Sie sind auf Seite 1von 8

1

AbstractAn attitude control system using reaction thrusters


for the upper stage of a launch vehicle is considered. The thruster
configuration (position and direction) determines control system
response, fuel consumption, effective torque and system fault
tolerance. We propose a procedure for finding the optimal
thruster configuration with desired control effectiveness over the
range of selected torque commands. An optimization technique
called Particle Swarm Optimization is used for the numerical
experiments. The validity of the solution is checked through
computer simulations.

Index Terms PWPF, RCS, PSO, Thruster Configuration,
Optimization Technique

I. INTRODUCTION
Up-to-date launch vehicles carry multiple satellites and
every satellite has its own demanding criteria for safely placing
itself into orbit. The precision of the satellite orbit is determined
by the attitude control efficiency of the upper-stage launch
vehicle. The Reaction Control System (RCS) is plentifully used
for attitude control means of the upper-stage launch vehicle in
space. For three-axis control of the launch vehicle, multiple
thrusters are required. In the case of the upper-stage launch
vehicle, there is a space restriction. The thruster must be set up
in the connection region that is between the upper part and the
lower part of the launch vehicle. The number of thrusters is also
limited because the pressure level of the fuel tank must be
maintained.
In the case of linear thruster, several theoretical works have
been done about thruster configuration design for limited fuel
flow rate, maximum robustness to thruster failure, maximum
margin of safety [1], [2]. On-off one-sided thruster may be hard
to analyze by linear algebraic theory, because it has a lot of
nonlinear characteristics. In this research, the control system is
assumed to use on-off one-sided thrusters. In general, control
system response, fuel consumption, effective torque and system
fault tolerance are affected by the thruster configuration
(position and direction).

Manuscript received January 31, 2003.
Authors Address : all authors are with department of aerospace engineering
,KAIST, Taejon , Korea.
Contacting Author: Min-Jea Tahk
E-mail : twhwang@fdcl.kaist.ac.kr
Taejon , youseong, guseong 373-1 KAIST, AE, FDCL
We propose a procedure for finding the optimal thruster
configuration. The optimal solutions are given through
optimizing a cost function related to control effectiveness. An
optimization technique called Particle Swarm Optimization
(PSO) is used for optimization.
Section II presents the general formulation of attitude
dynamics of launch vehicles. In section III we explain the
procedure of gain selection for quaternion feedback loop, the
process of characteristic parameter determination for the PWPF
modulator, and an example of the command distribution logic
for a thruster configuration of the previous design. Section IV
introduces a PSO optimization algorithm which is one of the
newest optimization techniques. Section V considers the
process of problem formulation and cost function definition. In
section VI computer simulations are performed to show
performance of the optimized configuration.

II. LAUNCH VEHICLE ATTITUDE DYNAMICS
The orbit motions of upper-stage launch vehicle are
composed of translational and rotational motion. Their
independent characteristics allow one to deal with two motions
separately. For attitude control of a vehicle, only rotational
equations are required. The rotational (attitude) motions of a
rigid spacecraft in space are described by Eulers equations (1)
[3].

|
B
dH
M H
dt
= +

(1)

We deduced the rotational motion equation from (1)

J J + =

(2)
where,
( , , )
T
p q r =

: Angular velocity of the vehicle w.r.t. inertial space


1 2 3
( , , )
T
=

: Total external torque


J : Inertia dyadic w.r.t. the center of mass of the vehicle

III. ATTITUDE CONTROL SYSTEM DESIGN
The attitude control system treated in this paper is a
quaternion feedback method. In the past, the open-loop type
algorithm or the Euler angle feedback was frequently used for
Upper-Stage Launch Vehicle Servo Controller Design
Considering Optimal Thruster Configuration
Tae Won Hwang, Chang-Su Park, Min-Jea Tahk, Hyochoong Bang
AIAA Guidance, Navigation, and Control Conference and Exhibit
11 - 14 August 2003, Austin, Texas
AIAA 2003-5330
Copyright 2003 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.


2
control system in space. For the quaternion feedback system,
the entire usable attitude related information is expressed in the
body axis. There is no discordant coordinates problem within
attitude angle and angular velocity information. Also, the
attitude conversion of the shortest course through an Euler axis
rotation is possible. The conversion equation from Euler angle
( , , ) to quaternion (
0 1 2 3
Q q q i q j q k = + + + ) is as
follows.

0
1
2
3
cos( ) cos( ) cos( ) sin( ) sin( ) sin( )
2 2 2 2 2 2
sin( ) cos( ) cos( ) cos( ) sin( ) sin( )
2 2 2 2 2 2
cos( ) cos( ) sin( ) sin( ) sin( ) cos( )
2 2 2 2 2 2
sin( ) cos( ) sin( ) cos( ) sin( ) cos
2 2 2 2 2
q
q
q
q




+

+
+
(
(
(
(
(
=
(
(
(
(
(

( )
2

(
(
(
(
(
(
(
(
(
(

(3)

The attitude error quaternion is the coordinate
transformation quaternion from the current attitude
B
i
q to the
command attitude
C
i
q . The conversion equation (3) gives
B
i
q and
C
i
q . Error quaternion
e
q is expressed as (4) from the
relationship of coordinate transformation. The upper script *
denotes the multiplication operation of two quaternions.

0 1 2 3 0
1 0 3 2 1
2 3 0 1 2
3 2 1 0 3
*
( ) ( )
c c c c b
c c c c b
c c c c b
c c c c b
C C B
e B i i
q q q q q
q q q q q
q q q q q
q q q q q
q q q q



= =
=
( (
( (
( (
( (
( (

(4)

The concept of quaternion control is to apply the quaternion
control toque to the vehicle. The quaternion control torque
vector has a magnitude which is proportional to the rotary angle
of the error quaternion and direction parallel to the Euler axis.
The feedback error angle is given as equation (5). When
the rotation angle is small, the error angle can be
expressed as (6).

e =

(5)
[ ]
1 2 3
2 , ,
e e e
q q q (6)
where,
[ ]
0 1 2 3
, , , cos ,(sin )
2 2
e e e e
q q q q e

=
(
(


1
0
2 cos ( )
e
q

= : Rotation angle ,
[ ]
1 2 3
, ,
T
e e e e =

=
1 2 3
sin sin sin
2 2 2
( , , )
T
e e e
q q q

: Euler Axis vector
It was proved in [4] that control command (7) makes the
closed-loop stable.
p d
u K K =
(7)

The block diagram of the RCS attitude tracking loop is
shown in Figure 1.

The attitude control system is composed of the quaternion
feedback loop, the PWPF modulator, and the command
distribution logic.
A. Feedback Loop Gain Selection
The control gains ,
p d
K K of equation (7) are acquired
from the solution of the LQR problem for which the
performance index is given by (9) and the state equation (8)
disregards the cross product term of the original equation (2).

State Equation:
,
0 1 0
,
0 0 1
J
x Ax Bu u
A B

= + =
| | | |
= =
| |
\ . \ .

(8)
Performance index:
2
2
2 1
2
(
1 0
1
,
0 0
)
t
u
c
V x d
Q R
c

= +
| |
= =
|
\ .

(9)
The optimal control gain and control command are expressed
by (10). The matrix P of (10) is the solution of the arithmetic
riccati equation. It can be solved analytically as follows:
Optimal Control & Gain :
1
1
2
0
T
p d
T T
G R B P K K c c
A P PA PBR B P Q
u G x

( ( = = =

+ + =
=
(10)
Closed characteristic transfer function :
2
(2 ) s c s c + +
(11)
Control feedback gains are determined as follows.
, 2 , 1, 2, 3
i i
i i
p d
i i
J J
K c K c i

= = =
i
J : Moment of inertia
i
: Maximum external torque
T
Flight
Dynamics
K
d
K
p

+
-
IMU
-
PWPF Modulator
q
b
q
c
U
on
U
off
1
-1

m
s + 1
K
m
Quaternion
Computation


Fig. 1. Block diagram of RCS attitude tracking loop


3
B. PWPF Modulator Design
Pulse Width Pulse Frequency(PWPF) modulators are used
for thruster control of numerous satellites and launch vehicles.
As in Fig.1, the PWPF modulator is composed of a first order
lag filter and Schmitt trigger. The pulse sequence (1,0,-1)
proportional to the input command size will drive the thruster
valves. The experimental results of control system which uses
PWPF agree with results of the linear analysis well. The
boundary limits of the input which maintains linear
characteristics are given as (12). The effective dead band of a
modulator is derived as (13) [5].

min
max
1
on
m
off
m
U
K
U
K
r
r
=
= +
(12)
min
( )
ln(1 )
on on off off
m m
m m
U U U U
K K
T

=
(13)

There is no relation between the PWPF modulator static
characteristics and the vehicles moment of inertia. The
parameters of the PWPF modulator are set as follows: The
sampling frequency of the control loop is 50 Hz. The excess
values of
max
r is 2%. The target error of
PWPF
is chosen as
0.5

. The DC gain of lag filter


m
K is selected as 3. When the
effective dead band is smaller than the minimum pulse width,
the internal parameters of PWPF are given as (14).

0.02
/( )
on m p PWPF
off m
m m m on off
U K K
U K
K U U
=

(14)
[ , , ] 0.5 0.02sec
T
PWPF PWPF m
= = =



C. Command Distributon Logic
The basic thruster configuration of the concept design is
illustrated in Fig. 2. There are 8 thrusters parallel to the Y-Z
plane of the vehicle. Positive pitching moment is acquired by
turning on thrusters number 5 and 6. Positive yawing moment
is acquired by turning on thrusters number 3 and 4. Positive
rolling moment is acquired by turning on thrusters number
1,3,5 and 7.
Thruster
Main
Nozzle
Thruster
Main
Nozzle
Payload
(satellite)
Thrusters(8 )
Main
Nozzle
Final
Stage
Payload
(satellite)
)
Main
Nozzle
Final
Stage Thruster
Main
Nozzle
Thruster
Main
Nozzle
Payload
(satellite)
Thrusters(8 )
Main
Nozzle
Final
Stage
Payload
(satellite)
)
Main
Nozzle
Final
Stage


The control command are given in 3 directions; pitch, yaw,
roll. As in Fig.2, 8 thrusters are needed to control the system.
Thus, we require another command distribution logic. This can
be solved by deriving a minimum norm type pseudo inverse
E
+
from the non-dimensional thruster control matrix E given
in (15). The thruster control matrix E is the contribution
measure of controllability for a thruster configuration.

Thruster Control matrix : E
1 1
2 2
2 2 0 0 2 2 0 0
1
3 3
0 0 2 2 0 0 2 2
4
1 1 1 1 1 1 1 1
7 7
8 8
p
y
r
E






( (
( (
( (
(
(
( (
(
(
= =
( (
(
(
( (
(


( (
( (
( (


(15)
Command Mixing Matrix :
1
( )
T T
E E EE
+
=
1
2
3
7
8
p
y
r
E

+
(
(
(
(
(
(
=
(
(
(
(

(
(
(

(16)

The possible state of one-sided thrusters is just ON or OFF.
The relation (16) is used with a threshold with the size of .
When an input is over (positive small value), the output
should be ON for the thruster. On the other hand, when an input
is smaller than , the output should be OFF. The
non-dimensional effective torque due to torque command is
derived as (17)
1
2
3
/ /
7
8
( ) ( )
e
e
p
p
y on off on off y
r
r e
E f E f E

+
(
(
(
(
(
(
(
(
= =
(
(
(
(
(
(


(
(
(

(17)

1 2
-1 -2
-3 3
-4 4
0.5m
0.5m
1 2
-1 -2
-3 3
-4 4
11 22
6 5
8 33
7 44
0.5m
0.5m
Main Nozzle
1.3m
1 2
-1 -2
-3 3
-4 4
0.5m
0.5m
1 2
-1 -2
-3 3
-4 4
11 22
6 5
8 33
7 44
0.5m
0.5m
1 2
-1 -2
-3 3
-4 4
0.5m
0.5m
1 2
-1 -2
-3 3
-4 4
11 22
6 5
8 33
7 44
0.5m
0.5m
Main Nozzle
1.3m

Fig.2. Launch vehicle modeling and thruster configuration of concept design


4
When a system adopts the linear two-sided thrusters which
has a large thrust limit, the torque command agrees exactly with
non-dimensional effective torque (17). However, our system
uses the on-off one-sided thrusters; there happens control cross
coupling phenomena. Smaller effective torque compared to the
torque command is produced when the simultaneous 3
direction command are given. The torque command from the
pulse modulator should be one of three values, -1,0,1. But the
non-dimensional effective torque can be of any value between
-1 and 1. All 27 command types can exist.

IV. OPTIMIZATION TECHNIQUE
In this section, we introduce a PSO algorithm which is one of
the newest optimization techniques. The PSO was originally
developed to graphically simulate the bird flock or insects. The
birds and insects form groups and move in a flock. They are
assumed to share all information of their group. Applying these
characteristics to optimization theory, the PSO technique was
born [6].
The advantage of the PSO technique is easier numerical
formula and more efficient convergence quality compared to
general evolutionary algorithms(GA, EP, ES..). The
assumption of random population is shared with general
evolutionary algorithms. But PSO does not need the original
characteristics of evolutionary algorithm such as reproduction,
mutation, and selection process. PSO adopt the concept of
position and velocity for each population.
The random velocity term is produced to aim the best points
of each population and the whole group in track. Finally, the
whole group will be placed in the vicinity of the optimal
solution. The risk of converging to a local minimum is avoided
by adjusting the velocity terms of several populations to escape
from the historically best point
The algorithm of the PSO technique has 6 steps and is briefly
explained as follows:

ParticleSwarmOptimization(PSO)Algorithm
1. Initialize a population of particles with random positions
and velocities
2. For each particle, evaluate the desired optimization fitness
function
3. Compare particle's fitness evaluation with particle's pbest.
Exchange the particle's fitness value and location with pbest
if it is better.
4. Compare particle's fitness evaluation with the population's
overall previous best, gbest. Exchange the particle's fitness
value and location with gbest if it is better.
5. Change the velocity and position of the particle according
to the following equations
1 2
( ) ( )
id id
id id best id best id
id id id
c Rand p x c Rand g x
x x

= + +
= +

6. Loop to step 2 until a criterion is met.

The PSO algorithm is applied to Dejong F1 cost function in
Fig.3. The randomly assigned velocity term of each population
give rise to individuals converging motion to the optimal
point.

De Jong F1 :
2 2
( , ) f x y x y = + , Optimal Solution:(0,0)

V. COST FUNCTION AND PARAMETER OPTIMIZATION
A. Cost Function Selection
The desirable thruster configuration means that effective
torque vector has the same direction as the original 27 torque
command and the torque vector has maximum magnitude in
each direction. A cost function which is the summation of inner
product of command torque and resultant torque with
weighting factors is introduced to determine the optimal
configuration. Tuning the weighting factor of each command
direction make it possible to give weight to a specific torque
direction.
As stated in above section, the thrusters must be set up in the
connection region that is between the upper part and the lower
part of the launch vehicle. The thruster must be located on a
circle with a radius 0.7071m where the main nozzle is the
center point. is the angle which determines the place where
the thruster is located. The angle between the thruster direction
vector and its projected vector in the Y-Z plane is called . The
angle of the projected vector in the Y-Z plane with respect to
the Y-axis is called .

Fig.4. Thruster configuration and definition of position/direction angle
(3-dimensional Rear view of upper-stage launch vehicle)


Fig.3. Movement of population when PSO algorithm is working


5
As you seen in the Fig.4, if a thruster location is decided in
the first quarter, the location of 3 thrusters in 3 other quarter
planes are decided automatically in accordance with the
symmetry rules. The symmetry about the X axis, the symmetry
about the Y axis, and the symmetry about the origin point will
satisfy the condition of zero mean torques in all 3 directions.
The assumption is frequently used in the actual thruster design.
The arrangement of two thrusters in the first quarter plane
accomplishes the thruster configuration.
The thruster position and direction angles for the first quarter
plane are limited as in Table 1. For application to other planes,
the results obtained from the first quarter must be converted
according to the converting logic of Table 1. If the result of
conversion exceeds 180, then it can be converted as the
angle in 180. The minimum angle of is limited to 30
considering the structural limitation.

TABLE I
BOUNDARY CONDITION FOR , , AND CONVERTING LOGIC
1
st
quarter 2
nd
quarter 3
rd
quarter 4
th
quarter
1
1
1
0 90
180 180
30 90


<





2 1
2 1
1 2


=
=
=
3 1
3 1
3 1
180
180



= +
= +
=


4 1
4 1
4 1
180
180



=
=
=



The external torque in this system is a multiplied value of
thrust of thrusters and torque arm length in pitch/yaw/roll
directions. The torque equation about a thruster configuration
in pitch/yaw/roll directions is expressed as follows.

1 2 3 7 8
where, [ , , , , , ]
2 , cos( ) , sin( )
cos( ) , sin( )
cos( ) , sin( )
p y v
y x v
r x y
h v
x h y h
T dn T dy
T dn T dx
T dy T dx
T T T T T T
dn m dx r dy r
T T T T
T T T T




=
=
= +
=
= = =
= =
= =

(18)

The thrust vector can be broken up into
h
T parallel to Y-Z
plane and
v
T parallel to X-axis. The distance from C.G. of the
vehicle to the lower end plane is the pitch/yaw direction torque
arm for
h
T (2m). The , dx dy which is dependent on the
location of the thruster on a circle is the another pitch/yaw
direction torque arm for
v
T . The rolling torque is generated on
Y-Z plane. The thrust
h
T is the only source of rolling torque.
The effective torque (18) can be non-dimensionalized by
dividing it by the maximum torque value of each axis. The
non-dimensional thruster control matrix E can be written as
(19). As there are 8 thruster, the dimension of E is 38 matrix.
i
X is the maximum torque value in the i direction.

1
2
3
7
8
,
p
p
p
y
y
y
r
r
r
X
E E
X
X

(
(
(
(
(
(
(
(
(
(
( = =
(
(
(
(
(
(
(
(
(
( (

(19)
8 8
1 1
8 8
1 1
8 8
1 1
(| | ) ( | | )
max( | | , | | )
2 2
(| | ) ( | | )
max( | | , | | )
2 2
(| | ) ( | | )
max( | | , | | )
2 2
i i i i
i i i i
i i i i
p p p p
p
i i
y y y y
y
i i
r r r r
r
i i
X
X
X



= =
= =
= =
+ +
=
+ +
=
+ +
=





The control commands from the pulse modulator are
separated into 3 axes. However, we must use this information to
control the switch gear of 8 thrusters. Thus, the inverse value of
E is needed to control the system. The command distributing
matrix
1
( )
T T
E E EE
+
= is derived as (16). It is the
minimum norm type pseudo inverse of E (minimum-power
controller). X is the dimensional conversion parameter which
is derived in (19). Considering the maximum torque
parameter X , the dimensional resultant torque is given as
follows.

/
, ( )
e e e
e e e
p p p p
p
y y y y on off y
r
r e r e r r e
X
X E f E
X



+
( ( (
(
( ( (
(
= =
( ( (
(
( ( (
(


(20)

The of function
/ on off
f is set to 0.3 in this research.

The 27 commands/responses against three pitch/yaw/roll
Radius=1
Sphere
Normalized
Command Torque n
Dimensional
Resultant Torque T
r
Radius=1
Sphere
Normalized
Command Torque n
Dimensional
Resultant Torque T
r
Fig.5. Normalized torque command and dimensional resultant torque


6
directions are possible to get from equation (20). The cost
function for optimization is set as (21). It is a summation of
inner product between normalized command torque vectors
and weighted dimensional resultant toque vectors. Minus sign
is applied for minimization. In this paper, we emphasized the
margin of rolling torque. Thus, we set higher weightings in
rolling torque direction. The optimal thruster configuration is
accomplished by searching for the location/direction angle
which minimizes the cost function.

27
1
k
k k r
k
J n w T
=
(
=
(

i
(21)
,
[ , , ] ,
e e e
T
r p y r k
where
T w weighting factor = =



To verify the selected cost function, the cost function is
evaluated for 3 different thruster configurations with 0 = . 3
cases are shown in Fig 6. The cost function value for each case
is -2259.8, -2417.8, -2110.4. Case C is previous configuration
design. The smallest cost function is acquired in case B. Thus,
case C is a near optimal configuration.

B. Parameter Optimization
As mentioned in the previous section, the uncertain
parameters are the 6 position/direction angles of the two
thrusters in the first quarter plane. PSO technique in Section IV
is applied to find the optimal solution. Boundary conditions are
as follows.

1 1 1
0 90 180 180 30 90 <



Case 1) Thruster establishment location is fixed (
1,2
0 = )
For some cases, the thruster establishment locations are
limited due to the fuel tank and valve locations. The thruster
nozzle positions are fixed and the nozzle directions are freely
chosen by the optimization process. Fig.7. is the optimal
thruster configuration with that case.

Case 2) Thruster establishment location is free (
1,2
is free)
If there are no constraints on the position of two thrusters in
first quarter plane, the optimal thruster configuration by PSO is
plotted as Fig. 8. It is configured to have maximum torque
quantity in all directions.



0 500 1000 1500 2000
-150
-100
-50
0
50
100
150
200 Swarm Optimization Results
Thruster Position/Direction Angle
E
s
t
i
m
a
t
e
d

O
p
t
i
m
a
l


P
o
s
i
t
i
o
n
/
D
i
r
e
c
t
i
o
n

(
D
e
g
)
Iteration Number

1

1

1

2

2

2
Fig.9. Variation of estimated parameters (angle) when PSO is applied

Fig.8. The optimal thruster configuration (
1,2
is free)

Fig .7. The optimal thruster configuration (
1,2
0 = )

(case A) (case B) (case C)

Fig.6. Configuration example for cost function validity test


7

Fig.9 and Fig 10 shows that it converges after about 1000
steps. PSO gives robust convergence irrespective of initial
proposition of the population. The sequential quadratic
programming (SQP), optimization technique provided
MATLAB, is applied to same cost function to confirm the
reliability of solution [7]. SQP could not match the PSO in
convergence ability.

VI. NUMERICAL SIMULATION
The assumed upper-stage launch vehicle specification and
thrust level in numerical simulations are as follows.

TABLE II
UPPER-STAGE LAUNCH VEHICLE SPECIFICATION
Mass
(Kg)
C.G.
(m)
length
(m)
radius
(m)
Ixx
(Kg-m
2
)
Iyy
(Kg-m
2
)
Izz
(Kg-m
2
)
1500 2.0 4.0 1.3 1267.5 2633.7 2633.7

TABLE III
THRUSTER SPECIFICATION
Thruster level 20N
No. of available Thruster 8 EA
Fuel limit MAXIMUM OPERATING TIME : 500 SEC

Numerical simulations are performed using Simulink in
MATLAB. The validity of the proposed optimal thruster
configuration is checked with the RCS attitude control loop
response and fuel consumption.
Fig 11 and Fig 12 shows the responses of quaternion feed
back loop and accumulated fuel consumption when using
normal operating thrusters. It shows that the proposed optimal
configuration is little bit superior to the existing configuration
from the view point of response characteristics. However, the
proposed configuration requires slightly more fuel
consumption than the previous one.



To check the response differences in respective
configuration in case of abnormal operating thruster case, we
assumed a situation where one of 8 thrusters went out. Fig.13
and Fig.14 shows the maximum overshoot case and the
minimum undershoot case in simulation results. Fig.13 is for
the previous concept design configuration and Fig.14 is for the
proposed optimal thruster configuration. It shows the optimal
thruster configuration can achieve smaller overshoot and
undershoot error in case of a trouble. Securing more effective
torque in every command direction made it possible to gain
more robustness about a thruster failure.

0 10 20 30 40 50 60 70 80 90 100
-20
0
20
40
60
80
100
120
140
160
180
EulerAngle: PWPFCoarseControl
Time
QuaternionFeedbackResult
Phi
Theta
Psi
Command

0 10 20 30 40 50 60 70 80 90 100
-60
-40
-20
0
20
40
60
80
EulerAngle: PWPFCoarseControl
Time
QuaternionFeedbackResult
Phi
Theta
Psi
Command
Fig.13. Existing thruster configuration: (One of 8 thrusters is not operating)
Euler angle response (Maximum overshoot, Minimum undershoot case)
0 10 20 30 40 50 60 70 80 90 100
-10
0
10
20
30
40
50
60
70
80
Euler Angle : PWPF Coarse Control
Time
Quaternion Feedback Result
Phi
Theta
Psi
Command
0 10 20 30 40 50 60 70 80 90 100
0
5
10
15
20
25
30
35
40
int (1/2 u
2
) : PWPF Coarse Control
Time
Quaternion Feedback Result
Control Force
Fig.12. Proposed thruster configuration: (Normal Operation)
Euler/Quaternion response and accumulated fuel consumption
0 10 20 30 40 50 60 70 80 90 100
-10
0
10
20
30
40
50
60
70
80
Euler Angle : PWPF Coarse Control
Time
Quaternion Feedback Result
Phi
Theta
Psi
Command
0 10 20 30 40 50 60 70 80 90 100
0
5
10
15
20
25
30
35
int (1/2 u
2
) : PWPF Coarse Control
Time
Quaternion Feedback Result
Control Force
Fig.11. Existing thruster Configuration: (Normal Operation)
Euler/Quaternion Response and accumulated fuel consumption
0 500 1000 1500 2000
-2600
-2500
-2400
-2300
-2200
-2100
-2000
Swarm Optimization Results
Cost Function
C
o
s
t
Iteration Number
J
Fig.10. Variation of cost function when PSO is applied
0 10 20 30 40 50 60 70 80 90 100
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Quaternion Angle : PWPF Coarse Control
Time
q0
q1
q2
q3
q0c
q1c
q2c
q3c
0 10 20 30 40 50 60 70 80 90 100
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Quaternion Angle : PWPF Coarse Control
Time
q0
q1
q2
q3
q0c
q1c
q2c
q3c


8

VII. CONCLUSION
This paper considers an attitude control system using
reaction thrusters for the upper stage of a launch vehicle. The
attitude control loop is designed based on error quaternion
feedback algorithm. We introduce a gain selection procedure
for quaternion feedback loop, a characteristic parameter
determination process for PWPF modulator and a way to set the
command distribution logic. The thruster configuration has a
big influence on control system response, fuel consumption,
effective torque and system fault tolerance. We proposed a
procedure to find the optimal thruster configuration that
guarantees an effective torque vector which has similar
direction to original torque command with maximum possible
magnitude. The proposed procedure can find optimal thruster
configuration which improves control loop efficiency even
when the position and direction of the thruster configuration are
limited.
The control gain determination process is demonstrated.
Computer simulation result show that the optimal thruster
configuration has more desirable control characteristics than
that of the previous thruster configuration.
ACKNOWLEDGMENT
The authors would like to acknowledge support from the
Korea Ministry of Science & Technology (MOST) under
Project Number M1-0138-00-0005, the space technology
development program.
REFERENCES
[1] R.S.Pena, R.Alonso, P.Anigstein,"Robust Optimal Soltuion to the
Attitude/Force Control Problem ", IEEE Transation on Aerospace and
Electronic System, Vol. 36, No.3 July 2000
[2] H.P.Jin, P.Wiktor, D.B.Debra, "An Optimal Thruster Configuration
Design and Evaluation for Quick Step", Control Eng. Practice., Vol. 3,
No.8 1995, pp. 1113-1118
[3] A.E. Bryson, Control of Space craft and Aircraft , Princeton University
Press, 1994
[4] B.Wie, J.Lu, "Feedback control logic for spacecraft eigenaxis rotations
under slew rate and control constraints", Journal of Guidance, Control,
and Dynamics V.18,1995, pp. 1372-9.
[5] B.N.Agrawal, R.S.Mcclelland, G.Song, "Attitude Control of Flexible
spacecraft Using Pulse-Width Pulse-Frequency Modulated Thrusters",
Space Technology, Vol. 17, No. 1, 1997, pp. 15-34
[6] Kennedy and Eberhart, Particle swarm optimization" Proc IEEE Int.
Conf Neural Networks, Nov 1995
[7] T.F.Coleman, Y.Jhang, "Optimization Tollbox User's Guide", Mathworks,
July. 2002

0 10 20 30 40 50 60 70 80 90 100
-20
0
20
40
60
80
100
120
EulerAngle: PWPFCoarseControl
Time
QuaternionFeedbackResult
Phi
Theta
Psi
Command
0 10 20 30 40 50 60 70 80 90 100
-20
-10
0
10
20
30
40
50
60
70
80
EulerAngle: PWPFCoarseControl
Time
QuaternionFeedbackResult
Phi
Theta
Psi
Command
Fig.14. Proposed thruster configuration: (One of 8 thrusters is not operating)
Euler angle response (Maximum overshoot, Minimum undershoot

Das könnte Ihnen auch gefallen