Professional Documents
Culture Documents
Proba-3 High Precision Formation Flying Control System
Proba-3 High Precision Formation Flying Control System
ABSTRACT
The PROBA-3 System consists of two independent,
PROBA-3 is the most challenging of the ESA's
6 degrees of freedom spacecraft flying in a high
technology demonstration missions and is devoted
elliptic Earth orbit; the nominal master from the
to fly for the first time a fully autonomous
GNC point of view is the Occulter Spacecraft (OSC)
Formation Flying (FF) mission. It will require
since it is the spacecraft that performs the fine
overcoming several difficulties ranging from the
actuations with Cold Gas thrusters for the relative
development of new technologies, to the application
position control hence it is the “active” spacecraft.
of advanced and robust control techniques and
The reference for the active spacecraft is the
implementation of a full on-board autonomy for
Coronagraph Spacecraft (CSC), which carries the
nominal FF manoeuvres.
relative position metrology system. These spacecraft
The main mission challenge is to keep the S/C
will autonomously break and reacquire the
formation, which functions as a distributed
formation at each perigee pass and demonstrate
telescope, with very high pointing and position
formation flying near apogee. The combined system
precision. This will be done near apogee with an
is expected to achieve a relative positioning
advanced GNC system that uses µm precision
accuracy of the order of 1mm over a separation
optical metrology (among other classical AOCS
range of 25m to 250m.
related sensors such as STR) and mN thrusters
complemented with reaction wheels for attitude
A Coronagraph System for inner corona observation
control.
is the selected payload. To allow the operation of
This paper shows an overview of the flight
this payload, the OSC satellite will fly in front of the
formation mission objectives and associated
Sun to cast a shadow across the CSC, eclipsing the
requirements. Afterwards, the H∞ design approach
Sun on a sustained basis.
for the fine relative position control during the
formation flying station keeping and manoeuvres is
PROBA-3 will additionally validate autonomous
shown.
rendezvous in a high elliptic orbit based on
navigation with optical camera. This technology is
1. INTRODUCTION required for a number of exploration missions like
1.1. PROBA-· MISSION OVERVIEW planetary sample returns.
PROBA-3 (PRoject for On-Board Autonomy) is the
most complex and ambitious of the ESA's The demonstration mission budget imposes several
technology demonstration missions and it is part of constrains on the development, verification and
the overall in Orbit Demonstration missions from operation approach. Efficiency will be maximized to
the ESA Technical Directorate. The main objective reduce costs under a novel pragmatic approach. The
of PROBA-3 is to demonstrate precision Formation PROBA-3 project will provide along with the
Flying techniques and technologies as precursor to demonstration mission a set of processes,
future ESA missions. As follow up of the PROBA-1 procedures, techniques and tools for development of
and PROBA-2 missions, the extension of the future Formation Flying missions.
autonomy concept for Formation Flying will also be
demonstrated.
1.2. PROBA-3 PERFORMANCE
It will fly several innovating technologies required REQUIREMENTS REVIEW
for Formation Flying of multiple spacecraft. The objective of the PROBA-3 mission can be
Formation Flying technology demonstration is a summarised in the form of performance
milestone for future small and large scale virtual requirements. The tables below (see Table 1 and
structure missions where these virtual structures will Table 2) show a summary of the values for the
allow small spacecraft, flying with a fixed relative Relative Displacement Error (RDE), the Absolute
geometry, to form giant structure-less instruments. Attitude Error (AAE) and the Absolute Attitude
Measurement Error (AAME). The definitions of the
Page 1 of 11
errors as presented in the tables below are taken integrated in the last term by SENER and
from in [1]. The following requirements for relative commanded from only one brain within the GNC
positioning (i.e., RDE) belong to a set of software. These two GNC modules are the
requirements, the so-called High-Precision Attitude Formation Flying GNC (FF-GNC, in charge of
and Position (HPAP), which have been defined to relative position control in nominal mission and also
include the maximum desirable precision on attitude of attitude in one experiment) and the Spacecraft
and relative position for the PROBA-3 mission. GNC (SC-GNC, in charge of attitude control in
Note that the lateral displacement error (RDE) has nominal mission and low level commanding to
been assumed to be approximately scaled with the actuators in all phases); the modes, states and the
relative Inter-Satellite Distance (ISD) between the schedule in both modules are controlled by the
two spacecraft. Formation Flying Management (FFM). The overall
responsibility for the FF Software, including the
The necessity to achieve even better performance development of FFM and FF-GNC has been
(mainly for some figures related to attitude) is under assumed by GMV (Spain) with SENER as
analysis in order to meet all the performance subcontractor that also takes the responsibility of the
requirements. Note that the attitude error of the Formation Flying Failure Detection Isolation and
CSC, which performs the metrology, will influence Recovery (FF-FDIR) function. The responsibility
the errors in the relative position thus leading to a for SC-GNC has been assumed by NGC Aerospace
coupling between attitude and position. (Canada) following a high level design and
specification by SENER.
Table 1: PROBA-3 Position Performance Requirements The GNC in both spacecraft will be very similar to
PROBA-3 each other where nominally the OSC is the master
Error Positioning Units Remarks
and the CSC the slave. The latter serves as the
Requirements
68%. Error along the origin of the reference coordinate system within
nominal formation which the formation control is performed. It is thus
RDEX 1.5 mm
direction (longitudinal the OSC that will be performing the translations for
error) the position control whereas the CSC will be
0.5 (ISD up to
40m)
drifting freely along the nominal orbit.
RDEY 68%. Lateral errors to the
0.73 (ISD up to
mm nominal formation
RDEZ
160m)
direction The FF-GNC includes all the functions for
1.2 (ISD up to determination of the formation translation
250m)
(relative position) and the formation attitude
(relative attitude), as well as the calculation of
Table 2: PROBA-3 Pointing Performance Requirements suitable and optimal paths to be applied in the
PROBA-3 control for achieving such position and
S/C Error Pointing Units Remarks attitude, and the computation of the control
Requirements commands. It receives the measurements
68%. Error
acquired from the metrology sensors and from
AAEX 300 arcsec around the line of
sight (LOS) the navigation module in SC-GNC for absolute
AAEY 68%. Across attitude and transmitted if necessary through
2.8 arcsec LOS, value in the ISL (Inter Satellite Link) to the FF-GNC in
CSC AAEZ each axis the companion spacecraft.
68%. Around
AAMEX 30 arcsec
LOS The SC-GNC includes the classical functions
AAMEY
1.25 arcsec
68%. Across of an Attitude and Orbit Control system
AAMEZ LOS (AOCS), with absolute attitude determination,
68%. Around paths calculation and absolute attitude control,
AAEX 900 arcsec
LOS
and actuators (including thrusters) low level
AAEY 68%. Across commanding. The commands to the thrusters
30 arcsec
AAEZ LOS
OSC are managed at this level in the so-called
68%. Around
AAMEX 300 arcsec
LOS
Actuators Manager, in such a way that the FF-
GNC transmits the commands in form of
AAMEY 68%. Across
AAMEZ
5 arcsec
LOS desired force (and torque in one experiment)
and they are decomposed in individual
commands to each of the thrusters within this
actuator manager. The same holds for the
1.3. GNC GENERAL ARCHITECTURE reaction wheels. It also includes the absolute
It has been decided that the GNC for both PROBA- position and velocity determination. The SC-
3 spacecraft is to be separated into two GNC is commanded by the FFM.
interconnected GNC modules, designed together,
although implemented by different companies, and
This separation is justified by the industrial simplification of the problem posing by tailoring it
organisation of PROBA-3 and mission to the mission specifics. However, at present, the
characteristics, where the two spacecraft must be decision has been taken to depart from the inertial
able to be operated in stack configuration and approach and to adopt an LVLH formulation for the
separately, where the SC-GNC is commanded by control design as that would lead to added benefits
ground without the FF-GNC and FFM in the loop. regarding the robustness analysis in later phases.
The nominal mission includes all modules working Additionally, it would also lead to a more general
together in the control loop to demonstrate the design approach applicable to other flight formation
Formation Flying capability. missions. The LVLH design approach is currently in
progress and simulation results are to be expected
The previous approach allows the on board GNC to soon. In this paper, the former inertial formulation
perform a wide set of manoeuvres and experiments, approach is presented and afterwards the steps and
either in Formation or in Stand Alone configuration implications towards the LVLH frame are briefly
and Safe configurations. As mentioned previously, described.
the GNC subsystems must perform various
experiments during the flight formation phase, The FF-GNC control and the SC-GNC control
leading to potentially different strategies to be together have the demanding task of maintaining a
developed. Two main distinct manoeuvres can be flight formation with a very high precision in
distinguished, which are denoted as the rigid and relative position and attitude described in section
loose manoeuvres. 1.2.
Rigid formation manoeuvres: where the The Guidance module will provide a position and
nominal attitude profile for each spacecraft has attitude profile, which is compared to the
to keep the two payload reference coordinate Navigation results. The Control will thus compute
frames parallel during the whole manoeuvre, in the necessary actions to follow the Guidance profile.
such a way that the high-accuracy metrology In order to improve the response and performance
system is able to operate throughout these during the manoeuvres, the controller consists of
manoeuvres. The formation will either change both a feed-forward term and feed-back term. The
the distance between S/C (resizing: from 150m feed-forward terms are derived directly from the
to 25m and back and from 150m to 250m and Guidance profiles and consist of the nominal forces
back) or the formation is rotated (retargeting and torques that are necessary to execute the
up to 30 degrees and back to the original manoeuvre. The feed-back term then compensates
position) or both at the same time. for any deviations during from the nominal path.
Formation loose manoeuvres: There are no For the FF-GNC control, the most demanding
stringent requirements on attitude and position manoeuvres are the rigid ones, where the high
to be met during the manoeuvre, only at the accuracy metrology has to be operating, within its
start and end of the manoeuvre. These narrow field of view. Throughout these rigid
manoeuvres will be applied with impulsive manoeuvres, only one spacecraft will be under
delta-Vs calculated in the Guidance module. active closed-loop position control while both
spacecraft will be actively controlled to keep the
These manoeuvres implying relative position attitude profiles as provided by the guidance
change will be executed by the OSC spacecraft that modules for each respective spacecraft, either for
carries the mN thruster system. The relative position the manoeuvres that the attitude has to be kept
measurements are taken by CSC and transmitted to inertial, Sun pointing or when a retargeting
the OSC via radiofrequency (with an Inter Satellite manoeuvre is commanded.
Link – ISL).
Modern robust control techniques, such as H∞ and
µ-analysis, are used to design and analyse the FF-
2. CONTROL DESIGN GNC controller. For the plant modelling, the
One of the discussions during the design phase has relative orbital dynamics can be simplified by
been the choice of the reference frame for the expressing the control problem with respect to the
problem posing and associated plant model. For nominal reference position of the OSC and
most rendezvous & docking missions a rotating considering the gravity gradient as a deterministic
orbital frame such as the Local Vertical Local perturbation that is to be compensated through an
Horizon (LVLH) formulation is more suitable. In additional feed-forward term. A very important non-
PROBA-3, the formation is to be maintained near- linear element is the presence of thrusters and its
inertially since it is related to the Sun direction, As a effects are carefully taken into account by designing
consequence, the use of an inertial formulation was an adequate modulation scheme and by making sure
thus at first adopted since it would lead to a that the H∞ design does not enter the frequency
range of the modulation. Naturally, the attitude and also retargeting, resizing and both together). The
position control are not completely independent objective is to achieve the same precision in
from each other for the formation control scheme. A pointing and position performances as in the
GNC architectural separation of attitude and nominal modes, where the attitude is controlled with
position exists hence the attitude dynamics is taken wheels and the relative position with mN level
into account in the FF-GNC design by both thrusters. In this experiment, the attitude control is
imposing performance requirements on the SC- fully integrated in the translational control within
GNC and to add the necessary robustness properties the FF-GNC for the OSC spacecraft. Consequently,
to the position controller synthesis and analysis. a simultaneous 6DoF design approach will be
adopted.
The attitude followed by the two spacecraft is
transferred at a certain frequency to the CSC 2.1. COUPLING BETWEEN POSITION
spacecraft to check the progress of the manoeuvre AND ATTITUDE
and to allow to the on-board Formation Flying- At system level, the decision has been made to
FDIR subsystem to decide what to do in case the separate the position and attitude control, which are
relative attitude and/or position exceeds the allowed respectively executed by the FF-GNC and SC-GNC.
thresholds. In fact, they are both designed by different parties
hence a combined and strict 6DoF approach is
The functions to be performed by the control discarded for nominal modes. Nevertheless, the
module within the FF-GNC during the loose reality is that there is an inevitable coupling between
formation manoeuvres is limited to take the the attitude and position dynamics hence a strict
reference delta-V provided by the Guidance module separation and independent design is not possible
to be decomposed in forces that will be commanded and some iteration between both parties will be
as a function of time. mandatory during the design and analysis process. In
order to cope with the coupling and to minimise its
In addition to this Formation Flying Control, each effect, the following two measures are introduced.
spacecraft has a set of standalone Attitude Control First, both attitude and position control are subject to
Modes. These modes are part of the SC-GNC, in requirements and restrictions such that they interact
which each satellite is commanded regardless the as little as possible with each other. Second, the
fact that it is part of a Formation or not. The attitude control is included in the position control
standalone Attitude Control Modes involve design as a perturbation source within the H∞ robust
autonomous Attitude Determination, path profiles design framework.
calculation and control functionalities to provide the
capabilities to get and maintain Sun pointing with a The most important requirement is to have different
set of simple and robust units and algorithms, to get controller bandwidths for the position and attitude
and maintain high accuracy inertial pointing, high control in order to avoid any dynamical interaction
accuracy Sun pointing (that is quasi-inertial leading to any undesired amplification of position or
pointing), to get and maintain target pointing (that is attitude dynamics. The attitude control is assumed to
used to face both spacecraft towards each other to be significantly faster than the position control.
acquire optical metrology) and to perform correction Consequently, any attitude related effects will
manoeuvres. behave as a high frequency disturbance.
These standalone modes are also active during the Ideally, when firing the thrusters, they would
formation phases, to allow a significant decoupling generate pure forces. But the presence of thruster
between Attitude Control and Translational Control. errors and the fact that the resultant net force vector
Note that this is decoupling in terms of calculation will not go exactly through the S/C mass centre,
of commands to thrusters and wheels, but not in the there will be residual perturbation torques that will
control problem posing, since the control at FF- affect the attitude. Consequently, the design of the
GNC level has to take into account the unavoidable position control should take this into account by
coupling with the attitude, including the attitude of avoiding large control forces and too much control
the other spacecraft. activity. This way, the attitude dynamics will be
perturbed less allowing the attitude controller to
There will be a challenging experimental control compensate for these impulsive thruster based
mode that will consist of a 6 degrees of freedom torque perturbations.
(6DoF) controller based only on thrusters as Additionally, the presence of a low frequency
actuators (i.e., no reaction wheels are included in the gravity gradient acceleration will lead to control
control loop in the active spacecraft) to be forces in a particular direction, which in turn may
performed by the OSC, while the other remains in lead to systematic torque perturbations in the low
inertial pointing (formation station keeping will be frequency region. Once this effect is characterised
demonstrated in orbit within this experiment, but and quantified, a specific disturbance rejection
requirement can thus be put on the attitude control extent be “visible” by the controller in the form of a
design. “hopping arc” signal.
The direction of the commanded forces depends on The “arcs” in the position error signal (see Fig.1)
the attitude. Therefore, any attitude deviation will can be considered a noise, which has to be
lead to small deviations in the desired force eliminated. The main problem lies with the time
direction. Considering the very stringent pointing derivative, which will have very large values that
requirements, these force errors are expected to be could significantly affect the performance of a linear
small and may perhaps be negligible for the controller. One could apply low pass filters to filter
commanded forces. In any case, this effect is to be out the undesired components but that could impose
studied and assessed. severe restrictions on the resulting allowable
controller bandwidth. Another option could be to
2.2. PLANT MODELLING explicitly generate a guidance profile to follow an
The OSC is subjected to a forced motion whereas estimated “arc-profile” and compute the appropriate
the CSC is free floating in a highly elliptic orbit. The error signal but this would introduce additional
relative orbital dynamics can be linearised around complexity and this is best avoided if possible. The
the CSC and could in principle serve as the plant approach that is currently proposed is to sample the
model of the OSC for the control synthesis. error signal at the points of the thruster actuation. In
However, it is easier and more convenient to centre other words, the error signal is sampled at the same
the plant dynamics at the origin of the OSC. After frequency as that of the PWM. A qualitative result is
all, in the nominal station keeping case, the OSC shown in Fig.2. Here, the blue signal represents the
needs to be controlled at a fixed relative position original “hopping” error signal. The red circles
with respect to CSC. The relative orbital dynamics indicate the points where the thrusters are actuated
will try to “push” the OSC away from its desired and here the instantaneous error signal is sampled.
position. Since the required position accuracy is at As a result, the green line is the signal that is fed to
mm level, these small variations will hardly affect the controller and it can be observed that the “arcs”
the magnitude and direction of the gravity gradient are no longer present. Both error signal and its time
acceleration around the nominal position. Therefore, derivative are now much smoother leading to cleaner
the relative orbital dynamics can be treated as a control commands. Naturally, this approach has the
deterministic acceleration field that is superposed on effect that the error signal is sampled at the PWM
the OSC. As a result, the plant model of the OSC frequency, which is indicated in the figure by the
can thus be reduced to a double integrator where the vertical dashed lines. But since the thruster PWM is
gravity gradient acceleration is estimated on-line and any way in the control loop, the bandwidth and
compensated through a feed-forward term. “effective” sample time is already driven by the
PWM. In any case, it must be highlighted that the
2.3. THRUSTER ACTUATION SCHEME Navigation Module does still make use of all
available measurement hence the Navigation does
Thrusters are a highly non-linear element whose take into account all the data along the arcs. It is
dynamical effects must be taken into account during only the Control that in a selective manner only
the design. Due to its non-linear properties, it is not chooses to take data at specific points. This concept
trivial to incorporate this actuator into a classical has been implemented in a simplified yet dedicated
linear design and analysis process. A common simulation environment for testing purposes and it
procedure is to apply a pulse width modulation has shown to work very well.
(PWM) scheme so that the linear control commands The describing Function technique can be employed
are converted into a series of pulses. to assess its impact on the stability margins. In any
case, in order to ensure a linear control behaviour
The presence of the gravity gradient acceleration using thrusters, the PWM frequency must be higher
will lead to control forces in a specific direction. The than that of the control bandwidth.
thrusters will therefore generate discrete pulses in
one direction to compensate a continuous force in
gravity gradient acceleration
the opposite direction. This leads to a “hopping” of
the position error signal as shown in Fig.1. x
frequencies do not coincide, the dynamics that Figure 1: Effect of thruster pulses within the gravity gradient
results from the thruster pulse pattern will to some acceleration field
G Plant model of the S/C and is represented
x by a double integrator.
WT This weighting function bounds the
complementary sensitivity function T and
time
is here primarily used to set an upper limit
of the bandwidth, which is critical in the
Figure 2: Error signal processing scheme through sampling at the presence of the thruster PWM.
PWM frequency
Additionally, it can be used to include
robustness properties but that aspect is
2.4.
H∞ SYNTHESIS AND PRELIMINARY
excluded in the current preliminary design.
RESULTS
The final objective is to design a robust controller WS This weighting function bounds the
based on H∞ synthesis and µ-analysis. The design sensitivity function S and it is here
process has started where the control problem is primarily used to impose a desired
analysed and adapted to the H∞ framework. As a suppression factor to the low frequency
first approach, a generalised plant model is obtained external disturbance forces.
with a set of suitable weighting functions but, at WKS This weighting function bounds the
least for now, without modelling the uncertainties maximum values of the control commands.
for robustness properties. The aim is to synthesize a When taking the transfer function from r to
preliminary H∞ controller in order to obtain both z2, the maximum control commands occur
qualitative and quantitative insight into the control at frequencies near the controller
problem and achievable performance. bandwidth, which are naturally much
A block diagram of the feedback system and higher than the operating frequency of the
weighting functions is shown in Fig.3. tracking command r, which is near-inertial
pointing. Nevertheless, this weighting
z1
WS function can be used to add additional poles
d
and zeros near and after the cross-over
Wd
WKS
z2 frequency for loop shaping and additional
filtering.
r
K G WT
z3 Wd This weighting function represents the
magnitude scaling and frequency range of
the external disturbance forces. Here, the
main disturbance of interest is the gravity
gradient acceleration.
n
Wn
Wn This weighting function represents the
magnitude scaling and frequency range of
Figure 3: Feedback block diagram with the weighting functions the Navigation errors.
for the H∞ synthesis
Magnitude (dB)
Magnitude (dB)
0 0
-10
Magnitude (dB)
Magnitude (dB)
-70
0 -75
suppression is chosen such that its influence on the Figure 4: Bode plots of the weighting functions
error signal is reduce to sub-mm level at this
maximum frequency within the WS weight function.
- Noise level: Currently, no quantitative information P
is available for the Navigation errors with high n
Wn z1
WS
accuracy metrologies. Consequently, some r
assumptions are to be made. The magnitude is d z2
WKS
assumed to be compatible with the mm-level
position accuracy. Since the Navigation uses explicit Wd WT
z3
z1 DS DG C d
DS C G 0 CS 0 0 DS D S DG D d 0 DS DG r
The synthesis results are shown in Fig.6. Here, the
C KS 0 D KS d
z2 0
z D D C
0 0 0 0
x
0 0
maximum value of the singular value decomposition
DT C G CT 0 0 0 0 DT DG Dd 0 DT DG n
3
v D C
G d
T G d
CG 0 0 Cn 0
I DG Dd Dn
DG u (SVD) of the input/output channels from r to zi and d
(2) to z1 are shown together with the corresponding
weighting functions. (Each of the input/output
The equations (1) and (2) represent a minimum channels is a 3x3 MIMO transfer function). The
realisation of the state-space model of the resulting H∞ norm of the complete transfer function
generalised plant P. It is expressed as the individual from all inputs to all outputs is equal to γ= 0.89.
state-space matrices A, B, C and D of the plant G
and weighting functions. The sub-indices of the
50
0
0
-50
Magnitude (dB)
Magnitude (dB)
-50
-100
-100
-150
-200
-150 S (bilin) T (bilin)
S (bilin back-transformed) -250 T (bilin back-transformed)
W-1 W-1
T
S
-200 -300
-6 -4 -2 0 2 -6 -4 -2 0 2
10 10 10 10 10 10 10 10 10 10
Frequency (rad/sec) Frequency (rad/sec)
50
0 0
-50
Magnitude (dB)
Magnitude (dB)
-50 -100
-150
-100 -200
-250
SGWd (bilin)
-150 KS (bilin) -300
SGWd (bilin back-transformed)
KS (bilin back-transformed)
-350
W-1 W-1
S
KS
-200 -400
-6 -4 -2 0 2 -6 -4 -2 0 2
10 10 10 10 10 10 10 10 10 10
Frequency (rad/sec) Frequency (rad/sec)
These results show that the main drivers that affect order in order to still provide enough suppression at
the controller synthesis are the combination of the the desired frequency while not conflicting with WT
upper bandwidth restriction in WT and the in terms of the allowable bandwidth region.
suppression factor in WS. In fact, WS had to be of 2nd
The resulting controller needs to be subjected to an requires a double integrator term to obtain the
order reduction, which consists of three steps. First, necessary suppression factor leading to a PII2D
a model reduction is applied based on the Hankel controller structure. It becomes clear that despite the
eigenvalues, which can be done directly with the low frequency of the gravity gradient term and
appropriate MATLAB tools. Second, the values of relatively low value, it is an important design driver
the poles are analysed and checked whether there are as the required position accuracy is very high. But it
any near zero indicating integral action. Since the has to be stressed that the imposed suppression
synthesis routine cannot obtain a pure integrator factor is very conservative as the feed-forward term
with poles at zero, it has to be imposed by hand. And was not taken into account. In reality, the necessary
thirdly, a check is made to see whether there are any suppression factor will be less as it only needs to
high frequency poles and zeros that can be compensate the residual errors of the feed-forward
eliminated altogether without affecting the controller compensation. In that case, it is possible to reduce
behaviour. The results are shown in Fig.7 after the controller further and eliminate the double
applying the order reduction. integral term and keeping only the PID term.
The order reduction has shown that there were two
poles near zeros indicating that the controller
r --> z1 - S vs. W-1
S
r --> z3 - T vs. W-1
T
100 50
50
0
-50
-50
Magnitude (dB)
Magnitude (dB)
-100 -100
-150
-150
-200
-200
-250 S, after order reduction T, after order reduction
W-1
S
W-1
T
-300 -250
-6 -4 -2 0 2 -6 -4 -2 0 2
10 10 10 10 10 10 10 10 10 10
Frequency (rad/sec) Frequency (rad/sec)
50
0 0
-50
Magnitude (dB)
Magnitude (dB)
-50 -100
-150
-100 -200
-250
-150 -300
KS, after order reduction SGWd, raw full order
-350
W-1
KS
SGWp, BST model reduction
-200 -400
-6 -4 -2 0 2 -6 -4 -2 0 2
10 10 10 10 10 10 10 10 10 10
Frequency (rad/sec) Frequency (rad/sec)
To validate the results in a time-domain simulation, well to represent sensor and actuator noise. The
a simple Simulink model has been created consisting simulation results for a 12 hour period are shown in
of the discretised controller, the S/C plant, a 5.8. Here, it can be seen that the position error is
representative PWM scheme applied to the control below 1mm hence the suppression factor is indeed
command to simulate the discontinuous thruster as effective as intended by the design process. (Note
effects and a sinusoidal external force perturbation that there is deliberately no feed-forward term,
representing the gravity gradient matching the which would in reality further improve the
maximum expected magnitude. Additionally, suppression). A zoom of the results shows the arcs
Gaussian noise values have been implemented as and “hopping” nature of the error signal, which is
caused by the combination of thrusters and external A more in-depth analysis is to be performed to
acceleration field. In Fig.9, both the real position assess what the optimal thruster PWM period is. At
error and the error signal as computed by the control the moment, 20 seconds has been chosen for station
are shown together. Here, it can be observed that the keeping but since the thruster modulation has a
proposed scheme of taking samples only at the critical impact on the performance, it needs to be
points of the thruster actuations works as intended as analysed thoroughly, mainly for retargeting
it eliminates the “arc” signal. manoeuvres where a value of up to 5 seconds could
be selected.
1
Simulation results: Position error vs time
The properties of the actuators and sensors need to
0.5
be consolidated (out of scope of Phase B) whenever
possible in order to be able to tune the controller
0
position error [mm]
-0.5
-1 synthesis accordingly.
-1.5 MIMO couplings need to be analysed in-depth and
-2
addressed during the design. Especially the coupling
between position and attitude. As a starting point,
-2.5
-3
0.44
0.38
0.36
0.34
The issue of robustness properties will be addressed
0.32 once the current system design is more mature
0.28
0.3
where both hardware (e.g. actuators and sensors) and
0.26
3.14 3.15 3.16 3.17 3.18 3.19
software (e.g. Navigation, Attitude control
time [sec]
x 10
4
algorithms) start to get consolidated. Adequate
Figure 8: Time domain simulation test (top: 12 hour period, modelling of the plant and its uncertainties will
bottom: Zoom )
become critical and µ-analysis will be used to
Real position error vs Control error signal analyse the robustness.
0.22 Real position error
Control error signal
0.2
2.6. LVLH FORMULATION
0.18
As mentioned previously, the decision has been
0.16
taken to adopt an LVLH formulation for the control
position error [mm]
0.14
design. This design approach has already started and
0.12 a promising controller candidate design is already
0.1 available. Time-domain simulations for validation
0.08 are pending and are to be expected soon. The main
0.06
differences and implications by adopting this new
2.36 2.365 2.37 2.375 2.38 2.385 2.39 2.395 2.4
approach will be described next.
time [sec] 4
x 10
3. CONCLUSION
A preliminary controller has been designed within
the H∞ framework based on an inertial formulation
for the problem posing. Although the design is at a
preliminary stage the results are very promising.
First, the design setup within the H∞ framework has
been developed and implemented in MATLAB
using the Robust Control Toolbox. Second, a
dedicated thruster PWM and error signal processing
has been proposed and implemented successfully.
And thirdly, a preliminary controller design has
been obtained yielding very useful insight into the
achievable performances, feasibility and tuning of
the weighting functions.