AIAA 99-4265

INITIAL CONDITIONS AND FUEL-

OPTIMAL CONTROL FOR FORMA-
TION FLYING SATELLITES
Srinivas R. Vadali, Hanspeter Schaub and Kyle T. Al-
friend
Texas A&M University, College Station, TX 77843-3141

AIAA Guidance, Navigation and

Control Conference
August 9–12, 1999 / Portland, OR
For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics
1801 Alexander Bell Drive, Suite 500, Reston, VA 22091
 AIAA-99-4265

INITIAL CONDITIONS AND FUEL-OPTIMAL CONTROL

FOR
FORMATION FLYING OF SATELLITES

S. R. Vadali, H. Schaub, and K. T. Alfriend

Department of Aerospace Engineering
Texas A&M University
College Station, TX 77843-3141

Abstract . generate periodic solutions. Recently, Bond6 developed

an alternate set of equations which does not have the
This paper deals with the formulation and solution
stability (secular drift) problem associated with Hill’s
of the initial condition determination and fuel-optimal
equations. The shape of the projection of the relative
control problem in regard to formation flying of
orbit perpendicular to the radial (zenith-nadir) direction
satellites. Unlike the relative station-keeping problem
is of interest for the purpose of optical interferometry.
for satellites in the same plane, formation flying entails
Kong et al.7 have studied various types of free and
the creation of proper out-of-plane relative motion such
forced orbits, using Hill’s equations, suitable for space
that the satellites in the formation satisfy some mission
based interferometes. Melton8 presents a state
requirements. Two methods for determining initial
transition matrix for relative motion between satellites
conditions of a satellite (Deputy), given the initial
in elliptic orbits in terms of a power series in
conditions of the chief satellite, for formation flying
eccentricity.
without thrust for a long period, are presented. The first
method matches the mean J 2 -induced angular drift The attractive solutions to Hill’s equations under
rates of the two satellites, and the second method is ideal conditions get disturbed when perturbations due to
based on imposing periodic boundary conditions on the the Earth’s oblateness or aerodynamic drag are included
relative position and velocity in a rotating coordinate in the model. The fuel consumption required for
system. A fuel-optimal impulsive thrusting scheme is maintaining a formation to fight these perturbations will
developed to establish a formation. A fuel-optimal, low- be prohibitive for more than a very short period of time.
thrust, variable Isp propulsion scheme is presented for The primary perturbation of interest is due to J 2 , which
orbit maintenance. causes, among others, three important effects: Nodal
regression, and drifts in perigee and the mean anomaly.
Introduction These effects, using orbit-averaged quantities, are given
below9:
Under ideal Two-Body assumptions, a satellite 2
(Deputy) can be kept at a constant distance from  = −1.5 J æç Re ö n cos i
Ω (1)
2ç
another satellite (Chief) in a circular orbit, without the è p
use of thrust, by choosing the right phasing and the

( )
inclination of the relative orbit. This type of a relative 2
orbit has been proposed for the LISA1 mission. Other æR ö
 = 0.75 J 2 çç e n 5 cos 2 i − 1
ω (2)
proposed formation flying missions are ST32, ORION3, è p
Auroral Lites4 , and Techsat-2123. Hill’s equations5

( )
have been used to study relative motion of rendezvous 2
2 æ Re ö
mechanics. The key to establishing a formation using M = n + 0.75 J 2 1 − e çç n 3 cos 2 i − 1
Hill’s equations is to choose initial conditions that è p
(3)
.
Copyright © 1999 by the American Institute of Aeronautics
and Astronautics, Inc. All rights reserved.

1
American Institute of Aeronautics and Astronautics
where Ω , is the longitude of the ascending node, ω , is
the argument of perigee, M , is the mean anomaly, e ,
is the eccentricity, i , is the inclination,
Equations of Relative Motion
The equation of motion of a satellite under the
p = a (1 − e 2 ) , n = µ / a 3 , and a , is the semi- influence of gravitational and thrust effects is given
major axis. below:
If these effects are not controlled, the satellite r = −φ r + u (4)
formation will break down. Differential drag is not a
major concern if the two satellites have similar where, r , is the position vector, φ , is the gravitational
Kechichian10
potential and φ r is its gradient. The thrust acceleration
aerodynamic characteristics. has
developed a method for studying relative motion in the
presence of oblateness and drag using a dragging and is denoted by u . The gravitational potential and its
precessing reference frame. The formulation presented gradient5 , including the contribution of J 2 , are shown
in this work is based on osculating elements rather than below:
the mean elements.
µ é J æ R2 ö öù
2
æ z2
φ = − ê1 − 2 çç e ÷÷
Out-of-plane motion can be created using a node ç 3 − 1÷ (5)
difference or an inclination difference between the rê 2 è r ø ç r2 ÷
satellites. It is much easier to deal with the node ë è ø
difference than the inclination difference because the
latter gives rise to differential drift rates in µ J 2 µRe2 é 6z æ 3 15 z 2 ö ù
Ω,ω, and M . φr = 3 r + ê 5 nˆ + çç 5 − 7 ÷÷r ú
r 2 ëê r èr r ø ú
Schaub and Alfriend11 developed an analytical
solution to the initial condition problem by enforcing to (6)
first order, equal mean nodal rates and equal mean The expressions for the energy and the angular
 + M . Unfortunately, the three equations, Eq. (1-3)
ω momentum are also useful.
cannot be satisfied simultaneously by two satellites with
1 T
different mean a , e, and i . The mean elements are E= v v+φ (7)
converted to corresponding osculating elements using a 2
first order approximation taken from Brouwer12.
H = r×v (8)
In this paper, the two conditions of Ref. 11 are
It is well known that for the above model, the energy
enforced numerically to determine the initial conditions
and the polar component of the angular momentum are
of the deputy, given the initial conditions of the chief.
conserved in the absence of thrust and drag.
An alternate constraint set is also developed in terms of
the differences in energy and the polar component of Herein, variables with subscript 0 are used to
the angular momentum, between the two satellites. A denote the conditions of the chief. Any variable
second technique is presented for the determination of connected with the deputy is denoted by a subscript 1.
initial conditions, using numerical integration of the The inertial relative displacement and velocity are
differential equations subject to periodic boundary defined as follows:
conditions on the relative position and velocity, as seen
in a rotating coordinate system. This technique δ r Z r1 J r0 (9)
produces relative orbits with much smaller drift rates δ v Z v1 J v 0 (10)
than the previous method but it is not convenient for use
in the control problem. The relative motion between two satellites in
general elliptic orbits is best visualized in a rotating
Subsequently, the first technique for determining frame constructed with the following coordinates:
the initial conditions is utilized to develop a multi-
impulse, minimum-fuel control scheme to establish the δ r T r0
formation. A low-thrust, variable Isp, minimum-fuel δxZ (11)
control scheme is also developed for periodically r0
correcting small errors in the formation due to other δ r T (H 0 “ r0 )
perturbations. This method of control can be δyZ (12)
implemented using plasma or ion propulsion.
H 0 “ r0

2
American Institute of Aeronautics and Astronautics
 δ rT H0 also be calculated using position and velocity
δzZ (13) information, they provide a means of checking
H0 transformations between mean and osculating orbit
The relative velocities in this rotating coordinate system elements. The initial condition problem can be stated as
follows:
are given by the following:
Determine the position and velocity vectors of the
deputy, given the position and velocity vectors of the
δ v Tr0 H δ r T v 0 (δ r T r0 )(δ r0T v 0 ) chief, to satisfy the following constraints:
δ x Z J (14)
 =Ω
r0 r03 Ω 0 1 (20)

δ v T (H 0 “ r0 ) H δ r T (H 0 “ r0 H H 0 “ v 0 )
δ y Z J  0 + M 0 = ω
ω  1 + M 1 (21)
H 0 “ r0
δ r T (H 0 “ r0 )(H 0 “ r0 ) T (H 0 “ r0 H H 0 “ v 0 ) Since δi is specified, there are six unknown parameters
3 to be selected with three constraints. Hence additional
H 0 “ r0 constraints can be imposed to get a unique solution.
(15) The constraints of Eq. (17-18) are equivalent to those of
Eqs. (20-21). The process of converting the mean

δ v TH0 H δ r TH  )
δ r TH 0 (H 0TH
δ z Z 0
J 0 elements to the respective osculating elements is treated
H0 H03 in Ref. 11. The conversion of the osculating elements
to position and velocity can be found in any textbook on
(16) Astrodynamics, such as the one by Battin13. The
We refer to relative motion along nonlinear constraint satisfaction problem is solved using
the nonlinear programming code NPOPT which is a
δ x , δ y , and δ z , respectively, as radial, along-
part of the SNOPT package14. Two examples are
track, and out-of-plane. treated in this section:

Establishment of Approximate Example-1:

Periodic Motion Using Analytical LM5883.7397OP
Solutions
MM
r (0) = 2274.2509 km
P
N2524.0672PQ
0
Schaub and Alfriend11 have presented analytical
results to establish orbital parameters of the deputy for a
J 2 invariant relative orbit. It can be shown that for
such orbits, the difference in the energies between the
LM-3.9273OP
deputy and chief, to first order in J 2 is MM PP
v (0) = 4.5469 km / sec
0

E1 − E 0 = δE =
µJ 2 δ H z
cos 4 i0 (1 + 5 cos 2 i0 )
N 5.0533 Q
5 The mean elements for the chief, calculated using
5 a 0 H z0
Brouwer’s theory are:
(17)
and the difference in the polar component of the angular a = 7152.9917 km, e =.0499, i = 0.8377,
momentum is
Ω = -4E - 7, ω = 0.5237, M = -1.0037E - 004
5
δH z = − H z0 tan i0 δi (18) Besides the constraints given by Eqs. (20-21), three
4 additional constraints are imposed as given below:
where, the inclination difference between the two
satellites is given by: δi Z.002 , δt Z.005 , δω Z.01 , and δM Z J.01
δi = i1 − i0 (19)
Since δH z and δE are known to be constants of The solution for the initial position and velocity vector
offsets, expressed in the inertial frame are
motion in the absence of other perturbations, and can

3
American Institute of Aeronautics and Astronautics
 δr (0) Z [ -1.320 - 0.1061 - 0.511] km This is an example of a near-polar orbit. Mean
elements are:
δv (0) = [-0.108E - 2 0.377E - 3 0.115E - 2] km / sec
a0 = 8876.786 km, e0 = 0.1781, and
Figure 1 shows the relative orbit and Fig.2 shows the i0 =1.49589. The geometric constraints are specified as
relative displacement. These figures represent data for
a period of 3 days.
δi Z tan J1 (1 / a0 ), δt Z 0, δω Z 0, and δM Z 0
The specification of the inclination difference as shown
above, produces the maximum out-of-plane motion at
1 the peak latitudes, in the amount of 1 km. The out-of-
0.5 plane displacement at the equator is zero.
Radial (km)

0
1.5
-0.5

-1

1
3

Out-of-plane (km)
1 0.5

-1
0
0.5
Along-Track (km) -2 0
-0.5
Out-of-Plane (km)
-0.5

Fig. 1. Relative Orbits for Example-1

-1
-20 -15 -10 -5 0 5 10 15 20
3.2 Radial (km)

3
Fig. 3. Relative Orbits for Example-2
2.8
1
Relative Displacement (km)

2.6
0.8
2.4
0.6
2.2
0.4
delta-z of the deputy (km)

2
0.2
1.8
0
1.6
-0.2
1.4
-0.4
1.2
0 0.5 1 1.5 2 2.5 3 -0.6
Time (days)
-0.8
Fig. 2. Relative Displacement for Example-1
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Z-position of the chief (km) 4
x 10
Fig. 4. Nature of the out-of-plane displacement for
Example-2
The first and the last relative orbits are shown in
Example-2:
Fig. 3 over a period of 10 days. Note that the axes are
not drawn to the same scale, in order to exaggerate the
r0 (0) = 7300 0 0 km out-of-plane motion. The relationship between δz ,
v 0 (0) = 0 0.6 8 km/s the out-of-plane displacement and z , the polar
component of the chief’s inertial displacement are

4
American Institute of Aeronautics and Astronautics
plotted in Fig. 4. This figure shows that indeed, the frame, in which Figs. 1, 3, and 5 are plotted. Although
maximum out-of-plane displacement of 1 km is the integration method is more accurate, it is
achieved at the extreme latitudes. inconvenient for computing optimal controls.

Establishment of Periodic Motion by 1.5

Numerical Integration
1
In this section the procedure to obtain periodic or
near-periodic relative motion between the chief and the
0.5
deputy is described. Given the initial conditions of the

Out-of-plane (km)
chief, initial conditions of the deputy are sought such
0
that the relative velocity and the relative position
vectors satisfy periodicity conditions at an unknown
-0.5
time (approximately the nodal period of the chief’s
orbit). Hence seven parameters have to be determined.
-1
The six constraints on the relative positions and
velocities are of the type given by Eqs. (22-23)
-1.5

δ y (0) = δ y (t f )
-30 -20 -10 0 10 20 30
(22) Radial (km)

δ x (0) = δ x (t f ) (23)
Fig. 5. Relative Orbits for Example-2 obtained using the
where tf is an undetermined parameter. In order to integration method
compare the solutions to the same problem using the the
current and the previous method, Example-2 is selected 1.5
for testing. Notice that in Fig. 4, there is a linear
relationship between δz and z . This requires two 1
constraints as shown below:
delta-z of the Deputy (km)

tan(δ i ) z0 - sin(i0 )δ z Z 0
0.5
(24)
tan(δ i ) z0 - sin(i0 )δ z Z 0 (25) 0

These constraints can be specified at the initial or final -0.5

times. Since there is one too many constraints, the
radial position periodic constraint is dropped. This
-1
gives rise to a problem with seven unknowns and seven
constraints. The initial guess for the solution is
-1.5
obtained from the solution to the same problem using -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
the method described in the previous section. Z-displacement of the Chief (km) 4
x 10

Fig. 6. Nature of the out-of-plane dispalcement for

The relative orbits for Example-2 for a period of 10 Example-2
days are shown in Fig. 5. Only the first and the last
orbits are shown in the figure for the sake of clarity.
Figure 6, shows the plot of the out-of-plane
displacement versus the polar component of the
displacement of the chief. It is interesting to note that
the relative orbits of Fig. 5 are larger compared to those
of Fig. 3 but the drift is much less. However, Fig. 6
shows that there is a slight differential nodal precession.
This is due to the way the constraints are specified. In
the first method, equal nodal rates are enforced. Nodal
precession is about the inertial z-axis. In the second
method, the constraints are in terms of the rotating

5
American Institute of Aeronautics and Astronautics
 The above matrix is obtained by using the average drift rates
Fuel-Optimal Multi-Impulse
for the elements. The B ( e) matrix is written in the same
Control form as given in Ref. 13, with the assumption that mean
orbit elements will be used to valuate it. This may not
be precise but proves useful.
In this section, the first method of determining
initial conditions, discussed above, is used to define a LM 2a e sin f
2
2a 2 p
0
OP
terminal constraint set for the multi-impulse fuel-
optimal control problem. Prussing and Chiu15 have
MM p sinH f Hr
( p + r ) cos f + re PP
utilized primer vector theory for computing optimal MM H H
0
r cosθ P
P
multi-impulse rendezvous maneuvers. Kumar and
B ( e) = M
M 0 0 P
r sin θ P
Seywald16 treat the problem of fuel-optimal station- H
keeping based on Hill’s equations. Smith17 has studied MM 0 0
H sin i P
P
the problem of satellite constellation maintenance in a
MM − p cos f ( p + r ) sin f − r sin θ cos i P
H sin i P
multi-impulse setting using genetic algorithms. Schaub
et al.18 present a feedback control scheme using the MMη( p cosHef − 2re) He
−η( p + r ) sin f PP
mean elements as well as the cartesian position and
velocity vectors. Ulybyshev19 uses the LQR approach
N He He
0
Q
for controlling the drifts in period and nodes. A
nonlinear, adaptive, tracking controller is proposed by
p
Queiroz et al20. In this paper, we use the mean element η = 1 − e2 , r = , and θ = ω + f
formulation and Gauss’ variational equations13. These 1 + e cos f
equations are quite convenient for the problem at hand
H Z µp is the scalar angular momentum. The true
due to the fact that during coasts, the mean
a , e, and i remain constant and the mean anomaly, f , is related to the mean anomaly through
Ω,ω, and M vary linearly. Gauss’s equations for the the following equations:
variations in the mean elements are written below: M = E − e sin E (27)
e = A (e) + B (e)u (26)
f 1+ e E
tan = tan (28)
where, e = [a e i Ω ω M ] T
and the control 2 1− e 2
u = [ur uθ uH ] . The control vector is defined using
T

components along the radial, tangential, and orbit In order to establish a formation, the mean elements
normal directions. Other quantities are as defined a , e, and i need satisfy Eqs. (19-21) and the mean
below:
Ω,ω, and M have to meet some specifications.
Assuming that the control is impulsive, the elements
LM 0 OP undergo jump discontinuities at the impulse application
times. During the coasting phases, they can be
MM 0
PP integrated analytically. The true anomaly and the other
time-varying quantities in the B ( e) matrix can be
MM 0
FR I 2
PP calculated using the equations presented above. The
. J G J n cos i
−15
H pK
e impulse application times and the delta-v
A ( e) = M PP
2
magnitude/directions become the free parameters. Since
MM FR I 2

PP
the fuel consumed is directly proportional to delata-v,
0.75 J G J nc5 cos i − 1h the net delta-v is to be minimized. Once the
MM H pK
e 2
2

MMn + 0.75J 1 − e FG R IJ nc3 cos i − 1hPPP

optimization process is over, the terminal position and
2 velocity states of the deputy can be obtained by
transformation from the mean elements. The maximum
H pK
2 e 2

N 2
Q number of impulses is unknown apriori, but is not more
than six since there are six elements to be changed in
general. Allowing for initial and final coasts and six
impulses, the total number of parameters is 25. In some

6
American Institute of Aeronautics and Astronautics
instances, a computed impulse magnitude may be too 5 0.55D-4 -0.184D-4 0.338D-2 3.38D-3
small and can be eliminated. 6 -0.42D-6 -0.51D-5 -0.98D-5 1.11D-5

The total delta-v required is 6.204 m/sec, which is

Example: primarily due to impulses 1 and 5. Both of them are
The mean elements of the chief are selected as predominantly in the normal direction. They occur
when θ is close to π . The B ( e) matrix shows that
e 0 = [7555km .05 480 0 100 1200 ] this is a time when inclination change is best performed.
The impulse times and the final time are
The corresponding orbital elements of the deputy
are , t i = [640.49, 813.467 , 3863.48 ,
6014.13, 7169.23 , 1047.85] sec
e1 Z [7555km .05057 48.0540 t f = 1.19602d4 sec
The variation of the semi-major axis is shown in Fig. 7.
-0.01 9.9981 119.9774 ] The first impulse overcorrects for the error and the
subsequent impulses bring the semi-major axis back to
The desired orbit element differences at the time of the desired value.
orbit establishment are 7555.1

δi =.006 , δΩ = δω = δM = 0
0 0
7555.08

7555.06
Semi-major axis (km)

The differences in the other orbit elements to

satisfy the two rate constraints can be determined from
Eq. (20-21)as 7555.04

δa = −.00193km, δe =.0005767 7555.02

7555
The deputy is initially disturbed from the desired
state by introducing large errors in a , i ,and Ω : 7554.98
0 2000 4000 6000 8000 10000 12000
a1 = a0 + δa − 100m Time (sec)

i1 = i0 + δi +.050 Fig. 7. Changes in the Semi-major axis

Ω1 = Ω 0 + δΩ −.01 0
48.07
e1 = e 0 , ω 1 = ω 0 , M1 = M 0
48.06

A six-impulse solution to the orbit establishment 48.05

problem is given below in Table 1:

Inclination (deg)

48.04
Table1
48.03
Impulse Magnitudes
# Radial Tangential Normal Total 48.02

(km/sec)
48.01
1 0.51D-4 0.40D-4 0.278D-2 2.78D-3
2 -0.57D-7 0.442D-7 0.758D-6 7.62D-7
48
3 0.153D-6 -0.125D-4 -0.26D-4 2.92D-5 0 2000 4000 6000 8000 10000 12000
Time (sec)
4 -0.11D-7 0.918D-8 -0.74D-8 1.65D-8

7
American Institute of Aeronautics and Astronautics
 considered. The formulation presented here is similar
Fig. 8. Changes in the inclination to that given in Ref. 21. Only the deputy is assumed to
Figure 8 shows that the changes in the inclination be active.
come in two installments. As mentioned before, the
delta-v expenditure is primarily due to these inclination The equations of motion including the thrust terms
changes. Figure 8 shows the resulting relative orbits are given below:
over a period of 3 days after the maneuver is completed.
r = v (29)
2εP
v = −φ r + u (30)
4 mg
2 2εεP T
m = − u u (31)
Radial (km)

0 g2
-2 where P , is the maximum power available, ε , is the
-4
efficiency factor, g , is the acceleration due to gravity
at sea level, m , is the mass of the satellite, and u is
8
6 the control vector defined as follows:
4
1
2
uT u =
Isp 2
0
-2
-4
-6 The controls are constrained through the relationship
Along-Track (km)
-8 0 0.5
-0.5 given below:
Out-of-Plane (km)

Fig. 9. Relative Orbits after the maneuver 1 1 1

2
≤ 2
≤ (32)
Isp max Isp Isp min 2

Orbit Maintenance Using The optimal control problem is posed as:

Power-Limited, Electric Propulsion
Minimize: J = − m(t f ) (33)
Electric propulsion using ion or plasma thrusters is
being considered for formation flying. Power- limited, subject to Eqs. (29-31) and the terminal constraints
low-thrust propulsion has already been demonstrated by
the DS-1 mission. The Techsat-2123 program has a Ψ (e(t f ), t f ) = 0 , where, e(t f ) , is the mean orbit
proposed mission plan, involving swarms of element specification for the deputy. The constraint set
microsatellites, each with an estimated mass of 77 kg is of dimension six and is full rank. It is the same used
and a power rating of 1 kw. The thruster rating for each in method-1 for the determination of initial conditions.
satellite is of the order of 0.1 N. The specific impulse, The final time, t f , is free. The variational
Isp, can vary between 1200-2500 sec. It is assumed
here that each satellite has one thruster that can be Hamiltonian is written as follows:
gimbaled to produce thrust in a desired direction.
There exist many works in this area. Kechichian21 e = λ r r + λ V v + + λ m m (34)
has presented an excellent treatment of orbit transfer The costate equations are
using low-thrust propulsion. Coverstone-Carroll and
Prussing22 consider the problem of cooperative power-
∂e
limited rendezvous between two satellites in λr = − (35)
neighboring circular orbits using Hill’s framework. In ∂r
this study both the satellites are given the freedom to ∂e
maneuver and cooperate to achieve better performance λv = − (36)
∂v
than that of an active/passive pair.
The controls are related to the costates as shown
Since the available thrust is severely limited for the below24:
Techsat 21 program, the Isp bounds must be

8
American Institute of Aeronautics and Astronautics
Unconstrained Solution: optimal control problem is solved by a direct approach
using the nonlinear programming code NPOPT. The
g λv
u = (37) transversality conditions on the other costates and the
2m λ m Hamiltonian are not utilized. There are seven unknown
initial costates and the final time, to be determined. The
nonlinear programming problem is posed with the
The convexity condition is performance index given by Eq. (33). The
2 performance index and the terminal constraints are
∂ H 4εP evaluated by integrating Eqs. (29-31 and 35-36). The
=− λm > 0 (38)
2 controls are evaluated using Eqs. (37), (39), or (40),
∂u g2
which ever is applicable. The terminal constraints are
given by Eqs. (19-21), three specifications on
Ω,ω, and M , and Eq. (41). The partial derivatives
Hence the solution as obtained above is optimal if
are computed using finite differences by the code.
λ m < 0 and the Isp constraint is not violated. For a
minimum-fuel problem, λ m is negative if the final time As an example, the initial conditions of the chief
is sufficiently long. However for the problem at hand, are chosen as those of Example-1 on page-3. The
this may not be the case. The optimal control when the desired mean orbit element differences are:
constraints are active are given below:
δi =.00057 , δΩ =.01 , δω =.01 , and δM = −.01
Constrained Solutions The initial conditions of the deputy are disturbed from
the desired, by perturbing its position by 100 m along
For λ m > 0 , the optimal solution requires minimum each inertial axis and the inertial velocity by 0.1 m/sec
along each axis. To correct for the initial errors, the
Isp. This solution is
minimum Isp of 1200 sec is required. The final time is
399.3 sec and the mass of fuel required is 0.0029 kg.
λv
u=− (39) The control accelerations required along the three
inertial directions are those shown in Fig. 10.
I sp min λTv λ v
-7
x 10
6

For λ m < 0 , there are two possibilities. If the 4

Z
X

unconstrained solution is such that I sp > I sp 2

Control Acceleration (km/sec 2)

then
max
the controls are given by 0

-2
λv
u=− (40) -4

I sp max λTv λ v Y
-6

-8
If the unconstrained solution is such that
I sp < I sp then the controls are the same as in Eq.
min
-10
0 50 100 150 200 250 300 350 400
Time (sec)
(39).
Fig. 10. Controls along the inertial axes
The transversality condition on the mass costate is

λ m (t f ) = −1 (41) The maximum control acceleration required is of the

Since the transformation between the position and 2
order of 0.0011 m/sec . This translates to a peak
velocity vectors and the mean elements is quite thrust magnitude of .085 N at beginning of life of the
complex, the free-final time, terminally constrained, satellite.

9
American Institute of Aeronautics and Astronautics
 Conclusions [8] Melton, R. G., “Time-Explicit Representation of
Relative Motion Between Elliptical Orbits,” Paper
In this paper the formulation and solution of the presented at the AAS/AIAA Astrodynamics
initial condition determination and fuel-optimal control Conference, August 4-7 1997, Sun Valley, Idaho.
problem in regard to formation flying of satellites was
presented. The emphasis was on creation of the out-of- [9] Vallado, D. A., Fundamentals of Astrodynamics
plane relative motion using both inclination and node and Applications, McGraw Hill Inc., 1997.
differences. Two methods for determining initial [10] Kechichian, J. A., “The Analysis of the Relative
conditions were presented. The first method matches the Motion in General Elliptic Orbits with Respect to a
mean J 2 -induced angular drift rates of the two Dragging and Precessing Coordinate Frame,” AAS
satellites, and the second method is based on imposing 97-733, presented at the AAS/AIAA
periodic boundary conditions on the relative position Astrodynamics Conference, August 4-7 1997, Sun
and velocity in a rotating coordinate system. Even Valley, Idaho.
though the second method is more accurate, the first
[11] Schaub, H. and Alfriend, K.T., “ J 2 Invariant
method can readily be used for control. A method to
determine multi-impulse fuel-optimal impulsive Relative Orbits for Spacecraft Formations”, NASA
maneuvers to establish formations was presented. A GSFC Flight Mechanics and Estimation
fuel-optimal, low-thrust, variable Isp propulsion scheme Conference, May 1999.
was presented for orbit maintenance. [12] Brouwer, D., “Solution to the Problem of Artificial
Satellite Theory Without Drag,” The Astronautical
Journal, Vol. 64, N0. 1274, 1959, pp. 378-397.

Acknowledgment [13] Battin, R. H., An Introduction to the Mathematics

and Methods of Astrodynamics, Published by
This research was supported by AFOSR under the AIAA. Inc., New York, NY., 1987.
Techsat21 program with Dr. Marc Jacobs serving as the
contract monitor. [14] Gill, P. E., Murray, W., and Saunders, M. A.,
“User’s Guide for SNOPT 5.3: A Fortran Package
for Large Scale Nonlinear Programming,” Dec 7,
References 1998
[1] Bender, P. et al., “LISA: Laser Interferometer [15] Prussing, J. E., and Chiu, J., “Optimal Multi-
Space Antenna,” Technical Report, European Impulse Time-Fixed Rendezvous Between Circular
Space Agency, MPQ 208. Orbits,” Journal of Guidance, Control, and
[2] http://spacetechnology3.jpl.nasa.gov/indexm.html Dynamics, Vol.9, No. 1, Jan.-Feb. 1986, pp. 17-22.

[3] How, J.P., “ORION: A Low Cost Demonstration of [16] Kumar, R.R., and Seywald, H., "Fuel-Optimal
Formation Flying in Space Using GPS,” AIAA-98- Stationkeeping via Differential Inclusions," Journal
4398. of Guidance, Control, and Dynamics, Vol. 18, No.
5, pp. 1156-1162.
[4] Hametz, M. E., Conway, D. J., and Richon, K., [17] Smith, J. E., “Application of Optimization
“Design of a Formation of Earth Orbiting Satellites: Techniques to the Design and Maintenance of
The Auroral Lites Mission,” Proceedings of the Satellite Constellations,” CSDL-T-1336, M.S.
1999 NASA GSFC Flight Mechanics and Thesis, MIT, June 1998
Estimation Conference, pp. 295-308.
[5] Kaplan, M.H., Modern Spacecraft Dynamics and [18] Scaub, H., Vadali, S. R., Junkins, J.L., and
Control, John Wiley, Inc., New York, NY. 1976. Alfriend, K. T., “Spacecraft Formation Flying
Control Using Mean Orbit Elements,” AAS 99-
[6] Bond, V. R., “A New Solution for the Rendezvous 310, presented at the 1999 Astrodynamics
Problem,” AAS 99-178, presented at the Conference, Alaska.
AAS/AIAA Space Flight Mechanics Meeting, [19] Ulybyshev, Y., "Long-Term Formation Keeping of
Breckenridge, Colorado, February 7-10, 1999. Satellite Constellation Using Linear Quadratic
[7] Kong, E. M., D. W. Miller, R. J. Sedwick, Controller," Journal of Guidance, Control, and
“Optimal Trajectories and Orbit Design for Dynamics, Vol. 21, No. 1, pp. 109-115.
Separated Spacecraft Interferometry”, MIT, SERC [20] Queiroz, M. et al., “Nonlinear Control of Multiple
Reort # 13-98, November 1998. Spacecraft Formation Flying,” AIAA 99-4270,

10
American Institute of Aeronautics and Astronautics
 Proceedings of the AIAA Guidance, Navigation,
and Control conference, Portland, OR, August
1999, and C

[21] Chobotov, V. A., Ed., Orbital Mechanics, Second

Edition, Published by AIAA. Inc., New York, NY.,
1996,Chapter 14, Contributed by Kechichian, J. A.,
pp 327-402.
[22] Coverstone-Carroll, V. and Prussing, J. E., Optimal
Cooperative Power-Limited Rendezvous Between
Neighboring Circular Orbits, Journal of Guidance,
Control, and Dynamics, Vol.16, No. 6, Nov.-Dec.
1993, pp. 1045-1054.
[23] http://www.vs.afrl.af.mil/VSD/TechSat21/
[24] Vadali, S.R., Nah, R., Braden, E., and Johnson Jr.,
I., " Fuel-Optimal Planar Interplanetary
Trajectories Using Low-Thrust Exhaust-Modulated
Propulsion," AAS 99-132, presented at the
AAS/AIAA Space Flight Mechanics Meeting,
Breckenridge, Colorado, February 7-10, 1999.

11
American Institute of Aeronautics and Astronautics