches 5, 8 have been developed. Both approaches are suitable for
additive noise case only, and they are able to reduce m back to 2n 1. To further reduce the
complexity, strategies for introducing fewer sigma points are exploited, known as the reduced
sigma point lters. Several simplex points selection strategy have been developed, including
the n 2 point minimal-skew simplex points 9, the spherical simplex points 10, 11, and
some enhanced higher-order extensions 12. Each of the above sigma point sets involves
a zero-valued central point, which is usually endowed with a negative weight so as to
minimize higher order eects, known as the scaled UT technique 13. In recent years, new
sigma point selection strategy involving no central-point is introduced, such as the Schmidt
orthogonalization-based simplex set 14, which includes n 1 equally weighted points. It
is also proved that 15 such negative weight-free, equally weighted sigma point sets are
numerically more stable and accurate as well as having a better eciency. In fact, both central-
point elimination and equal weight assignment improve the symmetry property of the sigma
point set, and a better symmetry property provides a better numerical behavior, as it has a
better capability to suppress the impact of the round o errors. In this article, we address the
construction strategies to make a best symmetric structure in simplex sigma point set.
Anyhow, applications of simplex sets have reduced the required sigma points to
50% of the traditional nonaugmented algorithms and have made a signicant improvement
in numerical eciency. In fact, the numerical eciency of simplex SPKF algorithm is
able to rival or even exceed its EKF counterpart if general formed Riccati equations and
numerical integration process are involved in the latter. However, for typical quaternion-
based spacecraft attitude problems, since simple analytical solutions and sparse matrices
exist for MEKF covariance propagation and measurement updates, general simplex SPKF
algorithm still compares unfavorably with MEKF in eciency.
Clearly, to develop a competitive algorithm alternative to the MEKF for practical
applications, we need a still further reduction of m. Recently, a new sigma point selection
Mathematical Problems in Engineering 3

strategy for a class of partially linear transformation problems has been proposed, namely,
the marginal unscented transformation 16. By exploiting the linear substructures in system
model and margining out the corresponding variables, the marginal UT algorithms only
needs a set of sigma points that adequately describes the statistics property of the nonlinear
part of the states. It provides a possible approach to further shrink the size of the sigma
point set. As attitude dynamics also has linear substructures in gyro bias drift model and
the observation equations, it is imaginable that we can also address the attitude and rate
estimator design with similar ideas.
The main contributions of this article include two parts. First, we have derived in
detail a marginal version of SPKF algorithm for typical 6-state attitude determination system.
The new algorithm uses merely 4 sigma points to give a complete attitude and angular rate
estimation; hence it is able to achieve a high numerical eciency that truly rivals the MEKF.
Second, we have proposed a new set of simplex sigma points for Euclidean Geometric space,
named the Geometric Simplex sigma point set. The new set has a symmetric structure, a lower
computational expense and is numerically more accurate. It would be of use in a variety of
3-dimensional modeled dynamic problems.
The organization of this paper proceeded as follows. First, we established a general
6-state stellar-inertial spacecraft attitude kinemics and measurement model and analyzed
the partially linear structure in the system. Then a marginal SPKF estimator is derived in
detail. Next, we looked into the asymmetrical properties of existing sigma point construction
algorithms and proposed the Geometric simplex set. Finally, we incorporated the proposed
sigma point set into the marginal ltering framework to congure a complete attitude
estimator, named the Marginal Geometric Sigma Point Filter, and inspected its performance
in simulation with comparisons to the traditional MEKF and Spherical Simplex SPKF.

2. Models and Traditional Filtering Frameworks

2.1. Attitude Dynamics and the MEKF Framework
For spacecraft attitude estimation, quaternion has been the most widely used attitude
T
parameterization. The quaternion is given by a 4-dimension vector dened as q  qT , q4  ,
with q q1 , q2 , q3 T  n
 sin/2 and q4  cos/2, where n
 is Euler axis and is the
rotation angle. Quaternion parameter satises a single constraint given by qT q  1. The
kinematics equation is given by

1
q  q, 2.1
2

where is the angular rate vector given from the gyros measurement meas by
compensating the gyro bias b:

 meas b ARW , 2.2

where ARW is a zero-mean Gaussian angular random walk noise with a covariance of
2
ARW I3 . b is often modeled as a rate random walk process with white noise RRW and a
2
covariance of RRW I3 :

b  RRW . 2.3
4 Mathematical Problems in Engineering

From 2.12.3, we can derive the discrete-time version of the above models with
numerical integration. In fact, numerically simpler closed-form solution exists for 2.1 if
we approximately consider the direction of as a constant vector during the propagation
interval T for each ltering circle from tk1 to tk :

  
q k/k1  qk1 qk1 , 2.4

where
 T


T
sin k1 /2 
 k1
q k1  k1  , cos 2.5
2 cos  /2 2
k1

with

 k1  meas 
k1 bk1 ,
  2.6
  
k1  k1 T, k1   k1 .

The approximation is tenable as far as the period T is small enough, which is usually well
satised in practice, and a second-order accuracy is guaranteed.
In actual calculation process, 2.5 seems too complex, and a 2nd-order approximation
T
is enough. To associate the quaternion q   nT sin/2, cos/2 to its pertinent
rotation vector  n in a straightforward way, we can choose an arbitrarily 3-dimensional
attitude parameter as the media. In this paper, we choose the Modied Rodrigues Parameters
MRPs recommended in 17 to give a 2nd-order, trigonometric function, and square-root
function-free approximation of 2.5:

4q
q   ,
1 q4
  T 2.7
8T , 16 2
q  
16 2

with 2  T . Clearly, such MRP-based expression is quite economic for computation. Then
the discrete-time propagation equations can be written as follows:

 T
T
 
8k1 , 16 k12


q   ,
k1
16 k1
2 2.8

 k/k1  b
b  k1 .

To cope with the unit-length constraint of quaternion, local error states also known
as the local disturbance states are introduced into lter design. We describe the local
Mathematical Problems in Engineering 5

attitude error as a 3-dimensional rotation vector a and the local gyro bias error as another 3-
dimensional vector b. Then the actual state vector processed by the lter is the 6-dimension
disturbance state:
 T
x  aT , bT . 2.9

 and b as the global states, which would

For clarity, we hence address the original states q
mainly serve as singular-point-free reference and storage for the lter. The global states and
the local error state are aliated as

 qa,
qq
2.10
 b,
bb

where q is the local error quaternion corresponding to a. Again we choose to use the MRP
approximation, as

4q
aq   ,
1 q4
 T  T 2.11
8a , 16 a2
qa 
16 a2 

with a2  aT a. Clearly, such an MRPs-based expression is free from any square-root or

trigonometric functions, economic in computation.
After all generality, the observation model in this article is established as an automatic
star sensor with quaternion measurements qmeas k . But in actual practice this information is
presented to the Kalman lter in a more convenient way as in terms of a 3-dimensional
parameter. Again we choose to use the MRP parameter, and then the star sensors observation
model is simply dened as the local error between the predicted and observed attitudes:


zmeas
k  hxk   a qmeas
k vk , 2.12

where

qmeas  1
q meas
k k/k1 qk , 2.13

and vk is the measurement noise covariance modeled as

Rk  r2 I3 . 2.14

With the above models, we can derive an MEKF estimator as in 1 and briey reviewed
in blocked matrix form in Table 4. Note that the most computational cost parts of the
6 Mathematical Problems in Engineering

algorithm involve the covariance propagation, measurement update, and the Kalman lter
gain computing, namely,

k/k1  k1 Pk1 k1 Qk , 2.15

T
PX X

 1
K k  PX HT
k/k1 k Hk P X
HT
k/k1 k R k , 2.16

k  I6 Kk Hk Pk/k1 ,
PX 2.17
X

where PXk
is the estimation of xs covariance, Kk is the Kalman lter gain, Hk is the observation
matrix, and k and Qk are, respectively, the transition matrix and the propagation noise
matrix, as

   T
k1 k1 QA QBA
k1  , Qk  k k . 2.18
033 I3 QBA
k QBk

General problem would involve numerical integration of Riccati equations to evaluate

the k matrix in 2.15, and complex derivations to evaluate the Hk matrix in 2.16 and
2.17. Fortunately, it is also found that simple analytical solutions exist for k1 , k1 see
Table 4 18, and Qk 19 as

   
1
QA  T 2
ARW 2
T 2
I3 ,
k
3 RRW
 
QBA 
1
2 T 2 I3 , 2.19
k
2 RRW

QBk  RRW
2
TI3 .

Moreover, Hk involves sparse and unit matrices as

Hk  I3 , 033 . 2.20

Hence the MEKF algorithm is numerically very ecient. The computational complex of a
typical stellar-inertial MEKF attitude estimator is evaluated in Table 4.

2.2. Traditional Sigma Point Filtering Framework

We consider now the application of SPKF to the system discussed above. A set of m sigma
points are chosen to approximate the statistic distribution of the 6-dimension disturbance
T
state x  aT , bT  :

 T
i  iT , iT , i  1, 2, . . . , m, 2.21
Mathematical Problems in Engineering 7

where i is the index, i , i represent the attitude and bias error, respectively, each point
is assigned with a weight W i , and all the weights satisfy the normalization constraint:


m
W i  1. 2.22
i1

For clarity, write the sigma points and their associated weight in matrix form, as

   

1 2 m
  1 , 2 , . . . , m
 , ,...,   ,
 1 , 2 , . . . , m
  2.23
  W 1 , W 2 , . . . , W m ,

 diag  .

For unbiased distribution of x n

, is constructed as

 SX  , 2.24

where SX is an arbitrary
square-root matrix of PX with SX SX T  PX  covx,x, also
denoted as S  P .   u , . . . , u 
X X 1 m nm
is a base set of sigma points, and it can
be dened in several dierent rules, depending on the sigma point construction strategy we
use. Anyhow,  has an unbiased mean and a unit covariance:

  T  0,
2.25
 W  T  Im .

The construction strategy of  is also the dominating dierentiation between dierent SPKF
algorithms. With the denitions above, the set X is able to capture the statistics of xs
distribution precisely up to 2nd-order.
The SPKF attitude estimator is as follows. At the beginning time of each ltering step
tk1 , the local error state is reset to zero:

 T
T
xk1  aTk1 , b  031 , 031 T . 2.26
k1

Then we construct the sigma points of Xk1 . As in 5, we would like to use the trapezoidal
approximation to avoid state augmentation; so actually SXk1 is computed as


X
SX
k1  PX
k1
Qk1 , 2.27
8 Mathematical Problems in Engineering

where

 
A T T2 2
X Qk1 , 033 ARW
2
I3 , 033
Qk1   2 6 RRW , 2.28
B T 2
033 , Qk1 033 , RRW I3
2

and it is equivalent to have the Qk in 2.18 implicitly propagated together with the sigma
points 5.
Next, propagate the m sigma points as follows:

i

i
k1   k1 k1 T,
 iT  T
8k1 , 16 k1
i2 2.29
i
qk1   i2
,
16 k1

i i
where k1  |k1 |, and

  T
iT i2
8k1 , 16 k1
i
qk1   ,
i2
16 k1 2.30
 1
i 
qk/k1  q
i i
qk1 qk1 .
k1

Thereby we have the propagated sigma points k/k1 and k/k1 , respectively, as

i
 4qk/k1
i i
k/k1  qk/k1   , 2.31
i
1 q4k/k1

i i
k/k1  k1 . 2.32

Note we have propagated the attitude-related sigma points i directly without adhering it

to the global state q k1 as in 5; so a decrease of computational complexity is achievable. This
approach stands as far as the time interval of T is guaranteed to be small enough in order that
both i and i can be taken as small rotation vectors, and the MRP approximation is valid.
Then the predicted mean of x is

xk/k1  k/k1  T . 2.33

Mathematical Problems in Engineering 9

Now we compute the predicted measurements. Noting that the sigma points have already

implicitly changed their reference from q 
k1 to qk/k1 during propagation, so the predicted
star sensor quaternion measurements are

 i  q
 i
q k/k1 k/k1 qk/k1 . 2.34

Following 2.12 and 2.13, it is straight forward to derive the observation model for the
SPKF as

i i
k/k1  k/k1 . 2.35

Denoted in matrix form as  1 , 2 , . . . , m , then we can get the predicted
measurement k/k1 as

k/k1  Zk/k1 WT . 2.36

Next we compute the covariance predictions. If we take xk/k1 as a bias of the sigma
points, then to appropriately compute the covariance, we must remove this bias at rst. Hence
the covariance prediction process would involve two steps:

k/k1  k/k1 11m xk/k1 , 2.37

W T
k/k1  Xk/k1 Xk/k1 ,
PX 2.38

where denotes the Kronecker product, and 11m  1, 1, . . . , 11m . Similarly, we give the
unbiased measurement points set as

Zk/k1  Zk/k1 11m k/k1 2.39

the innovation covariance and cross covariance matrix are then, respectively,

Zk/k1 T Rk ,
k/k1  Zk/k1
PZ
2.40
Zk/k1 T
k/k1  Xk/k1
PXZ

and the measurement update procedure is

 1
Kk  PXZ Z
k/k1 Pk/k1 , 2.41

xk  xk/k1 Kk kmeas k/k1 , 2.42
T
k  Pk/k1 Kk Pk/k1 Kk  .
PX 2.43
X Z
10 Mathematical Problems in Engineering

 and b
Finally, update the global states q  as

 q
q k k/k1 q
ak ,
2.44
b  k/k1 b
k  b k.

Above is the framework of an SPKF version attitude estimator. A blocked-form

procedure summary is listed in Table 5. It is not dicult to evaluate the computational
complexity of the SPKF estimator. Even when reduced-point algorithms, such as the spherical
simplex unscented Kalman lter, are adopted, the computing eort is still a double of the
MEKFs, as listed in Table 3.

3. Marginal Geometric Sigma Point Filters

3.1. Marginal Sigma Point Filtering Framework for Spacecraft Attitude and
Rate Estimations
Now we look into some special structures of the above estimator. First, noting 2.32, we
nd that the bias-related sigma points  1 , . . . , m  remain unchanged during the
whole process of the propagation, indicating that their mean would also remain unchanged
as b  k/k1  b k1  031 . In other words, no information has been introduced into the
state bs mean and covariance covb, b during the propagation. Therefore, once we are
able to capture the information of a, cova, a and covb, a, we have already obtained all the
information available during time propagation.
Then noting 2.35, we nd only the attitude-related sigma points k/k1 are explicitly
used to construct the measurement predictions k/k1 . We can write a formal expression of
this transform as

  hx  a. 3.1

In fact, 3.1 belongs to a special class of nonlinear transformation, namely, the partially linear
transformation 16. Clearly, for the measurement update process, the random variable  s
mean  , cov ,  , and cross covariance cova,   are all independent of b, and it is proved
that the cross covariance covb,   is also independent of covb, b up to the 2nd-order.
The above discussions leads to the following conclusion: as long as we can construct a
set of sigma points that matches the given mean estimations of a, b,  and the covariances
of P  cova, a and P
A BA
 covb, a, it is enough to capture the rst two moments
statistics properties of the random state x with a precision up to 2nd-order. Before moving
on to construct such a sigma point set, we should also note from 2.26, 2.37 and 2.39 that
with the reset and de-bias steps embeding the lter, actually we are on the assumption that
T T
we have an unbiased distribution as x  aT , bT   0T31 , 0T31  .
So now our goal becomes to construct a minimum set of sigma points, which is able
to fully capture all the available information given as a unbiased mean estimation, that is,
a  031 , b  031 , and b the covariance estimation as PA  cova, a PBA  covb, a.
Mathematical Problems in Engineering 11

To match the unbiased mean, we have


WT  Wm 1 2 m  0, 3.2

1 2 m
 WT  Wm  0. 3.3

As stated in 9, to fully capture the mean of an n-dimensional state vector, at least m  n 1
points are needed. Noting both a and b 3 , the minimum m which satises both 3.2 and
3.3 is m  4.
To further reduce the computational expense and make a better symmetry property,
assign equal weights W for all the sigma points. Then we have


m
Wi  mW  1. 3.4
i1

Clearly,

1 1
W  , 3.5
m 4
   
1 W 1
W 11m ,  Im . 3.6
m m

a with outer products approximation,

To match as covariance estimation PA  cova,
we have
 
1
W  T   T  PA . 3.7
m

Again we make use of a base set U 3m to help matching PA . Denoting U 

ua1 , ua2 , ua3 , ua4 , the following relationship should be satised:

U WT  0, 3.8

U W U T  I3 . 3.9

Substituting 3.5 and 3.6, it is straightforward to get


m
uai  0, 3.10
i1

U U T  mI3 . 3.11

As m  4, U must be a simplex set. In spite of the constraints in 3.8 and 3.9, we still
have the freedom to choose an arbitrary realization of U . Later we will propose a novel set
12 Mathematical Problems in Engineering

derived from a heuristic geometry approach. Now supposing that U is given, then we may
construct simply as

 SA U , 3.12

T
where SA is an arbitrary matrix square-root of PA fullling SA SA   PA . It is
straightforward to validate 3.12 as substituting it to 3.7:

   T
1
W  T  SA U SA U
m
    T
1
 SA U U T SA 3.13
m
   T
1
 SA mI3  SA  PA .
m

To match the cross covariance PBA is

 
1
BW  T  B T  PBA . 3.14
m

T
Here we propose a simple and convenient algorithm to compute B. Dene SBAk1
 PBA SA  ,
which we could get with the low computational cost Gaussian elimination:

 T PBA
k1
SBA
k1  P BA
S A
  A T . 3.15
Sk1

Then we have

B  SBA
k1 U . 3.16

Proof of 3.15 is straightforward as substituting it to 3.14:


T 1  BA   A  T
B    Sk1 U S U
m
  

1   T S A T
k1 U U 
SBA
m
  3.17
1  T  T
 PBA SA mI3  SA
m
 T  T
 PBA SA SA  PBA .
Mathematical Problems in Engineering 13

Thereupon, we have constructed a desired set of marginal sigma points as in 3.12 and 3.16.
Note that it is enough to use merely one base set U to construct the full-length Sigma point
set   T , BT  . So hereafter we would suppress the superscript of U as U.
T

Looking into 2.38, we nd

   BA T
PA , PAB PA ,
P
PX   . 3.18
PBA , PB BA
P , PB

With the proposed sigma points, the propagation and innovation steps of bs covariance
estimation PB are no longer necessary. Eliminating the PB -related term and making use of the
symmetric structure of the matrix, it is enough to have PA and PBA propagated. Accordingly,
it is only necessary to have the same matrices updated. There by, we will replace 2.38 and
2.43 with

k/k1  k/k1 k/k1  ,

W T
PA

k/k1  Bk/k1 k/k1  ,

W T
PBA
 1  T
PA
k  P A
k/k1 P AZ
k/k1 P Z
k/k1 P AZ
k/k1

 T 3.19
 PAA
k/k1 K AZ
k P AZ
k/k1 ,
 1  T
PBA
k  P BA
k/k1 P BZ
k/k1 P Z
k/k1 P AZ
k/k1

 T
 PBA
k/k1 K BZ
k P AZ
k/k1 .

To avoid state augmentation, we would like to have the propagation noise terms
incorporated into the lter with trapezoidal approximation. However, as PB is no longer used,
we have to seek for alternate approach. Denote

 ! 2     
! T T2
QA
Sk1
A
 Qk1  " ARW
2
RRW I3 ,
2
2 6
% 3.20

QB B T
Sk1  Qk1  RRW I3 .
2

Then similar to 11, we can add the noise terms directly to the sigma points with the help of
U:

k1  SAk1 U SQA QA
k1 U  Sk1 Sk1 U,
A
3.21

k1 U SQ U  Sk1 SQ U.
Bk1  SBA B BA B
3.22
14 Mathematical Problems in Engineering

Table 1: Simplex base sets for 3-dimensional space.

Sigma set U nm for n  3

0,
1/ 2, 1/ 2, 0, 0

(
U  0.5/ Wm  0, 1/ 6, 1/ 6, 2/ 6, 0

Spherical simplex 0, 1/ 12, 1/ 12, 1/ 12, 3/ 12

m5 0 < W0 < 1, Wm  1 W0 /n 1

1/ 2, 1/ 2, 0, 0


U  1/ W 1/ 6, 1/ 6, 1/ 3/2, 0
Schmidt orthogonal
1/ 12, 1/ 12, 1/ 12, 1/ 4/3
m4 W  1/n 1  1/4

1, 1 1 1
U  1, 1 1 1
Geometric simplex 1, 1 1 1

m4 W  1/n 1  1/4

Table 2: Residues of numerical mean estimation.

101 , 0, 0;
102 , 0, 0;
104 , 0, 0;

S 102 , 101 , 0; 103 , 102 , 0; 105 , 104 , 0;
103 , 104 , 101 104 , 105 , 102 106 , 107 , 104

SS 2.151464 1018 1.319254 1019 5.591741 1022

Mean SO 1.734723 1018 6.505213 1019 1.694066 1021
GS 0 precise 0 precise 0 precise

SS 3.878960 1018 4.065758 1019 8.271806 1024

Covariance SO 4.249187 1018 3.030437 1019 7.018860 1024
GS 1.734723 1018 0 precise 2.339620 1024
SS: Spherical simplex set. SO: Schmidt orthogonal set. GS: Geometric simplex set.

Note in 3.22 that the sign before SBQ is negative. The reason is that with a given covariance
estimation P and its square-root S, it is equivalent for us to choose either S or S when
constructing the sigma points. Both directions work because the given mean is unbiased.
But when we add noise terms directly to the sigma points as in 3.22, we must guarantee
k1 and SQ have a common sign. Recalling 25 as
that SBA B

 
1
QBA  2 T 2 I3 , 3.23
k1
2 RRW

here the negative sign in the right side of 3.23 clearly indicates opposite signs between SBA
k1
and SBQ .
Mathematical Problems in Engineering 15

Table 3: Comparisons of total arithmetic operations.

Square
Algorithm Phase Multiplies Adds
roots
Propagation 350 270 2
Measurement update 260 185 1
MEKF
One full lter step 610 455 3
One observation circle 1680 1265 9
Propagation 350 285 5
Measurement update 380 280 1
MGSPF
One full lter step 750 565 6
One observation circle 1860 1480 21
Propagation 810 565 8
Measurement update 455 370 1
SSUKF
One full lter step 1265 935 9
One observation circle 3695 2630 33

Including 1 propagation step and 1 measurement update step.

Including several propagation steps and 1 measurement update step. In this case we take a typical value as 4 propagations
for each 1 update.

3.2. The Geometric Simplex Sigma Points

Construction of the base set U plays an important role in simplex sigma point algorithms.
Existing strategies had mainly focused on the design of general operation ow for getting a
base set for any arbitrary dimension n. In 9, 10, direction-extending technique is developed
and used to build the minimal-skew and spherical simplex set. While in 14, Schmidt
orthogonalization is employed to develop a new set. Table 2 demonstrates both the spherical
simplex set 10 and the Schmidt orthogonal simplex set 14 for n  3. Close comparison
could reveal that both sets are equivalent after a sign shift except for the existence of an
additional central point in the spherical simplex set. Both sets are easy to be extended to
higher dimension space, and because all the points are equally weighted and equidistantly
placed on a hyper sphere, they are numerically stable over the increase of n.
However, both
sets
lack numerical accuracy, and they are complex to compute, as
irrational numbers 2, 6, and so on exist. Further, they do not have symmetric structures. A
fully symmetric set needs that for every point ui U, we can get another point uj U, i /
j
simply by element permutation or sign-changing point of the generator point ui 15.
Clearly, except for the rst dimension or describing in matrix language, the rst row of
U, not even element level symmetry is guaranteed in the spherical or Schmidt orthogonal
simplex sets.
In order to make a better symmetry property, we propose a new base set of sigma
points here as in the 3rd row of Table 1 and Figure 1. It is straight forward to nd that both
3.10 and 3.11 are completely satised. For clarity, we name this new set as the Geometric
Simplex Set, for it has a nice symmetrical structure as a tetrahedron in the 3-dimensional
Euclidean space Figure 1. The proposed new sigma point set has several benets.
i The new set is more intuitive to comprehend and apply, especially for the 3-
dimensional Euclidean space, the true space where we are, and the true space in which a
variety of dynamical problems as guidance, navigation, and so on take place.
16 Mathematical Problems in Engineering

Table 4: Pseudocode and computational expense evaluation of the MEKF with closed-form solutions.

Algorithm

x0  aT0 , b   0T , 1T , b

 T T  061 , q  0  031
0 0 31

0  a I3 , P0  b I3 ,
PA 2 B 2
Initialize
QA
k
 T ARW
2
1/3RRW
2
T 2 I3

QBA
k
 1/2RRW
2
T 2 I3 , QBk  RRW
2
T I3

 k1  k1
meas  k1 ,
b  k1 |, k1 
 k1  |  k1 T 6 5 1
T

q   T , 16
8 k1
2
 /16 k1
2
 6 2
k1 k1


n  k1
 k1 / 3

k1  I3 sin k1 
n 1 cos k1 
n2
49 30
Propagation I3 k1 k1
3
n k1
/6 2
n2
/2
k1  T/2I3 1 cos k1 /
 k1 
nk1 

k1 sin k1 / nk1 2

 k1  51 30

T/2I3 k1
2
/2 nk1  k1
 k1  3
nk1 2
 k1 
/6

k/k1  k1 Pk1 k1 k1 Pk1 k1

T T
PA A B
108 81
T
k1 PBA T k1 PBA T  QA 54 57
k1 k1 k1 k1 k

PBA
k/k1
 PBA T
k1 k1
PBk1 Tk1 QBA
k
54 48

PBk/k1  PBk1 QBk 3

     
q k/k1  qk1 qk1 , qk/k1  qk/k1 / qk/k1 22 15 1
 k/k1  b
b  k1

353 271 2

PZ
k/k1
 Hk PX HT Rk  PA
k/k1 k k/k1
Rk 3

PAZ
k/k1
 PA HT  PA
k/k1 k k/k1

PBZ
k/k1
 PBA HT  PBA
k/k1 k k/k1
1 1
KAZ
k
 PA
k/k1
PZ
k/k1
 , KBZ
k
 PBA PZ
k/k1 k/k1
 114 58

qmeas  1
q meas
k k/k1 qk 16 12
Measurement
update kmeas  4qmeas
k
/1 q4 meas
k
, meas
k
 kmeas 5 4

ak  KAZ meas  k  KBZ meas

, b 18 9
k k k k
T
PA
k
 PA
k/k1
KAZ
k
PAZ
k/k1
 27 27
T
PBA
k
 PBA
k/k1
KBZ
k
PAZ
k/k1
 27 27
T
PBk  PBk/k1 KBZ PBZ 
k/k1 k/k1
27 27

 q  ak , q
 q / q

qk k/k1 q k k k 22 15 1

b  k/k1 b
k  b k 3

256 185 1
Mathematical Problems in Engineering 17

Table 5: Pseudocode and computational expense evaluation of the MGSPF algorithm.

Algorithm

x0  aT0 , b   0T , 1T , b

 T T  061 , q  0  031
0 0 31

0  a I3 , P0  a b I3 m 4
Initialize PA 2 BA

QA
 QB (
Sk1  T 6ARW T RRW /12I3 Sk1  RRW T/2I3
2 2 2

T
SA
k1
 PA k1
, SBA
k1
 PBA
k1
/SA
k1
 25 13 3
QA QB
k1  SA
k1
Sk1 UGS , Bk1  SBA
k1
Sk1 UGS 0 54
 k1  k1
meas  k1 ,
b  k1 |, k1 
 k1  |  k1 T 6 5 1
T
  8
q  T , 16 2  /16 2  6 2
k1 k1 k1 k1
i i
For i  1 : m, k1    k1 k1 T 3m 3m
i iT i2 T i2
qk1  8 k1 , 16 k1  /16 k1  6m 2m
i iT i2 T i2
qk1  8k1 , 16 k1  /16 k1  6m 2m
1
Propagation i
qk/k1   
q
i
qk1 qk1
i
32m 24m
k1
i i i
k/k1  4qk/k1 /1 q4 k/k1  end 5m 1m
Bk/k1  Bk1 , Zk/k1  k/k1

ak/k1  1/m k/k1

 k/k1  b
11m 1m , b  k1  031 3 3m 1

k/k1  k/k1 11m ak/k1 3m

T
PA
k/k1
 1/m k/k1  k/k1  6 6m 6m 1
T
PBA
k/k1
 1/mBk/k1  k/k1  9 9m 9m 1
   ,   
q k/k1  qk1 qk1 q k/k1  qk/k1 / qk/k1 22 15 1
 k/k1  b
b  k1

For m  4 351 285 5

k/k1  1/mZk/k1 1m1 3 3m 1
Zk/k1  Zk/k1 11m k/k1 3m
PZ
k/k1
 1/mZk/k1 Zk/k1 Rk T
6 6m 9m 1 3
PAZ
k/k1
 1/mak/k1 Zk/k1 T
9 9m 9m 1
PBZ
k/k1
 1/mBk/k1 Zk/k1 T 9 9m 9m 1
1 1
KAZ
k
 PA PZ
k/k1 k/k1
 , KBZk
 PBA
k/k1
PZk/k1
 114 58
Measurement meas 1 meas
qk  qk/k1 qk 16 12
update
meas
k  4qmeas
k
/1 q4 meas
k
, meas
k
 kmeas k/k1 5 4
ak  KAZ measb
,  k  KBZ meas 18 12
k k k k
T
PAk
 PA
k/k1
K AZ
k
P AZ
k/k1
 27 27
T
PBA
k
 PBA
k/k1
K BZ
k
PAZ
k/k1
 27 27
T
PBk  PBk/k1 KBZ PBZ 
k/k1 k/k1
27 27

 q  q / q

qk k/k1 q
ak , q k k k 22 15 1
k  b
b  k/k1 b
k 3
For m  4 379 278 1
18 Mathematical Problems in Engineering

ii Lower computation expense and better round-o error behavior. The new set is
free from calculating any irrational numbers. Furthermore, as it is only constituted of 1,
we can replace the multiplication operations in 3.12, 3.16 with simple sign changes. By
elimination of both irrational number and multiplication, we made 3.12, 3.16 precise, and
free from round-o errors.
iii The Geometric simplex set has a better symmetrical structure, which would
help to further increase the numerical accuracy, including a single dimension symmetry
completely fullled or in matrix language, each row of U is constituted with symmetrically
distributed elements. b interdimensional symmetry or per mutational symmetry partly
fullled. Dene the generator point as u  1, 1, 1T , and construct new points from u by
permutation and sign-changing, altogether we can make 8 points occupying the 8 vertices of
the unit cube in Figure 1. Note U has included 4 of them with a symmetric structure, which is
enough to capture the random states rst two order moments mean and variance. Actually,
the other 4 points can be found in U, and clearly, for an unbiased problem, choosing either
U or U is equivalent.

Numerical Demonstration
Suppose that we have already obtained a 3-dimensional unbiased state a, covariance P, and
its coresponding square-root matrix S. Then we are going to generate a set of sigma points
with a base set U as  S U. As had been claimed, theoretically we should have



a  Wm 1 m  0,
3.24

W T
P   SS , T

 
where a and P represent the result by numerical computation. Then we can take the norm of
 
the residues | a 0| and | diagP P| as a criterion of a sigma sets numerical accuracy. Three
typical base sets, namely, the Spherical Simplex set, the Schmidt Orthogonal set and the new
proposed Geometry Simplex set are compared over a series of dierent S matrices with their
diagonal elements ranked form 102 106 .
The numerical experiment is programmed with double precision oat numbers in
MATLAB, and some typical results are listed in Table 2. As can be seen, numerical behaviors
of the Geometric simplex set are quite encouraging. On mean computation, both spherical
and Schmidt sets introduced a residue error at the scale of about 1016 of the diagonal
elements of S, while the geometric sets computed residue had always been precisely 0. On
covariance computation, the new sets accuracy is also signicantly superior. Biased sets are
also studied and the result is similar.

3.3. The Highly Efcient New Filter

Incorporating the Geometric Simplex sigma point set into the Marginal SPKF framework, we
would have a new nonlinear SPKF estimator for attitude estimation, namely, the Marginal
Geometric Sigma Point Filter MGSPF algorithm, summarized in Table 5. As a sigma point
ltering algorithm, the MGSPF has a signicant increase in numerical eciency, while it still
guarantees a same order precision as the traditional SPKF algorithms. A general comparison
Mathematical Problems in Engineering 19

0
1
1
0 1
1

Figure 1: Geometry simplex sigma points in 3-dimensional Euclidean space.

of computation expense is taken between the MEKF and MGSPF algorithms as listed in Tables
4 and 5, and the result is summarized in Table 3.
For the computing eort of the propagation phase, there is little dierence between
MGSPF and MEKF, both are about half of the SSUKFs. For measurement update phase, the
MGSPF takes some more arithmetic operations, but still only 80% of the SSUFK. In fact, if
we take into account that in most actual implementations, there exist more propagation steps
than observation steps, the total computational expense of MEKF and MGSPF would be very
close. It is clear that the MGSPF has achieved a truly rivalizing eciency as the MEKF, even
when simple analytical closed-form solutions are included in the MEKF, and they are almost
50% of the SSUKF.

4. Simulations
In this section we apply the proposed Marginal Geometric Sigma Point Filter MGSPF
algorithm to the typical stellar-inertial spacecraft attitude determination system with
numerical simulations. To give a comparison, the multiple extended Kalman lter MEKF
and a nonaugmented spherical simplex unscented Kalman lter SSUKF with trapezoidal
approximation of the propagation noise are also simulated.
Parameters of the simulated model are set as follows. The spacecrafts initiation
attitude is q0  0.1, 0.15, 0.2, 1T , or expressed in 3-1-2 Euler Angles as 14 ,15 ,21 . The
initial angular velocity is 0  0.05 /s, 0.1 /s, 0.15 /sT , and it runs a sinusoidal maneuver
at an Amplitude of 0.5 /s and periods of 100 s, 120 s, and 125 s for each axis. The gyroscope
is modeled as initial bias 3.4 /s, drift instability also known as the icker noise 0.001/s;
angular random walk ARW 1 103 /s1/2 , rate random walk 1.4 103 /s3/2 , and sampling
20 Mathematical Problems in Engineering

100

101

Attitude error  
102

0.001  bond
103

104

0 30 60 90 120 150
Sensor observation times samples

MEKF
SSUKF
MGSPF

Figure 2: Estimation error history of the attitude.

100

101
Gyro bias error  /s

102

0.001  /s bond

103

104

0 30 60 90 120 150
Sensor observation times samples

MEKF
SSUKF
MGSPF

Figure 3: Estimation error history of the gyroscope bias.

frequency 20 Hz. The star sensor is simulated with 1 accuracy as cross boresight 10 arc-
seconds and around boresight 30 arc-seconds, and the sensors update-rate is 5 Hz.
Mathematical Problems in Engineering 21

100

101

Gyro bias error  /s

Increasing RRW
2
10

0.001  /s bond

103

104

0 30 60 90
Sensor observation times samples

Figure 4: Inuence of the bias noise parameter in MGSPF.

The initial states of all lters are set equivalently as x0  aT0 , b  
 T T  061 , q
0 0
T  0  031 . The initial covariance PX is set with the attitude-related elements
0T31 , 1 , and b 0
2
0 10  I3 , and bias-related elements
PA PB0  0.1 /s2 I3 . For MGSPF, as PB0 is no longer used,

we equivalently set a PBA 0  PA B
0 P0 . Specic elements in Rk and Qk are chosen through
tuning, set as r  2 107 , ARW  4.5 104 , and RRW  4 104 , respectively, for all three
lters.
The simulation results of attitude estimation error and and gyro bias estimation error
are, respectively, illustrated in Figures 2 and 3. As the star sensor has a high precision and
a 5 Hz Data Update Rate, the three lters steady-state accuracies are close to each other;
so we mainly focus on the initial stage of the estimation process. As shown in Figure 2, of
the attitude estimation error, to converge to a value below 0.001, MEKF takes more than 60
star observations, SSUKF takes about 40, while the MGSPF takes only about 20. Meanwhile,
as in Figure 3 of the gyro bias estimation, to achieve an estimation precision of 0.001/s,
MEKF takes more than 50 star samples, SSUKF takes 30, and the MGSPF takes about 20.
This indicates that the MGSPF algorithm, once properly implemented, provides a better
performance than MEKF at a similar numerical expense, while it is able to achieve, if not
better, at least a comparable performance to traditional sigma point lters, at a signicant
lower expense.
We now address the issue of tuning. The main dierence between MGSPF and
traditional sigma point lters in parameter selection is mainly reected in the usage of RRW ;
so we focus on the eect of dierent RRW on the performance of MGSPF. As illustrated in
Figure 4, a larger RRW has the advantage of enabling a faster convergence or can also be said
as an enhancement of the lters tracking ability. On the other hand, a large RRW also has a
drawback of instable parameter estimation in steady state; that is, the estimation of b will
jump. An optimized value of RRW would be a tradeo between the two. In addition, as
the sigma points are created from the covariance matrices which are added with the noise
terms arisen from RRW , the scale of RRW should be kept in a reasonable range that would
not obscure the information contained in the covariances.
22 Mathematical Problems in Engineering

5. Conclusion
A new, minimum sigma points algorithm for spacecraft attitude and angular rate estimation
has been developed. By marginalizing out the linear substructures within the random walk
gyro bias model and the attitude involving, only observation model, the new algorithm needs
only 4 sigma points to give a complete 6-state attitude and angular rate estimation. The
algorithms computational expense is only 50% of the traditional SSUKF algorithm. It has
truly rivaled the MEKF algorithms computing speed even when simple analytical closed-
form solutions are included. Yet it is still able to achieve the same accuracy as traditional
unscented Kalman lters.
A new, symmetrical, and numerically more ecient simplex sigma set has been
presented. The new set is completely free from irrational numbers and is free from any
multiplication operations during sigma point construction. The new set introduces almost
none of round-o error for mean reference and smaller error for covariance reference. It
would be of use for the implementation in a variety of 3-dimensional Euclidean space
involving dynamical problems such as positioning and attitude estimation problems.
With the remarkable reduction in computational expense, the sigma point Kalman
lter would gain a signicant upgrading in its competitiveness as a candidate algorithm for
actual onboard implementation.

References
1 E. J. Leerts, F. L. Markley, and M. D. Shuster, Kalman ltering for spacecraft attitude estimation,
Journal of Guidance, Control, and Dynamics, vol. 5, no. 5, pp. 417429, 1982.
2 S. J. Julier and J. K. Uhlmann, Unscented ltering and nonlinear estimation, Proceedings of the IEEE,
vol. 92, no. 3, pp. 401422, 2004.
3 R. D. van Merwe, E. A. Wan, and S. I. Julier, Sigma-point Kalman lters for nonlinear estimation and
sensor-fusionapplications to integrated navigation, in Proceedings of the Guidance, Navigation, and
Control Conference (AIAA 04), vol. 3, pp. 17351764, Providence, RI, USA, August 2004.
4 S. J. Julier, J. K. Uhlmann, and H. F. Durrant-Whyte, New approach for ltering nonlinear systems,
in Proceedings of the American Control Conference (ACC 95), vol. 3, pp. 16281632, Seattle, Wash, USA,
1995.
5 J. L. Crassidis and F. L. Markley, Unscented ltering for spacecraft attitude estimation, Journal of
Guidance, Control, and Dynamics, vol. 26, no. 4, pp. 536542, 2003.
6 K. Lai, J. Crassidis, and R. Harman, In-space spacecraft alignment calibration using the unscented
lter, in Proceedings of the Guidance, Navigation, and Control Conference and Exhibit (AIAA 03),
Honolulu, Hawaii, USA, August 2003.
7 J. Cote and J. de Lafontaine, Magnetic-only orbit and attitude estimation using the square-root
unscented Kalman lter: application to the PROBA-2 spacecraft, in Proceedings of the Guidance,
Navigation, and Control Conference and Exhibit (AIAA 08), Honolulu, Hawaii, USA, August 2008.
8 R. van der Merwe and E. Wan, The square-root unscented Kalman lter for state and parameter-
estimation, in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing
(ICASSP 01), vol. 6, pp. 34613464, Salt Lake City, Utah, USA, May 2001.
9 S. J. Julier and J. K. Uhlmann, Reduced sigma point lters for the propagation of means and
covariances through nonlinear transformations, in Proceedings of the American Control Conference
(ACC 02), vol. 2, pp. 887892, Anchorage, Alaska, USA, May 2002.
10 S. J. Julier, The spherical simplex unscented transformation, in Proceedings of the American Control
Conference (ACC 03), vol. 3, pp. 24302434, Denver, Colo, USA, June 2003.
11 J. G. Castrejon-Lozano, L. R. Garca Carrillo, A. Dzul, and R. Lozano, Spherical simplex sigma-point
Kalman lters: a comparison in the inertial navigation of a terrestrial vehicle, in Proceedings of the
American Control Conference (ACC 08), pp. 35363541, 2008.
Mathematical Problems in Engineering 23

12 J.-F. Levesque, Second-order simplex sigma points for nonlinear estimation, in Proceedings of the
Guidance, Navigation, and Control Conference (AIAA 06), vol. 2, pp. 819830, Keystone, Colo, USA,
August 2006.
13 S. J. Julier, The scaled unscented transformation, in Proceedings of the American Control Conference
(ACC 02), vol. 6, pp. 45554559, Anchorage, Alaska, USA, May 2002.
14 W.-C. Li, P. Wei, and X.-C. Xiao, A novel simplex unscented transform and lter, in Proceedings of
the International Symposium on Communications and Information Technologies (ISCIT 07), pp. 926931,
Sydney, Australia, 2007.
15 I. Arasaratnam and S. Haykin, Cubature Kalman lters, IEEE Transactions on Automatic Control, vol.
54, no. 6, pp. 12541269, 2009.
16 M. R. Morelande and B. Ristic, Smoothed state estimation for nonlinear Markovian switching
systems, IEEE Transactions on Aerospace and Electronic Systems, vol. 44, no. 4, pp. 13091325, 2008.
17 F. L. Markley, Attitude error representations for Kalman ltering, Journal of Guidance, Control, and
Dynamics, vol. 26, no. 2, pp. 311317, 2003.
18 P. Singla, A new attitude determination approach using split eld of view star camera, M.S. thesis, Texas
A&M University, College Station, Tex, USA, 2002.
19 R. L. Farrenkopf, Analytic steady-state accuracy solutions for two common spacecraft attitude
estimators, Jounal of Guidance and Control, vol. 1, no. 4, pp. 282284, 1978.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 242396, 18 pages
doi:10.1155/2009/242396

Research Article
Spin-Stabilized Spacecrafts: Analytical Attitude
Propagation Using Magnetic Torques

Roberta Veloso Garcia,1 Maria Ceclia F. P. S. Zanardi,2

and Helio Koiti Kuga1
1
Space Mechanic and Control Division, National Institute for Space Research (INPE), Sao Paulo,
Sao Jose dos Campos 12227-010, Brazil
2
Department of Mathematics, Group of Orbital Dynamics and Planetology, Sao Paulo State University
(UNESP), Sao Paulo, Guaratingueta 12516-410, Brazil

Correspondence should be addressed to Roberta Veloso Garcia, [email protected]

Received 28 July 2009; Accepted 4 November 2009

Recommended by Antonio Prado

An analytical approach for spin-stabilized satellites attitude propagation is presented, considering

the inuence of the residual magnetic torque and eddy currents torque. It is assumed two
approaches to examine the inuence of external torques acting during the motion of the satellite,
with the Earths magnetic eld described by the quadripole model. In the rst approach is included
only the residual magnetic torque in the motion equations, with the satellites in circular or elliptical
orbit. In the second approach only the eddy currents torque is analyzed, with the satellite in
circular orbit. The inclusion of these torques on the dynamic equations of spin stabilized satellites
yields the conditions to derive an analytical solution. The solutions show that residual torque does
not aect the spin velocity magnitude, contributing only for the precession and the drift of the
spacecrafts spin axis and the eddy currents torque causes an exponential decay of the angular
velocity magnitude. Numerical simulations performed with data of the Brazilian Satellites SCD1
and SCD2 show the period that analytical solution can be used to the attitude propagation, within
the dispersion range of the attitude determination system performance of Satellite Control Center
of Brazil National Research Institute.

Copyright q 2009 Roberta Veloso Garcia et al. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
This paper aims at analyzing the rotational motion dynamics of spin-stabilized Earths
articial satellites, through derivation of an analytical attitude prediction. Emphasis is placed
on modeling the torques steaming from residual magnetic and eddy currents perturbations,
as well as their inuences on the satellite angular velocity and space orientation. A spherical
coordinated system xed in the satellite is used to locate the spin axis of the satellite in relation
2 Mathematical Problems in Engineering

to the terrestrial equatorial system. The directions of the spin axis are specied by the right
ascension  and the declination  as represented in Figure 1. The magnetic residual torque
occurs due to the interaction between the Earth magnetic eld and the residual magnetic
moment along the spin axis of the satellite. The eddy currents torque appears due to the
interaction of such currents circulating along the satellite structure chassis and the Earths
magnetic eld.
The torque analysis is performed through the quadripole model for the Earths
magnetic eld and the satellite in circular and elliptical orbits. Essentially an analytical
averaging method is applied to determine the mean torque over an orbital period.
To compute the average components of both the residual magnetic and eddy current
torques in the satellite body frame reference system satellite system, an average time in
the fast varying orbit element, the mean anomaly, is utilized. This approach involves several
rotation matrices, which are dependent on the orbit elements, right ascension and declination
of the satellite spin axis, the magnetic colatitudes, and the longitude of ascending node of the
magnetic plane.
Unlike the eddy currents torques, it is observed that the residual magnetic torque does
not have component along the spin axis; however, it has nonzero components in satellite
body x-axis and y-axis. Afterwards, the inclusion of such torques on the rotational motion
dierential equations of spin-stabilized satellites yields the conditions to derive an analytical
solution 1. The theory is developed accounting also for orbit elements time variation,
not restricted to circular orbits, giving rise to some hundreds of curvature integrals solved
analytically.
In order to validate the analytical approach, the theory developed has been applied
for the spin-stabilized Brazilian Satellites SCD1 and SCD2, which are quite appropriated
for verication and comparison of the theory with the data generated and processed by the
Satellite Control Center SCC of Brazil National Research Institute INPE. The oblateness
of the orbital elements is taken into account.
The behaviors of right ascension, declination, and spin velocity of the spin axis with
the time are presented and the results show the agreement between the analytical solution
and the actual satellite behavior.

2. Geomagnetic Field
It is well known that the Earths magnetic eld can be obtained by the gradient of a scalar
potential V 2; it means that

B  V, 2.1

with the magnetic potential V given by

k  
 rT n 1 
n
 m m
V r  , ,  rT gn cos m hm
n sen m Pn , 2.2
n1
r m0

n are the Gaussian coecients, Pn  are

where rT is the Earths equatorial radius, gnm , hm m

the Legendre associated polynomial and r, , mean the geocentric distance, the local
colatitudes, and local longitude, respectively.
Mathematical Problems in Engineering 3


K j

Spin plane
k  s
90

J
Equatorial plane
i

I

Figure 1: Orientation of the spin axis   J,

s: equatorial system I,  K,
 satellite body frame reference system
  
i, j, k, right ascension , and declination  of the spin axis.

In terms of spherical coordinates, the geomagnetic eld can be expressed by 2,

B  Br r B  B ,
 2.3

with

V 1 V 1 V
Br  , B  , B  . 2.4
r r rsen

For the quadripole model, it is assumed that n equals 1 and 2 and m equals 0, 1 and 2 in 2.2.
After straightforward computations, the geomagnetic eld can be expressed by 3, 4

 3  4
rT  rT 
Br  2 f1 , 3 f2 , , 2.5
r r
 3  4
rT  rT 
B  f3 , f4 , , 2.6
r r
)  3  4  4 *
1 rT  rT  rT 
B  f5 , f6 , 2 f7 , , 2.7
sen r r r

where the functions fi , i  1, 2 , . . . , 7, are shown in 3 and depend on the Gaussian
coecients g22 , h11 , h12 , h22 .
In the Equator reference system, the geomagnetic eld is expressed by 2


BX  Br cos B sen cos B sen , 2.8

BY  Br cos B sen sen B cos , 2.9

BZ  Br sen B cos , 2.10

4 Mathematical Problems in Engineering

where and are the right ascension and declination of the satellite position vector,
respectively, which can be obtained in terms of the orbital elements; Br , B , and B are given
by 2.5, 2.6, and 2.7, respectively.
In a satellite reference system, in which the axis z is along the spin axis, the
geomagnetic eld is given by 4, 5


B  Bxi By j Bz k, 2.11

where

Bx  BX sen BY cos ,

By  BX sen cos BY sen sen BZ cos , 2.12

Bz  BX cos cos BY cos sen BZ sen ,

with BX , BY , and BZ given by 2.82.10.

3. Residual and Eddy Currents Torques

Magnetic residual torques result from the interaction between the spacecrafts residual
magnetic eld and the Earths magnetic elds. If

m is the magnetic moment of the spacecraft

and B is the geomagnetic eld, then the residual magnetic torques are given by 2

Nr 

m B. 3.1

For the spin-stabilized satellite, with appropriate nutation dampers, the magnetic moment is
mostly aligned along the spin axis and the residual torque can be expressed by 5

N r  Ms k B, 3.2

where Ms is the satellite magnetic moment along its spin axis and k is the unit vector along
the spin axis of the satellite.
By substituting the geomagnetic eld 2.11 in 3.1, the instantaneous residual torque
is expressed by


N r  Ms Byi Bx j . 3.3

On the other hand, the eddy currents torque is caused by the spacecraft spinning motion. If

W is the spacecrafts angular velocity vector and p is the Foucault parameter representing
the geometry and material of the satellite chassis 2, then this torque may be modeled by 2



Ni  pB B W . 3.4
Mathematical Problems in Engineering 5

For a spin-stabilized satellite, the spacecrafts angular velocity vector and the satellite
magnetic moment, along the z-axis and induced eddy currents torque, can be expressed by
5, 6

 
N i  pW Bx Bzi By Bz j By2 Bx2 k . 3.5

4. Mean Residual and Eddy Currents Torques

In order to obtain the mean residual and eddy currents torques, it is necessary to integrate



the instantaneous torques N r and N i , given in 3.3 and 3.5, over one orbital period T as

 ti T  ti T

1

1

N rm  N r dt, N im  N i dt, 4.1
T ti T ti

where t is the time ti the initial time, and T the orbital period. Changing the independent
variable to the fast varying true anomaly, the mean residual and eddy currents torque can be
obtained by 4

 i 2  i 2

1
r2
1
r2
N rm  N r d, N im  N i d, 4.2
T i h T i h

where i is the true anomaly at instant ti , r is the geocentric distance, and h is the specic
angular moment of orbit.
To evaluate the integrals of 4.2, we can use spherical trigonometry properties,
rotation matrix associated with the references systems, and the elliptic expansions of the true
anomaly in terms of the mean anomaly 7, including terms up to rst order in the eccentricity
e. Without losing generality, for the sake of simplication of the integrals, we consider the
initial time for integration equal to the instant that the satellite passes through perigee. After
extensive but simple algebraic developments, the mean residual and eddy currents torques
can be expressed by 3, 6

pW 
N rm  Nrxmi Nrym j, N im  Nixmi Niym j Nizm k , 4.3
2

with

Ms
Nrxm  A sen cos B sen sen C sen ,
2
4.4
Ms
Nrym   D sen E cos 
2

and Nixm , Niym , Nizm as well as the coecients A, B, C, D, and E are presented in the
appendix. It is important to observe that the mean components of these torques depend on the
attitude angles ,  and the orbital elements orbital major semi-axis: a, orbital eccentricity:
e, longitude of ascending node: , argument of perigee: , and orbital inclination: i.
6 Mathematical Problems in Engineering

5. The Rotational Motion Equations

The variations of the angular velocity, the declination, and the ascension right of the spin axis
for spin-stabilized articial satellites are given by Euler equations in spherical coordinates 5
as

1
W  Nz ,
Iz
1
 Ny , 5.1
Iz W
1
 Nx ,
Iz WCos

where Iz is the moment of inertia along the spin axis and Nx , Ny , Nz are the components of
the external torques in the satellite body frame reference system. By substituting Nrm , given
in 4.3, in 5.1, the equations of motion are

dW
 0, 5.2
dt
d Nrym
 , 5.3
dt Iz W
d Nrxm
 , 5.4
dt Iz W cos

where it is possible to observe that the residual torque does not aect the satellite angular
velocity because its z-axis component is zero.
By substituting Nim , given in 4.3, in 5.1, the equations of motion are

dW pW
 Nizm , 5.5
dt 2Iz
d pW
 Niym , 5.6
dt 2Iz
d p
 Nixm . 5.7
dt 2Iz cos

The dierential equations of 5.25.4 and 5.55.7 can be integrated assuming that the
orbital elements I, , w are held constant over one orbital period and that all other terms
on right-hand side of equations are equal to initial values.
Mathematical Problems in Engineering 7

6. Analysis of the Angular Velocity Magnitude

The variation of the angular velocity magnitude, given by 5.5, can be expressed as:

dW Nizm p
 k dt, with k  . 6.1
W 2 Iz

If the parameter k is considered constant for one orbital period, then the analytical solution
of 6.1 is

W  W0 ekt , 6.2

where W0 is the initial angular velocity. If the coecient k < 0 in 6.2, then the angular
velocity magnitude decays with an exponential prole.

7. Analysis of the Declination and Right Ascension of Spin Axis

For one orbit period, the analytical solutions of 5.3-5.4 and 5.6-5.7 for declination and
right ascension of spin axis, respectively, can simply be expressed as,

 k1 t 0 , 7.1

 k2 t 0 , 7.2

with:

i for the case where the residual magnetic torque is considered in the motion
equations,

Nrym
k1  ,
Iz W o
7.3
Nrym
k2  ,
Iz Wo cos o

ii for the case where the eddy currents torque is considered in the motion equations,

pNiym
k1  ,
2Iz
7.4
pNiym
k2  ,
2Iz cos o

where W0 , 0 , and 0 are the initial values for spin velocity, declination, and right ascension
of spin axis.
8 Mathematical Problems in Engineering

Table 1: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
elliptical orbit and under the inuence of the residual magnetic torque index QER for declination and
right ascension and SCD1 in degrees.

Day SCC QER SCC QER SCC QER SCC QER

22/08/93 282.70 282.7000 0.0000 79.64 79.6400 0.0000
23/08/93 282.67 282.7002 0.0302 79.35 79.6399 0.2899
24/08/93 283.50 282.6999 0.8001 79.22 79.6394 0.4194
25/08/93 283.01 282.7004 0.3096 78.95 79.6395 0.6895
26/08/93 282.43 282.7015 0.2715 78.70 79.6399 0.9399
27/08/93 281.76 282.7019 0.9419 78.48 79.6398 1.1598
28/08/93 281.01 282.7019 1.6919 78.27 79.6393 1.3693
29/08/93 280.18 282.7024 2.5224 78.08 79.6392 1.5592
30/08/93 279.29 282.7036 3.4136 77.91 79.6396 1.7296
31/08/93 278.34 282.7043 4.3643 77.78 79.6397 1.8597
01/09/93 277.36 282.7044 5.3444 77.67 79.6391 1.9691

The solutions presented in 7.1 and 7.2, for the spin velocity magnitude, declination
and right ascension of the spin axis, respectively, are valid for one orbital period. Thus, for
every orbital period, the orbital data must be updated, taking into account at least the main
inuences of the Earths oblateness. With this approach, the analytical theory will be close to
the real attitude behavior of the satellite.

8. Applications
The theory developed has been applied to the spin-stabilized Brazilian Satellites SCD1
and SCD2 for verication and comparison of the theory against data generated by the
Satellite Control Center SCC of INPE. Operationally, SCC attitude determination comprises
8, 9 sensors data preprocessing, preliminary attitude determination, and ne attitude
determination. The preprocessing is applied to each set of data of the attitude sensors that
collected every satellite that passes over the ground station. Afterwards, from the whole
preprocessed data, the preliminary attitude determination produces estimates to the spin
velocity vector from every satellite that passes over a given ground station. The ne attitude
determination takes one week a set of angular velocity vector and estimates dynamical
parameters angular velocity vector, residual magnetic moment, and Foucault parameter.
Those parameters are further used in the attitude propagation to predict the need of attitude
corrections. Over the test period, there are not attitude corrections. The numerical comparison
is shown considering the quadripole model for the geomagnetic eld and the results of the
circular and elliptical orbits. It is important to observe that, by analytical theory that included
the residual torque, the spin velocity is considered constant during 24 hours. In all numerical
simulations, the orbital elements are updated, taking into account the main inuences of the
Earths oblateness.

9. Results for SCD1 Satellite

The initial conditions of attitude had been taken on 22 of August of 1993 to the 00:00:00 GMT,
supplied by the INPEs Satellite Control Center SCC. Tables 1, 2, and 3 show the results
Mathematical Problems in Engineering 9

Table 2: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
circular orbit and under the inuence of the residual magnetic torque index QCR for declination and
right ascension and SCD1 in degrees.

Day SCC QCR SCC QCR SCC QCR SCC QCR

22/08/93 282.70 282.7000 0 79.64 79.6400 0
23/08/93 282.67 282.7216 0.0516 79.35 79.6251 0.2751
24/08/93 283.50 282.7151 0.7849 79.22 79.6172 0.3972
25/08/93 283.01 282.6737 0.3363 78.95 79.6182 0.6682
26/08/93 282.43 282.5877 0.1577 78.70 79.6303 0.9303
27/08/93 281.76 282.4526 0.6926 78.48 79.6552 1.1752
28/08/93 281.01 282.2631 1.2531 78.27 79.6940 1.4240
29/08/93 280.18 282.0158 1.83588 78.08 79.7473 1.6673
30/08/93 279.29 281.7091 2.4191 77.91 79.8140 1.9040
31/08/93 278.34 281.3439 3.0039 77.78 79.8962 2.1162
01/09/93 277.36 280.9240 3.5640 77.67 79.9892 2.3191

with the data from SCC and computed values by the present analytical theory, considering
the quadripole model for the geomagnetic eld and the satellite in circular and elliptical orbit,
under inuence of the residual and eddy currents torques.
The mean deviation errors for the right ascension and declination are shown in Table 4
for dierent time simulations. The behavior of the SCD1 attitude over 11 days is shown in
Figure 2. It is possible to note that mean error increases with the time simulation. For more
than 3 days, the mean error is bigger than the required dispersion range of SCC.
Over the 3 days of test period, better results are obtained for the satellite in circular
orbit with the residual torque. In this case, the dierence between theory and SCC data has
mean deviation error in right ascension of 0.2444 and 0.2241 for the declination. Both
are within the dispersion range of the attitude determination system performance of INPEs
Control Center.
In Table 5 is shown the computed results to spin velocity when the satellite is under
inuence of the eddy currents torque, and its behavior over 11 days is shown in Figure 3. The
mean error deviation for the spin velocity is shown in Table 6 for dierent time simulation.
For the test period of 3 days, the mean deviation error in spin velocity was of 0.0312 rpm and
is within the dispersion range of the attitude determination system performance of INPEs
Control Center.

10. Results for SCD2 Satellite

The initial conditions of attitude had been taken on 12 February 2002 at 00:00:00 GMT,
supplied by the SCC. In the same way for SCD1, Tables 7, 8, and 9 presented the results with
the data from SCC and computed values by circular and elliptical orbits with the satellite
under the inuence of the residual magnetic torque and eddy currents torque.
The mean deviation errors are shown in Table 10 for dierent time simulations. For
this satellite, there is no signicant dierence between the circular and elliptical orbits when
considering the residual magnetic torque. The behavior of the SCD2 attitude over 12 days is
shown in Figure 4.
10 Mathematical Problems in Engineering

Table 3: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
circular orbit and under the inuence of the eddy current torque index QCI for declination and right
ascension and SCD1 in degrees.

Day SCC QCI SCC QCI SCC QCI SCC QCI

22/08/93 282.70 282.7000 0.0000 79.64 79.6400 0.0000
23/08/93 282.67 282.6848 0.0148 79.35 79.6490 0.2990
24/08/93 283.50 282.6723 0.8277 79.22 79.6457 0.4257
25/08/93 283.01 282.6621 0.3479 78.95 79.6352 0.6852
26/08/93 282.43 282.6538 0.2238 78.70 79.6220 0.9220
27/08/93 281.76 282.6468 0.8868 78.48 79.6092 1.1292
28/08/93 281.01 282.6412 1.6312 78.27 79.6001 1.3301
29/08/93 280.18 282.6366 2.4566 78.08 79.5942 1.5142
30/08/93 279.29 282.6331 3.3431 77.91 79.5913 1.6813
31/08/93 278.34 282.6305 4.2905 77.78 79.5904 1.8104
01/09/93 277.36 282.6287 5.2687 77.67 79.5909 1.9209

Table 4: Mean deviations for dierent time simulations for declination and right ascension and SCD1 in
degrees.

Time Simulation days 11 8 3 2

SCC QER 1.5882 0.5435 0.2566 0.0151
SCC QCR 1.0779 0.3587 0.2444 0.0258
SCC QCI 1.5400 0.5047 0.2710 0.0074
SCC QER 1.0896 0.8034 0.2364 0.1449
SCC QCR 1.1707 0.8172 0.2241 0.1376
SCC QCI 1.0653 0.7882 0.2416 0.1495

Table 5: INPEs Satellite Control Center Data index SCC and computed results for spin velocity, with the
satellite in circular orbit and under the inuence of the eddy currents torque index QCI in rpm.

Day WSCC WQCI WSCC WQCI

22/08/93 86.2100 86.2100 0.0000
23/08/93 86.0400 86.3156 0.2756
24/08/93 85.8800 86.4985 0.6185
25/08/93 85.8000 86.7144 0.9144
26/08/93 85.7300 86.9439 1.2139
27/08/93 85.6600 87.1719 1.5119
28/08/93 85.5800 87.3631 1.7831
29/08/93 85.5100 87.5296 2.0196
30/08/93 85.4400 87.6657 2.2257
31/08/93 85.3700 87.7658 2.3958
01/09/ 93 85.3100 87.8426 2.5326

Table 6: Mean deviations for dierent time simulations for spin velocity and SCD1 in degrees.

Time Simulation days 11 8 3 2

WSCC WQCI rpm 0.1475 0.1091 0.0312 0.0144
Mathematical Problems in Engineering 11

284 80
Right ascension SCD1 deg
283

Declination SCD1 deg

79.5
282

281 79

280 78.5
279
78
278

277 77.5
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
days days

QER QCI QER QCI

QCR INPE QCR INPE
a b

Right ascension Declination

1
Mean error deviation SCD1 deg

Mean error deviation SCD1 deg 0

0
0.5
1

2 1

3
1.5
4
2
5

6 2.5
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
days days

QER QER
QCR QCR
QCI QCI

c d

Figure 2: Evolution of the declination  and right ascension  of satellite spin axis for SCD1 and its
mean deviation error.

Over the test period of the 12 days with the satellite in elliptical orbit and considering
the residual magnetic torque, the dierence between theory and SCC data has mean deviation
error in right ascension of 0.1266 and 0.1358 in the declination. Both torques are within the
dispersion range of the attitude determination system performance of INPEs Control Center,
and the solution can be used for more than 12 days.
In Table 11 the computed results to spin velocity are shown when the satellite is
under the inuence of the eddy currents torque. The mean deviation error for the spin
velocity is shown in Table 12 for dierent time simulation. For the test period, the mean
deviation error in spin velocity was of 0.0253 rpm and it is within the dispersion range of
the attitude determination system performance of INPEs Control Center. The behavior of
the spin velocity is shown in Figure 5.
12 Mathematical Problems in Engineering

Table 7: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
elliptical orbit and under the inuence of the residual magnetic torque index QER for declination and
right ascension and SCD2 in degrees.

Day SCC QER SCC QER SCC QER SCC QER

12/02/02 278.71 278.710000 0.0000 63.47 63.470000 0.0000
13/02/02 278.73 278.709999 0.0200 63.45 63.469998 0.0200
14/02/02 278.74 278.710000 0.0300 63.42 63.470002 0.0500
15/02/02 278.74 278.710000 0.0300 63.39 63.470005 0.0800
16/02/02 278.72 278.709999 0.0100 63.36 63.470002 0.1100
17/02/02 278.68 278.709999 0.0300 63.33 63.470000 0.1400
18/02/02 278.63 278.710000 0.0800 63.31 63.470003 0.1600
19/02/02 278.57 278.710001 0.1400 63.29 63.470006 0.1800
20/02/02 278.50 278.710000 0.2100 63.27 63.470004 0.2000
21/02/02 278.42 278.709999 0.2900 63.25 63.470000 0.2200
22/02/02 278.33 278.710000 0.3800 63.24 63.470002 0.2300
23/02/02 278.23 278.710002 0.4800 63.23 63.470006 0.2400

9.25

9.2
SCD1 spin velocity rpm

9.15

9.1

9.05

8.95

8.9
0 1 2 3 4 5 6 7 8 9 10
Time days

INPE spin velocity

Computed W

Figure 3: Evolution of the spin velocity W for SCD1.

11. Mean Pointing Deviation

For the tests, it is important to observe the deviation between the actual SCC supplied and
the analytically computed attitude, for each satellite. It can be computed by


 cos1 i ic j jc k kc , 11.1

 indicates the unity vectors computed by SCC and ic , jc , kc  indicates the unity
where i, j, k
vector computed by the presented theory.
Mathematical Problems in Engineering 13

Table 8: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
circular orbit and under the inuence of the residual magnetic torque index QCR for declination and
right ascension and SCD2 in degrees.

Day SCC QCR SCC QCR SCC QCR SCC QCR

12/02/02 278.71 278.710000 0 63.47 63.470000 0
13/02/02 278.73 278.7113 0.01870 63.45 63.4692 0.0192
14/02/02 278.74 278.7127 0.02733 63.42 63.4683 0.0482
15/02/02 278.74 278.7141 0.0259 63.39 63.4673 0.0773
16/02/02 278.72 278.7155 0.0045 63.36 63.4664 0.1064
17/02/02 278.68 278.7168 0.0368 63.33 63.4654 0.1354
18/02/02 278.63 278.7180 0.0880 63.31 63.4646 0.1546
19/02/02 278.57 278.7191 0.1491 63.29 63.4638 0.1738
20/02/02 278.50 278.7200 0.2200 63.27 63.4631 0.1931
21/02/02 278.42 278.7207 0.3007 63.25 63.4625 0.2125
22/02/02 278.33 278.7212 0.3913 63.24 63.4621 0.2221
23/02/02 278.23 278.7215 0.4916 63.23 63.4618 0.2318

Table 9: INPEs Satellite Control Center Data index SCC and computed results with the satellite in
circular orbit and under the inuence of the eddy currents torque index QCI for declination and right
ascension and SCD2 in degrees.

Day SCC QCI SCC QCI SCC QCI SCC QCI

12/02/02 278.71 278.7100 0.0000 63.47 63.4700 0.0000
13/02/02 278.73 278.7170 0.0130 63.45 63.4921 0.0421
14/02/02 278.74 278.7261 0.0139 63.42 63.5119 0.0919
15/02/02 278.74 278.7371 0.0029 63.39 63.5268 0.1368
16/02/02 278.72 278.7497 0.0296 63.36 63.5352 0.1752
17/02/02 278.68 278.7635 0.0835 63.33 63.5370 0.2070
18/02/02 278.63 278.7772 0.1472 63.31 63.5345 0.2245
19/02/02 278.57 278.7912 0.2212 63.29 63.5302 0.2402
20/02/02 278.50 278.8044 0.3043 63.27 63.5285 0.2585
21/02/02 278.42 278.8159 0.3959 63.25 63.5334 0.2834
22/02/02 278.33 278.8253 0.4953 63.24 63.5477 0.3077
23/02/02 278.23 278.8321 0.6021 63.23 63.5724 0.3423

Table 10: Mean deviations for dierent time simulation for declination and right ascension and SCD2 in
degrees.

Time Simulation days 12 8 5 2

SCC QER 0.1266 0.0200 0.0180 0.0100
SCC QCR 0.1334 0.0247 0.0153 0.0093
SCC QCI 0.1875 0.0565 0.0139 0.0065
SCC QER 0.1358 0.0925 0.0520 0.0099
SCC QCR 0.1312 0.0894 0.0502 0.0096
SCC QCI 0.1925 0.1397 0.1088 0.0210
14 Mathematical Problems in Engineering

Table 11: INPEs Satellite Control Center Data index SCC and computed results of spin velocity, with the
satellite in circular orbit and under the inuence of the eddy currents torque index QCI in rpm.

Day WSCC WQCI WSCC WQCI

12/02/02 34.4800 34.4800 0.0000
13/02/02 34.4200 34.4942 0.0742
14/02/02 34.3700 34.4572 0.0872
15/02/02 34.3100 34.3561 0.04617
16/02/02 34.2600 34.1831 0.0769
17/02/02 34.2000 33.9323 0.2678
18/02/02 34.1400 34.6059 0.4659
19/02/02 34.0800 34.2108 0.1308
20/02/02 34.0200 33.7703 0.2497
21/02/02 33.9600 33.3067 0.6533
22/02/02 33.9000 32.8493 1.0508
23/02/02 33.8300 32.4199 1.4101

Table 12: Mean deviations for dierent time simulations for spin velocity and SCD1 in degrees.

Time simulation days 12 8 5 2

WSCC WQCI 0.0253 0.0060 0.0027 0.0039

Figures 6 and 7 present the pointing deviations for the test period. The mean pointing
deviation for the SCD1 for dierent time simulations are presented in Table 13. Over the
test period of 11 days, the mean pointing deviation with the residual magnetic torque and
elliptical orbit was 1.1553 , circular orbit was 1.2003 , and eddy currents torque with circular
orbit was 1.1306 . The test period of SCD1 shows that the pointing deviation is higher than the
precision required for SCC. Therefore for SCD1, this analytical approach should be evaluated
by a time less than 11 days.
For SCD2, the mean pointing deviation considering the residual magnetic torque and
elliptical orbit was 0.1538, residual magnetic torque and circular orbit was 0.1507, and eddy
current torque was 0.2160. All the results for SCD2 are within the dispersion range of the
attitude determination system performance of INPEs Control Center of 0.5 .

12. Summary
In this paper an analytical approach was presented to the spin-stabilized satellite attitude
propagation taking into account the residual and eddy currents torque. The mean
components of these torques in the satellite body reference system have been obtained and
the theory shows that, unlike the eddy currents torque, there is no residual torque component
along the spin axis z-axis. Therefore this torque does not aect the spin velocity magnitude,
but it can cause a drift in the satellite spin axis.
The theory was applied to the spin-stabilized Brazilian satellites SCD1 and SCD2 in
order to validate the analytical approach, using quadripole model for geomagnetic eld and
the satellite in circular and elliptical orbits.
Mathematical Problems in Engineering 15

278.9 63.6
Right ascension SCD2 deg
278.8 63.55

Declination SCD2 deg

278.7 63.5
278.6 63.45
278.5 63.4
278.4 63.35
278.3 63.3
278.2 63.25
278.1 63.2
0 2 4 6 8 10 12 0 2 4 6 8 10 12
days days

QER QCI QER QCI

QCR INPE QCR INPE

a b
Right ascension Declination
0.1 0.05
Mean error deviation SCD2 deg

0 Mean error deviation SCD2 deg 0

0.1 0.05
0.2 0.1
0.3 0.15
0.4 0.2
0.5 0.25
0.6 0.3
0.7 0.35
0 2 4 6 8 10 12 0 2 4 6 8 10 12
days days

QER QER
QCR QCR
QCI QCI

c d

Figure 4: Evolution of the declination  and right ascension  of satellite spin axis for SCD2 and its
mean deviation error.

Table 13: Mean pointing deviation for SCD1.

Time simulation days 11 8 3 2

QER 1.1553 0.8226 0.2448 0.1450
QCR 1.2003 0.8288 0.2326 0.1376
QCI 1.1306 0.8071 0.2503 0.1495

The result of the 3 days of simulations of SCD1, considering the residual magnetic
torque, shows a good agreement between the analytical solution and the actual satellite
behavior. For more than 3 days, the pointing deviation is higher than the precision required
for SCC 0.5 .
16 Mathematical Problems in Engineering

3.7

3.65

SCD2 spin velocity rpm

3.6

3.55

3.5

3.45

3.4

3.35
0 2 4 6 8 10 12
Time days

INPE spin velocity

Computed W

Figure 5: Evolution of the spin velocity magnitude W for SCD2.

2.5
SCD1 pointing deviation degrees

1.5

0.5

0
0 1 2 3 4 5 6 7 8 9 10
days

QER
QCR
QCI

Figure 6: Pointing deviation evolution in degrees for SCD1.

For the satellite SCD2, over the test period of the 12 days, the dierence between
theory when considering the residual or eddy currents torque and SCC data is within the
dispersion range of the attitude determination system performance of INPEs Control Center.
Thus the procedure is useful for modeling the dynamics of spin-stabilized satellite
attitude perturbed by residual or eddy currents torques but the time simulation depends on
the precision required for satellite mission.
Mathematical Problems in Engineering 17

0.45

0.4

SCD2 pointing deviation degrees

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0
0 2 4 6 8 10 12
days

QER
QCR
QCI

Figure 7: Pointing deviation evolution in degrees for SCD2.

Appendix
The coecients of the mean components of the residual magnetic torques, given by 2.9, are
expressed by


7 
7 
7 
7 
7 
7
A aia aib , B  bia bib , C  cia cib ,
i1 i1 i1 i1 i1 i1
A.1

7 
7 
7 
7
D aia aib , E bia bib ,
i1 i1 i1 i1

where aib , bib , cjb , i  1, 2, . . . , 7; j  1, . . . , 4, can be got by Garcia in 3. It is important to note
that the parcel bib is associated with the quadripole model and the satellite in an elliptical
orbit. For circular, orbit, bib is zero.
The mean components Nixm , Niym , Nizm of the eddy currents torque are expressed by


14 712 
18 426
Nixm  tr xi Nxi,
i1 i1


14 712 
53 765
Niym  tr yi Nyi, A.3
i1 i1

7
350 
21 435
Nizm  tr zi Nzi,
i1 i1

where tr xi, tr yi, tr zi, Nxi, Nyi, and Nzi are presented by Pereira 6.
18 Mathematical Problems in Engineering

The terms aib , bib , cjb , tr xi, tr yi, tr zi, Nxi, Nyi, and Nzi depend on orbital
elements a, e, I, , w and attitude angles , .

Acknowledgment
This present work was supported by CNPq National Counsel of Technological and Scientic
Development.

References
1 L. C. Thomas and J. O. Cappelari, Attitude determination and prediction of spin-stabilized satellites,
The Bell System Technical Journal, pp. 16561726, 1964.
2 J. R. Wertz, Spacecraft Attitude Determination and Control, D. Reidel, Dordrecht, The Netherlands, 1978.
3 R. V. Garcia, Satelites estabilizados por rotacao e torque magnetico residual, Dissertacao de Mestrado,
Faculdade de Engenharia, UNESP, Guaratingueta, Brazil, 2007.
4 M. C. Zanardi, Dinamica da atitude de satelites articiais, Tese de Livre Docencia, Universidade
Estadual Paulista, Guaratingueta, Brazil, 2005.
5 H. K. Kuga, W. C. C. Silva, and U. T. V. Guedes, Dynamics of attitude for spin stabilized satellites,
Tech. Rep. INPE-4403, INPE, Sao Jose dos Campos, Brazil, 1987.
6 A. J. Pereira, Propagacao da atitude de satelites estabilizados por totacao com torque induzido,
Trabalho de Conclusao de Curso, Faculdade de Engenharia, UNESP, Guaratingueta, Brazil, 2006.
7 D. Brouwer and G. M. Clemence, Methods of Celestial Mechanics, Academic Press, New York, NY, USA,
1961.
8 V. Orlando, H. K. Kuga, and U. T. V. Guedes, Flight dynamics leop and routine operations for SCD2,
the INPEs second environmental data collecting satellite, Advances in the Astronautical Sciences, vol.
100, no. 2, pp. 10031013, 1998.
9 H. K. Kuga, V. Orlando, and R. V. F. Lopes, Flight dynamics operations during LEOP for the INPEs
second environmental data collection satellite SCD2, Journal of the Brazilian Society of Mechanical
Sciences and Engineering, vol. 21, pp. 339344, 1999.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 173145, 9 pages
doi:10.1155/2009/173145

Research Article
Using of H-Innity Control Method in
Attitude Control System of Rigid-Flexible Satellite

Ximena Celia Mendez Cubillos and Luiz Carlos Gadelha de Souza

Space Mechanics and Control Division (DMC), National Institute for Space Research (INPE),
Avenida dos Astronautas 1758, P.O. Box 515, Sao Paulo, 12201-940 Sao Jose dos Campos, Brazil

Correspondence should be addressed to Luiz Carlos Gadelha de Souza, [email protected]

Received 1 July 2009; Revised 7 October 2009; Accepted 18 November 2009

Recommended by Tadashi Yokoyama

The attitude control systems of satellites with rigid and exible components are demanding more
and more better performance resulting in the development of several methods control. For that
reason, control design methods presently available, including parameters and states estimation,
robust and adaptive control, as well as linear and nonlinear theory, need more investigation to
know their capability and limitations. In this paper the investigated technique is H-Innity method
in the performance of the Attitude Control System of a Rigid-Flexible Satellite.

Copyright q 2009 X. C. Mendez Cubillos and L. C. G. de Souza. This is an open access article
distributed under the Creative Commons Attribution License, which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction
The rapidly complexity increase of systems and processes to be controlled has stimulated the
development of sophisticated analysis and design methods called advanced techniques. The
H-Innity H control theory, introduced by Zames 1, is one of the advanced techniques
and its application in several problems of control has been growing rapidly.
The employment of exible structures in the spatial area is another problem of control
system which has been growing up too. Flexible systems oer several advantages compared
with the rigid system. Some advantages are relatively smaller actuators, lower overall mass,
faster response, lower energy consumption, in general, and lower cost. With the study of the
Attitude Control System ACS of space structures with exible antennas and/or panel and
robotic manipulators, one becomes more complex when the dimensions of such structures
increase due to necessity to consider a bigger number of vibration modes in its model in
order to improve the model delity 2. Examples of projects that involve exible space
structures are the International Space Station ISS, the Lunar Reconnaissance Orbiter LRO,
the Lunar Crater Observation and Sensing Satellite LCROSS, the Hubble Space Telescope,
and so forth.
2 Mathematical Problems in Engineering

In Rigid-Flexible Satellite RFS the function of the ACS is to stabilize and orient
the satellite during its mission, counteracting external disturbances torques and forces. In
this paper is investigated multivariable control method H for attitude control of an RFS
consisting of a rigid body and two exible panels. The satellite modeling was built following
the Lagrangian approach and the discretization was done using the assumed-modes method.
The equations of motion obtained were written in its modal state space form.

2. The Rigid-Flexible Satellite Model

Figure 1 shows the picture of the satellite used in this work, which is composed of a rigid
body of cubic shape and two exible panels. The center of mass of the satellite is in the point
0 origin of the system of coordinates X, Y, Z that coincides with its main axis of inertia. The
elastic appendixes with the beam format are connected in the central body, being treated as
a punctual mass in its free extremity. The length of the panel is represented by L, the mass is
represented by m, and vx, t is the elastic displacement in relation to the axis Z. The moment
of inertia of the rigid body of the satellite in relation to the mass center is J0 . The moment of
inertia of the panel in relation to its own mass center is given by Jp .

3. Equations of Motion
In the Lagrang approach are considered the equation of motion of the satellite around in Y
and the elastic displacement of the panels. The Lagrange equations 3 for the problem can
be written in the following form:

 
d L L
 , 3.1
dt
 
d L L M
 0. 3.2
dt qi qi qi

In 3.1 is the torque of the reaction wheel, L  T V is the Lagrangian, and is the
rotation angle of the satellite around the axis Y. In 3.2 M is the dissipation energy associated
to the deformation of the panel qi it represents each one of the generalized coordinates of the
problem.
The beam deection variable vx, t is discretized using the expansion


n
vx, t  i xqi t 0 x L, 3.3
i1

where n represents the number of manners to be adopted in the discretization and i x
represents each one of the own modes of the system. The admissible functions i x are given
by 4

i x  coshai x cosai x i sinhai x sinai x, 3.4

Mathematical Problems in Engineering 3

Jp

J0

x
L

Figure 1: Satellite Model.

where

coshai L cosai L
i  , 3.5
sinhai L sinai L

and ai L are the eigenvalues of the free system and undamped.

For the complete system, the total kinetic energy T is given by T  TSatellite TPanel ;
therefore,
  L 
1  2  2  2 
T  J0 A
2
vx, t 2vx, tx x vx, t dx , 3.6
2 0

where is the density of the panels and A is the area. The dissipation energy function is

M  vx, t2 Kd , 3.7

where Kd is the dissipation constant. So L  T V is given by

  L 
1 
 2  2 
L  J0 A
2
vx, t 2vx, tx x vx, t dx vx, t2 K.
2
2 0
3.8

In 3.8 K is constant elastic of the panels. After some manipulations of 3.8 5 and
using the orthogonalization property of vibration modes of the beam 6, one has
L L
i j dx  1 if i  j, i j dx  0  j.
if i / 3.9
0 0
4 Mathematical Problems in Engineering

Finally, two equations are obtained. These equations represent the dynamics of
rotation motion of the satellite and the elastic displacement of the panels, respectively,
 

n 
n
1
1 a qi2 i a qi  . 3.10
i1 i1
J1

qi i 2 qi d qi c qi  q . 3.11

where the term nonlinear i in 3.10 is dened as centripetal rigidity, and in 3.11 q is a
Piezoelectric actuator adapted for the following simulations where will be considered i  1
one mode, and the constants are given by

2A K Kd
a , J1  J0 2Jp , c , d . 3.12
J1 A A

4. H-Innity Control Method

4.1. Introduction
Throughout the decades of 1980 and 1990, H-Innity control method had a signicant impact
in the development of control systems; nowadays the technique has become fully grown
and it is applied on industrial problems 5. In the control theory in order to achieve
robust performance or stabilization, the H-Innity control method is used. The control
designer expresses the control problem as a mathematical optimization problem nding the
controller solution. H techniques have the advantage over classical control techniques in
which the techniques are readily applicable to problems involving multivariable systems
with cross-coupling between channels; disadvantages of H techniques include the high
level of mathematical understanding needed to apply them successfully and the need for a
reasonably good model of the system to be controlled. The problem formulation is important,
since any synthesized controller will be optimal in the formulated sense.
The H name derives from the fact that mathematically the problem may be set in the
space H, which consists of all bounded functions that are analytic in the right-half complex
plane. We do not go to this length. The H norm is the maximum singular value of the
function; let us say that it can be interpreted as a maximum gain in any direction and at
any frequency; for SISO Single In, Single Out systems, this is eectively the maximum
magnitude of the frequency response. H method is also used to minimize the closed-
loop impact of a perturbation: depending on the problem formulation, the impact will be
measured in terms of either stabilization or performance. Thus, one concludes that the
procedures to project control systems are a dicult task due to the cited terms which are
conicting properties 7.

4.2. Modeling
This problem is dened by the conguration of Figure 2. The plant is a given system with
two inputs and two outputs. It is often referred to as the generalized plant. The signal w
is an external input and represents driving signals that generate disturbances, measurement
noise, and reference inputs. The signal u is the control input. The output z has the meaning
Mathematical Problems in Engineering 5

Plant
w z
G
u y

Compensador

Figure 2: Generalized Plant.

of control error and ideally should be zero. The output y, nally, is the observed output and
is available for feedback.
The project of control system is based given by

xt  Ax Bu,
4.1
yt  Cx Du.

A more general state space representation of the standard plant is

xt  Axt B1 wt B2 Wt,

zt  C1 xt D11 wt D12 ut,

yt  C2 xt D21 wt D22 ut,

4.2

xt A B1 B2 xt

zt  C1 D11 D12 .wt  P.

yt C2 D21 D22 ut

The solution of the corresponding H problem based on Riccati equations is implemented

requires the following conditions to be satised 8:

1 A, B2  is stabilizable and C2 , A is detectable,

2 D12 and D21 have full rank,

3 A-jwI B2 ; C1 D12  has full column rank for all w  R hence, D12 is tall,

4 A-jwI B1 ; C2 D21  has full column rank for all w  R hence, D21 is wide.

The augmented plant is formed by accounting for the weighting functions W1 , W2 , and W3 ,
as shown in Figure 3. In order to reach the acting objectives, the outputs were chosen to be
transfer weight functions: z1  W1 ; z2  W2 y; z3  W3 u.
6 Mathematical Problems in Engineering

z
z1 z2 z3

W1 W2 W3

w
K G y
e u

Figure 3: Plant with weighting functions for H design.

Table 1: Parameters.
Parameter Description Value
J0 Moments of inertia of the rigid body of the satellite 720 Kgm2
Jp Moment of inertia of the panel 40 Kgm2
K Constant elastic of the panels 320 Kgrad2 /s2
Kd Dissipation constant 0,48 Kgrad2 /s
L Length of the panel 2m
m Mass of the satellite 200 kg

The function cost of mixed sensibility is given for

S  I GK1 ,
W1 S

Ty1u1 
W2 R,
R  KI GK1 , 4.3
W3 T
T  GKI GK1 ,

where S is called sensibility, T is complementary sensitivity function, and R does not have
any name. The cost function of mixed sensibility is named alike, because it punishes S, R, and
T at the same time; it can also be said project requirement. The transfer function from w to
z1 is the weighted sensitivity function W1 S, which characterizes the performance objective of
good tracking; the transfer function from w to z2 is the complementary sensitivity function T,
whose minimization ensures low control gains at high frequencies, and the transfer function
from w to z3 is KS, which measures the control eort. It is also used to impose the constraints
on the control input for example, the saturation limits.

5. Simulations
The simulations were carried out by computational implementation of the software MatLab.
The initials conditions used here are  0.001 rad. and  0 rad./s. The values considered
for the physical parameters in the numerical simulation are presented in Table 1.
The procedure of the project of H is dierent from other control projects knowledge
such as LQR Linear Quadratic Regulator and LQG Linear Quadratic Gaussian 5; the
Mathematical Problems in Engineering 7

104 H angle time

0.4
0.3

rad.
0.2
0.1
0
0.1
0 1 2 3 4 5
s
a

104 Angular speed time

1
0
rad./s

1
2
3
4
0 1 2 3 4 5
s
b

Figure 4: Angle and Angular Speed.

dierence is the use of weighting functions W1 , W2 , and W3 , where W2  0 and the others are
given by

0.11 s/1002
W11  1 I2x2 ,
1 s/50002
5.1
2000
W31  I2x2 ,
s

where W1 punishes the error sign e, W2 punish the control sign u, and W3 punishes the exit
of the plant y; is a parameter obtained through successive attempts.

6. Results
First we analyze the open loop of the system through transmission zeros TZs and the close-
loop with H Control. The TZs are critical frequencies where signal transmission between
input and output is stopped. The importance of use of the TZs is given by their application in
robust control, because they are the zeros of a MIMO system. In Table 2 are represented the
values.
Following, the performances of H control in the ACS are observed in Figures 4 and 5.
Both graphs, in Figure 4, have existence of overshoot, in which they could commit the
system; however, the time of stabilization of both was of approximately 3.5 seconds. In other
words, in spite of the existence of overshoots, the control of the system, in a long time, was
reached.
8 Mathematical Problems in Engineering

107 H-elastic displacement time

5

Abs
5

10

15
0 0.1 0.2 0.3 0.4 0.5
s
a
6
10 H-variation elastic displacement time
15

10
Abs

5
0 0.1 0.2 0.3 0.4 0.5
s
b

Figure 5: Vibration of the Panels.

Table 2: Transmission Zeros.

Transmission Zeros
Open Loop 0.0244 5.7049i
2318.10
433.3
H Control
0.0 05.700i
0 0i

In Figure 5 the behavior of the vibration of the panels is presented. The displacement
of overshoot is of the order of 107 , in other words, very small. The time of stabilization
in the rst graph is about 0.5 seconds and for the second one is about 0.45 seconds. This
demonstrates that the control H possesses a good performance for angle and angular
velocity, as well as to control the vibration of the panels.

7. Conclusions
The problem of attitude control of satellites is not new and has been addressed by several
researchers using many dierent approaches. The H control method is one of the most
advanced techniques available today for designing robust controllers. One great advantage
with this technique is that it allows the designer to tackle the most general form of control
architecture wherein explicit accounting of uncertainties, disturbances, actuator/sensor
noises, actuator constraints, and performance measures can be accomplished. The system is
very dierent from the methods LQR and LQG, for example. However, a great disadvantage
is the experience and necessary abilities to design the form of the weighting functions and
Mathematical Problems in Engineering 9

the fact that the plant can increase. Basically, the success of the method depends on the correct
choice of the weight functions transfer.

Acknowledgments
The authors would like to thank CAPES and INPE/DMC. This work was supported by
CAPES thought the BrasilPortugal Cooperation Project PCT no. 241/09.

References
1 G. Zames, Feedback and optimal sensitivity: model reference transformations, multiplicative
seminorms, and approximate inverses, IEEE Transactions on Automatic Control, vol. 26, no. 2, pp. 301
320, 1981.
2 L. C. G. De Souza and S. A. Silva, Vibration control of a rigid-exible satellite during attitude
maneuver, in Proceeding of the 17th ASME Biennial Conference on Mechanical Vibration and Noise
(DETC 99), Las Vegas, Nev, USA, September 1999.
3 L. Meirovitch, Methods of Analytical Dynamics, Dover, Mineola, NY, USA, 1998.
4 J. L. Junkins and Y. Kim, Introduction to Dynamics and Control of Flexible Structures, AIAA, Washington,
DC, USA, 1993.
5 X. C. M. Cubillos, Investigation of Multivariable Control Techniques for Attitude Control of a RigidFlexible
Satellite, INPE, Sao Jose dos Campos, Brazil, 2008.
6 C. H. G. Hassmann, and A. Fenili, Attitude and vibration control of a satellite with a exible solar
panel using Lqr tracking with innite time, in Proceedings of the ASME International Congress of
Mechanical Engineering, Seattle, Wash, USA, November 2007.
7 M. G. Safonov, A. J. Laub, and G. L. Hartmann, Feedback properties of multivariable systems: the
role and use of the return dierence matrix, IEEE Transactions on Automatic Control, vol. 26, no. 1, pp.
4765, 1981.
8 O. H. Bosgra and H. Kwakernaak, Design methods for control systems, Notes for a Course of the
Dutch Institute of Systems and Control, Winter term 2001-2002.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 753653, 18 pages
doi:10.1155/2009/753653

Research Article
Hill Problem Analytical Theory to the Order Four:
Application to the Computation of Frozen Orbits
around Planetary Satellites

Martin Lara1 and Jesus F. Palacian2

1
Celestial Mechanics Division, Real Observatorio de la Armada, 11110 San Fernando, Spain
2
Departamento de Ingeniera Matematica e Informatica, Universidad Publica de Navarra,
31006 Pamplona, Spain

Correspondence should be addressed to Martin Lara, [email protected]

Received 7 June 2009; Accepted 24 August 2009

Recommended by Silvia Maria Giuliatti Winter

Frozen orbits of the Hill problem are determined in the double-averaged problem, where short and
long-period terms are removed by means of Lie transforms. Due to the perturbation method we
use, the initial conditions of corresponding quasi-periodic solutions in the nonaveraged problem
are computed straightforwardly. Moreover, the method provides the explicit equations of the
transformation that connects the averaged and nonaveraged models. A fourth-order analytical
theory is necessary for the accurate computation of quasi-periodic frozen orbits.

Copyright q 2009 M. Lara and J. F. Palacian. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
Besides its original application to the motion of the Moon 1, the Hill problem provides
a good approximation to the real dynamics of a variety of systems, encompassing the
motion of comets, natural and articial satellites, distant moons of asteroids, or dynamical
astronomy applications 24. Specically, the Hill model and its variations 59 are useful
for describing the motion about planetary satellites. In addition, the Hill problem is an
invariant model that does not depend on any parameter, thus, giving broad generality to the
results, whose application to dierent systems becomes a simple matter of scaling. Note that
Hills case of orbits close to the smaller primary is a simplication of the restricted three-body
problem, which in turn is a simplication of real models.
A classical result shows that low eccentricity orbits around a primary body are
unstable for moderate and high inclinations due to third-body perturbations 10. Almost
circular orbits close to the central body remain with low eccentricity in the long-term only
when the mutual inclination with the perturbing body is less than the critical inclination of
the third-body perturbations I  39.2 see 11 and references therein. Because of their
2 Mathematical Problems in Engineering

low eccentricity, high inclination orbits are precisely the candidate orbits for science missions
around natural satellites. Therefore, a good understanding of the unstable dynamics of the
Hill problem is required.
The study of the long-term dynamics is usually done in the double-averaged problem.
After removing the short- and long-period terms, and truncating higher-order terms, the
problem is reduced to one degree of freedom in the eccentricity and the argument of the
periapsis. As the double-averaged problem is integrable and the corresponding phase space
is a compact manifold, the solutions are closed curves and equilibria. The latter are orbits
that, on average, have almost constant eccentricity and xed argument of the periapsis, and
are known as frozen orbits.
To each trajectory of the double-averaged problem it corresponds a torus of
quasiperiodic solutions in the nonaveraged problem. The accurate computation of initial
conditions on the torus requires the recovery of the short- and long-period eects that
were eliminated in the averaging. This is normally done by trial and error, making iterative
corrections on the orbital elements, although other procedures can be applied 12.
Our analytical theory is computed with Deprits perturbation technique 13. The
procedure is systematic and has the advantage of providing the explicit transformation
equations that connect the averaged analysis with proper initial conditions of the
nonaveraged problem. A second-order truncation of the Hamiltonian shows that there are
no degenerate equilibria and, therefore, it is sucient to give the qualitative description of
the reduced system. However, the second-order truncation introduces a symmetry between
the direct and retrograde orbits that is not part of the original problem, and a third-order
truncation is required to reveal the nonsymmetries of the problem.
While, in general, the third-order theory provides good results in the computation
of quasiperiodic, frozen orbits, its solutions are slightly aected by long-period oscillations.
This fact may adversely aect the long-term evolution of the frozen orbits and it becomes
apparent in the computation of science orbits about planetary satellites, a case in which small
perturbations are enough for the unstable dynamics to defrost the argument of the periapsis.
Then, the orbit immediately migrates along the unstable manifold with an exponential
increase in the eccentricity.
We nd that a higher-order truncation is desirable if one wants to use the analytical
theory for computing accurate initial conditions of frozen orbits. The computation of the
fourth-order truncation removes almost all adverse eects from the quasiperiodic solutions,
and shows a high degree of agreement between the averaged and nonaveraged models even
in the case of unstable orbits.
Whereas the third-body perturbation is the most important eect in destabilizing
science orbits around planetary satellites, the impact of the nonsphericity of the central body
may be taken into account. The previous research including both eects has been limited
up to third-order theories see 14 and references therein, but from the conclusions of this
paper it may worth to develop a higher-order theory including the inhomogeneities of the
satellites gravitational potential.

2. Double-Averaged Hill Problem to the Fourth-Order

The equations of motion of the Hill problem are derived from the Hamiltonian

   
1 2 
H X X x X Wx, W r 2 3x2 , 2.1
2 2 r
Mathematical Problems in Engineering 3

where, in the standard coordinate system of Hills model, x  x, y, z is the position vector,
X  X, Y, Z is the vector of conjugate momenta, r  ||x||, and both the rotation rate of the
system  |||| and the gravitational parameter of the primary are set to 1 in appropriate
units.
The problem is of three degrees of freedom, yet admitting the Jacobi constant
H  C/2. Despite its nonintegrability, approximate solutions that explain the long-term
dynamics can be found by perturbation methods. Close to the central body the Hill problem
can be written as the perturbed two-body problem
   
1 1  2  2
H X Y Z
2 2 2
 x Y y X r 3x2 , 2.2
2 r 2

where the rst three terms of Hamiltonian 2.2 correspond to the Keplerian motion in the
rotating frame and  is a formal parameter introduced to manifest the importance of each
eect. Thus, the Coriolis term is a rst order eect and the third-body perturbation appears
at the second-order .
To apply perturbation theory, we formulate the problem in Delaunay variables
, g, h, L, G, H, where  is the mean anomaly, g is the argument of the periapsis,
h the

argument of the node in the rotating frame, L  a is the Delaunay action, G  L 1 e2 is
the modulus of the angular momentum vector, H  G cos I is its polar component, and a, e, I,
are usual orbital elements: semimajor axis, eccentricity, and inclination.
Our theory is based on the use of Lie transforms as described by Deprit 13, 15.
It has the advantage of connecting the averaged and original problems through explicit
transformation equations. After removing the short- and long-period terms we get the
transformed Hamiltonian
     
2 3 4
K  K0,0 K0,1 K0,2 K0,3 K0,4 , 2.3
2 6 24

where  L3 ,

1
K0,0  ,
2L2 
K0,1  K0,0 2,
   
1
K0,2  K0,0 2 3e2 2 3s2 15e2 s2 cos 2g ,
4
    
27
K0,3  K0,0 2s2 50 17s2 e2 15e2 s2 cos 2g ,
32
 
3    
K0,4  K0,0 3285s4 e4 cos 4g 12s2 3996 2940s2 4582 4035s2 e2 e2 cos 2g
512
 
8 784 708s2 9s4 144 926 941s2 244s4 e2
 
9 10728 15208s2 5007s4 e4 ,
2.4
4 Mathematical Problems in Engineering

and  H/L  c,  1 e2 is the eccentricity function, and we use the abbreviations
s sin I, c cos I. Details on the perturbation method and expressions to compute the
transformation equations of the averaging are given in the appendix.
The double-averaged Hamiltonian 2.3 depends neither on the mean anomaly nor
on the argument of the node. Therefore, the corresponding conjugate momenta, L and H,
are integrals of the reduced problem and the Hamiltonian 2.3 represents a one degree
of freedom problem in g and G. The equations of motion are computed from Hamilton
equations dg/dt  K/G, dG/dt  K/g,

dg 6 2   27   
 5c 2 5 c2 2 cos 2g 5c2 112 5 c2 2 cos 2g
dt G 4G
32   
2113c2 3285c4 3915 9165c4 2 1581 7791c2 4
128G
   
4 802c2 1095c4 19 2565c4 2 547 1744c2 4 cos 2g
  2.5
1095e2 s2 c2 2 cos 4g ,
)
dG 3 2 2
 e s 58 9 sin 2g
dt 4
*
2    
2 509 1095c 547 4035c sin 2g 1095e s sin 4g .
2 2 2 2 2
32

Once g and G are integrated for given initial conditions, the secular variations of  and h
are computed from simple quadratures derived from Hamilton equations dh/dt  K/H,
d/dt  K/L,

 
 
h  h0 K gt, Gt; H, L dt,   0 K gt, Gt; H, L dt. 2.6
H L

3. Qualitative Dynamics
The ow can be integrated from the dierential equations mentioned previously, 2.5.
However, since the system dened by 2.5 is integrable, the ow is made of closed curves
and equilibria, and it can be represented by contour plots of Hamiltonian 2.3. Thus, for
given values of the dynamical parameters L and Hor and we can plot the ow
in dierent maps that are function of g, G. Figure 1 shows an example in semiequinoctial
elements e cos g, e sin g, where we note a hyperbolic point corresponding to an unstable
circular orbit, and two elliptic points corresponding to two stable elliptic orbits with e  0.2
and periapsis at g  /2, respectively.
Delaunay variables are singular for zero eccentricity orbits, where the argument of
the periapsis and the mean anomaly are not dened, and for equatorial orbits, where the
Mathematical Problems in Engineering 5

0.6

0.4

0.2

e sin g
0

0.2

0.4

0.6

0.6 0.4 0.2 0 0.2 0.4 0.6

e cos g

Figure 1: Flow in the doubly reduced phase space.

argument of the node is not dened. Hence, it is common to study the reduced phase space
in the variables introduced by Coey et al. 16, see also 8:

1 
1  es cos g, 2  es sin g, 3  2 1 2 , 3.1
2

that dene the surface of a sphere

1  2
21 22 23  1 2 3.2
4

of radius R  1/2 1 2  the sphere representation misses the case G  H  0, irrelevant

in astrodynamics.
Then, after dropping constant terms and scaling, Hamiltonian 2.3 writes

 
3022 9 22
K  12 2 25 24 15 2
2 2 2
4

2 
3815 9528 2 9 4 6 343 1709 2 2 18244 3.3
64

 2 42
2
6 293 821 1470
2 2
3285 4 .
2
6 Mathematical Problems in Engineering

The ow on the sphere is obtained from Liouville equations i  {i ; K}, i  1, 2, 3, where the
dot means derivative in the new time scale. Then,
)
3  
1  2 64 5 82 5 2 72 5 22 5 2
16

2 
3815 18244 9528 2 9 4 62 343 1709 2 3.4
64

  2 2  4 *
2
6 293 821 1470
2 2
3285 ,

)
3   
2  1 6082 4 2 128 576 2 343 1709 2
16
3.5
   2 2  4 *
2
320 360 2 293 1470 2 10952 ,

)   2 *
3 2
3  1 2 320 360 2 293 1470 2 8212 1095 , 3.6
8

with the constraint 1 1 2 2 3 3  0, derived from 3.2.

Equations 3.43.6 show that circular orbits 1  2  0, 3  R, the north pole
of the sphere are always equilibria. Equations 3.5 and 3.6 vanish when 1  0, 2 /
 0, but
3.4 vanishes only when
  
10952 4 2 320 360 2 802 2565 2 2
   3.7
192 216 2 362 35 2 6 612 8  0.

Equation 3.7 is a polynomial equation of degree 8 in , therefore admitting eight roots. Note
that, for the accepted values of  1, 3.7 is of the form A21 A22 x A23 x3 A24 x4  0 that
admits a maximum of three real roots, according to Descartes rule of signs.
The real roots of 3.7 veried by the dynamical constraint || 1 are also
equilibria. The root  1 marks a change in the number of equilibria due to a bifurcation
 > 1 could be a root but not an equilibrium. Then, the number of equilibria changes when
crossing the line

27 45 3 5076 1473 2 4730 4 27375 6 3.8
4
423 767 2 1470 4

obtained setting  1 in 3.7 that establishes a relation between the dynamical parameters
and corresponding to bifurcations of circular orbits. Figure 2 shows that this line denes
two regions in the parameters plane with dierent number of equilibria in phase space.
Circular orbits in the outside region of the curve are stable. When crossing the line given
by 3.8 the number of real roots of 3.7 with dynamical sense increases such that a pitchfork
Mathematical Problems in Engineering 7

Bifurcation line of circular orbits

0.7

0.6

0.5

0.4


0.3

0.2

0.1

0
0.5 0 0.5

 1 e2 cos I

Figure 2: Regions in the parameters plane with dierent numbers of equilibria.

1
Stable circular orbtis

0.5
Unstable circular orbtis
1 e2 cos I

Enceladus
Callisto
Europa

0 Stable elliptic orbtis :g  2

0.5

Stable circular orbtis

1
0 0.02 0.04 0.06 0.08 0.1


Figure 3: Bifurcation lines in the parameter plane.

bifurcation takes place: circular orbits change to unstable and two stable elliptic orbits appear
with periapsis, respectively at g  /2, as in the example of Figure 1.
Note that the curve given by 3.8 notably modies the classical inclination limit
cos2 I > 3/5 for circular orbits stability. However, we cannot extend the practical application
of the analytical theory to any value of . It is common to limit the validity of the Hill problem
approximation to one third of the Hill radius rH  31/3 . Then < rH /33/2  1/9, including
most of the planetary satellites of interest. Figure 3 shows the bifurcation lines of circular
orbits in the validity region of the parameters plane with the values of corresponding to
low altitude orbits around dierent planetary satellites highlighted.
A powerful test for estimating the quality of the analytical theory is to check the
degree of agreement of the bifurcation lines of the analytical theory with those computed
numerically in the nonaveraged problem. To do that we compute several families of three-
dimensional, almost circular, periodic orbits of the Hill nonaveraged problem that bifurcate
from the family of planar retrograde orbits at dierent resonances. For variations of the Jacobi
constant the almost circular periodic orbits evolve from retrograde to direct orbits through the
180 degrees of inclination. At certain critical points, almost circular orbits change from stable
to unstable in a bifurcation phenomenon in which two new elliptic periodic orbits appear.
8 Mathematical Problems in Engineering

0.78 0.77

0.76 Stability 0.78 Instability

1 e2 cos I

0.79

1 e2 cos I
Enceladus

Enceladus
0.74

Europa

Europa
0.8
0.72
0.81

0.7

Instability 0.82 Stability
0.68
0.83
0.02 0.04 0.06 0.08 0.1 0.02 0.04 0.06 0.08 0.1
 
a b
Figure 4: Comparison between the bifurcation line of circular, averaged orbits full line, and the curve of
critical periodic orbits dots.

Table 1: Initial orbital elements of an elliptic frozen orbit for  0.0470573,  0.422618.

Theory a Hill units e I deg g deg h 

Classical 0.130342 0.674094 55.0995 90 0 0
2nd order 0.130342 0.648065 55.6915 90 0 0
3rd order 0.130515 0.637316 56.1798 90 0 0
4th order 0.130538 0.634803 56.2813 90 0 0

The computation of a variety of these critical points helps in determining stability regions for
almost circular orbits 17.
The tests done show that the fourth-order theory gives good results for < 0.05. As
presented in Figure 4, the bifurcation line of retrograde orbits clearly diverges from the line of
corresponding critical periodic orbits for higher values of , and it may be worth developing
a higher-order theory that encompasses also the case of Enceladus.

4. Frozen Orbits Computation

Hills case of orbits close to the smaller primary is a simplication of the restricted three-
body problem, which in turn is a simplication of real models. Therefore, the nal goal of
our theory is not the generation of ephemerides but to help in mission designing for articial
satellites about planetary satellites, where frozen orbits are of major interest.
For given values of the parameters , , determined by the mission, a number of
frozen orbits may exist. A circular frozen orbit, either stable or unstable, exists always and
the computation of real roots || 1 of 3.7, if any, will provide the eccentricities of
the stable elliptic solutions with frozen periapsis at g  /2. To each equilibrium of the
doubly reduced phase space it corresponds a torus of quasiperiodic solutions in the original,
nonaveraged model. In what follows we present several examples that justify the eort in
computing a fourth-order theory to reach the quasiperiodicity condition in the Hill problem.

4.1. Elliptic Frozen Orbits

We choose  0.0470573,  0.422618. If we rst try the classical double-averaged solution,
the Hamiltonian 2.3 is simplied to K0,0 K0,1 2 /2K0,2 , and the existence of elliptic
Mathematical Problems in Engineering 9

Classical averaging Undoing 2nd order averaging

0.131 0.131
0.1305 0.1305
a 0.13 a 0.13
0.1295 0.1295
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
58 58
I deg

I deg
56 56
54 54
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0.7 0.7
e 0.66 e 0.66
0.62 0.62
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
g deg

g deg
85 85
90 90
95 95
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0.62 0.62

0.64 0.64
e sin g

e sin g

0.66 0.66

0.68 0.68

0.7 0.7

0.72 0.72

0.05 0 0.05 0.05 0 0.05

e cos g e cos g

a b
Figure 5: Long-term evolution of the orbital elements of the elliptic frozen orbit.

frozen orbits reduces to the case 2 < 3/5, g  /2. The eccentricity of the elliptic frozen
1/4
solutions is then computed from  5 2 /3 obtained by neglecting terms in in 3.7.
Thus, for the given values of and , and taking into account that we are free to choose
the initial values of the averaged angles , h, we get the orbital elements of the rst row
of Table 1. The left column of Figure 5 shows the long-term evolution of the instantaneous
orbital elements for this case, that we call classical averaging, in which we nd long-period
oscillations of more than four degrees in inclination, more than fteen in the argument of
periapsis, and a variation of 0.06 in the eccentricity.
When computing a second-order theory with the Lie-Deprit perturbation method
we arrive exactly at the classical Hamiltonian obtained by a simple removal of the short-
period terms and the classical bifurcation condition that results in the critical inclination
of the third-body perturbations I  39.2 10, 11. However, now we have available the
transformation equations to recover the short- and long-period eects, although up to the
rst order only. After undoing the transformation equations we nd the orbital elements of
the second row of Table 1, where we see that all the elements remain unchanged except for
10 Mathematical Problems in Engineering

Undoing 3th order averaging Undoing 4th order averaging

0.131 0.131
0.1305 0.1305
a 0.13 a 0.13
0.1295 0.1295
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
58 58
I deg

I deg
56 56
54 54
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0.7 0.7
e 0.66 e 0.66
0.62 0.62
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
g deg

g deg
85 85
90 90
95 95
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0.62 0.62

0.64 0.64
e sin g

e sin g

0.66 0.66

0.68 0.68

0.7 0.7

0.72 0.72

0.05 0 0.05 0.05 0 0.05

e cos g e cos g

a b
Figure 6: Long-term evolution of the orbital elements of the elliptic frozen orbit.

the eccentricity and inclination. The long-term evolution of these elements is presented in
the right column of Figure 5, in which we notice a signicant reduction in the amplitude of
long-period oscillations: 2.5 in inclination, around 10 in the argument of the periapsis, and
0.04 in eccentricity.
The results of the third- and fourth-order theories are presented in the last two rows
of Table 1 and in Figure 6. The higher-order corrections drive slight enlargements in the
semimajor axis. While both higher-order theories produce impressive improvements, we note
a residual long-period oscillation in the elements computed from the third-order theory left
column of Figure 6. On the contrary, the orbital elements of the frozen orbit computed with
the fourth-order theory are almost free from long-period oscillations and mainly show the
short-period oscillations typical of quasiperiodic orbits.

4.2. Circular Frozen Orbits

If we choose the same value for but now  0.777146, frozen elliptic orbits do not exist any
longer and the circular frozen orbit is stable. Both the third and fourth-order theories provide
Mathematical Problems in Engineering 11

0.008 0.008
0.006 0.006
e 0.004 e 0.004
0.002 0.002
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
a b

0.005 0.005
e sin g

e sin g
0 0

0.005 0.005

0.004 0.002 0 0.002 0.004 0.004 0.002 0 0.002 0.004

e cos g e cos g

c d

Figure 7: Long-term evolution of the orbital elements of the circular stable frozen orbit. a and c third-
order theory. b and d fourth-order theory.

good results, but, again, the third-order theory provides small long-period oscillations in the
eccentricity whereas the fourth-order theory leads to a quasiperiodic orbit see Figure 7.
For  0.0339919 and  0.34202 the circular frozen orbit is unstable. Due to the
instability, a long-term propagation of the initial conditions from either the third or the
fourth theory shows that the orbit escapes following the unstable manifold with exponential
increase in the eccentricity. But, as Figure 8 shows, the orbit remains frozen much more time
when using the fourth-order theory. A variety of tests performed on science orbits close to
Galilean moons Europa and Callisto showed that the fourth-order theory generally improves
by 50% the lifetimes reached when using the third-order theory.

4.3. Fourier Analysis

Alternatively to the temporal analysis mentioned previously, a frequency analysis using the
Fast Fourier Transform FFT shows how initial conditions obtained from dierent orders of
the analytical theory can be aected of undesired frequencies that defrost the orbital elements.
12 Mathematical Problems in Engineering

Undoing 3rd order averaging Undoing 4th order averaging

0.1052 0.1052
a 0.1050 a 0.1050
0.1048 0.1048
0.1046 0.1046
0 5 10 15 20 25 0 5 10 15 20 25
Years Years
71 71
I deg

I deg
69 69
67 67
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0.4 0.4
0.3 0.3
e e
0.2 0.2
0.1 0.1
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Years Years

0 0

0.05 0.05

0.1 0.1
e sin g

e sin g

0.15 0.15

0.2 0.2

0.25 0.25

0.3 0.3

0.3 0.25 0.2 0.15 0.1 0.05 0 0.3 0.25 0.2 0.15 0.1 0.05 0
e cos g e cos g
a b

Figure 8: Long-term evolution of the orbital elements of the circular, unstable, frozen orbit.

Thus, Figure 9 shows the FFT analysis of the instantaneous argument of the periapsis
of the elliptic orbit in the example mentioned previously. Dots correspond to initial conditions
obtained from the double-averaged phase space after a classical analysisthat is equivalent
to the second-order analytical theoryand the line corresponds to initial conditions obtained
from the fourth-order analytical theory after undoing the transformation. While most of the
frequencies match with similar amplitudes, in the magnication of the right plot we clearly
appreciate a very low frequency of 0.15 cycles/year with a very high amplitude in the
classical theory that is almost canceled out with the fourth-order approach. The semiannual
frequency remains in both theories because it is intrinsic to the problem. It is due to the third-
body perturbation and it cannot be avoided.
Figure 10 shows a similar analysis for the instantaneous eccentricity of the stable
circular orbit mentioned previously. Now, dots correspond to the fourth-order theory and
the line to the third-order one both after undoing the transformation equations. While the
Mathematical Problems in Engineering 13

0.5

log10 amplitude
0
2
log10 amplitude

1
0 0.5
1
2
3 1
4
5 1.5
0 10 20 30 40 50 60 70 0 0.5 1 1.5 2 2.5 3
Cycles/year Cycles/year

a b

Figure 9: a FFT analysis of the instantaneous argument of the periapsis of the elliptic solution. b
Magnication over the low frequencies region.

1
1.5
log10 amplitude
2
log10 amplitude

1
2.5
2
3 3
4
5 3.5
6
7 4
0 10 20 30 40 50 60 70 0 0.5 1 1.5 2 2.5 3
Cycles/year Cycles/year

a b

Figure 10: a FFT analysis of the instantaneous eccentricity of the elliptic solution. b Magnication over
the low frequencies region.

third-order theory provides good results, reducing the amplitude of the undesired frequency
to low values, the fourth-order theory practically cancels out that frequency.
An FFT analysis of unstable circular orbits has not much sense because of the time
scale in which the orbit destabilizes.

5. Conclusions
Frozen orbits computation is a useful procedure in mission designing for articial satellites.
After locating the frozen orbit of interest in a double-averaged problem, usual procedures for
computing initial conditions of frozen orbits resort to trial-and-error interactive corrections,
or require involved computations. However, the explicit transformation equations between
averaged and nonaveraged models can be obtained with analytical theories based on the Lie-
Deprit perturbation method, which makes the frozen orbits computations straightforward.
Accurate computations of the initial conditions of frozen, quasiperiodic orbits can
be reached with higher-order analytical theories. This way of proceeding should not be
undervalued in the computation of science orbits around planetary satellites, a case in which
third-body perturbations induce unstable dynamics.
14 Mathematical Problems in Engineering

Higher-order analytical theories are a common tool for computing ephemeris among
the celestial mechanics community. They are usually developed with specic purpose,
sophisticated algebraic manipulators. However, the impressive performances of modern
computers and software allow us to build our analytical theory with commercial, general-
purpose manipulators, a fact that may challenge aerospace engineers to use the safe, well-
known techniques advocated in this paper.

Appendix
Let T : x, X x , X , where x are coordinates and X their conjugate momenta, be
+ i
a Lie transform from new primes to old variables. If W  i  /i! Wi 1 x, X is
its generating function expanded as a power series in a small parameter , a function
+ i
F  /i! Fi,0 x, X can be expressed in the new variables as the power series T : F 
+ i i
i  /i! F0,i x , X  whose coecients are computed from the recurrence
 
 i , -
Fi,j  Fi 1,j1 Fk,j1 ; Wi 1k , A.1
0ki k

where {Fk,j1 ; Wi 1k }  x Fk,j1 X Wi 1k X Fk,j1 x Wi 1k , is the Poisson bracket.

Conversely, the coecients Wi 1 of the generating function can be computed step by step
from A.1 once corresponding terms F0,i of the transformed function are chosen as desired.
In perturbation theory it is common to chose the F0,i as an averaged expression over some
variable, but it is not the unique possibility 18. Full details can be found in the literature
19, 20.
To average the short-period eects we write Hamiltonian 2.2 in Delaunay variables
as
     
2 3 4
H  H0,0 H1,0 H2,0 H3,0 H4,0 ,
2 6 24 A.2

where H0,0  1/2L2 , H1,0  H, H2,0  r 2 {1 3cosf g cos h c sinf g sin h2 }, and
H3,0  H4,0  0. Note that the true anomaly f is an implicit function of .
Since the radius r never appears in denominators, it results convenient to express
Hamiltonian A.2 as a function of the ellipticinstead of the trueanomaly u by using the
ellipse relations r sin f  a sin u, r cos f  acos u e, r  a1 e cos u.
After applying the Delaunay normalization 21 up to the fourth-order in the
Hamiltonian, we get
     
 2 3 4
H  H0,0  H0,1 H0,2 H0,3 H0,4 , A.3
2 6 24

where, omitting primes,

1
H0,0  ,
2L2
H0,1  H0,0 2c,
Mathematical Problems in Engineering 15
 
2  
H0,2  H0,0 4 6e2 2 3s2 3s2 cos 2h
8
   
15e2 2s2 cos 2g 1 c2 cos 2g 2h 1 c2 cos 2g 2h ,
 
453    
H0,3  H0,0 e2 1 c2 cos 2g 2h 1 c2 cos 2g 2h ,
8
 
34
H0,4  H0,0
512
  
16 47 282c2 63c4 144 227 90c2 59c4 e2
   
18 227 610c2 701c4 e4 24s2 558 270c2 109 555c2 e2 e2 cos 2g
   
24s2 216 56c2 8 161 59c2 e2 11 701c2 e4 cos 2h
   
481 c2 338 90c 90c2 91 185c 185c2 e2 e2 cos 2g 2h
   
481 c2 338 90c 90c2 91 185c 185c2 e2 e2 cos 2g 2h

6s4 56 472e2 701e4 cos 4h 1710s4 e4 cos 4g
    
60s2 18 37e2 e2 1 c2 cos 2g 4h 1 c2 cos 2g 4h
   
1140s2 e4 1 c2 cos 4g 2h 1 c2 cos 4g 2h
   
285e4 1 c4 cos 4g 4h 1 c4 cos 4g 4h .
A.4

The generating function of the transformation is W  W2 1/2W3 , where

 
2
W2  L
192
    
4 2 3s2 3e 5 32 S1,0,0 9e2 S2,0,0 e3 S3,0,0
  
6s2 e 3 5 32 S1,0,2 S1,02  9eS2,0,2 S2,02  e2 S3,0,2 S3,02 
 2   
6s2 1 15eS1,2,0 9 6 S2,2,0 eS3,2,0
 2   
6s2 1 15eS1,2,0 9 6 S2,2,0 eS3,2,0
 2   
31 c2 1 15eS1,2,2 9 6 S2,2,2 eS3,2,2
16 Mathematical Problems in Engineering
 2   
31 c2 1 15eS1,2,2 9 6 S2,2,2 eS3,2,2
 2   
31 c2 1 15eS1,2,2 9 6 S2,2,2 eS3,2,2
 2   
31 c2 1 15eS1,2,2 9 6 S2,2,2 eS3,2,2 ,
 
3
W3  L
256
  
72es2 13 32 S1,0,2 S1,0,2  24e2 s2 17 42 S2,0,2 S2,0,2 

88e3 s2 S3,0,2 S3,0,2  6e4 s2 S4,0,2 S4,0,2 

   
36e 1 13 82 1 c2 S1,2,2 1 c2 S1,2,2
   
36e 1 13 82 1 c2 S1,2,2 1 c2 S1,2,2
 2   
12 1 17 6 82 1 c2 S2,2,2 1 c2 S2,2,2
 2   
12 1 17 6 82 1 c2 S2,2,2 1 c2 S2,2,2
 2   
4 1 e 11 6 1 c2 S3,2,2 1 c2 S3,2,2
 2   
4 1 e 11 6 1 c2 S3,2,2 1 c2 S3,2,2
 2  
3 1 e2 1 c2 S4,2,2 1 c2 S4,2,2
 2  
3 1 e2 1 c2 S4,2,2 1 c2 S4,2,2 .

A.5

We shorten notation calling Si,j,k sini u j g k h.

The Lie transform of generating function W can be applied to any function of
+
Delaunay variables F  i i /i! Fi   , g  , h , L , G , H  . Since W1  0, up to the third-order
in the small parameter recurrence A.1 gives

   
2 3
F  F0 {F0 ; W2 } {F0 ; W3 }. A.6
2 6

Specically, this applies to the transformation equations of the Delaunay variables

themselves, where F0   , g  , h , L , G , H   and Fi 0 for i > 0.
+
A new application of the recurrence A.1 to the Hamiltonian K  0i4 i /i! Ki,0 ,
where Ki,0 H0,i of A.3, allows to eliminate the node up to the fourth-order, obtaining
the double-averaged Hamiltonian 2.3. Note that K0,4 corrects previous results in 22. The
generating function of the transformation is V  V1 V2 2 /2 V3 , where, omitting double
Mathematical Problems in Engineering 17

primes,
 
3    
V1  L 4 6e2 s2 sin 2h 5 1 c2 e2 sin 2g 2h 5 1 c2 e2 sin 2g 2h ,
64
   
3 
V2  L 6c 2 17e2 s2 sin 2h 5 2 9c 1 c2 e2 sin 2g 2h
128
 
51 c2 2 9ce2 sin 2g 2h ,
 
9
V3  L
32768
    
16s2 456 104c2 8 193 754c2 e2 47 7831c2 e4 sin 2h

2s4 232 416e2 1803e4 sin 4h
    
321 c2 e2 2 323 285c 780c2 527 1135c 2125c2 e2 sin 2g 2h
    
321 c2 e2 2 323 285c 780c2 527 1135c 2125c2 e2 sin 2g 2h
    
220s2 e2 4 11e2 1 c2 sin 2g 4h 1 c2 sin 2g 4h
   
4520s2 e4 1 c2 sin 4g 2h 1 c2 sin 4g 2h
   
385e4 1 c4 sin 4g 4h 1 c4 sin 4g 4h .
A.7

The new Lie transform of generating function V can be applied to any function of
Delaunay variables, and, specically, to the Delaunay variables themselves. For any 
  , g  , h , L , G , H   the transformation equations of the Lie transform are computed, up
to the third-order, from
   
  2 3
 1 2 3 , A.8
2 6

where
, -
1   ; V1 ,
, -
2   ; V2 {1 ; V1 }, A.9
, - ,, - -
3   ; V3  ; V2 ; V1 {1 ; V2 } {2 ; V1 }.

Acknowledgments
This work was supported from Projects ESP 2007-64068 the rst author and MTM
2008-03818 the second author of the Ministry of Science and Innovation of Spain is
18 Mathematical Problems in Engineering

acknowledged. Part of this work has been presented at 20th International Symposium on
Space Flight Dynamics, Annapolis, Maryland, USA, September, 2428 2007.

References
1 G. W. Hill, Researches in the lunar theory, American Journal of Mathematics, vol. 1, no. 2, pp. 129147,
1878.
2 M. L. Lidov, The evolution of orbits of articial satellites of planets under the action of gravitational
perturbations of external bodies, Planetary and Space Science, vol. 9, no. 10, pp. 719759, 1962,
translated from Iskusstvennye Sputniki Zemli, no. 8, p. 5, 1961.
3 M. Henon, Numerical exploration of the restricted problem. VI. Hills case: non-periodic orbits,
Astronomy and Astrophysics, vol. 9, pp. 2436, 1970.
4 D. P. Hamilton and A. V. Krivov, Dynamics of distant moons of asteroids, Icarus, vol. 128, no. 1, pp.
241249, 1997.
5 Y. Kozai, Motion of a lunar orbiter, Publications of the Astronomical Society of Japan, vol. 15, no. 3, pp.
301312, 1963.
6 M. L. Lidov and M. V. Yarskaya, Integrable cases in the problem of the evolution of a satellite orbit
under the joint eect of an outside body and of the noncentrality of the planetary eld, Kosmicheskie
Issledovaniya, vol. 12, pp. 155170, 1974.
7 D. J. Scheeres, M. D. Guman, and B. F. Villac, Stability analysis of planetary satellite orbiters:
application to the Europa orbiter, Journal of Guidance, Control and Dynamics, vol. 24, no. 4, pp. 778787,
2001.
8 M. Lara, J. F. San-Juan, and S. Ferrer, Secular motion around synchronously orbiting planetary
satellites, Chaos, vol. 15, no. 4, pp. 113, 2005.
9 M. E. Paskowitz and D. J. Scheeres, Orbit mechanics about planetary satellites including higher
order gravity elds, in Proceedings of the Space Flight Mechanics Meeting, Copper Mountain, Colo,
USA, January 2005.
10 Y. Kozai, Secular perturbations of asteroids with high inclination and eccentricity, The Astronomical
Journal, vol. 67, no. 9, pp. 591598, 1962.
11 R. A. Broucke, Long-term third-body eects via double averaging, Journal of Guidance, Control and
Dynamics, vol. 26, no. 1, pp. 2732, 2003.
12 M. E. Paskowitz and D. J. Scheeres, Design of science orbits about planetary satellites: application to
Europa, Journal of Guidance, Control and Dynamics, vol. 29, no. 5, pp. 11471158, 2006.
13 A. Deprit, Canonical transformations depending on a small parameter, Celestial Mechanics, vol. 1,
no. 1, pp. 1230, 1969.
14 M. Lara, Simplied equations for computing science orbits around planetary satellites, Journal of
Guidance, Control, and Dynamics, vol. 31, no. 1, pp. 172181, 2008.
15 A. Deprit and A. Rom, The main problem of articial satellite theory for small and moderate
eccentricities, Celestial Mechanics, vol. 2, no. 2, pp. 166206, 1970.
16 S. Coey, A. Deprit, and E. Deprit, Frozen orbits for satellites close to an Earth-like planet, Celestial
Mechanis and Dynamical Astronomy, vol. 59, no. 1, pp. 3772, 1994.
17 M. Lara and D. Scheeres, Stability bounds for three-dimensional motion close to asteroids, Journal
of the Astronautical Sciences, vol. 50, no. 4, pp. 389409, 2002.
18 A. Deprit, The elimination of the parallax in satellite theory, Celestial Mechanics, vol. 24, no. 2, pp.
111153, 1981.
19 S. Ferrer and C. A. Williams, Simplications toward integrability of perturbed Keplerian systems,
Annals of the New York Academy of Sciences, vol. 536, pp. 127139, 1988.
20 J. F. Palacian, Dynamics of a satellite orbiting a planet with an inhomogeneous gravitational eld,
Celestial Mechanis and Dynamical Astronomy, vol. 98, no. 4, pp. 219249, 2007.
21 A. Deprit, Delaunay normalisations, Celestial Mechanics, vol. 26, no. 1, pp. 921, 1982.
22 J. F. San-Juan and M. Lara, Normalizaciones de orden alto en el problema de Hill, Monografas de la
Real Academia de Ciencias de Zaragoza, vol. 28, pp. 2332, 2006.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 396267, 14 pages
doi:10.1155/2009/396267

Research Article
Collision and Stable Regions around Bodies with
Simple Geometric Shape

A. A. Silva,1, 2, 3 O. C. Winter,1, 2 and A. F. B. A. Prado1, 2

1
Space Mechanics and Control Division (DMC), National Institute for Space Research (INPE),
Sao Jose dos Campos 12227-010, Brazil
2
UNESP, Universidade Estadual Paulista, Grupo de Dinamica Orbital & Planetologia,
Av. Ariberto Pereira da Cunha, 333-Guaratingueta 12516-410, Brazil
3
Universidade do Vale do Paraba (CT I / UNIVAP), Sao Jose dos Campos 12245-020, Brazil

Correspondence should be addressed to A. A. Silva, aurea [email protected]

Received 29 July 2009; Accepted 20 October 2009

Recommended by Silvia Maria Giuliatti Winter

We show the expressions of the gravitational potential of homogeneous bodies with well-dened
simple geometric shapes to study the phase space of trajectories around these bodies. The
potentials of the rectangular and triangular plates are presented. With these expressions we study
the phase space of trajectories of a point of mass around the plates, using the Poincare surface of
section technique. We determined the location and the size of the stable and collision regions in
the phase space, and the identication of some resonances. This work is the rst and an important
step for others studies, considering 3D bodies. The study of the behavior of a point of mass
orbiting around these plates 2D, near their corners, can be used as a parameter to understand
the inuence of the gravitational potential when the particle is close to an irregular surface, such
as large craters and ridges.

Copyright q 2009 A. A. Silva et al. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
The aim of this paper is to study the phase space of trajectories around some homogeneous
bodies with well-dened simple geometric shapes. Closed-form expressions derived for the
gravitational potential of the rectangular and triangular plates were obtained from Kellogg
1 and Broucke 2. They show the presence of two kinds of terms: logarithms and arc
tangents. With these expressions we study the phase space of trajectories of a particle around
two dierent bodies: a square and a triangular plates. The present study was made using the
Poincare surface of section technique which allows us to determine the location and size of
the stable and chaotic regions in the phase space. We can nd the periodic, quasiperiodic and
chaotic orbits.
2 Mathematical Problems in Engineering

Some researches on this topic can be found in Winter 3 that study the stability
evolution of a family of simply periodic orbits around the Moon in the rotating Earth-Moon-
particle system. He uses the numerical technique of Poincare surface of section to obtain
the structure of the region of the phase space that contains such orbits. In such work it
is introduced a criterion for the degree of stability. The results are a group of surfaces of
section for dierent values of the Jacobi constant and the location and width of the maximum
amplitude of oscillation as a function of the Jacobi constant. Another research was done
by Broucke 4 that presents the Newtons law of gravity applied to round bodies, mainly
spheres and shells. He also treats circular cylinders and disks with the same methods used
for shells and it works very well, almost with no modications. The results are complete
derivations for the potential and the force for the interior case as well as the exterior
case.
In Sections 2 and 3, following the works of Kellogg 1 and Broucke 2, we show the
expressions for the potential of the rectangular and triangular plates, respectively. In Section 4
we use the Poincare surface of section technique to study the phase space around the plates.
In Section 5 we show the size and location of stable and collision regions in the phase space.
In the last section, we have some nal comments.

2. The Potential of the Rectangular Plate

Let us consider a homogeneous plane rectangular plate and an arbitrary point P 0, 0, Z, not
on the rectangle. Take x and y axes parallel to the sides of the rectangle, and their corners
referred to these axes are Ab, c, Bb , c, Cb , c , and Db, c . Let Sk denote a typical
element of the surface, containing a point Qk located in the rectangular plate with coordinates
xk , yk . See Figure 1.
The potential of the rectangular plate can be given by the expression

 c  b
GSk dxdy
U  G 
rk c b x2 y 2 Z 2
2.1
) c      c    *
 2
 G ln b b y2 Z2 dy ln b b2 y2 Z2 dy ,
c c

where G is the Newtons gravitational constant, is the density of the material, and rk is the
distance between the particle and the point Qk .
In evaluating the integrals we nd

b d3  b d1  c d3  c d1 
U  G c ln c ln  b ln b ln 
b d4  b d2  c d2  c d4 
  2.2
b c b c b c bc
Z tan1 tan1 tan1 tan1 ,
Z d4 Z d3 Z d2 Z d1
Mathematical Problems in Engineering 3

D c C

Sk
Qk
x
b O b

A c B

Figure 1: Rectangular plate is on the plane x, y.

where

d12  b2 c2 Z2 ,

d22  b c2 Z2 ,
2

2.3
d32  b c Z2 ,
2 2

d42  b2 c Z2
2

are the distances from P 0, 0, Z to the corners A, B, C, and D, respectively.

3. The Potential of the Triangular Plate

We will give the potential at a point P 0, 0, Z on the Z-axis created by the triangle shown in
Figure 2 located in the xy-plane. The side P1 P2 is parallel to the x-axis. The coordinates of P1
and P2 are r1 x1 , y1  and r2 x2 , y2 , but we have that y1  y2 and x1 > x2 > 0. The distances
are given by d12  x12 y12 Z2 and d22  x22 y12 Z2 , where d1 is the distance from P 0, 0, Z
to the corner P1 x1 , y1  and d2 is the distance from P 0, 0, Z to the corner P2 x2 , y1 . Using
the denition of the potential, we have that the potential at P 0, 0, Z can be given by

.
     /
x1 d1 1 1 Z 1 2 Z
U  G y1 ln Ztan Ztan |Z|12 , 3.1
x2 d2 d1 d2

where 1  x1 /y1 , 2  x2 /y2 , and 12  1 2 represent the angle of the triangle at the
origin O and it is showed in Figure 2.
The potential of this triangle at the point P 0, 0, Z on the Z-axis must be invariant
under an arbitrary rotation of the triangle around the same Z-axis. Therefore, 3.1 should
also be invariant under this rotation and their four terms are individually invariant, where
they can be expressed in terms of invariant quantities, such as the sides and the angle of the
triangle. The potential of an arbitrary triangular plate P1 P2 P3 Figure 3 can be obtained by
4 Mathematical Problems in Engineering

P2 x2 , y1  P1 x1 , y1 

y1  y2 Qk
r2 r1

12
2 1

2
x
O

Figure 2: Triangle OP1 P2 located in the xy-plane x, y.

y
P2

r12

r2
r23 r1 P1
x
O
r3 r31

P3

Figure 3: Triangular plate P1 P2 P3 .

the sum of three special triangles of the type used in this section. So, the result will have three
logarithmic terms and three arc-tangent terms. So, the potential is given by

.



C12 d1 d2 r12 C23 d2 d3 r23 C31 d3 d1 r31
U  G ln ln ln
r12 d1 d2 r12 r23 d2 d3 r23 r31 d3 d1 r31
      / 3.2
1 Nd1 1 Nd2 1 Nd3
tan tan tan SignZ ,
D1 D2 D3

where the numerator and the denominators are

N  ZC12 C23 C31 ,


D1  Z2 r12 D23 D31 D12 C12 C31 ,
 3.3
D2  Z2 r22 D31 D12 D23 C23 C12 ,

D3  Z2 r32 D12 D23 D31 C31 C23 .
Mathematical Problems in Engineering 5

The symbol SignZ in 3.2 is the sign of the variable Z and rij i, j  1, 2, 3 is the distance
between Pi and Pj . In particular, the dot-product Dij  ri rj cos ij and the cross-product
Cij  ri rj sin ij of the vectors ri and rj are two invariants, where ij  i j . In order
to obtain more details see Broucke 2. This result is invariant with respect to an arbitrary
rotation around the Z-axis.
An important generalization of 2.2 and 3.2 can be done, for the case where the
point P has arbitrary coordinates X, Y, Z instead of being on the Z-axis. For the rectangle,
this task is rather easy, because the four vertices A, B, C, and D have completely arbitrary
locations. In order to have more general results it is sucient to replace b, b , c, c  by
b X, b X, c Y, c Y  in 2.2. For the triangle, it is sucient to replace all the vertex-
coordinates xi , yi  by xi X, yi Y  in 3.2. These generalizations give us expressions
for the potential U of the rectangle and the triangle as a function of the variables X, Y, Z.
Then, it is possible to compute the components of the acceleration by taking the gradient of
the potential. It is important to note that, in taking the partial derivatives of UX, Y, Z, the
arguments of the logarithms and the arc tangents were treated as constants by Broucke 2.
This simplies the work considerably. The general expressions of the acceleration allow us to
study orbits around these plates, and this study is very important to obtain knowledge that
will be necessary to study the cases of three-dimensional solids, such as polyhedra.

4. Study of the Phase Space around Simple Geometric Shape Bodies

With the potential determined, we study the phase space of trajectories of a point of mass
around the plates. In this section we use the Poincare surface of section technique to study the
regions around the rectangular and triangular plates. In this part of the paper we explore the
space of initial conditions. The results are presented in Poincare sections x, x, from which
one can identify the nature of the trajectories: periodic, quasiperiodic, or chaotic orbits. We
can nd the collision regions and identify some resonances.
In order to obtain the orbital elements of a particle at any instant it is necessary to
know its position x, y and velocity x, y that corresponds to a point in a four dimensional
phase space. The conservation of the total energy of the system implies in the existence of
a three-dimensional surface in this phase space. For a xed value of the total energy only
three of the four quantities are needed, for example, x, y, and x, since the other one y is
determined, up to the sign, by the total energy. By dening a plane, that we choose y  0,
in the resulting three-dimensional space, the values of x and x can be plotted every time
the particle has y  0. The ambiguity of the sign of y is removed by considering only those
crossing with a xed sign of y. The section is obtained by xing a plane in the phase space
and plotting the points when the trajectory intersects this plane in a particular direction
Winter and Vieira Neto, 5. This technique is used to determine the regular or chaotic
nature of the trajectory. In the Poincare map, if there are closed well-dened curves then
the trajectory is quasiperiodic. If there are isolated single points inside such islands, the
trajectory is periodic. Any fuzzy distribution of points in the surface of section implies that
the trajectory is chaotic Winter and Murray 6.
In order to study the regions around simple geometric shape bodies, two cases were
considered. In the rst case we used a square plate, where all of its sides are   2 and the
initial position of the test particle is chosen arbitrarily in the right side of the plate. The motion
of the particle is in the counterclockwise direction. In the second case we used an equilateral
triangular plate, with sides   1 and one of its sides is parallel to the y-axis. For the triangle,
the initial position of the test particle is in the right side of the plate and its motion is in the
6 Mathematical Problems in Engineering

counterclockwise direction. Both of the cases consider a baricentric system, where the plates
are centered in the origin of the system. The constants used are the Newtons gravitational
constant and the density of the material, respectively, G   1.
In this paper, the Poincare surface of section describes a region of the phase space
for two-body problem plate-particle. The numerical study makes use of the Burlisch-Stoer
method of integration with an accuracy of O 1012 . The Newton-Raphson method was used
to determine when the trajectory crosses the plane y  0 with an accuracy of O 1011 . There
is always a choice of surfaces of section and, in our work, the values of x and x were computed
whenever the trajectory crossed the plane y  0 with y > 0.
In this section, Figure 4 represents a set of Poincare surfaces of section around the
triangular plate with values of energy, E, varying from 0.41 to 0.46, at intervals of 0.01. We
considered approximately seventy starting conditions for each surface of section, but most of
them generated just a single point each one. In that case the particle collided with the plate.
Figures 4a to 4f show regions where the initial conditions generated well-dened
curves. These sections represent quasiperiodic trajectories around the triangular plate. The
center of the islands for each surface of section corresponds to one periodic orbit of the
rst kind circular orbit. In Figure 4a, for example, we can see a group of four islands
corresponding to a single trajectory with E  0.41 and x0  0.8422. The center of each
one of those islands corresponds to one periodic orbit of the second kind resonant orbit.
In the same gure there is a separatrix that represents the trajectory with E  0.41 and
x0  0.83176 where the particle has a chaotic movement conned in a small region.
Figure 4d shows that the group of four islands is disappearing and a new group of ve
islands appears. This new group of islands corresponds to a trajectory with E  0.44 and
x0  0.80281. The region of single points outside the islands corresponds to a chaotic sea
where the particle collides with the plate. This is a region highly unstable.
Figure 5a shows a Poincare surface of section around the triangular plate for energy,
E  0.4. With a zoom in the region 1 we can see a fuzzy but conned distribution of points
that corresponds to the separatrix Figure 5b generated with x0  0.80816. With a zoom
in the region 2 we can see islands Figure 5c that were generated with x0  0.81596 and
x0  0.83351.
A similar structure of the phase space is found for the square plate when analyzed its
Poincare surface of section.

5. Collision and Stable Regions

In this section we show a global vision of the location and size of the stable and collision
regions around the triangular and square plates, using Poincare map. The values of x, when
x  0, are measured with the largest island of stability quasiperiodic orbit for each value
of energy Winter 3. With the diagrams Figures 6 and 7 we can see the evolution of the
stability for the family of periodic orbits for dierent values of energy constant.
The surfaces of section are plotted at intervals of 0.02 for the energy constant E. For
the triangular plate, the energy varies from 0.01 to 0.47 and, for the square plate, from 0.13
to 1.09. The values of the energy E  0.47 for the triangle and E  1.09 for the square
represent the region that there is no island of stability within the adopted precision.
Figure 6a shows the stable and collision regions around the triangular plate. The
white color represents the region that is considered stable, neglecting the small chaotic
regions that appear in the separatrices. The collision region is represented by the dark
gray color and it indicates the region where the particle collides with the plate due to its
Mathematical Problems in Engineering 7

0.3 0.3
E  0.41 E  0.42
0.2 0.2

0.1 0.1

x 0 x 0

0.1 0.1

0.2 0.2

0.3 0.3
1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4
x x
a b

0.3 0.3
E  0.43 E  0.44
0.2 0.2

0.1 0.1

x 0 x 0

0.1 0.1

0.2 0.2

0.3 0.3
1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4
x x
c d

0.3 0.3
E  0.45 E  0.46
0.2 0.2

0.1 0.1

x 0 x 0

0.1 0.1

0.2 0.2

0.3 0.3
1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4
x x
e f

Figure 4: A group of Poincare surfaces of section of trajectory around the triangular plate for dierent
values of energy, 0.46 E 0.41 at intervals of 0.01.
8 Mathematical Problems in Engineering

E  0.4
0.2
Region 1

0.1

x 0

0.1 Region 2

0.2

1.5 1.35 1.2 1.05 0.9 0.75

x
a

b

c

Figure 5: a Poincare surfaces of section around the triangular plate for energy, E  0.4; b a zoom in
region 1 that represents the separatrix; c a zoom in region 2 that represents a group of islands.

gravitational attraction. The light gray color corresponds to a prohibited region where there
are no starting conditions for those values of energy. The triangular plate is located to the right
of the limit of the collision border. Figure 6b corresponds to a zoom of the previous one in
order to have the best visualization of the regions.
Figure 7a summarizes the same studies around the square plate and the color codes
are the same as mentioned for the triangle. The square plate is located to the left of the limit
of the collision border. Figure 7b corresponds to a zoom of the previous one.
Mathematical Problems in Engineering 9

0.5 0.5

0.4 0.4

n
sio
0.3 0.3 Prohibited region
E

lli
Prohibited region

Co
0.2 0.2

0.1 0.1
le Stable
ab
St

100 80 60 40 20 0 11 9 7 5 3 1
X X
a b

Figure 6: a Study of the regions around the triangular plate: stable area white, collision area dark
gray, and prohibitedarea light gray; b zoom of a region in a.

1.2 1.2

1 1
Col

Prohibited region
0.8 0.8 Co
lisio

lli
E

sio
E

Prohibited region
n

0.6 0.6 n

0.4 0.4 Stable

0.2 Stable 0.2

0 5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 10
X X
a b

Figure 7: a Study of the regions around of the square plate: stable area white, collision area dark gray,
and prohibited area light gray; b zoom of a region in a.

6. Trajectories around Triangular and Square Plates

In the last section was shown a global vision of a family of orbits around the geometric plates.
In this section we will show some trajectories of a particle around triangular and square
plates. The initial conditions position and velocity of the particle that orbit the plates are
given by:


x, y, z  x0 , 0, 0,
   6.1

V x, V y, V z  0, 2E U, 0 ,

where E is the total energy of the system and U is the potential energy generated by the plate.
10 Mathematical Problems in Engineering

Table 1: Value of the energy E and the initial position of the particle x0  for the triangular and square
plates.

Figures E x0
8 0.40 0.83
9 0.41 0.83176
10 0.60 1.4

1.5 E  0.4

0.5

Y 0

0.5

1.5

1.5 1 0.5 0 0.5 1 1.5

X

Figure 8: Trajectory around the triangular plate: quasiperiodic.

Table 1 shows the value of the initial position of the particle x0  and its correspondent
value of the energy for each trajectory around the triangular and square plates. Figures 8 and
9 show a quasiperiodic orbit and a chaotic orbit in a region of the separatrix, respectively,
around the triangular plate. Figure 10 shows a quasiperiodic orbit around the square plate.
The trajectories in these gures show that the eect of the potential due to the plates
can be compared with the eect of the Earths attening J2 , generating trajectories as a
precessing elliptic orbit.
In this section we will explore the regions very close to the vertex of the geometric
plates and verify the behavior of the particle in this situation due to the potentials of the
plates. So, it is showed an analysis of the semimajor axis and the eccentricity of the orbit.
In the next example is considered a triangular plate as a central body, the value of the
energy is E  0.1 and the initial position for the particle is x0  1.14 and y0  z0  0,
that generate a quasiperiodic orbit. Figure 11 shows the trajectory around the plate, where
A, B, and C are the vertexes of the triangular plate and Figure 12 shows the behavior of the
semimajor axis versus time. The points in Figure 11 correspond to the enumerated regions in
Figure 12. For better visualization of Figure 11, there is a zoom of the region around the plate.
Figure 13 shows a zoom of Figure 12 with the details of the semimajor axis peaks along six
orbital periods. The results show the following features.
The codes 1a, 2a, 3a, 4a, 5a, and 6a are the maximum values of the semimajor axis that
occur when the particle is close to the vertex A during six complete orbits. The codes 1 to 6
correspond to the six close approaches of the particle to the vertex A. The results show that
Mathematical Problems in Engineering 11

1.5 E  0.41

0.5

Y 0

0.5

1.5

1.5 1 0.5 0 0.5 1 1.5

X

Figure 9: Trajectory around the triangular plate: chaotic orbit separatrix.

E  0.6
4

Y 0

4 2 0 2 4
X

Figure 10: Trajectory around the square plate: quasiperiodic.

the semimajor axis increases as the particle approaches to the vertex A. From Figures 11 to
13 we see the same codes for passages of the particle close to the other two vertexes, B and
C. Since the trajectory of the particle does not get so close to the vertexes, B and C, then the
increase on the semimajor axis is not so large in these cases.
With these analyses we can verify that the proximity of the particle to the corners of the
plate changes the behavior of the trajectory. In the studied cases, the orbits become eccentric
and precess. For the eccentricity we could verify a similar behavior as that found for the
semimajor axis.
The studies with the square plate show the same behavior as obtained for the triangle.
12 Mathematical Problems in Engineering

10 2

8
1 3 2 C
6 4

4 5
Y Y 0 B
2
C A 6
0 A B 5
1 1
2 34
6 5
2

4 2
4 2 0 2 4 6 8 10 2 1 0 1 2
X X
a b

Figure 11: a Trajectory of the particle around the triangular plate. b Zoom of the trajectory near the
plate.

16
5a
4a
14 6a
3a
12

a 10 2a

1a
8

6
2c 3c 4c 5c 5b 6b

4
0 100 200 300 400 500
T

Figure 12: Semimajor axis of the orbit.

7. Final Comments
In this paper we used closed form solutions for the gravitational potential for two simple
geometric shape bodies in order to study the behavior of a particle around each one of them.
The development was applied to the square and to the triangular plates. We used the Poincare
surface of section technique to study the phase space of trajectories of a particle around those
plates. We identied dierent kinds of orbits: periodic, quasiperiodic, and chaotic orbits. We
found a collision region and identied some resonances.
The results showed that there is a region of starting conditions where it is possible
to have orbits around the plates before the collision. The location and size of the stable and
collision regions were measured for each value of energy and it showed us the evolution
of the stable regions in the phase space. We have seen that the Poincare surface of section
technique is an easy and powerful way to identify the nature of the orbits.
Mathematical Problems in Engineering 13

14 14

12 12

a 10 a 10 2a

1a
8 8
2c
6 6

0 1 2 3 4 5 6 7 66 68 70 72 74 76
T T
a b

4a
14 14
3a
12 12

a 10 a 10

8 8

3c
6 6 4c

136 138 140 142 144 146 206 208 210 212 214
T T
c d

5a
14 6a
14

12 12

a 10 a 10

8 8

6b
6 5c 5b 6

276 278 280 282 284 286 348 350 352 354 356
T T
e f
Figure 13: Zoom of Figure 12. Peak of semimajor axis in six orbital periods.

It was observed that the corners of the plates have a signicant inuence in the
behavior of the trajectory, mainly when the particle has a close approach to them. The orbits
become eccentric and precess due their gravitational eld.
This study opens the way to obtain the potential and the trajectories around three-
dimensional bodies with irregular shapes, such as asteroids and comets.
14 Mathematical Problems in Engineering

Acknowledgment
The authors wish to express their appreciation for the supports provided by the Brazilian
agencies: FAPESP, CAPES, and CNPq. They are very grateful to Roger A. Broucke for the
information and discussions.

References
1 O. D. Kellogg, Foundations of Potential Theory, Dover, New York, NY, USA, 1929.
2 R. A. Broucke, Closed form expressions for some gravitational potentials: triangle, rectangle, pyramid
and polyhedron, in Proceedings of AAS/AIAA Spaceight Mechanics Meeting, Albuquerque, NM, USA,
February 1995.
3 O. C. Winter, The stability evolution of a family of simply periodic lunar orbits, Planetary and Space
Science, vol. 48, no. 1, pp. 2328, 2000.
4 R. A. Broucke, The gravitational attraction of spheres and spherical shells, in Advances in Space
Dynamics, O. C. Winter and A .F. B. A. Prado, Eds., vol. 2, pp. 190233, INPE, Sao Jose dos Campo,
Brazil, 2002.
5 O. C. Winter and E. Vieira Neto, Distant stable direct orbits around the Moon, Astronomy and
Astrophysics, vol. 393, no. 2, pp. 661671, 2002.
6 O. C. Winter and C. D. Murray, Atlas of the Planar, Circular, Restricted Three-Body Problem. I. Internal
Orbit, vol. 16 of QMW Maths Notes, Queen Mary and Westeld College, London, UK, 1994.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 181360, 23 pages
doi:10.1155/2009/181360

Research Article
Dynamical Aspects of an Equilateral Restricted
Four-Body Problem

Martha Alvarez-Ramrez1 and Claudio Vidal2

1
Departamento de Matematicas, UAM-Iztapalapa, A.P. 55-534, 09340 Iztapalapa, Mexico, Mexico
2
Departamento de Matematicas, Facultad de Ciencias, Universidad del Bo Bo, Casilla 5-C,
Concepcion, VIII-Region 4081112, Chile

Correspondence should be addressed to Claudio Vidal, [email protected]

Received 26 July 2009; Revised 10 November 2009; Accepted 9 December 2009

Recommended by Tadashi Yokoyama

The spatial equilateral restricted four-body problem ERFBP is a four body problem where a mass
point of negligible mass is moving under the Newtonian gravitational attraction of three positive
masses called the primaries which move on circular periodic orbits around their center of mass
xed at the origin of the coordinate system such that their conguration is always an equilateral
triangle. Since fourth mass is small, it does not aect the motion of the three primaries. In our
model we assume that the two masses of the primaries m2 and m3 are equal to and the mass m1
is 1 2. The Hamiltonian function that governs the motion of the fourth mass is derived and it
has three degrees of freedom depending periodically on time. Using a synodical system, we xed
the primaries in order to eliminate the time dependence. Similarly to the circular restricted three-
body problem, we obtain a rst integral of motion. With the help of the Hamiltonian structure,
we characterize the region of the possible motions and the surface of xed level in the spatial as
well as in the planar case. Among other things, we verify that the number of equilibrium solutions
depends upon the masses, also we show the existence of periodic solutions by dierent methods
in the planar case.

Copyright q 2009 M. Alvarez-Ramrez and C. Vidal. This is an open access article distributed
under the Creative Commons Attribution License, which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work is properly cited.

1. Introduction
Dynamical systems with few bodies three have been extensively studied in the past, and
various models have been proposed for research aiming to approximate the behavior of real
celestial systems. There are many reasons for studying the four-body problem besides the
historical ones, since it is known that approximately two-thirds of the stars in our Galaxy
exist as part of multistellar systems. Around one-fth of these is a part of triple systems,
while a rough estimate suggests that a further one-fth of these triples belongs to quadruple
or higher systems, which can be modeled by the four-body problem. Among these models,
the conguration used by Maranhao 1 and Maranhao and Llibre 2, where three point
masses form at any time a collinear central conguration Euler conguration, see 3, is
2 Mathematical Problems in Engineering

of particular interest not only for its simplicity but mainly because in the last 10 years, an
increasing number of extrasolar systems have been detected, most of them consisting of a
sun and a planet or of a sun and two planets.
We study the motion of a mass point of negligible mass under the Newtonian
gravitational attraction of three mass points of masses m1 , m2 , and m3 called primaries
moving in circular periodic orbits around their center of mass xed at the origin of the
coordinate system. At any instant of time, the primaries form an equilateral equilibrium
conguration of the three-body problem which is a particular solution of the three-body
problem given by Lagrange see 4 or 3. Two of these primaries have equal masses and
are located symmetrically with respect to the third primary.
We choose the unity of mass in such a way that m1  1 2 and m2  m3  are the
masses of the primaries, where 0, 1/2. Units of length and time are chosen in such a
way that the distance between the primaries is one.
For studying the position of the innitesimal mass, m4 , in the plane of motion of
the primaries, we use either the sideral system of coordinates, or the synodical system of
coordinates see 5 for details. In the
synodical coordinates, the three point masses m1 ,
m2 , and m3 are xed at  3, 0, 0,  3/21 2, 1/2, 0, and  3/21 2, 1/2, 0,
respectively. In this paper, the equilateral restricted four-body problem shortly, ERFBP consists
in describing the motion of the innitesimal mass, m4 , under the gravitational attraction of the
three primaries m1 , m2 , and m3 . Maranhaos PhD thesis 1 and the paper 2 by Maranhao
and Llibre studied a restricted four body problem, where three primaries rotating in a xed
circular orbit dene a collinear central conguration.
In the ERFBP, the equations of motion of m4 in synodical coordinates x, y, z are

x 2y  x ,

y 2x  y , 1.1

z  z ,

where

 1 2 1 2
 x, y, z  x y2 ,
2 1 2 3

 2
1  x 3 y2 z2 , 2

! 2   1.2
! 3 1 2
 " x 1 2 y z2 ,
2 2

! 2  
! 3 1 2
3  " x 1 2 y z2 .
2 2

We remark that the ERFBP becomes the central force problem when  0, and m1  1 is
situated in the origin of the system, while  1/2 results in the restricted three-body problem
with the bodies m2 and m3 of mass 1/2.
Mathematical Problems in Engineering 3

Our paper is organized as follows: Section 2 is devoted to describing the most

important dynamical phenomena that governs the evolution of asteroid movement and states
the problem under consideration in the present study. In Section 3 reductions of the problem
are discussed and a comprehensive treatment of streamline analogies is given. Section 4 is
devoted to the principal qualitative aspect of the restricted problemthe surfaces and curves
of zero velocity, several uses of which are discussed. The regions of allowed motion and
the location and properties of the equilibrium points are established. We describe the Hill
region. The description of the number of equilibrium points is given in Section 5, and in the
symmetrical case i.e.,  1/3, we describe the kind of stability of each equilibrium. In
Section 6, the planar case is considered. There, we prove the existence of periodic solutions
as a continuation of periodic Keplerian orbits, and also when the parameter is small and
when it is close to 1/2. Finally, in Section 8 we present the conclusions of the present work.
Next, we will enunciate some four-body problem that has been considered in the
literature. Cronin et al. in 6, 7 considered the models of four bodies where two massive
bodies move in circular orbits about their center of mass or barycenter. In addition, this
barycenter moves in a circular orbit about the center of mass of a system consisting of these
two bodies and a third massive body. It is assumed that this third body lies in the same
plane as the orbits of the rst two bodies. The authors studied the motion of a fourth body of
small mass which moves under the combined attractions of these three massive bodies. This
model is called bicircular four-body problem. Considering this restricted four-body problem
consisting of Earth, Moon, Sun, and a massless particle, this problem can be used as a model
for the motion of a space vehicle in the Sun-Earth-Moon system. Several other authors have
considered the study of this problem, for example, 811 and references therein. The quasi-
bicircular problem is a restricted four body problem where three masses, Earth-Moon-Sun,
are revolving in a quasi-bicircular motion i.e. a coherent motion close to bicircular also has
been studied, see 12 and references therein. The restricted four-body problem with radiation
pressure was considered in 13, while the photogravitational restricted four body problem
was considered in 14.

2. Statement of the Problem

It is known that equilateral congurations of three-bodies with arbitrary masses m1 , m2 , and
m3 on the same plane, moving with the same angular velocity, form a relative equilibrium
solution of the three-body problem see e.g., 4 or 3. More precisely, we consider three
particles of masses m1 , m2 , and m3 called primaries each describing, at any instant, a circle
around their center of masses which is xed at the origin, with the same angular velocity
and such that its conguration at any instant is an equilateral triangle see Figure 1. Now,
we consider an innitesimal particle m4 attracted by the primaries m1 , m2 , and m3 according
to Newtons gravitational law. Let r be the position vector of m4 .
The equations of motion can be written as


r  U, 2.1

where   denotes derivative with respect to t and

m1 m2 m3
U  Ur; t, m1 , m2 , m3   , 2.2
r r1 t r r2 t r r3 t
4 Mathematical Problems in Engineering

m4

m2
m1

x3  0
m3

Figure 1: The equilateral restricted four body problem in inertial coordinates.

with r1 t, r2 t, and r3 t representing the position of each primary, respectively. To remove
the time dependence of the system 2.1, we consider the orthonormal moving frame in R3 ,
given by {e1 , e2 , e3 } where

e1  e1 t  eit , e2  e2 t  ie1 , e3  e3 t  0, 0, 1 2.3

with i2  1. This orthonormal moving frame corresponds to the synodical system. Then,
2.1 can be written as

U
x1 2x2 2 x1  ,
x1
U
x2 2x1 2 x2  , 2.4
x2
U
x3  ,
x3

where

m1 m 2 m 3
U  Ux1 , x2 , x3   ,
d1 d2 d3
  2
d1  x1 1 2 x2 1 x32 ,
 2.5
2  2
d2  x1 2  x2 2 x32 ,
  2
d3  x1 3 2 x2 3 x32 ,

where rj t  eit j , with j  j ij for j  1, 2, 3. Applying the above notation, we can write
r  x1 ix2 e1 x3 e3 , r1  1 e1 , r2  2 e1 , r3  3 e1 , and so r rj  x1 ix2  x3 e3 j for
j  1, 2, 3.
Mathematical Problems in Engineering 5

We perform the reparametrization of time d  dt, then the system 2.4 is

transformed into

1 W
x1 2x2 x1  ,
2 x1
1 W
x2 2x1 x2  , 2.6
2 x2
1 W
x3  ,
2 x3

where the dot denotes the derivative with respect to , and the potential W is given by

m 1 m 2 m3
W  Wx1 , x2 , x3   2.7
1 2 3

with

  2
1  x1 1 2 x2 1 x32 ,
  2
2  x1 2 2 x2 2 x32 , 2.8
  2
3  x1 3 2 x2 3 x32 .

If we dene 1  m1 /M, 2  m2 /M, and 3  m3 /M, where M  m1 m2 m3 , the equations

of motions 2.4 become

M W
x1 2x2 x1  ,
2 x1
M W
x2 2x1 x2  , 2.9
2 x2
M W
x3  ,
2 x3

where

M1 M2 M3
W  Wx1 , x2 , x3   . 2.10
1 2 3

For simplicity, we will consider an equilateral triangle of side 1 and so we obtain that
M/2  1.
6 Mathematical Problems in Engineering

x3 m4

m3  m2 
x2

m1  1 2

x1

Figure 2: The equilateral restricted four body problem in a rotating frame.

3. Equations of Motion and Preliminary Results

From 2.9, we deduce that the equations of motion of the ERFBP in synodical coordinates
are given by the system of dierential equations

x1 2x2  x1 ,

x2 2x1  x2 , 3.1

x3  x3 ,

where

1 2
 x1 , x2 , x3   x1 x22 Wx1 , x2 , x3 ,
2
3.2
1 2
W  Wx1 , x2 , x3   ,
1 2 3

with

 2
1  x1 3 x22 x32 ,

! 2  
! 3 1 2
2  " x1 1 2 x2 x32 ,
2 2 3.3

! 2  
! 3 1 2
3  " x1 1 2 x2 x32 .
2 2

Analogously to the circular three-body problem, we can verify that the system 3.1
possesses a rst Jacobi type integral given by

1 2
C x1 x22 x32 x1 , x2 , x3 . 3.4
2

Thus we have the following result.

Mathematical Problems in Engineering 7

Proposition 3.1. The Jacobi-type function 3.4 is a rst integral of the ERFBP for any value of .

Proof. Dierentiating 3.4 with respect to the time, we get

dC W W W
 x1 x1 x2 x2 x3 x3 x1 x1 x2 x2 x1 x2 x3 , 3.5
dt x1 x2 x3

and using 2.9 we can reduce the obtained expression to

   
dC W W W
 x1 x1 2x2 x2 x2 2x1 x3
dt x1 x2 x3
3.6
W W W
x1 x1 x2 x2 x1 x2 x3  0.
x1 x2 x3

Hence C is a constant of motion.

In order to write the Hamiltonian formulation of the ERFBP we introduce the new
variables

x  x1 , y  x2 , z  x3 ,
3.7
X  x y, Y  y x, Z  z.

Hence, it is veried that system 3.1 is equivalent to an autonomous Hamiltonian system

with three degrees of freedom with Hamiltonian function given by

 1 2 
H  H x, y, z, X, Y, Z  X Y 2 Z2 yX xY W. 3.8
2

Therefore, the Hamiltonian system associated is

x  y X, X  Y Wx

y  x Y, Y  X Wy 3.9

z  Z, Z  Wz .

Of course, the phase space where the equations of motion are well dened is
)  )  
  3 1
M x, y, z, X, Y, Z R \
3
3, 0, 0 , 1 2 , , 0 ,
2 2
 * * 3.10
3 1
1 2 , , 0 R3 ,
2 2

where the points that have been removed correspond to binary collisions between the
massless particle and one of the primaries.
8 Mathematical Problems in Engineering

Additionally, the spatial ERFBP admits the planar case as a subproblem, that is, z 
Z  0 is invariant under the ow dened by 3.9.
On the other hand, we see that there are two limiting cases in the ERFBP, which we
described below.
a If  0, we obtain a central force problem, with the body of mass m1  1 at the origin
of the coordinates.
b If  1/2, we obtain the circular restricted three-body problem, with masses m2  m3 
1/2.
Note that  1/3 corresponds to the symmetric case, that is, where the masses of the
primaries are all equal to 1/3.
It is easily seen that the equations of motion 3.9 are invariant by the symmetry
 
S: x, y, z, X, Y, Z, x, y, z, X, Y, Z, . 3.11

This means that if   x, y, z, X, Y , Z is a solution of the system 3.9,
then t  x, y, z, X, Y , Z is also a solution. We note that
this symmetry corresponds to a symmetry with respect to the xz-plane. In the planar case,
the symmetry corresponds to symmetry with respect to the x-axis.

4. Permitted Regions of Motion

In this section, we will see that the function x, y, z allows us to establish regions in the
x, y, z space, where the motion of the innitesimal particle could take place. We will use
similar ideas to those developing in 15, 16.
By using 3.4, the surface of zero velocity is dened by the set
 
RC : x, y, z R3 such that x, y, z  C, for any level C. 4.1

This set
corresponds to the so-called Hill region. We note that C 0 implies RC 
R3 \ { 3, 0, 0,  3/21 2, 1/2, 0,  3/21 2, 1/2, 0}. That is, the region of all
possible motions is given by the whole phase space and so the innitesimal particle is free to
move; in particular escape solutions are permitted.
In the spatial case, the surfaces that separate allowed and nonallowed motions are
called zero-velocity surfaces, and for the planar case the set that separates the allowed and
nonallowed motions is called zero-velocity curve. The shape and size of zero velocity sets
C  x, y, z depend on C and . They correspond to the boundary of the Hill regions. The
zero-velocity set RC  is dened by the equation

1 2 1 2
 x y 2  
2 2  2  2
x 3 y2 z2 x  3/2 1 2 y 1/2 z2

  C,
 2  2
x  3/2 1 2 y 1/2 z2
4.2
Mathematical Problems in Engineering 9

10 10

5 5

0 0
z z

5 5
2 2
1 1
10 0 y 10 0 y
1 1
2 1 2 1
0 1 2 0 1 2
2 2
x x
a b

10

z 0

5
2
10 1
2 0
1
0 1 y
x
1 2 2

c

Figure 3: Evolution of zero-velocity surface in the three-dimensional ERFBP for  1/3. a C  1/4. b
C  1/2. c C  3/4.

only for C < 0 and any value of . Next, we give a list of all possible situations that may
appear when this condition is fullled.

1 z on the RC in which case x2 y2 2C, this means,that around the

z-axis the variables x, y must be asymptotic to a circle of radius 2C.

2 x or resp., y or  on the RC , when C .

3 For |C| very large this implies that x, y can be suciently close to one of the
primaries, or the innitesimal mass is close to innity.

4 Since x2 y2 is a factor of on RC , then small values for C are not allowed.
10 Mathematical Problems in Engineering

10

1
5
0.5
z 0 z 0

0.5
5
2
1 2
1 2
0 y 1
10 1 1
2 0 0
2 1 0 x 1 y
1 2 1
x 2 2
a b
2

y
0

2
2 1
0 1 1
x 2 0.5
0
0.5
1
c

Figure 4: Evolution of zero-velocity surface in the three dimensional ERFBP for  1/3. a C  1. b
and c C  1.6 under dierent points of view.

By simplicity, we will only show zero-velocity surfaces for the case  1/3 and
dierent values of the integral of motion C. Figures 3, 4, 5, and 6 show evolution of zero-
velocity surfaces for several C values.

4.1. The Planar Case

As we mentioned in last section, the set {z  Z  0} is invariant under the ow, and so
the motion of the innitesimal body lies on the xy plane that contains the primaries. In
Figure 7, we show the evolution of the function in the planar case for dierent values
of the parameter .
Next we show the evolution of the Hills regions as well as the zero velocity curves,
for  1/3 and many values of the Jacobian constant C; the permissible areas are shown on
Figures 8, 9, 10, and 11 shading.
In Figure 12, we show the behavior of level curves in the planar case for some values
of and for dierent energy levels.
Mathematical Problems in Engineering 11

1.5
1
0.5
2 y
0
0.5
1
1
1.5
z 0 1.5 1
0.5
0 0.5 1
1 1.5
3
2
1
2 0
3 2 1
1 2 y 1 z
0 1 2 3
x 3 2
a b

0.4

0.2

z 0

0.2
1
0.5
0.4 0 y
0.5
0.5 1
0 0.5 1
x
c

Figure 5: Evolution of zero-velocity surface in the three dimensional ERFBP for  1/3. All cases
correspond to C  1.7 under dierent points of view.

3
2 4
1
3 0 2
1 y 0
2
2
1 3 2
0 3 4
2
y 1 y 1 4 2
0 0 2
2 3 x 1 4 3
3 2
0 1 2 1 z 3 3 x
3 1 1 z
3 2 1 1 z
x
3 1
1
3 3
a b c

Figure 6: Evolution of zero-velocity surface in the three dimensional ERFBP for  1/3. a C  2. b
C  3. c C  5.
12 Mathematical Problems in Engineering

 0.266318
0  1/16

3.5 3
1 2.5 1 1
2.5 2.5
0.5 2 0.5 2 0.5
1.5 1.5 1.5
1 0 1 0 1 0
0.5 0.5 0.5
0 0.5 0 0.5 0 0.5
0.5 0.5 0.5
1 1 1 1 1 1
a b c
 1/3  0.4  1/2

2.5 3
1 2.5 1 2.5 1
2 0.5 2 0.5 2 0.5
1.5 1.5 1.5
1 0 1 0 1 0
0.5 0.5 0.5
0 0.5 0 0.5 0 0.5
0.5 1 0.5 1 0.5 1
1 1 1
d e f

Figure 7: Evolution of the graph of x, y on the xy plane for dierent values of .

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0

0.5 0.5 0.5

1 1 1

1.5 1.5 1.5

1.5 1 0.5 0 0.5 1 1.5 1.5 1 0.5 0 0.5 1 1.5 1.5 1 0.5 0 0.5 1 1.5
a b c

Figure 8: Evolution of zero-velocity curves and Hills region in the planar ERFBP for  1/3, where
shading represents permissible areas. a C  1.6, b C  1.6775, c C  1.6795.

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0

0.5 0.5 0.5

1 1 1

1.5 1.5 1.5

1.5 1 0.5 0 0.5 1 1.5 1.5 1 0.5 0 0.5 1 1.5 1.5 1 0.5 0 0.5 1 1.5
a b c

Figure 9: Evolution of zero-velocity curves and Hills region in the planar ERFBP for  1/3, where
shading are permissible areas. a C  1.7, b C  1.75, c C  1.765.
Mathematical Problems in Engineering 13

2 2 2

1 1 1

0 0 0

1 1 1

2 2 2
2 1 0 1 2 2 1 0 1 2 2 1 0 1 2
a b c

Figure 10: Evolution of zero-velocity curves and Hills region in the planar ERFBP for  1/3, where
shading are permissible areas. a C  1.775. b C  1.8. c C  2.

4 1
2
2 0.5
1

0 0 0

1
2 0.5
2
4 1
2 1 0 1 2 4 2 0 2 4 1 0.5 0 0.5 1
a b c

Figure 11: Evolution of zero-velocity curves and Hills region in the planar ERFBP for  1/3, where
shading are permissible areas. a C  3, b and c both gures correspond to C  5 with dierent
window size.

5. Equilibrium Solutions
It is veried that the equilibrium solutions of the system 3.9 or equivalently 3.1 are given
by the critical points of the function  x, y, z or simply they are the solutions of the
following system of equations:

 

 x 3 3 1 1
1 2 x 1 2  x,
13 2 23 33
 
 y y 1/2 y 1/2
1 2 3  y, 5.1
1 23 33
 
1 2
3 3 z  0.
13 2 3

From the last equation we see that the coordinate z must be zero, so the critical points are
restricted to the plane xy, and are given by the solutions of the rst two equations.
14 Mathematical Problems in Engineering

2 2 2

1 1 1

0 0 0

1 1 1

2 2 2
2 1 0 1 2 2 1 0 1 2 2 1 0 1 2
0  1/16  0.266318

a b c

2 2 2

1 1 1

0 0 0

1 1 1

2 2 2
2 1 0 1 2 2 1 0 1 2 2 1 0 1 2
 1/3  0.4  1/2

d e f

Figure 12: Energy level curves for some values of the parameter in the planar case.

It is known see 17 that the number of equilibrium solutions of the system 5.1 is 8,
9 or 10 depending on the values of the masses, m1 , m2 and m3 which must be positive. Six of
them are out of the symmetry axis i.e., out of the x-axis, therefore on the axis of symmetry
we must have 2, 3 or 4. From the analysis done it follows that the number of the equilibrium
solutions depends on the parameter . This implies that nding the critical points is a non-
trivial problem, and this is one of the main dierences with the problem studied by Maranhao
in his doctoral thesis 1, because there, the number of critical points did not depend on the
parameter .
The critical points on the axis y  0 are the zeros of the function

   

 x 3 x 3/2 1 2
F x  1 2  3 2 x, 5.2
x 3 23

where

! 2
  ! 3
 " 1 5.3
1  x 3, 2  3  x 1 2 .
2 4
Mathematical Problems in Engineering 15

Table 1: Number of critical points on the x-axis.

0 0 < <  < < 1/2  1/2

2 2 3 4 3

30
5
25
4
20
3
15
2 10

1 5
0 0
2 1.5 1 0.5 0 0.8 1 1.2 1.4 1.6 1.8
x x
a b

Figure 13: Graph of F1/3 . a x < 3/3. b x > 3/3.

An explicit computation shows that in the limit case problems the number of
equilibrium points corresponding to the system 5.1 is as follows.

a The function 5.2 with  0 results in

x
F0 x  x 5.4
|x|3

whose zeros are x  1 and x  1, and so there are two equilibrium points.

b Taking  1/2 in 5.2 becomes

x
F1/2 x  x 5.5
x2 1/43/2

with zeros given by x   3/2, x  0 and x   3/2. We conclude that there are
three equilibrium points.

From numerical simulations we get that the number of critical points along the xaxis
is given in Table 1. Observe that : 0.266318 is the bifurcation value.
In the symmetric case when all the masses are equals i.e.,  1/3 we have that the
graph of F1/3 is similar to the one shown in Figure 13. As a consequence, there are exactly 4
equilibrium solutions on the x-axis, and therefore there are exactly 10 equilibrium solutions.
Of course, 0, 0, 0 is an equilibrium solution.
16 Mathematical Problems in Engineering

In general for any equilibrium solution of the form x0 , y0 , 0, the linearized system
3.9 in the planar case give us that the characteristic polynomial is
    
CA   2 Wzz x0 , y0 , 0 4 2 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 2
    
1 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 5.6

Wxy
2
x0 , y0 , 0 ,

whose roots are

  1
 Wzz x0 , y0 , 0 ,  , 5.7
2

where is given by

 4 2a c 2 a c2 4b2 8a c 5.8

with a  Wxx x0 , y0 , 0, b  Wxy x0 , y0 , 0 and c  Wyy x0 , y0 , 0. A very simple result is the
following.

Lemma 5.1. The roots of p  2 A B are real and negative if and only if A > 0, B > 0 and
 A2 4B 0.

Associating to our characteristic polynomial 5.6 we have

 
A  2 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 ,
     5.9
B  1 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 Wxx x0 , y0 , 0 Wyy x0 , y0 , 0 Wxy
2
x0 , y0 , 0 .

Now, we remark that

 
 1 2
Wzz x0 , y0 , 0  3 3 < 0. 5.10
13 2 3

Consequently we have the following result:

Corollary 5.2. In the spatial ERFBP for any equilibrium solution x0 , y0 , 0 we ( have at least two
pure imaginary eigenvalues associated to the linear part, which are given by  Wzz x0 , y0 , 0i.

From this corollary we deduce that to study the nonlinear stability in the Lyapunov
sense of each equilibrium solution of the spatial ERFBP is not a simple problem, because
we need to take into account the existence or not of resonance in each situation. Leandro
in 17 studied the spectral stability in some situations according to the localization of the
equilibrium solution along the symmetry-axis. In a future work we intend to study the
Lyapunov stability of each equilibrium.
Mathematical Problems in Engineering 17

5.1. Analysis of the Symmetrical Case,  1/3

As we have said previously in the symmetrical case i.e.,  1/3 there
are 10 equilibrium

solutions and one of them is 0, 0, 0. Here we have 1  2  3  1/ 3, a  b  3 3/2 and
c  0. Consequently, the characteristic roots are

 
1  3 3i, 2  3 3i,
   
1 1
3  6 3 4 4 6 3, 4  6 3 4 4 6 3, 5.11
2 2
   
1 1
5  6 3 4 4 6 3, 6  6 3 4 4 6 3.
2 2

Therefore, we have the following result.

Corollary 5.3. In the symmetrical spatial ERFBP the equilibrium solution 0, 0, 0 is unstable in the
Lyapunov sense.

In general, it is possible to prove that the equilibrium solutions on the xaxis are
x1  0.9351859666722429, x2  0.23895830919534947 and x3  1.1799984048894328, and
by symmetry it follows:

Corollary 5.4. In the symmetrical spatial ERFBP all the equilibrium solutions are unstable in the
Lyapunov sense.

According to 17 we have the following corollary.

Corollary 5.5. In the symmetrical planar ERFBP all the equilibrium solutions are unstable in the
Lyapunov sense.

6. Continuation of Periodic Solutions in the Planar Case

In this section we prove the existence of periodic solutions in the ERFBP for suciently
small in the planar case and by the use of the Lyapunov Center Theorem when is close to
1/2. In order to nd periodic orbits of our problem we will use the continuation method
developed by Poincare which is one of the most frequently used methods to prove the
existence of periodic orbits in the planar circular restricted three-body problem see 15.
This method has been also used by other authors in dierent problems. In Meyer and Hall 5,
we nd a good discussion of the Poincare continuation method to dierent n-body problem
see also 18.
In our approach we will continue circular and elliptic solutions of the Kepler problem
with the body xed in the origin of the system with mass 1. We know that all the orbits of the
Kepler problem with angular momentum zero are collision orbits with the origin. We assume
that
the angular momentum
is not zero and we study the orbits that have positive distance of
 3/2, 1/2 and  3/2, 1/2. In the following lemma we resume the kind of orbits that
we will consider.
18 Mathematical Problems in Engineering

Lemma 6.1. Fixed a > 0 there exists a nite number of elliptic orbits with semi-major axis a, such
that its trajectories
are periodicin the rotating system and pass through the singularity of the other
primaries  3/2, 1/2 or  3/2, 1/2.

The proof of this lemma can be found in 19.

6.1. Continuation of Circular Orbits

In this section we show that circular solutions of the unperturbed Kepler problem can be
continued to periodic solutions of the ERTBP for small values of . We introduce the polar
coordinates given as x  r cos , y  r sin , thus x  r cos r sin and y  r sin r cos .
So, X  r cos r 1 sin and Y  r sin r 1 cos , consequently X 2 Y 2  r 2
2
r 2  1 and yX xY  r 2  1. Thus, the Hamiltonian 3.8 now is

 2
r 2 r 2 1  6.1
H r 2 1 V r, ,
2

where

 
1 2 1 1
V r,   , 6.2
1 2 3

where

 2

1  r cos 3 r 2 sin2 ,

 2
2  r cos
3/2 1 2 r sin 1/22 , 6.3

 2
3  r cos 3/2 1 2 r sin 1/22 .

The new coordinates are not symplectic. In order to obtain a set of symplectic coordinates
r, , R,  we dene R  r radial velocity in the sideral system and  r 2  1 angular
momentum in the sideral system, then H is

R2 /r 2
H V r, . 6.4
2

When  0 we have that

R2 2 /r 2 1
H , 6.5
2 r
Mathematical Problems in Engineering 19

is the Hamiltonian of the Kepler problem in polar coordinates. So, if is a small parameter,
the Hamiltonian 3.8 assumes the form

R2 2 /r 2 1 
H O . 6.6
2 r

For  0, the Hamiltonian system associated is

2 1
r  R, R  ,
r3 r2
6.7

 2 1,  0.
r

Let  c be a xed constant. For c /

 1, the circular orbit R  0, r  c2 is a periodic
solution with period |2c /1 c |. Linearizing the r and R equations about this solution
3 3

gives

r  R, R  c6 r, 6.8

which has solutions of the form expit/c3 , and so the nontrivial multipliers of the circular
orbits are expi2/1 c3  which are not 1, provided 1/1 c3  is not an integer. Thus we
have proved the following theorem see details in 5.

Theorem 6.2. If c /  1 and 1/1 c3  is not an integer, then the circular orbits of the Kepler problem
in rotating coordinates with angular momentum c can be continued into the equilateral restricted four
body problem for small values of .

6.2. Continuation of Elliptic Orbits

In Section 3, we saw that the ERFBP has the S-symmetry which when exploited properly
proves that some elliptic orbits can be continued from the Kepler problem. The main idea is
given in the following lemma, which is a consequence of the uniqueness of the solution of the
dierential equations and the symmetry of the problem.

Lemma 6.3. A solution of the equilateral restricted problem which crosses the line of syzygy (the x-
axis) orthogonally at a time t  0 and later at a time t  T/2 > 0 is T -periodic and symmetric with
respect to the line syzygy.

That is, if xt and yt is a solution of the equilateral restricted four body problem
such that y0  x0  yT/2  xT/2  0, where T > 0, then this solution is T -periodic
and symmetric with respect to the x-axis.
In Delaunay variables l, g, L, G, an orthogonal crossing of the line of sizygy at a time
t0 is

lt0   n, gt0   m, n, m integers. 6.9

20 Mathematical Problems in Engineering

These equations will be solved using the Implicit Function theorem to yield the following
theorem see details in 5.

Theorem 6.4. Let m, k be relatively prime integers and T  2m. Then the elliptic T -periodic
solution of the Kepler problem in rotating coordinates which satises

m
l0  , g0  , L3 0  6.10
k

and does not go through  3/2, 1/2 and  3/2, 1/2 can be continued into the equilateral
restricted four body problem for small . This periodic solution is symmetric with respect to the line of
syzygy.

Proof. The Hamiltonian of the ERFBP in Delaunay coordinates for suciently small is

1 
H G O , 6.11
2L2

and the equations of motion are

1  
l  3
O , L  0 O ,
L 6.12
 
g  1 O , G  0 O .

Let L30  m/k, and let lt, , , gt, , , Lt, ,  and Gt, ,  be the solution which goes
through l  , g  , L  , G arbitrary at t  0; so, it is a solution with an orthogonal
crossing of the line of syzygy at t  0.
From 6.12 lt, , 0  t/3 , gt, , 0  t . Thus, lT/2, L0 , 0  1 k and
gT/2, L0 , 0  1 m, and so when  0, this solution has another orthogonal crossing at
time T/2  m. Also,

l l  1/3
k k4
t 3
m m
det


 det

/
 0. 6.13

g g
t tT/2, LL0 , 0 1 0

Thus, the theorem follows by the Implicit Function theorem.

6.3. Application of the Lyapunov Center Theorem

For  1/2, we have three equilibrium solutions on the x-axis which are P1   ( 3/2, 0,

P2  0,
( 0 and P   3/2, 0. At the point P , the associated eigenvalues are 75 8 2
3 2
and 75 8 2. Therefore, this equilibrium is unstable and by Lyapunovs Center Theorem
see 5, we obtain the following theorem.
Mathematical Problems in Engineering 21

y
4

x
4 2 2 4

Figure 14: Circular orbit for c  2 with  102 .

y
y y
1
0.5 0.5
0.5

x
1 0.5 0.5 1 1.5
x x
0.5 0.5 0.5 0.5
0.5

1
0.5 0.5
1.5

a b c

Figure 15: The circular orbit associated to c  9/10 for  0 continued to  104 which is not circular for
 102 .

Theorem 6.5. There exists a one-parameter family of periodic orbits of the ERFBP emanating from
the Euler equilibrium (for (1/2). Moreover, when approaching the equilibrium point along the

family, the periods tend to 2/ 3 8 2.

7. Numerical Results
In the Section 8, we established theorems on the continuation of periodic solutions from
the Keplers problem in rotating coordinates to the ERFBP. In this section, we present some
numerical experiments that illustrates the thesis of Theorem 6.2.
To nd those circular orbits we rst selected an angular momentum c such that c  /1
and 1/1 c3  / Z. By varying c we generated a set of initial conditions for Kepler problem
in rotating coordinates given by the system 3.9 taking  0. We have chosen y0  0 and
X0  0 for all orbits, ensuring that we were following a family of symmetric orbits; we have
taken into account the fact that circular orbits satisfy r  c2 .
We have noticed that for values of c  2, 3, 4, 5, 6, 7, 8, 9, 10 with  102 the orbit is
close to the circular orbit, see Figure 14.
22 Mathematical Problems in Engineering

However, the circular orbits associated to c 0, 1 is close to the circular orbit if 104 ,
for instance c  9/10 can be continued for small and of the order 104 but not for higher
values. The orbits obtained as a consequence of numerical simulations are shown in Figure 15.

8. Conclusions and Final Remarks

The spatial equilateral restricted four-body problem ERFBP is considered. This model
of four-body problem, we have that three masses, moving in circular motion such that
their conguration is always an equilateral triangle, the fourth mass being small and not
inuencing the motion of the three primaries. In our model we assume that two masses of
the primaries m2 and m3 are equal to and the mass m1 is 1 2. In a synodical systems of
coordinates the dynamics obeys to the system of dierential equations

x 2y  x ,

y 2x  y , 8.1

z  z ,

where

 1 2 1 2
 x, y, z  x y2 ,
2 1 2 3

 2
1  x 3 y2 z2 ,

! 2   8.2
! 3 1 2
2  " x 1 2 y z2
2 2

! 2  
! 3 1 2
3  " x 1 2 y z2 .
2 2

In Section 4 it is devoted to give the principal qualitative aspect of the restricted

problemthe surfaces and curves of zero velocity, several uses of which are discussed.
The regions of permitted motion and the location and properties of the equilibrium points
are established. We describe the Hill region. The description of the number of equilibrium
points is given in Section 5, and in the symmetrical case i.e.,  1/3 we are describing
the kind of stability of each equilibrium. In Section 6 the planar case is considered. Here, we
prove the existence of periodic solutions as continuation of periodic Keplerian orbits, when
the parameter is small and when it is close to 1/2. Finally, in Section 7 we present some
numerical experiments that illustrates the thesis of theorem concerning with the continuation
of circular orbits of the Kepler problem to the ERFBP with small enough.
In a work in progress we intend to continue the study of the ERFBP in dierent aspects
of its dynamics. For example, the behavior of the ow near the singularities collisions. The
study of the escapes solutions i.e., the unbounded solutions. Existence of chaos under the
Mathematical Problems in Engineering 23

construction of a shift map. We desired to get periodic solutions under the use of numerical
methods.

Acknowledgments
The rst author was partially supported by CoNaCyT Mexico Grant 32167-E. The authors
would like to thank the anonymous referees for their very careful reviews of the paper that
include many important points and will improve signicantly the clarity of this paper.

References
1 D. Maranhao, Estudi del ux dun problema restringt de quatre cossos, Ph.D. thesis, UAB, Barcelona, Spain,
1995.
2 D. L. Maranhao and J. Llibre, Ejection-collision orbits and invariant punctured tori in a restricted
four-body problem, Celestial Mechanics & Dynamical Astronomy, vol. 71, no. 1, pp. 114, 1998/99.
3 A. Wintner, The Analytical Foundations of Celestial Mechanics, vol. 5 of Princeton Mathematical Series,
Princeton University Press, Princeton, NJ, USA, 1941.
4 R. H. Battin, An Introduction to the Mathematics and Methods of Astrodynamics, AIAA Education Series,
AIAA, Reston, Va, USA, 1999.
5 K. R. Meyer and G. R. Hall, Introduction to Hamiltonian Dynamical Systems and the N-Body Problem, vol.
90 of Applied Mathematical Sciences, Springer, New York, NY, USA, 1992.
6 J. Cronin, P. B. Richards, and L. H. Russell, Some periodic solutions of a four-body problem. I, Icarus,
vol. 3, pp. 423428, 1964.
7 J. Cronin, P. B. Richards, and I. S. Bernstein, Some periodic solutions of a four-body problem. II,
Icarus, vol. 9, pp. 281290, 1968.
8 C. Simo, G. Gomez, A. Jorba, and J. Masdemont, The bicircular model near the triangular libration
points of the RTBP, in From Newton to Chaos (Cortina dAmpezzo, 1993), vol. 336 of NATO Advanced
Science Institutes Series B: Physics, pp. 343370, Plenum, New York, NY, USA, 1995.
9 A. Jorba, On practical stability regions for the motion of a small particle close to the equilateral points
of the real earth-moon system, in Proceedings of the 3rd International Symposium (HAMSYS-98) Held
at Patzcuaro, vol. 6 of World Scientic Monograph Series in Mathematics, pp. 197213, World Scientic,
River Edge, NJ, USA, 2000.
10 A. F. B. De Almeida Prado, Numerical and analytical study of the gravitational capture in the
bicircular problem, Advances in Space Research, vol. 36, no. 3, pp. 578584, 2005.
11 A. L. Machuy, A. F. B. A. Prado, and T. J. Stuchi, Numerical study of the time required for the
gravitational capture in the bi-circular four-body problem, Advances in Space Research, vol. 40, no.
1, pp. 118124, 2007.
12 M. A. Andreu, Dynamics in the center manifold around L2 in the quasi-bicircular problem, Celestial
Mechanics & Dynamical Astronomy, vol. 84, no. 2, pp. 105133, 2002.
13 T. J. Kalvouridis, M. Arribas, and A. Elipe, Parametric evolution of periodic orbits in the restricted
four-body problem with radiation pressure, Planetary and Space Science, vol. 55, no. 4, pp. 475493,
2007.
14 T. J. Kalvouridis and K. G. Hadjifotinou, Bifurcations from planar to three-dimensional periodic
orbits in the photo-gravitational restricted four-body problem, International Journal of Bifurcation and
Chaos, vol. 18, no. 2, pp. 465479, 2008.
15 V. Szebehely, Theory of Orbits, Academic Press, New York, NY, USA, 1967.
16 J. Lundberg, V. Szebehely, R. S. Nerem, and B. Beal, Surfaces of zero velocity in the restricted problem
of three bodies, Celestial Mechanics, vol. 36, no. 2, pp. 191205, 1985.
17 E. S. G. Leandro, On the central congurations of the planar restricted four-body problem, Journal
of Dierential Equations, vol. 226, no. 1, pp. 323351, 2006.
18 C. L. Siegel and J. K. Moser, Lectures on Celestial Mechanics, vol. 18 of Die Grundlehren der mathematischen
Wissenschaften, Springer, New York, NY, USA, 1971.
19 R. F. Arenstorf, Periodic solutions of the restricted three-body problem representing analytic
continuations of Keplerian elliptic motions, American Journal of Mathematics, vol. 85, pp. 2735, 1963.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 740460, 24 pages
doi:10.1155/2009/740460

Research Article
Nonsphericity of the Moon and
Near Sun-Synchronous Polar Lunar Orbits

Jean Paulo dos Santos Carvalho,1 Rodolpho Vilhena de Moraes,1

and Antonio Fernando Bertachini de Almeida Prado2
1
UNESPUniversidade Estadual Paulista, CEP 12516-410, Guaratingueta-SP, Brazil
2
Division of Space Mechanics and ControlINPE, CEP 12227-010, Sao Jose dos Campos, SP, Brazil

Correspondence should be addressed to Rodolpho Vilhena de Moraes, [email protected]

Received 30 July 2009; Revised 28 October 2009; Accepted 3 December 2009

Recommended by Tadashi Yokoyama

Herein, we consider the problem of a lunar articial satellite perturbed by the nonuniform
distribution of mass of the Moon taking into account the oblateness J2  and the equatorial
ellipticity sectorial term C22 . Using Lie-Hori method up to the second order short-period terms
of the Hamiltonian are eliminated. A study is done for the critical inclination in rst and second
order of the disturbing potential. Coupling terms due to the nonuniform distribution of mass of
the Moon are analyzed. Numerical simulations are presented with the disturbing potential of rst
and second order is. It an approach for the behavior of the longitude of the ascending node of a
near Sun-synchronous polar lunar orbit is presented.

Copyright q 2009 Jean Paulo dos Santos Carvalho et al. This is an open access article distributed
under the Creative Commons Attribution License, which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work is properly cited.

1. Introduction
In this paper, we consider the problem of a lunar articial satellite of low altitude taking
into account the oblateness J2  and the equatorial ellipticity sectorial term C22  of the
Moon. The Lie-Hori 1 perturbation theory method up to the second order is applied to
eliminate the short-period terms of the disturbing potential. The perturbation method up
to the second order is applied to analyze coupling terms. In this work, the long-period
term of the disturbing potential is analyzed. A formula is developed to compute the critical
inclination when the perturbations due to the nonsphericity of the Moon as a function of the
terms of the zonal and sectorial harmonics occur.
An approach is done for a special type of orbit, denominated Sun-synchronous orbit of
Moons articial satellites. The Sun-synchronous orbit is a particular case of an almost polar
orbit. The satellite travels from the North Pole to the South Pole and vice versa, but its orbital
2 Mathematical Problems in Engineering

Table 1: Magnitude orders for J2 and C22 .

C20 J2 C22
Earth 103 2 106
Moon 2 104 2 105

plane is always xed for an observer that is posted in the Sun. Thus the satellite always passes
approximately on the same point of the surface of the Moon every day in the same hour. In
such a way the satellite can transmit all the data collected for a lunar xed antenna, during
its orbits. An analysis of Sun-synchronous orbits considering the nonuniform distribution of
mass of the Moon is done for the longitude of the ascending node with an approach based on
Park and Junkins 2.
In 35 an analytical theory, is developed to study the orbital motion of lunar articial
satellites using the method of transformation of Lie 6, 7 as a perturbation method. The main
perturbation is due to the nonspherical gravitational eld of the Moon and the attraction of
the Earth. The disturbing body is in circular orbit with the disturbing function developed in
polynomial of Legendre up to the second order. In 811 an analytical theory, is developed
with numerical applications taking into account the nonuniform distribution of mass of the
Moon and the perturbation of the third body in elliptical orbit Earth is considered. The
disturbing function is expanded in Legendre associated functions up to the fourth order.
This paper is developed based on 2, 4, where the perturbation theory method of Lie-
Hori up to the second order is used. Our contribution is characterized by a We developed
of a new formula for the critical inclination of second order; b we x g and h to assure the
condition of frozen orbits; c we showed that the coecients J2 and C22 aect the variation of
the eccentricity strongly it aects the eccentricity directly in the second order contributing
especially the C22 term to increasing the variation of the eccentricity mainly for small
inclinations; d the coupled perturbations nonuniform distribution of mass of the Moon J2
and C22  and third-body P2 help to control the variation of the eccentricity for low-altitude
polar orbits; e we presented a new formula to compute inclinations for Sun-synchronous
orbits when it is taking into account the harmonic J2 and C22 in the rst-order potential.
This paper has seven sections. In Section 2, the terms due to nonsphericity of the Moon
are presented while Section 3 is devoted to the Hamiltonian of the system. In Section 4, an
approach concerning the critical inclinations is used. In Section 5, an approach concerning the
Sun-synchronous lunar orbits is used. Numeric applications are done in Section 6. Section 7
is devoted to the conclusions.

2. Nonsphericity of the Moon

Besides the fact that the Moon is much less attened than the Earth, it also causes
perturbations in space vehicles. Table 1 presents orders of magnitude for some zonal and
sectorial harmonics compared with the same parameters for the Earth. The term C20 describes
the equatorial bulge of the Moon, often referred to as the oblateness. The coecient C22
measures the ellipticity of the equator.
The space vehicle is a point of mass in a three-dimensional orbit with orbital elements:
a semimajor axis, e eccentricity, i inclination, argument of periapsis, longitude
of the ascending node, and n mean motion given by the third Keplers law n2 a3  Gm0 ,
where m0 is the mass of the Moon.
Mathematical Problems in Engineering 3

Then, we will present the

Hamiltonian formalism using Delaunay canonical variables

7 dened as L  a, G  L 1 e2 , H  G cos i, l  M mean anomaly, g  argument of
periapsis, and h  longitude of the ascending node.
The force function, the negative of the total energy as used in physics, is given by

F  H  V T, 2.1

here, V is the negative of the potential energy, and T is the kinetic energy. The force function
can be put as 12:

H RT  R, 2.2
r 2a

or

2
H R. 2.3
2L2

the function R, comprising all terms of V except the central term, is known as the disturbing
potential. The term due to the unperturbed potential is given by

2
H0  . 2.4
2L2

Considering the lunar equatorial plane as the reference plane, the disturbing potential
can be written in the form 13 of

  n  2  3 
5
R0  R0  R0 
VM  Jn Pn sin C22 P22 sin cos 2 C31 sin cos ,
r n2 r r r
2.5

where is the Lunar gravitational constant, R0 is the equatorial radius of the Moon R0 
1738 km, Pn represent the Legendre polynomial, Pnm represent the associated Legendre
polynomial, the angle is the latitude of the orbit with respect to the equator of the Moon, the
angle is the longitude measured from the direction of the longest axis of the Moon, where
  22 , since  is the longitude reckoned from any xed direction, and 22 is the longitude
of the Moons longest meridian from the same xed direction. However, 22 will contain the
time explicitly see 4, 13, for a detailed discussion. Using spherical trigonometry, we have
sin  sin i sinf g, where f is the true anomaly.
The following assumptions have been made 4, 14 and will be used in this work: a
the motion of the Moon is uniform librations are neglected; b the lunar equator lies in the
ecliptic we neglect the inclination of about 1.5 of the lunar equator to the ecliptic, and the
inclination of the lunar orbit to the ecliptic of about 5 ; c the longitude of the lunar longest
meridian is equal to the mean longitude of the Earth librations are neglected; d the mean
longitude of the Earth, , is equal to 22 . See De Saedeleer 4 for a detailed discussion.
4 Mathematical Problems in Engineering

Since the variables and appear only as a combination of , where 

nM t const, with nM being the lunar mean motion, the degree of freedom can be reduced by
choosing as a new variable h  . A new term must then be added to the Hamiltonian in
order to get h  H/H  nM . The Hamiltonian is still time-dependent through . Since
the longest meridian is always pointing toward the Earth, it is possible to choose a rotating
system whose x-axis passes through this meridian.
Regarding the Earths potential, the dominant coecient is J2 . The rest are of higher-
order terms 15. In contrast to the Earth, the rst harmonics of the Lunar potential are all
almost of the same order see Table 1. This fact complicates the choice of the harmonic where
the potential can be truncated and this makes its choice a little arbitrary. The inuence of the
Earth and of the nonsphericity of the Moon on the stability of lunar satellites was also pointed
out by 16 but sectorial harmonics were not considered. In terms of the orbital elements, the
Legendre associated functions for zonal up to J5 and sectorial terms C22 and C31 , where the
values for the spherical harmonic coecients are given in the Appendix C, can be written in
the following form 13, 14:

 1  
P2 sin  3sin2 isin2 f g 1 ,
2
 5  3 
P3 sin  sin3 isin3 f g sini sin f g ,
2 2
 35 4  15  3
P4 sin  sin isin4 f g sin2 isin2 f g ,
8 4 8
 63 5  35  15 
P5 sin  sin isin5 f g sin2 isin3 f g sini sin f g ,
8 4 8
 
P22 sin  3 3sin2 isin2 f g ,

      
P22 sin cos 2  6 2 cos2 f 2 sin2 f 2 sin 2f 3 1 sin2 isin2 f g ,
   
 15 2 3  3 15 2 
P31 sin cos  s cosh cos f g s c sinh sin f g
8 2 4 8
15 2  15 
s cosh cos 3f 3g s2 c sinh sin 3f 3g ,
8 8
2.6

where

 
 cos g cosh cosi sin g sinh,
 
 sin g cosh cosi cos g sinh, 2.7

s  sini, c  cosi.
Mathematical Problems in Engineering 5

In this paper, are taken into only the terms due to J2 and C22 accout. The zonal
perturbation due to the oblateness J2 is dened by 3 H20  /r 3 P20 sin , where
 J2 R20 . However, the disturbing potential is

 
H20  3
1 3cos2 i 3sin2 i cos 2f 2g . 2.8
4r

Substituting the relation  n2 a3 , using the Cayleys tables 17 to express the true
anomaly in terms of the mean anomaly, and with some algebraic manipulations, we get

3  2   
H20  5e 2 cos2 i 2 5e2 cos 2g 3l 7e 7ecos2 i cos 2g 3l
8
   
17e2 cos2 i 17e2 cos 2g 4l e ecos2 i cos 2g l 2.9
  
1 2 2 1 2
9 cos i
2
e cosl e e cos2l
2
n2 .
3 9 3 3

For the sectorial perturbation, we get 13, 14

 2 2   
H22  6 cos f 62
sen 2
f 12sen 2f 3 3s 2
sen 2
f g , 2.10
r3

where  C22 R20 R0 is the equatorial radius of the Moon; R0  1738 km.
With some manipulations, we get

45 2
H22  n
16

   
2 2  1 2 2 
e 2
cos i 1 cos 2l 2g 2h cos i 1 e 2
cos 2l 2g 2h
5 3 5
   
2 2  1 2 2 
cos i 1 e 2
cos 2l 2g 2h e cos i 12 cos 2l 2g 2h
5 3 5
17 2  7 
e cos i 12 cos 4l 2g 2h ecos i 12 cos 3l 2g 2h
5 5
17 2  7 
e cos i 12 cos 4l 2g 2h ecos i 12 cos 3l 2g 2h
15 15
7  17 
ecos i 12 cos 3l 2g 2h e2 cos i 12 cos 4l 2g 2h
5 5
7  17 
ecos i 12 cos 3l 2g 2h e2 cos i 12 cos 4l 2g 2h
15 15
1  1 
ecos i 12 cos l 2g 2h ecos i 12 cos l 2g 2h
15 5
6 Mathematical Problems in Engineering

1  1 
ecos i 12 cos l 2g 2h ecos i 12 cos l 2g 2h
5 15
 
2 2  9 2
sin2 i e2 cos 2g 2l e cos2l 2h
3 5 5
7  1  9 6
e cos 2g 3l e cos 2g l e2 cos2l 2h cos2he2
5 5 5 5

4 17 2  6 6
cos2h e cos 2g 4l e cosl 2h e cosl 2h ,
5 5 5 5
2.11

where the disturbing potential is written in the form R  H20 H22 .

3. The Hamiltonian System

We nd in the literature several papers that use the method of the average to calculate
perturbations of long-period on articial satellites of the Moon, such as 1823. However,
our objective here is to compute analytically secular and periodic perturbations up to the
second order and, using this, to analyze the coupling terms relating the harmonic coecients.
The Lie-Hori 1 perturbation method is applied to eliminate short-period terms of the
Hamiltonian.
In 24, a dierent approach is proposed for the canonical version of Hori method. The
reference 24 showed that the ordinary dierential equation with an auxiliary parameter t
as independent variable, introduced through Hori auxiliary system, can be replaced by a
partial dierential equation in time t.
In what follows, rst the Lie-Hori 1 method will be shortly presented and then
applied to the problem of the orbital motion of the satellite around the Moon.
Consider the mth order equation of the algorithm of the perturbation method proposed
by Hori 1:

{H0 , Sm } m  Hm , 3.1

where braces stand for the Poisson brackets, H0 is the undisturbed Hamiltonian, and m
is a function obtained from the preceding orders, involving H0 , Hm , Sk , Hk , and Hk , k 
1, . . . , m 1. All of these functions are written in terms of the new set of canonical variables
,  and dened through the following equations:

   k 
H x, y  H0 x, y Hk x, y ,
k1
   

H ,  H0 , k Hk , , 3.2
k1
  k 
S ,  Sk , ,
k1
Mathematical Problems in Engineering 7

where x, y is the original set of canonical variables, Hx, y is the original Hamiltonian,
H ,  the new Hamiltonian and S,  is the generating function of the canonical
transformation, x, y , . The transformation is such that the new canonical system
has some advantages for the solution.

In order to determine the functions Sm and Hm , Hori introduces an auxiliary

parameter t through the following system of canonical equations 1:

di H0 di H0
 ,  , i  1, . . . , n. 3.3
dt i dt i

Accordingly, 3.1 reduces to

dSm
 m Hm , 3.4
dt

with m written in terms of the general solution of the system 3.3, involving 2n constants

of integration. Equation 3.4 has two unknown functions: Sm and Hm .
The Poisson brackets are dened as

, - x y x y x y x y x y x y
x, y  3.5
L l l L G g g G H h h H

with respect to the classical Delaunay variables set l, g, h, L, G, H. Since only l, g, h, L are
explicitly present in the Hamiltonian, the partial derivatives with respect to L, G, H are
computed as /L  /L 2 /eL/e, /G  /eL/e c/Ls/i,
/H  1/Ls/i where the bracket indicates the derivative with respect to L
occurring explicitly, and c, s, are c  cos i, s  sin i,  1 e2 .
To solve 3.4, we separated in secular and periodic part as it is done by the principle of
the mean 7. Using 3.3 and 3.4, Hori 1 writes the equations that dene the integration
theory based on average principle to determine the new Hamiltonian and the generating
function as follows.
Order zero:

H0  H0 . 3.6

Order one:

H1  H1s ,
 3.7
S1  H1p d.
8 Mathematical Problems in Engineering

Order two:

1, -
H2  H2s H1 H1 , S1 s ,
2
 
1, - 3.8
S2  F2p H1 H1 , S1 p d,
2

where, at each function, the index s represents the secular part of the function and the index
p the periodic part of the function. See Hori 1 for a detailed discussion.
The Hamiltonian of the dynamical system associated to the problem of the orbital
motion of the satellite around the Moon can be written in the following form:

H  H0 H20 H22 , 3.9

where

2
H0  nM H, 3.10
2L2

The term nM H is added to reduce the degree of freedom, since the mean longitude of
the Earth is time-dependent 14. Here, the term nM H is taken as order zero as suggested by
Breiter 25.
Now, doing

H20  H1 ,
3.11
H22  H2 ,

we get,

H  H0 H1 H2 . 3.12

With the purpose of applications of the perturbation method, the terms of the
Hamiltonian are written in the following form:

0 2
H0  nM H,
2L2 3.13
1
H1  H1 H2 .

The disturbing terms are represented in the rst order of the applied method. To
eliminate the short-period terms of equation 3.13, the method of Lie-Hori 1 perturbation
theory is applied. In this work long-period terms are calculated, substituting the result in
the planetary equations of Lagrange 26. The equations of motion are integrated and nally
Mathematical Problems in Engineering 9

the results analyzed. With a simplied model for the disturbing potential it is possible to do
analyses for the orbital motion of the satellite.
Applying the method of Hori 1 to our problem to eliminate the terms of short period,
we get the following:
Order zero:

2
H0  H0  nM H. 3.14
2L2

Order one:

  2
1 1 1
H1  H1  H1 dl,
s 2 0
3.15
 
1  1
S1  H1p d  H1 H1 dl,

0
where  H0 /L.
Order two:
 2 
1  1  1 1 1

H2  H1 H1 , S1  H1 H1 , S1 dl. 3.16
2 s 2 0 2

4. Critical Inclination
We consider now the problem of a lunar articial satellite with low altitude taking into
account the oblateness J2  and the equatorial ellipticity sectorial term C22  of the Moon.
The rst order long-period disturbing potential order of the method of perturbation theory
obtained by the Hori method algorithm can be written as

1 2
k1  n 6cos2 i 3e2 2 18 cos2he2 18 cos2he2 cos2 i 12 cos2h
8
4.1
12 cos2hcos2 i 9cos2 ie2 ,

where  J2 R20 e  C22 R20 and k1  H1 .

We observe that, at the order considered, the disturbing potential has the terms due to
the oblateness , that are secular, and terms due to the equatorial ellipticity of the Moon ,
that appear multiplied for cos2h.
Taking into account 4.1 a formula for the critical inclination is found. In fact,
substituting 4.1 in the planetary equations of Lagrange 26 and solving the equation
dg/dt  0, we get

6 cos2h
cos2 i  . 4.2
5 2 cos2h
10 Mathematical Problems in Engineering

72

70

68

i 66

64

62

60

0 1 2 3 4 5 6
h

Figure 1: Variation of the critical inclination with respect to the longitude of the ascending node where
i-degree and h-rad.

Table 2: Critical inclination for the potential of rst order, where:  613.573;  67.496.

Longitude of the ascending Critical inclination for direct Critical inclination for
node h orbits ic  retrograde orbits ic 
1 rad 61.10 118.90
2 rad 59.98 120.02
/2 58.56 121.45
/3 60.69 119.31
72.83 107.17

this formula was already obtained by De Saedeleer and Henrard 5 and was here derived in
an independent way, observing that in 5  C22 R20 . Thus, when we consider the terms due
to the oblateness J2  and the equatorial ellipticity of the Moon C22 , the critical inclination
depends on the longitude of the ascending node. Figure 1 represents the variation between
the inclination and the longitude of the ascending node. Table 2 represents the values of the
critical inclination for some values of the ascending node.
Now, let us consider the second-order disturbing potential k2  H2 where is given in
the Appendix A order of the method of perturbation theory; the potential k2 presents:

a coecients of second order J22 , C22

2
,
b coupling terms between J2 and C22 .

Plugging the equations for the potential in the planetary equations of Lagrange and
solving the equation dg/dt  0, we present a new formula to compute the critical inclination
taking into account the J2 and C22 terms of the second-order disturbing potential. It is a
function of two variables: the argument of the periapsis g and the longitude of the ascending
node h. When the sectorial term C22 is considered, the rst order disturbing potential is a
function of the longitude of the ascending node and of both longitude of the ascending node
and argument of the periapsis to the second order potential. The new formula is given in the
Appendix B.
Mathematical Problems in Engineering 11

We observe that, at the second order, the disturbing potential presents terms due to the
oblateness  and to the equatorial ellipticity of the Moon , that also appear multiplied
by periodic functions. Here, terms of couplings between the oblateness and the equatorial
ellipticity of the Moon J2 , C22  and terms of second order appear. Several scenarios can be
considered. For instance frizzing orbits with particular values of h and g, let us say h  /2
and g  3/2, we get a value of 53.46 for the critical inclination taking into account equation
given in the Appendix B. Therefore, the critical inclination taking into account equation 4.2
is 58.56 see Table 2.

5. Sun-Synchronous Lunar Orbit

Now, an approach is presented for a Lunar Sun-synchronous orbit. The Moon rotates with
angular rate about 360 for 27, 32 days, while the Earth rotates with angular rate about 360
by day. The perturbation caused by the orbital precession has been studied historically for
orbits centered in the Earth because of near polar orbits the precession is about one degree per
day and to provide attractive Sun-synchronous orbits for many missions around the Earth.
Considering Sun-synchronous orbits for lunar satellites we show that it is not possible to
produce near polar orbits. The precession of the ascending node due to the nonsphericity of
the Moon, when only the eect of the J2 is considered, in 3.3 is well known in Brouwer
theory 27 that the precession of the longitude of the ascending node is given by

d 3 J2 R20 n cos i
 . 5.1
dt 2 a2 1 e2 2

The Moons orbital period is about 27, 32 days and the Earths orbital period is about
365, 26 days. Then, for a Sun-synchronous orbit, we have, in lunar day 2:
 
d 27, 35
 360 /lunar day  26, 92657 /lunar day. 5.2
dt 365, 26

An inclination for a Sun-synchronous orbit was presented by Park and Junkins 2
using 5.1 and 5.2, with the following initial conditions: a  1837.63 km; e  0. The
calculated inclination is is  144.82 . This inclination is not feasible for producing near-polar
orbits, because this orbit does not pass suciently near the poles. Considering the disturbing
potential given by 4.1 and substituting in the Lagrange planetary equations 26 to calculate
the variation of the longitude of the ascending node, we get

d 3 n 3n  
 2
cosi  cosi 2 3e2 cos2h , 5.3
dt 2 a2 1 e2  2a2 1 e2

where  J2 R20 ,  C22 R20 . This new equation 5.3 gives the precession of the longitude of
the ascending node due to nonsphericity of the Moon when the eect of the J2 and the C22 are
considered. In rst approximation the periodic terms due to the J2 are negligible, however,
for the C22 term in the rst approximation appears the periodic term cos2h. When  0, we
obtain the classic solution given by 5.1.
12 Mathematical Problems in Engineering

170

160

IS 150

140

0 0.5 1 1.5 2 2.5 3

h

Figure 2: Variation of the inclination Sun-synchronous IS in degree with longitude of the ascending node
h-rad, a0  1838 km, e0  0.038.

In what follows, the variation of the longitude of the ascending node will be analyzed
to obtain polar or near-polar orbits for some special cases. Using 5.2 and 5.3 we get a new
formula to compute the inclination of Sun-synchronous orbits for Moons satellites of low
altitude:

 3/2
1.327307409107 a2 1 e2
is  arccos   , 5.4
n 2 e2 3e4 cos2h

where n is given in rad/s,  67.496 km2 ;  613.573 km2 . Equation 5.4 gives the
inclination depending on the semi-major axis, the eccentricity and on the longitude of
the ascending node. For Sun-synchronous orbits, considering a  1837.63 km; e  0;
h  /2 the calculated inclination is is  132.35 . This inclination is not also ideal for
near-polar orbits. Thus, the obtained results are still distant from a polar Sun-synchronous
orbit, but it is important to consider the term due to the Moons equatorial ellipticity to
get more realistic results. Figure 2 represents the variation of Sun-synchronous inclinations
with respect to the longitude of the ascending node where a0  1838 km and e0 
0.038. It can be observed in Figure 2 that, xing h  /2, we get an inclination of
about 132 .

6. Applications for Low Altitude Satellites

The disturbing potential rst order k1 and second order k2 is substituted in the
Lagranges planetary equations 26 and numerically integrated. Considering the disturbing
potential due to the nonsphericity of the Moon J2 and C22 , numerical applications long-
period potential are performed to analyze the variation of the eccentricity for dierent initial
conditions of the inclination.
Mathematical Problems in Engineering 13

0.038

0.036

0.034

e 0.032

0.03

0.028

0.026
0 500 1000 1500 2000
t

Order one
Order two

Figure 3: Initial conditions: a0  1838 km, e0  0.038, i0  30 , g0  3/2, h0  /2 and t-days.

0.038

0.03798

0.03796

e 0.03794

0.03792

0.0379

0.03788

0 500 1000 1500 2000

t

Order one
Order two

Figure 4: Initial conditions: a0  1838 km, e0  0.038, i0  90 , g0  3/2, h0  /2 and t-days.

Figures 3 and 4 represent the comparison for dierent orders of the disturbing
potential. The variation of the eccentricity for lunar satellites in low altitude is constant
at rst order. This happens because the coecients J2 and C22 do not aect the variation
rate of the eccentricity as we can verify in Lagranges planetary equations. Therefore,
it is important to insert more terms in the potential to get more realistic results as, for
example, to study the lifetimes of low altitude Moon articial satellites 28, considering
the zonal terms J2 , J3 and J5 and the sectorial terms C22 and C31 in the disturbing potential.
When taking into account the disturbing potential of second order considering the eect
14 Mathematical Problems in Engineering

38

36

180 i/
34

32

30
0 200 400 600 800 1000
t

Order one

Figure 5: Initial conditions: a0  1838 km, e0  0.038, i0  30 , g0  3/2, h0  /2 and t-days. Fixing a
value h  /2 we nd a value of 58.56 for the critical inclination.

of the nonuniform distribution of mass of the Moon J2 and C22 , the coecients J2
and C22 aect the variation rate of the eccentricity as we can verify in the Lagranges
planetary equations. For the second order the results shows a small variation of the
eccentricity for larger inclinations and an accentuated increase for small inclinations. This
is due to the C22 term that aects the eccentricity of the satellite directly in second
order.
The expression of the eccentricity is presented in the following form de/dt 
sin2g 2 sin2g 2 sin2g 2h sin2g 2h where
the terms due to the oblateness  and the equatorial ellipticity  appear multiplied by
periodic functions, terms of couplings between J2 and C22 and terms of second order of
the type J22 and C22 2
. Figures 3 and 4 shows the temporal variation of the eccentricity for
cases where initial conditions are obtained from the frozen orbits condition. For instance
frizzing orbits with particular values of h and g, let us say h  /2 and g  3/2.
Another factor that also contributes for the variation of the eccentricity, using the potential
up to the second order, is the presence of the coecients of second order terms J22 , C22 2

and the coupling terms between J2 and C22 . The choice of the initial inclination is very
important to assure a frozen orbit when it is taken into account the second order disturbing
potential.
Figures 5 and 6 show the inclination suering a periodic variation that depends of
the longitude of the ascending node, as we can verify by 4.2. Figure 5 shows a variation
of the inclination for a value below of the critical inclination around 8 degrees in 100 days,
while Figure 6 presents a more accentuated variation for values of the inclination above of
the critical inclination around 50 degrees in 200 days.
Considering the second-order disturbing potential, Figures 7 and 8 represent the
variation rate for the inclination. The variation represented is due to the initial values given
for the argument of the periapsis and for the longitude of the ascending node. The given
initial values for g and h are for the condition of frozen orbits. For instance frizzing orbits
with particular values of h and g, let us say h  /2 and g  3/2. The same comments done
for the eccentricity, including those for the coupling terms zonal and sectorial, are valid for
the variation of the orbital inclination.
Mathematical Problems in Engineering 15

110

100

180 i/
90

80

70

0 200 400 600 800 1000

t

Order one

Figure 6: Initial conditions: a0  1838 km, e0  0.038, i0  65 , g0  3/2, h0  /2 and t-days. Fixing a
value h  /2 we nd a value of 58.56 for the critical inclination.

30

29.999

29.998

29.997
i
29.996

29.995

29.994

0 200 400 600 800 1000

t

Order two

Figure 7: Initial conditions: a0  1838 km, e0  0.038, i0  30 , g0  3/2, h0  /2 and t-days.

Numerical applications with the rst and second order disturbing potential are done
taking into account the nonsphericity of the Moon and perturbations from the third-body
in elliptical orbit Earth is considered considering the term P2 of the Legendre polynomial
and the eccentricity of the disturbing body up to the second order. Figures 9, 10 and
16 Mathematical Problems in Engineering

65.002502

65.001999

65.001503
i

65.000999

65.000504

65
0 200 400 600 800 1000
t

Order two

Figure 8: Initial conditions: a0  1838 km, e0  0.038, i0  65 , g0  3/2, h0  /2 and t-days.

0.048
0.047
0.046
0.045
0.044
0.043
e
0.042
0.041
0.04
0.039
0.038
0.037
0 100 200 300 400 500 600 700 800 900
t day

k1 k2 i0  20 i0  72
i0  30 i0  90

Figure 9: Initial conditions: a0  1838 km, e0  0.038, g  3/2, h  /2, emax  0.05. Considered
Perturbations: P2 , J2 , C22  0.

11 represent the sum of the potential of the rst order with the second order k1 k2.
/ 0 and C22  0 the small inclinations cause small oscillations in
Figure 9 shows that, if J2 
the variation of the eccentricity and Figure 10 shows the eect caused by the C22 term,
when J2  0, where the small inclinations cause a large increase in the variation of the
eccentricity, as well as we can visualize in Figure 11, where it is taken into account the
P2 , J2 , and C22 terms. Considering the disturbing potential up to second order, the terms
J2 and C22 that does not cause perturbation in the eccentricity in rst order, appear as
Mathematical Problems in Engineering 17

0.075
0.07
0.065
0.06
0.055
e
0.05
0.045
0.04
0.035
0.03
0 100 200 300 400 500 600 700 800 900
t day

k1 k2 i0  20 i0  72
i0  30 i0  90

Figure 10: Initial conditions: a0  1838 km, e0  0.038, g  3/2, h  /2, emax  0.05. Considered
Perturbations: P2 , J2 , C22 .

0.045

0.04

0.035

e 0.03

0.025

0.02

0.015
0 100 200 300 400 500 600 700 800 900
t day

k1 k2 i0  20 i0  72
i0  30 i0  90

Figure 11: Initial conditions: a0  1838 km, e0  0.038, g  3/2, h  /2, emax  0.05. Considered
Perturbations: P2 , J2 , C22 .

disturbing term of the type J22 and C22 2

and terms of coupling of the type J2 C22 that aects
the eccentricity of the satellite directly. The terms appear due to the perturbation method
used.
The temporal variation of the eccentricity is strongly aected by the initial inclination
i0 . As it can be observed by Figure 11, for i0 < 48.6 the variation of the eccentricity presents
great amplitude but, for i0 > 48.6 , the variation has small amplitude.
Figures 12 and 13 represent the variation of the eccentricity for a lunar satellite of
low altitude considering dierent terms in the disturbing potential. Figure 12 considered
18 Mathematical Problems in Engineering

0.09

0.08

e 0.07

0.06

0.05
0 100 200 300 400 500 600
t

Order one

Figure 12: Initial conditions: a0  1935.79 km, e0  0.05, i  90 , g  3/2, h  /2, emax  0.09 and t-days.
Perturbations P2 .

0.054

0.053
e
0.052

0.051

0.05
0 1000 2000 3000 4000
t

Order one

Figure 13: Initial conditions: a0  1935.79 km, e0  0.05, i  90 , g  3/2, h  /2, emax  0.09 and t-days.
Perturbations P2 J2 C22 .

third body perturbation and Figure 13 considered nonsphericity of the Moon and third body
P2  perturbation. A comparison is done between the perturbations for the case where the
inclination is 90 polar orbit. Therefore we can conclude that, besides the term due to
the J2 , the sectorial term C22 should also be considered in the disturbing potential to get
more realistic results. For lunar satellites of low altitude it is impracticable to consider real
applications taking into account only the perturbation of the third body P2 . In fact, taking
into account only the perturbation of the third body P2 , the eccentricity of the satellite
increases causing escape from the Moon, or crashing to the Moon in 600 days see Figure 12.
Mathematical Problems in Engineering 19

We observe that, for a Moons articial satellite orbiting in low altitude, the combination of
the two perturbations help to control the variation of the eccentricity, mainly for inclinations
of 90 see Figure 13.
These results agree with those obtained by dAvanzo et al. 29 when it is considered
just the eects of the zonal harmonics J2 J5 and a rst order potential.

7. Conclusions
Using Lie-Hori method, the disturbing potential due to the nonsphericity of the gravitational
eld of the Moon is obtained up to rst and second order. The disturbing potential is
substituted in Lagranges planetary equations and they are numerically integrated. Analyses
for the variations of the orbital elements are done. Terms of couplings between the oblateness
and the equatorial ellipticity of the Moon J2 , C22  and terms of second order of type J22
2
and C22 are obtained. A formula is developed to compute the critical inclination when
the eect of the C22 equatorial ellipticity term is considered in the Hamiltonian in rst
and second order. The critical inclination can be strongly aected by the coecient due
to the equatorial ellipticity of the Moon and by the longitude of the ascending node. The
formula for the critical inclination for the second order is a function of two variables:
the argument of the periapsis and the longitude of the ascending node. At the rst order
this formula is a function of the longitude of the ascending node only. For Lunar low
altitude satellites LLAS, it is important to take into account both, the terms due to
the oblateness and terms with the equatorial ellipticity of the Moon to get more realistic
results.
The variation of the longitude of the ascending node is analyzed looking Lunar Sun-
synchronous orbits and near-polar orbits. A new formula is obtained to compute inclinations
of Lunar Sun-synchronous orbits when the terms due to the oblateness of the Moon J2 and
C22  are taken into account. The presented formula to the inclination depends on the semi-
major axis, eccentricity and on the longitude of the ascending node of the satellite. The term
due to the eect of the C22 must be considered for the case of a lunar satellite to analyze the
precession of the longitude of the ascending node.
For a LLAS, when it is considered only the nonsphericity of the Moon in the disturbing
potential, and at the rst order, the orbital eccentricity of the satellite is constant along the
time. This happens since the coecients J2 and C22 , at this order, do not aect the variation
rate of the eccentricity directly, therefore it is important to insert more terms in the potential
as, for example, the zonal terms J3 , J5 and the sectorial term C31 , to get more realistic results.
At the second order, small variations are present. In this case, the small variations of the
eccentricity are due to a combination of the following factors: a initial conditions given to
get frozen orbits b to couplings terms between the oblateness and the equatorial ellipticity
of the Moon J2 , C22  and c terms of second order of type J22 and C222
. For small inclinations,
second-order terms including coupled terms are greater than 1st order terms. This happens
because the coecients J2 and C22 , at second order, aect the variation of the eccentricity
directly.
To study lifetime of LLAS, due to the characteristics of the mass distribution of the
Moon, it is necessary to take into account up the second order of the disturbing potential
and develop up to the second order the Lie-Hori algorithm. In fact, at rst order, the
coecients do not aect the eccentricity directly while at the second order the coecients
J2 and C22 aects the eccentricity directly and thus contributing eciently to more complete
studies.
20 Mathematical Problems in Engineering

Appendices
A. Appendix A
Let us consider the second-order disturbing potential k2  H2 order of the method of
perturbation theory, where k2  H2 , c  cos i, s  sin i:
     
1 15 4 11 2 2 14 2 32 2 8
k2  27 e 2
e c 2
e e c
16a2 2 15 30 15 15 15 15

2 17 2 13 2 4 
e e c 12 cos 2g 2h
45 15 30 15
   
15 4 11 2 14 2 32 8
e2 e2 c2 e e2 c
2 15 30 15 15 15 15

2 17 13 4 
e2 e2 c 12 cos 2g 2h
45 15 30 15
   
643 2 36 1 
s e2 c 12 cos 4g 2h
6 643 4
   
643 36 1 
c 12 e2 s2 cos 4g 2h
6 643 4
   
643 36  643 36
e2 2 c 14 cos 4g 4h c 14 e2
16 643 16 643
 1  
2 cos 4g 4h 2 s2 c 12 2 e2 cos 2g 4h
2
1  
2 s2 c 12 2 e2 cos 4h 2g
2
   
15 2 13
2 e2 2 e 2 e2 2 c2
2 4
   
17 2 11
5e2 2 2 e2 e2 2
6 3 12
   
 3215 4 36 2 
s2 cos 2g s e2 cos 4g
24 643 5
 
307 2 56 148 199 2
s e2 e c2
2 921 2763 307

407 2 172 20 2863 2
e e cos2h
921 2763 921 2763
 
705 2 484
e 2 s4 cos4h
8 6345
    
3401 2 175 2 139 2 10 229 2 2 37 2 4
e e e c
24 18 6 9 12 18
    
4019 2 55 2 26 442 2 266 2 2 22 2 2
e e e c
12 3 9 9 9 9
     
11003 2 223 2 467 2 16 19 497 2 2 2
e e 2 e n .
72 18 18 9 18 36
A.1
Mathematical Problems in Engineering 21

B. Appendix B
Plugging the equations for the potential in the planetary equations of Lagrange and solving
the equation dg/dt  0, we present a new formula to compute the critical inclination taking
into account the J2 and C22 terms of the second-order disturbing potential. We get

cosi
    
 sqrt 28408563 cos 2h 2g 56754363 cos 6g 2h 78645782 2 cos 8g 4h
 
33903363 cos 6g 12947043 cos 6g 4h 2251724322 2 cos2h
  
28408563 cos 2h 2g 1159172763 cos 4g 2h 96644162 2 cos 2h 2g
  
864112143 cos 8g 2h 179877603 cos 4g 2h 11574003 cos 6g 2h
 
1768279523 cos4h 50121842 2 cos 4h 2g 1159172763 cos 4g 2h
  
229746243 cos 6g 2666430722 2 cos 4g 762801603 cos 4g 4h
  
864112143 cos 8g 2h 1420405083 cos 4g 6h 1885165923 cos 8g
  
681470403 cos 4g 313883283 cos 8g 4h 78645782 2 cos 8g 4h
 
235412463 cos 8g 6h 314349842 2 cos 8g 2h 163782203 cos2h
  
53873282 2 cos 6g 2h 180979563 cos 2g 6h 50121842 2 cos 4h 2g
  
48356763 cos 2h 2g 1577082242 2 cos 4g 2h 235412463 cos 8g 6h
 
24829202 2 cos 6g 4h 1369123323 cos6h 180635043 cos 4h 2g
 
48356763 cos 2h 2g 338078403 cos2h 62381883 cos 6g 6h
 
76171203 cos 2g 3584131304 92003044 24829202 2 cos 6g 4h
  
179877603 cos 4g 2h 42351363 cos 2g 1577082242 2 cos 4g 2h
  
180635043 cos 4h 2g 1420405083 cos 4g 6h 12984482 2 cos 2g
  
56754363 cos 6g 2h 11574003 cos 6g 2h 36577442 2 cos 6g
  
471408122 2 cos 8g 941649844 cos 8g 6h 941649844 cos 8g 6h
  
277475764 cos 6g 2h 5161678804 cos 4g 2h 165829684 cos 6g 6h
  
165829684 cos 6g 6h 2063206324 cos 4g 6h 5161678804 cos 4g 2h
  
460135444 cos 2g 6h 460135444 cos 2g 6h 2063206324 cos 4g 6h
 
5727135844 cos 4g 4h 813385802 2 319063683 13199764 cos 6g 8h
  
353118694 cos 8g 8h 13199764 cos 6g 8h 353118694 cos 8g 8h
22 Mathematical Problems in Engineering
  
1099291324 cos 8g 4h 1099291324 cos 8g 4h 280975684 cos 4h 2g
 
280975684 cos 4h 2g 3550428963 23033524 cos 2g 8h
  
23033524 cos 2g 8h 5727135844 cos 4g 4h 277233124 cos 2g

1804285264 cos8h 7540785364 cos4h 6240879684 cos6h

  
3789418324 cos 4g 2124777424 cos 8g 419640484 cos 6g
 
3816903364 cos2h 1536069244 cos 4g 8h 1536069244 cos 4g 8h
  
44089924 cos 6g 4h 277475764 cos 6g 2h 44089924 cos 6g 4h
  
396004 cos 4g 18010564 cos 2g 190262164 cos 2h 2g
  
190262164 cos 2h 2g 1573148884 cos 8g 2h 1573148884 cos 8g 2h
 
180979563 cos 2g 6h 314349842 2 cos 8g 2h
 
710935642 2 cos4h 569589122 2 cos 4g 4h 569589122 2 cos 4g 4h
  
53873282 2 cos 6g 2h 62381883 cos 6g 6h 313883283 cos 8g 4h
  
12947043 cos 6g 4h 1855092483 cos 4g 96644162 2 cos 2h 2g
  
762801603 cos 4g 4h 60 722 cos 2h 2g
  
60 722 cos 2h 2g 118982 cos 4g 4h
  
118982 cos 4g 4h 158642 3966 cos 4g 2h
   
158642 3966 cos 4g 2h / 1056 17282 cos 2h 2g
   
1056 17282 cos 2h 2g 366662 cos 4g 4h 366662 cos 4g 4h
  
1104 6002 5762 cos 2g 5424 9922 cos2h
  
2 
40740 cos 4g 27316 cos4h 7488 50868 4608 .
2 2 2
5
B.1

C. Appendix C
Model given by Akim and Golikov 30 for the spherical harmonic coecients is

J2  2.09006496 104 , J4  2.32815 106 , C22  2.447305 105 ,

C.1
J3  5.48445 106 , J5  3.169113 106 , C31  2.871327 105 .
Mathematical Problems in Engineering 23

Acknowledgment
This work was accomplished with support of the FAPESP under the contract no. 2007/04413-
7 and 2006/04997-6, SP-Brazil, and CNPQ 300952/2008-2.

References
1 G. I. Hori, Theory of general perturbations with unspecied canonical variables, Publications of the
Astronomical Society of Japan, vol. 18, no. 4, pp. 287296, 1966.
2 S.-Y. Park and J. L. Junkins, Orbital mission analysis for a lunar mapping satellite, Journal of the
Astronautical Sciences, vol. 43, no. 2, pp. 207217, 1995.
3 B. De Saedeleer and J. Henrard, Orbit of a lunar articial satellite: analytical theory of perturbations,
in Transits of Venus: New Views of the Solar System and Galaxy, IAU Colloquium, no. 196, pp. 254262,
Cambridge University Press, Cambridge, Mass, USA, 2005.
4 B. De Saedeleer, Analytical theory of a lunar articial satellite with third body perturbations,
Celestial Mechanics & Dynamical Astronomy, vol. 95, no. 14, pp. 407423, 2006.
5 B. De Saedeleer and J. Henrard, The combined eect of J2 and C22 on the critical inclination of a
lunar orbiter, Advances in Space Research, vol. 37, no. 1, pp. 8087, 2006.
6 A. Deprit, Canonical transformations depending on a small parameter, Celestial Mechanics, vol. 1,
pp. 1230, 1969.
7 S. Ferraz Mello, Canonical Perturbation Theories, Springer, New York, NY, USA, 2007.
8 J. P. dos Santos Carvalho, R. V. de Moraes, and A. F. B. de Almeida Prado, Semi-analytic theory of a
moon articial satellite considering lunar oblateness and perturbations due to a third-body in elliptic
orbit, in Proceedings of the 7th Brazilian Conference on Dynamics, Control and Applications, pp. 5157,
Presidente Prudente, Brazil, 2008.
9 J. P. dos Santos Carvalho, R. V. de Moraes, and A. F. B. de Almeida Prado, Moon articial satellites:
lunar oblateness and earth perturbations, in Proceedings of the International Conference on Mathematical
Problems in Engineering, Aerospace and Sciences (ICNPAA 08), pp. 10951106, Cambrige Scientic,
Genoa, Italy, 2009.
10 J. P. dos Santos Carvalho, R. V. de Moraes, and A. F. B. de Almeida Prado, A study on resonance
for a lunar articial satellite, in Proceedings of the 8th Brazilian Conference on Dynamics, Control and
Applications, pp. 16, Bauru, Brazil, 2009.
11 J. P. dos Santos Carvalho, R. V. de Moraes, and A. F. B. de Almeida Prado, Non-sphericity of the
moon and critical inclination, in Proceedings of the 32nd Congresso Nacional de Matematica Aplicada e
Computacional, Cuiaba, Brazil, 2009.
12 W. M. Kaula, Theory of Satellite Geodesy: Applications of Satellites to Geodesy, Dover, New York, NY, USA,
2000.
13 M. Radwan, Analytical approach to the motion of a lunar articial satellite, Astrophysics and Space
Science, vol. 283, no. 2, pp. 133150, 2003.
14 G. E. O. Giacaglia, J. P. Murphy, and T. L. Felsentreger, A semi-analytic theory for the motion of a
lunar satellite, Celestial Mechanics, vol. 3, no. 1, pp. 366, 1970.
15 A. Elipe and M. Lara, Frozen orbits about the moon, Journal of Guidance, Control, and Dynamics, vol.
26, no. 2, pp. 238243, 2003.
16 Z. Knezevic and A. Milani, Orbit maintenance of a lunar polar orbiter, Planetary and Space Science,
vol. 46, no. 11-12, pp. 16051611, 1998.
17 A. Cayley, Tables of the developments of functions in the theory of elliptic motion, Memoirs of the
Royal Astronomicai Society, vol. 29, p. 191, 1861.
18 R. A. Broucke, Long-term third-body eects via double averaging, Journal of Guidance, Control, and
Dynamics, vol. 26, no. 1, pp. 2732, 2003.
19 I. V. Costa and A. F. B. de Almeida Prado, Orbital Evolution of a Satellite Perturbed by a Third Body,
in Advances in Space Dynamics, A. F. Bertachini and A. F. B. de Almeida Prado, Eds., p. 176, 2000.
20 R. C. Domingos, R. V. de Moraes, and A. F. B. de Almeida Prado, Third-body perturbation in the
case of elliptic orbits for the disturbing body, Advances in the Astronautical Sciences, vol. 129, part 3,
pp. 26392652, 2008.
21 A. F. B. de Almeida Prado, Third-body perturbation in orbits around natural satellites, Journal of
Guidance, Control, and Dynamics, vol. 26, no. 1, pp. 3340, 2003.
24 Mathematical Problems in Engineering

22 C. R. H. Solorzano, Third-body perturbation using a single averaged model, M.S. dissertation, National
Institute for Space Research INPE, Sao Jose dos Campos, Brazil, 2002.
23 C. R. H. Solorzano and A. F. B. de Almeida Prado, Third-body perturbation using a single averaged
model, in Advances in Space Dynamics 4: Celestial Mechanics and Astronautics, H. K. Kuga, Ed., pp.
4756, 2004.
24 S. Da Silva Fernandes, Notes on Hori method for canonical systems, Celestial Mechanics & Dynamical
Astronomy, vol. 85, no. 1, pp. 6777, 2003.
25 S. Breiter, Second-order solution for the zonal problem of satellite theory, Celestial Mechanics and
Dynamical Astronomy, vol. 67, pp. 237249, 1997.
26 J. Kovalevsky, Introduction to Celestial Mechanics, Bureau des Longitudes, Paris, France, 1967.
27 D. Brouwer, Solution of the problem of articial satellite theory without drag, The Astronomical
Journal, vol. 64, pp. 378397, 1959.
28 W. K. Meyer, J. J. Buglia, and P. N. Desai, Lifetimes of Lunar Satellite Orbits, Tech. Rep. N-TP-3394
94, p. 27771, NASA STI/Recon, 1994.
29 P. DAvanzo, P. Teolatto, and C. Ulivieri, Long-term eects on lunar orbiter, Acta Astronautica, vol.
40, no. 1, pp. 1320, 1997.
30 E. Akim and A. Golikov, Combined model of the lunar gravity eld, in Proceedings of the 12th
International Symposium on Space Flight Dynamics (ESOC 97), pp. 357360, Darmstadt, Germany, June
1997.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 347835, 12 pages
doi:10.1155/2009/347835

Research Article
GPS Satellites Orbits: Resonance

Luiz Danilo Damasceno Ferreira1

and Rodolpho Vilhena de Moraes2
1
Department of Geomatic, Federal University of Parana (UFPR), 81531-909 Curitiba, PR, Brazil
2
UNESP-Univ Estadual Paulista, Guaratingueta, CEP 12516-410, SP, Brazil

Correspondence should be addressed to Rodolpho Vilhena de Moraes, [email protected]

Received 29 July 2009; Accepted 28 September 2009

Recommended by Antonio Prado

The eects of perturbations due to resonant geopotential harmonics on the semimajor axis of
GPS satellites are analyzed. For some GPS satellites, secular perturbations of about 4 m/day can
be obtained by numerical integration of the Lagrange planetary equations considering in the
disturbing potential the main secular resonant coecients. Amplitudes for long-period terms
due to resonant coecients are also exhibited for some hypothetical satellites orbiting in the
neighborhood of the GPS satellites orbits. The results are important to perform orbital maneuvers
of GPS satellites such that they stay in their nominal orbits. Also, for the GPS satellites that are not
active, the long-period eects due to the resonance must be taken into account in the surveillance
of the orbital evolutions of such debris.

Copyright q 2009 L. D. D. Ferreira and R. V. de Moraes. This is an open access article distributed
under the Creative Commons Attribution License, which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work is properly cited.

1. Introduction
The period of the orbits of the GPS satellites is about 12 hours, and the main perturbations
acting on their orbits are caused by the nonuniform distribution of the Earths mass, by
the lunar and solar gravitational attractions and by the solar radiation pressure. In this
paper, it is analyzed just some perturbations due to resonant terms of the geopotential
coecients. The resonance considered here is the 2 : 1 commensurability between the
orbital period of the GPS satellites and the period of the Earths rotation. As it is pointed
out by Hugentobler 1 the resonance leads a daily drift rate in semimajor axis of up to
7 m/day.
Lagrange planetary equations, describing the temporal variation of the orbital
elements, are used here to analyze the orbital perturbations of the GPS satellites under the
inuence of resonant coecients.
2 Mathematical Problems in Engineering

2. Disturbing Potential
The geopotential acting on Earths articial satellite can be expressed as 2

n 
GM  
ae
n

R Cnm cos m Snm sin mPnm sin . 2.1
a n2 m0 a

Here GM is the geogravitational constant, r is geocentric distance of the satellite, ae is

the semimajor axis of the adopted Earths ellipsoid, a is the orbital semimajor axis, Cnm and
Snm are the spherical harmonic coecients. Pnm denotes the associated Legendre functions,
, are the satellites geocentric latitude and the longitude, n and m are, respectively, the
degree and order of the harmonic coecients.
The geocentric distance, the latitude, and the longitude of the satellite can be expressed
in terms of the orbital elements and 2.1 becomes 2, 3

n 
 GM  
ae  
n n
R Rnmpq  Fnmpq i Gnpq eSnmpq , , M, T , 2.2
a n2 m0 a p0 q

where Fnmpq i and Gnpq e are, respectively, functions of the satellites orbital inclination i
and eccentricity e, represents the right ascension of the orbital ascending node, is the
argument of perigee, M is the mean anomaly, and T is the Greenwich sidereal time. The
function Snmpq , , M, T  can be expressed as

    nm,even
Cnm Cnm
Snmpq , , M, T   cos A sin A , 2.3
Snmn Snmn nm,odd

and the argument A is given by

 
A  n 2p n 2p q M m T . 2.4

The functions Fnmpq i and Gnpq e are presented as tables 2 and can be adapted to
be used in computers.

3. Resonance in GPS Satellites

Resonance is associated with small divisors. For articial Earth satellites whose orbital
periods are in commensurability with the period of the Earths rotation resonance can occur
when 1, 3, 4
  
A  n 2p n 2p q M m T  0, 3.1

where T is the Earths sidereal rotation.

For the GPS satellites where the commensurability is 2 : 1, there are bounds among the
parameters n, m, p, q.
Mathematical Problems in Engineering 3

Table 1: Resonant parameters.

Degree n Order m p q
3 2 1 0
4 4 1 0
2 2 1 1
4 2 2 1

Taking into account that ,  T , 3.1 can be put as


n 2p q M mT  0, 3.2

since n  M
 2T , we get

n 2p q 2
 m. 3.3

Considering harmonics of order and degree up to order 4, Table 1, presents some

values for the parameters satisfying the resonance condition for the GPS satellites.
Taking into account in the summations the conditions n m even and n m odd
and putting 5

Cnm  Knm cos mnm ,

3.4
Snm  Knm sin mnm ,

where

Knm  2
Cnm S2nm ,
  3.5
1 Snm
nm  tan1 ,
m Cnm

equation 2.2 can be written in general form as 1

 nm,even
GM  ae n cos A
Rnmpq  Fnmp iGnpq e Knm , 3.6
a a sin A nm,odd

where

 
A  n 2p n 2p q M m T nm . 3.7

In order to analyze the orbital perturbations of GPS satellites due to the resonant
coecients presented in Table 1, the Lagrange planetary equations will be used. However,
4 Mathematical Problems in Engineering

Drift rates in the semi major axis m/day

3

4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Satellites

Figure 1: Drift rates in semimajor axis due to the resonant geopotential coecient 32.

Table 2: Maximum values.

Resonants coecients Maximum drift rates m/day
C32 , S32 4.0
C44 , S44 1.5
C 22 , S22 1.5
C42 , S42 0.002

here this analysis will be concentrated initially in the secular eects of such perturbation on
the orbital semimajor axis. Therefore, we have 2

da 2 Rnmpq
 , 3.8
dt na M

with n being the mean motion of the satellite.

Special care must be taken to compute Rnmpq /M. In fact, the parameters n, m, p, q
given by Table 1 must be considered and each set of them gives a unique solution for 3.8.
A numerical integration of 3.8 was performed for a period of one day using IGS/POE
The International GNSS Service Precise Orbital Ephemeredes given on June, 07, 2007.
Transformations were performed to get the corresponding orbital elements.
Figures 1, 2, 3, and 4 show the daily variations of the semimajor axis a for all GPS
satellites.
Figure 1 gives the variation due to the coecients C32 , S32 , which are responsible by
the greatest variation in the semimajor axis that, in this case, corresponds to the satellite
PRN 06. Figures 2, 3, and 4 present, respectively, the daily variations as function of the
coecients C44, S44 , C22 , S22 , C42 , S42 . Figure 5 represents the total daily variation due to the
above considered coecients.
Table 2 shows the maxima and minima values for the daily variation of the semimajor
axis according to the resonant coecients.
Mathematical Problems in Engineering 5

1.5

Drift rates in the semi major axis m/day

1

0.5

0.5

1.5
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Satellites

Figure 2: Drift rates in semimajor axis due to the resonant geopotential coecient 44.

2
Drift rates in the semi major axis m/day

1.5

0.5

0.5

1.5
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Satellites

Figure 3: Drift rates in semimajor axis due to the resonant geopotential coecient 22.

Figures 6 and 7 show the semimajor axis variation for the satellites PRN 02 and PRN
06 for a time interval of 200 days and taking into account the harmonic coecients C32 , S32 .
During this period and for the considered initial condition, it can be observed variations of
about 600 m and 680 m, respectively.

4. Long-Period Perturbations
The eects of the resonance are enhanced when long periods are considered. Table 3 and
Table 4, Figures 8, 9, 10, 11, 12, and 13 show some simulations using hypothetical satellites
and a particular method to study eects of resonance on the orbits of articial satellites 69.
6 Mathematical Problems in Engineering

0.04

Drift rates in the semi major axis m/day

0.03

0.02

0.01

0.01

0.02
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Satellites

Figure 4: Drift rates in semimajor axis due to the resonant geopotential coecient 42.

3
Drift rates in the semi major axis m/day

5
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Satellites

Figure 5: Drift rates in semimajor axis due to all resonant geopotential coecients, Table 1.

Table 3 contains the amplitude and period of the variations of orbital elements for
hypothetical satellites considering low eccentricity, small and high inclination, and the
inuence of the harmonics J20 and J22
Figures 8 and 9 represent, respectively, the temporary variation of the semimajor axis
and of the eccentricity of articial satellites of the GPS type in the neighborhood of the 2 :
1 resonance region when the harmonics J20 and J22 are considered. By Figure 10, assuming
several values for the semimajor axes in the neighborhood of the 2 : 1 resonance, it can be
observed distinct behavior for their temporary variations.
Figure 11 represents the temporary variation for the semimajor axis in the neighbor-
hood of the resonance 2 : 1 considering the inuence of the harmonics J20 and J32 for a satellite
of the GPS type with inclination of about 55 .
Mathematical Problems in Engineering 7

700

600

Drift in the semi major axis m

500

400

300

200

100

0
0 23 46 69 93 116 139 162 185 200
Time interval days

Figure 6: Semimajor axis variation PRN 02 due to the resonant geopotential coecient 32, for a time
interval of 200 days.

100
Drift in the semi major axis m

200

300

400

500

600

700
0 23 46 69 93 116 139 162 185 200
Time interval days

Figure 7: Semimajor axis variation PRN 06 due to the resonant geopotential coecient 32, for a time
interval of 200 days.

Table 4 gives the inuence of the resonance due to the harmonics J20 and J32
considering dierent inclinations, including that of the GPS type satellite. It can be observed
that the amplitudes of the variations are smaller when compared with those when the
harmonics J20 and J22 are taken into account. The more the satellites approach the region
that was dened as a resonant region, the more the variations increase.
Figure 12 represents the temporary variation of the semimajor axis of a satellite of the
GPS type considering resonance due to the harmonics J20 , J22 , and J32 .
8 Mathematical Problems in Engineering

Table 3: Amplitude and period of perturbations due to the 2 : 1 resonance: J2 J22 .

Orbital elements Amplitude Period

ao  26561.770 km e i amax emax imax   t days
ao 0.430 0.01 4 12 km 0.014 8.7 104 2500
ao 0.718 0.01 4 12.5 km 0.0145 8.6 104 7000
ao 0.290 0.01 4 4.5 km 0.0065 3.6 104 2000
4
ao 0.500 0.01 4 3 km 0.0045 2.8 10 1000
ao 2.5 0.01 4 6.5 km 0.0085 4 104 1700
ao 5 0.01 4 1.4 km 0.0023 1.3 106 500
3
ao 0.05 55 1.8 km 0.005 2 10 3200
ao 0.718 0.005 55 2.75 km 0.006 2.8 103 3900
ao 1.770 0.005 55 1.2 km 0.0045 1.4 103 2000
3
ao 5.129 0.005 55 900 m 0.0033 10 1500
ao 3.729 0.005 55 1.8 km 0.0055 1.9 103 2500
ao 0.05 63.4 7.5 km 0.003 10.8 103 1000
ao 2 0.05 63.4 4 km 0.0015 5.7 103 1000
3
ao 0.05 87 7 km 0.0037 13.9 10 2000
ao 2 0.05 87 3.5 km 0.002 10.3 103 1900

Table 4: Amplitude and period of perturbations due to the 2 : 1 resonance: J2 J32 .

Orbital elements Amplitude Period

ao  26561.770 km e i amax emax imax   t days
ao 3.77 0.01 4 120 m 0.0004 1.78 103 500
ao 0.718 0.01 4 245 m 0.0009 3.9 103 1000
ao 0.929 0.01 4 110 m 0.00045 1.78 103 500
3
ao 5 0.01 4 100 m 0.0003 1.15 10 250
ao 2.5 0.005 4 100 m 0.0009 4.6 105 500
ao 0.429 0.005 4 700 m 0.006 2.4 x 104 1500
ao 0.929 0.005 55 1.9 km 0.0085 6.8 104 16000

ao 1.73 0.005 55 250 m 0.0017 1.13 103 16000
ao 5 0.005 55 30 m 0.00055 2.3 104 1500
5
ao 3.77 0.005 55 47 m 0.00035 3.43 10 500
ao 0.429 0.005 87 110 m 0.00085 2.29 104 1800
ao 2.23 0.005 87 45 m 0.00035 8.59 105 800
ao 3.77 0.005 87 20 m 0.00014 4 105 500

Figure 13 represents the time variation of the orbital semimajor axis of a satellite with
eccentricity about 0.01 and inclination about 4 orbiting in a region near the 2 : 1 resonance.
It is remarkable the oscillation of the semimajor axis in the region between by a  26560.0 km
and a  26562.48 km. For instance, a variation of about 10 m in the initial semimajor axis a 
26562.48 km brings up a variation of about more than 10 km in its amplitude in a period of
about 2000 days.
Mathematical Problems in Engineering 9

0.4

0.3

0.2

0.1

km
0

0.1

0.2

0.3

0.4

0.5
0 5000 10000 15000 20000
t day

a  26562.2 km
a  26559.27 km

Figure 8: a versus time, e  0.005, i  55 .

0.008

0.007

0.006

0.005
e
0.004

0.003

0.002

0.001
0 5000 10000 15000 20000
t day

a  26562.2 km
a  26559.27 km

Figure 9: Eccentricity versus time, e  0.005.

5. Conclusions
From the obtained results it can be seen that for the GPS satellites no negligible perturbations
are provoked by resonant tesseral coecients. The daily variation of the semimajor axis due
to the C32 , S32 is less than 4 m/day and those values found for a time interval of 200 days
10 Mathematical Problems in Engineering

1.5

0.5

a km
0.5

1.5

2
0 0.5 1 1.5 2
104
days

Figure 10: a versus time, e  0.005, i  55 , a0  26561.18 km.

0.5
a  26562.7 km

0.5
km

1.5

2
0 4000 8000 12000 16000 20000
t day

Figure 11: a versus time, e  0.005, i  55 , considering the harmonics J20 and J32 .

680 m are compatible with the results presented by specialized literature 7 m/day and
1200 m about this subject 1.
Taking into account the total eects of daily resonant perturbations, it can be observed
that for some GPS satellites these eects are enhanced and for another satellites are attenuated
but all of these eects are smaller than the amplitudes mentioned above. It was shown also
that the eects of the resonance are very important for the analysis of long-period behavior
of the GPS satellites orbits.
Mathematical Problems in Engineering 11

0.02
0
0.02
0.04
0.06
0.08
0.1

km
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0 2000 4000 6000 8000 10000 12000
t day

a  26562.49 km
a  26558 km

Figure 12: Time variation of semimajor axis considering the harmonics J20 , J22 , and J3,2 .

Resonance 2:1 eects on the semimajor axis

26565
26564
26563
26562
26561
26560
26559
a km

26558
26557
26556
26555
26554
26553
26552
26551
26550
26549
0 2000 4000 6000 8000 10000 12000
Time day

a  26562.49 km a  26559.9 km
a  26562.2 km a  26562.48 km

Figure 13: Time variation of the semimajor axis considering the harmonics J20 and J22 for e  0.005, i  4 .

An important aspect to be considered is the necessity to perform orbital maneuvers of

GPS satellites in such way that they stay in their nominal orbits. Also, for the GPS satellites
that are not active, the long-term eects due to the resonance must be taken into account in
the surveillance of the orbital evolutions of such debris.

Acknowledgment
This work was partially supported by CNPq under the contracts No. 305147/2005-6 and
300952/2008-2
12 Mathematical Problems in Engineering

References
1 U. Hugentobler, Astrometry and satellite orbits: theoretical consideration and typical applications,
Geodatisch-geophysikalische Arbeiten in der Schweiz, vol. 57, 1998.
2 W. M. Kaula, Theory of Satellite Geodesy, Blaisdell, Waltham, Mass, USA, 1966.
3 G. Seeber, Satellite Geodesy: Foundations, Methods and Applications, Walter de Gruyter, Berlin, Germany,
2003.
4 D. Ineichen, G. Beutler, and U. Hugentobler, Sensitivity of GPS and GLONASS orbits with respect to
resonant geopotential parameters, Journal of Geodesy, vol. 77, no. 7-8, pp. 478486, 2003.
5 D. Delikaraoglu, On Principles, Methods and Recent Advances in Studies Towards a GPS-Based Control
System for Geodesy and Geodynamics, NASA Technical Memorandum 100716, Greenbelt, Md, USA, 1989.
6 P. H. C. L. Lima Jr., Sistemas ressonantes a altas excentricidades no movimento de satelites articiais, Doctoral
thesis, ITA, Sao Jose dos Campos, 1998.
7 P. H. C. L. Lima Jr., R. Vilhena de Moraes, and S. S. Fernandes, Semi analytical method to study
geopotential perturbations conssidering high eccentric resonant orbits, in Dinamics of Natural and
Articial Celestial Bodies, H. Pretka-Ziomek, E. Wnuk, P. K. Seidelmann, and D. Richardson, Eds., pp.
407413, Kluwer, Dordrecht, The Netherlands, 2001.
8 J. K. S. Formiga, Study of resonances in the orbital motion of articial satellites, Master dissertation, Fac-
uldade de Engenharia do Campus de Guaratingueta, Universidade Estadual Paulista, Guaratingueta,
Brazil, 2005.
9 R. Vilhena de Moraes, J. K. S. Formiga, P. H. C. L. Lima Jr., and H. K. Kuga, Orbital perturbations:
resonance eects, in Proceedings of 6th International Symposium of the IAA on Small Satellites for Earth
Observation, p. 8, Berlin, Germany, April 2007, paper IAA-B6-0717P.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 510759, 22 pages
doi:10.1155/2009/510759

Research Article
Some Initial Conditions for Disposed Satellites of
the Systems GPS and Galileo Constellations

Diogo Merguizo Sanchez,1 Tadashi Yokoyama,1

Pedro Ivo de Oliveira Brasil,2 and Ricardo Reis Cordeiro3
1
Departamento de Estatstica, Matematica Aplicada e Computacao, IGCE, UNESP campus Rio Claro,
Caixa Postal 178, CEP: 13506-900 Rio Claro, SP, Brazil
2
Departamento de Fsica, IGCE, UNESP campus Rio Claro, CEP: 13506-900 Rio Claro, SP, Brazil
3
Departamento de Fsica, Centro de Ciencias Exatas e Tecnologicas, Universidade Federal de Vicosa, UFV,
CEP: 36571-000 Vicosa, MG, Brazil

Correspondence should be addressed to Diogo Merguizo Sanchez, [email protected]

Received 31 July 2009; Accepted 20 October 2009

Recommended by Silvia Maria Giuliatti Winter

Through the averaged equations we revisit theoretical and numerical aspects of the strong
resonance that increases the eccentricity of the disposed objects of GPS and Galileo Systems. A
simple view of the phase space of the problem shows that the resonance does not depend on the
semi-major axis. Therefore, usual strategies of changing altitude raising perigee do not work. In
this problem we search for a set of initial conditions such that the deactivated satellites or upper-
stages remain at least for 250 years without penetrating in the orbits of the operational satellites. In
the case that Moons perturbation is not signicant, we can identify, in the phase space, the regions
where eccentricity reaches maximum and minimum values so that possible risks of collision can
be avoided. This is done semi-analytically through the averaged system of the problem. Guided
by this idea, we numerically found the ,  values of the real unaveraged problem. In particular,
for the Galileo case, the theoretical results predicted in the averaged system are in good agreement
with numerical results. We also show that initial inclination of the Moon plays an important role
in the search of these conditions.

Copyright q 2009 Diogo Merguizo Sanchez et al. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
Broadly speaking, the GPS, GLONASS, and Galileo 1 systems are satellite constellations
which were designed mainly for positioning and navigation purposes. The rst members
of GPS block I originally were designed to have inclination of 63.4 degrees with respect
to the equator, distributed in three orbital planes, each one separated from 120 degrees in
the longitude of the node. The altitude is 20,200 km. The GLONASS members are similar,
with slightly lower altitude 19,100 km, orbital period  11:15 h. The European GALILEO
2 Mathematical Problems in Engineering

system is still in construction, and the inclination of the satellites will be 55 or 56 degrees, with
altitude of 23,615 km. All the three systems have rather similar altitudes. In order to avoid
risks of collision and following American Govern instructions about debris mitigation, there
are some recommendations that the disposal satellites and upper-stages should be deposited
at least 500 km above or below the semisynchronous orbit 2.
In a constellation of a navigation system, the members must be kept under precise
requirements of functionality. However, after some time, they have to be deactivated since
some level of these requirements cannot be fullled for long time. The destination of these
deactivated objects is a problem since they must be moved into some disposal regions in order
to preclude collisions with operational members of the constellation. While these vehicles can
be designed a priori to transport additional propellant at some nonnegligible cost to be used
in some planned maneuvers to insert them in the disposal regions; the same is not true for the
upper-stage. In some cases block IIF of GPS system, due to design restrictions, this upper-
stage cannot be easily guided to the disposal region. It must perform several operations after
the satellite is injected in the constellation. All these operations change its nal parameters
3. Since the inclination of these vehicles is near to 55 or 56 degrees, the eccentricity suers
strong variations and even an initially circular orbit can become highly eccentric so that
they can cross very easily the orbit of the operational satellites. What is interesting and also
problematic is the fact that the rate of growing of the eccentricity is very sensitive to the initial
parameters of the disposal orbit eccentricity, argument of the perigee, and longitude of the
node. In this work, based on a theoretical framework, we present a set of initial conditions
argument of the perigee, longitude of the node for GPS and Galileo systems such that the
disposed objects can remain at the least 250 years with small eccentricity 0.01 or 0.02 without
causing any risk to the operational satellites.
The above strategy of keeping small eccentricity can generate some additional
problem: after some time, the disposed vehicles will accumulate and a graveyard of these
objects will be created. Therefore, a risk of collisions amongst themselves is a crucial problem,
since the products of these extra collisions are almost untrackable fragments that may oer
more risks to the operational elements of the constellation.
According to Jenkin and Gick 4, the strategy in the opposite direction, that is,
exploiting the growth of the eccentricity in order to dilute disposal orbit collision risk has
some interesting points to be considered: the percentage of disposed vehicles that will reenter
in the atmosphere can be increased. Another advantage observed is, although eccentricity
growth strategy increases the collision risk in the constellation, that in some cases this risk
can be reversed with proper choice of the initial disposal eccentricity.
In this sense, we briey started the investigation of some initial conditions that can
cause large increase of the eccentricity, for a minimum time interval, considering also dierent
initial inclinations of the Moons orbit see Appendix B. Our calculations show that the
growth of the eccentricity is rather sensitive to the Moons position inclination and semi-
major axis.

2. Methods
2.1. Disturbing Function of the Sun
As we want to highlight some theoretical aspects, it is instructive to write the main disturbing
forces in terms of the orbital elements.
Mathematical Problems in Engineering 3

In this section we obtain the averaged disturbing function of the Sun. Following the
classical procedure 5, in a reference center xed in the Earth equator, the disturbing function
of the Sun is

 
1 r r
R  k M
2
, 2.1
|r r | |r |3

where M is the mass of the Sun, k2 is the gravitational constant, and r, r are position vector
of the satellite and the Sun, respectively.
Expanding 2.1 in powers of r/r  up to order 2, we have

 
r2 1 3
R  k2 M cos 2
S , 2.2
r3 2 2

where S is the angular distance between the satellite and the Sun. We use the classical
notations a, e, I, f, , and , for semi-major axis, eccentricity, inclination, true anomaly,
argument of the perigee, and longitude of the node. The same set is used for the Suns
elements, adding the index .
In order to write cosS in terms of orbital elements, we have see Figure 1

x x y y z z
cosS  . 2.3
r r r r r r

Considering classical relations of the two-body problem, we write cosS in terms of f,

f , ,  , ,  , I, I as follows:

1 
cosS  1 cosI1 cosI  cos f f  
4
1 
1 cosI1 cosI  cos f f  
4
1 
1 cosI1 cosI  cos f f   2.4
4
1 
1 cosI1 cosI  cos f f  
4
1    
sinI sinI  cos f f  cos f f  ,
2

or in a compact form as follows:

cosS  Aa Bb Cc Dd Ee, 2.5

4 Mathematical Problems in Engineering

Satellite

Sun
f

S

f  Equator


I
I

Figure 1: Geometry of the problem.

where

1
A 1 cosI1 cosI ,
4

a  cos f f   ,

1
B 1 cosI1 cosI ,
4

b  cos f f   ,

1
C 1 cosI1 cosI ,
4 2.6

c  cos f f   ,

1
D 1 cosI1 cosI ,
4

d  cos f f   ,

1
E sinI sinI ,
2
 
e  cos f f  cos f f  .

In order to get rid of the short period variations, we have to obtain the averaged
system and the rigorous procedure is to apply the classical von-Zeipel or Horis method
5, 6. In the present case, as our interest is only to examine the long-term behavior, without
Mathematical Problems in Engineering 5

retrieving the contribution of the short period terms eliminated during the averaging process,
the secular disturbing function can be found simply considering 7, 8

 2
1
R   R dl, 2.7
2 0

where l is the mean anomaly of the satellite. Consider

R  R 

k2 M a2

2r3

 
3 2
P A2 B2 C2 D2 2E2
2 3

3  3 
A2 Z cos 2 2f  2 2 2 C2 Z cos 2 2f  2 2 2
2 2

3  3 
B2 Z cos 2 2f  2 2 2 D2 Z cos 2 2f  2 2 2
2 2

3   3  
Z E2 2CD cos 2 2f  2 Z E2 2AB cos 2 2f  2
2 2
  
3Z E2 AD BC cos2 3P E2 AC BD cos 2f 2

3P AB CD cos2 2  3ACZ cos2 2 2 

 
3ADP cos 2f 2 2 2 3EP A D cos 2f 2 

3EP A B C D cos   3EZA C cos2  

 
3AEZ cos 2 2f  2  3BCP cos 2f 2 2 2

3BDZ cos2 2 2  3EP B C cos 2f 2 

3EZB D cos2   3BEZ cos 2f 2 2 

 
3CEZ cos 2 2f  2  3DEZ cos 2 2f  2  ,
2.8

where P  1 3/2e2 , Z  5/2e2 .

6 Mathematical Problems in Engineering

Performing a second and similar average with respect to the mean anomaly of the Sun,
we get

2 2
   k M a
R 3
4r


P
1 3cos2 I 3cos2 I  9cos2 Icos2 I 
4
3 
Zsin2 I 1 3cos2 I  cos2
2
3
P sin2 Isin2 I  cos2 2 
2
3 2.9
Z1 cosI2 sin2 I  cos2 2 2 
8
3
Z sinI sinI 1 cosI cosI  cos2  
2
3P sinI cosI sinI  cosI  cos  
3  2
Z 1 cos2 I sin2 I  cos2 2 2 
8

3
Z sinI1 cosI sinI  cosI  cos2   .
2

In the above expression, the orbit of the Sun is assumed to be a Keplerian circular orbit. The
elliptic case is briey discussed in Appendix A.

2.2. Oblateness Disturbing Function

For the oblateness, the disturbing function truncated at second order of RP /r is

 
k2 MT R2P 1 3 2
U2  J2 sin , 2.10
r3 2 2

where RP , J2 , and are equatorial radius of the planet, oblateness coecient, and geocentric
latitude of the satellite, respectively. If we proceed in the exact same way as we did before,
we have from the geometry of Figure 1

 
sin  sinI sin f . 2.11
Mathematical Problems in Engineering 7

Once is eliminated, the average of U2 with respect to the mean anomaly of satellite is 9, 10


1 2
U2   U2 dl,
2 0
2.12
1   3/2
RJ2  U2   n2 J2 R2P 3cos2 I 1 1 e2 ,
4

where n is the mean motion of the satellite.

2.3. Some Special Resonances

For close satellites, usually the oblateness is the dominant part. In this case, the main
frequencies of the system are given by

3nJ2 R2P 
5cos2 I 1 ,
4a2 1 e2 2
2.13
3nJ2 R2P
cosI.
2a2 1 e2 2

The ratio of these two frequencies is

2 cosI
 k. 2.14
1 5cos2 I

Note that for k  integer, we have the special resonances which do not depend on the
semi-major axis. These resonances usually aect the eccentricity 8. For k  2, we have
2 0 for I  56.06 or I  110.99 . Another classical resonance occurs when I  63.4 , so
that 0.

3. Effects of 2 and Resonances

In order to see the eects of the resonances which aect GPS and Galileo satellites, the
osculating equations of a satellite will be integrated. For the moment, as disturbers, we
consider only the Sun and the oblateness the complete Cartesian equations involving the
remaining disturbers will be given in Section 4. Note that the resonant conditions to be used
this time are extracted from the averaged system as presented in the precedent section I 
56.06 , I  63.4 . Figures 2 and 3 show the eects of both resonances on the eccentricity and
on the resonant angles. Note that an initial small eccentricity reaches a signicant increase.
8 Mathematical Problems in Engineering

0.5 360

0.4
270
Eccentricity

2
0.3
180
0.2
90
0.1

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
years years

a b

Figure 2: Time evolution of the eccentricity a and the critical angle b. Initial conditions: a  4.805 RT
30,647 km, e  0.005, I  56.06 , and other elements equal to zero. In the simulations, we consider only
Sun and oblateness as disturbers.

0.25 180
0.2
Eccentricity

0.15
90

0.1

0.05

0 0
0 1000 2000 3000 0 1000 2000 3000
years years

a b

Figure 3: Same as Figure 1. Initial conditions: a  4.7 RT  29,977 km, e  0.005, I  63.4 ,  90 , and
other elements equal to zero.

0.7

0.6

0.5
Eccentricity

0.4

0.3

0.2

0.1

0
180 90 0 90 180
2
2 R
Figure 4: Level curves of Hamiltonian R 2 J2 , showing the eccentricity variation versus resonant angle.
Mathematical Problems in Engineering 9

0.5

0.4

Eccentricity
0.3

0.2

0.1

0
0 400 800 1200 1600 2000

80

40
2

40

80
0 400 800 1200 1600 2000
years

Figure 5: Time evolution of the osculating eccentricity top and the osculating critical angle bottom for
a disposal GPS satellite. Note that the minimum of the eccentricity occurs when 2 is crossing zero
in decreasing direction from positive to negative value while maximum occurs when 2 is crossing
zero, but in increasing direction from negative to positive value. Initial conditions: a  3.5 RT 22,324 km,
e  0.005, I  56.06 , and other elements equal to zero; IL  28.58 .

Let us pay more attention to the case I  56.06 which is the inclination of the members
of the Galileo constellation. For this inclination, the dominant term in the R is cos2
 . Neglecting the remaining terms of R   , the Hamiltonian of the problem is

k2 M a2 P 
F  RJ2 3
1 3cos2 I 3cos2 I  9cos2 Icos2 I 
2r 8
 3.1
3
Z sinI sinI 1 cosI cosI  cos2   .
4

(
Let us take L  k2 MT ma, G  L 1 e2 , H  G cosI, l  mean anomaly,  g, and
 h the set of the Delaunay variables. After a trivial Mathieu canonical transformation 11,

G, H, ,  P1 , P2 , 1 , 2 , 3.2

where

G G
1  2 , P1  , 2  , P2  H , 3.3
2 2
10 Mathematical Problems in Engineering

then we have

 
2
2
2   k M a P P1 P2 2 P1 P2 2
R 13 3cos 2
I  9 cos2 I 
2r3 8

4P12 4P12

 1/2
  3.4
3 P1 P2 2 P 1 P2
Z 1 sinI  1 cosI  cos1 ,
4 4P12 2P1
  3/2
P1 P2 2 L2 4P 21
2 J2  1 n2 J2 R2
R 3 1 1 .
P
4 4P12 L2

Since Suns orbit is a Keplerian one, we also considered   0,   0. In Pi , i  variables,

our Hamiltonian is a one degree of freedom problem, whose dynamics is very similar to
the very well-known Lidov-Kozai resonance. In Figure 4, we consider an initial eccentricity
e0  0.005 and semi-major axis a  4.805 RT 30,647 km. This gure is very instructive: note
that in the bottom part there is a large region where the satellite remains some nite time
with very small eccentricity. These are the exactly region we are looking for. It corresponds
to the region where 2 0. On the other hand, we have the counterpart of this situation
at the top of the gure: very high eccentricity, which occurs again for 2 0. We can
separate these two congurations and have a close view of these two cases. Only to conrm
our reasoning, let us integrate the problem in Cartesian coordinates. We also have to decrease
the eect of the Moons perturbation since in the averaged analysis we considered only R 2
and R2 J2 . To do that, we consider convenient value for the semi-major axis. Figure 5 initial
conditions: a  3.5 RT  22323 km, e  0.005, I  56.06 , and other elements equal to zero;
Moon inclination IL  18.28  shows that the minimum of eccentricity occurs when 2
crosses zero from positive to negative values decreasing direction, while maximum occurs
when 2 crosses zero from negative to positive values increasing direction. It is worth
noting that if the semi-major axis is high, then the eect of the Moon cannot be neglected,
so that the problem is no more a one degree of freedom problem. In this case the search of
the ,  pair, such that eccentricity remains small, must be done integrating the complete
equations of the motion.

4. Special ,  Initial Conditions for Galileo Case

In the previous section, we considered only the eects of the Sun and of the oblateness.
Moreover, in the presence of the resonance, the main eects are governed by the long term
variations, so that we eliminated the short period terms. From a theoretical point of view, this
averaged system is quite ecient to highlight the basic dynamics that aects the eccentricity
of the GPS and Galileo satellites. However, for a more complete and realistic study, we need
to include more disturbers.
In this section we want to nd some special initial conditions such that the satellites
can remain stable for at least 250 years with very small eccentricity without causing any
risk of collision to the operational elements of the constellation. The strategy to search these
Mathematical Problems in Engineering 11

360

270

180

90

0
0 90 180 270 360

Figure 6: Black dots: represent ,  values such that a satellite with a  30,647 km remains at least 250
years with eMAX 0.01. Green dots: the same, but eMAX 0.02. Blue dots: curve satisfying 2  0.
Moons inclination: IL  18.28 . Note that most of the stable black dots ,  points satisfy 2  0,
2 with 0,  .

particular initial conditions is guided from the theoretical approach described in the previous
section.
In this section we integrate the osculating elements of a disposal satellite of the Galileo
system under the eect of the Sun, Moon, and the oblateness.
The equations for the osculating elements exact system, including Moon, are

   
k 2 M m r r r r rL rL
r  r k 2 M k ML
2
PJ2 , 4.1
r3 |r r |3 |r |3 |r rL |3 |rL |3
 
3x 15 z2 x
PJx  k 2
MJ2 R2P , 4.2
2r 5 2 r7
 
3y 15 z2 y
PJy  k 2
MJ2 R2P , 4.3
2r 5 2 r7
 
9z 15 z3
PJz  k 2
MJ2 R2P , 4.4
2r 5 2 r7

where PJ2 is the acceleration due to the oblateness, whose Cartesian components are given
by PJx , PJy , PJz 10. The second and third terms in 3.3 are the contributions of the Sun
and the Moon, respectively, and M, m, ML are the masses of the Earth, satellite, and Moon,
12 Mathematical Problems in Engineering

360

270

180

90

0
0 90 180 270 360

Figure 7: Same of Figure 6, but now IL  28.58 . Note that the distribution of the stable ,  is very
sensitive under changes in IL .

0.006 360
0.005
270
Eccentricity

0.004
180
0.003

0.002 90

0.001 0
0 50 100 150 200 250 0 50 100 150 200 250
years years

a b

Figure 8: Time evolution of the eccentricity a and the critical angle b obtained from integration of 4.1.
Initial conditions: a  4.805 RT , e  0.005, I  56.06 . Initial , : 24 , 0 : black, 23 , 2 : red, 18 , 8 :
blue. These initial conditions were extracted from the red square shown in Figure 6. Initial conditions of
the Moon: aL  380, 367.2 km, eL  0.0276, IL  18.28 , L  12.11 , L  92 , and lL  337 . As expected,
the eccentricity remains less than 0.005 for at least 250 years.

respectively. The position vectors of the satellite, Sun, and Moon are indicated by r, r , and
rL .
As we said before, we take 500 km above of the nominal altitude of the constellation.
The initial elements are xed to a  4.805RT 30,647 km, e  0.005, l  0 , and I  56.06 .
We consider two cases for the Moons inclination, I  18.28 and 28.58 . We show that the
initial value of the inclination is important as shown in Figures 6 and 7. In these gures, we
show the pair ,  such that the disposal object remains at least 250 years with eccentricity
smaller than 0.01, so that there is no risk of collision with any member of the constellation.
The black region corresponds to initial conditions such that the maximum eccentricity is
less than 0.01. In the green region, the maximum eccentricity is less than 0.02. The two
Mathematical Problems in Engineering 13

1 360
Eccentricity 0.8 270

2
0.6
180
0.4
0.2 90

0 0
0 50 100 150 200 250 0 50 100 150 200 250
years years

a b

Figure 9: Same as Figure 8 but considering initial conditions from the red small circle of Figure 6. Initial
conditions: a  4.805 RT 30,647 km, e  0.005, I  56.06 ,  90 ,  180 . This time eccentricity grows
fastly since the corresponding initial condition was not taken from the black region of Figure 6.

360

270

180

90

0
0 90 180 270 360

Figure 10: Same of Figure 6, but now a  26,060 km.

straight lines represent the exact condition 2  k in particular we only plot the
case k  0. Note that, in special, the black dots Figure 6 are in fact located in places
predicted from the previous theoretical model. For the remaining gures, the black dots
are slightly shifted upward from the line 2  0. We believe that this is caused by
the strong perturbation of the Moon. Figure 8 shows the time evolution of the eccentricity
for integration whose initial conditions are obtained from Figure 6 small square in the
bottom. As expected, the eccentricity remains very low, while if we take ,  outside
the marked regions in Figure 6 or Figure 7, a signicant increase is veried as shown in
Figure 9. The initial conditions ,  used in this case correspond to the red circle given
in Figure 6.
14 Mathematical Problems in Engineering

360

270

180

90

0
0 90 180 270 360

Figure 11: Same of Figure 10, but now IL  28.58 .

5. ,  Conditions for GPS Case

Here let us consider the GPS system. Again we take I  18.28 and I  28.58 for the Moons
inclination. As before, the importance of the Moons inclination is clear.

6. Tesseral and Sectorial Harmonics

Up to now, we have not considered tesseral and sectorial harmonics. Since GPS
satellites have orbital period near to 12:00 h, the inclusion of such harmonics must
be examined when drawing the gures of Section 5. While the numerical values of
the coecients of these harmonics are very small compared to the zonal harmonics,
due to the 1n : 2 resonance  is the rotation velocity of the Earth, this contribution
could cause some nonnegligible eects. We show very briey these perturbations.
There are several models of the geopotential, including most sophisticated recursive
formulae to generate very high-order JGM-3 12, EGM96S, EGM96 13, etc.. In
this work, we do not need high order model, so that we consider only few terms.
The disturbing function for the general geopotential can be written in the form 14,
15

) *
  n  
k 2 MT k 2 MT  RP n    
RP n  
V Jn Pn0 sin Jnm cos mmn Pnm sin ,
r r n2
r n2 m1
r
6.1

where Jnm , nm are numerical coecients and Pnm are the associated functions of Legen-
dre.
Mathematical Problems in Engineering 15

Zz

Satellite

Figure 12: Geometry of the problem.

Figure 12 describes the fundamental axes of the following reference: the potential 6.1
is referred to the x, y, z which is an equatorial system xed on the Earth; therefore, it rotates
with respect to X, Y, Z which is an inertial system, as follows:

r : the position vector of the satellite.

I: inclination.
: geocentric latitude.
: longitude.
: longitude of node.

According to Figure 12, some simple geometrical relations can be written as


x  r cos cos,

y  r cos sin,

z  r sin ,

x2 y 2 2xy 6.2
cos2  , sin2  ,
x2 y 2 x2 y 2

x2 y2
cos2   , sin2   ,
x2 y 2 x2 y 2

x2 y2  r 2 cos2 .

Let us dene nm  k2 MP /rRP /rn Jnm cos m nm Pnm sin.

16 Mathematical Problems in Engineering

1 1
0.8 0.8
Eccentricity

Eccentricity
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 250 500 750 1000 0 250 500 750 1000
years years

Sun moon RJ2 Sun moon RJ2 22

Sun moon RJ2 22 32 33 Sun moon RJ2 32
Sun moon RJ2 33

a b

Figure 13: Time evolution of eccentricity of GPS satellite. Initial conditions: a  27,059.74 km, e  0.005, 
58 ,  154 . The initial conditions of the Moon correspond to the epoch August 1, 2001. The contribution
of individual and all nm is shown through dierent colors. The dierences appear only after 370 years.

360

270

180

90

0
0 90 180 270 360

Figure 14: Same of Figure 10, including 22 , 32 , and 33 .

Therefore,

3k 2 MP R2P  2 
22  J 22 x y 2
cos222  2xy sin222  . 6.3
r5
Mathematical Problems in Engineering 17

360

270

180

90

0
0 90 180 270 360

Figure 15: Same of Figure 14, but now IL  28.58 .

Proceeding in the similar way, we obtain

15k 2 MP R3P  
32  J32 x2 y2 z cos232  2xyz sin232  ,
r7
6.4
15k 2 MP R3P    
33  J33 x x2 3y 2 cos333  y 3x2 y2 sin333  .
r7

Note that the zonal terms m  0 were already considered before. In Figure 12 the
system x, y, z is xed on the Earth while X, Y, Z is an inertial system, so that x, y, z
rotates with respect to X, Y, Z. Therefore, to have nm referred to X, Y, Z, we consider the
trivial rotation:

x cos sin 0 X

y  sin cos 0 Y , 6.5

z 0 0 1 Z

where  t 0 ,  2/day.
Therefore considering only 22 , 32 , and 33 , the force to be added in 4.1 is


P  gradXY Z 22 32 33 . 6.6

Once we have introduced XYZ system, it is clear that several angular combinations
like 2 appear in the above nm , when 6.3 and 6.4 are expressed in the classical orbital
18 Mathematical Problems in Engineering

elements. Due to the nominal semi-major axis of the GPS satellites, 2 will generate long-
term variations due to their proximity of the 2 : 1 exact resonance. In fact, the orbital period
of GPS satellite is about 12:00 h.
To express nm in terms of the orbital elements is trivial; however, the best way
to see the eects is to keep these terms in Cartesian coordinates without any expansion
in eccentricity or inclination, as the former increases to high values, while the second is
essentially high from the beginning.
The eects of these additional terms are shown in Figure 13. The interaction of
several arguments coming from 22 , 32 , 33 , and also those related to the lunar disturbing
function, give rise to new resonant combinations. It is interesting that the presence of
dierent nm cause signicative dierences in the behavior of the eccentricity, however, in
the beginning up to t  370 years, all the curves are almost coincident. In particular, our
numerical experiments show that if eccentricity remains small, the eects of nm are not
signicant.
Figures 14 and 15 show that the ,  initial conditions are similar to those
gures of Section 4. Note that dierences when nm are included are almost neg-
ligible as expected according to what we learned from Figure 13. However, in the
strategy of exploiting the growth of the eccentricity, the initial ,  can be sensitive,
depending on the time integration and on the number of the harmonics considered
in the geopotential. In Appendix B, we show similar gures, where ,  values
are the initial conditions that cause fast increase of the eccentricity in less than 250
years.

7. Conclusion
With the averaged equations, we showed the dynamics of the 2 resonance. The
reason of the increase of the eccentricity is essentially due to this resonance which
does not depend on the value of the semi-major axis. Therefore, any change of the
semi-major axes raising the perigee of the decommissioned object will not remove
from the resonance. After showing the existence of some initial conditions in the
,  domain where the eccentricity can remain very small based on the averaged
simplied model, we used the complete set of equations to search this pair in , 
plane. The importance of the Moons inclination becomes clear as shown in Figures 6,
7, 10, and 11. We obtained these initial values for GALILEO and GPS systems. For
completeness, we also derived a rst-order averaged system in the eccentricity of the
third body Appendix A. Then several additional resonances appear although their eects
are not so relevant for the navigation system. The search of the ,  pair for the
maximum increase of the eccentricity can be done straightforward following the same
procedure we used for minimum eccentricity. For completeness, in the disturbing function
of the geopotential we also included terms coming from J22 , J32 , and J33 . However,
their contributions for the rst 370 years are not signicant. In a separate paper, we
intend to show the corresponding Figures 10 and 11 considering the second strategy of
exploiting the increase of the eccentricity. Our experiments show that the eect of J22 ,
J32 , and J33 terms becomes more visible only after some hundred years. Therefore, their
contributions in Figures 14 and 15 are almost negligible. However, considering long-
time integration, their eects and interaction with solar and lunar perturbations become
important.
Mathematical Problems in Engineering 19

Appendices
A. Solar Disturbing Function (Up to First Order in e)
Here we give the complete expression of the averaged disturbing function up to rst order in
eccentricity of the third body:

k2 M3 a3
R1  e
2a3

 
9 2 9
P A2 B2 C2 D2 2E2 cos l A2 Z cos2 l 2 2 2 
2 3 4
9 9
A2 Z cos2 3l 2 2 2  B2 Z cos2 l 2 2 2 
4 4
9 9
B2 Z cos2 3l 2 2 2  C2 Z cos2 3l 2 2 2 
4 4
9 9
C2 Z cos2 l 2 2 2  D2 Z cos2 3l 2 2 2 
4 4
9 9 
D2 Z cos2 l 2 2 2  Z E2 2CD cos2 3l 2 
4 4
9  9 
Z E2 2CD cos2 l 2  Z E2 2AB cos2 l 2 
4 4
9  9 
Z E2 2AB cos2 3l 2  Z E2 AD BC cosl 2
4 2
9  9 
Z E2 AD BC cosl 2 P E2 AC BD cosl 2 
2 2
9  9
P E2 AC BD cos3l 2  P AB CD cosl 2 2 
2 2
9 9
P AB CD cosl 2 2  ACZ cosl 2 2 2 
2 2
9 9
ACZ cosl 2 2 2  ADP cosl 2 2 2 
2 2
9 9
ADP cos3l 2 2 2  EP A D cosl 2  
2 2
9 9
EP A D cos3l 2   EP A B C D cosl  
2 2
9 9
EP A B C D cosl   EZA C cosl 2  
2 2
9 9
EZA C cosl 2   AEZ cos2 l 2  
2 2
9 9
AEZ cos2 3l 2   BCP cosl 2 2 2 
2 2
20 Mathematical Problems in Engineering

360

270

180

90

0
0 90 180 270 360

Figure 16: Black dots represent ,  values such that a satellite with a  26,559.74 km reaches e 0.6 in
250 years. Green dots: the same, but eccentricity reaches e 0.5 in 250 years. Blue dots: curve satisfying
2  270. Moons inclination: IL  18.28 .

9 9
BCP cos3l 2 2 2  BDZ cosl 2 2 2 
2 2
9 9
BDZ cosl 2 2 2  EP B C cosl 2  
2 2
9 9
EP B C cos3l 2   EZB D cosl 2  
2 2
9 9
EZB D cosl 2   BEZ cosl 2 2  
2 2
9 9
BEZ cos3l 2 2   CEZ cos2 3l 2  
2 2
9 9
CEZ cos2 l 2   DEZ cos2 3l 2  
2 2

9
DEZ cos2 l 2   .
2
A.1

Note that all the cosines in the above relations have l in the argument. Since the
inclination of satellite is xed about 55 or 56 , each possible resonance can occur for one
particular value of the semi-major axis. However, considering the nominal altitude of the
GPS and Galileo satellites, and since e is small, none of the above combinations of angles is
signicant.
Mathematical Problems in Engineering 21

360

270

180

90

0
0 90 180 270 360

Figure 17: Same of Figure 16, but now IL  28.58 .

1 360
0.8
Eccentricity

0.6
270
0.4
0.2
0 180
0 50 100 150 200 250 0 50 100 150 200 250
years years

a b

Figure 18: Time evolution of eccentricity a and critical angle b of a GPS satellite with a  26,559.74 km
whose eccentricity reaches the value of 0.6 after 200 years. Initial conditions: e  0.005, I  56.06 ,  0 ,
 270 , IL  18.28 .

B. The Eccentricity Increasing Strategy

Following the same strategy to obtain Figures 10, 11, 14, and 15, we obtain similar gures,
but now the pair ,  represents the initial condition of an orbit whose eccentricity attains
e 0.5 in less than 250 years. As before, all orbits start with initial a  26,559.74 km, e  0.005,
and I  56.06 . The maximum eccentricity usually is reached after t  200 years.

Acknowledgments
The authors thank CNPQ, FAPESP, and FUNDUNESP. An anonymous referee is gratefully
thanked for very useful comments.
22 Mathematical Problems in Engineering

References
1 ESA, Galileo: Mission High Level Denition, 2002, http://ec.europa.eu/dgs/energy transport/
galileo/doc/galileo hld v3 23 09 02.pdf.
2 NASA, Guidelines and assessment procedures for limiting orbital debris, NASA Safety Standard
1740.14, Oce of Safety and Mission Assurance, Washington, DC, USA, 1995.
3 A. B. Jenkin and R. A. Gick, Collision risk posed to the global positioning system by disposed upper
stages, Journal of Spacecraft and Rockets, vol. 43, no. 6, pp. 14121418, 2006.
4 A. B. Jenkin and R. A. Gick, Dilution of disposal orbit collision risk for the medium earth orbit
constellations, The Astronomical Journal, vol. 126, pp. 398429, 2005.
5 D. Brouwer and G. M. Clemence, Methods of Celestial Mechanics, Academic Press, New York, NY, USA,
1961.
6 G. Hori, Theory of general perturbation with unspecied canonical variable, Publications of the
Astronomical Society of Japan, vol. 18, pp. 287296, 1966.
7 H. Kinoshita and H. Nakai, Secular perturbations of ctitious satellites of uranus, Celestial Mechanics
and Dynamical Astronomy, vol. 52, no. 3, pp. 293303, 1991.
8 T. Yokoyama, Dynamics of some ctitious satellites of Venus and Mars, Planetary and Space Science,
vol. 47, no. 5, pp. 619627, 1999.
9 Y. Kozai, The motion of a close earth satellite, The Astronomical Journal, vol. 64, pp. 367377, 1959.
10 T. Yokoyama, Possible eects of secular resonances in Phobos and Triton, Planetary and Space Science,
vol. 50, no. 1, pp. 6377, 2002.
11 C. Lanczos, The Variational Principles of Mechanics, Mathematical Expositions, no. 4, University of
Toronto Press, Toronto, Canada, 4th edition, 1970.
12 B. D. Tapley, M. M. Watkins, J. C. Ries, et al., The joint gravity model 3, Journal of Geophysical Research,
vol. 101, no. 12, pp. 2802928049, 1996.
13 F. G. Lemoine, et al., The development of the joint NASA GSFC and NIMA geopotential model
EGM96, Tech. Rep. NASA/TP-1998-206891, NASA/GSFC, 1998.
14 P. R. Escobal, Methods of Orbit Determination, John Wiley & Sons, New York, NY, USA, 1965.
15 O. Montenbruck and E. Gill, Satellite Orbits: Models, Methods and Applications, Springer, Berlin,
Germany, 2005.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 794578, 16 pages
doi:10.1155/2009/794578

Research Article
Quality of TEC Estimated with Mod Ion Using GPS
and GLONASS Data

Paulo de Oliveira Camargo

Department of Cartography, Faculty of Science and Technology, Sao Paulo State University (UNESP),
Rua Roberto Simonsen 305, 19060-900 Presidente Prudente, SP, Brazil

Correspondence should be addressed to Paulo de Oliveira Camargo, [email protected]

Received 1 July 2009; Accepted 28 September 2009

Recommended by Silvia Maria Giuliatti Winter

One of the largest sources of error in positioning and navigation with GNSS is the ionosphere, and
the associated error is directly proportional to the TEC and inversely proportional to the square of
the signal frequency that propagates through the ionosphere. The equatorial region, especially in
Brazil, is where the highest spatial and temporal value variations of the TEC are seen, and where
these various features of the ionosphere, such as the equatorial anomaly and scintillation, can be
found. Thus, the development and assessments of ionospheric models are important. In this paper,
the quality of the TEC was evaluated, as well as the systematic error in the L1 carrier and the inter-
frequency biases of satellites and receivers estimated with the Mod Ion, observable from GPS and
integration with the GLONASS, collected with dual frequency receivers.

Copyright q 2009 Paulo de Oliveira Camargo. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
The Global Navigation Satellite System GNSS is one of the most advanced technologies
and has revolutionized the activities related to navigation and positioning from space
technology. A main component of GNSS is the Global Positioning System GPS, developed
by the United States, and Globalnaya Navigatsionnaya Sputnikowaya Sistema GLONASS
of responsibility of the Republic of Russia. A relevant fact is that, in December 2005, the
rst GALILEO satellite was launched, which is being developed by the European Union
and must come into operation in 2013. The GNSS is composed by the so-called Satellite
Based Augmentation System SBAS, such as the Wide Area Augmentation System WAAS
in USA, European Geostationary Navigation Overlay Service EGNOS in Europe, Multi-
functional Satellite-Based Augmentation Service MSAS in Japan, and Satellite Navigation
Augmentation System SNAS in China 1.
2 Mathematical Problems in Engineering

In general, most users simply use the GNSS system to get their coordinates, without
being committed to details, but for the quality of information precision provided by it,
except aviation. This quality information is very optimistic, demanding attention from users.
However, certain applications require the knowledge of the various processes related to
the system. The mitigation of the eects of the atmosphere troposphere and ionosphere
over GNSS observables requires knowledge of the signal analysis and its behavior in
the atmosphere, requiring interaction with other sciences such as Aeronomy, Meteorology,
among others. Accordingly, this interaction allows studies related to the behavior of the
ionosphere and the troposphere to be made from GNSS observables. In Brazil, the ionosphere
shows a very complex behavior, for being located near the geomagnetic equator, requiring the
development of models and appropriate studies for the region 112.
With the Selective Availability SA deactivation, in the case of GPS, the error due
to the ionosphere has become a major source of systematic error in positioning, especially
in periods of high solar activity for one frequency GNSS users, in the conventional point
positioning as well as relative positioning. Another eect that aects considerably the GNSS
signals is ionosphere scintillation, a result of propagating the signal through a region in which
there are irregularities in the density of electrons.
The error due to the ionosphere depends on the Total Electron Content TEC present
in the ionosphere and in signal frequency. Users of, at least, dual frequency receivers can
make corrections of this eect, using the ionospheric free linear combination. This observable
eliminates rst-order ionospheric eect. Users of single frequency receivers, however, need
to correct the systematic eect observables due to the ionosphere. The quantication of
this eect can be done by 3, 4, 12: coecients transmitted by navigation messages, using
the Klobuchar model; observations collected with one or dual frequency GNSS receivers of
Ionosphere map Exchange format IONEX archives obtained from Global Ionospheric Maps
GIM, which provide values of vertical TEC VTEC in a grid with spatial resolution of
5 2.5 in longitude and latitude, respectively, and temporal resolution of 2 hours 13.
In the geodetic community, one of the models used and implemented in commercial
software, to minimize the eects of the ionosphere on GPS observables, is the Klobuchar
model. This model, also called Broadcast model, estimates the systematic error due to
the ionosphere to one frequency receivers 14 and its coecients are transmitted by GPS
satellites in navigation messages. However, this model removes around 50% to 60% of the
total eect 15, 16. Being more appropriate for use in regions of middle latitudes, which is
the more predictable ionospheric region, where the ionosphere has a more regular behavior.
However, this is not an appropriate model to be used in Brazil, where there is high variation
in the density of electrons as well as in South America. So with the need to have a more
eective correction strategy of the ionosphere eect, several models were developed by
various research centers and universities, using observations collected with dual frequency
GPS receivers. In terms of South America, we can quote the La Plata Ionospheric Model
LPIM model, developed at Astronomical and Geophysical Sciences Faculty of Universidad
Nacional de La Plata FCAG/UNLP, Argentine 2 and the Regional Model of Ionosphere
Mod Ion developed in FCT/UNESP, Brazil 3.
The accuracy of VTEC values in the nal IONEX les grid 11 days of latency is 28
TEC units TECUs and for rapid les <24 hours of 29 TECU 17. Ciraolo et al. 18, in
a calibration process, determined the interfrequency bias IFB of a pair of receivers, which
ranged from 1.4 to 8.8 TECU. This paper aims to assess the quality of TEC and the error in the
L1 carrier estimated with the Mod Ion, from GPS observables and integration with the ones
from GLONASS, collected with dual frequency receivers.
Mathematical Problems in Engineering 3

The paper is organized as follows. In Section 2 there is a brief description of impact

of the ionosphere on the propagation of GNSS signals; Section 3 describes the equations,
based on the geometry-free linear combination of observables collected with dual frequency
receivers, used in the Mod Ion, as well as gets the TEC and the systematic error due to
the ionosphere in the L1 carrier, and some aspects of adjustment by the least squares; the
results and analysis of the experiments in order to verify the quality of the TEC provided by
model, as well as the IFB of the satellites and receivers are presented in Section 4; based on
the experiments, conclusions and future works will be presented in Section 5.

2. Impact of the Ionosphere on the Propagation of GNSS Signals

The terrestrial atmosphere, for practical purposes, can be considered as a set of gas layers,
spherical and concentric to the Earth. Its structure is related to various thermal, chemical, and
electromagnetic elements. These combined parameters vary depending on the time, latitude,
longitude, time of year, and solar activity.
With respect to the propagation of electromagnetic waves, the Earths atmosphere
is divided into ionosphere and troposphere. In this division, the troposphere is the layer
between the Earths surface up to 50 km in height. It is composed of neutral particles, and
the highest concentration of gas is found on up to a height of 12 km, consisting of nitrogen,
oxygen, carbon dioxide, argon, water vapor, among others. The propagation of the signal in
the troposphere depends mainly on the water vapor content, air pressures, and temperature.
For frequencies below 30 GHz, the refraction does not depend on the frequency of the signal
transmitted 16.
The ionosphere is dened as the portion of the upper atmosphere, where there is
sucient ionization to aect the propagation of radio waves 19. Unlike the troposphere,
it is a dispersive medium; that is, in this case, signal propagation depends on the frequency. It
is characterized mainly by the formation of ions and electrons, and it starts at around 50 km,
extending to approximately 1000 km in height.
In the region covered by the ionosphere, the electron density is sucient to alter the
propagation of electromagnetic waves. The ions and free electrons in the ionosphere are
mainly created by the process of photo ionization. The ionospheric photo ionization is the
absorption of solar radiation, predominantly in the range of extreme ultraviolet and X-rays
by neutral atmospheric elements 1921. The ionosphere as a dispersive means aects the
modulation and phase of the carrier, causing, respectively, a delay and an advance 16. The
delay is also referred to as ionospheric delay and increases the apparent length of the path
traveled by the signal.
The troposphere eects on GNSS signals are usually reduced by processing techniques
or determined directly by models. Since it is not possible to assess the atmospheric pressure
and temperature along the route of the signal through the neutral layer, there are several
models available, which correct for 92% to 95% of this eect 22. In contrast, the ionosphere
eect, which depends on frequency and, hence, on the refractive index proportional to the
TEC, that is, to the number of electrons present along the path between the satellite and the
receiver. If the TEC values were constant, the eects caused by the ionosphere would be easy
to determine. The problem is that the TEC varies in time and space, in relation daytime,
season, solar cycle, geographical location of the receiver and Earths magnetic eld, and so
forth. Besides the refraction eect, these variations can cause the receiver to go out of tune
with the satellite, by weakening the signal strength, the specic case of the phenomenon
known as scintillation.
4 Mathematical Problems in Engineering

Table 1: Maximum vertical ionospheric range error m.

Frequency 1st-order eect 1/f 2  2nd-order eect 1/f 3  3rd-order eect 1/f 4 
L1 32.5 0.036 0.002
L2 53.5 0.076 0.007
L0 0.0 0.026 0.006

The ionosphere eects are divided into eects of 1st, 2nd, and 3rd order. Table 1 shows
the maximum error in the vertical direction, which can be expected for the GPS L1 , L2 carriers
and for the ionospheric free linear combination L0 . For inclined directions, the inuence
increases 1.
The error or eect of 1st order, due to the ionosphere in phase Ifr
s
 and pseudorange
Igr  along the satellite direction s and receiver antenna r, is given according to the TEC
s

and the frequency of the signal f 3, 16:

40.3
s
Ifr  TEC, 2.1
f2

40.3
s
Igr  TEC. 2.2
f2

According to 2.1 and 2.2, we can see that the errors due to the ionosphere
for the phase and pseudorange have the same magnitude but opposite signs. Both are
proportional to the TEC and inversely proportional to the square of the frequency of the
carrier. The TEC unit TECU is given in electrons per square meter el/m2  and the
constant 40.3 m Hz2 el/m2 1 . The eect of rst order can be obtained from the free geometry
linear combination using observables collected with GPS receivers and/or dual frequency
GPS/GLONASS, and the remaining error represents a few centimeters 1.
The eect of second order of the ionosphere depends on, besides the TEC and the
frequency, geomagnetic induction at the point where the signal passes through the layer of
the ionosphere and the angle of the signal in the geomagnetic induction vector. Unlike the
eect of rst order which is the same and has opposite signals to the phase and pseudorange,
the one of the second order of the phase is half of the second-order eect of the group 23.
But the eect of third order does not depend on the magnetic eld, but is a function
of maximum density of electrons, at the phase the eect is equivalent to one third of the
pseudorange eect 23.

2.1. Regular Variations of the TEC

The regular temporal changes of the TEC include daytime and seasonal variations and cycles
of long periods. The daytime variation is mainly due to Sunlight, that is, solar radiation.
Throughout the day, the density of electrons depends on the local time, with its peak
occurring between 12 and 16 local times 24. In the low latitude equatorial region, a second
peak occurs in the hours preceding midnight, especially in periods close to the equinoxes and
to the summer and during periods of high solar activity.
Mathematical Problems in Engineering 5

Seasons also inuence variation in electron density, due to the change in the zenithal
angle of the sun and the intensity of the ionization ow, characterizing seasonal variations.
During the equinoxes, the eects of the ionosphere are bigger, whereas in the solstices, they
are smaller 5.
Changes in long-period cycles, with cycles of approximately 11 years, are associated
with the occurrence of sunspots and the increase of ionization and thus the TEC is
proportional to the number of spots.
The geographic location also inuences the variation of the density of electrons in the
ionosphere, because the overall structure of the ionosphere is not homogeneous. It changes
with latitude, due to the variation of the zenithal angle of the Sun, which inuences directly,
the level of radiation, which changes, in turn, the density of electrons in the ionosphere.
The equatorial regions are characterized by a high density of electrons and have a high
spatial variation. The regions of middle latitudes, however, are considered relatively free
from ionospheric abnormalities, presenting a more regular behavior, close to that described
by theoretical models. The ionosphere over the north and south poles, alternatively, known
as polar or high latitudes ionosphere, is extremely unstable 20. More details on the changes
of regular TEC can be obtained, for example, from 19, 20.

3. Regional Ionosphere Model (Mod Ion)

The Mod Ion was developed in FCT/UNESP to represent the ionosphere in an analytical
way 3. The parameters of the model are estimated from data collected with dual frequency
GNSS receivers. With the introduction of several receivers it was possible to also estimate the
systematic error due to satellites and receivers, called Dierential Code Bias DCB or IFB,
caused by the signal route on the hardware of satellites, until it was spread out on space, and
on antenna cables and hardware of receivers, until the signal decorrelation.
The adjustment by the Least Squares Method LSM with constraints is used in bath
in the process of estimating the parameters of the model. The GNSS observable used in the
calculation of the TEC or the systematic error due to the ionosphere in the L1 carrier is the
pseudorange ltered by the carrier phase 25. The original observable can also be used as
well as the carrier phase.

3.1. Ionospheric Model

Models that use GNSS data are based on the geometry-free linear combination of observables
collected with dual frequency receivers. In the derivation of the model, errors due to
nonsynchronism of the satellite and receiver, ephemeredes and the tropospheric refraction
are neglected, since their eects contaminate both frequencies the same way and do not aect
the validity of results.
The model is based on the dierence between the pseudoranges of the carriers L2 and
L1 , with frequencies f2 and f1 of signals generated by the satellites that are part of the GNSS
3:

 
s
P2r P1r
s
 I2r
s
I1r
s
Ssp2 Ssp1 Rp2 Rp1 p21 . 3.1
6 Mathematical Problems in Engineering

From 2.2 we have

f12 f22 f12 f22

s
I2r I1r
s
 40.3 TECs  I1r
s
, 3.2
f12 f22 f22

thus

 s   
F TEC P2r P1r
s
 TECsr F TEC Ssp2 Ssp1 Rp2 Rp1 F TEC p21 3.3

or

 s   
F I1 P2r P1r
s
 I1r
s
F I1 Ssp2 Ssp1 Rp2 Rp1 F I1 p21 . 3.4

Equation 3.3 is the observation equation of Mod Ion used to calculate the TEC in the
satellite/receiver direction. The unknowns Ssp2 Ssp1  and Rp2 Rp1  represent, respectively,
the IFBs of satellites and receivers, and P 21 represents another dierential remaining errors
multipath, receiver noise, etc., where F TEC  f12 f22 /40.3f12 f22 , in general representing a
constant for the GPS satellites and particularly for each of the GLONASS satellites.
By 3.4, one can calculate the ionospheric delay, that is, ionospheric error I1r s
 in the
L1 carrier, in the satellite/receiver direction, with F I1  f22 /f12 f22 .
The TEC or the ionospheric delay along the path of the satellite/receiver can be
obtained according to the VTEC or the vertical ionospheric delay I1v , by the expression,
assigned as standard geometric mapping function 1/cos z , which provides the slant factor,
like this

VTEC
TECsr  3.5
cos zs
r

or

I1v
s
I1r  3.6
cos zs
r

for a receiver r, zs is the zenithal angle of the signal path from the satellite s to a
ionospheric point located in a ionospheric layer, for example, of 400 km of height. Then

 s VTEC   
F TEC P2r P1r
s
 F TEC
S s
S s
Rp2 Rp1 F TEC p21 3.7
cos zs
r
p2 p1

or

 s I1v   
F I1 P2r P1r
s
 F I1 Ssp2 Ssp1 Rp2 Rp1 F I1 p21 . 3.8
cos zs
r
Mathematical Problems in Engineering 7

Due to the periodic nature of the eect, to model the diurnal behavior of the VTEC or
the error in the L1 carrier 10 use the series


n4
, -
VTEC or I1V  a1 a2 Bs aj cosiBs  aj 1 siniBs 
i1
j2i 1
3.9

m4
, -
an 2 3 h
2
aj cosihs  aj 1 sinihs  .
i1
j2i 10

The variable Bs represents geographic latitude of the subionospheric point projection

of a point on ionospheric layer on the earth surface and variable hs is given as

2 
hs  t 14h , 3.10
T

where T represents the 24-hour period and t the local time of the subionospheric point.
The total number of parameters of the model is given by 4 4 3 r s, where the

4 4 3 represent the coecients of the series, r is the receivers IFB, a total equal to the number
of receivers used in the network, and s is the satellites IFB, which is equal to the number of
satellites tracked to determine the parameters of the model.
In adjustment by least squares, matrix A shows rank deciency, equal to two. This
implies that satellite or receivers IFBs have to be determined for two of them, one regarding
GPS and the other regarding GLONASS. Thus, the constraints were imposed in one of the
GPS/GLONASS receivers.

4. Experiments, Results, and Analysis

The experiments were performed at the Laboratory of Space Geodesy of the FCT/UNESP,
where 4 dual frequency GPS/GLONASS receivers were connected to a TRM 55971.00 Zephyr
GNSS Geodetic Model 2 Antenna, using a splitter with 4 outputs. Data were collected for 15
days in the year 2007 132 to 137, 153 to 157, and 173 to 177 using 2 Topcon TPS HYPER
GGD H826 and H819, TRIMBLE NTR5, and LEICA GRX1200 GGPRO receivers. Two
experiments were conducted, the rst using only GPS observables, and the second aiming
the integration of GPS and GLONASS systems. The experiments were conducted using data
in RINEX format, with observables collected every 15 seconds, with 20 degrees elevation.
The precise ephemerides and satellite clocks of the International GNNS Service IGS were
used. It is worth mentioning that all experiments passed the quality control of adjustment
and, according to the Dst geomagnetic index, the observables were collected in condition of
weak geomagnetic storm 30 nT to 50 nT, and that on days in question they did not exceed
25 nT.
Receiver H826 was chosen as a reference for estimation of IFBs, and relatively
constrained as zero and weight tending to innity, since the value of the receivers IFB is
unknown. Some GLONASS satellites did not participate in some days, for being under main-
tenance in the quoted period http://gge.unb.ca/Resources/GLONASSConstellationPlot
.pdf.
8 Mathematical Problems in Engineering

IFB-satellites
10
5
0

TEC TECU
5
10
15
20
25
30
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Experiments

G1 G7 G13 G20 G26

G2 G8 G14 G21 G27
G3 G9 G16 G22 G28
G4 G10 G17 G23 G29
G5 G11 G18 G24 G30
G6 G12 G19 G25 G31

Figure 1: GPS satellite IFBError in TEC.

IFB-receivers
5
0
5
TEC TECU

10
15
20
25
30
35
40
45
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Experiments

H819
NETR
LEIC

Figure 2: GPS receiver IFBError in TEC.

It was yet adopted as a criterion for rejection of the observables, the standard one,
which represents the upper limit of change in the TEC for static users, given by 0.1 1016
el/m2  per second 26. This value represents 0.085 Hz in L1 GPS and corresponds to
0.0163 m/s of change in pseudoranges due to the ionospheric eect. The dierences between
consecutive linear combinations bigger than 0.0163 m/s imply the rejection of the observables
used to estimate the parameters of the model.

4.1. Satellites and Receivers IFBs Obtained with GPS Observables

Using established procedures, the satellite and receiver IFBs were estimated, as well as the
coecients of the series that allows the calculation of the VTEC considering only the GPS
observables. On Figures 1 and 2, the satellites and receivers IFBs are presented, in TECU.
Experiments 1 to 5 correspond to the days of the year 133 to 137 of 2007, 6 to 10 correspond
to days 153 to 157/2007, and 11 to 15 correspond to days 173 to 177/2007.
Mathematical Problems in Engineering 9

IFB-satellites
2
1

Error in L1 m
0
1
2
3
4
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Experiments

G1 G7 G13 G20 G26

G2 G8 G14 G21 G27
G3 G9 G16 G22 G28
G4 G10 G17 G23 G29
G5 G11 G18 G24 G30
G6 G12 G19 G25 G31

Figure 3: GPS satellite IFBError in L1 .

Analyzing Figure 1, we nd that the satellites IFBs showed a similar behavior, except
for satellite 2 G2 in the 4th and 5th experiments, and satellite 16 G16, whose variation was
10.658 TECU. This behavior shows that the systematic errors of the satellites are not stable.
It is worth mentioning that the IFBs of satellites include the IFB of receiver H826 which was
adopted as reference. The Root Mean Square RMS error indicates that you can estimate the
IFBs of satellites with precision, better than 2.735 TECU.
The receiver IFB of H826 was constrained as being zero, the values receivers IFBs
were estimated in relation to the receiver. On Figure 2, we observed that receivers showed
a behavior very similar and stable, but with dierent values for each. Receiver H819 features
an IFB very close to the one adopted as reference, with an average of 0.500 TECU. For
receivers NETR and LEIC, the values were, for the trial period, respectively, 41.510 and
42.564 TECU. The variation of the IFB of receivers was around 9 times less than the ones of
satellites, indicating the stability of the receivers. The RMS indicates that one can estimate the
IFBs of receivers with accuracy better than 0.329 TECU.
Regarding IFBs in L1 carrier, which represent the systematic error that aects GPS
observables made in L1 , values can be obtained using 2.2 or through the Mod Ion. To
determine the error in GPS L2 carrier can also be used by the same equation or multiply the
error in L1 by the constant 1.64694. Figures 3 and 4 show IFBs due to satellites and receivers
for L1 carrier, in units of m, respectively.
The error in the L1 carrier due to satellites showed RMS better than 0.444 m and the
receivers better than 0.054 m.

4.2. VTEC Obtained with GPS Observables

From geometry-free linear combination see 3.3, applying the correction of IFBs due to
satellites and receivers can get a set of values of VTEC for each of the 4 receivers. To calculate
the dierences of VTEC, they took as reference the value obtained from the 19 coecients
estimated for the series see 3.9, which analytically represents the ionosphere. Figures 5 to
7 show the discrepancies of VTEC in the quoted period. For each of the experiments it is also
presented the values of VTEC determined analytically, and used as reference.
10 Mathematical Problems in Engineering

IFB-receivers
0

Error in L1 m
2

8
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Experiments

H819
NETR
LEIC

Figure 4: GPS receiver IFBError in L1 .

Analyzing Figures 5, 6, and 7, we see that the values of VTEC did not exceed 30 units,
because the experiments were conducted over a period of low solar activity. The receivers
of the same manufacturer H826 and H819 have the same behavior for the discrepancies
of VTEC, and the daily average is less than 0.227 VTEC units, the RMS indicates that
the precision with which the VTEC is estimated is better than 2.365 units, and the biggest
discrepancy was 10.917 units. The behavior of receiver NTER is noisier and the daily
average and RMS of discrepancies are better than, respectively, 0.244 and 2.396 VTEC units.
Regarding receiver LEIC, there is a little higher value than the other, with daily average of
discrepancies of up to 0.337 units and RMS of 2.713 VTEC units. The modeling also shows
a systematic error, as all the daily average of discrepancies show the same bias, that is, the
same signal.

4.3. Satellites and Receivers IFBs Obtained with GPS/GLONASS Observables

In this experiment for the weight of GLONASS observables was assigned a scale factor of 1/2
on the GPS, because fundamental frequency is half the frequency of the GPS. The experiment
was conducted only with data collected between the days of the year from 133 to 137 of 2007,
and receiver LEIC did not participate because it did not collect GLONASS observables in
this period. In Figure 8, it is presented the IFBs due to GPS and GLONASS satellites and, in
Figure 9, the ones due to receivers.
On Figure 8, we nd that the GPS satellites IFBs had a very similar behavior,
contrary to what occurred with the most part of GLONASS satellites. For GPS satellites,
in relation to IFBs determined only with GPS observables, the biggest dierence to the
RMS was 1.675 TECU G14. The GLONASS satellites showed the biggest variation in
the determination of IFBs, reaching 36.191 TECU, with the RMS value of 12.727 TECU
R14.
GPS receiver IFBs Figure 9 are much more stable than those of the GLONASS
receivers. In the previous experiment, the biggest dierence of the averages for the GPS did
not exceed 0.567 TECU, with RMS of 0.164 TECU. The IBFs related to GLONASS observables
have much variation in the order of up to 13.167 TECU, with values very dispersed in
relation to the average, with RMS being 3.994 and 4.364 TECU, respectively, for H819 and
NETR.
Mathematical Problems in Engineering 11

MOD ION-GPS

25

VTEC
15
5
5
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

a
Discrepancies-H826 GPS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

b
Discrepancies-H819 GPS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

c
Discrepancies-NETR GPS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

d
Discrepancies-LEIC GPS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

e

Figure 5: VTEC and discrepancies GPS: 133 to 137/2007.

4.4. VTEC Obtained with GPS/GLONASS Observables

To evaluate the quality of the TEC obtained with Mod Ion, a modeling was conducted using
the GPS and GLONASS observables simultaneously. Figure 10 shows the daily dierence in
the quoted period, including the modeled VTEC.
12 Mathematical Problems in Engineering

MOD ION-GPS

25

VTEC
15
5
5
153 153.5 154 154.5 155 155.5 156 156.5 157 157.5 158
Day of year/2007

a
Discrepancies-H826 GPS
15
10
5
VTEC

0
5
10
15
153 153.5 154 154.5 155 155.5 156 156.5 157 157.5 158
Day of year/2007

b
Discrepancies-H819 GPS
15
10
5
VTEC

0
5
10
15
153 153.5 154 154.5 155 155.5 156 156.5 157 157.5 158
Day of year/2007

c
Discrepancies-NETR GPS
15
10
5
VTEC

0
5
10
15
153 153.5 154 154.5 155 155.5 156 156.5 157 157.5 158
Day of year/2007

d
Discrepancies-LEIC GPS
15
10
5
VTEC

0
5
10
15
153 153.5 154 154.5 155 155.5 156 156.5 157 157.5 158
Day of year/2007

e

Figure 6: VTEC and discrepancies GPS: 153 to 157/2007.

The behavior of the VTEC discrepancies Figure 10 regarding receivers H826, H819,
and NETR are similar, the daily average is less than 0.891 VTEC units and RMS is better
than 4.929 VTEC. The biggest discrepancy was 30.409 VTEC units, related to the inuences
of GLONASS satellites.
Mathematical Problems in Engineering 13

MOD ION-GPS

25

VTEC
15
5
5
173 173.5 174 174.5 175 175.5 176 176.5 177 177.5 178
Day of year/2007

a
Discrepancies-H826 GPS
15
10
5
VTEC

0
5
10
15
173 173.5 174 174.5 175 175.5 176 176.5 177 177.5 178
Day of year/2007

b
Discrepancies-H819 GPS
15
10
5
VTEC

0
5
10
15
173 173.5 174 174.5 175 175.5 176 176.5 177 177.5 178
Day of year/2007

c
Discrepancies-NETR GPS
15
10
5
VTEC

0
5
10
15
173 173.5 174 174.5 175 175.5 176 176.5 177 177.5 178
Day of year/2007

d
Discrepancies-LEIC GPS
15
10
5
VTEC

0
5
10
15
173 173.5 174 174.5 175 175.5 176 176.5 177 177.5 178
Day of year/2007

e

Figure 7: VTEC and discrepancies GPS: 173 to 177/2007.

5. Conclusions and Future Works

Results showed that, when using only GPS observables, you get the estimation of satellites
IFBs with RMS better than 2.735 TECU, and better than 0.329 TECU for those of receivers.
14 Mathematical Problems in Engineering

IFB-satellites
20

TEC TECU
0
20
40
60
80
1 2 3 4 5
Experiments

G1 G9 G18 G26 R4
G2 G10 G19 G27 R6
G3 G11 G20 G28 R7
G4 G12 G21 G29 R8
G5 G13 G22 G30 R10
G6 G14 G23 G31 R14
G7 G16 G24 R1 R23
G8 G17 G25 R3

Figure 8: GPS and GLONASS satellite IFBError in TEC.

IFB-receivers
5
0
5
10
TEC TECU

15
20
25
30
35
40
45
1 2 3 4 5
Experiments

H819-GPS H819-GLONASS
NETR-GPS NETR-GLONASS

Figure 9: GPS and GLONASS receiver IFBError in TEC.

It represents, respectively, in L1 carrier, errors of 0.444 m and 0.054 m, respectively, for

satellites and receivers used. The RMS obtained for the estimation of the VTEC is better than
2.713 units, which for L1 carrier is an error of 0.441 m. The RMS obtained by integrating GPS
and GLONASS was better than, respectively, for the satellites and receivers IFBs and VTEC
of 12.727, 4.364 TECU, and 4.929 VTEC units, representing an error of 6.568 m in L1 carrier,
0.709 m, and 0.800 m respectively. Out of experiments conducted, it is concluded that the GPS
observables show better quality than when combined with GLONASS, and compatible with
the nal values determined with GIM, which is about 28 TECU.
Also, new experiments will be conducted using data collected in times of minimum
and maximum solar activity, because the solar cycle is the period of minimal activity and
periods of irregularities in the ionosphere, having the performance of the model analyzed
in the result of absolute and relative positioning with one frequency receivers. In the case of
relative positioning, the resolution of ambiguities will also be evaluated.
Mathematical Problems in Engineering 15

MOD ION-GPS/GLONASS

25

VTEC
15
5
5
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

a
Discrepancies-H826 GPS/GLONASS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

b
Discrepancies-H819 GPS/GLONASS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

c
Discrepancies-NETR GPS/GLONASS
15
10
5
VTEC

0
5
10
15
133 133.5 134 134.5 135 135.5 136 136.5 137 137.5 138
Day of year/2007

d

Figure 10: VTEC and discrepancies GPS/GLONASS: 133 to 137/2007.

About receivers IFBs are necessary to develop methodologies to calibrate and

constraint them in the modeling process. It will also be implemented the calculations of eects
of 2nd and 3rd orders in the model, in order to provide all the eects of the ionosphere to the
users of GPS, GLONASS and, in the future, GALILEO.
And nally, with the modernization and expansion of GNSS networks in Brazil, it is
possible to produce maps of the ionosphere in terms of TEC and/or eects on L1 carrier.

Acknowledgments
The author thanks CNPq for the nancial support Process: 482052/2007-4 and Research
Productivity GrantPQ Process: 309332/2007-9.
16 Mathematical Problems in Engineering

References
1 G. Seeber, Satellite Geodesy, Walter de Gruyter, Berlin, Germany, 2003.
2 C. A. Brunini, Global ionospheric model from GPS measurement, Ph.D. thesis, Facultad de Ciencias
Astronomicas y Geofsicas, Universidad Nacional de La Plata, La Plata, Argentina, 1998.
3 P. de Oliveira Camargo, Modelo regional da ionosfera para uso em posicionamento com receptores de uma
frequencia, Ph.D. thesis, Universidade Federal do Parana, Curitiba, Brazil, 1999.
4 P. de Oliveira Camargo, J. F. G. Monico, and L. D. D. Ferreira, Application of ionospheric corrections
in the equatorial region for L1 GPS users, Earth, Planets and Space, vol. 52, no. 11, pp. 10831089, 2000.
5 E. S. Fonseca Jr., O sistema GPS como ferramenta para avaliacao da refracao ionosferica no Brasil, Ph.D.
thesis, Escola Politecnica da Universidade de Sao Paulo, Sao Paulo, Brazil, 2002.
6 M. Fedrizzi, Estudo do efeito das tempestades magneticas sobre a ionosfera utilizando dados do GPS, Ph.D.
thesis, Instituto Nacional de Pesquisas Espaciais, Sao Jose dos Campos, Brazil, 2003.
7 M. T. Matsuoka and P. de Oliveira Camargo, Evaluation of functions for modelling of the eect of the
ionospheric refraction in the propagation of the GPS signals, in Proceedings of the 16th International
Technical Meeting of the Satellite Division of the Institute of Navigation (ION GPS/GNSS 03), pp. 1736
1741, Portland, Ore, USA, 2003.
8 M. T. Matsuoka, P. de Oliveira Camargo, and W. R. Dal Poz, Declnio do numero de manchas solares
do ciclo 23: reducao da atividade ionosferica e melhora da performance do posicionamento com
GPS, Boletim de Ciencias Geodesicas, vol. 10, no. 2, pp. 141157, 2004.
9 W. R. Dal Poz and P. de Oliveira Camargo, Consequencias de uma tempestade geomagnetica no
posicionamento relativo com receptores GPS de simples frequencia, Boletim de Ciencias Geodesicas,
vol. 12, no. 2, pp. 275294, 2006.
10 C. R. de Aguiar and P. de Oliveira Camargo, Modelagem em tempo real do erro sistematico das
observaveis GPS devido a ionosfera, Boletim de Ciencias Geodesicas, vol. 12, no. 1, pp. 101120, 2006.
11 I. S. Batista, M. A. Abdu, J. R. Souza, et al., Unusual early morning development of the equatorial
anomaly in the Brazilian sector during the Halloween magnetic storm, Journal of Geophysical Research
A, vol. 111, no. 5, Article ID A05307, 2006.
12 M. T. Matsuoka, Inuencia de diferentes condicoes da ionosfera no posicionamento com GPS: avaliacao na
regiao brasileira, Ph.D. thesis, Faculdade de Ciencias e Tecnologia, Universidade Estadual Paulista,
Presidente Prudente, Brazil, 2007.
13 S. Schaer, Mapping and predicting the earths ionosphere using the Global Positioning System, Ph.D.
dissertation, University of Bern, Bern, Switzerland, 1999.
14 J. A. Klobuchar, Ionospheric time-delay algorithm for single-frequency GPS users, IEEE Transactions
on Aerospace and Electronic Systems, vol. 23, no. 3, pp. 325331, 1987.
15 S. P. Newby and R. B. Langley, Three alternative empirical ionospheric modelsare they better than
GPS broadcast model? in Proceeding of the 6th International Geodetic Symposium on Satellite Positioning,
vol. 1, pp. 240244, Columbus, Ohio, USA, 1992.
16 A. Leick, GPS Satellite Surveying, John Wiley & Sons, New York, NY, USA, 1995.
17 IGS, International GNSS Service, 2009, http://igscb.jpl.nasa.gov/components/prods.html.
18 L. Ciraolo, F. Azpilicueta, C. Brunini, A. Meza, and S. M. Radicella, Calibration errors on
experimental slant total electron content TEC determined with GPS, Journal of Geodesy, vol. 81,
no. 2, pp. 111120, 2007.
19 K. Davies, Ionospheric Radio, Peter Peregrinus, London, UK, 1990.
20 L. F. McNamara, The Ionosphere: Communications, Surveillance, and Direction Finding, Krieger, Malabar,
Fla, USA, 1991.
21 V. W. J. H. Kirchho, Introducao a Geofsica Espacial, EDUSP, Sao Paulo, Brazil, 1991.
22 D. Wells, N. Beck, D. Delikaraoglou, et al., Guide to GPS Positionin, Canadian GPS Associates,
Frederiction, Canada, 1986.
23 D. Odijk, Fast Precise GPS Positioning in the Presence of Ionospheric Delays, Netherlands Geodetic
Commission, Delft, The Netherlands, 2002.
24 I. Webster, A regional model for prediction of ionospheric delay for single frequency users of the global
positioning system, M.S. thesis, University of New Brunswick, New Brunswick, Canada, 1993.
25 X. X. Jin, Theory of carrier adjusted DGPS positioning approach and some experimental results, Ph.D. thesis,
Delft University of Technology, Delft, The Netherlands, 1996.
26 J. A. Klobuchar, Ionospheric eects on GPS, in Global Positioning System: Theory and Applications,
vol. 1, chapter 12, pp. 485515, American Institute of Aeronautics and Astronautics, Washington, DC,
USA, 1996.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 856138, 9 pages
doi:10.1155/2009/856138

Research Article
The Impact on Geographic Location Accuracy
due to Different Satellite Orbit Ephemerides

Claudia C. Celestino,1 Cristina T. Sousa,2 Wilson Yamaguti,1

and Helio Koiti Kuga3
1
Space Systems Division (DSE), Instituto Nacional de Pesquisas Espaciais (INPE),
12227-010 Sao Jose dos Campos, SP, Brazil
2
Mathematics Department (DMA), University of Sao Paulo State (UNESP-FEG),
12516-410 Guaratingueta, SP, Brazil
3
Space Mechanics and Control Division (DMC), Instituto Nacional de Pesquisas Espaciais (INPE),
12227-010 Sao Jose dos Campos, SP, Brazil

Correspondence should be addressed to Helio Koiti Kuga, [email protected]

Received 30 July 2009; Accepted 29 October 2009

Recommended by Tadashi Yokoyama

The current Brazilian System of Environmental Data Collection is composed of several satellites
SCD-1 and 2, CBERS-2 and 2B, Data Collection Platforms DCPs spread mostly over the
Brazilian territory, and ground reception stations located in Cuiaba and Alcantara. An essential
functionality oered to the users is the geographic location of these DCPs. The location is
computed by the in-house developed GEOLOC program which processes the onboard measured
Doppler shifts suered by the signal transmitted by the DCPs. These data are relayed and stored
on ground when the satellite passes over the receiving stations. Another important input data to
GEOLOC are the orbit ephemeris of the satellite corresponding to the Doppler data. In this work,
the impact on the geographic location accuracy when using orbit ephemeris which can be obtained
through several sources is assessed. First, this evaluation is performed by computer simulation of
the Doppler data, corresponding to real existing satellite passes. Then real Doppler data are used to
assess the performance of the location system. The results indicate that the use of precise ephemeris
can improve the performance of GEOLOC by reducing the location errors, and such conclusion can
then be extended to similar location systems.

Copyright q 2009 Claudia C. Celestino et al. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
The Brazilian System of Environmental Data Collection SBCDA can be organized in three
subsystems: the space subsystem, the data collection ground subsystem, and the tracking
and control ground segment 1, 2. The space subsystem is composed of SCD-1, SCD-2,
CBERS-2, and CBERS-2B satellites. The ground subsystem of data collection is composed of
2 Mathematical Problems in Engineering

Figure 1: Typical meteorological data collection platform.

hundreds of Data Collection Platforms DCPs that are deployed on ground, xed or mobile
see Figure 1. The INPEs tracking and control ground segment is composed of the Satellite
Control Center and Ground Stations in Cuiaba and Alcantara.
In this system, the satellite works as a message retransmitter, that is, providing a
communication link between data collection platform DCP and a reception ground station.
One of the functionalities oered by the system is the Geographic Location of the DCPs. The
location is computed by the in-house developed GEOLOC program which processes the
Doppler shifts suered by the signal transmitted by the DCPs, together with a statistical least-
squares method 3, 4. The DCPs transmit data signals to the satellites in the UHF frequency
band. Aboard the satellite, the DCP is identied and the payload data and the received
frequency data are relayed to the tracking reception station. Then, the Geographical Location
software GEOLOC developed by INPE is fed with Doppler shift data and the corresponding
satellite orbit ephemeris. In general the most common format of orbit ephemeris exchange
is the Two Line Elements TLE set 5, 6. The TLEs can be obtained through the Satellite
Control Center at INPE or through Internet, for example, 5. The TLE format is composed
of seven parameters and time and is basically dened by orbital mean motion, eccentricity,
inclination, right ascension of the ascending node, perigee argument, mean anomaly, and a
modied ballistic coecient 5, 6. Figure 2 shows a representation of Two Line Elements
format.
To evaluate the impacts on geographic localization errors generated by the GEOLOC
software, satellite orbit ephemerides from several sources were used such as: accurate PVT
Position, Velocity, Time, TLEs from INPEs Control Center, and TLEs from Internet. Then
actual passes of the satellite were used to compare the errors considering such dierent orbital
data sources for the location software. To foresee the magnitude of the expected errors, a
comparison with ideal simulated case is also performed as detailed in the next section.
Mathematical Problems in Engineering 3

Drag term or
1st derivative of mean radiation pressure
Name of satellite motion or ballistic coecient coecient
Element number
11 characters & check sum
International Epoch year & 2nd derivative of mean Ephemeris
designator day fraction motion, usually blank type

NOAA 6
1 11416U 84123 A 86 50.28438588 0.00000140 00000-0 67960-4 0 5293
2 11416 98.5105 69.3305 0012788 63.2828 296.9658 14.24899292 346978

Satellite Inclination Eccentricity Mean anomaly

number Argument Mean motion
Right ascension
of the ascending of perigee Revolution number
node at epoch & check sum

Figure 2: Presentation of the two line elements. Source: 5

It is expected that conclusions arising from this work can be extended to similar
systems using Doppler data and articial satellites for location purposes.

2. Data and Work Outline

In order to get a benchmark to allow the analysis of location errors due to orbit ephemeris,
the work followed the following steps.
i A survey on the satellite passes samples and the feasibility of obtaining the
accurate orbit ephemeris PVT format were made to select the test period.
ii Three reference DCPs no. 113, no. 32590, and no. 109 were selected whose
locations are: DCP113 12.0960 S, 77.0400 W, DCP32590 15.5550 S, 56.0698 W, and
DCP109 5.1860 N, 52.6870 W. Such locations are accurate to the GPS level, that is, 10 to
30 m. The nominal frequency is 401.650 MHz UHF. The Doppler shift data corresponding
to the passes were obtained by the reception stations of Cuiaba and/or Alcantara, consistent
with the location of the DCPs. The test period was from November, 21st to 27th, 2008.
iii The Doppler shift data of SCD-2 satellite passes were considered. The orbit
inclination of this satellite is 25 , that is, a near equatorial orbit. The orbit ephemerides in
Two Line Element TLE format were obtained from both the Satellite Control Center CCS
at INPE 79 and Internet at 5.
iv The accurate orbit ephemerides PVT format were provided by the INPE CCS
Orbit Determination System 7 with steps of one minute for each considered day. The
GEOLOC location software was executed with this accurate ephemerides. Then, the accuracy
of the SCD2 satellite orbit ephemerides was veried.
v To foresee the magnitude of the error when using TLEs instead of accurate PVTs,
another program simulated the errors for the ideal case.
vi The GEOLOC location software was run using the PVT, TLEs from CCS, and TLEs
from Internet, covering 4 SCD2 satellite actual passes. For each satellite pass, considering
the established period above, the reference DCP locations are done using the Doppler shift
measurements and the ephemerides obtained from two dierent sources TLEs from both
CCS and Internet. The location errors are compared with the ones obtained using accurate
PVT state vector, considered as reference.
4 Mathematical Problems in Engineering

Instant T1

P2
Instant T2

P1

Figure 3: Location cones.

3. The System for Geographic Location of DCPs

During the pass of a satellite, the signals transmitted from the DCPs are immediately relayed,
through the satellite, to the reception stations. In the reception station, the signals are received
and the Doppler shifted frequencies, computed on board, are collected. The dierence
between the frequency of the received signal and the nominal frequency supplies the Doppler
shift. For each Doppler measure it corresponds to a solid cone of location, whose intersection
with the terrestrial sphere represents the possible positions of the transmitter. The intersection
of the two cones in the altitude sphere supplies two possible position solutions, in a single
pass Figure 3.
With the apriori knowledge of an approximate position, it is possible to distinguish
which is the correct one. However, because the Doppler measurement is corrupted by several
error sources, a direct solution is not possible. Therefore, one should use statistical methods
to solve the problem.
To do so, the parameters of position location and the ephemeris of the satellite are
related through expression 3, 4:

   
x X x X y Y y Y z Z z Z
hx    2 b0 b1 t, 3.1
x X 2 y Y z Z 2

where x  x, y, z and X, Y, Z are the position coordinates of the transmitter and the satellite
respectively; x, y, z and X, Y , Z represent the velocity vectors of the transmitter and the
satellite; b0 and b1 are the bias and drift associated with each Doppler curve; t is time interval
from the time of closest approach.
Mathematical Problems in Engineering 5

Orbit

Cuiaba
113

Doppler
shift
Observed
Inection
point

b0
Corrected Time

b0 b1 t

Figure 4: Typical SCD-2 pass and Doppler curve.

Figure 4 shows an example of a typical pass of satellite SCD-2 over DCP 113 and
Cuiaba reception station. In this gure, the blue line represents positive Doppler shifts, and
the red line the negative ones, with respect to DCP 113. The yellow circles are visibility circles
where the contact with satellite is possible. Also a Typical Doppler curve is depicted.

4. TLE-S Accuracy: CCS and Internet

With the aim at studying the impact of ephemeris accuracy of SCD2 satellite on the location
system, some tests were made to compare the accuracy of dierent sources of orbit ephemeris.
Four satellite passes on November 22-23, 2008 were used, and these ephemerides were
compared with TLEs from CCS and Internet.
The CCS carries out precise orbit determination based on ground tracking data of type
ranging and Doppler 7. The orbit determination using ranging is called ranging
solution, and the orbit determination using Doppler data is called Doppler solution. The
dierences between both solutions agree to the level of 15 m 10. The orbit ephemerides are
provided as Inertial True of Date True Of Date PVTs Position, Velocity, Time, equally
spaced at one-minute intervals. To recover the ephemeris to the desired time, the PVTs from
CCS are interpolated using a sixth-degree Lagrange polynomial.
CCS provides also TLEs of SCD2. The TLEs obtained from CCS are generated using an
approximate period of one week data. Analyses have shown that the errors, compared to the
PVTs, are zero mean with standard deviations around 300 m, however, sometimes presenting
peaks of errors of 1000 to 1500 m in along-track transversal components 8, 10, 11. Such
accuracies are consistent with the ones presented elsewhere 12.
6 Mathematical Problems in Engineering

Table 1: Satellite SCD2 passes considered in the comparisons of the orbit ephemeris dierences.

Date Initial time End time

Pass
Nov. 2008 UTC UTC
1 22 17:27:44 17:35:18
2 22 19:13:49 19:25:56
3 23 16:40:25 16:52:32
4 23 18:27:00 18:40:08

The TLEs obtained by Internet are refreshed with a not well-dened periodicity but, in
general, fresh TLEs are available every 2 to 3 days. It is claimed that accuracy 6, 12 enough
for tracking purposes is provided. The orbital ephemerides of many satellites are broadcast by
NASA in the TLE format in the electronic addresses of Internet, for example, 5. To recover
the ephemerides for the desired time, the model SGP4 6 is conventionally used.
All the ephemerides are in Inertial True of Date True Of Date system. The inertial
coordinates are transformed to the ECEF system Earth Centered Earth Fixed of WGS-84
system, taking into account the polar motion and equation of the equinoxes, which are the
most relevant corrections. The SCD2 satellite passes considered in the tests are presented in
Table 1.
From the former considerations, the PVTs from CCS are taken as reference to the
comparisons. Figures 5 a and b show the dierences between the ephemerides using TLE
from Internet NORAD and from CCS, respectively, in terms of the radial R, transversal or
along-track T, and normal N components.
Considering the dierent ephemerides from PVTs and TLEs of CCS, one observes that
all R, N, and T component dierences are inferior to around 450 m. On the other hand, the
dierences between the PVTs and Internet TLEs present values similar for the R and N com-
ponents; however, the transversal component T presents much higher values near 1000 m.
Therefore, for the TLEs from Internet, there is a more pronounced dierence in the
transversal component T. It is very likely that it can cause a biased error on the geographic
location, because of the usage of dierent source of orbit ephemeris.

5. Test Results
In the tests, SCD-2 satellite and three DCPs in dierent locations were used. The SCD-2
satellite is in a quasi-circular low orbit around of Earth, with 25 inclination, and altitude
of 750 km. The 3 DCPs are DCP109, DCP113, and DCP32590.

5.1. Ideal Case

Considering a ctitious ideal case, the Doppler measurements for DPCs 109 and 32590
were simulated computationally using the more precise PVTs provided by the CCS Satellite
Control Center as orbit ephemeris, with SCD2 satellite relayed by the reception station of
Cuiaba. On the other hand, TLEs from Internet and from CCS were used as orbit ephemeris
in the GEOLOC location software, and the results are presented in Table 2. This test gives an
idea of the errors magnitude expected when using ephemeris of dierent accuracies. For days
21 and 22, of November 2008, the Table shows that these errors are mostly near 1 km, which,
as expected, is a rather direct translation of ephemerides dierences as shown in Figure 5.
Mathematical Problems in Engineering 7

TLE NORAD
Time hours
16:30 17:06 17:42 18:18 18:54 19:30
1000

500

Error m
0

500

1000
a

TLE CCS
Time hours
16:30 17:06 17:42 18:18 18:54 19:30
1000

500

Error m
0

500

1000

R
N
T
b

Figure 5: Radial R, Normal N, and Transversal T deviations between CCS PVTs and TLEs from
Internet a and from CCS b.

Table 2: Location errors for a simulated ideal case.

DPC Date Time Internet CCS

no. Nov. 2008 UTC TLE error km TLE error km
109 21 16:29:50 1.07 0.59
109 21 18:14:20 0.67 0.42
109 22 08:31:30 0.99 0.28
109 22 10:19:50 0.83 0.20
109 22 17:27:20 0.88 0.32
109 22 19:13:50 0.85 0.29
109 22 21:02:00 0.61 0.66
32590 21 09:16:40 0.59 0.38
32590 21 11:02:50 0.49 0.27
32590 21 18:09:40 0.94 0.23
32590 22 08:29:50 0.64 0.46
32590 22 10:15:40 1.04 0.32
32590 22 15:36:10 1.25 0.51
32590 22 17:22:30 1.28 0.17
32590 22 19:09:40 0.24 0.42
8 Mathematical Problems in Engineering

DCP109-November, 2008 DCP113-November, 2008

2 2
1 1
Error km

Error km
0 0
22 23 24 25 26 27 28 22 23 24 25 26 27 28
1 1
2 2
3 3
4 4
Time days Time days

a b
DCP32590-November, 2008
2
1
Error km

0
1 22 23 24 25 26 27 28
2
3
4
Time days

Error PVT CCS

Error PVT NORAD
c

Figure 6: Dierence between location errors using the PVT location compared to locations from CCS TLE
and NORAD TLE.

5.2. Real Case

In this case, Doppler data relayed by satellite SCD2 and corresponding to days from
November 2228, 2008, were retrieved from the SBCDA archives for the three DCPs 109, 113,
and 32590. For this actual test case, Doppler data were processed by the GEOLOC software,
and the results are presented in Figure 6. Figure 6 shows the dierence in location errors
using all three sources of orbit ephemeris: PVTs from CCS, TLEs from CCS, and TLEs from
Internet. The gure contains, for all 3 DCPs, the location error dierences between CCS and
PVT, and TLE and PVT. That is, they show the dierences in location when using 3 dierent
sources of orbital ephemeris. Notice that DCP 32590 presents the highest dierences when
using TLEs from Internet NORAD, similar to what occurred in the simulated ideal case.

6. Conclusions
This article presents the impact on the accuracy when dierent sources of orbit ephemeris
with heterogeneous accuracies are used to perform location of ground transmitters using
the Doppler shifted data recorded on board satellites. Three dierent sources were analyzed
i precise PVTs Position, Velocity, Time arising from INPEs Control Center orbit
determination system, ii TLEs computed by INPEs Control Center on weekly basis, and
iii TLEs obtained freely from Internet. For a restricted period of 2-day comparison, analyses
show the dierences in accuracy which could arise between the dierent orbit ephemeris
sources. The comparisons showed that the ephemerides have minor and similar dierences
Mathematical Problems in Engineering 9

in radial and normal components. Nevertheless, extending the comparison to the transverse
along-track component, the dierence ranges from 500 to 1000 meters. Therefore, it is
straightforward to conclude that this is the major contributing error to the location system. A
simulated location session using ideal measurements, but with orbit ephemeris of dierent
accuracy, showed clearly the along track error being transposed to the location error at
similar levels. Actual data conrmed the same behavior. A future work can make use of more
precise orbit ephemerides available via on-board GPS, to conrm the location errors steaming
from ephemeris with dierent accuracies. At the end, the results point that usage of precise
ephemeris can improve the performance the GEOLOC location system, and such conclusion
can then be extended to similar systems.

Acknowledgment
The authors thank CNPq Grant no. 382746/2005-8 and FAPESP Grant no. 2005/04497-0
for the nancial support.

References
1 W. Yamaguti, et al., The Brazilian environmental data collecting system: status, needs, and study
of continuity of the Data Collecting Mission, SCD-ETD-002, Instituto Nacional de Pesquisas
Espaciais, Sao Jose dos Campos, Brazil, 2006.
2 W. Yamaguti, V. Orlando, and S. P. Pereira, Brazilian data collecting system: status and future plans,
in Proceedings of the 14th Remote Sensing Brazilian Symposium (SBSR 09), pp. 16331640, Natal, Brazil,
2009.
3 C. T. de Sousa, Geolocation of transmitters by satellites using Doppler shifts in near real time, Ph.D.
dissertation, Space Engineering and Technology, Space Mechanics and Control Division, INPE
Instituto Nacional de Pesquisas Espaciais, Sao Jose dos Campos, Brazil, 2000.
4 C. T. de Sousa, H. K. Kuga, and A. W. Setzer, Geo-location of transmitters using real data, Doppler
shifts and least squares, Acta Astronautica, vol. 52, no. 912, pp. 915922, 2003.
5 Celestrak, November 2008, http://www.celestrak.com/.
6 F. R. Hoots and R. L. Roehrich, Models for propagation of NORAD element sets, Spacetrack Report
no. 3, Peterson AFB, Colorado Springs, Colo, USA, December 1980.
7 H. K. Kuga, Flight dynamics at INPE, in Nonlinear Dynamics, Chaos, Control and Their Applications to
Engineering Sciences, J. M. Balthazar, D. T. Mook, and J. M. Rosario, Eds., vol. 1, pp. 306311, AAM
American Academy of Mechanics, 1997.
8 R. V. de Moraes, H. K. Kuga, and D. Y. Campos, Orbital propagation for Brazilian satellites using
NORAD models, Advances in Space Research, vol. 30, no. 2, pp. 331335, 2002.
9 H. K. Kuga, On using the two-lines ephemerides from NORAD in the CBERS-1 satellite orbit model,
in Proceedings of the Brazilian Conference on Dynamics, Control and Their Applications (DINCON 02), vol.
1, pp. 925930, Sao Jose do Rio Preto, Brazil, 2002.
10 H. K. Kuga, A. R. Silva, and R. V. F. Lopes, Analysis of accuracy of on-board CBERS-2B satellite GPS
orbital ephemerides, in Proceedings of the 14th Remote Sensing Brazilian Symposium (SBSR 09), pp.
20572064, Natal, Brazil, 2009.
11 H. K. Kuga and V. Orlando, Analysis of on-board orbit ephemeris impact on CBERS-2 image
processing, in Proceedings of the 18th International Symposium on Space Flight Dynamics (ISSFD 04),
pp. 543546, Munich, The Netherlands, October 2004, European Space Agency, Special Publication
ESA SP.
12 R. Wang, J. Liu, and Q. M. Zhang, Propagation errors analysis of TLE data, Advances in Space
Research, vol. 43, no. 7, pp. 10651069, 2009.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 147326, 19 pages
doi:10.1155/2009/147326

Research Article
Simulations under Ideal and Nonideal Conditions
for Characterization of a Passive Doppler
Geographical Location System Using Extension of
Data Reception Network

Cristina Tobler de Sousa,1 Rodolpho Vilhena de Moraes,1

and Helio Koiti Kuga2
1
UNESP-University, Estadual Paulista, Guaratingueta, CEP 12516-410, SP, Brazil
2
INPE-DMC, CP 515, S.J. Campos, CEP 12245-970, Brazil

Correspondence should be addressed to Helio Koiti Kuga, [email protected]

Received 30 July 2009; Accepted 5 November 2009

Recommended by Maria Zanardi

This work presents a Data Reception Network DRN software investigation considering
simulated conditions inserting purposely errors into the Doppler measurements, satellites
ephemeris, and time stamp, to characterize the geographical location software GEOLOC
developed by Sousa 2000 and Sousa et al. 2003. The extension of reception stations in Brazilian
territory can result in more precise locations if the network is considered in the GEOLOC. The
results and analyses were rst obtained considering the ground stations separately, to characterize
their eects in the geographical location GL result. Six conditions were investigated: ideal
simulated conditions, random and bias errors in the Doppler measurements, errors in the satellite
ephemeris, and errors in the time stamp in order to investigate the DRN importance to get
more accurate locations; an analysis was performed considering the random errors of 1 Hz in
the Doppler measurements. The results are quite satisfactory and also show good compatibility
between the simulator and the GEOLOC using the DRN.

Copyright q 2009 Cristina Tobler de Sousa et al. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
This work presents the validation, through software simulations, of the passive Doppler
GEOLOC system using extended DRN. The Doppler measurements data of a single satellite
pass over a Data Collecting Platform DCP, considering a network of ground reception
stations, is the rule of the DRN. The DRN uses an ordering selection method that merges
the collected Doppler shift measurements through the stations network in a single le. The
preprocessed and analyzed measurements encompass the DCP signal transmission time and
2 Mathematical Problems in Engineering

the Doppler shifted signal frequency received on board the satellite. Thus, the assembly to
a single le of the measurements collected, considering a given satellite pass, will contain
more information about the full Doppler eect behavior while decreasing the amount of
measurement losses, as a consequence of an extended visibility between the relay satellite
and the reception stations.
The simulations software was developed to produce simulated Doppler measure-
ments according to several error simulation scenarios. A test scenario composed by
SCD-2 Brazilian Data Collecting satellite and NOAA-17 National Oceanic Atmospheric
Administration satellite passes, a single DCP and ve ground receiving stations were used.
This software is integrated to the developed GEOLOC system that uses the method of near
real-time just after data reception LOCATION of transmitters through satellites 1.
Nowadays there are more than 600 xed and moving DCPs transmitting several
types of payload data meteorological, hydrological, agricultural, deforestation, CO2 gas
concentration, etc. through Low Earth Orbit LEO satellites. In Brazil, near real-time
LOCATION of transmitters and its monitoring through satellites is particularly useful for
monitoring moving DCPs 2 drifting buoys in sea or rivers, to track displacements and
habits of animals by xing minitransmitters on them 3, for checking if the DCP is still in
place, or for insuring that goods reaches the destiny, to monitor emergency location and
rescue of aircraft and ships 4, and others.
The relay satellite measures the Doppler shift suered by the DCP transmitted signal,
which in turn, together with the payload data, downlinks the Doppler measurements to
ground receiving stations. Such Doppler shift measurements are freely available passive,
being further processed to compute the DCP location through the location software.
The results and analysis with table and graphics are represented under six conditions:
1 ideal conditions from simulated Doppler shift measurements;
2 random Gaussian errors in the Doppler measurements;
3 bias errors in the Doppler measurements;
4 errors in the satellite ephemeris;
5 errors in the time stamp;
6 errors of 1 Hz in the Doppler measurements using DRN.
For the Doppler shift measurements simulations we considered two transmission
intervals: either 600 Doppler measurements per pass good statistical condition with lots of
measurements, approximately one measurement every second or 7 Doppler measurements
per pass realistic condition and not so good statistical scenario, approximately one
measurement every 90 seconds.

2. Geographical Location with Doppler Shift Measurements

Basically the satellite receives the UHF signals from DCPs and relays such signals to ground
reception stations in range. The Doppler shift measurements are computed in the ground
station. In the Data Collection Mission Center, the Doppler measurements are sorted and
merged to input them to the geographic location software, which provides the DCPs location.
When the transmitter and the reception stations are inside the satellite visibility circle
of around 5000 km diameter for 5 minimum elevation angle, the nominal UHF frequency
signals periodically sent by the transmitter are received by the satellite and immediately
Mathematical Problems in Engineering 3

DCPs

Cuiaba
station

Figure 1: Brazilian Environmental data collection system and Cuiaba and Alcantara station visibility
circles.

realtime sent down to the reception station. The platforms installed on ground xed or
mobile are congured for transmission intervals of between 40 to 220 seconds. In a typical
condition, in which both transmitter and receiver are close enough, this period can last up
to 10 minutes. The DCP messages retransmited by the satellites and received by the Cuiaba
and Alcantara stations are sent to the Data Collection Mission Center located at Cachoeira
Paulista, for processing, storage, and dissemination to the users, as seen in Figure 1.
The dierence between the received signal frequency and the nominal frequency
supplies the Doppler shift. The basic principle of transmitter location considers that for each
signal transmitted a location cone is obtained Figure 2. The satellite is in the cone vertex
and its velocity vector v lies in the symmetry axis. Two dierent cones of location intercept
the surface and its intersection contains two possible transmitter positions. To nd which
of the two ambiguous positions is the correct one, additional information is required, as
for example, the knowledge of an initial approximate position. A second satellite overpass
removes any uncertainties.
The transmitter geographic location can be determined by means of the Doppler
shift of the transmitted frequency due to the relative velocity between the satellite and the
transmitter. The satellite velocity relative to the transmitter V cos in vacuum conditions,
denoted by , is given by the Doppler eect 5 equation

 
fr ft
 c, 2.1
ft

where fr is the frequency value as received by the satellite; ft is the reference frequency
sent by the transmitter; fr ft  is the Doppler shift due to the relative velocity satellite
transmitter; c is the speed of light; is the angle between the satellite velocity vector V and
the transmitter position relative to the satellite.
The satellite ephemeris generator uses the SGP8 Special General Perturbation model
of North American Aerospace Defense Command NORAD 6, 7, for obtaining the satellite
4 Mathematical Problems in Engineering

Instant T1

P2 Instant T2

P1

Figure 2: Location cones.

orbit at the measured Doppler shift times. The updated ephemeris are used within the
Doppler eect equation to model the observations. Given the observations modeled as

y  hx v, 2.2

where y is the set of Doppler shifts measured data; v is a noise vector assumed zero mean
Gaussian; hx is the nonlinear function relating the measurements to the location parameters
and function of the satellite ephemeris:
     
x X x X y Y y Y z Z z Z
hx    2 b0 b1 t, 2.3
x X2 y Y z Z2

where x, y, z and X, Y, Z are the satellite and transmitter coordinates position; b0 drift
and b1 drift rate are constants associated with each Doppler curve to account for unknown
bias in the Doppler measurements and a possible drift in the transmitter oscillator; t is the
diference time between the rst Doppler measurement and the current one.
The nonlinear least squares solution 8 is

H1 x  y1 , 2.4

where x  x x, and H1 is a triangular matrix. The method turns out to be iterative as we

take the estimated value x as the new value of the reference x successively until x goes to
zero. The H1 matrix is the result of the Householder orthogonal 9 transformation T such
that

  1/2 
H1 S0
T , 2.5
0 W 1/2 H

where H is the partial derivatives matrix h/xxx of the observations relative to the state
parameters latitude, longitude, height, bias, drift, drift rate around the reference values;
Mathematical Problems in Engineering 5

STA 1
2 N, 61 W
Panama
Venezuela Guyana STA 5
Colombia
Guyana French Guiana 2.3 S, 45 W
Suriname
STA 2
Ecuador 5 S, 35 W

Peru DCP
12 S, 50 W
Bolivia Brazil

STA 4
15 S, 57 W
Paraguay

STA 3
30 S, 54 W
Uruguay

Figure 3: Reception stations conguring ideal geometry.

W 1/2 is the square root of the measurements weight matrix, and S1/2
0 is the square root of the
information matrix.
The y1 is such that

  1/2 
y1 S0 x0
T , 2.6
y2 W 1/2 y

where y is the residuals vector. The nal cost function is

3 32 3 32
J  3y1 H1 x3 3y2 3 2.7

with y2 2  Jmin , where Jmin is the minimum cost. The whole detailed procedure is fully
described in 10

3. Results
In this section we show the results of the simulated Doppler shift measurements for ideal
conditions, inserting random errors in the Doppler measurements, bias errors in the Doppler
measurements, errors in the satellite ephemeris, errors in the time stamp and realistic random
errors in the Doppler measurements using DRN.
For this analysis we initially considered the results using ideal conditions from
simulated Doppler shift measurements of a transmitter located in the center of Brazil at
latitude 12.1200 S and longitude 310.1100 W and ve simulated ground reception stations
conguring the ideal geometry as illustrated in Figure 3.
6 Mathematical Problems in Engineering

Table 1: Ideal conditions.

Number of Mean location error km

Satellite
samples per pass STA 1 STA 2 STA 3 STA 4 STA 5

SCD-2 600 7.E-5 3.E-5 4.E-5 1.E-5 6.E-5 2.E-5 4.E-5 2.E-5 6.E-5 2.E-5
7 3.E-4 1.E-4 7.E-4 4.E-4 3.E-4 1.E-4 5.E-4 7.E-4 6.E-4 4.E-4

NOAA-17 600 6.E-5 2.E-5 1.E-5 4.E-5 1.E-5 3.E-5 3.E-5 3.E-5 3.E-5 1.E-5
7 2.E-4 1.E-4 2.E-4 1.E-4 3.E-4 1.E-4 1.E-4 7.E-4 2.E-4 8.E-4

The data were gathered from March 10 to 19, 2008. The SCD-2 ephemerides were
provided by the Control Center of INPE and the NOAA-17 ephemerides were obtained from
Internet at http://www.celestrak.com/.

3.1. Simulated Doppler Shift Measurements in Ideal Conditions

Table 1 shows results under ideal conditions, without any errors ideal.
In the third column mean location error we have the geographical location mean
errors from the transmitter nominal position. They are listed for the 5 reception stations
STA1, STA2, STA3, STA4, STA5 of Figure 3, in terms of mean and standard deviations. From
Table 1, we can observe that the mean location errors and their deviation standards for both
SCD-2 and NOAA-17 satellites, in the case of 600 samples Doppler measurements per pass,
are around 105 km. This is an evidence that the developed GEOLOC software Sousa, 2000
using data that simulates ideal conditions provides precise results.
The les corresponding to high sampling rate 600 samples, which represent a
transmission rate of one burst per second yield results one order of magnitude more precise
than the les resulting that from low sampling rate 7 samples or around one transmission
burst each 90 s. This emphasizes that a large number of measurements imply a better
statistical result as expected.

3.2. Random Errors in the Simulated Doppler Measurements

In this section we present the analyses and results of simulated les under nonideal
conditions inserting purposely errors in the Doppler shifts.
With the aim of verifying the eect of inserting Doppler measurements errors, the
estimator of bias was turned o. Thus, the GEOLOC assumes that there is no bias in the
Doppler measurements.
The random Gaussian errors zero mean with 1 Hz, 10 Hz, and 100 Hz standard
deviations were inserted in the Doppler measurements. The corresponding results are
presented in Tables 2a and 2b as follows.
From Tables 2a and 2b we can observe that as the random errors inserted in the
Doppler measurements increase, the corresponding mean location errors increase.
We can also verify in the mean errors column that the resulting errors for STA 4 were
smaller than for the other stations. This is a consequence of a higher number of measurements
obtained by this station per each satellite pass and occurs because the fourth reception station
is nearby the transmitter, and so they are almost inside the same satellite visibility circle of
around 5000 km diameter for 5 minimum elevation angle.
Mathematical Problems in Engineering 7

Table 2

a Random errors in the doppler measurements for satellite SCD-2.

Number of Random errors in the Mean location error for

Satellite samples per Locations processed
simulated Doppler reception stations STA
sum of the 5 stations
pass measurements Hz 1 to STA 5 km
0.06 0.02
0.03 0.01
600 1 394 0.03 0.05
0.02 0.04
0.04 0.06

0.36 0.27
0.44 0.22
7 1 408 0.39 0.10
0.31 0.14
0.33 0.22

0.16 0.12
0.14 0.11
600 10 385 0.23 0.05
SCD-2 0.13 0.04
0.34 0.06

2.54 1.51
3.39 2.87
7 10 386 2.43 1.84
2.33 1.78
2.91 1.82

1.68 0.98
1.43 0.03
600 100 385 2.53 1.81
1.12 0.29
2.62 0.48

34.15 23.39
30.74 22.31
7 100 408 24.59 11.46
23.67 12.97
27.37 17.40
8 Mathematical Problems in Engineering
b Random errors in the Doppler measurements for Satellite NOAA-17.

Number of Random errors in the Mean Location error

Satellite Locations processed
samples per simulated Doppler for reception stations
sum of the 5 stations
pass measurements Hz STA 1 to STA 5 km
0.04 0.09
0.06 0.06
NOAA-17 600 1 173 0.06 0.06
0.09 0.08
0.08 0.01
0.31 0.24
0.30 0.17
7 1 157 0.47 0.14
0.29 0.18
0.30 0.15

0.19 0.01
0.18 0.02
600 10 173 0.10 0.02
0.12 0.02
0.28 0.14

2.87 1.00
3.87 2.06
NOAA-17 7 10 154 3.58 1.60
2.00 1.11
3.56 1.18

1.44 0.57
1.69 0.81
600 100 173 1.69 0.31
1.24 0.62
2.71 0.71

34.71 24.55
35.97 25.32
7 100 153 37.86 20.07
21.80 11.69
24.25 15.20

Also when we consider the low sampling 7 measurements compared to the high
sampling 600 measurements cases, for the same reason the location errors were about one
order of magnitude higher. Thus, greater measurements quantity during one satellite pass
will provide a better location result.
The conclusion is as expected: to improve the quality of the presented geographical
location results, considering the reception stations independently, it would be necessary to
install more receptions stations around Brazil territory, nearby the transmitters, to get full
Doppler curve reconstitutions without loss of Doppler shift measurements.
Mathematical Problems in Engineering 9

3.3. Bias (Systematic) Errors in the Simulated Doppler Measurements

This section presents the results of purposely inserted bias errors eect in the simulated
Doppler measurements. For the simulation we used the SCD-2 and NOAA-17 satellites
and bias errors of 1 Hz, 10 Hz, and 100 Hz. The results are presented in Table 3 as
follows.
We can observe in Tables 3a and 3b that, due to the inserted errors in the simulated
Doppler measurements being systematic errors and not random errors nature, the results
considering high and low sampling are in the same order of magnitude. Thus we conclude
that the location errors resulting from bias errors are inuenced in a lesser extent by the rate
of the sampling.
Also, considering the dierences among the random and biased simulations results,
we can conclude from to Tables 2 and 3 that the random errors are ltered out. They
produce mean locations results that tend to zero, mainly for high sampling rate 600 measures
per pass. The location errors maintain some consistency and proportionality; thus, high
sampling random errors of 1, 10, and 100 Hz result in location errors of 0.02, 0.2, and 2 km,
respectively. For low sampling rate 7 measures per pass the obtained results are also
consistent: random errors of 1, 10, and 100 Hz produce location errors of 0.2, 2, and 20 km,
respectively.
These results obtained were expected, due to the errors intrinsic characteristics. We can
conclude that the smaller the measured Doppler error, the smaller becomes the location error,
especially for high sampling rate. The inclusion of biased errors produces more degradation
in the location results than that for random errors.
The GEOLOC location process with tendency bias in the Doppler measurements was
calculated without estimating the drift in the measurements; however, these tendencies may
be removed if the bias estimator is turned on in the location algorithm.

3.4. Errors in the Satellite Ephemerides

In this section we present the transmitters location error results considering inserted errors
of 10 km in the satellites ephemerides. This analysis was performed to verify how the
accuracy in the satellites ephemerides impacts the location accuracy. We also included
errors of observation Doppler measurements of 10 Hz random and/or bias. The results
considering ephemerides errors adding and not adding observation errors are described in
Table 4.
Observing Tables 4a and 4b we verify that for the obtained locations, both
using SCD-2 and NOAA-17, the largest errors appear when we insert all errors types
simultaneously see the last line for each satellite. From the table, we can see that the
largest error contribution is due the errors inserted in the ephemerides. The random errors
are ltered by the least squares algorithm and produce marginal inaccuracy. The systematic
errors bias result in a greater nal inaccuracy, and the added error levels are similar to those
obtained in Table 3 for bias errors.
We can then conclude that the precision in the two-lines elements ephemerides
produces direct impact in the location accuracy, that is, an error of 10 km in the ephemerides
results in location errors of the same order of magnitude. Therefore considering the analyzed
satellites, we can claim that, in low sampling rate conditions, precisions of 1-2 km order can
only be obtained if the ephemerides errors are less than 2 km and the random and biased
errors lower than 10 Hz.
10 Mathematical Problems in Engineering

Table 3

a Bias errors in the simulated Doppler measurements for satellite SCD-2.

Number of Bias errors in the Mean location error for

Satellite samples per Locations processed
simulated Doppler reception stations
sum of the 5 stations
pass measurements Hz STA1 to STA5 km
0.30 0.03
0.27 0.01
600 1 385 0.25 0.05
0.21 0.05
0.22 0.07

0.38 0.09
0.44 0.07
7 1 410 0.30 0.08
0.23 0.02
0.39 0.04

2.99 0.21
2.78 0.52
600 10 385 2.46 0.21
2.10 0.48
SCD-2 2.27 0.70

3.47 0.34
3.34 0.97
7 10 419 2.87 0.46
2.77 0.19
2.90 0.14

21.37 4.88
23.76 6.07
600 100 385 23.82 8.74
20.00 4.80
24.02 8.64

28.72 7.91
30.37 10.05
7 100 420 26.14 9.80
25.48 5.40
26.63 6.83
Mathematical Problems in Engineering 11
b Bias errors in the simulated Doppler measurements for satellite NOAA-17.

Number of Bias errors in the Mean location error for

Satellite Locations processed
samples per simulated Doppler reception stations
sum of the 5 stations
pass measurements Hz STA1 to STA5 km
0.27 0.01
0.24 0.02
NOAA-17 600 1 173 0.31 0.06
0.21 0.05
0.22 0.05
0.32 0.07
0.28 0.07
0.28 0.07
7 1 155 0.24 0.05
0.34 0.05

2.71 0.89
2.41 0.43
600 10 173 2.49 0.39
2.12 0.24
2.19 0.47

3.69 0.59
3.64 0.52
NOAA-17 7 10 161 3.64 0.52
3.22 0.18
3.49 0.91

26.52 2.48
23.56 3.11
600 100 173 27.75 7.81
20.42 2.36
21.994 9.72

28.79 9.93
29.05 5.05
7 100 158 28.23 8.71
21.97 5.72
31.82 10.88

Comparing the presented Tables 2, 3, and 4, we can observe that the mean location
error from the ve simulated reception stations considering a transmission burst every 1s
600 samples per pass is lower than for a burst every 90 s 7 samples per pass. This shows
the importance of having a high rate DCP transmission burst to gather more Doppler data
and a better location result.
Also, the obtained location results using NOAA-17 satellite are better than those using
the SCD-2 satellite. This occurs because we simulate random errors of 0.1 Hz to the NOAA-17
and 1 Hz to the SCD-2 in the Doppler shift measurements, which is more consistent with their
respective on board oscillator accuracies 10.
12 Mathematical Problems in Engineering

Table 4

a Satellite ephemeris and observed error for satellite SCD-2.

Random errors in
Mean location
Number of Errors in the the simulated Locations
error for reception
Satellite samples per satellite Doppler processed sum
stations STA 1 to
pass ephemeris km measurements of the 5 stations
STA 5 km
Hz
8.48 0.14
8.57 0.13
600 10 No 357 8.53 0.11
8.38 0.13
8.67 0.14

8.84 0.18
8.72 0.11
7 10 410 8.83 0.18
8.59 0.11
8.75 0.07

8.79 0.23
8.71 0.26
600 10 10 random 357 8.74 0.18
8.67 0.18
8.68 0.27

9.19 0.35
9.01 0.27
7 10 410 9.64 0.14
SCD-2 9.77 0.12
9.23 0.32

10.53 0.26
10.96 2.42
600 10 10 biased 357 10.60 0.38
10.76 0.24
10.78 0.43

11.41 0.41
11.10 0.32
7 10 412 11.17 0.21
11.58 0.12
11.50 0.31
Mathematical Problems in Engineering 13
a Continued.

Random errors in
Mean location
Number of Errors in the the simulated Locations
Satellite samples per error for reception
satellite Doppler processed sum
stations STA 1 to
pass ephemeris km measurements of the 5 stations
STA 5 km
Hz
10.89 0.32
10.89 0.39
10 random 10.63 0.24
600 10 10 biased 357 10.76 0.26
10.81 0.30

11.30 0.37
11.58 0.33
7 10 410 11.99 0.31
11.73 0.23
11.88 0.34
10.65 0.36
10.62 0.37
600 10 10 biased 174 10.44 0.22
10.79 0.26
10.68 0.33

11.72 0.33
11.82 0.32
7 10 155 11.02 0.28
11.04 0.23
11.84 0.36

SCD-2 10.69 0.48

10.62 0.37
10 random 174 10.45 0.14
600 10 10 biased 10.21 0.12
10.86 0.34

11.97 0.32
11.47 0.28
7 10 148 11.10 0.36
11.05 0.31
11.36 0.39
14 Mathematical Problems in Engineering
b Satellite ephemeris and observed error for satellite NOAA-17.

Random errors in
Mean location
Number of Errors in the the simulated Locations
error for reception
Satellite samples per satellite Doppler processed sum
stations STA 1 to
pass ephemeris km measurements of the 5 stations
STA 5 km
Hz
8.32 0.10
8.22 0.10
600 10 No 174 8.57 0.11
8.12 0.10
8.58 0.12

8.77 0.14
8.87 0.12
NOAA-17 7 10 164 8.70 0.19
8.66 0.11
8.78 0.12

8.62 0.24
8.69 0.27
600 10 10 random 174 8.67 0.15
8.55 0.14
8.71 0.28
9.56 0.21
9.15 0.21
7 10 160 9.11 0.31
9.58 0.10
9.19 0.31

10.65 0.36
10.62 0.37
600 10 10 biased 174 10.44 0.22
10.79 0.26
10.68 0.33

11.72 0.33
11.82 0.32
NOAA-17 7 10 155 11.02 0.28
11.04 0.23
11.84 0.36

10.69 0.48
10.62 0.37
10 random 10.45 0.14
600 10 10 biased 174 10.21 0.12
10.86 0.34

11.97 0.32
11.47 0.28
7 10 148 11.10 0.36
11.05 0.31
11.36 0.39
Mathematical Problems in Engineering 15

The SCD-2 satellite ephemerides used are computed just once a week by the INPE
Control Center of Sao Jose of Campos. In order to improve the quality of the obtained results
it would be necessary that the ephemerides be calculated with a higher precision, that is, it is
recommended a daily computation instead of a week one. The NOAA satellite ephemerides
used are obtained with high precision by Collecte Localisation Satellites CLS/Argos using
their own method and not supplied to the public. The two-line format NOAAs ephemerides
supplied by Internet site, like SCD-2s, are also imprecise, according to 1.
From the tables results, we conclude that it is fundamental to minimize the errors
in the satellite ephemerides, because the location algorithm cannot compensate for them.
The errors inserted in the satellites ephemeris are approximately similar in magnitude to the
resulting errors in the transmitters location.

3.5. Errors in the Time Stamp

In this section we present the results and analysis of inserted errors of 1 and 0.1 seconds
ahead in each transmitted signal frequency time stamp. This analysis was accomplished to
verify how the accuracy in time impacts the location accuracy. The results are described in
Table 5.
Observing Table 5 we verify that the mean location error values increase in the same
order the errors in the time stamp increase. We can also conclude that a time delay of up to
1 s regarding the uplink signal transmitted to the satellite can cause an inaccuracy of up to
10 km in the location.
Thus, to avoid losing the quality on the transmitters geographic location results
the measurements time tagging should be very stable; otherwise we will have one more
unexpected error in the nal location result.

4. Simulation Using the (Data Reception Network) DRN

In Section 3 we showed the analyses of the location results using collected measurements of
several reception stations during a satellite pass and the impact of inserting errors on the
transmitters geographical locations. Now the main goal of this present section is to show the
location results when merging the Doppler measurements of all reception stations during the
satellite pass in the location algorithm. We notice that using this approach, the nal location
results enhance considerably.
In this section the simulations consider a Data Reception Network DRN composed of
ve reception stations according to Figure 3. We insert random errors of 1 Hz in the Doppler
shift measurements and consider one transmission each 90 s for a period of 10 minutes of the
SCD-2 and the NOAA-17 satellite passes.
Tables 6 and 7 present the location errors considering ve reception stations separately
and as a network DRN during March 10 and 11, 2008.
In Tables 6 and 7, the rst and second columns of the tables present the day and hour
of each satellite pass. The next six columns present the location errors inserting random error
of 1Hz in the simulated Doppler measurements, over sign / the respective measurements
amount obtained during the satellite pass.
The last column shows the location errors using the DRN network. In the cells without
results, the symbol means that there were no data to compute the location. As shown in
the last column, when we use the DRN to merge data from the ve reception stations, we can
obtain results that were not present before, as well as improved results.
16 Mathematical Problems in Engineering

Table 5: Errors inserted in the time stamp.

Number of Mean location error for

Locations processed
Satellite samples per Error in the time seg reception stations STA
sum of the 5 stations
pass 1 to STA 5 km
1.54 0.13
1.16 0.30
600 1 385 1.02 0.46
0.71 0.23
0.90 0.23

9.35 0.16
8.96 0.41
7 1 411 8.88 0.36
SCD-2 7.48 0.17
8.28 0.18

0.18 0.39
0.11 0.19
600 0.1 385 0.10 0.13
0.08 0.05
0.09 0.09

1.03 0.14
1.05 0.50
7 0.1 411 1.11 0.02
0.89 0.09
0.97 0.29
1.29 0.10
1.68 0.20
600 1 173 1.68 0.30
2.13 0.13
2.33 0.23

11.81 0.17
8.87 0.14
7 1 166 7.94 0.19
NOAA-17 9.51 0.13
9.51 0.13

0.47 0.19
0.62 0.21
600 0.1 173 0.62 0.31
0.64 0.16
0.22 0.24

1.26 1.28
0.85 0.17
7 0.1 170 1.09 0.29
0.65 0.17
0.65 0.27
Mathematical Problems in Engineering 17

Table 6: Location error considering the DRN and SCD-2.

Location error km with simulated

Doppler measurements with 1 Hz
March 2008 random error and transmission /Number of samples per pass:
burst of 90 s, using the following
reception stations
Day Hour STA 1 STA 2 STA 3 STA 4 STA 5 All 5 stations
10 10 0.28/6 0.31/7 0.16/7 0.39 /9 0.46/9 0.14/38
10 12 0.12/6 0.08/6 0.16/9 0.12/9 0.11/8 0.05/38
10 14 0.36/3 0.41/4 0.15/8 0.30/8 0.22/6 0.05/29
10 16 0.61/5 0.29/8 0.16/7 0.17/6 0.13/26
10 17 0.31/5 0.51/5 0.18/7 0.13/8 0.11/25
10 19 0.30/8 0.31/7 0.25/6 0.14/8 0.25/8 0.13/37
10 21 0.15/7 0.03/5 0.47/6 0.38/7 0.02/25
11 08 0.07/6 0.21/7 0.09/6 0.14/7 0.07/26
11 09 0.26/7 0.08/7 0.35/7 0.17/9 0.35/9 0.08/39
11 11 0.08/5 0.07/6 0.06/8 0.19/8 0.18/8 0.03/35
11 13 1.36/4 0.95/4 0.13/8 0.15/8 0.19/6 0.12/30
11 15 0.19/5 0.06/8 0.20/8 0.09/6 0.04/27
11 17 0.61/4 0.35/6 0.06/8 0.16/8 0.32/7 0.03/33
11 18 1.79/8 0.13/6 0.99/7 1.55/9 0.69/7 0.03/37
11 20 0.17/8 0.17/6 0.61/3 0.30/7 0.14/7 0.12/31

Table 7: Location error considering the DRN and NOAA-17.

Location error km with simulated

Doppler measurements with 1 Hz
March 2008 random error and transmission /Number of samples per pass:
burst of 90 s, using the following
reception stations
Day Hour STA 1 STA 2 STA 3 STA 4 STA 5 All 5 stations
10 00 0.09/6 0.05/9 0.06/6 0.05/9 0.11/8 0.02/38
10 02 0.14/4 0.10/5 0.38/5 /1 0.07/15
10 12 / 0.08/7 0.26/4 0.14/6 0.08/7 0.06/24
10 13 0.17/7 0.20/5 0.25/6 0.43/8 0.12/6 0.07/32
11 00 0.13/4 0.50/7 0.48/4 0.19/7 0.41/7 0.28/29
11 02 0.52/6 0.66/5 0.22/7 0.55/5 0.03/23
11 11 0.49/5 0.57/3 0.64/4 0.34/4 0.13/16
11 13 0.11/7 0.17/7 0.15/6 0.26/8 0.10/8 0.05/36

In most of the results the locations obtained for satellite passes with a larger number
of measurements are better than those for smaller number of measurements.
As we see in the last column, the quality of the obtained results was improved when
considering measurements collected by all reception stations compared to one single station
alone. Therefore, there is an enhancement in the location results when generated using the
DRN. For example we notice in Table 7 that Stations 2 and 5 STA 2 and STA 5 on day 10
18 Mathematical Problems in Engineering

at 02 oclock have only 0 and 1 data points, respectively, insucient for a location. When we
merge the whole data set all other stations, the Doppler curve is fully recovered and the
location error drops improves drastically to 0.07 km.

5. Conclusions
We conclude that the developed GEOLOC is operating appropriately as we can see by the
obtained results under ideal conditions, say, without errors in the Doppler measurements,
ephemeris, and time.
Considering the random and biased simulations results, we conclude that the random
errors are ltered out by the least squares algorithm. They produce mean locations results
that tend to zero error, mainly for high sampling rate 600 measures per pass. The
simulation results considering biased errors yield errors that equally degrade the location
for both sampling rates 600 and 7. The inclusion of biased errors degrades the location
results more than the random errors. From the tables with results inserting errors in the
satellite ephemerides, we can conclude that it is fundamental to minimize such errors,
because the location system cannot compensate for them. The satellites ephemeris errors are
approximately similar in magnitude to their resulting transmitters location errors.
The simulation results using the DRN showed that to improve the location results
quality it would be necessary to have more Reception Stations than the existing Cuiaba,
Cachoeira Paulista, and Alcantara, spread over the Brazilian territory, to increase the data
amount. Then, on the other hand, it improves the geometrical coverage between satellite and
DCPs, and recovers better the full Doppler curves, yielding as a consequence more valid and
improved locations.

Acknowledgment
The authors thank FAPESP Process no. 2005/04497-0 for the nancial support.

References
1 C. T. Sousa, H. K. Kuga, and A. W. Setzer, Geo-Location of transmitters using real data, Doppler
shifts and Least Squares, Acta Astronautica, vol. 52, no. 9, pp. 915922, 2003.
2 M. Kampel and M. R. Stevenson, Heat transport estimates in the surface layer of the Antarctic polar
front using a satellite tracked drifterrst results, in Proceedings of the 5th International Congress of the
Brazilian Geophysical Society, Sao Paulo, Brazil, September 1997.
3 C. M. M. Muelbert, et al., Movimentos sazonais de elefantes marinhos do sul da ilha elefante,
Shetland do sul, antartica, observacoes atraves de telemetria de satelites, in Proceedings of the 7th
Seminars on Antartica Research, p. 38, USP/IG, Sao Paulo, Brazil, 2000, program and abstracts.
4 Techno-Sciences, COSPAR/SARSAT, October 2000, http://www.technosci.com/.
5 H. Resnick, Introduction to Special Relativity, John Wiley & Sons, New York, NY, USA, 1968.
6 F. R. Hoots and R. L. Roehrich, Models for propagation of NORAD element sets, Spacetrack Report
3, Aerospace Defense Command, Peterson Air Force Base, Colorado Springs, Colo, USA, 1980.
7 R. Vilhena de Moraes, A. Silva, and H. K. Kuga, Low cost orbit determination using GPS, in
Proceedings of the 5th International Conference on Mathematical Problems in Engineering and Aerospace
Sciences, S. Sivasundaram, Ed., pp. 495502, Cambridge Scientic, Cambridge, UK, 2006.
8 G. J. Bierman, Factorization Methods for Discrete Sequential Estimation, Academic Press, New York, NY,
USA, 1977.
Mathematical Problems in Engineering 19

9 C. L. Lawson and R. J. Hanson, Solving Least Squares Problems, Prentice-Hall, Englewood Clis, NJ,
USA, 1974.
10 C. T. Sousa, Geolocalizacao de transmissores com satelites usando desvio doppler em tempo quase real,
PhD dissertation, Space Engineering and Technology, Space Mechanics and Control Divison, INPE-
Instituto Nacional de Pesquisas Espaciais, Sao Jose dos Campos, Brazil, 2000.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 140963, 12 pages
doi:10.1155/2009/140963

Research Article
A Discussion Related to Orbit Determination Using
Nonlinear Sigma Point Kalman Filter

Paula Cristiane Pinto Mesquita Pardal,1 Helio Koiti Kuga,1

and Rodolpho Vilhena de Moraes2
1
INPE (National Institute for Space Research), DMC, Avenue dos Astronautas, 1.758. Jd. Granja,
CEP 12227-010, Sao Jose dos Campos, SP, Brazil
2
UNESP (Univ Estadual Paulista), Guaratingueta, CEP 12516-410, SP, Brazil

Correspondence should be addressed to Paula Cristiane Pinto Mesquita Pardal,

[email protected]

Received 30 July 2009; Accepted 9 November 2009

Recommended by Tadashi Yokoyama

Herein, the purpose is to present a Kalman lter based on the sigma point unscented transfor-
mation development, aiming at real-time satellite orbit determination using GPS measurements.
First, a brief review of the extended Kalman lter will be done. After, the sigma point Kalman lter
will be introduced as well as the basic idea of the unscented transformation, in which this lter is
based. Following, the unscented Kalman lter applied to orbit determination will be explained.
Such explanation encloses formulations about the orbit determination through GPS; the dynamic
model; the observation model; the unmodeled acceleration estimation; also an application of this
new lter approaches on orbit determination using GPS measurements discussion.

Copyright q 2009 Paula Cristiane Pinto Mesquita Pardal et al. This is an open access article
distributed under the Creative Commons Attribution License, which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction
In orbit determination of articial satellites, the dynamic system and the measurements
equations are of nonlinear nature. It is a nonlinear problem in which the disturbing forces
are not easily modeled. The problem consists of estimating variables that completely specify
the body trajectory in the space, processing a set of information pseudorange measurements
related to this body. A tracking network on Earth or through sensors, like the GPS receiver
onboard Topex/Poseidon T/P satellite, can collect such observations.
The Global Positioning System GPS is a powerful and low cost means to allow
computation of orbits for articial Earth satellites by means of redundant measurements. The
T/P is an example of using GPS for space positioning.
2 Mathematical Problems in Engineering

The Extended Kalman Filter EKF implementation in orbit estimation, under

inaccurate initial conditions and scattered measurements, can lead to unstable or diverging
solutions. For solving the problem of nonlinear nature, convenient extensions of the Kalman
lter have been sought. In particular, the unscented transformation was developed as a
method to propagate mean and covariance information through nonlinear transformations.
The Sigma Point Kalman Filter SPKF appears as an emerging estimation algorithm applied
to nonlinear system, without needing linearization steps.

2. Extended Kalman Filter

The real-time estimators are a class of estimators which fulll the real time requirements.
They are recursive algorithms and produce sequentially the state to be estimated. Among
them, the Kalman lter and its variations are outstanding.
The Kalman lter is the recursive estimator most used nowadays because it is easy to
implement and to use on digital computers. Its recursiveness leads to lesser memory storage,
which makes it ideal for real-time applications. The EKF is a nonlinear version of the Kalman
lter that generates reference trajectories which are updated at each measurement processing,
at the corresponding instant. Details can be found in Brown and Hwang 1.
Due to the complexity of modeling the articial satellites orbit dynamics accurately,
the EKF is generally used in works of such nature. The EKF algorithm always brings up to
date reference trajectory around the most current available estimate.
The KF lter consists of phases of time and measurement updates. In the rst, state
and covariance are propagated from one previous instant to a later one, meaning that they
are propagated between discrete instants of the system dynamics model. In the second one,
state and covariance are corrected for the later instant corresponding to the measurement
time, through the observations model. This method has, therefore, recursive nature and it
does not need to store the measurements previously in large matrices.
Exploiting the assumption that all transformations are quasilinear, the EKF simply
linearizes all nonlinear transformations and substitutes the Jacobian matrices for the linear
transformations in the Kalman lter equations. Although the EKF maintains the elegant and
computationally ecient recursive update form of the Kalman Filter, it suers a number of
serious limitations.
The rst limitation is that linearized transformations are only reliable if the error
propagation can be matched approximated by a linear function. If this condition does
not hold, the linearized approximation can be extremely poor. At best, this undermines
the performance of the lter. The second is that linearization can be applied only if
the Jacobian matrix exists. However, this is not always the case. Some systems contain
discontinuities, singularities, and the states themselves are inherently discrete. And the
last is that calculating Jacobian matrices can be a very dicult and error-prone process.
The Jacobian equations frequently produce many pages of dense algebra that must be
converted to code. This introduces numerous opportunities for human coding errors that
may degrade the performance of the nal system in a manner that cannot be easily
identied and debugged. Regardless of whether the obscure code associated with a linearized
transformation is or is not correct, it presents a serious problem for subsequent users who
must validate it for use in any high integrity system.
Summarizing, the Kalman lter can be applied to nonlinear systems if a consistent set
of predicted quantities can be calculated. These quantities are derived by projecting a prior
Mathematical Problems in Engineering 3

estimate through a nonlinear transformation. Linearization, as applied in the EKF, is widely

recognized to be inadequate, but the alternatives incur substantial costs in terms of derivation
and computational complexity. Therefore, there is a strong need for a method that is probably
more accurate than linearization but does not incur the implementation nor computational
costs of other higher-order ltering schemes. The Unscented Transformation was developed
to meet these needs.

3. Sigma Point Kalman Filters

If the dynamics system and the observation model are linear, the conventional Kalman lter
can be used fearlessly. Although, not rarely, the dynamic systems and the measurement
equations are nonlinear, convenient extensions of the Kalman Filter have been sought.
The SPKF is a new estimator that allows similar performance than the Kalman lter for
linear systems and it elegantly extends to nonlinear systems, without the linearization steps.
These lters are a new approach to generalize the Kalman lter for nonlinear process and
observation models. A set of weighted samples, called sigma points, is used for normalizing
mean and covariance of a probability distribution. This technique is claimed to lead to a lter
more accurate and easier to implement than the EKF or a second-order Gaussian lter.
The SPKF approach is described by Van Merwe et al. 2 as follows.

1 A set of weighted samples is deterministically calculated, based on mean and

covariance decomposition of a random variable. One minimum need is that the
rst- and second-order momentums are known.
2 The sigma points are propagated through the real nonlinear function, using only
functional estimation, that is, analytical derivatives are not used to generate a
posteriori set of sigma points.
3 The later statistics are calculated using propagated sigma points functions and
weights. In general, they assume the form of a simple weighted average of the mean
and the covariance.

3.1. Basic Idea: The Unscented Transformation

A recent method to calculate the statistics of a random variable that passes through a
nonlinear transformation is the unscented transform UT. The UT builds on the principle
that it is easier to approximate a probability distribution than it is to approximate an arbitrary
nonlinear function, which is detailed in Julier and Uhlmann 3. The approach is simple:
a set of points sigma points are chosen so that their mean and covariance are x and Pxx ,
according to Julier et al. 4, 5, and Julier and Uhlmann 6. The nonlinear function is applied
to each point, in turn, to yield a cloud of transformed points. The statistics of the transformed
points, mean y, and covariance Pyy predicted, can then be calculated to form an estimate of
the nonlinearly transformed mean and covariance.
There are several fundamental dierences to the particle lters, although this method
bears resemblance it. First, the sigma points are deterministically chosen so that they exhibit
certain specic properties like a given mean and covariance, and are not drawn at random.
Second, sigma points can be weighted in ways that are inconsistent with the distribution
interpretation of sample points in a particle lter.
4 Mathematical Problems in Engineering

The n-dimensional random variable x, where n is the state vector dimension, with x
mean and Pxx covariance, is approximated by 2n 1 weighted points, given by

0  x,
 
i  x n Pxx ,
i
3.1
 
i n  x n Pxx
i

(
in which R,  n Pxx i is the ith row or column of the root square matrix of n Pxx ,
and Wi is the weight associated to the ith point

W0  ,
n 
1
Wi  , i  1, . . . , n, 3.2
2n 
1
Wi n  , i  1, . . . , n.
2n 

The transformation in the prediction step of the EKF occurs as follows.

1 Transform each point through the function to yield the set of transformed sigma
points
 
yi  f i . 3.3

2 The observations mean is given by the weighted average of the transformed points


2n
y Wi yi . 3.4
i0

3 The covariance is the weighted outer product of the transformed points


2n
 
Pyy  Wi yi y yi yT . 3.5
i0

As the algorithm works with a nite number of sigma points, it naturally lends itself to
being used in a black box ltering library. Given a model with dened inputs and outputs,
a standard routine can be used to calculate the predicted quantities as necessary for any given
transformation. The computational cost of the algorithm is the same order of magnitude as
the EKF. The most expensive operations are to calculate the matrix square root and the outer
products which are required to compute the covariance of the projected sigma points.
Mathematical Problems in Engineering 5

Initial covariance and state vector

Calculation of Generating the

weights sigma points

Propagation of sigma points:

mean and covariance of the
transformed sigma points

Prediction
Prediction of state, measurement, and
covariance: to transform the sigma
points according to the dynamics and
the observation models

Up to date/correction:
state and covariance up to date

K k 1

Figure 1: Modied EKF, leading to UKF.

Any set of sigma points that encodes the mean and covariance correctly calculates
the projected mean and covariance correctly to the second order. Therefore, the estimate
implicitly includes the second-order bias correction term of the truncated second-order
lter, but without the need to calculate any derivatives.
The algorithm can be used with discontinuous transformations. Sigma points can
pass over a discontinuity and, thus, can approximate the eect of a discontinuity on the
transformed estimate.

3.2. The Unscented Kalman Filter

Using UT, the Kalman lter processes, as summarized in the following steps.
The steps shown in Figure 1 are detailed next.

1 To predict the new state system xk 1 | k and its associated covariance Pk 1 | k,
taking into account the eects of the process white gaussian noise.
2 To predict the expected observation zk 1 | k and its residual covariance
innovation P k 1 | k, considering the eects of the observation noise.
3 To predict the cross correlation matrix Pxz k 1 | k.
6 Mathematical Problems in Engineering

These steps are put in order in the Kalman lter with the restructuring of dynamics,
state vector, and observations models. First, the state vector is added of the noise vector wk ,
with dimension q 1, in order to obtain a vector of dimension na  n q,

 
xk
x k 
a
. 3.6
wk

The dynamics model is rewritten in function of xa k as

xk 1  fxa k, 3.7

and the UT uses 2na 1 sigma points, generated by

   
xk | k Pk | k Px k | k
x k | k 
a
, P k | k 
a
. 3.8
0q1 Px k | k Qx k | k

The matrices in the principal diagonal of Pa k | k are the variances, and the ones out
of it are the correlations between the state dynamic errors and the Gaussian process noises.
There are several extensions and modications that can be done in this basic method
to consider specic details for one given application. In the next section, it will be presented
a discussion of the orbit determination, in real time, using UKF.

3.3. Comparing EKF and UKF

The conventional nonlinear lters, such as the linearized or the extended Kalman Filter, many
times have a poor performance when applied to nonlinear problems, due to two known
diculties.
The linearization of the dynamic and the measurements models can lead to a highly
instable performance of the lter if the time discretization is not enough small.
The derivation of the Jacobian is not simple in most applications, and usually it makes the
implementation dicult.
The UKF has more advantages, when compared to the EKF, as Lee and Alfriend 7, 8
wrote, in the following aspects.

i It allows more stable and accurate estimates of mean and covariance.
ii It can estimate discontinuous functions.
iii No explicit derivation of the Jacobian and/or Hessian matrix is necessary.
iv It is suitable for parallel processing.

4. Using UKF on Orbit Determination

Before presenting a discussion of the orbit determination, in real time, using UKF, some points
need to be outlined: the orbit determination; the dynamic model; the observations model.
Mathematical Problems in Engineering 7

4.1. The Orbit Determination

The orbit determination will be based on GPS technology, whose working principle is based
on the geometric method. In such method, the observer knows the set of satellites position in
the reference system, obtaining its own position in the same reference frame. Figure 2 presents
the basic parameters used by GPS for user position determination.
Here, R GPSi is the position of the ith GPS satellite in the reference system; i is the
pseudorange, the user satellite position in respect to the ith GPS satellite; ru is the user satellite
position in the reference system.

4.2. The Dynamic Model

In the case of orbit determination via GPS, the ordinary dierential equations that represent
the dynamic model are as follows:

r  v
,

r
  3 a
v  w,

r 4.1
b  d,

d  0 wd

with variables given in the inertial reference frame. In the equations above, r is the vector
containing the position components x, y, z; v  is the vector of velocity components; a 
represents the modeled perturbations; w  is the white noise vector with covariance Q; b is the
user clock bias; d is the user clock drift; wd is the white noise on the drift rate with variance
Qd .

4.3. The Observations Model

The nonlinear equation of the observations model is given by:

zk  hk xk , t k , 4.2

where zk is the vector of m observations; hk xk  is the nonlinear function of state xk , with
dimension m; k is the vector of observation errors with dimension m.

4.4. Estimation of the Unmodeled Accelerations

Some spacecraft missions require precise orbit knowledge to support payload experiments.
Sometimes after launch, ground-based orbit determination solutions do not provide the
level of accuracy expected. After verifying all known dynamic models, there may be a
residual signature in the orbit as result of unmodeled accelerations. This leads to attempt to
estimate anomalous accelerations during the orbit t, if sucient data exist. If successful, the
acceleration estimates can improve the t residuals, and also results in better orbital position
estimates, as can be seen in Soyka and Davis 9.
8 Mathematical Problems in Engineering

GPS

GPS

User satellite


R GPS

GPS

r
u

GPS

Figure 2: The geometric method.

Unmodeled accelerations may have many reasons: truncation of geopotential eld;

limitations of modeling solar pressure, Earth albedo, Earth infrared radiation, drag, and
others. Some of these accelerations can be corrected through the use of higher delity
dynamic and physical modeling, while others require postlaunch calibration.
The use of periodic accelerations, with a period near once per revolution of the satellite
orbit, has been used within precision orbit determination programs to improve the accuracy
of the derived ephemeris.

4.4.1. Anomalous Accelerations Modeling

When dening an anomalistic or periodic acceleration, one must consider three aspects: the
subarc interval, the type of function, and the coordinate frame.

Subarc Interval
The subarc interval is the time of duration or number of revolutions for a given acceleration
to be active. As its name implies, it is usually a subset of the total arc. A reason to break an
arc into a subarc is to allow for better overall ts.
Mathematical Problems in Engineering 9

Type of Function
The underlying mathematical function of an acceleration function is usually a constant, a
sine, or a cosine function.
The constant function is the most basic: a constant force in a specic direction. And
the periodic functions sine or cosine have amplitude, frequency, and phase associated with
them. The periodic functions are written as

 
acc  A sin t A or acc  B cos t B , 4.3

where A and B are amplitudes, is the frequency, t is the time elapsed since the start of the
periodic function reference point or subarc interval, A and B are the phase osets. Either of
these accelerations can be rewritten as:

acc  A sint B cost, 4.4

where for a sine acceleration with phase, A  A cos A ; B  B sin A , and, for a cosine
acceleration with phase, A  B cos B ; B  B sin B . When estimated, the amplitudes A
and B will adjust themselves to produce an eective phase oset.

Coordinate Frame
The selection of the start of the subarc can be important, especially for noncircular orbits.
Conventionally, equator crossings, argument of perigee, mean anomaly, or orbit angle have
been used as reference point.

4.5. Discussion of the Application

Now that the purpose of the discussion is known, it is possible to present a discussion about
the subject.
In order to generate the UKF, it is necessary to rewrite the Kalman lter from UT. First,
the state vector is increased, with the measurements noise vector wk , yielding a vector with
dimension na  n q. The increased versions of the state and the covariance are

   
xk Pk 0
x k 
a
, Pak  . 4.5
wk 0 Qk

This increase can still contain k , the Gaussian process noise, of dimension l 1. The
new covariance matrix would have Rk in the principal diagonal, the covariance of such noise,
and the new state vector dimension would be na  n q l, according to Lee and Alfriend
10.
10 Mathematical Problems in Engineering

Next, the increased set of sigma points is built:

a0,k  xak ,

ai,k  xak na Pak , i  1, . . . , na , 4.6
i

ai,k  xak na Pak , i  na 1, . . . , 2na
i

with  2 na  na , where control the sigma points scattered about the mean xak , and it
is usually chosen small, inthe interval 104 1, as Jwo and Lai 11 wrote; provides an
extra degree of freedom;  na Pak  is the ith row or column of the root square matrix of
i
na Pak .
In the propagation step, the state vector and covariance predicted are calculated based
on the mean and the covariance of the propagated sigma points, transformed from the state
vector and dynamical noises

+a
2n
xk 1  Wi xi,k 1 ,
i0
+a 
2n  T
Pk 1  Wi xi,k 1 xk 1 xi,k 1 xk 1 . 4.7
i0

The prediction of the observation vector and the innovation matrix, P

k 1
, is done the
same way. That means, the observation and the innovation are predicted from mean and
covariance of the transformed sigma points:


2na
m
yk 1  Wi yi,k 1 , 4.8
i0

where yi,k 1 represents the sigma vectors propagated through the nonlinear equation of the
observation model, yielding the transformed sigma points from the state vector and the
dynamics noise, shown before.
In the up to date correction step of measurement, the Kalman gain, K k 1 , is
calculated based on the correlation matrix between the measurement and the observation,
xy
Pk 1 , and the innovation matrix, both predicted


2na   T
c
k 1 
P Wi yi,k 1 yk 1 yi,k 1 yk 1 ,
xy  1 i0
Kk 1  Pk 1 P
k 1 , with 4.9
2
na   T
c
Wi xi,k 1 xk 1 yi,k 1 yk 1 .
xy
Pk 1 
i0
Mathematical Problems in Engineering 11

Finally, the up to date state and covariance are


x k 1  xk 1 K k 1 yk 1 yk 1 ,
4.10
P k 1  Pk 1 K k 1 P
k 1 K k 1 ,
T

where y is the vector eectively measured in the instant k 1.

The process is repeated for the next instant, and the up to date mean from state x k 1 
and covariance will be used to deterministically generate the sigma points of the next instant.

5. Final Comments
The SPKF estimation technique has been investigated in very many dierent applications.
After analyzing the investigations, some comments may be done.
i The main advantages of the SPKF are: easy to implement; computationally strong;
high accuracy.
ii The motivation for applying such technique herein: recent nonlinear state estimate
techniques will be applied to the specic problem of orbit determination using
real data from GPS, instead of simulated data. This can improve the results, when
compared to the EKF, and get better the estimated state variables accuracy.
iii alman lter may be used as the estimation algorithm. However, not rarely,
the dynamic systems and the measurements equations are of nonlinear nature.
For solving such a problem, convenient extensions of the Kalman lter have
been sought. The EKF implementation in orbit estimation, under inaccurate
initial conditions and scattered measurements, can lead to unstable solutions.
Nevertheless, the unscented transformation was developed as a method to
propagate mean and covariance information through nonlinear transformations.
The SPKF appears as an emerging estimation algorithm applied to nonlinear
system, without linearization steps.

Acknowledgment
The authors wish to express their appreciation for the support provided by FAPESP The
State of Sao Paulo Research Foundation, under Contract 07/53256-1.

References
1 R. G. Brown and P. Y. C. Hwang, Introduction to Random Signals and Applied Kalman Filtering, John
Wiley & Sons, New York, NY, USA, 3rd edition, 1985.
2 R. D. Van Merwe, E. A. Wan, and S. I. Julier, Sigma-point kalman lters for nonlinear estimation
and sensor-fusionapplications to integrated navigation, in Collection of Technical PapersAIAA
Guidance, Navigation, and Control Conference, vol. 3, pp. 17351764, Providence, RI, USA, August 2004.
3 S. J. Julier and J. K. Uhlmann, New extension of the Kalman lter to nonlinear systems, in Signal
Processing, Sensor Fusion, and Target Recognition VI, vol. 3068 of Proceedings of SPIE, pp. 182193,
Orlando, Fla, USA, April 1997.
12 Mathematical Problems in Engineering

4 S. J. Julier, J. K. Uhlmann, and H. F. Durrant-Whyte, New approach for ltering nonlinear systems,
in Proceedings of the American Control Conference, vol. 3, pp. 16281632, Seattle, Wash, USA, June 1995.
5 S. Julier, J. Uhlmann, and H. F. Durrant-Whyte, A new method for the nonlinear transformation of
means and covariances in lters and estimators, IEEE Transactions on Automatic Control, vol. 45, no.
3, pp. 477482, 2000.
6 S. J. Julier and J. K. Uhlmann, Unscented ltering and nonlinear estimation, Proceedings of the IEEE,
vol. 92, no. 3, pp. 401422, 2004.
7 D.-J. Lee and K. T. Alfriend, Precise real-time orbit estimation using the unscented Kalman lter,
Advances in the Astronautical Sciences, vol. 114, supplement, pp. 18351854, 2003.
8 D.-J. Lee and K. T. Alfriend, Adaptive sigma point ltering for state and parameter estimation,
in Collection of Technical PapersAIAA/AAS Astrodynamics Specialist Conference, vol. 2, pp. 897916,
Providence, RI, USA, August 2004.
9 M. T. Soyka and M. A. Davis, Estimation of periodic accelerations to improve orbit ephemeris
accuracy, Advances in the Astronautical Sciences, vol. 108, pp. 11231140, 2001.
10 D.-J. Lee and K. T. Alfriend, Sigma point ltering for sequential orbit estimation and prediction,
Journal of Spacecraft and Rockets, vol. 44, no. 2, pp. 388398, 2007.
11 D.-J. Jwo and C.-N. Lai, Unscented Kalman lter with nonlinear dynamic process modeling for GPS
navigation, GPS Solutions, vol. 12, no. 4, pp. 249260, 2008.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 537256, 11 pages
doi:10.1155/2009/537256

Research Article
Orbital Dynamics of a Simple Solar
Photon Thruster

Anna D. Guerman,1 Georgi V. Smirnov,2

and Maria Cecilia Pereira3
1
University of Beira Interior, 6201-001 Covilha, Portugal
2
University of Porto, 4169-007 Porto, Portugal
3
Aeronautical Institute of Technology, 12.228-900 Sao Jose dos Campos, Brazil

Correspondence should be addressed to Anna D. Guerman, [email protected]

Received 29 July 2009; Accepted 13 November 2009

Recommended by Maria Zanardi

We study orbital dynamics of a compound solar sail, namely, a Simple Solar Photon Thruster
and compare its behavior to that of a common version of sailcraft. To perform this analysis,
development of a mathematical model for force created by light reection on all sailcraft elements
is essential. We deduce the equations of sailcrafts motion and compare performance of two
schemes of solar propulsion for two test time-optimal control problems of trajectory transfer.

Copyright q 2009 Anna D. Guerman et al. This is an open access article distributed under the
Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
The use of solar pressure to create propulsion can minimize spacecraft on-board energy
consumption during a mission 1, 2. Modern materials and technologies made this
propulsion scheme feasible, and many projects of solar sails are now under development,
making solar sail dynamics the subject of numerous studies.
So far, the most extensively studied problem is the orbital maneuver of a Flat Solar Sail
FSS, Figure 1. In this case, the control is performed by turning the entire sail surface with
respect to the Sun direction. This changes the radiation pressure and results in evolution of
the vehicle trajectory. Some of the many missions studied are described in 313.
The use of a compound solar sail, or Solar Photon Thruster SPT, was proposed by
Tsander long ago 1, 2, but the study of this spacecraft began quite recently 1418. The SPT
consists of a parabolic surface which concentrates the solar radiation pressure on a system
of smaller mirrors. The control eort in such system is produced by displacement of a small
mirror with respect to the parabolic surface. The sail axis is supposed to be oriented along
2 Mathematical Problems in Engineering

the Sun-sailcraft direction. There exist several versions of compound solar sails. Forward 14
described two types of a compound sail, namely, Simple Solar Photon Thruster SSPT and
Dual Reection Solar Photon Thruster DRSPT. The few existent studies on SPT dynamics
consider the latter scheme.
In order to assess the dynamical characteristics and to compare the SPT performance to
the most studied version, FSS, one should study the application of these propulsion schemes
for orbital transfer and/or maintenance for various missions. To perform this analysis, a
coherent mathematical model for force acting on such a structure due to solar radiation
pressure is essential. There are many results concerning radiation pressure force and torque
models for a sunlit body, and in many studies they are applied successively to develop a force
model for an FSS. However, the usual approach cannot be applied for a Solar Photon Thruster
due to multiple light reections on the SPT elements.
Some attempts to develop a mathematical description for SPT force have been made
before, mostly for a Dual Reection Solar Photon Thruster. In 2 the model for an ideally
reecting DRSPT is described. In 19 this model is extended for nonideal DRSPT. Meanwhile,
both of these models are based on the supposition that all the incoming light ux is reected
consequently on each one of the DRSPT elements and then leaves the system, which is not
true 20. Moreover, the results of 20 show that the existing shadowing and related energy
dissipation diminish signicantly the DRSPT eciency, making dubious the advantages of
this propulsion scheme compared to FSS.
In the present article, we focus on the other version of the compound scheme of solar
propulsion, Simple Solar Photon Thruster Figure 2. We develop a mathematical model for a
solar radiation force acting on SSPT and provide a comparative study of trajectory dynamics
and control for the FSS and SSPT schemes. In our analysis, we use the following assumptions.

i Solar radiation pressure follows inverse-square variation law.

ii The only gravitational eld is that of the Sun, and this eld is central Newtonian.

iii The sails are ideal reectors all photons are perfectly reected.

We derive the equations of motion for the SPT and compare the orbital behavior of FSS and
SPT studying two test problems: Earth-Mars transfer and Earth-Venus transfer.

2. Equations of Motion
To write down the equations of motion for a solar sail spacecraft, we introduce two right-
oriented Cartesian frames with their origin in the center of mass of the spacecraft O as
follows:

i Oxyz is the coordinate frame attached to the spacecraft; the axes Ox, Oy, and Oz
are the central principal axes of the spacecraft.

ii OXYZ is the orbital frame, its axis OZ is directed along the radius vector of the
point O with respect to the center of mass of the Sun, and the axis OY is orthogonal
both to OZ and to the velocity of the point O.

We determine the position of the coordinate frame Oxyz attached to the spacecraft with
respect to the orbital frame using the transition matrix between these frames, aij .
Mathematical Problems in Engineering 3

We use a set of canonical units which implies that the radius of the Earths orbit is
1 AU, and the period of its revolution is 2. The equations of orbital motion can be written in
the form

..
r  r

a s, 2.1
r3

where as is the acceleration due to the radiation pressure.

To complete the equations of motion, we have to calculate also the force produced by
the Sun radiation pressure.

2.1. Flat Solar Sail

The interaction of the solar radiation ow with a at perfectly reecting surface has been
studied earlier 1, 2. By the symmetry of falling and reected ows, the total solar radiation
force is directed along the symmetry axis of the sail and produces no torque with respect
to any point of this axis, including the sails center of mass Figure 1. This force can be
expressed as 1, 2

 2
P  2

n ,
n S 2, 2.2
r

where n is the normal to the sail surface and points to the Sun,
is the unit vector of the
parallel light ow i.e., it opposes the Sun-sailcraft direction, its coordinates in the Oxyz
frame are
 x , y , z   a31 , a32 , a33 , S is the total area of the sail, and  4.563
106 N/m2 is the nominal solar radiation pressure constant at 1 AU. The solar radiation force
projections onto the spacecraft-connected and orbital coordinate frames are

2
Px  Py  0, Pz  2S a sign a33 ,
r 2 33
2.3

PX  2S 2 a13 a233 sign a33 , PY  2S 2 a32 a233 sign a33 , PZ  2S 2 |a33 |a233 ,
r r r

respectively.
If

n lies in the OXZ plane, then the components of the radiation force are

PX  S sin 2|cos |, PY  0, PZ  2S cos2 |cos |, 2.4
r2 r2

where is the angle between the vector

n and the OZ axis.
The above expressions are standard and appear in numerous studies of the propulsion
eort of a Flat Solar Sail.
4 Mathematical Problems in Engineering

Y
x
dS

Z
z 1

Figure 1: Flat Solar Sail FSS.

1 y

z
f  a/2
v F0
d
2

Figure 2: Solar Photon Thruster SPT.

2.2. Solar Photon Thruster

We consider here another system that is shown in Figure 2. It consists of a parabolic collector
and a control mirror director. When reection is ideal and the collector axis is exactly
aligned with the Sun-sailcraft direction, the collector concentrates the sunlight in the center
of the director. In order to minimize the solar radiation torque that causes perturbations of
the sailcraft orientation, the director should be located at the sailcrafts center of mass. This
scheme of solar propulsion seems to be more reliable with respect to small misalignments of
the sail axis than the DRSPT scheme studied in 1517 which uses a collimator.
In the analysis, we assume the control mirror small enough to disregard the inuence
of its shadow. We also suppose that the SPT axis is aligned exactly along the Sun direction.
We consider the parabolic surface described in the reference frame Oxyz by the
equation

 
x2 y2 2a z f  0, x2 y2 R2 , 2.5
Mathematical Problems in Engineering 5

where a is the parameter of the paraboloid, f is the focal distance, and R is the radius of the
sails projection on the plane Oxy. Since f  a/2, equation 2.5 reduces to

1  2
z a x2 y 2 . 2.6
2a

The sunlight is directed along the vector

 0, 0, 1. Suppose that it is reected on the




element of the parabolic surface dS, containing the point  x i y j z k with z satisfying
2.6. The force produced by the falling light is given by



d P 1 
,
n dS, 2.7

where

n is the normal to the sunlit side of the sail surface

x i y j ak
n   , 2.8
x2 y2 a2

and is the intensity of the light ow at the current point of the orbit

 . 2.9
r2

The ray reected from the element dS of the collectors surface has the direction

1 satisfying


2
1 
n
, n. 2.10

Reection of light from the element of surface dS produces the force


d P 2 

1
,
n dS. 2.11

Finally, this ray is reected at the focus on the directors surface with the normal


x , y , z . The force produced by the reected light can be written as


d P 3  d P 2 , 2
d P 4 
,
n dS. 2.12

Here


2 

1 2
1,
2.13

is the direction of the ray reected from the control mirror.

6 Mathematical Problems in Engineering

The reection of the light on the parabolic surface is unique if the normal to the director
does not cross this surface, so the control angle must be greater than half the angular aperture,
that is,

 
R
tan1 . 2.14
zR

Here is the angle between vector

and the sail axis cos  z , and zR  1/2aa2 R2 
is the z-coordinate of the collectors border. Finally we arrive at the restriction

2aR
|tan | . 2.15
a2 R2

Multiple reections on the collector destroy the collectors lm and produce a considerable
disturbing torque, and so have to be avoided. Therefore condition 2.15 has to be satised
during the orbital maneuver.
The elementary force created by interaction of light with parabolic surface and mirror
is

dP  dP 1 dP 2 dP 3 dP 4  dP 1 dP 4
    2.16

,
n dS
2 ,n d
2 ,
n dS.

After integration, we obtain

  
a2 R2
Px  2 2 R x z 1 2 2 ln 1 2
2
,
r R a
  
a2 R2
Py  2 2 R y z 1 2 2 ln 1 2
2
, 2.17
r R a
  
 a2 R2
Pz  2 2 R z 1 2z
2 2 2
ln 1 2 .
r R2 a

If the control mirror moves in the OXZ plane x  sin , y  0, z  cos , then the
components of the light pressure force in the orbital coordinate frame are

  
a2 R2
PX  2 R 1 2 2 ln 1 2
2
sin 2, PY  0,
r R a
    2.18
a2 R2
PZ  2 2 R cos 2 ln 1 2 cos 2 .
2 2
r R a
Mathematical Problems in Engineering 7

If  R/a  1, then it is possible to simplify 2.18. One can use Taylors formula to obtain
the expressions

  
R2 R3
PX  2 R 1 2 o
2
sin 2, PY  0,
r a a3
   2.19
R2 R3
PZ  2 2 R sin 2 cos 2 o
2 2
.
r 2a a3

If 2  R2 /a2 is negligible i.e., the sail is almost plane, we get

PX  S sin 2, PY  0, PZ  2 S sin2 , 2.20
r2 r2

where S  R2 is the eective sail area, that is, the area of the sail projection on the plane Oxy.
Formulas 2.20 are similar to those used in 1113 for a dierent scheme of SPT, so one can
expect qualitative similarity of the results for small , at least for the maneuvers that require
control angles within limits 2.15.

3. In-Plane Orbital Motion

To compare the principal characteristics of SPT and FSS we studied two test time-optimal
control problems of solar sail dynamics, namely, the time-optimal Earth-Mars and Earth-
Venus transfers 3, 21 for both systems. Since our goal is to compare qualitative behavior
of the above systems, we choose the simplest formulation for orbital transfer problem. In
both cases, we assume that the planet orbits are circular and coplanar and that the spacecraft
moves in the ecliptic plane, starting from the Earth-orbit at 1 AU with Earth-orbital velocity.
We nd the control law that guarantees the fastest transfer to the planets orbit.
This model of orbital dynamics results in the following equations of motion in the orbit
plane 3:

w w2 1 uw
r  u,  , u  2 asZ , w  asX . 3.1
r r r r

Here is the polar angle, and u and w are the radial and transversal components of sail
velocity, respectively.
For the FSS, the components of the light pressure acceleration onto the axis of orbital
coordinate frame OXYZ are

 3 
asX  S|cos | sin 2, asZ  2 Scos . 3.2
mr 2 mr 2
8 Mathematical Problems in Engineering

1.5

0.5

0
1.5 1 0.5 0 0.5 1 1.5

Figure 3: Earth-Mars transfer trajectories.

0.64
0.62
0.6
Control rad

0.58
0.56
0.54
0.52
0.5
0.48
0 50 100 150
Time days

Figure 4: Earth-Mars transfer control eort for SSPT for  0.125.

For the SPT the light pressure acceleration is given by

  
a2 R2
asX  R 1 2 2 ln 1 2
2
sin 2,
mr 2 R a
    3.3
a2 R2
asZ  2 2 R cos 2 ln 1 2 cos 2 .
2 2
mr R a

The control angle is limited by condition 2.15. In this case the sail surface has to follow
the Sun direction.

4. Results
The time-optimal problems for Earth-Mars transfer and Earth-Venus transfer are studied
numerically using the interactive software from 22. This optimization software developed
for personal computers running under MS Windows operating systems is based on the
penalty function approach and oers to the user a possibility to eectively solve optimal
control problems. During the interactive problem-solving process, the user can change the
penalty coecients, change the precision inuencing the stopping rule, and choose/change
the optimization algorithms. The system includes various gradient-free algorithms used at
Mathematical Problems in Engineering 9

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1

Figure 5: Earth-Venus transfer trajectories.

0.8
0.9

1
Control rad

1.1
1.2
1.3

1.4
1.5

0 50 100 150
Time days

Figure 6: Earth-Venus transfer control eort for SSPT for  0.125.

the beginning of the optimization, as well as more precise conjugate gradient and Newton
methods applied at the nal stage in order to obtain a precise solution.
The sail parameter is assumed to be S/m  0.0843 for both systems it corresponds
to the acceleration due to solar radiation pressure equal to 1 mm/s2 at the Earth-orbit.
The Earth-Mars transfer trajectories for the FSS and the SPT are shown in Figure 3.
The continuous line corresponds to the SPT trajectory and the dot-dashed line to the FSS
trajectory. We consider the ratio  R/a  0.125. The best possible transfer time for the FSS
M
is TFSS  2.87 166.7 days, and for the SPT it is TSPT
M
 2.71 157.5 days, so SSPT maneuver
is 5.6% faster than that of the FSS one. Figure 4 shows the variation of the SSPT control angle
for the optimal transfer; for FSS the respective control is well known 3.
For the Earth-Mars problem, we also study the inuence of SPT sail ratio . The
increase of results in longer maneuver time TSPT : for  0.25 it is TSPTM
 2.76, and for
 0.5 the maneuver time is TSPT  2.96. For greater values of , the control angle attains
M

the limits described by restriction 2.15 more frequently.

Analyzing the maneuver to Venus orbit for these two sailcraft schemes Figures 5 and
6, we have established that FSS reaches the objective in 181.2 days TFSSV
 3.12, while the
SPT performs this maneuver in 158.6 days TSPT  2.73. In this case, the eciency of SPT is
V

more signicant; SPT reaches Venus orbit 12.5% faster than FSS.
10 Mathematical Problems in Engineering

5. Conclusions
The problems of orbital dynamics and control are studied for two systems of solar propulsion:
a Flat Solar Sail FSS and a Simple Solar Photon Thruster SSPT. We develop a mathematical
model for force acting on SSPT due to solar radiation pressure, taking into account multiple
reections of the light ux on the sailcraft elements. We derive the SSPT equations of motion.
For in-plane motions of an almost at sail with negligible attitude control errors, these
equations are similar to those used in the previous studies of DRSPT.
For these two solar propulsion schemes, FSS and SSPT, we compare the best time
response in two test problems Earth-Mars transfer and Earth-Venus transfer. Our analysis
showed a better performance of SSPT in terms of response time. The result was more
pronounced for Earth-Venus transfer that can be explained by the greater values of the
transversal component of the acceleration developed by SSPT compared to those of FSS.

Acknowledgments
The authors are grateful to Vladimir Bushenkov for fruitful discussions and his help in use of
the optimization software 22. This work was supported by FCT-Portuguese Foundation for
Science and Technology project CODIS-PTDC/CTE-SPA/64123/2006 and project CoDMoS
Portugal-Brazil collaboration program, FCT-CAPES.

References
1 E. H. Polyahova, Space Flight with Solar Sail, Nauka, Moscow, Russia, 1986.
2 C. R. McInnes, Solar Sailing: Technology, Dynamics and Mission Applications, Praxis Series, Springer,
Berlin, Germany, 1999.
3 B. Ya. Sapunkov, V. A. Egorov, and V. V. Sazonov, Optimization of Earth-Mars solar sail spacecraft
trajectories, Cosmic Research, vol. 30, no. 2, pp. 155162, 1992.
4 M. Otten and C. R. McInnes, Near minimum-time trajectories for solar sails, Journal of Guidance,
Control, and Dynamics, vol. 24, no. 3, pp. 632634, 2001.
5 G. Colasurdo and L. Casalino, Optimal control law for interplanetary trajectories with nonideal solar
sail, Journal of Spacecraft and Rockets, vol. 40, no. 2, pp. 260265, 2003.
6 D. N. Sharma and D. J. Scheeres, Solar-system escape trajectories using solar sails, Journal of
Spacecraft and Rockets, vol. 41, no. 4, pp. 684687, 2004.
7 B. Dachwald, Minimum transfer times for nonperfectly reecting solar sailcraft, Journal of Spacecraft
and Rockets, vol. 41, no. 4, pp. 693695, 2004.
8 G. Mengali and A. A. Quarta, Optimal three-dimensional interplanetary rendezvous using nonideal
solar sail, Journal of Guidance, Control, and Dynamics, vol. 28, no. 1, pp. 173177, 2005.
9 B. Dachwald, Optimal solar-sail trajectories for missions to the outer solar system, Journal of
Guidance, Control, and Dynamics, vol. 28, no. 6, pp. 11871193, 2005.
10 B. Dachwald, W. Seboldt, and L. Richter, Multiple rendezvous and sample return missions to near-
Earth objects using solar sailcraft, Acta Astronautica, vol. 59, no. 811, pp. 768776, 2006.
11 M. Macdonald, G. W. Hughes, C. McInnes, A. Lyngvi, P. Falkner, and A. Atzei, GeoSail: an elegant
solar sail demonstration mission, Journal of Spacecraft and Rockets, vol. 44, no. 4, pp. 784796, 2007.
12 B. Wie, Thrust vector control analysis and design for solar-sail spacecraft, Journal of Spacecraft and
Rockets, vol. 44, no. 3, pp. 545557, 2007.
13 J. Bookless and C. McInnes, Control of Lagrange point orbits using solar sail propulsion, Acta
Astronautica, vol. 62, no. 2-3, pp. 159176, 2008.
14 R. L. Forward, Solar photon thrustor, Journal of Spacecraft and Rockets, vol. 27, no. 4, pp. 411416,
1990.
15 C. R. McInnes, Payload mass fractions for minimum-time trajectories of at and compound solar
sails, Journal of Guidance, Control, and Dynamics, vol. 23, no. 6, pp. 10761078, 2000.
Mathematical Problems in Engineering 11

16 G. Mengali and A. A. Quarta, Earth escape by ideal sail and solar-photon thrustor spacecraft,
Journal of Guidance, Control, and Dynamics, vol. 27, no. 6, pp. 11051108, 2004.
17 G. Mengali and A. A. Quarta, Time-optimal three-dimensional trajectories for solar photon thruster
spacecraft, Journal of Spacecraft and Rockets, vol. 42, no. 2, pp. 379381, 2005.
18 B. Dachwald and P. Wurm, Design concept and modeling of an advanced solar photon thruster, in
Proceedings of the 19th AAS/AIAA Space Flight Mechanics Meeting, Savannah, Ga, USA, February 2009,
paper no. AAS 09-147.
19 G. Mengali and A. A. Quarta, Compound solar sail with optical properties: models and
performance, Journal of Spacecraft and Rockets, vol. 43, no. 1, pp. 239245, 2006.
20 A. D. Guerman and G. Smirnov, Comment on compound solar sail with optical properties: models
and performance, Journal of Spacecraft and Rockets, vol. 44, no. 3, pp. 732734, 2007.
21 P. A. Tychina, V. A. Egorov, and V. V. Sazonov, Optimization of the ight of a spacecraft with a solar
sail from Earth to Mars with a perturbation maneuver near Venus, Cosmic Research, vol. 40, no. 3, pp.
255263, 2002.
22 G. Smirnov and V. Bushenkov, Curso de Optimizacao: Programacao Matematica, Calculo de Variacoes,
Controlo Optimo, Escolar Editora, Lisbon, Portugal, 2005.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 303604, 12 pages
doi:10.1155/2009/303604

Research Article
Alternative Transfers to the NEOs 99942 Apophis,
1994 WR12, and 2007 UW1 via Derived
Trajectories from Periodic Orbits of Family G

C. F. de Melo,1 E. E. N. Macau,2 and O. C. Winter3

1
UFABC, Universidade Federal do ABC, Santo Andre, 09210-170, SP, Brazil
2
INPE, Instituto Nacional de Pesquisas Espaciais, Sao Jose dos Campos, 12227-010, SP, Brazil
3
UNESP, Universidade Estadual Paulista, Guaratingueta, 12516-410, SP, Brazil

Correspondence should be addressed to C. F. de Melo, [email protected]

Received 30 July 2009; Accepted 11 December 2009

Recommended by Maria F. P. S. Zanardi

Swing-by techniques are extensively used in interplanetary missions to minimize fuel consump-
tion and to raise payloads of spacecrafts. The eectiveness of this type of maneuver has been
proven since the beginning of space exploration. According to this premise, we have explored the
existence of a natural and direct link between low Earth orbits and the lunar sphere of inuence to
get low-energy transfer trajectories to the Near Earth Objects NEOs 99942 Apophis, 1994 WR12,
and 2007 UW1 through swing-bys with the Moon. The existence of this link is related to a family of
retrograde periodic orbits around the Lagrangian equilibrium point L1 predicted for the circular,
planar, restricted three-body Earth-Moon-particle problem. The trajectories in this link are sensitive
to small disturbances. This enables them to be conveniently diverted reducing so the cost of the
swing-by maneuver. These maneuvers allow a gain in energy sucient for the trajectories to escape
from the Earth-Moon system and to stabilize in heliocentric orbits between the Earth and Venus or
Earth and Mars. Therefore, the trajectories have sucient reach to intercept the NEOs orbits.

Copyright q 2009 C. F. de Melo et al. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
The dynamics of the circular restricted three-body Earth-Moon-particle problem predicts the
existence of the retrograde periodic orbits around the Lagrangian equilibrium point L1. Such
orbits belong to the so-called Family G 1, and starting from them it is possible to dene a
set of trajectories that form a natural round trip link between the Earth and Moon and a link
between the Earth, the Moon and the distant space of the Earth sphere of inuence. These
links occur even for more complex dynamical systems as the complete Sun-Earth-Moon-
particle problem. In the last dynamical system, many kinds of transfers can be exploited.
2 Mathematical Problems in Engineering

For example, the round trip link is useful for transfers between Low Earth Orbits LEOs
and Low Lunar Orbits LLOs, including polar LLOs, since some of its trajectories pass a
few hundreds of kilometers from the Earths surface and a few dozens of kilometers from the
Moons surface Figure 1 2. Moreover, the round trip link is also useful for transfer between
two LEOs with dierent altitudes and inclinations 3. In these cases the Moons gravitational
eld, which is near the Earth, works as an extra impulse and provides an ecient method
to minimize the fuel consumption to be used in plane change maneuvers. This is possible,
because the trajectories of the links can lead spacecrafts in a natural way to accomplish swing-
bys with the Moon. In this paper, we use the link between the Earth, the Moon and the distant
space of the Earth sphere of inuence and the energy gain of the swing-by with the Moon to
get escape trajectories and to design missions to NEOs.
In general, a swing-by, or a yby, consists of the alteration of small celestial bodies
orbits comets, asteroids, or spacecrafts when they have a close approach with a planet
or a moon. Literature presents many studies about this subject with general descriptions,
applications in interplanetary missions and in the Earth-Moon System. For example, Lawden
4, Minovitch 5, and Broucke 6 present general descriptions of the mechanics of gravity-
assisted maneuvers. Successful examples in interplanetary missions were the Voyager 1 and
2 which accomplished several swing-bys with the visited planets to gain energy 7, and
Ulysses solar probe which accomplished a swing-by with Jupiter to get a perpendicular orbit
to the ecliptic and observe the Suns Poles 8. Among the many studies on swing-bys in the
Earth-Moon system, we can highlight Dunham and Davis 9, who designed missions with
multiple swing-bys with the Moon; Upho 10, who gave a description about lunar gravity-
assisted maneuvers, and Prado 11, who presented a study to use yby with the Moon to
accomplish transfers between Earth orbits with the same altitudes and dierent inclinations.
Among the missions that accomplished swing-bys with the Moon, we should mention the
ISEE-3 which after completing its mission in 1982 was renamed ICE International Cometary
Explorer and accomplished several lunar encounters to gain energy and intercept the comets
Giacobini-Zinner in September 1985 and Halley, in March 1986 12. On the other hand, also
in general, the literature does not present works focused on getting the trajectory that will
accomplish the swing-by. In this work we do not only use the swing-by to obtain transfer
trajectories, but we have included a study that involves the use of trajectories derived from
periodic orbits to accomplish the swing-by in an ecient way. This allows the reduction of
the swing-by maneuver costs.
Our results, considering trajectories derived from periodic orbits of Family G and the
quasiperiodic orbits that oscillate around them, have allowed us to present the parameters to
design Earth-NEOs transfers. The trajectories studied in this work can oer a saving up to
4% in relation to the conventional methods used to send spacecrafts to these asteroids.
This article is laid out according to the following order: in Section 2, the dynamical
systems and the orbits of Family G are described; in Section 3, the set of trajectories of
interest are dened. In Section 4, the studies about the mechanism of the proposed transfers
between Earth and NEOs are shown and, in Section 5, we present and discuss the results. The
conclusions on the work are presented in Section 6.

2. Earth-Moon Link
The equations of motion of a small particle of negligible mass, m, moving under the
gravitational inuence of two bodies with preponderant masses, written in a normalized
Mathematical Problems in Engineering 3

rotating coordinate system x, y, also-called synodic system, are

U U
x 2ny  , y 2nx  , 2.1
x y

where n is the mean motion and U is a scalar function given by

n2  2 1 2
U x y2 , 2.2
2 r1 r2

1 and 2 are the reduced masses of the two preponderant bodies, called primaries. r12 
x 2 2 y2 and r22  x 1 2 y2 are the distances between the primaries and m.
Equation 2.1 dene the well-known restricted three-body problem in which the primaries
have circular orbits around their common centre of mass 1, 13.
Equation 2.1 presents a special number of solutions that can be achieved searching
for points in synodic frame where x  y  x  y  0. There are ve points with these
features and they are called of Lagrangiam equilibrium points, Li , i  1, . . . , 5. The locations of
these points are obtained solving the simultaneous nonlinear equations U/x  U/y  0.
Figure 1 shows the locations of these points for Earth-Moon-particle problem. In this case,
1  Earth  0.9878494, 2  Moon  0.0121506, and n  1.
In the Earth-Moon-particle problem, groups of periodic orbits around the Earth, the
Moon and the Li points are known. We have considered a particular family of periodic orbits
around L1 known as Family G 13. In the normalized synodic frame, one of the groups of
initial conditions that allow us to obtain the Family G orbits has the form

 
x0 , y0 , z0 , x0 , y0 , z0  x0 , 0, 0, 0, y0 , 0 , 2.3

where Moon RE  < x0 < xL1, with RE  RE /384,400 km  0.016572, RE  6, 370 km is the
average radius of the Earth, xL1  0.836893 is the abscissa of the Lagrangian equilibrium
point L1, and 9.389476 y0 601.045381 1, which corresponds to a variation between
9.607 km/s and 614.964 km/s. The average Earth-Moon distance 384,400 km is chosen as
the unit of length of the normalized system. According to 2.3, for any orbit of Family G,
in t  0, the Earth, the particle and the Moon are aligned in this order, that is, in inferior
conjunction. All orbits in Family G are unstable and this feature is very important for the
transfers here exploited. Figure 1 also shows a typical orbit of Family G.
Periodic orbits as the one shown in Figure 1 and the quasiperiodic orbits that oscillate
around them are tangents to the Earth and Moon low orbits. In other words, they dene a link
between the Earth and the lunar sphere of inuence. This property allied with the instability
of Family G orbits allows the design of controlled swing-bys with the Moon to gain sucient
energy to escape from the Earth-Moon system 2. Moreover, we remark that these properties
allow the reduction of the swing-by maneuver costs besides providing the necessary energy
to win the Earths and the Moons gravitational elds 3, 14. Controlled swing-bys also allow
the use of this energy to obtain trajectories inclination changes 3.
In order to exploit these properties and to design more ecient and realistic Earth-
NEOs transfers, the Suns gravitational eld must be introduced. That is, new equations of
motion must be considered, this time, taking into account the Sun, the Earth and the Moon
4 Mathematical Problems in Engineering

1.5
Typical periodic
orbit of family G
1
L4

0.5

L3 L1 L2
y 0
Earth
Moon
0.5

L5
1

1.5
1.5 1 0.5 0 0.5 1 1.5
x

Figure 1: Locations of Lagrangian equilibrium points and a typical periodic orbit of Family G in synodic
frame x, y.

mutual gravitational attractions, besides the characteristics of the Earths orbit around the Sun
eccentricity and the Moons orbit around the Earth eccentricity and inclination. Thus, we
dene the full four-body Sun-Earth-Moon-particle problem, whose 12 equations of motion in
a Cartesian coordinates system X, Y, Z with origin in a xed point of the space are


4
j 
Ri  3
Rj Ri , 2.4
j1 Rji
i
j/

1/2
where Rij  |Rj Ri |  Xj Xi 2 Yj Yi 2 Zj Zi 2  , with j 
/ i, are the distances
between ith and jth bodies. The eccentricity of the Earths orbit and the eccentricity and
inclination of the Moons orbit are introduced through initial conditions.
The link between the Earth and the lunar sphere of inuence continues existing, even
when the dynamical system given by 2.4 is considered. This can be seen in Figure 2a
that shows a trajectory found considering the full four-body problem. Figure 2b shows
the inclination gain relative to the ecliptic after the swing-by with the Moon. Note that the
trajectory inclination change is about 52 degrees. As a rst conclusion, we can say that the
Moon gravitational eld acts like a natural propeller, without fuel consumption. In the next
section, we are going to exploit this property in a controlled way. That is, taking advantage
of the existing natural link between the Earth and the lunar sphere of inuence to obtain a
change in inclination and the energy gain needed for a determined sort of mission.

3. Escape Trajectories
The existence of the link between the Earth and the lunar sphere of inuence for the full
four-body problem is essential to the design of the Earth-Moon system escape trajectories.
Mathematical Problems in Engineering 5

Trajectory G
Apogee Moons orbit
1
Encounter with
0.5 the moon
Earth
0
z
0.5
1
1.5
1
0.5
0
1.5 1 0.5
0.5 1
y
0
0.5
x 1 1.5
a
180
Inclination relative to the ecliptic degrees

160

140

120

100

80

60

40

20

0
0 5 10 15 20 25 30 35 40
Time days

b

Figure 2: a Derived trajectory of a Family G periodic orbit considering the full four-body problem in
geocentric frame x , y , z . b Trajectory inclination change as function of the time.

For this dynamical system, the initial conditions of an escape trajectory in a geocentric frame
x , y , z  are 3
 
x0 , y0 , z0 , x0 , y0 , z0  x0 , 0, z0 , 0, y0 , 0 , 3.1

where
x0  RE h0  cosiMoon ,

y0  0,

z0  RE h0  siniMoon ,

3.2
x0  0,

y0  VI ,

z0  0.
6 Mathematical Problems in Engineering

Quantity VI is called injection velocity. Considering a circular LEO, VI  VCE V1 ,

where VCE is the LEOs velocity, VCE  GMEarth /RE h0 1/2 , G is the gravitational constant,
MEarth is the Earths mass, and h0 is the LEOs altitude. iMoon is the inclination of the Moons
orbit. V1 is the velocity increment that must be provided to the spacecraft. V1 is a function
of the h0 and VCE and its value is given by

VCE 
V1  2.3340 106 h0 0.8085 0.0001 . 3.3
2

Quantity assumes values that will take into account the relative position among
the Sun, the Earth, and the Moon. In other words, the values assumed by reect the
inuence of the eccentricities of the Earths orbit and, mainly, of the Moons orbit in the
distance between them. Equations to estimate the values assumed by as a function
of the true anomalies of the Earths and Moons orbits are given by de Melo et al.
3.
Here, we are interested in Earth-Moon escape trajectories. Then, we can consider the
Earth and the Moon in circular orbits, with the average radius of the Moons orbit equal to
384,400 km. This does not represent a signicant loss of precision. This way, we can assume
that  0 in 3.3 and, then, we just analyze the quantity whose values will determine
the spacecraft position relative to the Moon and the Earth during the passage by the lunar
sphere of inuence and the swing-by. For 1.50 104 km/s 1.50 104 km/s, we
will have collision trajectories with the Moon, for < 1.50 104 km/s, the trajectories will
have periselenium in the region close to the Moon, between itself and the Earth anterior
region. For > 1.50 104 km/s, the periselenium will be in the posterior region to the
Moon relative to the Earth. For < 1.50 104 km/s, it is possible to design transfers
between LEOs and LLOs of any altitudes and also between two Earth orbits of dierent
inclinations and altitudes, both at a low cost 2, 3. Our interest, however, is the trajectories
generated for > 1.50 104 km/s. In those cases, the gain of energy with the swing-by
guarantees the necessary velocity increment to win the Earths and the Moons gravitational
elds.
The spacecraft must reach the velocity VI when the Earth, the spacecraft and the Moon
are aligned in this order. This is necessary because the trajectory is derived from a Family G
orbit, and this is the only situation in which the orbits of Family G are tangent to the circular
LEOs. Note that this condition does not impose signicant restrictions in practical purpose.
For instance, if a satellite is in a LEO with an altitude of 200 km, this condition will occur every
1.48 hour, approximately, which is the period of the LEO. Figure 3 shows a typical escape
trajectory derived from Family G orbits. From now on, we will call them of escape trajectories
G.
It is important to note that the trajectories evolution of the Earth-Moon and escape
links shown here Figures 2 and 3 do not suer any intervention since they start from the
LEO. That is, the inclination and the energy gain occur naturally during the passage through
the lunar sphere of inuence. However, it is possible to take advantage of the inherent
instability of these trajectories to control the spacecraft passage through the lunar sphere of
inuence 2, 15. This idea will be exploited in the next section to generate specic escape
trajectories for determined missions.
Mathematical Problems in Engineering 7

1 Trajectory G
Apogee Moons orbit
0.5
Encounter with
0
z the moon Earth
0.5
1
1.5
1
0.5
1.5 0
1 0.5 y
0.5 1
0
x 0.5 1 1.5
a
0.2

0
Energy J/kg

0.2

0.4

0.6
0 20 40 60 80 100 120 140
Time days

b

Figure 3: a Typical escape trajectory from the Earth-Moon system. b Energy gain relative to the Earth
as function of the time for the trajectory. Note that it is a typical swingy-by energy gain.

4. Escape Trajectories to Design Earth-NEOs Transfers

The basic idea is to nd transfer trajectories between the Earth and the NEOs starting from
escape trajectories derived from Family G periodic orbits. This will be done to three NEOs:
the 99942 Apophis, the 1994 WR12, and the 2007 UW1. All these are NEOs of the Aten
asteroids class and they are dened by having semimajor axes of less than one astronomical
unit 1 AU is equal to 149.6 106 km. Some data about these objects are provided in Tables
1 and 2. Table 1 shows the orbital elements: semimajor axis a, eccentricity e, inclination
i, longitude of ascending node , argument of perihelion w, and the orbital period
P. Table 2 brings other information as average orbital velocity Vave , Escape velocity Vesc ,
Aphelion, Perihelion, Mass, and diameter. In Table 3, the three next Closest Point Approaches
CPA with the Earth are presented for 99942 Apophis, the 1994 WR12, and the 2007 UW1.
Once the mission target is determined, it is necessary to nd the point in the asteroids
orbit where the spacecraft will reach it. However, our goal here is not to nd an optimal
trajectory exactly. We intend to show that transfer trajectories derived from periodic orbits
around L1, generated after swing-bys with the Moon, can reach an NEO with Vs smaller
8 Mathematical Problems in Engineering

Table 1: NEOs orbital elements.

Asteroid a UA e i deg deg w deg P yr
Apophis 0.992 0.191 3.331 304, 5 126, 7 0.89
1994WR12 0.757 0.397 6.864 62, 85 205, 9 0.66
2007UW1 0.907 0.121 8.224 26, 04 146, 5 0.86

Inclination relative to the ecliptic. Source: http://neo.jpl.nasa.gov/.

Table 2: Some characteristics of the NEOs orbits.

Asteroid Vave km/s Vesc m/s Aphelion UA Perihelion UA Mass kg Diameter m
Apophis 30.728 0.144 1.099 0.746 2.7 1010 270
1994WR12 34.706 0.050 1.058 0.455 2.0 109 110260
2007UW1 31.509 0.047 1.017 0.798 1.3 109 80190
Source: http://neo.jpl.nasa.gov/.

than the ones required by trajectories that leave from the Earth directly. In order to do so, we
only considered trajectories that are tangents to the NEOss orbits. In order to nd an escape
trajectory that will reach an NEOs orbit, for example, an algorithm integrates a number
of them starting from the interception point backward in time until they reach the closest
point approach with the Moon during the swing-bys. These integrations are supplied for
initial conditions determined starting from previous numerical simulations of trajectories
that escape naturally from the Earth-Moon system, as the one shown in Figure 3. In these
simulations, a group of trajectories that have a point closer to the NEOs orbit is selected. The
positions and the velocities of these trajectories are analyzed and adjusted to form a group of
initial conditions for the integrations beginning in the interception point of the NEOs orbit.
When the closest point approach with the Moon is found, the algorithm calculates a small
V to determine which escape trajectory G is more suitable for the transfer. In other words,
it calculates the components values of the spacecrafts initial condition in the LEO 3.1. The
algorithm provides a group of solutions with VTotal , V1 , and Vintermediary , transfer time
and other quantities of interest. Besides the Vintermediary applied in the closest point approach
with the Moon, the algorithm also calculates another Vintermediary applied in the rst apogee
of trajectory G. This apogee is reached about eight days after the departure from the LEO and
prior to the passage by the lunar sphere of inuence Figure 3. This is the best point to do
velocity and inclination changes. In this point, the trajectories G velocities are very small; so
small Vs can provide signicant velocity and inclination changes.
A similar procedure is used to nd the transfer trajectory between the Earth and the
NEO for Patched-conic approximation. In this method, the transfers are designed seeking for
a heliocentric conic arc that links an LEO to an orbit around the asteroid 16.

5. Results
The results of the intercept missions to the NEOs Apophis, 1999 WR12, and 2007 UW1
are in Tables 4 and 5. Table 4 presents the values found for the missions conceived by
escape trajectories G, and Table 5 presents the correspondent values found for the missions
conceived by Patched-conic approach. For each sort of mission, the Vs applied in the
departure from the LEO, for occasional velocity and inclination changes, the spacecraft
velocities relative to the asteroids in the closest point approach CPA, and the transfer times
Mathematical Problems in Engineering 9

Table 3: Three next closest point approaches.

Asteroid
Date Apophis 1994WR12 2007UW1
Date 2013/01/09.48850 2017/11/22.55300 2020/04/24.60600
1st
Minimum possible distance UA 0.096662 0.001781 0.137462
Date 2029/03/06.05209 2019/11/24.23939 2026/10/07.03002
2nd
Minimum possible distance UA 0.112651 0.001731 0.002622
2039/06/0.832120
Date 2036/04/14.60516 2021/12/05.05848
2039/10/22.06710
3rd
0.099407
Minimum possible distance UA 0.000254 0.001540
0.001451
Sources: http://neo.jpl.nasa.gov/.

Table 4: Missions data of the Earth-NEOs transfer by escape trajectories G.

V1
Vint 1

Vint 2 VTotal T1 V1 T2 V2
Asteroid
km/s km/s km/s km/s days km/s days km/s
Apophis 3.148 0.010 0.018 3.176 81 0.800 225 0.550
1994WR12 3.147 0.018 3.165 53 3.700 103 3.600
2007UW1 3.149 0.018 0.012 3.169 153 0.490

Departure, 1st apogee, CPA with the Moon,

spacecraft velocity relative to the asteroid in T 1, and
spacecraft
velocity relative to the asteroid in T 2.

Table 5: Missions data of the Earth-NEOs transfer by Patched-conic approach.

Asteroid V1 km/s Vint

1 km/s VTotal km/s T 1 days V1 km/s T 2 days V2 km/s
Apophis 3.280 0.020 3.308 155 1.200 198 1.350
1994WR12 3.225 0.035 3.260 131 4.100
2007UW1 3.250 0.028 3.278 111 0.690

Departure, middle curse,

spacecraft velocity relative to the asteroid in T 1, and
spacecraft velocity relative to the
asteroid in T 2.

were calculated. The altitude of the circular departure LEO was considered equal to 200 km,
and the altitude of the CPA with the asteroids was considered equal to 1 km. For the transfers
between the Earth and Apophis and the Earth and 1994 WR12, it was possible to nd two
CPAs between the spacecraft and the asteroids. In Tables 4 and 5, we call these times of T1 for
the 1st CPA and T2 for the 2nd CPA. It is interesting to note that spacecraft velocities relative
to the asteroids are dierent in each CPA, and, depending on the mission goals, the choice of
one of them could represent an extra fuel saving in an occasional insertion maneuver of the
spacecraft into an orbit around the asteroid.
Figure 4a shows the Earths and the Apophiss orbits and the heliocentric ellipse arc
generated by the Patched-conic approach. The times of the 1st and 2nd CPA are 155 and 198
days after the departure from the LEO, respectively, and the distances from the Earth to these
points are 11.2 106 km and 113.6 106 km, respectively. The values of T 1 and T 2 of the other
transfers are in Tables 4 and 5.
By analogy, Figure 4b shows the escape trajectory G found for the Earth-Apophis
transfer. The 1st and the 2nd CPAs take place 86 and 225 days after the departure from the
LEO, and the distances from the Earth to these points are 52.27 106 km and 204.45 106 km,
respectively.
10 Mathematical Problems in Engineering

2e 008
1st CPA Transfers trajectory

1.5e 008

2nd CPA
1e 008

5e 007
Y km
Sun
0
Apophis orbit
5e 007

1e 008

Earths orbit
1.5e 008

2e 008
0
1.5e 008

1e 008

1.5e 008
5e 007

1e 008

2e 008
2e 008

5e 007

X km
a

2e 008
Trajectory G
1.5e 008

2nd CPA
1e 008 1st CPA

5e 007
Y km

Sun
0
Apophis orbit

5e 007

1e 008

Earths orbit
1.5e 008

2e 008
0
1.5e 008

1e 008

1.5e 008
5e 007

1e 008

2e 008
2e 008

5e 007

X km
b

Figure 4: a Earths and Apophiss orbits and the transfer trajectory found for Patched-conic transfer. b
Earths and Apophiss orbits and transfer trajectory G heliocentric frame.
Mathematical Problems in Engineering 11

The analysis of Tables 4 and 5 shows that the procedure described in Section 4 to
generate transfer trajectories between the Earth and NEOs, derived from periodic orbits
of Family G, requires Vs of 2% up to 4% less than the transfers conceived by Patched-
conic approach. Besides, the spacecraft velocities relative to the asteroids in the CPAs, for
the trajectories G, are the smallest. This can also provide fuel saving in occasional insertion
maneuvers into an orbit around the asteroids. However, the smallest relative velocities
always correspond to the longest transfer times.

6. Conclusions
In this work, a procedure capable of generating transfer trajectories between the Earth and
the NEOs has been presented. These trajectories are derived from the periodic orbits around
the Lagrangian point L1 and escape from the Earth-Moon system after a swing-by with the
Moon.
In terms of VTotal , those required by escape trajectories G are, in general, fewer than
the ones required by conventional transfers Patched-conic, between 2% up to 4%. Besides,
the spacecraft velocities relative to the asteroids are also, in general, less than those found by
the conventional methods.
With regard to the transfer time, we verify that in two cases Apophis and 1994 WR12
it was possible to nd two CPAs. The time longest always corresponds to the smallest relative
velocity in CPA for trajectories G. We also verify that there are not any discrepancies between
the transfer times of the two considered methods.
This way, we can conclude that the escape trajectories G presented in this work are a
good alternative to design future missions destined to the NEOs.
Finally, future studies of techniques to accomplish a swing-by with the Earth and, thus,
to gain more energy and a larger reach could provide the planning of transfers to more distant
objects, including the internal planets.

Acknowledgments
The authors are grateful to the FAPESP Fundacao de Amparo a Pesquisa do Estado de Sao
Paulo, processes 2005/05169-7 and 2006/04997-6, and the CNPq Conselho Nacional para
Desenvolvimento Cientco e Tecnologico Brazil, for the nancial support.

References
1 R. A. Broucke, Periodic orbits in the restricted three-body problem with Earth-Moon masses, Tech.
Rep. 32-1168, JPL, Los Angeles, Calif, USA, 1968.
2 C. F. de Melo, E. E. N. Macau, O. C. Winter, and E. V. Neto, Alternative paths for insertion of probes
into high inclination lunar orbits, Advances in Space Research, vol. 40, no. 1, pp. 5868, 2007.
3 C. F. de Melo, E. E. N. Macau, and O. C. Winter, Strategies for plane change of Earth orbits using
lunar gravity and derived trajectories of family G, Celestial Mechanics & Dynamical Astronomy, vol.
103, no. 4, pp. 281299, 2009.
4 D. Lawden, Perturbation maneuvers, Journal of the British Interplanetary Society, vol. 13, no. 5, 1954.
5 M. A. Minovitch, A method for determining interplanetary free-fall reconnaissance trajectories,
Technical Memo TM-312-130, JPL, 1961.
6 R. A. Broucke, The celestial mechanics of gravity assist, in Proceedings of AIAA/AAS Astrodynamics
Conference, Minneapolis, Minn, USA, August 1988.
12 Mathematical Problems in Engineering

7 C. E. Kohlhase and P. A. Penzo, Voyager mission description, Space Science Reviews, vol. 21, no. 2,
pp. 77101, 1977.
8 R. Carvell, Ulysses-the Sun from above and below, Space, vol. 1, pp. 1855, 1985.
9 D. Dunham and S. Davis, Optimization of a multiple lunar swing-by trajectory sequence, Journal of
the Astronautical Sciences, vol. 33, no. 3, pp. 275288, 1985.
10 C. Upho, The art and science of lunar gravity assist, orbital mechanics and mission design,
Advances in the Astronautical Sciences, vol. 69, pp. 333346, 1989.
11 A. F. B. A. Prado, Orbital control of a satellite using the gravity of the moon, Journal of the Brazilian
Society of Mechanical Sciences and Engineering, vol. 28, no. 1, pp. 105110, 2006.
12 R. Farquhar, Detour a Comet, Journey of the International Cometary Explorer, Planetary Report, vol.
5, no. 3, pp. 46, 1985.
13 C. D. Murray and S. F. Dermott, Solar System Dynamics, Cambridge University Press, Cambridge,
Mass, USA, 1999.
14 K. Deb, N. Padhye, and G. Neema, Interplanetary trajectory optimization with swing-bys using
evolutionary multi-objective optimization, in Proceedings of the 2nd International Symposium on
Intelligence Computation and Applications (ISICA 07), L. Kang, Y. Liu, and S. Zeng, Eds., vol. 4683 of
Lecture Notes in Computer Science, pp. 2635, Wuhan, China, September 2007.
15 E. E. N. Macau and C. Grebogi, Control of chaos and its relevancy to spacecraft steering,
Philosophical Transactions of the Royal Society of London. Series A, vol. 364, no. 1846, pp. 24632481, 2006.
16 V. A. Chobotov, Orbital Mechanics, AIAA, 1991.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 159287, 10 pages
doi:10.1155/2009/159287

Research Article
Controlling the Eccentricity of Polar Lunar Orbits
with Low-Thrust Propulsion

O. C. Winter,1, 2 D. C. Mourao,1 C. F. de Melo,3 E. N. Macau,2

J. L. Ferreira,4 and J. P. S. Carvalho1
1
Grupo de Dinamica Orbital & Planetologia, UNESP, Universidade Estadual Paulista, Avenida Ariberto
Pereira da Cunha, 333, CEP, 12516-410, Guaratingueta, SP, Brazil
2
Instituto Nacional de Pesquisas EspaciaisINPE, Avenida dos Astronautas 1752, CEP, 12227-010,
Sao Jose dos Campos-SP, Brazil
3
Universidade Federal do ABC, UFAB, Santo Andre, 09210-170, SP, Brazil
4
Instituto de Fsica, Universidade de BrasliaUnB, CP 04455, CEP, 70919-970, Braslia, DF, Brazil

Correspondence should be addressed to O. C. Winter, [email protected]

Received 30 July 2009; Accepted 31 December 2009

Recommended by A. F. B. De Almeida Prado

It is well known that lunar satellites in polar orbits suer a high increase on the eccentricity due to
the gravitational perturbation of the Earth. The nal fate of such satellites is the collision with the
Moon. Therefore, the control of the orbital eccentricity leads to the control of the satellites lifetime.
In the present work we study this problem and introduce an approach in order to keep the orbital
eccentricity of the satellite at low values. The whole work was made considering two systems: the
3-body problem, Moon-Earth-satellite, and the 4-body problem, Moon-Earth-Sun-satellite. First,
we simulated the systems considering a satellite with initial eccentricity equals to 0.0001 and
a range of initial altitudes between 100 km and 5000 km. In such simulations we followed the
evolution of the satellites eccentricity. We also obtained an empirical expression for the length
of time needed to occur the collision with the Moon as a function of the initial altitude. The results
found for the 3-body model were not signicantly dierent from those found for the 4-body model.
Secondly, using low-thrust propulsion, we introduced a correction of the eccentricity every time it
reached the value 0.05.

Copyright q 2009 O. C. Winter et al. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
Recently, several nations presented plans to reach the Moon. Satellites have been launched
and many more are planned for following years see, e.g., 1. The expectations are that in
the near future there will be a lunar base. The lunar poles are particularly of interest since
it seems to be where water can be found. Therefore, long living satellites in polar lunar
orbits will be needed. It is well known that lunar satellites in polar orbits suer a strong
gravitational perturbation from the Earth. That eect is a natural consequence of the Lidov-
Kozai resonance.
2 Mathematical Problems in Engineering

It is well known that the Lidov-Kozai resonance introduces equilibrium congura-

tions. In the case of lunar polar orbits disturbed by the Earths gravitational eld, this can
be used as an advantage to implement constellations of satellites with elliptic highly inclined
orbits 2, 3. On the other hand it causes instability for near circular highly inclined orbits.
Wytrzyszczak et al. 4 studied the regular and chaotic motion of geosynchronous satellites
disturbed by the Moons gravitational eld. They found that the chaotic nature of high
inclination satellites is caused due to the signicant eccentricity growth caused by the Lidov-
Kozai resonance.
Similarly, the nal fate of polar lunar near circular satellites is the collision with the
Moon. Therefore, the control of the orbital eccentricity leads to the control of the satellites
lifetime.
In this paper we propose the control of the eccentricity using an electrical thruster,
similar to the one that is in development at the University of Braslia. Electric propulsion is
basically a technique of space propulsion which involves the conversion of electrical power
into the kinetic power or thrust of the exhaust beam of ionized particles. The ability to
obtain high exhaust velocities with ionized particles enables plasma thrusters to perform
high specic impulse mission in space 5. The main innovation of the thruster that is being
developed at the University of Braslia is the use of a permanent magnet, saving energy
during the mission. Preliminary results in the laboratory show that it is possible to obtain
more than 100 mN with this technology. Inspired on this thruster project, in this work, we
assume a constant exhaust velocity, and we can control the switch of the thruster during the
mission.
In the present work we introduce an approach in order to keep the orbital eccentricity
of lunar polar satellites at low values. The approach is based on the use of low-thrust
propulsion in order to introduce a correction of the eccentricity.
In the next section we introduce the Lidov-Kozai resonance. In Section 3 we show the
evolution of the eccentricities form our numerical integrations. The approach proposed to
control the eccentricity and its application is presented in Section 3. In the nal section we
present our nal comments.

2. The Lidov-Kozai Resonance

Lidov 6, studying the dynamics of articial satellites, and Kozai 7, studying the dynamics
of asteroids, independently discovered what is now called the Lidov-Kozai resonance.
Following, we introduce the basic features of such resonance.
In this section we adopted a simple model see, e.g., 2 for the orbital evolution of an
articial satellite disturbed by a third body in circular equatorial orbit around the primary. It
was obtained by double averaging the system 8 taking into account the disturbing function
expanded in Legendre polynomials up to second order and the eccentricity of the disturbing
body also up to the second order. The disturbing function of the problem is averaged
independently over the mean longitudes of the satellite and the third body. The standard
denition for average used in this work is

 2
1
F  FdM, 2.1
2 0

where M is the mean anomaly, which is proportional to the time.

Mathematical Problems in Engineering 3

Following such approach one can nd the double averaged disturbing function given
by see, e.g., 9

3Gm1 m2 a2   2
R 2 e sin2
i e 2
5 cos 2 3sin2
i , 2.2
16a2E

where a, e, , and i are, respectively, the semimajor axis, eccentricity, argument of pericenter
and inclination, G is the gravitational constant, aE is the semimajor axis of the Earth with
respect to the Moon, and m1 and m2 are the masses of the Earth and Moon respectively.
Substituting R in Lagranges planetary equations see, e.g., 10, we nd

da
 0, 2.3
dt
de 15e ( 
 1 e2 n2E 2 3eE2 sin2sin2 i, 2.4
dt 16n

dw 3 2 3eE2 n2E   
 1 5 cos2 i e2 5 cos 2 cos2 i e2 1 2.5
dt 16n 1 e2

where n is the mean motion and  m1 /m1 m12 .

Considering the case when de/dt  0 and d/dt  0 one can nd three rst integrals:

a  a0 , 2.6

1 e2 cos2 i  k1 , 2.7
 
2
e 2
sin2 i sin2 w  k2 , 2.8
5

where a, e, i, and w are the semi-major axis, eccentricity, inclination, and argument of
pericenter of the satellite, and a0 , k1 , and k2 are the constants of motion. This system has a
set of xed points given by
 
5
w  90 or 270 , e
2
cos2 i  1. 2.9
3

Therefore, for a system with e, i, and satisfying conditions 2.10, the satellite would
be in what can be called a frozen orbit; that is, apart from short-period oscillations, the orbit
would be kept xed in size and location.
A simple analysis of 2.2 and 2.3 shows that 9
i for k2 > 0 and any value of k1 , w circulates,
ii for k2 < 0 and k1 < 3/5, w librates around 90 or 270 ,
iii for k2  0 and w  90 or 270 , i  i 39.2 ,
where i  39.2 is the critical value, which corresponds to the frozen orbit.
4 Mathematical Problems in Engineering

1
0.9
0.8
0.7
0.6
0.5

e
0.4
0.3
0.2
0.1
0
0 50 100 150 200 250 300 350
w deg

e0  0.02 e0  0.6

e0  0.2 e0  0.8
e0  0.4 e0  0.99

Figure 1: Sample of the satellites orbital evolution in a diagram e versus w. There are two sets of initial
values of eccentricity and inclination eo , io  : One for low inclination, io  20 thick lines and other for
high inclination, io  56.2 , thin lines. The values of the initial eccentricities are represented by the colour
code in the bottom of the gure. The argument of pericentre circulates in the case of initial low inclination,
while librates in the case of initial high inclination.

So, that is the Lidov-Kozai resonance. When i > i , the system behaves like a
pendulum, with stable xed points, librations around such points and circulation.
Following another approach let us consider a satellite orbiting the Moon and disturbed
by the Earth in an elliptical orbit with respect to the Moon. Taking into account only the term
P2 of the Legendre polynomial, the disturbing potential is given by

     
45 2 3 2 1 2
R a E e2 sin2 i cos 2w e2 cos2 i n2E eE2 , 2.10
32 5 3 3 3

where eE and nE are the eccentricity, and mean motion of the Earth with respect to the
Moon, and E  mE /mE mM , where mE and mM are the masses of the Earth and Moon,
respectively.
One can identify the Lidov-Kozai resonant features by numerically integrating
Lagranges planetary equations for the temporal variation of the argument of pericentre, w,
and the eccentricity, e, with the disturbing potential given by 2.10. In Figure 1 we present a
sample of these numerical integrations in a diagram e versus w. There are two sets of initial
values of eccentricity and inclination eo , io . One for low inclination io  20  and other for
high inclination io  56.2 . This gure shows a clear dependence of the eccentricity on the
argument of pericentre for an orbit with high inclination. All satellites inclined to the orbital
plane of the third body the Earth by more than 39.2 , the critical inclination, experience a
considerable growth of eccentricity. The Earth causes the Lidov-Kozai resonance driving the
eccentricity growth. For all trajectories from the set with initial inclination higher than the
critical value, the argument of pericentre librates, while it circulates for trajectories from the
set with initial inclination lower than the critical value.
Mathematical Problems in Engineering 5

3. Eccentricity Growth
In this section we present numerical integrations considering two dynamical systems: the 3-
body problem, Moon-Earth-satellite, and the 4-body problem, Moon-Earth-Sun-satellite. In
all simulations the satellite is initially in polar orbit i  90 .
In the case of the 3-body problem, considering a coordinate system centered in the
barycenter of the Earth-Moon system X, Y, Z, the equations of motion of the satellite are
given by

 2


x i

mi
x G 3
x ,
i1 |xi x|

2
mi 
y G 3 y i y , 3.1
i1  yi y 

z   G  z ,
2
mi z
3 i
i1 |zi z|

where m is mass and the index i  1 refers to the Earth and i  2 refers to the Moon. In this
system, the equations of motion for the moon and the Earth are given by

2
mi 

xi  G 3 x j x i ,
 i xj xi 
j1, j /

2 


yi 
mi
G 3 y j y i , 3.2
j1, j 
/i
yj yi 

z   G 
2
mi
i  3 z j z i .
i
j1, j / zj zi 

First, we simulated the system, integrating numerically 3.1 and 3.2, considering
a satellite with initial eccentricity equals to 0.0001 and a range of initial altitudes between
100 km and 5000 km. Figure 2 shows the evolution of the satellites eccentricity for the 3-body
simulations, considering altitudes h  100, 200, 500, 1000, and 5000 km. The plots show an
exponential evolution of the eccentricity. We computed the time needed in order to reach
the eccentricity that corresponds to the collision of the satellite with the Moon. A t of the
collision time, Tcollision , in Earth days, as a function of the altitude, h, is given by the expression:

4 2
Tcollision  2693e0.062h3.9210 h
. 3.3

The same set of simulations was performed considering the 4-body problem, adding the
perturbations of the Sun. Now we considered a new coordinate system X  Y  Z , centered at
the barycenter of Sun-Earth-Moon system. In this case we integrate numerically the equations
similar using prime in the variables to equations 3.1 and 3.2, but we add the index i  3,
where the fourth term refers to Sun.
6 Mathematical Problems in Engineering

0.05

0.04

Eccentricity
0.03

0.02

0.01

0
0 500 1000 1500 2000 2500 3000 3500 4000
Time days

100 1000
200 5000
500

Figure 2: Time evolution of the eccentricity for the 3-body problem. The colour code indicates the initial
altitude in kilometers. The time is in Earth days.

However, the results found for the 4-body model were not signicantly dierent from
those found for the 3-body model. The empirical expression for the length of time needed to
occur the collision with the Moon as a function of the initial altitude is given by
4 2
Tcollision  2494e0.063h3.9410 h
. 3.4

A comparison of the two sets of simulations and 2.5 and 2.6 is shown in Figure 3.

4. Controlling the Eccentricity

In order to control the satellites eccentricity we will use low-thrust propulsion. Following
the work of Sukhanov 11, we use the locally optimal thrust for each orbital element. This
development is based on the performance index, through the minimization of a functional in
the direction of the orbital element to be changed. In our case, the eccentricity is the parameter
to be minimized. The result is a vector, called Lawdens primer vector, P, which gives the
direction of the thruster to be turned on.
The eccentricity of the satellite relative to the Moon is given by
%
c2
e 1 h, 4.1
Gm2

where c is the magnitude of the angular momentum, and h is the integral energy. The primer
vector is given by
 

1 r 2

p P v vn , 4.2
Gm2 e a
Mathematical Problems in Engineering 7

4000

3500

3000

Collision time days

2500

2000

1500

1000

500
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Height km

Figure 3: Collision time as a function of the initial altitude. The red crosses are for the 3-body simulations
and green are for the 4-body simulations. The blue curve corresponds to 2.5 and the purple curve
corresponds to 2.6.

0.052 2400
500 500
0.05 2350
Radial distance km

0.048
2300
0.046
E

2250
0.044
2200
0.042

0.04 2150

0.038 2100
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Time days Time days

a b

Figure 4: Temporal evolution of the eccentricity a and of the orbital radius b. In this simulation the
initial altitude was 500 km and the thrust value used was 0.2 N.

where
r ,

v , and

v n are the vector position, velocity, and tangential velocity relative to the
Moon, and a and P are the semimajor axis and semilatus rectum relative to the Moon.
Then, we have that the acceleration components to change eccentricity are given by
 
1 1 rp2
pr  pa , pn  pa , 4.3
e e a

where pa  c2 /, pr , and pn are the radial and the tangential components, respectively.
The approach we are proposing is a very simple one. The idea is to introduce a
correction on the eccentricity every time it reaches a certain limit. The procedure is as follows.
8 Mathematical Problems in Engineering

90
80
70
60

Fuel kg
50
40
30
20
10
0
0 0.2 0.4 0.6 0.8 1
Life time years

0.1 N 100 km 0.2 N 500 km 0.3 N 1000 km

0.1 N 200 km 0.2 N 1000 km 0.3 N 5000 km
0.1 N 500 km 0.2 N 5000 km 0.4 N 200 km
0.1 N 1000 km 0.3 N 100 km 0.4 N 500 km
0.1 N 5000 km 0.3 N 200 km 0.4 N 1000 km
0.2 N 100 km 0.3 N 500 km 0.4 N 5000 km
0.2 N 200 km

Figure 5: The propellant consumption per year of lifetime for the whole set of simulations, that is, dierent
initial altitudes and dierent values of the thruster.

First x the nominal eccentricity, eo , and the maximum acceptable increase in eccentricity, e,
according to the mission design. Then, turn on the thruster every time the condition

e > eo e 4.4

is satised and turn o the thruster when e > eo .

Following, we present the results of some simulations assuming eo  0.04 and e 
0.01. These simulations were made considering a set of dierent thrust values, from 0.1 N up
to 0.4 N. In each run we measured the length of time TThruster , needed to correct the eccentricity
value from e  0.04 to e  0.05. From these results we obtained empirical expressions of
TThruster as a function of the initial altitude and as a function of the thrust value. As an example,
in Figure 4 is shown the temporal evolution of the eccentricity and of the orbital radius for a
satellite with an initial altitude of 500 km and using a thruster of 0.2 N.
In Figure 5 we present the propellant consumption per year of lifetime for the whole
set of simulations, that is, dierent initial altitudes and dierent values of the thruster. The
time intervals that the thrusters are turned on and o are shown in Figures 6 and 7.

5. Final Comments
In the present work we have studied the problem of polar lunar satellites in near circular
orbits under the gravitational perturbations of the Earth and the Sun. The problem is
dominated by the Lidov-Kozai resonance, which forces satellites with near circular orbits
to have an exponential growth of its eccentricity. In order to keep the satellite with low
eccentricity we proposed to use low-thrust propulsion every time the eccentricity reaches a
limiting eccentricity, acceptable by the requirements of the satellite mission. The results show
Mathematical Problems in Engineering 9

5000

4500

4000

3500

Height km
3000

2500

2000

1500

1000

500

0
0 20 40 60 80 100 120 140 160 180 200
Time interval days

O 0.3 N on
0.1 N on 0.4 N on
0.2 N on

Figure 6: Time interval that the thrusters are turned on and o, for the whole set of simulations.

5000

4500

4000

3500
Height km

3000

2500

2000

1500

1000

500

0
0 20 40 60 80 100 120 140 160 180 200
Time interval days

0.1 N on 0.3 N on
0.2 N on 0.4 N on

Figure 7: Zoom of Figure 6.

that the satellites lifetime can be reasonably extended several years at a not so expensive
cost. Therefore, it is shown that low-thrust propulsion is very adequate for this kind of
purpose.

Acknowledgment
This research was supported by CNPq, Fapesp, and the Brazilian Space Agency.
10 Mathematical Problems in Engineering

References
1 B. H. Foing and P. Ehrenfreund, Journey to the Moon: recent results, science, future robotic and
human exploration, Advances in Space Research, vol. 42, no. 2, pp. 235237, 2008.
2 T. A. Ely, Stable constellations of frozen elliptical inclined lunar orbits, Journal of the Astronautical
Sciences, vol. 53, no. 3, pp. 301316, 2005.
3 T. A. Ely and E. Lieb, Constellations of elliptical inclined lunar orbits providing polar and global
coverage, Journal of the Astronautical Sciences, vol. 54, no. 1, pp. 5367, 2006.
4 I. Wytrzyszczak, S. Breiter, and W. Borczyk, Regular and chaotic motion of high altitude satellites,
Advances in Space Research, vol. 40, no. 1, pp. 134142, 2007.
5 J. H. Souza, Estudo da dinamica de partculas em um propulsor a plasma do tipo Hall com mas permanentes,
M.S. thesis, Universidade de Braslia, Braslia, Brazil, 2006.
6 M. L. Lidov, Evolution of articial planetary satellites under the action of gravitational perturbations
due to external bodies, Iskusstviennye Sputniki Zemli, vol. 8, pp. 545, 1961 Russian.
7 Y. Kozai, Secular perturbations of asteroids with high inclination and eccentricity, Astrophysical
Journal, vol. 67, no. 9, pp. 591598, 1962.
8 A. F. B. A. Prado, Third-body perturbation in orbits around natural satellites, Journal of Guidance,
Control, and Dynamics, vol. 26, no. 1, pp. 3340, 2003.
9 M. A. Vashkovyak and N. M. Teslenko, Peculiarities of the orbital evolution of the Jovian satellite
J34 Euporie, Astronomy Letters, vol. 33, no. 11, pp. 780787, 2007.
10 J. Kovalevsky, Introduction to Celestial Mechanics, Bureau des Longitudes, Paris, France, 1967.
11 A. A. Sukhanov, Lectures on Astrodynamics, INPE, Sao Jose dos Campos, Brazil, 5th edition, 2007.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 471804, 36 pages
doi:10.1155/2009/471804

Research Article
Internal Loading Distribution in Statically Loaded
Ball Bearings Subjected to an Eccentric Thrust Load

Mario Cesar Ricci

Space Mechanics and Control Division (DMC), Brazilian Institute for Space Research (INPE),
Sao Jose dos Campos 12227-010, Brazil

Correspondence should be addressed to Mario Cesar Ricci, [email protected]

Received 29 July 2009; Accepted 18 November 2009

Recommended by Antonio Prado

Rolling-element bearings are simple machine elements of great utility used both in simple
commercial devices as in complex engineering mechanisms. Because of being a very popular
machine element, there is a lot of literature on the subject. With regard to the behavior of internal
loading distribution, elastic deformations at point or line contacts, and geometric parameters
under loading, although there are many works describing the parameters variation models, few
works show such variations in practice, even under simple static loadings. In an attempt to cover
this gap some studies are being developed in parallel. Particularly in this work, a new, iterative
computational procedure is introduced which calculates internal normal ball loads in statically
loaded single-row, angular-contact ball bearings, subjected to a known thrust load which is applied
to a variable distance lever arm or eccentricity from the geometric bearing center line. Numerical
examples results for a 218 angular-contact ball bearing have been compared with those from the
literature. Fifty gures are presented showing geometrical features and the following parameters
variations as functions of the thrust load and eccentricity: contact angle, contact ellipse parameters,
normal ball loads, distances between groove curvature centers, normal and axial deections, and
loading zones.

Copyright q 2009 Mario Cesar Ricci. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
Ball and roller bearings, generically called rolling bearings, are commonly used machine
elements. They are employed to permit rotary motions of, or about, shafts in simple
commercial devices such as bicycles, roller skates, and electric motors. They are also used
in complex engineering mechanisms such as aircraft gas turbines, rolling mils, dental drills,
gyroscopes, and power transmissions.
The standardized forms of ball or roller bearings permit rotary motion between two
machine elements and always include a complement of ball or rollers that maintain the
shaft and a usually stationary supporting structure, frequently called a housing, in a radially
or axially spaced-apart relationship. Usually, a bearing may be obtained as a unit, which
2 Mathematical Problems in Engineering

Figure 1: An angular-contact ball bearing courtesy of SKF Industries.

a Small angle b Large angle

Figure 2: Angular-contact ball bearings.

includes two steel rings; each of which has a hardened raceway on which hardened balls or
rollers roll. The balls or rollers, also called rolling elements, are usually held in an angularly
spaced relationship by a cage, also called a separator or retainer.
There are many dierent kinds of rolling bearings. This work is concerned with single-
row angular-contact ball bearings Figure 1 that are designed to support combined radial and
thrust loads or heavy thrust loads depending on the contact angle magnitude. The bearings
having large contact angle can support heavier thrust loads. Figure 2 shows bearings having
small and large contact angles. The bearings generally have groove curvature radii in the
range of 52%53% of the ball diameter. The contact angle does not usually exceed 40o .
This work is devoted to the study of the internal loading distribution in statically
loaded ball bearings. Several researchers have studied the subject as, for example, Stribeck
1, Sjovall 2, Jones 3, and Rumbarger 4, to cite a few. The methods developed by
Mathematical Problems in Engineering 3

them to calculate distribution of load among the balls and rollers of rolling bearings can be
used in most bearing applications because rotational speeds are usually slow to moderate.
Under these speed conditions, the eects of rolling-element centrifugal forces and gyroscopic
moments are negligible. At high speeds of rotation these body forces become signicant,
tending to alter contact angles and clearance. Thus, they can aect the static load distribution
to a great extension.
Harris 5 described methods for internal loading distribution in statically loaded
bearings addressing pure radial, pure thrust centric and eccentric loads, combined radial
and thrust load, which uses radial and thrust integrals introduced by Sjovall 2, and for ball
bearings under combined radial, thrust, and moment load, initially due to Jones 3.
The method described by Harris for eccentric thrust load, initially due to Rumbarger
4, is an approximate, direct method, based in a single-row, 90o thrust bearing and in thrust
and moment integrals whose values are obtained from tables and graphics, as functions of
eccentricity and pitch diameter. The maximum ball load is given directly and no computer
is necessary. Although it is not entirely appropriate, the method was used by Harris to nd
approximations for the maximum ball load magnitude and for the extension of the loading
zone in the 218 angular-contact ball bearing.
We can see that there are many works describing the parameters variation models
under static loads but few show such variations in practice, even under simple static loadings.
The author believes that the lack of practical examples in the literature is mainly due to the
inherent diculties of the numerical procedures that, in general, deal with the resolution of
several nonlinear algebraic equations that must be solved simultaneously.
In an attempt to cover this gap studies are being developed in parallel 614.
Particularly in this work is described a new, precise method for internal load distribution
computation in statically loaded, single-row, angular-contact ball bearings subjected to a
known external thrust load which is applied to a variable distance lever arm or eccentricity
from the geometric bearing center line. It must be solved iteratively using a digital computer
and can be thought as a particular case of the Jones method, with null external radial
load and external moment load given by the product of the thrust load by the eccentricity.
Unlike Rumbargers method, it is adequate to angular-contact bearings, and theoretically
and numerically more precise. The novelty of the method is in the choice of the set of the
nonlinear equations, which must be solved simultaneously. The author did not nd in the
literature the resolution of this problem using the same set of equations.
The dierence between the method described here and the method described by
Harris for eccentric thrust load mainly comes from the fact that Rumbargers method, for
sake of simplicity, makes use of the pitch radius, de /2, as lever arm, instead of the inner
contact radius, dcj /2, in the r.h.s. of the moment equationsee 4.19 for comparison and
secondarily by the fact that it uses the pitch radius instead of the locus of the centers of the
inner ring raceway groove curvature radii, Ri , in the computations of the load distribution
factor, , in 4.10 and of the extension of load zone, l , in 4.11. These approximations are
guarantee of the straightforwardness but obviously they introduce errors in the normal ball
loads determination. However, at rst glance appears that the method for thrust bearing
is more attractive than the method of this paper because it supplies results more directly
whereas no computer is necessary. But, despite the simplicity of the former, comparative
analyses between the results show signicant dierences in the magnitudes of the maximum
ball load and extension of the loading zone.
4 Mathematical Problems in Engineering

Pd

da de di db do

Figure 3: Radial cross-section of a single-row ball bearing.

r
D

Figure 4: Cross-section of a ball and an outer race showing race conformity.

2. Geometry of Ball Bearings

In this section, the principal geometrical relationships for an unloaded ball bearing are
summarized. The radial cross section of a single-row ball bearing shown in Figure 3 depicts
the diametral clearance and various diameters. The pitch diameter, de , is the mean of the inner-
and outer-race diameters di and do , respectively, and is given by

1
de  di do . 2.1
2

The diametral clearance, Pd , can be written as

Pd  do di 2D. 2.2

Race conformity is a measure of the geometrical conformity of the race and the ball in a plane
passing through the bearing axis also named center line or rotation axis, which is a line
passing through the center of the bearing perpendicular to its plane and transverse to the
race. Figure 4 depicts a cross section of a ball bearing showing race conformity expressed as

r
f . 2.3
D
Mathematical Problems in Engineering 5

Pd
4
D D
A
2 r0 Pd
r0 2 A
ri ri 2

Pd
4

Pe
2
a Initial position b Shifted position

Figure 5: Cross-section of a radial ball bearing showing ball-race contact due to axial shift of inner and
outer rings.

Radial bearings have some axial play since they are generally designed to have a diametral
clearance, as shown in Figures 5a and 5b, that shows a radial bearing with contact due to
the axial shift of the inner and outer rings when no measurable force is applied. The radial
distances between the curvature centers of the two races are the same in Figures 5a and
5b. Denoting quantities which referred to the inner and outer races by subscripts i and o,
respectively, this radial distance value can be expressed as A Pd /2, where A  ro ri D
is the curvature centers distance in the shifted position given by Figure 5b. Using 2.3 we
can write A as

A  BD, 2.4

where B  fo fi 1 is known as the total conformity ratio and is a measure of the combined
conformity of both the outer and inner races to the ball.
The contact angle, , is dened as the angle made by a line, which passes through
the curvature centers of both the outer and inner raceways and that lies in a plane passing
through the bearing rotation axis, with a plane perpendicular to the bearing axis of rotation.
The free-contact angle, f , Figure 5b is the contact angle when the line also passes through
the points of contact of the ball and both raceways and no measurable force is applied. From
Figure 5b, the expression for the free-contact angle can be written as

A Pd /2
cos f  . 2.5
A

From 2.5, the diametral clearance, Pd , can be written as


Pd  2A 1 cos f . 2.6
6 Mathematical Problems in Engineering

de D cos
2
de D cos de /2
2

CL

Figure 6: Cross-section of a ball bearing.

Free endplay, Pe , is the maximum axial movement of the inner race with respect to the outer
when both races are coaxially centered and no measurable force is applied. Free endplay
depends on total curvature and contact angle, as shown in Figure 5b, and can be written as

Pe  2A sin f . 2.7

Considering the geometry of two contacting solids ellipsoids in a ball bearing, we can arrive
at the two quantities of some importance in the analysis of contact stresses and deformations.
The curvature sum, 1/R, and curvature dierence, , are dened as

1 1 1
 ,
R Rx Ry
  2.8
1 1
R ,
Rx Ry

where

1 1 1
 ,
Rx rax rbx
2.9
1 1 1
 ,
Ry ray rby

with rax , rbx , ray , and rby being the radii of curvature for the ball-race contact.
A cross section of a ball bearing operating at a contact angle is shown in Figure 6.
Equivalent radii of curvature for both inner- and outer-race contacts in, and normal to, the
direction of rolling can be calculated from this gure. Considering x the direction of the
Mathematical Problems in Engineering 7

motion and y the transverse direction, the radii of curvature for the ball-inner-race contact
are

D
rax  ray  ,
2
de D cos 2.10
rbx  ,
2 cos

rby  fi D  ri .

The radii of curvature for the ball-outer-race contact are

D
rax  ray  ,
2
de D cos 2.11
rbx  ,
2 cos

rby  fo D  ro .

Let

D cos
 . 2.12
de

Then

D1
rbx  ,
2
  
1  1 1 1 1 1 1 2
  4  , 2.13
R i rax rbx ray rby D fi 1
  
1 1 1 1 1/fi 2/ 1
i  R   ,
rax rbx ray rby 4 1/fi 2/ 1

for the ball-inner-race contact, and

D1
rbx  ,
2
  
1  1 1 1 1 1 1 2
  4 , 2.14
R o rax rbx ray rby D fo 1
  
1 1 1 1 1/fo 2/ 1
o  R   ,
rax rbx ray rby 4 1/fo 2/ 1

for the ball-outer-race contact.

8 Mathematical Problems in Engineering

3. Contact Stress and Deformations

When two elastic solids are brought together under a load, a contact area develops; the shape
and size of which depend on the applied load, the elastic properties of the materials, and
the curvatures of the surfaces. For two ellipsoids in contact the shape of the contact area is
elliptical, with a being the semimajor axis in the y direction transverse direction and b being
the semiminor axis in the x direction direction of motion.
The elliptical eccentricity parameter, k, is dened as

a
k . 3.1
b

From Harris 5, k can be written in terms of the curvature dierence, , and the elliptical
integrals of the rst and second kinds K and E, as

%
2K E1 
Jk  , 3.2
E1 

where
 /2
  1/2
1
K 1 1 2 sin2 d,
0 k
 /2
  1/2 3.3
1
E 1 1 2 sin
2
d.
0 k

A one-point iteration method which has been used successfully in the past 15 is used, where

kn 1  Jkn . 3.4

When the ellipticity parameter, k, the elliptic integrals of the rst and second kinds, K and E,
respectively, the normal applied load, Q, Poissons ratio, , and the modulus of elasticity, E,
of the contacting solids, are known, we can write the semimajor and semiminor axes of the
contact ellipse and the maximum deformation at the center of the contact, from the analysis
of Hertz 16, as

 1/3
6k 2 EQR
a ,
E
3.5
 1/3
6EQR
b ,
kE

 2 1/3
9 Q
K , 3.6
2ER kE
Mathematical Problems in Engineering 9

where

2
E    . 3.7
1 a2 /Ea 1 b2 /Eb

4. Static Load Distribution under Eccentric Thrust Load

Methods to calculate distribution of load among the balls and rollers of rolling bearings
statically loaded can be found in various papers 5, 17. The methods have been limited to,
at most, three degrees of freedom in loading and demand the solution of a simultaneous
nonlinear system of algebraic equations for higher degrees of freedom. Solution of such
equations generally necessitates the use of a digital computer. In certain cases, howeverfor
example, applications with pure radial, pure thrust, or radial and thrust loading with nominal
clearancethe simplied methods will probably provide suciently accurate calculational
results.
Having dened a simple analytical expression for the deformation in terms of load in
the previous section, it is possible to consider how the bearing load is distributed among the
rolling elements. Most rolling-element bearing applications involve steady-state rotation of
either the inner or outer race or both; however, the speeds of rotation are usually not so great
as to cause ball or roller centrifugal forces or gyroscopic moments of signicant magnitudes.
In analyzing the loading distribution on the rolling elements, it is usually satisfactory to
ignore these eects in most applications. In this section the load-deection relationships for
ball bearings are given, along with a specic load distribution consisting of an eccentric thrust
load of statically loaded rolling elements.

4.1. Load-Deection Relationships for Ball Bearings

From 3.6 it can be seen that for a given ball-raceway contact point loading

Q  K 3/2 , 4.1

where

%
 2ER
K  kE . 4.2
9K3

The total normal approach between two raceways under load separated by a rolling element
is the sum of the approaches between the rolling element and each raceway. Hence

n  i o . 4.3
10 Mathematical Problems in Engineering

1  0

2 3
j1
j
2

j

j
3
de

Figure 7: Ball angular positions in the radial plane that is perpendicular to the bearings axis of rotation;
 2/Z, j  2/Zj 1.

Therefore,

 3/2
1
Kn  , 4.4
1/Ki 2/3 1/Ko 2/3

Q  Kn n3/2 . 4.5

4.2. Ball Bearings under Eccentric Thrust Load

Let a ball bearing with a number of balls, Z, symmetrically distributed about a pitch
circle according to Figure 7, be subjected to an eccentric thrust load. Then, a relative axial
displacement, a , and a relative angular displacement, , between the inner and outer ring
raceways may be expected. Let  0 be the angular position of the maximum loaded ball.
Figure 8 shows the initial and nal curvature centers positions at angular position
, before and after loading, whereas the centers of curvature of the raceway grooves are
xed with respect to the corresponding raceway. If a and are known, then the total axial
displacement, t , at angular position , is given by

t  a Ri cos , 4.6

where

de 
Ri  fi 0.5 D cos f 4.7
2

expresses the locus of the centers of the inner ring raceway groove curvature radii.
Mathematical Problems in Engineering 11

a Ri cos

Initial position, inner

raceway groove
curvature center

Final position,
A inner raceway groove A Pd /2
curvature center
f s  A n

Outer raceway groove

curvature center xed

Figure 8: Initial and nal curvature centers positions at angular position , with and without applied load.

Also,

max t 0  a Ri . 4.8

From 4.6 and 4.8, one may develop the following relationship:


1
t  max 1 1 cos 4.9
2

in which
 
1 a
 1 . 4.10
2 Ri

The extension of the loading zone is dened by

 
a
l  cos1 . 4.11
Ri

From Figure 8,
 
 A Pd /2
 cos1 , 4.12
A n

t  A n  sin A sin f . 4.13

From 2.5 and 4.12, the total normal approach between two raceways at angular
position , after the thrust load has been applied, can be written as
 
 cos f
n  A 1 . 4.14
cos
12 Mathematical Problems in Engineering

From Figure 8 and 4.14 it can be determined that s, the distance between the centers
of the curvature of the inner and outer ring raceway grooves at any rolling-element position
, is given by

 cos f
s  A n  A . 4.15
cos

From 4.6, 4.13, and 4.14, yields, for  j ,


sin j f
a Ri cos j A  0, j  1, . . . , Z. 4.16
cos j

From 4.5, and 4.14 one yields, for  j,

 3/2
cos f
Qj  Knj A 3/2
1 , j  1, . . . , Z. 4.17
cos j

If the external thrust load, Fa , is applied at a point distant e from the bearings axis of
rotation, then for static equilibrium to exist


Z
Fa  Qj sin j , 4.18
j1

1 Z
M  eFa  dcj Qj sin j cos j , 4.19
2 j1

where dcj de D cos j .

Substitution of 4.17 into 4.18 yields

 3/2

Z cos f
Fa A 3/2
Knj sin j 1  0. 4.20
j1
cos j

Similarly,

 3/2
A3/2 
Z cos f
eFa Knj dcj cos j sin j 1  0. 4.21
2 j1 cos j

Equations 4.16, 4.20, and 4.21 are Z 2 simultaneous nonlinear equations with
unknowns a , , and j , j  1, . . . , Z. Since Knj and dcj are functions of nal contact angle, j ,
the equations must be solved iteratively to yield an exact solution for a , , and j .
Mathematical Problems in Engineering 13

5. Numerical Results
A numerical method the Newton-Rhapson method was chosen to solve the simultaneous
nonlinear equations 4.16, 4.20, and 4.21. Choosing the rolling bearing, input must be
given the geometric parameters di , do , D, Z, ri , and ro , in accordance with Figures 3 and 5,
and the elastic properties Ea , Eb , a , and b . Next, the following parameters must be obtained:
fi , fo , B, A, j j  1, . . . , Z, E , de , Pd , f ,and Ri .
The interest here is to observe the behavior of an angular-contact ball bearing under
a known thrust load which is to be applied statically to a variable distance lever arm or
eccentricity, e, from the geometric bearing center line. Then, given a thrust load and the
initial estimates for a , , and j , j  1, . . . , Z, for each distance e, varying from zero up to
a given maximum eccentricity, the values 1/R|i , 1/R|o , i , o , ki , ko , Ki , Ko , Ei , Eo , Ki , Ko ,
and Kn are calculated for each ball, according to previous sections, and new values for a , ,
and j are obtained. The new j values are compared with old ones, and if the dierence is
greater than a minimal error, then new values for 1/R|i , 1/R|o , i , o , ki , ko , Ki , Ko , Ei , Eo , Ki ,
Ko , and Kn are calculated for each ball, and again new values for a , , and j are obtained.
If the dierence is lesser than the error then a new value for e is taken. If e is the last valid
value, then a new thrust load value is acquired and the procedure is repeated up to the last
valid thrust load value, when the program ends.
To show an application of the theory developed in this work, a numerical example
is presented here. It was chosen the 218 angular-contact ball bearing that was also used by
Harris5. Thus, the results generated here can be compared to a certain degree with Harris
results. The input data for this rolling bearing were the following:
inner raceway diameter: di  0.10279 m,
outer raceway diameter: do  0.14773 m,
ball diameter: D  0.02223 m,
ball number: Z  16,
inner groove radius: ri  0.01163 m,
outer groove radius: ro  0.01163 m,
modulus of elasticity for both balls and races: E  2.075 1011 N/m2 ,
poissons ratio for both balls and races:  0.3.

The remaining parameters have been calculated yielding:

inner race conformity: fi  0.523166891587944,
outer race conformity: fo  0.523166891587944,
total conformity ratio: B  0.046333783175888,
initial curvature centers distance: A=0.00103 m,
eective elastic modulus: E  228021978021.978 N/m2 ,
angular spacing between rolling elements:  22.5 ,
angular position of rolling elements: j  22.5 j 1, j  1, . . . , 16,
bearing pitch diameter: de  0.12526 m,
diametral clearance: Pd  0.00048 m,
free-contact angle: f  39.915616407992260 ,
radius of locus of inner raceway groove curvature centers: Ri  0.063025 m.
14 Mathematical Problems in Engineering

Eccentric thrust load-218 angular-contact ball bearing

0.25 Fa  17800 N

Relative angular misalignment between

inner and outer ring raceways,  
M  Fa e Nm

0.2

0.15

0.1

0.05

0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 9: Relative angular misalignment, , for 17,800 N thrust load, as a function of the Moment, M.

For each thrust load value, the initial estimates for a , , and j were the following:
axial deection: a  105 m,
misalignment angle:  102 rd,
contact angle: j  1.1f , j  1, . . . , 16.

5.1. Numerical Results for a 17,800 N Thrust Load

Since it is the qualitative behavior of solutions that is the interest, the results are presented
here in graphical form.
Initially, for comparative purposes with the Harris work, a specic thrust load Fa 
17, 800 N was chosen to be applied, and the following graphical results are presented as
functions of the moment, M  Fa e:
i relative angular displacement, Figure 9,
ii partial axial displacement, Ri cos Figure 10,
iii relative axial displacement, a Figure 11,
iv total relative axial deection, t Figure 12,
v loading zone, l Figure 13,
vi distance between loci of inner and outer raceway groove curvature centers, s
Figure 14,
vii maximum elastic compression at the ball/inner-race contact, i Figure 15,
viii maximum elastic compression at the ball/outer-race contact, o Figure 16,
ix total normal ball deection, n Figure 17,
x ball-raceway normal load, Q Figure 18,
xi contact angle, Figure 19,
xii semimajor axis of the ball/inner-race contact area, ai Figure 20,
Mathematical Problems in Engineering 15

104
Eccentric thrust load-218 angular-contact ball bearing
1

and outer rings due to the relative angular

Relative axial displacement between inner
Fa  17800 N
0
M  Fa e Nm

misalignment , Ri cos m

 22.5
 45
 67.5
 90

 112.5  135
 157.5

 180

1
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 10: Partial axial deection, Ri cos , for each ball and 17,800 N thrust load, as a function of the
Moment, M.

105
Eccentric thrust load-218 angular-contact ball bearing
4
inner and outer ring raceways, a m

Fa  17800 N
Relative axial displacement between

3
M  Fa e Nm

4
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 11: Axial deection, a , for 17,800 N thrust load, as a function of the Moment, M.

xiii semiminor axis of the ball/inner-race contact area, bi Figure 21,

xiv semimajor axis of the ball/outer-race contact area, ao Figure 22,
xv semiminor axis of the ball/outer-race contact area, bo Figure 23,
xvi elliptical eccentricity parameter for ball/inner-race contact, ki Figure 24,
xvii elliptical eccentricity parameter for ball/outer-race contact, ko Figure 25.

The graphics above, with exception of Figures 9, 11, and 13, show one curve for each
ball angular position.
Figures 9 and 10 show the relative angular misalignment, , and the partial axial
deection for each ball, Ri cos , respectively. It is observed that there is an approximately
linear relationship between the misalignment angle, , and applied moment, M, for moment
16 Mathematical Problems in Engineering

104 Eccentric thrust load-218 angular-contact ball bearing

1
Fa  17800 N 0
 45  22.5
M  Fa e Nm

Total axial deection, t m

 67.5

 90
0
 112.5

 135

 157.5

 180

1
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 12: Total axial deection, t , for 17,800 N thrust load, as a function of the Moment, M.

Eccentric thrust load-218 angular-contact ball bearing

180
Fa  17800 N
M  Fa e Nm  157.5
160
Harris 2001
92.86
Loading zone, l  

140  135

120  112.5

100
 90

80
 67.5 Ricci
53.66
60

100 200 300 400 500 600 700 800 900

Moment, M Nm

Figure 13: Loading zone, l , for 17,800 N thrust load, as a function of the Moment, M.

values ranging from zero up to about 600 Nm, which corresponds to a distance e of
approximately 33.7 mm. Keeping the load constant and increasing the lever arm, e, above
this value, it can be observed a deeper increase in the misalignment angle and, therefore, in
the resultant axial deection, Ri .
From Figure 9 it can be observed that for an applied moment of 900 Nm e  50.6 mm
the angular misalignment can be as high as a quarter of degree.
As already been waited for, from Figure 10 it can be observed that the partial axial
deection is symmetrical with respect to the horizontal axis null displacement and that the
displacement is null for the balls located at  90 .
Figure 11 shows the axial deection, a . It is observed that the axial deection, a ,
is approximately constant for moment values where the relationship between and M is
approximately linear, that is, from zero up to about 600 Nm e  33.7 mm. For higher
moment values the axial deection falls abruptly and becomes negative in the vicinity of
Mathematical Problems in Engineering 17

103 Eccentric thrust load-218 angular-contact ball bearing

1.1

Distance between loci of inner and outer

raceway groove curvature centers, s m
Fa  17800 N 0
1.09
M  Fa e Nm  22.5
1.08  45

1.07  67.5

1.06
1.05
1.04  90
 112.5
1.03
 135
1.02  157.5
A
1.01  180

0 100 200 300 400 500 600 700 800 900

Moment, M Nm

Figure 14: Distance between curvature centers, s, for 17,800 N thrust load, as a function of the Moment, M.

105 Eccentric thrust load-218 angular-contact ball bearing

3.5 F  17800 N
Maximum local elastic compression at the

a 0
M  Fa e Nm  22.5
3
ball/inner-race contact, i m

2.5  45
 67.5
2  90
 112.5
1.5

1
 135
0.5  157.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 15: Maximum normal elastic compression at the ball/inner race contact, i , for 17,800 N thrust load,
as a function of the Moment, M.

800 Nm e  44.9 mm. The deeper increase in due to the increase in the lever arm forces
the decrease of a to preserve the force and moment static balances.
Figure 12 shows the total axial deection, t . It can be observed that the total axial
deection, t , is the axial deection, a , in two situations: under centric thrust load e  0,
where all balls have the same axial deection 3.6011095400455 105 m, and under eccentric
thrust load for balls located at  90 . Increasing from zero the lever arm, an almost
linear increase decrease in the total axial deection is observed for the balls whose angular
positions satisfy || < 90 || > 90 . This relation is approximately linear up to vicinity of
M  600 Nm when the ball located at  180 occurs to be unloaded, that is, t   180   0
for M  588.9687 Nm e  3.3088 102 m.
From Figure 12 it is observed that for eccentricity of about 50 mm the total axial
deection of the most heavily loaded ball can reach one tenth of millimeter.
18 Mathematical Problems in Engineering

105 Eccentric thrust load-218 angular-contact ball bearing

3.5 F  17800 N

Maximum local elastic compression at the

a 0
M  Fa e Nm
3  22.5

ball/outer-race contact, o m

2.5  45
 67.5
2  90
 112.5
1.5

1
 135
0.5  157.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 16: Maximum normal elastic compression at the ball/outer-race contact, o , for 17,800 N thrust load,
as a function of the Moment, M.

105 Eccentric thrust load-218 angular-contact ball bearing

7
Fa  17800 N 0
6 M  Fa e Nm  22.5
Total ball deection, n m

5  45
 67.5
 90
4
 112.5

2
 135
1  157.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 17: Total ball deection, n , for 17,800 N thrust load, as a function of the Moment, M.

The Figure 13 shows the loading zone, l . The increase of the moment above 587.4 Nm
or lever arm above 3.3 102 m causes the decrease of the loading zone from initial value
l  180 , with the successive unloading of the balls pairs located at  157.5 M 
609.448 Nm,  135 M  661.1407 Nm,  112.5 M  729.9584 Nm,  90
M  803.9741 Nm, and  67.5 M  873.7125 Nm, respectively. Going ahead cause the
unloading of the balls pair located at 45o . However, it is not advisable to go beyond M 
900 Nm, once the radial displacements between curvature centers start to acquire micrometer
order values and they cannot more be disregarded.
Figure 13 shows a substantial dierence between results found in this work and those
found by Harris. While Harris found a loading zone of 92.86o p. 252 for an eccentricity of
50.8 mm, this work found a loading zone of 53.66o . Considering the last result as reference,
this represents an error of 73% in the loading angle, meaning that Harris calculation has
underestimated the eect of the moment M.
Mathematical Problems in Engineering 19

Eccentric thrust load-218 angular-contact ball bearing

10000
Fa  17800 N
9000 M  F e Nm Ricci

Normal applied load, Q N

a
94.45 N
8000
Harris 2001 0
7000
58.78 N
6000
 112.5  22.5
5000  157.5
 90
4000  45
 135
3000
 67.5
2000
1000
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 18: Normal ball load, Q, for 17,800 N thrust load, as a function of the Moment, M.

Eccentric thrust load-218 angular-contact ball bearing

46
Fa  17800 N Ricci
0 44.32
44 M  Fa e Nm
 67.5  45  22.5

42
Contact angle,  

40
Harris 2001
 90 41.6
38  112.5
f
36  135

34  157.5

32  180

30
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 19: Contact Angle, , for 17,800 N thrust load, as a function of the Moment, M.

Figure 14 shows the distance between loci of inner and outer raceway groove
curvature centers, s. It can be observed that the distance, s, under centric thrust load e  0,
is the same for all balls 1.053468971830 103 m. Increasing from zero the lever arm, an
almost linear increase decrease in the distance, s, is observed for the balls whose angular
positions satisfy || < 90 || > 90 . This relation is approximately linear up to vicinity of
M  600 Nm when the ball located at  180 occurs to be unloaded, that is, s  180   A
for M  588.9687 Nm e  3.3088 102 m.
The increase of the moment above 588.9687 Nm or lever arm above 3.3088 102 m
causes the decrease of the loading zone, as already explained, with the successive unloading
of the ball pairs. At the points where the unloading occurs it is observed that the distance s
falls below of the distance between centers of curvature, A, for the unloaded bearing.
Figures 15 and 16 show the maximum normal elastic compressions at the ball/inner-
race and ball/outer-race contacts, i and o , respectively. It can be observed that i and o ,
under centric thrust load e  0, are the same for all balls 1.18852986717367 105 m for i
20 Mathematical Problems in Engineering

103 Eccentric thrust load-218 angular-contact ball bearing

3.5 Fa  17800 N
0

Semimajor axis of the ball/inner-race

M  Fa e Nm  22.5
3  45
 67.5
 90

contact area, ai m

2.5

1.5  112.5

 135
1
 157.5
0.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 20: Semimajor axis of the ball/inner-race contact area, ai , for 17,800 N thrust load, as a function of
the Moment, M.

104 Eccentric thrust load-218 angular-contact ball bearing

5
Fa  17800 N 0
Semiminor axis of the ball/inner-race

4.5
M  Fa e Nm  22.5
4  45
 67.5
contact area, bi m

3.5
 90
3
2.5
2
 112.5
1.5  135

1  157.5

0.5  180

0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 21: Semiminor axis of the ball/inner-race contact area, bi , for 17,800 N thrust load, as a function of
the Moment, M.

and 1.15836731583185105 m for o  and that the deformation for the maximum loaded ball,
in both cases, can reach values as high as 36 m for moment about 900 Nm.
Figure 17 shows the total normal ball deection, n , that can be obtained by summing
the maximum normal elastic compressions on the inner and outer races, i and o , or by
subtracting A from s, once n  s A > 0 also. It can be observed that n , under centric
thrust load e  0, is the same for all balls 2.3468971830055 105 m and that the total
normal elastic deformation for the maximum loaded ball can reach values as high as 70 m
for moment about 900 Nm.
Figure 18 shows the normal ball load, Q. It can be observed that the normal ball
load, Q, under centric thrust load e  0, is the same for all balls 1, 681.663561507042 N.
Increasing from zero the lever arm, an almost linear increase decrease in the normal ball
Mathematical Problems in Engineering 21

103 Eccentric thrust load-218 angular-contact ball bearing

3.5 Fa  17800 N
0

Semimajor axis of the ball/outer-race

M  Fa e Nm
3  22.5
 45
 67.5

contact area, ao m

2.5
 90

1.5  112.5

1  135

 157.5
0.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 22: Semimajor axis of the ball/outer-race contact area, ao , for 17,800 N thrust load, as a function of
the Moment, M.

104 Eccentric thrust load-218 angular-contact ball bearing

5
Fa  17800 N
Semiminor axis of the ball/outer-race

4.5 0
M  Fa e Nm  22.5
4  45
 67.5
 90
contact area, bo m

3.5
3
2.5
2  112.5

1.5  135

1  157.5

0.5
 180
0
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 23: Semiminor axis of the ball/outer-race contact area, bo , for 17,800 N thrust load, as a function of
the Moment, M.

load is observed for the balls whose angular positions satisfy || < 90 || > 90 . This
relation is approximately linear up to vicinity of M  600 Nm when the ball located at 
180 occurs to be unloaded, that is, Q  180   0 for M  589.18 Nm e  3.31 102 m.
Figure 18, as well as Figure 13, shows a substantial dierence between results found
in this work and those found by Harris. While Harris found a 5, 878 N magnitude for the
maximum normal ball load p. 252, for an applied load eccentricity of 50.8 mm, this work
found a 9, 445 N maximum normal ball load. This represents an error of 62.2% in the
normal load, meaning that the Harris calculation has underestimated the normal load for
the maximum loaded ball.
Figure 19 shows the contact angle, . It can be observed that the contact angle, , under
centric thrust load e  0, is the same for all balls 41.417986227161386o . Increasing from
zero the lever arm, an almost linear increase decrease in the contact angle is observed for
22 Mathematical Problems in Engineering

Eccentric thrust load-218 angular-contact ball bearing

8.2

Elliptical eccentricity parameter for ball/

Fa  17800 N  180
8.19
M  Fa e Nm
8.18  157.5
Unload
 135

inner-race contact, ki
8.17  112.5

8.16
8.15
8.14
 90
8.13  67.5
 45
8.12
 22.5
8.11 0

8.1
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 24: Elliptical eccentricity parameter for ball/inner-race contact, ki , for 17,800 N thrust load, as a
function of the Moment, M.

Eccentric thrust load-218 angular-contact ball bearing

6.93
Fa  17800 N 0
Elliptical eccentricity parameter for ball/

6.925  22.5
M  Fa e Nm  45
 67.5
6.92
outer-race contact, ko

6.915
6.91
 90
6.905  112.5

6.9
 135
6.895 Unload

6.89  157.5

6.885  180

6.88
0 100 200 300 400 500 600 700 800 900
Moment, M Nm

Figure 25: Elliptical eccentricity parameter for ball/outer-race contact, ko , for 17,800 N thrust load, as a
function of the Moment, M.

the balls whose angular positions satisfy || < 90 || > 90 . This relation is approximately
linear up to vicinity of M  600 Nm when the ball located at  180 occurs to be unloaded,
that is,   180   f for M  589.18 Nm e  33.1 mm.
Figure 19, as well as Figure 18 and Figure 13, shows a substantial dierence between
results found in this work and to those found by Harris. While Harris has assumed a contact
angle magnitude of 41.6o for all balls p. 252, under a 50.8 mm applied load eccentricity,
contact angles ranging from 44.31727851159821o to 16.16919216282055o were found in this
work while were varied from  0 to 180o , respectively. This represents errors between
6.1% and 157.3% in the contact angles determination, meaning that the Harris calculation
has underestimated strongly overestimated the contact angles for balls located at angular
positions satisfying || < 45 || > 45 .
Mathematical Problems in Engineering 23

104
Eccentric thrust load-218 angular-contact ball bearing
1

Partial axial deection for the maximum

0.9  2 104 N
 1.8 104 N
0.8
 1.6 104 N

loaded ball, Ri m

0.7
 1.4 104 N
0.6  1.2 104 N

0.5  1 104 N
 0.8 104 N
0.4  0.6 104 N
0.3  0.4 104 N
 0.2 104 N
0.2
Fa  0 N
0.1
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 26: Partial axial displacement for the maximum loaded ball, Ri , as a function of lever arm, e.

Figures 20 and 22 show the semimajor axes of the ball/inner-race and ball/outer-race
contact areas, ai and ao , respectively. It can be observed that ai and ao , under centric thrust
load e  0, are the same for all balls 2.069901480072 mm for ai and 2.025827993682 mm for
ao  and that the major axes for the maximum loaded ball, in both cases, can reach values as
high as 7.4 mm for moment about 900 Nm.
Figures 21 and 23 show the semiminor axes of the ball/inner-race and ball/outer-
race contact areas, bi and bo , respectively. It can be observed that bi and bo , under
centric thrust load e  0, are the same for all balls 0.254108993896064 mm for bi and
0.293013306181356 mm for bo  and that the major axes for the maximum loaded ball, in both
cases, can reach values as high as 0.9 mm for moment about 900 Nm.
Figures 24 and 25 show the elliptical eccentricity parameters for ball/inner-race and
ball/outer-race contact, ki and ko , respectively. It can be observed that ki and ko , under centric
thrust load e  0, are the same for all balls 8.1457 for ki and 6.9138 for ko . Increasing from
zero the lever arm, an almost linear increase decrease in the parameter ki is observed for
the balls whose angular positions satisfy || > 90 || < 90 , and an almost linear increase
decrease in the parameter ko is observed for the balls whose angular positions satisfy || <
90 || > 90 . These relations are approximately linear up to vicinity of M  600 Nm when
the ball located at  180 occurs to be unloaded. It can be observed that ki   180  
8.1631 and ko   180   6.9024 when M  588.9687 Nm e  3.3088 102 m.
The increase of the moment above 588.9687 Nm or lever arm above 3.3088 102 m,
causes the successive unloading of the ball pairs. At the points where the unloading occurs
the values of the parameters ki and ko remain roughly equal to those indicated in the
preceding paragraph for  180 . So, it can be observed that the contact ellipse of the
inner race is slightly more eccentric than that of the contact ellipse of the outer race. For
M  900 Nm, for example, while ki varies numerically from 8.11, for the most heavily loaded
ball, to 8.37, for the minimum loaded ball, ko varies from 6.83 to 6.37, respectively.

5.2. Numerical Results for Thrust Load Ranging from 0 up to 20,000 N

Graphics for various thrust loads also are shown. The following graphics present curves for
thrust loads ranging from 0 up to 20, 000 N as functions of lever arm, e:
24 Mathematical Problems in Engineering

105
Eccentric thrust load-218 angular-contact ball bearing
5

 1.6 104 N

Relative axial deection, a m

4
 1.8 104 N
 2 104 N
3

1
 1.2 104 N
 1 104 N  1.4 104 N
0  0.8 104 N
 0.6 104 N
1  0.4 104 N Fa  0 N
 0.2 104 N
2
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 27: Relative axial displacement, a , as a function of lever arm, e.

104
Eccentric thrust load-218 angular-contact ball bearing
1 4  2 10 N
Total axial deection for the maximum

0.9  1.6 104 N

 1.8 104 N
0.8  1 104 N
loaded ball, t   0 m

 0.4 104 N  1.4 104 N

0.7
0.6  0.8 104 N

0.5
0.4
0.3
0.2
0.1  1.2 104 N
Fa  0 N
 0.2 104 N  0.6 104 N
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 28: Total relative axial deection for the maximum loaded ball, t   0, as a function of lever
arm, e.

i partial axial displacement for the maximum loaded ball, Ri Figure 26,
ii relative axial displacement, a Figure 27,
iii total relative axial deection for the maximum loaded ball, t   0 Figure 28,
iv total relative axial deection for the minimum loaded ball, t   180  Figure 29,
v loading zone, l Figure 30,
vi distance between loci of inner and outer raceway groove curvature centers for the
maximum loaded ball, s  0 Figure 31,
vii distance between loci of inner and outer raceway groove curvature centers for the
minimum loaded ball, s  180  Figure 32,
viii total normal ball deection for the maximum loaded ball, n   0 Figure 33,
Mathematical Problems in Engineering 25

105
Eccentric thrust load-218 angular-contact ball bearing
4
 2 104 N

Total axial deection for the minimum

3  1.8 104 N

loaded ball, t   180  m

 1.6 104 N
2

1
Fa  0 N

0
 0.2 104 N
1  1.2 104 N
 1 104 N

2  1.4 104 N  0.8 104 N

 0.6 104 N
3
 0.4 104 N
4
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 29: Total relative axial deection for the minimum loaded ball, t   180 , as a function of lever
arm, e.

Eccentric thrust load-218 angular-contact ball bearing

180

 2 104 N Fa  0 N
160
Loading zone, l  

140
 0.2 104 N
120

100

80

60

0.034 0.036 0.038 0.04 0.042 0.044 0.046 0.048 0.05

Distance from the bearings axis of rotation, e m

Figure 30: Loading zone, l , as a function of lever arm, e.

ix total normal ball deection for the minimum loaded ball, n   180  Figure 34,
x ball-raceway normal load for the maximum loaded ball, Q  0 Figure 35,
xi ball-raceway normal load for the minimum loaded ball, Q  180  Figure 36,
xii contact angle for the maximum loaded ball,   0 Figure 37,
xiii contact angle for the minimum loaded ball,   180  Figure 38,
xiv semimajor axis of the ball/inner-race contact area for the maximum loaded ball,
ai   0 Figure 39,
xv semiminor axis of the ball/inner-race contact area for the maximum loaded ball,
bi   0 Figure 40,
xvi semimajor axis of the ball/outer-race contact area for the maximum loaded ball,
ao   0 Figure 41,
26 Mathematical Problems in Engineering

103
Eccentric thrust load-218 angular-contact ball bearing
1.1

for the maximum loaded ball, s  0 m

Distance between the curvature centers
 2 104 N
1.09
 1.6 104 N
 1.2 104 N  1.8 104 N
1.08
 1.4 104 N
 1 104 N
1.07

1.06

1.05

1.04

1.03
4  0.4 104 N  0.6 104 N  0.8 104 N
Fa  0 N  0.2 10 N

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 31: Distance between loci of inner and outer raceway groove curvature centers for the maximum
loaded ball, s  0, as a function of lever arm, e.

103 Eccentric thrust load-218 angular-contact ball bearing

1.06
for the minimum loaded ball, s  180  m
Distance between the curvature centers

 1 104 N
1.055
 1.2 104 N

1.05  1.4 104 N

 1.6 104 N
1.045  1.8 104 N
 2 104 N
1.04

1.035
 0.6 104 N
1.03
 0.4 104 N
 0.8 104 N
1.025  0.2 104 N
Fa  0 N

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

Distance from the bearings axis of rotation, e m

Figure 32: Distance between loci of inner and outer raceway groove curvature centers for the minimum
loaded ball, s  180 , as a function of lever arm, e.

xvii semiminor axis of the ball/outer-race contact area for the maximum loaded ball,
bo   0 Figure 42,
xviii semimajor axis of the ball/inner-race contact area for the minimum loaded ball,
ai   180  Figure 43,
xix semiminor axis of the ball/inner-race contact area for the minimum loaded ball,
bi   180  Figure 44,
xx semimajor axis of the ball/inner-race contact area for the minimum loaded ball,
ao   180  Figure 45,
xxi semiminor axis of the ball/inner-race contact area for the minimum loaded ball,
bo   180  Figure 46,
Mathematical Problems in Engineering 27

105
Eccentric thrust load-218 angular-contact ball bearing
5
 2 104 N

maximum loaded ball, n   0 m

4.5  1.2 104 N

Total normal ball deection for the

 1.8 104 N
 1.4 104N
4  1.6 104 N
3.5
3
2.5
2
1.5
1
4
0.5  0.4 104 N  0.8 10 N
4 Fa  0 N
 0.2 104 N  0.6 104 N  1 10 N
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 33: Total normal ball deection for the maximum loaded ball, n   0, as a function of lever arm, e.

105
Eccentric thrust load-218 angular-contact ball bearing
2.5
 1 104 N
minimum loaded ball, n   180  m
Total normal ball deection for the

 1.2 104 N
2  1.4 104 N

 1.6 104 N
1.5
 1.8 104 N

1  2 104 N

0.5
 0.2 104 N  0.6 104 N
Fa  0 N
 0.4 104 N  0.8 104 N
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 34: Total normal ball deection for the minimum loaded ball, n   180 , as a function of lever
arm, e.

xxii elliptical eccentricity parameter of the ball/inner-race contact area for the
maximum loaded ball, ki   0 Figure 47,
xxiii elliptical eccentricity parameter of the ball/outer-race contact area for the
maximum loaded ball, ko   0 Figure 48,
xxiv elliptical eccentricity parameter of the ball/inner-race contact area for the minimum
loaded ball, ki   180  Figure 49,
xxv elliptical eccentricity parameter of the ball/outer-race contact area for the minimum
loaded ball, ko   180  Figure 50.

Due to the size quite extensive of the paper comments about the gures will be omitted
from now on.
28 Mathematical Problems in Engineering

Eccentric thrust load-218 angular-contact ball bearing

4500

Ball-raceway normal load for the maximum

 0.2 104 N  1.4 104 N  1.8 104 N
4000  0.4 104 N  1.2 104 N  1.6 104 N
 0.6 104 N  1 10 N
4

3500

loaded ball, Q  0 N

 0.8 104 N

3000

2500
 2 104 N
2000

1500

1000

500
Fa  0 N
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Distance from the bearings axis of rotation, e m
Figure 35: Ball-raceway normal load for the maximum loaded ball, Q  0, as a function of lever arm, e.

Eccentric thrust load-218 angular-contact ball bearing

2000
Ball-raceway normal load for the minimum

 2 104 N
1800
 1.8 104 N
loaded ball, Q  180  N

1600
 1.6 104 N
1400
 1.4 104 N
1200  1.2 104 N
1000  1 104 N

800  0.8 104 N

 0.6 104 N
600
 0.4 104 N Fa  0 N
400
 0.2 104 N
200
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Distance from the bearings axis of rotation, e m
Figure 36: Ball-raceway normal load for the minimum loaded ball, Q  180 , as a function of lever arm,
e.

Eccentric thrust load-218 angular-contact ball bearing

43
 2 104 N
 1.8 104 N
Contact angle for the maximum

 1.4 10 N
4
42.5
 1.6 104 N
loaded ball, Q  0  

 1.2 104 N
42  1 104 N

41.5

41

40.5
Fa  0 N
 0.2 104 N  0.4 104 N  0.6 104 N  0.8 104 N
40

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

Distance from the bearings axis of rotation, e m
Figure 37: Contact angle for the maximum loaded ball,   0, as a function of lever arm, e.
Mathematical Problems in Engineering 29

Eccentric thrust load-218 angular-contact ball bearing

41.4  2 104 N

 1.8 104 N
41.2

Contact angle for the minimum

loaded ball, Q  180   
 1.6 104 N
41
 1.4 104 N
40.8  1.2 104 N

40.6  1 104 N

40.4

40.2
 0.6 104 N Fa  0 N
 0.2 104 N
40  0.4 104 N  0.8 10 N
4

39.8
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Distance from the bearings axis of rotation, e m

Figure 38: Contact angle for the minimum loaded ball,   180 , as a function of lever arm, e.

103 Eccentric thrust load-218 angular-contact ball bearing

contact area for the maximum loaded ball,

3.5  1.8 104 N  2 104 N

 1.4 104 N
Semimajor axis of the ball/inner-race

 1.6 10 N
4
 1.2 104 N
3
 1 104 N

2.5
ai   0 m

1.5

 0.8 104 N
1  0.6 104 N
 0.4 104 N
0.5
 0.2 10 N
4
Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 39: Semimajor axis of the ball/inner-race contact area for the maximum loaded ball, ai   0, as a
function of lever arm, e.

6. Conclusions
The importance of this work lies in the fact that it uses a new procedure for gettinnng
numerically, accurately, and quickly the static load distribution of a single-row, angular-
contact ball bearings, subjected to a known thrust load which is applied to a variable
distance from the geometric bearing center line. Precise applications, as for example,
space applications, require a precise determination of the static loading. Models available
in literature are approximate and often are not compatible with the desired degree of
accuracy. This work can be extended to determine the loading on high-speed bearings where
centrifugal and gyroscopic forces are not discarded. The results of this work can be used in
the accurate determination of the friction torque of the ball bearings, under any operating
condition of temperature and speed.
30 Mathematical Problems in Engineering

104
Eccentric thrust load-218 angular-contact ball bearing
5

contact area for the maximum loaded ball,

4  1.6 10 N

Semiminor axis of the ball/inner-race

4.5  1.4 104 N  2 104 N

 1.2 104 N  1.8 104 N

4
3.5  1 104 N

bi   0 m
3
2.5
2
1.5
1  0.6 104 N  0.8 104 N
 0.4 104 N
0.5  0.2 104 N Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 40: Semiminor axis of the ball/inner-race contact area for the maximum loaded ball, bi   0, as a
function of lever arm, e.

103 Eccentric thrust load-218 angular-contact ball bearing

contact area for the maximum loaded ball,

3.5  1.8 104 N  2 104 N

 1.4 104 N
Semimajor axis of the ball/outer-race

 1.6 104N
4
 1.2 10 N
3
 1 104 N
2.5
ao   0 m

1.5

1
 0.8 104 N
 0.6 104 N
0.5  0.4 104 N
Fa  0 N
 0.2 104 N
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 41: Semimajor axis of the ball/outer-race contact area for the maximum loaded ball, ao   0, as a
function of lever arm, e.

Symbols
a: Semimajor axis of the projected contact, m
A: Distance between raceway groove curvature centers, m
b: Semiminor axis of the projected contact, m
B: fo fi 1, total curvature
d: Raceway diameter, m
da : Bearing outer diameter, m
db : Bearing inner diameter, m
dc : Contact diameter, m
de : Bearing pitch diameter, m
D: Ball diameter, m
Mathematical Problems in Engineering 31

104
Eccentric thrust load-218 angular-contact ball bearing
5

contact area for the maximum loaded ball,

4  2 10 N
 1.6 104 N

Semiminor axis of the ball/outer-race

4.5  1.4 104 N  1.8 104 N
4
 1.2 10 N
4
 1 104 N
3.5

bo   0 m
3
2.5
2
1.5  0.8 104 N
 0.6 104 N
1  0.4 104 N
0.5  0.2 104 N Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 42: Semiminor axis of the ball/outer-race contact area for the maximum loaded ball, bo 0, as a
function of lever arm, e.

103
Eccentric thrust load-218 angular-contact ball bearing
2.5
contact area for the minimum loaded ball,

 2 104 N
Semimajor axis of the ball/inner-race

 1.8 104 N
2
 1.6 104 N
 1.4 104 N
 1.2 104 N
ai   180 

1.5

 1 104 N
1

 0.2 104 N
 0.4 104 N
0.5
 0.6 104 N
Fa  0 N  0.8 104 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 43: Semimajor axis of the ball/inner-race contact area for the minimum loaded ball, ai   180 ,
as a function of lever arm, e.

e: Eccentricity of loading, m
E: Modulus of elasticity, N/m2
E : Eective elastic modulus, N/m2
E: Elliptic integral of second kind
f: Raceway groove radius D
F: Applied load, N
k: a/b
K: Load-deection factor, N/m3/2
K: Elliptic integral of rst kind
M: eFa
Pd : Diametral clearance, m
Pe : Free endplay, m
32 Mathematical Problems in Engineering

104 Eccentric thrust load-218 angular-contact ball bearing

 2 104 N

contact area for the minimum loaded ball,

2.5  1.8 104 N

Semiminor axis of the ball/inner-race

 1.6 104 N
 1.4 104 N
 1.2 104 N
2
 1 104 N

bi   180  m
1.5

1
 0.4 104 N
 0.2 104 N  0.6 104 N
0.5  0.8 104 N

Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 44: Semiminor axis of the ball/inner-race contact area for the minimum loaded ball, bi   180 ,
as a function of lever arm, e.

103 Eccentric thrust load-218 angular-contact ball bearing

contact area for the minimum loaded ball,

 2 104 N
Semimajor axis of the ball/outer-race

2  1.8 104 N
 1.6 104 N
 1.4 104 N
1.5  1.2 104 N
ao   180 

 1 104 N
1

 0.4 104 N
0.5  0.2 104 N  0.6 104 N
 0.8 104 N
Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 45: Semimajor axis of the ball/outer-race contact area for the minimum loaded ball, ao   180 ,
as a function of lever arm, e.

Q: Ball-raceway normal load, N

r: Raceway groove curvature radius, solids curvature radius, m
s: Distance between loci of inner and outer raceway groove curvature centers, m
R: Curvature radius, radius of locus of raceway groove curvature centers, m
Z: Number of rolling elements
: Contact angle, rad, o
f : Free-contact angle, rad, o
: D cos /de
: Curvature dierence
: Deection or contact deformation, m
: Angular spacing between rolling elements, rad, o
: Load distribution factor
Mathematical Problems in Engineering 33

104 Eccentric thrust load-218 angular-contact ball bearing

contact area for the minimum loaded ball,

3  2 104 N

Semiminor axis of the ball/outer-race

 1.8 104 N
 1.6 104 N
 1.4 104 N
 1.2 104 N

bo   180 
2
 1 104 N

1  0.4 104 N
 0.2 104 N  0.6 104 N
 0.8 104 N

Fa  0 N

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 46: Semiminor axis of the ball/outer-race contact area for the minimum loaded ball, bo   180 ,
as a function of lever arm, e.

Eccentric thrust load-218 angular-contact ball bearing

8.165
 0.2 104 N
the maximum loaded ball, ki   0
the ball/inner-race contact area for

 0.4 104 N  0.6 104 N

Elliptical eccentricity parameter of

8.16  0.8 104 N Fa  0 N

 1 104 N
8.155

8.15

8.145

8.14

8.135  1.8 104 N

 2 104 N  1.4 104 N
8.13
 1.6 104 N  1.2 104 N
8.125
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 47: Elliptical eccentricity parameter of the ball/inner-race contact area for the maximum loaded
ball, ki   0, as a function of lever arm, e.

: Bearing misalignment angle, rad, o

: Poissons ratio
: Auxiliary angle
: Azimuth angle, rad, .

Subscripts:
a refers to solid a or axial direction.
b refers to solid b.
x, y refers to coordinate system.
i refers to inner raceway.
34 Mathematical Problems in Engineering

Eccentric thrust load-218 angular-contact ball bearing

 1.8 104 N
 2 104 N

the maximum loaded ball, ko   0

 1.6 104 N

the ball/outer-race contact area for

Elliptical eccentricity parameter of
 1.4 10 N
4

6.92  1.2 104 N

6.915

6.91

6.905  0.8 104 N  1 104 N

 0.4 104 N
 0.2 104 N  0.6 104 N

Fa  0 N
6.9
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 48: Elliptical eccentricity parameter of the ball/outer-race contact area for the maximum loaded
ball, ko   0, as a function of lever arm, e.

Eccentric thrust load-218 angular-contact ball bearing

8.165
 0.6 104 N  0.8 104 N
the minimum loaded ball, ki   180 
the ball/inner-race contact area for
Elliptical eccentricity parameter of

 0.2 104 N
 0.4 104 N Fa  0 N
8.16
 1 104 N

 1.2 104 N
8.155
 1.4 104 N

 1.6 104 N
8.15  1.8 104 N

 2 104 N

8.145
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 49: Elliptical eccentricity parameter of the ball/inner-race contact area for the minimum loaded ball,
ki   180 , as a function of lever arm, e.

j refers to rolling-element position.

n refers to direction collinear with normal load, integer number.
o refers to outer raceway.
t refers to total axial deformation.

Acknowledgments
The author thanks the nancial support provided by the Brazilian Institute for Space Research
INPE, the Brazilian Scientic and Technological Development Council CNPq, The State of
Mathematical Problems in Engineering 35

Eccentric thrust load-218 angular-contact ball bearing

6.915

for the minimum loaded ball, ko   180 

 2 104 N

Elliptical eccentricity parameter of

 1.8 104 N

the ball/outer-race contact area

 1.6 104 N

6.91  1.4 104 N

 1.2 104 N

 1 104 N

6.905
 0.2 104 N  0.4 104 N

Fa  0 N  0.6 104 N
 0.8 104 N
6.9
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Distance from the bearings axis of rotation, e m

Figure 50: Elliptical eccentricity parameter of the ball/outer-race contact area for the minimum loaded ball,
ko   180 , as a function of lever arm, e.

Sao Paulo Research FAPESP, and The Coordination of Development of Higher-Level Sta
CAPES. And the author thanks also Jose Pelogia da Silva for doing Figures 2 up to 7.

References
1 R. Stribeck, Ball bearings for various loads, Transactions of the ASME, vol. 29, pp. 420463, 1907.
2 H. Sjovall, The load distribution within ball and roller bearings under given external radial and axial
load, Teknisk Tidskrift, Mekaniska, p. 9, 1933.
3 A. Jones, Analysis of Stresses and Deections, New Departure Engineering Data, Bristol, Conn, USA,
1946.
4 J. Rumbarger, Thrust bearings with eccentric loads, Machine Design, 1962.
5 T. Harris, Rolling Bearing Analysis, John Wiley & Sons, New York, NY, USA, 4th edition, 2001.
6 M. C. Ricci, Ball bearings subjected to a variable eccentric thrust load, in Booklet of Abstracts of the
8th Brazilian Conference on Dynamics, Control and Applications (DINCON 09), Bauru, Brazil, May 2009.
7 M. C. Ricci, Ball bearings subjected to a variable eccentric thrust load, in Proceedings of the 8th
Brazilian Conference on Dynamics, Control and Applications (DINCON 09), Bauru, Brazil, May 2009.
8 M. C. Ricci, Internal loading distribution in statically loaded ball bearings, in Program and Abstracts
of the 1st International Conference on Computational Contact Mechanics (ICCCM 09), pp. 2122, Lecce,
Italy, September 2009.
9 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial and thrust load, including the eects of temperature and t, in Proceedings of World Academy
of Science, Engineering and Technology (WCSET 09), vol. 57, Amsterdam, The Netherlands, September
2009.
10 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial and thrust load, in Book of Abstracts of the 6th International Congress of Croatian Society of
Mechanics (ICCSM 09), p. 163, Dubrovnik, Croatia, September 2009.
11 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial and thrust load, in Proceedings of the 6th International Congress of Croatian Society of Mechanics
(ICCSM 09), Dubrovnik, Croatia, September 2009.
12 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial, thrust, and moment load, in Abstracts of the 60th International Astronautical Congress, Daejeon,
South Korea, October 2009.
13 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial, thrust, and moment load, in Proceedings of the 60th International Astronautical Congress, p. 163,
Daejeon, South Korea, October 2009.
36 Mathematical Problems in Engineering

14 M. C. Ricci, Internal loading distribution in statically loaded ball bearings subjected to a combined
radial, thrust, and moment load, including the eects of temperature and t, in Proceedings of the 11th
Pan-American Congress of Applied Mechanics, Foz do Iguacu, Brazil, January 2010.
15 B. J. Hamrock and W. J. Anderson, Arched-outer-race ball-bearing considering centrifugal forces,
NASA Report TN D-6765, NASA, Moett Field, Calif, USA, 1972.
16 H. Hertz, On the contact of elastic solids and on hardness, in: D.E. Jones, G.A. Schott (Eds.), Miscellaneous
Papers by Heinrich Hertz, Macmillan, London, 1896.
17 B. J. Hamrock and W. J. Anderson, Rolling-element bearings, NASA Report RP 1105, NASA, Moett
Field, Calif, USA, 1983.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 384071, 16 pages
doi:10.1155/2009/384071

Research Article
The Determination of the Velocities after
Impact for the Constrained Bar Problem

Andre Fenili,1 Luiz Carlos Gadelha de Souza,2 and Bernd Schafer3

1
Center for Engineering, Modeling and Applied Social Sciences (CECS),
Federal University of ABC (UFABC), Aerospace Engineering, Av. dos Estados,
5001. Bloco B/Sala 936, 09210-580 Santo Andre, SP, Brazil
2
National Institute for Space ResearchINPE, Av. dos Astronautas,
1758, 12201-940 Sao Jose dos Campos, SP, Brazil
3
German Aerospace CenterDLR, Institute of Robotics and Mechatronics, 82234 Wessling, Germany

Correspondence should be addressed to Luiz Carlos Gadelha de Souza, [email protected]

Received 27 May 2009; Accepted 23 November 2009

Recommended by Antonio Prado

A simple mathematical model for a constrained robotic manipulator is investigated. Besides the
fact that this model is relatively simple, all the features present in more complex problems are
similar to the ones analyzed here. The fully plastic impact is considered in this paper. Expressions
for the velocities of the colliding bodies after impact are developed. These expressions are
important in the numerical integration of the governing equations of motion when one must
exchange the set of unconstrained equations for the set of constrained equation. The theory
presented in this work can be applied to problems in which robots have to follow some prescribed
patterns or trajectories when in contact with the environment. It can also de applied to problems
in which robotic manipulators must handle payloads.

Copyright q 2009 Andre Fenili et al. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
There are several ways to deal with the problem of interaction between bodies. Impact
dynamics and continuous contact between bodies can both be included in the mathematical
model of the constrained problem, or just one of these eects can be considered. It depends,
obviously, on the characteristics of the studied problem.
The investigations about the contact between bodies include at least two dierent
kind of analysis 1: one associated with the beginning of contact and one associated with its
termination. In the rst analysis, the distance between the bodies must be checked in order to
know when contact occurs; in the second analysis, once the contact is established, the reaction
2 Mathematical Problems in Engineering

normal; compression force between the bodies must be checked. In the second analysis,
contact nishes when the contact force is equal to zero.
One of the hardest parts in the study of contact problems involves the dierent models
that must be developed for contact and noncontact situations and the switching between
these models when integrating the equations of motion 2, 3. The unconstrained problem
and the constrained problem do not have the same number of degrees of freedom. Dynamic
systems when constrained have less degrees of freedom than when unconstrained.
The transition between constrained and unconstrained motions is sometimes called
contact including impact and sometimes called just impact mostly when the bodies
separate after the collision. When contact occurs, the new velocities of the bodies involved
must be known in order to generate the initial conditions to the second part constrained
problem of the numerical integration. In the constrained problem, the concept of coecient
of restitution is very important 4.

2. Geometric Model of the System and Governing Equations of Motion

The problem discussed here is depicted in Figure 1. According to this gure, in a part of
its trajectory, the free end of the bar moves along the constraint represented by the mass
named mw . All the movements occur in the horizontal plane. When contact occurs, impact
and bouncing are also allowed to occur.
The mass in which the rigid bar is pivoted ms  oscillates when excited by the
movement of the bar free and constrained. In the axis Z, passing through the connection
between the bar and ms perpendicular to the paper sheet, there is a prescribed moment,
M , acting to turn the bar.
The dashed lines represent the position of the masses in which the springs and
dampers are free of forces. The dotted line represents the position from which one starts
to count the angular displacement, .
In physical terms, this system may represent a robot with a translational joint and a
rotational joint; mw can be thought as an obstructing wall on the robots trajectory or some
object this robot must handle or interact with, and M can be thought as an external torque
provided by a dc motor.
According to 5, the constrained governing equations of motion for this system are
given by

mb ms ys cs ys ks ys mb dAcmb 2 sin mb dAcmb cos FN  0,

mw yw cw yw kw yw FN  0, 2.1


Ib,cm mb dAcmb
2
mb dAcmb ys cos FN  cos  M ,

and the constraint condition is given by

d ys yw  sin  0, 2.2
Mathematical Problems in Engineering 3

kw cw

mw
yw

mb

d

ys
A
X

ks cs

Figure 1: Oscillating constrained bar.

vwy  yw
mw P2
vP 2,x
vP 2,y
w /2
ys

P1

y mb , Ib ,cm

x
ys

dAcmb

ys

vsy  ys

ms

Figure 2: Velocities.
4 Mathematical Problems in Engineering

F  kw F
w  F  cw  P
x, P
y
x
B

y
B x
P
y
P

x
P

x
A 4
M

y
A

4
M x
A
x
C

y
C
F  ks F
s  F  cs  A
 x, A
 y

Figure 3: Impulses.

where Ib,cm represents the bar moment of inertia around its center of mass, mb represents
the mass of the bar, dAcmb represents the distance from A to the cm of the bar, cw represents
the damping coecient of mw , cs represents the damping coecient associated with mass
ms , kw represents the stiness coecient of mass mw , ks represents the stiness coecient
associated with ms , and FN represents the amplitude of the normal force. It is assumed the
there are no friction forces involved and  represents the total length of the bar.
Equations 2.1 are the equations of motion for ys , yw , and . Equation 2.2 is an
additional relationship between the generalized coordinates ys , and yw when contact
occurs. Equations from 2.1 to 2.2 provide four equations and four unknowns ys , , yw ,
and FN  considering the constrained problem and three equations and three unknowns ys ,
, and yw  considering the unconstrained problem. In the unconstrained case, 2.2 does not
apply and FN  0.

3. The Contact Case

In contact, for this problem, there is the loss of one degree of freedom. In other words, one
of the variables is dependent on all the others. The best choice is the elimination of the
generalized coordinate yw , which is not always present into the system represented by the
Mathematical Problems in Engineering 5

4 0.6
3 0.5
2
0.4

dyw /dt m/s

dys /dt m/s

1
0.3
0
0.2
1
0.1
2
3 0

4 0.1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

a b
30 12

25 10
d/dt rad/s

20 8

FN N 6
15
4
10
2
5
0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 4: ys , yw , , and FN considering kw  10 Nm.

oscillating bar 6. The new set of equations 5 is given by

1 
ys a1 cs cw ys a1 ks kw ys a1 cw  cos a1 kw  sin a1 a2 2 sin
a1 mt a3 cos
2

mb cw  2 dAcmb cos3 mb kw  2 dAcmb sin cos2

mb kw ddAcmb cos2 mw cs mb dAcmb cw ys cos2

a2 cos
mw ks mb dAcmb kw ys cos2 a1 kw d  M ,
a1 mt a3 cos2
1 
cw mt  a2 cos2 kw mt  a2  sin cos kw dmt  a2  cos
a1 mt a3 cos2
a2 mb dAcmb mw mt  a2 2 sin cos kw mt  a2  a2 ks ys cos
mt
cw mt  a2  a2 cs ys cos  M .
a1 mt a3 cos2
3.1
6 Mathematical Problems in Engineering

4 0.6
3 0.4
2
0.2
dys /dt m/s

dyw /dt m/s

1
0
0
0.2
1
2 0.4

3 0.6
4 0.8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s
a b
30 14
12
25
10
d/dt rad/s

20
FN N
8
15 6

10 4
2
5
0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 5: ys , yw , , and FN considering kw  400 Nm.

The fully plastic impact case is considered here for the calculation of the velocities
immediately after contact. Separation will take place when the normal force is zero.
As soon as these two variables are known, the remaining variable, yw , is also known
through 2.2. Equations 3.1 represent, respectively, the time behavior of the generalized
coordinates ys and during the contact condition. In 5, an analytical expression to the
reaction force, FN , is also presented.

4. The Determination of the Velocities after Contact (Impact)

The equations for the impact are formulated for point P see Figure 2 for the representation
of the velocities of the three bodies where, for sake of clarity, it is distinguished between
Point P1 belonging to the wall and point P2 belonging to the bar. Figure 3 shows the free body
diagram for the three rigid bodies indicating not the forces at the points of connection or
contact but rather indicating the equivalent linear impulses due to impact. All these quantities
are marked with an overhead symbol hat, for example, Px , which is the linear impulse of
the equivalent force Px . The physical dimension is the same as the linear momentum, that is,
N s, except for the angular impulse M4 whose unit is Nm s.
Mathematical Problems in Engineering 7

4 0.5
3 0.4
0.3
2
0.2

dyw /dt m/s

dys /dt m/s

1 0.1
0 0
1 0.1
0.2
2
0.3
3 0.4
4 0.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

a b
25
20

20
15
d/dt rad/s

15
FN N 10
10
5
5
0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 6: ys , yw , , and FN considering kw  1000 Nm.

For each of the three rigid bodies, we can formulate now the linear impulse/linear
momentum equations in the two directions x and y. Additionally, for the rotating bodies,
we have the equivalent angular impulse/angular momentum equation in z-direction,
formulated w.r.t. to the respective centre of mass.
To better distinguish between velocities right before and right after impact, they are
denoted with superscripts after and before. Their two components in x- and y-
directions are indicated by corresponding subscripts x and y.
And, to be more general, it is also allowed initially for the rigid bodies with masses
ms and mw to rotate as well. The respective angular velocities therefore will be denoted by
with appropriate indices. Later, this additional degree of freedom will be kinematically
constrained.
For the wall, it is obtained that

mw vwy
mw vwy  Py Fw ,

mw vwx
mw vwx  Px Bx , 4.1
w 

Iw w
Iw w  By Py  sin .
2
8 Mathematical Problems in Engineering

2 3.5
3
1.5
2.5

dyw /dt m/s

dys /dt m/s

1
2
0.5 1.5
1
0
0.5
0.5
0
1 0.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

a b
8
7 7
6 6
d/dt rad/s

5 5
FN N 4
4
3
3
2
2
1
1
0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 7: ys , yw , , and FN considering M  5Nm.

For the bar, it is obtained b  that

mb vby
mb vby  y Py ,
A

mb vbx
mb vbx  x Px ,
A
4.2
4 A
Ib,cm b Ib,cm b  M  x dAcmb sin A
 y dAcmb cos

Px  dAcmb  sin Py  dAcmb  cos .

And, nally, for the lower rigid body with mass ms , it is obtained that

ms vsy
ms vsy y C
 A  y Fs ,

ms vsx
ms vsx x C
 A x , 4.3

4 A
Is s Is s  M  x bsy C
 y bsx ,
Mathematical Problems in Engineering 9

5 5
4
4
3

dyw /dt m/s

dys /dt m/s

3
2
1 2
0
1
1
0
2
3 1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

a b
12
12
10 10
d/dt rad/s

8 8
FN N
6 6

4 4

2 2

0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 8: ys , yw , , and FN considering M  10 Nm.

 x and Fs are going through the center of mass. The geometric
assuming that the directions of C
quantities bsx and bsy , not shown in Figure 3, denote the distances of the respective linear
impulses measured from the center of mass.
These equations simplied if the following assumptions are made.

1 The external two linear impulses Fw and Fs , and the angular impulse M
4 are small
compared with the internal impulses; therefore, they can be neglected.
2 The rotational motion of both, the wall and the lower rigid body, is omitted;
therefore, one has w  0 and s  0.
3 The wall is allowed to move only in the vertical direction, as well as the lower rigid
body; therefore, vwx  0 and vsx  0.
4 The contact surface between the lower rigid body and the left or right vertical
guiding surface not shown in the gures is assumed ideally smooth; therefore,
 y  0.
C
5 The contact zone between the free end of the bar and the wall surface is also
assumed ideally smooth; therefore, Px  0. Otherwise, if this surface is rough, we
have to account for an additional velocity relationship, for example, given by the
denition of the coecient of restitution in x-direction.
10 Mathematical Problems in Engineering

8 6
6 5
4
4

dyw /dt m/s

dys /dt m/s

2
3
0
2
2
1
4
6 0

8 1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

a b
18 20
16
14 15
d/dt rad/s

12
10 FN N 10
8
6 5
4
2 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time s Time s

c d

Figure 9: ys , yw , , and FN considering M  20 Nm.

Applying these assumptions, the following set of equations is obtained:

mw vwy
mw vwy  Py , 4.4

Bx  Px  0, 4.5

2
By  sin Py , 4.6
w

mb vby
mb vby  y Py ,
A 4.7

mb vbx
mb vbx  x Px  A
A x , 4.8

 x dAcmb sin A
Ib,cm b Ib,cm b  A  y dAcmb cos Py  dAcmb  cos , 4.9

ms vsy
ms vsy y ,
 A 4.10

C x ,
x  A 4.11

bsy  0. 4.12
Mathematical Problems in Engineering 11

In order to calculate the velocities at the point of impact, P , only 4.4 and 4.7 to
4.10 are of interest. Additionally, it is needed to establish some kinematic relationships. For
the bar center of mass, one has

vb  vs b rASb
 T
 b dAcmb sin , vsy b dAcmb cos 4.13
 T
 vbx , vby ,

where the length of the vector rAcmb is just |rAcmb |  dAcmb . Equation 4.13 is valid for the
velocity right before and after impact. For the free end of the bar, it is obtained equivalently

vP 1  vs b rAP1
 T
 b  sin , vsy b  cos 4.14
 T
 vP 1x , vP 1y

with |rAP1 |  . In the same way as 4.13, equation 4.14 is valid for the velocity right before
impact and right after. During impact, one has the additional equation, which relates the
velocities before and after impact at point P , in the direction normal to the contact surface,
that is, in y-direction:

vP 1y vP 2y
y  4.15
vP 1y vP 2y

with

vP 1y  vsy

b  cos ,

vP 1y  vsy

b  cos ,
4.16
vP 2y  vwy

 yw ,

vP 2y  vwy

 yw .

In the following, it is assumed that there is a fully plastic impact, that is, the impacting
bodies maintain steady contact as far as the contact force is repulsive otherwise, they will
separate. This leaves y  0, and hence

vP 1y  vP 2y 4.17

or

vwy  vsy b  cos  vby

b  dAcmb  cos . 4.18
12 Mathematical Problems in Engineering

With these equations, it is possible to calculate all the velocities right after impact,
given the velocities before impact. Additionally, but not needed here, it is also possible
to calculate the appropriate linear impulses. To summarize, one has the following eight


equations to determine all the ve velocities right after impact vsy , vbx , vby
, vwy , b , as well
as the impulses A  y , Py :
x , A

mw vwy
mw vwy  Py , 4.19

mb vby
mb vby  y Py ,
A 4.20

mb vbx
mb vbx x ,
A 4.21

 x dAcmb sin A
Ib,cm b Ib,cm b  A  y dAcmb cos Py  dAcmb  cos , 4.22

ms vsy
ms vsy y ,
 A 4.23

vbx  b dAcmb sin , 4.24

vby  vsy b dAcmb cos , 4.25

vwy  vsy b  cos  vby

b  dAcmb  cos . 4.26

 y is simply obtained from 4.23 or by adding

Initially, all the impulses are obtained. A
the two 4.19 and 4.20, giving


 y  ms vsy
A
ms vsy
 mb vby
mb vby
mw vwy
mw vwy . 4.27

 x also goes simply with 4.21,

A

 x  mb v mb v
A 4.28
bx bx

and Py is simply obtained directly from 4.19 or by adding 4.20 and 4.23

 
Py  mw vwy

mw vwy
 mb vby
mb vby
ms vsy
ms vsy . 4.29

Comparing 4.27 with 4.29, it is observed that both equations yield the same result
for the linear momenta before and after impact. To determine now the velocities right after
impact, one can rely on 4.22, 4.24, 4.25, 4.26, and 4.27 or 4.29, which is the same.
Mathematical Problems in Engineering 13

Replacing vbx , vby and vwy , one arrives at the two equations for the unknown velocities vsy

and b :

ms mb mw vsy  ms vsy mb vby mw vwy b mb dAcmb mw  cos ,
 
Ib,cm mb dAcmb
2
sin2 mw  dAcmb cos2 b
4.30

 vsy ms dAcmb mw  dAcmb  cos Ib,cm b ms vsy

dAcmb cos

mb vbx dAcmb sin mw vwy  dAcmb  cos .

And with vby
 vsy b dAcmb cos , and vbx

 b dAcmb sin , these equations can
nally be expressed by means of the independent velocities, vsy
, vwy , and b , right before
impact:

ms mb mw vsy  ms mb vsy mw vwy mb b dAcmb cos b mb dAcmb mw  cos .
 
Ib,cm mb dAcmb
2
sin2 mw  dAcmb cos2 b


 vsy ms dAcmb mw  dAcmb  cos Ib,cm mb dAcmb
2
sin2 b

ms vsy dAcmb cos mw vwy  dAcmb  cos
4.31

With the abbreviations

mtot  ms mb mw ,

Itot  Ib,cm mb dAcmb

2
sin2 mw  dAcmb cos2 ,

r1  ms mb vsy mw vwy mb b dAcmb cos ,
 4.32

r2  ms vsy dAcmb cos mw vwy  dAcmb  cos Ib,cm mb dAcmb
2
sin2 b ,

1  mb dAcmb mw  cos ,

2  ms dAcmb mw  dAcmb  cos ,

one nally obtains

r1 Itot r2 1
vsy  ,
1 2 mtot Itot
4.33
r1 2 r2 mtot
b  .
1 2 mtot Itot
14 Mathematical Problems in Engineering

The denominator of these two equations then is written as

 
1 2 mtot Itot  ms mb dAcmb
2
mtot Ib,cm mb mw  2dAcmb cos2 dAcmb
2

4.34
ms mw  2 cos2 m2b dAcmb
2
sin2 .

In order to check 4.33, one case is investigated; that is, for  90 , we should
maintain the simple translational impact between the combined rigid body consisting of the
two masses ms and mb and the wall with mass mw . For the fully plastic impact, one then
obtains from 4.33 with 1  0 and 2  0:

r1 ms mb vsy mw vwy
vsy   90    ,
mtot mtot
4.35
r2 Ib,cm mb dAcmb
2
b   90    b ,
Itot Itot

where the rst equation for the translational motion coincides with the result governed from
simple impact of two rigid bodies.

5. Numerical Results
The values for the parameters used in the numerical simulations that follow are presented
in Tables 1 and 2. The time step considered in the integration of the governing equations of
motion is kept constant and equal to 0.0001 s. The fourth-order Runge-Kutta is the numerical
integrator used. Two dierent classes of simulation are investigated.
The constant torque with dierent amplitudes was chosen because it is the simplest
one, and in order to make the bar rotate always in the same direction and fulll 360 . Any
other kind of excitation e.g., like a sinusoidal one with maximum amplitude of 180o , for
instance can be chosen without problem. In the simulation runs, the motion of the bar starts
always in its horizontal position to the right, that is, with  0 .
The very beginning of contact is considered here as a fully plastic impact with impact
time t 0 and with e  0, where e represents the coecient of restitution. Contact nishes
when FN  0. No friction or contact is considered, up to this point of the investigation,
between ms and the guide it slides through or between mb and mw .

5.1. Considering Different Values of kw

When rst contact takes place, mw is at rest. The second contact only shown here for the
simulations varying M  will happen with mw presenting some velocity. The bar is able to
develop many turns and, in fact, there are possibilities for it to reach many contact conditions
as the time evolves.
According to Figures 4, 5, 6, 7, 8, and 9, the amplitude of FN jumps at the beginning
of contact, from zero no contact to a value associated with the impact force between the
bodies. The contact force evolves with time according to the system states and properties.
The value of FN at the instant of impact does not necessarily represent the biggest value for
Mathematical Problems in Engineering 15

Table 1: Numerical values considered in the numerical simulations for dierent values of kw .

Parameter Value Unity

mb 2.00 Kg
ms 5.00 Kg
mw 10.00 Kg
ks 5.00 Nm
10.00
kw 400.00 Nm
1000.00
cs 7.00 Ns/m
cw 1.00 Ns/m
 1.00 m
d 0.60 m
dAcmb 0.50 m
M 10.00 Nm
Ib,cm 0.1667 Kg/m2

Table 2: Numerical values considered in the numerical simulations for dierent values of M .

Parameter Value Unity

mb 2.00 Kg
ms 1.00 Kg
mw 1.00 Kg
ks 400.00 Nm
kw 5.00 Nm
cs 7.00 Ns/m
cw 7.00 Ns/m
 1.00 m
d 0.60 m
dAcmb 0.50 m
5
M 10 Nm
20
Ib,cm 0.1667 Kg/m2

the contact force, as can be seen in these gures. A sudden change in velocity, when collision
takes place, can be veried clearly in these gures.

5.2. Considering Different Values of M

Table 2 shows numerical values considered in the numerical simulations for dierent values
of M .

6. Conclusions
To conclude, it is important to say that the time step used in the numerical integration and
the choice of the integrator are very important aspects to be considered. New numerical
16 Mathematical Problems in Engineering

integrators can be tested in the course of this investigation and results compared to the ones
presented here.
An important consideration not to be forgotten when dealing with problems
presenting some sort of constraint is that more than one set of governing equations of motion
must be integrated to cover all the system dynamics. The set of equations that governs the
system dynamics when the constraint condition is active is dierent from the one that governs
the unconstrained movement of the system. One of these sets is always generating the states
for the other.
In this context, the determination of the velocities after contact impact is very
important. The velocity expressions presented in 4.33 are the necessary corrections one
must do when considering the fully plastic impact case. If this correction is not taken into
consideration in the numerical integration of the governing equations, the system will gain
energy after impact, which is not true.
It is important to realize also that the number of degrees of freedom involved
changes from one set of equations to the other. The necessity for changing from one set
of governing equations to another according to the systems requirements of contact or
noncontact conditions represents a source of integration errors, since the integrator is faced
with singularities.
The problem presented in this paper and the procedures developed for its analysis can
be extended to many other systems and situations including more complex ones. The theory
presented here can be applied to problems in which robots have to follow some prescribed
patterns or trajectories when in contact with the environment like in painting activities, for
instance, or the ROKVISS experiment at DLR.
The next steps are the development of the analytical expressions for the velocities after
impact considering any value for the coecient of restitution and the inclusion of friction
forces between ms and the left and right vertical guiding surfaces; and between the free end
of the bar and mw .

References
1 F. Pfeier and C. Glocker, Multibody Dynamics with Unilateral Contacts, Wiley Series in Nonlinear
Science, John Wiley & Sons, New York, NY, USA, 1996.
2 C. Lanczos, The Variational Principles of Mechanics, Mathematical Expositions, no. 4, University of
Toronto Press, Toronto, Canada, 4th edition, 1970.
3 E. T. Whittaker, A Treatise on the Analytical Dynamics of Particles and Rigid Bodies, Cambridge University
Press, Cambridge, UK, 1965.
4 N. A. Fufaev and J. I. Neimark, Dynamics of Nonholonomic Systems, American Mathematical Society,
1972.
5 A. Fenili, L. C. G. Souza, and B. Schafer, A mathematical model to investigate contact dynamics in
constrained robots, in Proceedings of the 6th International Symposium on Dynamic Problems of Mechanics
(DINAME 05), D. A. Rade and V. Steen Jr., Eds., Ouro Preto, Brazil, February-March 2005.
6 B. Schafer, B. Rebele, and A. Fenili, Space robotics contact dynamics investigations and numerical
simulations: ROKVISS, in Proceedings of the 15th CISM-IFToMM Symposium on Robot Design, Dynamics
and Control, 2004.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 897570, 11 pages
doi:10.1155/2009/897570

Research Article
Gravitational Capture of Asteroids by Gas Drag

E. Vieira Neto and O. C. Winter

FE-Guaratingueta, Departamento de Matematica, UNESPUniversidade Estadual Paulista,
12.516-410 SP, Brazil

Correspondence should be addressed to E. Vieira Neto, [email protected]

Received 30 July 2009; Revised 6 November 2009; Accepted 1 December 2009

Recommended by Silvia Maria Giuliatti Winter

Several irregular satellites of the giant planets were found in the last years. Their orbital
conguration suggests that these satellites were asteroids captured by the planets. The restricted
three-body problem can explain the dynamics of the capture, but the capture is temporary. It is
necessary some kind of dissipative eect to turn the temporary capture into a permanent one.
In this work we study an asteroid suering a gas drag at an extended atmosphere of a planet to
turn a temporary capture into a permanent one. In the primordial Solar System, gas envelopes
were created around the planet. An asteroid that was gravitationally captured by the planet got its
velocity reduced and could been trapped as an irregular satellite. It is well known that, depending
on the time scale of the gas envelope, an asteroid will spiral and collide with the planet. So, we
simulate the passage of the asteroid in the gas envelope with its density decreasing along the time.
Using this approach, we found eective captures, and have a better understanding of the whole
process. Finally, we conclude that the origin of the irregular satellites cannot be attributed to the
gas drag capture mechanism alone.

Copyright q 2009 E. Vieira Neto and O. C. Winter. This is an open access article distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

1. Introduction
The giant planets of the Solar System have two kinds of satellites, the regular and the irregular
ones 1. This denition was based on the orbits of the satellites. The regular satellites have
their orbits near the equatorial plane, the eccentricity is near zero, and they are close to their
planets. The irregular ones are out of plane, many are retrograde, the eccentricities are above
0.1, and they are far from their planets.
Since the discovery of two new Uranus moons in 1997 by Gladman et al. 2, 98 other
satellites were found. These new satellites, plus the old irregulars satellites, are almost all
retrograde, as we can see in Table 1.
The orbits of the irregular satellites suggest that they were captured by their planets.
That is, they were formed elsewhere in the Solar System like any other asteroid, and later
they might have had a close encounter with a planet and could have been captured.
2 Mathematical Problems in Engineering

Table 1: Number of irregular satellites of the giant planets 3.

Planet Prograde Retrograde Total

Jupiter 6 49 55
Saturn 8 27 35
Uranus 1 8 9
Neptune 3 4 7

Using the dynamics of the three-body problem it is possible to explain the gravitational
capture of these satellites. But the gravitational capture is temporary see, for example, 4.
There are works that studied the capture time 5 or the directions of capture 6 in the case
of the three-body problem. Thus, it is necessary some kind of additional eect to turn the
temporary capture into a permanent one.
In the literature we can nd several mechanisms which were proposed in order to
turn the capture permanent. It could be a pull-down mechanism due to the mass growth
of the planet 79, or a capture through n-body interactions 1012. In this work we will
explore the gas drag mechanism. In the early history of the solar system, in the last stage of
the giant planets formation, a gas envelope was formed around each one of them 13. This
gas envelope could make a ying by asteroid to loose enough energy in order to became
a prisoner of the planet. Other authors already discussed this possibility see, for example,
14. Although the theory and numerical simulation have some success in explaining the
prograde satellites, the retrograde ones are not well explained. This is because the retrograde
satellites feel a stronger headwind which make them to spiral inward faster and collide with
the planet sooner than the prograde ones. This suggests distinct conditions for the gas drag
capture of the prograde and retrograde satellites.
Cuk and Burns 14 studied the case of a prograde satellite capture, using the orbital
and physical parameters of Himalia Jupiters satellite. They used a static surface density gas
envelope with edges to modify the size of the envelope. They vary the position of the edge,
but maintaining the surface density of the gas envelope constant. They have discussed the
eect of a gas envelope with decaying density in their work, but made no simulation. Also in
the work of Cuk and Burns 14, particles were started inside the gas envelope, mainly using
Himalias initial conditions, and integrated backwards in time.
In the work of Pollack et al. 15 they simulate the formation of a giant planet and
characterized the phases: 1 accretion of solid material; 2 accretion of solid material and
gas with a small rate; 3 runaway gas accretion. They concluded that the overall time scale is
determined by the second phase, and the rst and third phase are very quick. They suppose
that the envelope did not collapse in the phase of runaway gas accretion as long as the solar
nebula could supply the planet with gas rapidly enough. In their work they do not estimate a
time interval for the third phase, but their Figure 1a give us a clue of this time interval and
it seems to last less than a thousand years. In our work we will consider the very last stage,
when there is no more gas in the solar nebula to feed the planetary gas envelope and this
envelope collapses. We suppose the duration of this stage to be about a hundred years and in
this stage the density of the gas envelope will vary until vanish.
In this work we simulate a gas envelope with no edges, but its surface density varies
linearly with time. We started the integration of the asteroid orbit outside the gas envelope in
a proper region as an heliocentric orbit, and integrated it forward in time until it reaches the
gas envelope around the planet. We also point out that this work is in two dimensions, while
the work of Cuk and Burns 14 was three-dimensional.
Mathematical Problems in Engineering 3

360 1
330
300

Jupiter-particle energy
270
240
210

deg
180 0
150
120
90
60
30
0 1
4 4.5 5 5.5 6 6.5 7
Semi-major axis AU

Figure 1: Grid of initial conditions of the asteroid with respect to the Sun. In the plot we can see which
initial conditions that lead to negative two-body energy w.r.t. Jupiter, that is, the asteroid was temporarily
captured by Jupiter. The white region in the middle of the plot are trajectories which lead to a two-body
energy greater than one. For each asteroid initial condition in the a space, Jupiter begins its orbit at
5.2 AU from the Sun with  0 .

In our work we will discuss the very last stage of the gas envelope when it collapses
onto the planet. Thus, the surface density of the gas will change from some accepted initial
value down to zero, vanishing the gas. This conguration is favorable for the retrograde
satellites. Our goal is to investigate how to get permanent gravitational capture for the
retrograde satellites.
We will rst discuss, in Section 2, the possible heliocentric orbital region that could
make an asteroid to be captured by Jupiter. In Section 3 we will expose the gas model and
show the results of our simulations. Finally, we discuss our results in Section 4.

2. Possible Origins for the Captured Asteroids

In order to study the asteroids initial conditions which lead to capture by the planet, we used
the circular planar Sun-Jupiter system, and made a grid of initial conditions with semi-major
axis and true longitude . The others asteroids initial conditions parameters were xed
in such a way that the asteroid had a circular planar orbit with respect the Sun. The semi-
major axis a was varied from 4 to 7 Astronomical Units AU and the true longitude 
from 0 to 360 . For each trajectory in this grid, we measured the two-body energy of the
asteroid relative to Jupiter from time t  0 to time t  105 years and the minimum value was
stored. The two-body energy is not constant in the three-body system and will vary due to
the Suns perturbation. We were interested in the negatives values of the two-body energy,
which congures the temporary gravitational capture 16, since negative value of this energy
corresponds to an osculate elliptical orbit. Other possible option for this study would be the
minimum distance between asteroid and Jupiter and a passage through the Hill sphere, but
the energy was chosen because it takes the position and velocity of the particle with respect
to the planet, and we can avoid a close passage with very high speed which is not interesting
to our experiment.
In the search of proper initial conditions which leads to gravitational capture we did
not consider the gas drag. The idea in this rst moment is just to nd the path for an encounter
with Jupiter.
4 Mathematical Problems in Engineering

AU
0

8
8 6 4 2 0 2 4 6 8
AU

Figure 2: Asteroid initial conditions which lead to negative two-body energy w.r.t. Jupiter, taken from
Figure 1, in polar coordinates.

The numerical integrations we made in this work used an integrator that uses Gauss-
Radau spacings described by E. Everhart17.
The result for this simulation can be seen in Figure 1, where we separated the initial
conditions in the grid that lead to positive and negative two-body energy. The black points
indicate the initial conditions of trajectories which achieved a minimal negative two-body
energy at some point along the integration, while the other color dots correspond to those
trajectories whose minimum energy was positive.
On Figure 2 we plotted only the initial conditions which lead to negative two-body
energy of the asteroid with respect to Jupiter, in polar coordinates. From these two gures we
can see the horseshoe libration zone 18 and the chaotically-unstable region around Jupiters
orbit, which is due to overlap of mean motion resonances 19.
In Figures 1 and 2 we can see the existence of two dense regions with initial conditions
that lead to negative values of the two-body energy. These initial conditions are the ones
that lead the asteroid to pass close enough to the planet and have a temporary gravitational
capture.

3. Dynamics and Numerical Results

In order to turn a temporary capture into a permanent one, we used a drag force proportional
to the velocity of the asteroid relative to the velocity of the gas, the cross section of a spherical
asteroid, and the local density of the gas 20, 21:

F d  Cd R2 rvrel


v rel , 3.1
2
Mathematical Problems in Engineering 5

where Cd is the drag coecient, R is the asteroid radius, vrel is the relative velocity of the
asteroid with respect to the gas, and
 
r
r  0 , 3.2
r0

where r is the gas density at a distance r from the centre of the planet, and is the exponent
that gives how the gas density decay with the distance from the planet. In this work we used
 1, as in Cuk and Burns 14. We also used

0
0  3.3
H0

as the gas density at r0 , with 0 being the gas surface density, and H0 the gas envelope height.
We considered r0  100 Jupiter Radius JR, H0  0.05r0 , as in Cuk and Burns 14. The
asteroidal radius was took as R  6 km. We choose this small radius because we are interested
in the retrograde satellites and this will reduce the decaying eect of the gas. Besides, this is
a representative radius value for the retrograde satellites observed around Jupiter. The gas
velocity rotation around the planet was chosen to be 90% of the Keplerian velocity.
In order to change the density of the gas envelope during the passage of the asteroid
we implemented the following surface density time function:

0 t0 , t < t0


0 t  0 t0  t t0 , t0 t t 1 3.4




0, t > t1 ,

where is the time rate of the decreasing density. An exponential or sigmoid decay law will
change only slightly our results.
The gravitational model used was the general three-body model, that is, the mass of
the asteroid were considered. The density of the asteroid was considered 1.5 g/cm3 , and the
gas density surface considered was 0 t0   103 g/cm2 , as in Cuk and Burns 14.
The gas envelope is formed after the consumption of all the gas from the solar nebula.
The lack of gas to fuel the planet makes the envelope collapses into the planet 20, 22. It is
believed that this collapse is quick and took only some hundreds of years to occur. That is,
the surface density of the gas goes to zero in times of the order of 102 years.

3.1. Asteroids Passage through the Gas Envelope

In order to study the passage of the asteroid through the gas envelope we use the results
from Figure 1 to choose an adequate set of initial conditions. The initial conditions used are
in the same plane of the gas envelope around Jupiter. We xed a time interval of 100 years
t  |t1 t0 | for the gas to vanish. The gas has its density changed according to 3.4, from
0 t0  to zero in a time interval of t  100 years, and after this period there is no more gas
around the planet, 0 t  0 t1   0.
The eect of the change of the surface gas density on the asteroid trajectory will be
analysed by xing the initial orbital conguration of the asteroid and Jupiter, but selecting
dierent values of initial integration time from t  100 years to t  0 year, with steps of 0.1
6 Mathematical Problems in Engineering

year, while time variation of the gas envelope has ever the initial time t0  100 years and
nal time t1  0 year. That is, for each trajectory the asteroid and Jupiter have the same orbital
elements with respect to the Sun, but the orbital integration of the system starts at dierent
times, reaching the gas envelope at dierent surface density values 0 .
During the integration of the trajectory we monitored the relative two-body energy of
the asteroid relative to Jupiter to verify its sign. Initially the asteroid is in heliocentric orbit
and has a positive relative two-body energy with respect to Jupiter. We stoped the integration
if one of the following happens: i the asteroid passes through the gas envelope; ii the
asteroid collides with Jupiter; iii the integration time surpass t  50 years.
In the rst case, to know that the asteroid had passed through the gas envelope around
Jupiter, we measured the relative two-body energy of the asteroid with respect to Jupiter
during the encounter. The two-body energy turn into negative when the asteroid is close to
Jupiter. Then, the relative two-body energy turn to positive again when the asteroid is distant
from the planet.
For the second case, as we do not consider the Galilean satellites in the experiment,
the collision with Jupiter is considered when the distance Jupiter-asteroid is smaller than the
Jupiters radius.
The third case happens if one of the following occurs: a the asteroid is permanently
captured; b the asteroid is deected by the gas envelope; c the asteroid is temporarily
captured.
Case a is considered when the asteroid entered in the gas envelope, that is, its relative
two-body energy was turned into negative, in a time that the surface density of the gas has
a value between 0 t0  e 0. Due to the gas, the asteroid suers a drag and loose energy. But
the gas surface density vanishes, the asteroid stop to loose its energy and, if it lost enough
energy, it became a permanent satellite.
In case b the asteroid never reach a negative relative two-body energy with respect to
Jupiter. In these kind of trajectories the asteroid passes in the outer layers of the gas envelope
and it is deected.
Finally, in case c the asteroid arrives at the planet when the surface density of the gas
is already zero, or very low, and do not loose enough energy to be eectively captured.
We tested four initial conditions for semi-major axis at 6.0 AU, and other four initial
conditions for semi-major axis at 4.5 AU. Each one with a dierent true longitude value  for
the asteroid. These initial conditions were obtained from Figure 1. In Figure 3 we can see the
results of the trajectories of the asteroid which have initial semi-major axis of 6.0 AU. None
of the trajectories was eectively captured by the planet, many trajectories entered the gas
envelope and escaped quickly. Other trajectories never entered the gas envelope as shown in
Figures 3c and 3d, where there is a region in the interval 100, 40 years with no points.
This occurs because the trajectories are deected by the gas. In Figure 3d, the lack of points
for the region 7, 0 happens because the trajectories are with negative two-body energy, but
the gas has low surface density and the trajectories will escape again due to the fact that the
dissipative eect is too weak.
In Figure 4 we show the results for other four initial conditions, but now for a 
4.5 AU. The plots in the Figure show the time spent by the asteroid inside the gas envelope
with negative two-body energy. Only for Figure 4c eective captures occur, as showed
by the blue squares. In the other three cases there are only collisions and passages near the
planet.
We also note in the plots the existence of upward spikes. These spikes occur due to
the trajectories transition between trajectories which just pass through the gas envelope and
Mathematical Problems in Engineering 7

70 70

Time in the envelope years

Time in the envelope years

60 60
50 50
40 40
30 30
20 20
10 10
0 0
100 80 60 40 20 0 100 80 60 40 20 0
Asteroid initial time years Asteroid initial time years

a b

70 70
Time in the envelope years

Time in the envelope years

60 60
50 50
40 40
30 30
20 20
10 10
0 0
100 80 60 40 20 0 100 80 60 40 20 0
Asteroid initial time years Asteroid initial time years

Passage Passage
Collision Collision
c d

Figure 3: Length of time that the asteroid spent inside the gas envelope for each initial time simulation.
Four initial conditions for semi-major axis of 6.0 AU: a  70 ; b  110 ; c  250 ; d  290 .

trajectories that collide with the planet. These transition trajectories spend more time around
the planet making more loops than the normal trajectories.
We studied some trajectories of Figure 4c and show them in Figure 5. The trajectories
where condition i is satised, that is, a passage through the gas envelope, occurs when the
asteroid initiated its trajectory when the gas density is close to zero. In this case the gas is
so sparce that the asteroid almost does not feel it and escape see Figure 5, trajectory d.
Other possibility is when the asteroid initiated its trajectory just after the beginning of the gas
density variation and the gas density is strong enough to make the asteroid to scatter in the
extended atmosphere of the planet see Figure 5, trajectory a.
Condition ii, the collisions, mainly occurs for initial time interval 40, 26.1 years
of Figure 4c. In this time interval the gas is not so dense and allows the asteroid to enter in
the envelope, but it is a trap for the asteroid which has not enough energy to escape from the
gas envelope, then it spirals and collides with the planet see Figure 5, trajectory b.
Condition iii, integration time greater than 50 years, happens in the interval
26.0, 21.6 years of Figure 4c. Figure 5, for trajectory c, shows a trajectory with initial
time at t  25.0 years. The asteroid enters the gas envelope and it is trapped by the gas. This
trajectory has its semi-major axis reduced, the gas vanish and the trajectory is stabilized.
8 Mathematical Problems in Engineering

70 70

Time in the envelope years

Time in the envelope years

60 60
50 50
40 40
30 30
20 20
10 10
0 0
100 80 60 40 20 0 100 80 60 40 20 0
Asteroid initial time years Asteroid initial time years

Passage Passage
Collision Collision
a b

70 70
Time in the envelope years

Time in the envelope years

60 60
50 50
40 40
30 30
20 20
10 10
0 0
100 80 60 40 20 0 100 80 60 40 20 0
Asteroid initial time years Asteroid initial time years

Passage Passage
Collision Collision
Capture

c d

Figure 4: Length of time that the asteroid spent inside the gas envelope for each initial time simulation.
Four initial conditions for semi-major axis of 4.5 AU: a  70 ; b  110 ; c  250 ; d  290 .

3.2. Captured Orbits

In order to understand what happens with the permanently captured orbits, we computed
the average of the semi-major axis and eccentricity of the captured trajectories, presented in
Figure 4c. The average was made in the time interval between 0 and 50 years, that is, after
the gas had vanished.
In Figure 6 we present the behavior of the averaged semi-major axis and eccentricity
after the gas vanishes. For the rst trajectories, the gas is denser at the time the asteroid
arrives, which make its semi-major axis to decrease more, stabilizing the orbit with a semi-
major axis close to 10 Jupiters radius. With a less dense gas at the time of arrival of the
asteroid, the average value of the semi-major axis is close to 320 Jupiters radius.
The average value of the eccentricity is also shown in Figure 6. Most of the trajectories
have eccentricities close to 0.95. The few ones which are less than 0.8 are the ones which are
closer to the planet.
Mathematical Problems in Engineering 9

0.2
a
0.4

0.6

y AU
0.8 a

1
b
1.2
c
1.4 d
d
0.5 0.4 0.3 0.2 0.1 0 0.1 0.2
x AU

Figure 5: Four trajectories of Figure 4c, all in a non rotating planetocentric reference frame with x pointing
to the Vernal point. a Trajectory for gas density at t  80 years. The asteroid scatters in the gas and goes
away. b Trajectory for gas density at t  40 years. The asteroid enters in the gas envelope, but the gas
is strong and makes the asteroid to collide with the planet. c Trajectory for gas density at t  25 years.
The asteroid enters in the gas envelope in a time that the gas is just strong enough to make the asteroid
to reduce its energy and not collide with the planet. After the gas vanishes the capture is permanent. d
Trajectory for gas density at t  18 years. The asteroid has a temporary gravitational capture but the gas
is too sparce and the asteroid goes away.

350 1

300 0.95

250 0.9

200 0.85
a JR

150 0.8

100 0.75

50 0.7

0 0.65
26 25.5 25 24.5 24 23.5 23 22.5 22 21.5
Initial time years

a
e

Figure 6: Averaged values of semi-major axis left side and eccentricityright side in the time interval
between 0 and 50 years for the captures generated by gas drag.

We also pointed that the inclination of the asteroids during all the time after the gas
was vanished for all captured trajectories were retrograde, that is, the inclinations never
changes from i  180 with respect to Jupiter.
This process of capture is not suitable for the prograde capture. In the work of Cuk
and Burns 14 they show that the time needed to capture a Himalia satellite type is between
104 to 106 years. The process described here is rapid and the density of the gas envelope is
10 Mathematical Problems in Engineering

not enough to make a asteroid to lose orbital energy and turn it into a permanent prograde
satellite.
If we compare this and the work of Cuk and Burns 14, using the results of Pollack
et al. 15, the capture of the prograde satellites might have occurred in the second phase
of planetary formation, while the capture of the retrograde satellites might have occurred
only in the third stage. In the work of Pollack et al. 15 they estimate a time between
106 years and 8 106 years in the second phase. Thus, the model of capture with gas drag
show us that the capture of prograde satellites are much more probable than the retrograde
ones. This contradicts the observations. There are more retrograde satellites observed than
prograde ones. Therefore, the origin of the prograde and retrograde irregular satellites cannot
be attributed to the gas drag capture mechanism alone.

4. Conclusions
We found a region of initial conditions, close to Jupiter, which produces temporary
gravitational capture around Jupiter. We took eight sets of initial conditions from this region
and varied the initial time for the integration in a dynamics which includes a gas envelope
with decreasing density along the time to a moment that it vanishes. These simulations gave
us information about what might be the outcomes of the evolution of such system. The
possible outcomes are: the scatter of the asteroid, the temporary gravitational capture, the
collision with Jupiter, and the permanent capture. Seven out of eight sets of initial conditions
did not give eective capture. Only one set presented a few permanent captures due to the
gas drag.
We have shown that the permanent capture of an asteroid in retrograde orbit around
Jupiter due to a vanishing gas envelope is possible. We have the knowledge that this happens
with low probability, but considering the supply of planetesimals available in the early
time of the solar system formation, this kind of event could have happened. Although
comparing with the probability of prograde capture it is less probable. The observations show
more retrograde satellites than prograde ones. This is a problem for the gas drag capture
mechanism as a whole.
It is necessary to look more regions of the space of semi-major axis, eccentricity, and
inclination in order to have enough information about the gravitational capture with gas drag
and also the time for the gas envelope to vanish needs to be tested. In order to implement this
it is necessary to use some Monte Carlo method due to the huge range of free parameters.
The captures studied in this work, which were shown in Section 3.2, are close to Jupiter
compared with the real irregular satellites. Some could cross the Galilean satellites orbits and
collide with them. It is necessary to improve the gas envelope decaying mechanism to obtain
more information of the process, but in this work we tried to proof this concept and further
work is under development.

Acknowledgments
This work was funded by FAPESP and CNPq. These supports are gratefully acknowledged.

References
1 J. B. Pollack, J. A. Burns, and M. E. Tauber, Gas drag in primordial circumplanetary envelopes: a
mechanism for satellite capture, Icarus, vol. 37, no. 3, pp. 587611, 1979.
Mathematical Problems in Engineering 11

2 B. J. Gladman, P. D. Nicholson, J. A. Burns, et al., Discovery of two distant irregular moons of
Uranus, Nature, vol. 392, no. 6679, pp. 897899, 1998.
3 D. Jewitt and N. Haghighipour, Irregular satellites of the planets: products of capture in the early
solar system, Annual Review of Astronomy and Astrophysics, vol. 45, pp. 261295, 2007.
4 K. Tanikawa, Impossibility of the capture of retrograde satellites in the restricted three-body
problem, Celestial Mechanics, vol. 29, no. 4, pp. 367402, 1983.
5 E. Vieira Neto and O. C. Winter, Time analysis for temporary gravitational capture: satellites of
Uranus, Astronomical Journal, vol. 122, no. 1, pp. 440448, 2001.
6 D. S. de Oliveira, O. C. Winter, E. Vieira Neto, and G. de Felipe, Irregular satellites of Jupiter: a study
of the capture direction, Earth, Moon and Planets, vol. 100, no. 3-4, pp. 233239, 2007.
7 T. A. Heppenheimer and C. Porco, New contributions to the problem of capture, Icarus, vol. 30, no.
2, pp. 385401, 1977.
8 E. Vieira Neto, O. C. Winter, and T. Yokoyama, The eect of Jupiters mass growth on satellite capture
Retrograde case, Astronomy and Astrophysics, vol. 414, no. 2, pp. 727734, 2004.
9 E. Vieira Neto, O. C. Winter, and T. Yokoyama, Eect of Jupiters mass growth on satellite capture
the prograde case, Astronomy and Astrophysics, vol. 452, no. 3, pp. 10911097, 2006.
10 G. Colombo and F. A. Franklin, On the formation of the outer satellite groups of Jupiter, Icarus, vol.
15, no. 2, pp. 186189, 1971.
11 C. B. Agnor and D. P. Hamilton, Neptunes capture of its moon Triton in a binary-planet gravitational
encounter, Nature, vol. 441, no. 7090, pp. 192194, 2006.
12 H. S. Gaspar, O. C. Winter, and E. Vieira Neto, Irregular satellites of Jupiter: capturecongurations of
binary asteroids, submitted to Monthly Notices of the Royal Astronomical Society.
13 J. I. Lunine and D. J. Stevenson, Formation of the Galiean satellites in a gaseous nebula, Icarus, vol.
52, pp. 1439, 1982.
14 M. Cuk and J. A. Burns, Gas-drag-assisted capture of Himailas family, Icarus, vol. 167, pp. 369381,
2004.
15 J. B. Pollack, O. Hubickyj, P. Bodenheimer, J. J. Lissauer, M. Podolak, and Y. Greenzweig, Formation
of the giant planets by concurrent accreation of solid and gas, Icarus, vol. 124, pp. 6285, 1996.
16 E. Vieira Neto, O. C. Winter, and C. Melo, The use of the two-body energy to study problems of
escape/capture, in Dynamics of Populations of Planetary Systems, Z. Knezevic and A. Milani, Eds.,
Proceedings IAU Colloquium, no. 197, pp. 439444, Cambridge University Press, Cambridge, UK,
2005.
17 E. Everhart, An ecient integrator that uses Gauss-Radau spacings, in Dynamics of Comets: Their
Origin and Evolution, A. Carusi and G. B. Valsecchi, Eds., vol. 115 of Proceedings of IAU Colloquium, no.
83, pp. 185202, Astrophysics and Space Science Library, Dordrecht, The Netherlands, 1985.
18 C. D. Murray and S. F. Dermott, Solar System Dynamics, Cambridge University Press, Cambridge, UK,
1999.
19 J. Wisdom, The resonance overlap criterion and the onset of stochastic behavior in the restricted
three-body problem, Astronomical Journal, vol. 85, pp. 11221133, 1980.
20 M. Podolak, W. B. Hubbard, and J. B. Pollack, Gaseous acreation and the formation of giant planets,
in Protostars and Planets III, E. H. Levy and J. I. Lunini, Eds., pp. 11091147, University of Arizona
Press, Tucson, Ariz, USA, 1993.
21 I. Adashi, C. Hayashi, and K. Nakazawa, The gas drag eect on the elliptic motion of a solid body
in the primordial nebula, Progress in Theorical Physics, vol. 56, no. 3, pp. 17561771, 1976.
22 S. J. Weidenschilling, Planetesimals from Stardust, in From Stardust to Planetesimals, Y. J. Pendleton
and A. G. G. M. Tielens, Eds., vol. 122 of ASP Conference Series, p. 281, Astronomical Society of the
Pacic, Santa Clara, Calif, USA, 1997.
Hindawi Publishing Corporation
Mathematical Problems in Engineering
Volume 2009, Article ID 859678, 14 pages
doi:10.1155/2009/859678

Research Article
Atmospheric Reentry Dynamics of Conic Objects

J. P. Saldia,1 A. Cimino,1 W. Schulz,1 S. Elaskar,1, 2 and A. Costa1, 2

1
Departmento de Aeronautica, Universidad Nacional de Cordoba, Velez Sarseld 1611,
Cordoba 5000, Argentina
2
CONICET, Rivadavia 1917, Argentina

Correspondence should be addressed to W. Schulz, [email protected]

Received 25 June 2009; Revised 27 September 2009; Accepted 4 November 2009

Recommended by Silvia Maria Giuliatti Winter

One of the key issues in a reentry risk analysis is the calculation of the aerodynamic coecients.
This paper presents a methodology to obtain these coecients and couple it to a code that
computes re-entry trajectories considering six degrees of freedom. To evaluate the dierent ight
conditions encountered during the natural re-entry of conical objects, the Euler Equations for
gasdynamics ows are used. A new scheme TVD Total Variation Diminishing is incorporated to
a nite volume unstructured cell-centred formulation, for application to three-dimensional Euler
ows. Finally, numerical results are obtained for a conical body at dierent attack angles and Mach.
With these results, the calculation of the trajectories during atmospheric re-entry is completed.

Copyright q 2009 J. P. Saldia et al. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

1. Introduction
In the case of natural reentries non-controlled, the orbital evolution of an object can only
be monitored, with no or limited ability to control risks. The time window for reentry of a
satellite is usually provided with a standard error of 10% to 20% of the remaining orbital
lifetime. For the controlled reentries, it is required to simulate the dierent scenarios until the
right window for the mission is found, being that the total or partial disintegration, or the
landing on a safe place.
In this paper, the main objective is to conduct numerical simulations of the supersonic
ow regime on a conical body, thereby using a code developed at the Department of
Aeronautics of the National University of Cordoba, Argentina 1, 2. This code uses the
technique of nite volumes for solving Euler equations. The spatial discretization of the
domain is done through a mesh of unstructured tetrahedral volumes. It has implemented
a new technique for choosing the limiter functions that can reduce the articial viscosity
without the loss of strength robustness of the Total Variation Diminishing schemeTVD 3
6.
2 Mathematical Problems in Engineering

The drag and lift coecients, CD and CL , obtained by the numerical simulation of
compressible ows are used in a code that allows to evaluate the trajectories considering six
degrees of freedom 7. As a result of this research, the trajectories of reentry into the Earths
atmosphere for conical objects having dierent initial ight conditions are presented.

2. Methodology
2.1. Description of the Numerical Scheme for Compressible Flow
The three-dimensional Euler equations can be written as

U
F  0, 2.1
t

where U is the vector of conservative variables, and F is the 3D vectorial ow.

The temporal change of the conservative variables can be expressed as


t lfaces
Un 1  Un F nl Al , 2.2
V ol l1 l

where the ux of the conservative variables F has been replaced by the numerical ux tensor
F . Vol indicates the volume where the integration is performed, nl is the outward normal to
the control surface Al .
Equation 2.2 allows the use of a locally aligned system of coordinates whose unit
vector i coincides with the normal to the face l of the cell, and the unit vectors j and k are
tangential directions. To achieve second-order accuracy, the numerical ux at the interface
between cells l and l 1 in the direction normal to the face l is calculated by 8

fi fi 1 1  m m
fi 1/2  i 1/2 Ki 1/2 , 2.3
2 2 m

m
where fi and fi+1 are the physical uxes normal to the face in each cells, Ki 1/2 is the m-th right
eigenvector, and m i 1/2
is, in the original Harten-Yee scheme 911, dened as
 
 m  m
m
i 1/2  g m
i g m
i 1  i 1/2 m
i 1/2 i 1/2 , 2.4

being

     
S    m  m 2.5
gim  max 0, min m m
i 1/2 i 1/2 , S  i1/2 i1/2
,
2

S  sign m i 1/2 , 2.6
m
gi 1 gi
m
if m 0
i 1/2 /
m
i 1/2  m 2.7

0
i 1/2
if m  0,
i 1/2
Mathematical Problems in Engineering 3

where m i 1/2
is the jump of the conserved variables across the interfaces between cells i and
i 1, m
i1/2
is the m-th eigenvalue of the Jacobian matrix, gim is the limiter function, and S is
the sign function of the corresponding eigenvalue 811. Since the local Riemann problem
is solved with rotated data, the eigensystem is calculated in the locally aligned coordinate
frame.
The limiter function given in 2.5 is known as minmod 36. The minmod selects the
minimum possible value, so that the scheme is TVD. The other end is the limiter function
superbee that ponders the contribution of the high-order ux 3. The only implementation of
the superbee function leads to an excessively compressive scheme which it is not very robust
for general practical aerospace applications 6.
In the numerical solution of the three-dimensional Euler equations, ve wave families
appear. If the ve wave families are enumerated in correspondence with their speed, being
one the slowest and ve the fastest, it can be demonstrated that for waves of the families two
to four, the characteristic velocities at both sides of the discontinuity are the same and equal
to the velocity discontinuity 3, 5. This property makes it very dicult to solve theses waves
accurately. Generally they are solved diusely because the numerical methods incorporate a
large amount of articial viscosity to track the contact discontinuity.
In this work, the possibility of implementing dierent limiter functions for dierent
wave families is explored. The objective is to improve the numerical resolution of the
discontinuities associated with the families two to four using compressive limiter functions
superbee, and without losing robustness mainly due to the use of diusive limiter functions
minmod for the wave families one and ve. This technique implements the utilization of
the superbee limiter function only in linear degenerate waves and the minmod function in
nondegenerated nonlinear waves 1, 2.
To introduce in the numerical uxes calculations the limiter function superbee, 2.5 is
replaced by the following expression:

0 if m m < 0
i 1/2 i1/2
gim    2.8
max0, min2k, 1, mink, 2 1 m m
if m m 0
2 i1/2 i1/2 i 1/2 i1/2

being

 
 m  m
i 1/2 i 1/2
k  . 2.9
 m  m
i1/2 i1/2

To improve the overall scheme robustness, the implementation of dierent limiter

functions is carried out only in those cells interfaces where the greater relative intensities
of the discontinuities in central waves are registered, and using the conventional Harten-Yee
TVD scheme in all other cases. Notice that the comparison among the intensity of the waves
cannot be made using directly the coecients of the spectral decomposition m i 1/2
 since
these coecients depend on the module assigned to each eigenvector.
4 Mathematical Problems in Engineering

In the local coordinate system adopted for computing the numerical uxes across each
face, the corresponding eigenvectors are given by 5

1
1 0 0 1
u
uc 0 0 u c

v
K1 
v ,
K2 

,
K3 
1 , K4 
0 , K5 
v ,


w w 0 1 w

u2 v 2 w 2
H uc v w H uc
2
2.10

where H is the stagnation enthalpy; u, v and w are the velocity vector components, and c is
the sound velocity. It can be deduced from 2.10 that 1i 1/2 , 2i 1/2 and 5i 1/2 measure the
density jump in the waves 1, 2, and 5, respectively, and that 3i 1/2 and 4i 1/2 measure the
momentum jump in waves three and four. To compare these jumps it became necessary to
select reference values for the density and velocity. Thus,

         
 1   2   3   4   5 
i 1/2  i 1/2  i 1/2  i 1/2  i 1/2 
I1  , I2  , I3  , I4  , I5  . 2.11
ref ref ref uref ref uref ref

In this investigation, ref  0.5 i i 1  is taken as density reference, and as the

velocity reference of the average of the sound velocities at the cells uref  0.5 ci ci 1 ,
where ci is the sound velocity. The parameters Ii permit to measure the wave intensities.
Finally, if the maximum of I1 , I5 is higher than the maximum of I2 , I3 , I4 , the
conventional Harten-Yee TVD scheme is used; otherwise, the values of gi2 , gi3 , gi4 , are
calculated with the limiter function superbee and gi1 , gi5 , with the limiter function minmod.
For the evaluation of gim and gi 1 m
in 2.4, it is necessary to calculate the spectral
decompositions of the conservative variables increments at the interfaces i 1/2, i 1/2,
and i 3/2. In the context of three-dimensional not structured meshes of tetrahedrons, the
identication of the cells i and i 1 is intuitive they are two cells that share a face but
the determination of the points i 1 and i 2 is not direct. If two tetrahedrons that share a
face are analyzed, the nodes not belonging to the common face can be used as representative
points for i 1 and i 2. Then, these points can be used as imaginary cells. In this work these
ideas have been implemented, being the nodal values calculated as a pondered average of
the conservative variables between all cells that are in contact with the nodes i 1 and i 2.
Such pondered average is given by

+n
i1 Ucell i /dGCcell i node k 
Unode k  + n , 2.12
i1 1/dGCcell i node k 

where dGCcell i node k is the distance that separates the gravity center of the cell i from the node
k, and n is the cell number in contact with the node k.
Mathematical Problems in Engineering 5

The treatment of the boundary conditions is carried out through the imaginary cells
technique 24. Five dierent types of boundaries are considered 1 subsonic inlet; 2
supersonic inlet, 3 subsonic exit, 4 supersonic exit, 5 nonpenetration solid boundary
and symmetry.

2.2. Reentry Equations of Motion

The choice of a suitable set of coordinates and parameters of the trajectory to describe the
movement of an object in atmospheric reentry is inherent to any investigation of guided
spacecraft. To analyze a reentry trajectory it is appropriated to describe the motion of the
center of mass using a set of elements known as Flight Coordinates 12.
Thus, the ight coordinates are described by the six orbital elements: magnitude of
the position vector, r, longitude, , latitude, , magnitude of the velocity vector, v, ight-path
angle, , and heading angle, azimuth of the velocity. At every moment this object is under


the inuence of a total force, F , composed by the gravitational force, F G , the aerodynamic


force, A, and the force of propulsion, T :

F  T A F G. 2.13

The gravitational force is always present. For nonpowered ight, the propulsion force
is zero, while for ights outside the atmosphere, the aerodynamic force vanishes.
To derive the equations of motion, we must use an Earth-xed reference system. The
kinematics equations of motion are 12

dr
 v sin ,
dt

d v cos cos
 , 2.14
dt r cos

d v cos sin
 .
dt r

It is desirable to separate the aerodynamic force into two components and dene the

tangential component of the nongravitational force, F T , along the velocity vector, and the


normal component, F N , orthogonal to the velocity at the aerodynamic plane. When we have


in plane ights, the normal vector F N is in the plane  r ,

v , the vertical plane, and there is
no lateral force. However, it is possible to create a lateral component of this force, which has
the eect of changing the orbital plane. The non-gravitational force is then decomposed into
a component on the vertical plane and orthogonal to the velocity vector, and a component
orthogonal to this plane using the bank angle, .
6 Mathematical Problems in Engineering

The force equations are 12:

dv 1 1 
 FT FG sin 2 r cos sin cos cos sin sin ,
dt m m

d 1 1 v2
v  FN cos FG cos cos 2v cos cos
dt m m r


2 r cos cos cos sin sin sin ,
2.15

d 1 FN sin v2 
v  cos cos tan 2v tan sin cos sin
dt m cos r

2 r
cos sin cos .
cos

Here is the rotation of the Earth that appears because we have to consider a reference
system xed on the planet.

2.3. Attitude Equations

Knowing the attitude of a space object means knowing the orientation of an axis system
connected to the vehicle related to a vertical reference system. To specify the orientation
of a rigid body in space, three independent parameters are needed. These parameters are
commonly known as roll, pitch and yaw, the Euler angles.
However, the use of the Euler angles to compute the attitude evolution of a spacecraft
is limited: the equations of motion in attitude have singularities for certain values of the pitch
angle, namely /2. This limitation was solved with the substitution of the Euler angles with
a set of variables known as quaternions 13.
The quaternions are denoted as q  q1 , q2 , q3 , q4 . The components of q are dened
in terms of the Euler angles using the convention xyz or sequence 321 which is found in
Goldstein 14.
Kinematics equations of motion have the following form 14:

1
q1  3 q2 2 q3 1 q4 ,
2
1
q2  3 q1 1 q3 2 q4 ,
2
2.16
1
q3  2 q1 1 q2 3 q4 ,
2
1
q4  1 q1 2 q2 3 q3 .
2
Mathematical Problems in Engineering 7

The attitude dynamic equations of motion express the temporal dependence of the
angular velocity related to the applied torques:

d


I  N
I
, 2.17
dt

where I is the inertia matrix, and N the aerodynamic torque. Thus, with I1 , I2 , I3
corresponding to the moments of inertia about the main axes of the vehicle, the dynamic
equations of attitude are 13:

N1 I2 I3 2 3
1  ,
I1
N2 I3 I1 3 1
2  , 2.18
I2
N3 I1 I2 1 2
3  .
I3

2.4. Aerodynamic Forces

If the vehicle under consideration operates on a symmetry condition, the velocity vector
denes this plane of symmetry. So the attitude of the object is properly described by the
attack angle, , which is the angle between the velocity vector relative to the atmosphere and
a vehicles baseline, normally the longitudinal axis.
The aerodynamic force is decomposed into two components: the force opposite to the


direction of motion, called Drag -and part of the non-gravitational force F T in 2.15, and the

orthogonal component, called Lift part of F N :

1
D CD atm Ssat V 2 ,
2
2.19
1
L  CL atm Ssat V 2 .
2

Here, CD and CL are called drag and lift coecients, respectively, atm is the atmospheres
density, Ssat is the satellites area and V is the relative velocity between the space object
and the atmosphere. For each object it is necessary to calculate specic coecients at every
moment of the trajectory. The code developed in the Department of Aeronautics of the UNC
1 evaluates these parameters for the conical object into consideration.
The rst phase of the calculation is to identify the objects surface areas. So, the nodes
of the tetrahedrons, whose faces form the surface of the object, are identied in the code input
le.
The forces acting on each face have a magnitude equal to the pressure divided by
the area. The direction is the incoming normal to the surface, and the point of application
corresponds to the geometric center of the face. In this way, the resultant force on the body
is calculated as the vector sum of all forces acting on the discretized surface. Finally, the
8 Mathematical Problems in Engineering

resulting force components in directions parallel and perpendicular to the ow velocity

vector are projected, in order to obtain the drag and lift coecients, respectively.

3. Results
The numerical simulations were performed using as a core calculation code the one
developed at UNC 1. For delineation of the geometric bodies and meshing, we used the
application GID 8.0 with temporary license issued by the manufacturer.
GID has been developed as an interface for geometric modelling, meshing, income
data, and display results of all types of numerical simulation programs. The dierent menus
can be modied according to specic user needs. The graphical interface adapted to the code
is designed to allow the entrance of initial conditions, to mark the objects surface for the
calculation of forces, to dene the number of iterations, and other parameters inherent to the
code. The ultimate objective of the use of GID is writing the data le that enters the UNCs
code and the subsequent display of results, from reading the output le written by the code.
The limit of iterations in the code can be determined by the number of steps, or
the limit time independently. When it comes to any of these, the program completes its
implementation.
The motivation of the simulations performed is to obtain the main aerodynamic
characteristics drag and lift coecients of a conical body. They are used in the calculation
of the trajectory during reentry into the atmosphere 15. Due to the fact that the original
code did not have the calculation of the pressure forces on a body, the necessary sentences
to perform a simple summation of forces on the faces of the tetrahedrons lying on the object
were written. For the identication of these faces, the graphical interface developed in GID is
used, which has an option to mark the surface of the body.
The angle of the cone 10  was chosen arbitrarily, in order to guarantee that this value
is small. The cones length 1 m was chosen to facilitate the calculations.
It is known that the aerodynamic characteristics of a body with the given geometry
depend, in the case of nonviscous ow, only on the Mach number and the incidence angle of
the free ow, without considering heat transfer or changes in the properties of air. Thats the
reason why these variables are used as independent ones.
Two sets of simulations were performed. The rst case is for a cone-shaped object
that enters in the atmosphere by sharp narrow side forward, while the second one was
calculated in the assumption that the vehicle moves by wide obtuse side forward. Both
calculations were made for several attack angles and Mach numbers.
It is important to note that the pressure distribution around the body depends only
on the free ow Mach number. Then to obtain the CD and CL coecients, it is necessary to
modify only the free ow Mach number. For this reason the velocity is modied by keeping
pressure and density constant and equal to those used in the remaining cases. The speed in
the z direction is maintained equal to zero so there is ow symmetry around the xy plane.

3.1. Sharp Side Forward Case

The dimensions of the volume meshing are length in direction x  2.2 m, length in direction y
 3 m, distance to surface point of impact in the axis x  0.2 m, angle of revolution  180 . The
characteristics of the mesh are number of tetrahedrons around 130,000 and number of nodes
around 29,000.
Mathematical Problems in Engineering 9

Pressure Pa
122370 Pressure Pa
530390
47700

25440

a b

Figure 1: Pressure distribution on the object: a Mach  1, attack angle  0 ; b Mach  4, a.a.  15 .

The interface with GID requested that a new mesh was drawn every time a new
calculation was made, because it needed a new data le for every case considered. For this
reason, and because a nonstructured mesh was used, every case had a slightly dierent
number of elements, due to arbitraries processes during meshing.
The criterion used for the meshing is based on
i using the symmetry of the ow on a plan to reduce the volume of mesh, and on
two planes in the case of zero attack angle,
ii concentration of all elements near the surface of the cone, in particular areas
considered critical for the calculation: impact point and base,
iii signicant reduction of elements in remote areas of the body, to reduce computa-
tional cost.
Figures 1a and 1b display some examples of results obtained for the pressure eld
over a conical object considering two combinations of Mach number and attack angle.
The implemented numerical scheme has the capacity to simulate compressible ows
in subsonic, transonic, and supersonic regimens. From Figure 1a, it is possible to note, for
Mach number equal 1, the concentration of sonic waves near of the body edge; however, the
shock wave has not yet been formed. The shock wave can be seen clearly in Figure 1b when
the free Mach ow is 4. Furthermore Figure 1 shows that the pressure distribution, including
at the cone base, is dierent for transonic from that of supersonic ows , this explains the
dierent behaviors of the CD and CL coecients for dierent regimes.
Figure 2 shows the drag coecients for the simulated cases, while Figure 3 shows the
lift coecients. In all cases it is considered the cone angle equal to 10 . The reference area used
is the front section of the cone. It is important to note that the viscous eects are neglected to
obtain theses gures.
Note from Figure 2 that drag coecient is increasing accordingly as the attack angle
of the cone is increasing too, this phenomenon occurs until 75 . From Figures 2 and 3, it is
possible to observe that both coecients, drag and lift, have lower variations for height Mach
numbers and the greater variations occur at subsonic and transonic ows. For an attack angle
of 45 , it is produced that the highest lift and the lower occur for an angle of 90 , for this test
the lift is negative except for very reduced Mach number.
Figure 4 show some of the results of the reentry simulations. Figure 4a plots the
variation of attack angle during the reentry trajectory for initial values of this angle
10 Mathematical Problems in Engineering

2.5

Drag coecient
2

1.5

0.5

0
0 1 2 3 4 5 6
M

0 60
15 75
30 90
45

Figure 2: Sharp Side Forward: Drag coecient as a function Mach for dierent angles of attack.

1.5

1
Lift coecient

0.5

0
0 1 2 3 4 5 6

0.5
M

15 60
30 75
45 90

Figure 3: Sharp Side Forward: Lift coecient as a function Mach for dierent angles of attack.

corresponding to 0 , 15 , and 30 , while Figure 4b presents the trajectory, meaning the
altitude as a function of the time, for these angles.
We can notice the moment when the objects sense the presence of the atmosphere
around 1500 seconds and begin to experience high variation of the attack angle during
their descent. The rst trend is to decrease the initial attack angle. However, although the
atmosphere promotes an oscillatory movement around the zero attack angle condition, this
is very unstable for objects reentering sharp side forward, and it is possible that for certain
initial conditions, the atmosphere turns the object and reverses its attitude.
In what concerns the time evolution of the altitude, it could be seen that it is not
aected until the height of 150 km. This is a limitation of the atmospheric model that has
information on the atmospheric density just until this height. Since the aerodynamic forces
Mathematical Problems in Engineering 11

90 250000

60
200000
Attack angle deg

30

Altitude m
150000

0
100000
30

50000
60

90 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Time s Time s
a b

Figure 4: Sharp Side Forward: a Attack angles depending on the time; b Altitude. Initial attack angles:
0 blue, 15 red and 30 green.

depend directly on this parameter, it is assumed that the atmosphere is so faint over 150 km
that it does not aect the trajectory. Anyway, under this limit, all trajectories show bouncing
movements, indicating that the attitude variation aects the objects dynamics.

3.2. Wide Side Forward Case

In this case the characteristics of the mesh are number of tetrahedrons around 160,000 and
number of nodes around 32,000. An additional criterion used for the meshing is based on the
concentration of all elements over the cones surface, in particular areas considered critical
for the calculation: impact point and base. Also smaller elements were used in the zones
nearby the cone base and along the axis of the cone in order to determine more precisely
detached shock waves. Apart from that, the mesh was made thinner over the sides of the cone
compared to the aerodynamic shadow in order to more accurately determine the expansion
waves.
To simulate the atmospheric reentry of a cone shaped object by wide side forward, it is
takes into account that this object has an attack angle around 180o . Some of the results found
for this condition can be seeing on Figures 5 and 6.
Figure 5 shows the drag coecients as a function of the Mach number for attack angles
105 , 120 , 135 , 150 , 165 , and 180 . It can be seen that this parameter presents higher values
on the condition wide side forward compared with the same attitude angles on a sharp
side forward situation for attack angles around the stability condition 180 in this case and
0 in the sharp side forward as expected. For attack angles equal to or bigger than 135
corresponding to 45 in the other case, the drag coecient is very similar. On the other side,
the lift coecients tend to be smaller or maintain the same values Figure 6.
The result of these dierences can be noticed in Figures 7a and 7b. For a conic
object that reenters the terrestrial atmosphere wide side forward, small angles of attack here
around 180  tend to preserve their magnitude during the reentry trajectory as corresponding
12 Mathematical Problems in Engineering

2.5

Drag coecient
2

1.5

0.5
0 1 2 3 4 5 6
M

180 135
165 120
150 105

Figure 5: Wide Side Forward: Drag coecient as a function Mach for dierent angles of attack.

0.5
Lift coecient

0
0 1 2 3 4 5 6

0.5

M

165 120
150 105
135

Figure 6: Wide Side Forward: Lift coecient as a function Mach for dierent angles of attack.

to an equilibrium conguration. However, when the initial attitude is as far from 180 as 30 ,
equilibrium is lost during reentry and we found again the oscillations around the stability
condition.
The most important consequence of the wide side forward reentry is over the altitude
evolution. It can be seen in Figure 7b that all the cases present nearly ballistic paths.
Although the far from the 180 initial attack angle condition, the less ballistic is the trajectory
inside the dense part of the atmosphere, as can be seen by the dark green line that corresponds
to an initial attack angle of 150 , the trajectories of initial attack angles equal to 180 and 175
are superimposed.
Mathematical Problems in Engineering 13

270 250000

240
200000
Attack angle deg

210

Altitude m
150000

180
100000
150

50000
120

90 0
0 500 1000 1500 2000 0 500 1000 1500 2000
Time s Time s
a b

Figure 7: Wide Side Forward: a Attack angles depending on the time; b Altitude. Initial attack angles:
180 dark blue, 175 dark red and 150 dark green.

4. Conclusions
In this paper, a series of numerical simulations have been conducted using a code
developed at UNC 1, 2, which solves the Euler equations by the method of nite volume,
using unstructured 3D tetrahedral meshes. The code implements a new technique on the
introduction of limiting functions, which aims to decrease the numerical viscosity, that is,
increasing the contact discontinuities capture accuracy without loss of robustness regarding
other TVD methods.
The goal was the calculation of aerodynamic characteristics of a cone under the eects
of dierent ight conditions, to attach the results to other code that calculates the dynamic of
atmospheric reentry 7.
It was possible to perform simulations of reentry trajectories for a specic cone under
the inuence of various aerodynamic eects for a variety of initial attack angles. Simulations
show, as expected, that the trajectories are more aected when the object has initially a sharp
side forward conguration.
From the obtained results, was compared the numerical slope of the normal force
coecient with analytical results. In none of the cases the errors exceed 6%. This good
accuracy of the numerical results permits to induce that the error by not considering viscous
eects in the calculation of the aerodynamic coecients is low, and the obtained trajectories
and attack angles evolution during the reentry are reliable.
Although the methodology implemented has been shown to be suitable for calculating
the reentry trajectories inside the terrestrial atmosphere, it will be improved with the
inclusion of viscous eects in the simulation of the aerodynamic ow in future works.

References
1 O. A. Falcinelli, S. A. Elaskar, and J. P. Tamagno, Reducing the numerical viscosity in non-structured
three-dimensional nite volumes computations, Journal of Spacecraft and Rockets, vol. 45, no. 2, pp.
406408, 2008.
14 Mathematical Problems in Engineering

2 S. A. Elaskar, O. A. Falcinelli, J. P. Tamagno, and J. P. Salda, Further applications of scheme for
reducing numerical viscosity: 3D hypersonic ow, Journal of Physics: Conference Series, vol. 166, Article
ID 012018, 13 pages, 2009.
3 C. Hirsch, Numerical Computation of Internal and External Flows, Computational Methods for Inviscid and
Viscous Flows, vol. 2, John Wiley & Sons, New York, NY, USA, 1992.
4 P. K. Sweeby, High resolution schemes using ux limiters for hyperbolic conservation laws, SIAM
Journal on Numerical Analysis, vol. 21, no. 5, pp. 9951011, 1984.
5 E. F. Toro, Riemann Solvers and Numerical Methods for Fluid Dynamics: A Practical Introduction, Springer,
Berlin, Germany, 2nd edition, 1999.
6 R. J. LeVeque, Numerical Methods for Conservation Laws, Lectures in Mathematics ETH Zurich,
Birkhauser, Basel, Switzerland, 2nd edition, 1992.
7 W. Schulz and P. Moraes Jr., Reentry trajectory simulation of a small ballistic recoverable satellite,
in Advances in Space Dynamics 4: Celestial Mechanics and Astronautics, H. K. Kuga, Ed., vol. 4, Instituto
Nacional de Pesquisas Espaciais, Sao Jose dos Campos, Brazil, 2004.
8 B. Udrea, An advanced implicit solver for MHD, Ph.D. thesis, University of Washington, Seattle, Wash,
USA, 1999.
9 A. Harten, On a class of high resolution total-variation-stable nite dierence schemes, Tech. Rep.,
New York University, New York, NY, USA, 1982.
10 A. Harten, High resolution schemes for hyperbolic conservation laws, Journal of Computational
Physics, vol. 49, no. 3, pp. 357393, 1983.
11 H. C. Yee, R. F. Warming, and A. Harten, Implicit total variation diminishing TVD schemes for
steady-state calculations, Journal of Computational Physics, vol. 57, no. 3, pp. 327360, 1985.
12 N. X. Vinh, Optimal Trajectories in Atmospheric Flight, Elsevier Scientic, Amsterdam, The Netherlands,
1981.
13 J. R. Wertz, Ed., Spacecraft Attitude Determination and Control, D. Reidel, Dordrecht, The Netherlands,
1978.
14 H. Goldstein, Classical Mechanics, Addison-Wesley Series in Physic, Addison-Wesley, Reading, Mass,
USA, 2nd edition, 1980.
15 J. P. Saldia, W. Schulz, and S. Elaskar, Calculo de la Dinamica de Reentrada de Objetos Espaciales,
in Proceedings of the 14th Coloquio Brasileiro de Dinamica Orbital (CBDO 08), November 2008.