Conference PaperPDF Available

Magnetic Detumbling of Fast-tumbling Picosatellites


Abstract and Figures

The problem of pure magnetic detumbling of a fast-tumbling picosatellite is considered. A new weighted B-dot control algorithm is proposed. The algorithm enables power reduction while not sacrificing detumbling performance. Analytical expressions relating the maximal expected rotational rate to the minimum sampling time required are presented. Simulation results demonstrate the practical benefits of the proposed approach for picosatellites.
Content may be subject to copyright.
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
Magnetic Detumbling of Fast-tumbling Picosatellites
Robert Fonoda, Eberhard Gilla
aDepartment of Space Engineering, Delft University of Technology, Kluyverweg 1, Delft, The Netherlands 2629 HS
Corresponding Author:
The problem of pure magnetic detumbling of a fast-tumbling picosatellite is considered. A new
weighted B-dot control algorithm is proposed. The algorithm enables power reduction while not sacri-
ficing detumbling performance. Analytical expressions relating the maximal expected rotational rate to
the minimum sampling time required are presented. Simulation results demonstrate the practical benefits
of the proposed approach for picosatellites.
Keywords: Picosatellite, magnetic detumbling, B-dot, attitude control, stability
1. Introduction
The work presented in this paper is motivated by the
development of Delfi-PQ satellite [1]. With Delfi-PQ,
Delft University of Technology has entered the class of pi-
cosatellites and aims at pushing the boundaries of satellite
miniaturisation further. This next level of satellite minia-
turisation generates new research challenges and offers
innovation opportunities. The attitude control and deter-
mination subsystem of the Delfi-PQ is an essential corner-
stone and is, among others, responsible for the mission-
critical detumbling. It is anticipated, that the deployment
of Delfi-PQ satellite to low Earth orbit (LEO) will be ac-
complished by a newly designed, spring-loaded and not
in-flight tested, deployment system. Therefore, high ini-
tial angular rates up to 180 [deg/s] can be expected. Such
high rates will severely impact the communication and
power subsystem’s functionalities.
In the past decades, a fully magnetic attitude con-
trol became a viable and interesting option, especially for
low-cost nanosatellites [2–5]. However, developments of
fully magnetic control for highly miniaturised picosatel-
lites are still in their infancy. Only recently, preliminary
simulation results on successful magnetic detumbling of a
slow-tumbling (6[deg/s]) picosatellite (5×5×5[cm])
have been reported in [6].
Pure magnetic detumbling of a fast spinning picosatel-
lite creates new challenges, especially due to limited on-
board processing and sensing capabilities, magnetorquers
saturation, scarce power resources, and inherent under-
actuation. Driven by these constraints, it is obvious that
the detumbling shall be done in a robust and efficient way.
The widely adopted and simple-to-implement B-dot
algorithm [7] is, in theory, able to detumble a rotating
satellite from any initial angular rate down to approxi-
mately twice the orbital rate. However, if the sensing or
actuation cycle of the detumbling algorithm is not suffi-
ciently small, the satellite might actually spin up instead
of detumbling. Based on Nyquist criterion and rigorous
controllability analysis, we provide analytic expressions
relating the critical rotational rate of the satellite to the
minimum sampling time of the control and sensing loop
necessary for successful detumbling.
The B-dot algorithm was recently modified by
Avanzini and Giulietti [8] and a reasonable static B-dot
gain was derived yielding a quasi-minimum detumbling
time. However, for fast tumbling satellites, this B-dot gain
needs to be parametrised to efficiently utilise the mag-
netorquers duty cycle throughout the entire detumbling
phase. In this work, we propose a new weighted B-dot
control algorithm, where the B-dot gain does not remain
constant, but is parametrised by a so-called tumble pa-
rameter. This parameter is able to estimate the satellite’s
absolute tumbling rate using magnetometer readings only.
Additionally, the vector version of this parameter is able,
in a robust manner, to indicate whether the satellite is
detumbled or not. A simple sensor filtering approach is
introduced to mitigate the impact of sensor noises and bi-
ases caused by on-board electronics. Monte Carlo simula-
tion results are presented to demonstrate the effectiveness
of the proposed detumbling algorithm.
Notations: In this paper, bold italic face denotes vec-
tors and matrices; (·)Tstands for transposition; Rn×mde-
notes a set of n×mreal matrices, Rnrepresents the set
of ndimensional real column vectors; Nthe set of natural
numbers (including {0}); k·kstands for Euclidean vector
norm; |·|denotes absolute value; and N(µ,Σ)denotes a
density function of a multivariate normal distribution with
a mean vector µand covariance matrix Σ.
IAC-18-C1.3.11 Page 1 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
2. Background
In this section, we will introduce the essential back-
ground that allows to study and design a magnetically
controlled satellite.
2.1 Coordinate Frames
To represent satellite dynamics, we will use two main
reference frames, namely an Earth centred inertial (ECI)
reference frame and a satellite body-fixed (body) refer-
ence frame.
The origin of the ECI frame is located in the centre
of mass (CoM) of the Earth. The z-axis of this reference
frame points to the north pole of the ecliptic, the x-axis
points towards the vernal equinox (first point of Aries),
and the y-axis completes the right-handed orthogonal sys-
tem. This reference frame will be denoted as FI.
The origin of the body frame is located in the CoM
of the satellite. The axes of this frame coincide with the
satellite’s principal axes of inertia. Hereafter, the body
frame will be denoted as FB.
2.2 Rotational Dynamics
The rotational dynamics of a rigid body satellite
around its CoM are given by Euler’s moment equation:
B/I (t) = ωB
B/I (t)×IωB
B/I (t) + τB
where ×denotes the cross product, ωB
B/I R3repre-
sents the angular velocity vector of frame FBrelative to
FIdecomposed in FBcoordinates, IR3×3is the in-
ertia matrix, and τB
ext R3is the vector of total external
torque acting on the satellite in FBcoordinates. To avoid
extensive notation, the angular velocity vector, ωB
B/I , will
be simplified to ωB,[ωB
The total external torque in (1) can be decomposed as1
ext =τB
where τB
cis the net disturbance torque and τB
dis the con-
trol torque. The net disturbance torque is divided into the
following four components:
a- is the torque due to residual atmospheric drag,
s- is the torque due to solar and albedo radiation pres-
g- is the gravity gradient torque due to asymmetric
distribution of the mass moment of inertia,
m- is the torque due to residual magnetic dipole and
Lorenz forces.
The explicit physical model of the above disturbance
torques can be found, for instance, in [9, 10].
2.3 Magnetic Actuation
A magnetic dipole vector mB
cR3in FB, generated
by 3 orthogonally mounted electromagnetic coils aligned
with the satellite’s principal axes of inertia, can be mod-
elled with reasonable accuracy as [11]
where Njis the number of windings around the j-th coil,
ijis the current provided in the j-th coil, and Ajis the
area of the j-th coil. More accurate magnetic dipole
model can be found in [12].
A non-zero magnetic dipole vector (4), generated by
satellite body-fixed magnetorquers, will interact with the
local magnetic field and generate a control torque vector
caccording to
where bB
ER3is the local Earth magnetic field vector
expressed in terms of FBcoordinates.
The torque τB
ccan be used, among others, to damp
the angular momentum of the satellite close to zero or to
de-spin reaction wheels.
2.4 Sensor Model
The local Earth magnetic field bB
Ecan be measured
on-board the satellite using a triaxial magnetometer sen-
sor. In this paper, for noise filtering purposes and with-
out loss of generality, we assume availability of multiple
magnetometers on-board. The model of the i-th magne-
tometer reading, z(i)
raw R3, at discrete time instances
tk,kTs,kN, can be expressed as follows:2
raw (tk) = hT(i)
B/S bB
E(tk) + b(i)
bias +v(i)(tk)iLSB ,(6)
where v(i)R3is a random vector satisfying
v(i)∼ N(0,Σi)and aiming at modelling sensor noise
of the i-th magnetometer, b(i)
bias R3is the true static bias
of the i-th magnetometer, and T(i)
B/S R3×3is a rotation
matrix from FBto the i-th sensor frame. Note that TB /S
can be also used to account for sensor misalignment and
scaling errors. Finally, [·]LSb denotes a rounding function
driven by the resolution of the magnetometer with respect
to the least significant bit (LSB).
1To further simplify notation, the time-dependency ”(t)” of variables will be omitted if there is no ambiguity.
2Readings from multiple sensors are assumed to occur simultaneously or with minimum delay. The sampling time Tswill be defined in Section 4.
IAC-18-C1.3.11 Page 2 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
It is important to note that the above model excludes
time-varying biases and assumes magnetorquers being
switched off and properly desaturated when taking mea-
surements3. The effect of the magnetorquers on the mea-
surements can be easily incorporated into (6) by adding a
coupling matrix multiplied by the commanded magnetic
dipole moment mB
cas done in [13].
2.5 Problem Formulation
In this paper, we aim at reducing the angular rate
of the satellite from a non-zero initial value ωB(t0),
06=0below a desired absolute angular rate ωdes >0
for all three axes, i.e.,
i(tdet)ωdes ,i∈ {x, y, z},(7)
where tdet 0denotes the detumbling time defined as the
time when the condition (7) is satisfied for the first time.
The above aim is to be achieved using a pure magnetic
control and feedback provided from the three-axis magne-
tometers only. The resulting algorithm shall be simple to
implement on a low-performance on-board computer and
shall take into account physical limitations of the actua-
tors and sensors. The algorithm shall also be able to “as-
sess” the rotational speed of the satellite and shall be able
to make efficient use of the magnetorquers (especially at
the beginning of the detumbling phase) while not increas-
ing the detumbling time significantly.
Note that, due to physical limitations of the considered
setup (pure magnetic actuation and sensing), the satel-
lite’s angular rate can be decreased only until it reaches
values of approximately twice the orbital rate [14]. Re-
cent research suggests that a rotational rate of about 1.8
of the orbital rate can be achieved around the axis of max-
imum moment of inertia [15]. To make (7) realistic for all
axes, we suggest ωdes to be selected such that it satisfies
ωdes 2ωo, where ωois the orbital rate.
3. Controller Design and Stability
In this section, we introduce a new weighted B-dot
controller and briefly summarise the well-know Lyapunov
stability results applied on a rigid-body satellite dynamics
actuated by the proposed control law.
3.1 Weighted B-dot Control Algorithm
Many variants of the B-dot controller have been pre-
sented in the literature. The simplest one is given by a
static linear feedback of the form [7]
where kc
b>0is the constant controller gain, ˙
Eis the
time derivative of bB
E(hence the name B-dot), and mB
is the desired dipole moment to be generated by the mag-
netorquers. In an ideal case mB
d. In practice,
however, the commanded dipole moment might be differ-
ent from the desired one. This is due to magnetorquers’
physical limitations (e.g., saturation) which need to be ad-
dressed. Control efficiency of the B-dot controller im-
proves as the orbit inclination angle increases [16].
Although (8) is simple to implement [17], the choice
of the control gain kc
bis not straightforward. A high
value may lead to frequent saturation of the magnetor-
quers, whereas a low value might lead to an inefficient use
of the magnetorquers and very long detumbling time tdet.
Additionally, a static gain is not a good practical solution
for both, very high and at the same time low angular rates,
as the desired dipole moment is proportional to the abso-
lute change of the magnetic field, i.e., kmB
dk ∝ k˙
which directly relates to the angular rate of the satellite.
In this paper, we adopt the structure of the modified
B-dot controller proposed by Avanzini and Giulietti [8],
where the desired dipole moment is computed as
where ˙
Eis the time derivative of the unit vector ˆ
Ek. This unit vector is parallel to the Earth’s
magnetic field vector bB
E, expressed in FB. In practice,
the time derivative of ˙
Eis difficult to measure directly.
However, as will be shown in the next section, ˙
be estimated by differentiating the magnetic field vector
obtained from the body-mounted magnetometers.
In (9), instead of using a static control gain as in [8],
we parametrised the control gain by p, the so called tum-
ble parameter, as follows
b(p) = k
ϕp +ε,0p1,(10)
where 0ϕ < is the rate factor, 0< ε 1is the
tuning parameter, and k
bis the sub-optimal static gain de-
rived in [8] as
b= 2ωo(1 + sin(ξ)) Imin,(11)
where ω0is the orbital rate (mean motion), ξis the incli-
nation of the orbit plane relative to the geomagnetic equa-
tor and Imin is the minimum mass moment of inertia of
the satellite. It is obvious that for ϕ= 0 and ε= 1,
(9) becomes equivalent with the modified B-dot law of
Avanzini and Giulietti [8]. A pure static gain k˙
b(p) = k
results in a quasi-minimum detumbling time [8].
3This assumption arises from the fact that we limit the magnetorquers duty cycle and allow extra time for the magnetorquers to desaturate, see
the discussion in Section 4.
IAC-18-C1.3.11 Page 3 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
The tumble parameter prepresents the satellite’s abso-
lute sensable angular rate and can be related to the satel-
lite angular velocity as (if |ωB
i|≤|ωmax |,i∈ {x, y, z})
where ωmax is the maximal absolute rotational rate of the
satellite that the on-board magnetic sensor suite is able to
sense without aliasing effects, see the discussion on ωmax
in Section 4. The on-board computation of pusing mag-
netometer data will be detailed in the next section.
Note that the sign of the desired moment (9), on a
per-axis basis, is opposite to the rate of change of the
magnetic field along the respective axis and its magni-
tude is inversely proportional to the tumbling parameter
p. The motivation behind this design choice is to reduce
the power consumption of the magnetorquers, especially
during the early stages of the detumbling mode (most
critical phase), while not sacrificing the detumbling per-
formance measured by tdet. As will be shown later, by
proper choice of ε, the variable control gain k˙
b(p)can be
tuned to approach the sub-optimal gain k
bas the detum-
bling phase approaches its end (ttdet).
Remark 1. Similarly as the classical B-dot controller
given in (8), the weighted B-dot controller (9) does not re-
quire any knowledge of the inertia Iand is able to detum-
ble the satellite to approximately twice the orbital rate.
3.2 Stability Sketch of the Weighted B-dot Controller
We will neglect the disturbance torques (3) from the
following analysis as their influence during the detum-
bling phase is usually very minor. Hence, for τB
the equation of motion (1) simplifies to
From a Lyapunov stability perspective, the aim is to
drive ωB0as t→ ∞ for any ωB
stability). To proceed, we define a positive definite Lya-
punov function candidate as
V(ωB) = 1
where the right hand side of (14) represents the true ki-
netic energy over the rigid body. Note that the inertia
matrix Iis symmetric positive definite and has positive
eigenvalues called principal moments of inertia [9].
To apply the 2-nd theorem of Lyapunov, we take the
time-derivative Bd/dt(·)of the kinetic energy (14) as
(note that Iis constant in FBand the right hand side of
the Lyapunov function is frame-independent)
V(ωB) = 1
ωB)TIωB+ (ωB)TI˙
= (ωB)TωB×IωB+τB
= (ωB)TτB
It is obvious from (15) that, if τB
d=0, the change
of kinetic energy is due to the control torque only and it
decreases only if the dot product of the angular velocity
and control torque is negative, i.e., (ωB)TτB
Using the basic theorem of kinematics, we have
dtTI/B ˆ
E=TI/B ˙
where TI/B is a transformation matrix from FIto FB.
Since ˙
Evaries at most at angular rate of approximately
2ωo, for high angular rates ωB, the rate of change of the
magnetic field in the body frame is mainly due to the ro-
tation of the satellite. Hence, for kωBk  2ωo, the right
hand side of (16) can be approximated as
E≈ −ωB׈
Substituting the control torque (5) into (15) and using the
fact that aT(b×c) = bT(a×c), yields
V(ωB) = (mB
Substituting (17) in (9) and assuming mB
case), the time derivative of Lyapunov function yields
V(ωB) = ωB׈
Despite the fact that k˙
b(p)>0for all p[0,1], it is ob-
vious from (19) that ˙
Vis only negative semi-definite, i.e.,
V0. The kinetic energy associated with the compo-
nent of ωBparallel to bB
E(or equivalently to ˆ
E) cannot
be reduced. In other words, the control law (9) can effec-
tively remove the satellite’s angular momentum whenever
ωBis not parallel to bB
E, while leaving it unchanged when
Eis zero [18]. This is due to the fact that the mag-
netorquers cannot provide torque along bB
Eas well as the
magnetometer cannot sense this motion [7].
Thanks to the the orbital motion of the satellite, the
Earth’s magnetic field vector direction, ˆ
E, does not re-
main fixed and global asymptotic stability (i.e., kinetic
energy strictly decreasing) of the weighted B-dot con-
troller can be proven by averaging-based analysis [18],
by Lyapunov-like lemma derived from corollary to Bar-
balat’s lemma [8], or by Krasovskii-LaSalle theory of pe-
riodic systems [2, 18].
Remark 2. A natural design choice of τB
cin (15) could
be to generate torque according to
c=kωωB, kω>0.(20)
This choice ensures that (15) is strictly negative, thus
asymptotic stability guaranteed. However, a control law
based on (20) would require angular rate measurements
or its estimate. The latter can be accomplished by a
IAC-18-C1.3.11 Page 4 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
computationally simple estimator based on magnetome-
ter data only [19]. Unfortunately, (20) is not applicable
in combination with pure magnetic actuation as the mag-
netorquers are only able to generate torques orthogonal
to the Earth’s magnetic field, see (5).
4. Discretised Algorithm
The weighted B-dot control law (9) needs to be imple-
mented in the satellite’s on-board computer as a discrete-
time algorithm. In this section, the most important aspects
of the on-board implementation are discussed.
4.1 Noise Filtering
One of the most cost-efficient way to filter out noise
from the raw measurements (6) is to take the (weighted)
average as follows
E(tk) =
raw (tk) + ˜
where nsis the total number of available sensors, ˜
is the filtered Earth’s magnetic field in FBat discrete
time tk,wi0are non-negative weights satisfying
w1+. . . +wns= 1,˜
S/B R3×3is an estimate of the
rotation matrix from the i-th sensor frame to FB, i.e., of
B/S )1, and ˜
bias R3is an estimate of the i-th sen-
sor bias vector b(i)
bias. Both ˜
S/B and ˜
bias need to be esti-
mated by an adequate calibration technique. The weights
wican be used to reward or penalise measurements based
on the accuracy of the respective sensor or fully disregard
faulty sensors by setting the respective weight to zero.
Ideally, sensors shall be placed apart to allow averaging
out undesired magnetic fields generated by various on-
board electronics.
4.2 Normalised B-dot Computation
If two consecutive measurements at tk1and tk
are available and the Nyquist criterion holds (see Sec-
tion 4.5), then the normalised B-dot vector in (9) can be
computed as
E(tk) = ˆ
where Ts,tktk1is a constant sampling time be-
tween two consecutive measurements and ˆ
E(·)is the
normalised magnetic field computed as
E(·) = ˜
Clearly, both ˙
Eand ˆ
Eare estimates. The tilde symbol
is left out above these vectors to simplify notation.
4.3 Tumble Parameter
To compute the scalar tumble parameter pfor (9), we
use the following discrete-time filter
p(tk) = αTs
E(tk)k+ (1 α)p(tk1),(23)
where α > 0is the filter time-constant, ˙
Eis computed
as in (22), and the initial value of the filter, p0,p(t0),
satisfies 0p01.
It is obvious from (22) and (23) that
0≤ k˙
Ek ≤ 2/Tsp[0,1],(24)
thus (23) satisfies the condition on pin (10). Thanks to
the utilisation of the normalised B-dot vector in (23), the
tumble parameter becomes altitude insensitive.
4.4 Discrete-time Controller
In this paper, we will assume constant currents for
each axis. Only the current direction and duration will
be controlled. Therefore, the absolute dipole moment
achievable in any direction and per axis is
Once the desired magnetic dipole moment vector
d;z]Tis computed using (9), the
amount of time each magnetorquer shall be activated is
determined by
i=δTsmin 1,|mB
¯mi!, i ∈ {x, y, z},(26)
where ton
i0is the “on-time” (duration) of the i-th mag-
netorquer and 0< δ < 1is the duty cycle of the magne-
torquers within one control loop. To simplify implemen-
tation, we assume that one control cycle equals exactly to
the duration of the measurement sampling time Ts.
The current direction in which the respective magne-
torquer shall be activated is given by
isign mB
d;i, i ∈ {x, y, z},(27)
where don
i∈ {−1,0,1}is the i-th magnetorquer’s cur-
rent direction and mpol
iaccounts for the winding polarity
i∈ {−1,1})or failure (mpol
i= 0) of the i-th mag-
We assume that all the computations related to (26)
and (27) are done with minimum delay and the magnetor-
quers are commanded immediately after that. The mag-
netorquers are switched off once the time they were sup-
posed to be active has passed. The duty cycle δshall be
IAC-18-C1.3.11 Page 5 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
selected carefully to allow enough time, (1 δ)Ts, for
the magnetorquers to de-saturate and to account for the
delayed control phenomenon. For the latter, see the dis-
cussion in the next subsection.
4.5 Sampling Time Selection
It is evident from (16) that the satellite’s angular rate
is directly linked with the change of the normalised vector
of the Earth magnetic field in FB. From the Nyquist crite-
rion [10], the maximum frequency of changes in the mag-
netic field that can be sensed without experiencing alias-
ing is 1/(2Ts). Therefore, the maximal rotational rate that
can be measured from two consecutive measurements is
π/Ts[rad/s], which yields to the following sampling time
selection criterion
where ωmax is the maximal expected rotational rate of the
satellite in any axis and any direction.
From a control theory perspective, we need to care-
fully analyse the effect of the sampling time Tsand the
duty cycle δon the stability of the discretised B-dot con-
trol loop. In the subsequent analysis, we will follow sim-
ilar steps as done in [17]. First, we will assume a rotation
around the z-axis only, i.e., ωB= [0 0 ωB
z]T. In ad-
dition to that, we also assume the measurements being
error-free, i.e., ˜
E(·) = bB
E(·), and that (17) is a valid
approximation. Then, we can analytically expressed two
consecutive magnetic field measurements as:
E(tk1) =
E(tk) =
zTs) sin(ωB
zTs) 0
zTs) cos(ωB
zTs) 0
0 0 1
Subtracting the above two vectors yields
zTs)1 sin(ωB
zTs) 0
zTs) cos(ωB
zTs)1 0
0 0 0
It is clear that the above measurement vectors have the
same magnitude, i.e., kbB
E(tk1)k. Hence,
the normalised B-dot vector can be expressed as
E(tk) = 1
The instantaneous torque at time t=tk, generated by
the desired magnetic dipole moment mB
d(tk), is given by
(32). By examining the last row, it is evident that in or-
der the torque τB
c;zbeing opposite to the angular rate ωB
zTs|needs to be smaller than π. This leads again to the
Nyquist criteria (28).
Suppose that the dipole moment mB
d(tk)is applied
for the entire duty cycle of the magnetorquers, i.e., during
the interval t[tk, tk+δTs). By the end of this inter-
val, the magnetic field in the body frame has changed to
E(tk+δTs)as well as the instantaneous torque gener-
ated by the magnetorquers changes to τB
c(tk+δTs). In
order to reduce the rotation, the integral of the torque ap-
plied over the entire period t[tk, tk+δTs)should be
negative. Suppose that k˙
b(p)is constant during this in-
terval, then for the z-axis, the negative torque condition
[sin(ωzt)sin (ωz(δTs+t))] dt
= 2κcos2(ωzδTs)cos(ωzδTs)
x(tk1))2+ (bB
Suppose that the z-axis is rotating at rate ωmax, i.e.,
ωz=ωmax. Then, to satisfy the inequality (30), the sam-
pling time Tsneeds to satisfy the following condition
c(t) = mB
E(t) = k˙
y(tk1)(1 cos(ωB
zTs)) + bB
x(tk1) sin(ωB
zTs)1) + bB
y(tk1) sin(ωB
x(tk1))2+ (bB
IAC-18-C1.3.11 Page 6 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
Note that the above process can be repeated for rota-
tions around the remaining two axes.
To summarise the above analysis, in order to avoid
aliasing and to be able to apply (on average) a torque in
opposite direction to ωB, the sampling time Tsshall be
selected such that both conditions, (28) and (31), are si-
multaneously satisfied for the maximal expected rotation
rate ωmax. For δ= 0.5, the two conditions become equiv-
alent. The negative torque criterion becomes a driving
factor if δ > 0.5.
4.6 Stopping Criterion and Tuning
To decide when the detumbling mode shall stop, one
can use a simple threshold logic applied on the scalar
tumble parameter pgiven by (23). However, in order to
make sure that all axes have detumbled below a desired
rate ωdes, we propose a vector version of (23) defined as
pv(tk) = αTs
E(tk)|+ (1 α)pv(tk1),(33)
where pv,[pv
z]Tis the vector-valued tumble
parameter satisfying pv(t0),pv
0. The rest of the param-
eters are the same as in (23). Note that the absolute value
|·|in (33) is applied element-wise.
The satellite is considered to be detumbled once the
following inequality is satisfied (in all axes)
i(t)¯p, i∈ {x, y, z},(34)
for a certain period of time t[tkNw, tk], where
NwNis a confirmation window. Nwis introduced
to enhance robustness. In (34), ¯p > 0is a threshold pa-
rameter tuned such that the above inequality is met when
the angular rate of the satellite is approximately ωdes in
all axes. Care has to be taken when setting ¯pin order
not to exit the detumbling mode prematurely. Thanks to
the use of the normalised B-dot in (33), the tuning of ¯p
becomes altitude independent.
The threshold logic (34) provides some hints for the
variable control gain’s (10) parameters selection. A large
rate factor ϕ1shall be chosen to penalise power
consumption at the initial phase of the detumbling. If
ϕ < 1/3¯p, then the parameter εcan be selected as
ε= 1 3ϕ¯p. This choice ensures that the B-dot gain is
close to the sub-optimal value, i.e., k˙p(p)k
˙p, when the
satellite is near to be detumbled.
5. Simulation Results for Delfi-PQ
In this section, we will demonstrated the performance
of the proposed B-dot algorithm to detumble a fast-
tumbling picosatellite. We consider the current picosatel-
lite project of the Delft University of Technology, the
Delfi-PQ project [1], as the baseline for our simulation
5.1 Simulation Parameters
We simulate a LEO and Sun synchronous Keplerian
orbit and use the world geodetic system 1984 (WGS84)
standard. The orbit-related parameters are given in Ta-
ble 1. These orbital parameters imply an orbital period
of approximately 1h and 32 min. The 12-th generation
of the international geomagnetic reference field (IGRF)
model was used to model the Earth’s magnetic field vec-
tor in the body frame as
where TIGRF /B is a rotation matrix from the IGRF to the
body frame FB.
Table 1: Orbit-related parameters
Parameter Value Unit
Altitude 350 [km]
Eccentricity 0 [-]
Inclination 96.85 [deg]
RAAN 310 [deg]
Earth magnetic field IGRF 12 [T]
Air density (low solar) 2.01 ·1012 [kg·m3]
Table 2 gives an overview of the satellite-related pa-
rameters and their considered uncertainty ranges which
were considered for the Monte Carlo (MC) simulations.
The listed parameters represent realistic sensor and actu-
ator characteristics. We consider two identical-type sen-
sors being available on-board. Both sensors are corrupted
with noises and suffer from unknown biases and quantiza-
tion errors. For the magnetorquers, we model the effect of
fall/rise times on the resulting magnetic dipole moment.
IAC-18-C1.3.11 Page 7 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
Table 2: Satellite-related parameters
Parameter Nominal value Unit Uncertainty (1σ)
Mass 600 [g] 16.67%
Inertia principal moments {Ix, Iy, Iz} {1.731,1.726,0.264}[g·m2] 21.67%
Magnetorquer dipole moment∗∗ 0.002 [A·m2] 15%
Residual dipole moment vector∗∗∗ magnitude 0.0001 [A·m2] 10%
Centre of pressure offset from CoM [5.4 2.08.2] [mm] 10%
Drag coefficient 2.1 [-] N/A
Cross-sectional area {x, y, z}face {92.1,122.9,25.2}[cm2] N/A
Rise/fall times of the magnetorquers 0.01 [s] N/A
Magnetometers’ noise rms∗∗ 500 [nT] N/A
Magnetometers’ resolution∗∗ 300 [nT/LSb] N/A
Magnetometers’ bias vector∗∗∗ magnitude 400 [nT] N/A
16.67% due to the mass variations plus 5% individual uncertainty per axis; ∗∗ for all three axes;
∗∗∗ direction uniformly distributed
The detumbling algorithm-related parameters are
given in Table 3. The selected sampling time complies
with both (28) and (31) criteria for our maximal expected
angular rate of 180 [deg/s]. The threshold parameter ¯p
was tuned such that detumbling shall stop when ωdes 5
[deg/s] in all axes. The confirmation window Nwwas
selected to reflect a 30 minutes confirmation time. We as-
sume that the magnetometers are not pre-calibrated, i.e.,
we do not account for the static biases which are present
in the simulations.
Table 3: Detumbling algorithm-related parameters
Parameter Value Parameter Value
Ts0.25 pv
δ0.6 ¯p8.1250 ·103
α1/200 k
b1.2074 ·106
w1=w21/2 ˜
bias =˜
bias [0 0 0]T
For all simulations, the satellite is assumed to be de-
tumbled (tded) when the condition given in (7) is satisfied
for ωdes = 5 [deg/s] (for all axes simultaneously). Where
applicable, we also present the detumbling time estimated
by the algorithm, see Section 4.6.
5.2 Sample Run Demonstration
Before turning to a statistical MC analysis, we present
results from a sample run example. In this example, the
satellite has nominal parameters and an initial rate of 180
[deg/s] in all three axes. The simulation runs for 16 orbits.
Figure 1 depicts the time behaviour of satellite’s angu-
lar rates. It can be seen that the proposed algorithm is able
to detumble a fast-tumbling picosatellite within less than
a day. The algorithm confirmed detumbling with a delay
of approximately half an hour. During the confirmation
period, the satellite was still in active detumbling mode.
Therefore, the angular rates around the detumbling con-
firmation are actually smaller than the desired rate ωdes.
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
Angular rates [deg/s]
Detumbled after: 12.2 orbits = 18.6 hrs
Algorithm confirmed after: 12.5 orbits = 19.1 hrs
B-4 [deg/s]
Figure 1: Satellite angular rate in body frame and detum-
bling performance from 180 [deg/s] in all three axes.
The weight that multiplies the sub-optimal B-dot gain,
i.e., k˙p(p)/k
˙p= (ϕp+ε)1, and the scalar tumble param-
eter are depicted in Fig. 2. The measured magnetic field
vector and the vector-valued tumble parameter together
with the threshold parameter are shown in Fig. 3.
IAC-18-C1.3.11 Page 8 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
1Weighting of the optimal B-dot gain
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
Scalar tumble parameter
Figure 2: Sub-optimal B-dot gain weight (top) and scalar
tumble parameter (bottom).
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
Figure 3: Measured magnetic field vector (top) and
vector-valued tumble parameter (bottom, y-axis in log
It is obvious from (25) and (26) that the magnetor-
quers power consumption is proportional to the amount of
on-times. Hence, summing the magnetorquers on-times,
i(tk), i ∈ {x, y, z },
can give us an indication about the power consumption of
the algorithm. Figure 4 shows the commanded on-times
per axis. The total opening time value, Ton
tot, is also shown
for each axis.
0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
Figure 4: Normalised magnetorquer ON-times and sum
of the opening ON-times until tdet.
5.3 Monte Carlo Analysis
In this subsection, we will present results based on
500 Monte Carlo simulation runs. For each run, we used
different noise seeds and varied some satellite-related pa-
rameters according to the normal distribution and uncer-
tainty characteristics given in Table 2. All simulations
started with an initial angular rate of 180 [deg/s] in all
three axes.
Table 4 shows the worst-case magnitudes estimate for
various disturbance torques. It can be seen that the so-
lar radiation pressure and gravity gradient disturbances
have the lowest magnitude, hence they will not be con-
sidered for the MC analysis. Note that the considered at-
mospheric disturbance corresponds to a low solar activity
scenario (air density: ρlow = 2.01 ·1012 [kg·m3]).
Table 4: Worst-case disturbance torque magnitudes
Disturbance Magnitude [N·m]
Aerodynamic drag 1.91 ·108
Gravity gradient 2.31 ·109
Residual dipole 5.03 ·109
Solar radiation 8.82 ·1010
Figure 5 depicts the histogram and a Gaussian fit of
true detumbling times tdet obtained from the MC simula-
tions. It can be observed, than on average, the proposed
weighted B-dot law is able to detumble the satellites in
less than 15 orbits.
IAC-18-C1.3.11 Page 9 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
5 10 15 20 25
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
14 orbits / 21.4 hours
14.8 orbits / 22.6 hours
3.2 orbits / 4.8 hours
Detumbling time
Gaussian fit
Figure 5: Histogram of the true detumbling time tdet .
Figure 6 depicts the detumbling time correlations with
the principal moments of inertia and respective maximal
dipole moments of the magnetorquers. As expected, there
is a significant positive correlation with increasing mass
moment of inertia, but relatively small negative correla-
tion with the available dipole moment of the magnetor-
1 1.5 2 2.5
R = 0.55
1 2 3
R = 0.47
1 2 3 4
R = 0.52
1.5 2 2.5
R = -0.02
1.6 1.8 2 2.2
R = -0.03
1.5 2 2.5
R = -0.12
Figure 6: Detumbling time correlation with principal mo-
ments of inertia (top subplots) and with maximal dipole
moments of the magnetorquers (bottom subplots).
In order to compare the proposed B-dot controller
with the modified B-dot controller of Avanzini and Giuli-
etti [8], we run additional 500 MC simulations for the
same scenario as described above, but employing (9) with
k˙p(p) = k
˙p. The results are compared in Table 5. As can
be seen from this table, there is only a slight increase in
the mean detumbling time, namely 0.45%, while the aver-
age power consumption of the magnetorquers is reduced
Table 5: Detumbling performance comparison for k˙p(p)
and k
˙pfrom 180 [deg/s] (around all three axes). Results
are given in hours.
Weighted B-dot B-dot of [8]
µ σ µ σ η
tdet 22.63 4.83 22.52 4.85 0.45 %
tot;x10.98 2.50 11.72 2.51 -6.32 %
tot;y11.36 2.88 12.02 2.87 -5.43 %
tot;z12.48 2.83 12.87 2.86 -3.05 %
tot;i34.82 7.17 36.61 7.72 -4.88 %
µ= mean; σ= standard deviation; η= relative mean error
To better appreciate the power consumption reduction,
we have calculated the average sum of ON-times per orbit
basis, denoted as µ(Ton
orb;i), where
i(tk), i ∈ {x, y, z }
The values of µ(Ton
orb;i)for each axis are compared in Fig-
ure 7 for the two considered B-dot laws. A reduction in
average power consumption can be observed (especially
at the beginning of the detumbling) for the weighted B-
dot controller. It shall be noted that no tuning has been
performed to optimise ϕand εfor power efficiency or de-
tumbling time.
0 5 10 15 20 25 30
Weighted B-dot
B-dot of G.Avanzini & F.Giulietti
0 5 10 15 20 25 30
0 5 10 15 20 25 30
Number of orbits (1 orbit = 1.53 hours; 1 day = 15.73 orbits)
Figure 7: Comparison of the average on-times per orbit
IAC-18-C1.3.11 Page 10 of 11
69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright c
2018 by Dr. Robert Fonod. Published by the IAF, with permission and released to the IAF to publish in all forms.
6. Conclusions
The problem of pure magnetic detumbling of a fast-
tumbling picosatellite was studied. A new weighted B-
dot controller has been proposed. This controller enables
power consumption reduction while not significantly in-
creasing the time it takes to detumble. Monte Carlo sim-
ulation results have revealed that it is possible to save al-
most 5% of power while sacrificing only about 0.5% of
the detumbling performance when compared to the same
control strategy with constant control gain.
The authors would like to thank the members of the
Delfi-PQ team for their helpful discussions on practical
aspects of the detumbling algorithm implementation.
[1] S. Radu, M. S. Uluda ˘
g, S. Speretta, J. Bouwmeester,
E. Gill, and N. Chronas Foteinakis. Delfi-PQ: The
First Pocketqube of Delft University of Technol-
ogy. In International Astronautical Congress, 2018.
[2] R. Wisniewski. Satellite attitude control using only
electromagnetic actuation. PhD thesis, Aalborg
University, December 1996.
[3] J. T. Gravdahl. Magnetic attitude control for satel-
lites. In Conference on Decision and Control, vol-
ume 1, pages 261–266. IEEE, 2004.
[4] C.D. Maclean, D. Pagnozzi, and J. Biggs. Com-
putationally light attitude controls for resource lim-
ited nano-spacecraft. In International Astronautical
Congress, 2011. Paper IAC–11-C1.6.8.
[5] M. Yu. Ovchinnikov, D.S. Roldugin, and V.I.
Penkov. Three-axis active magnetic attitude control
asymptotical study. Acta Astronautica, 110:279–
286, 2015.
[6] D. Baldacchino, D Debattista, R Rotin, D. Cachia,
M. A. Azzopardi, S. G. Fabri, and M. K. Bugeja. Re-
view and Feasibility of Active Attitude Control and
Detumbling for the UoMBSat-1 PocketQube Pico-
Satellite. In Mediterranean Conference on Control
and Automation, pages 1237–1243. IEEE, 2017.
[7] A.C. Stickler and K. T Alfriend. Elementary mag-
netic attitude control system. Journal of Spacecraft
and Rockets, 13(5):282–287, 1976.
[8] G. Avanzini and F. Giulietti. Magnetic detumbling
of a rigid spacecraft. Journal of Guidance, Con-
trol, and Dynamics, 35(4):1326–1334, 2012. doi:
[9] M. J. Sidi. Spacecraft dynamics and control: a prac-
tical engineering approach, volume 7. Cambridge
University Press, 2000.
[10] J. R. Wertz. Spacecraft attitude determination and
control, volume 73. Springer Science & Business
Media, 2012.
[11] L.J. Kamm. Magnetorquer-a satellite orientation de-
vice. ARS Journal, 31(6):813–815, 1961.
[12] J. Li, M. Post, T. Wright, and R. Lee. Design of atti-
tude control systems for cubesat-class nanosatellite.
Journal of Control Science and Engineering, 2013:
4, 2013.
[13] F.L. Markley. Attitude control algorithms for the
solar maximum mission. In Guidance and Control
Conference, pages 59–69. AIAA, August 1978.
[14] J. Reijneveld and D. Choukroun. Attitude control
system of the Delfi-n3Xt satellite. Progress in Flight
Dynamics, GNC, and Avionics, 6:189–208, 2013.
[15] M. Yu Ovchinnikov, D.S. Roldugin, S.S. Tkachev,
and V.I. Penkov. B-dot algorithm steady-state mo-
tion performance. Acta Astronautica, 146:66–72,
[16] H.C. Chang, W.L. Chiang, and Y.Y. Lian. Efficiency
investigation of conventional satellite initial acquisi-
tion control with the consideration of orbital motion.
Procedia Engineering, 67:128–139, 2013.
[17] P. Reijneveld. Design of the Attitude Determination
and Control Algorithms for the Delfi-n3Xt. Mas-
ter’s thesis, Delft University of Technology, January
[18] M. Lovera. Magnetic satellite detumbling: The b-
dot algorithm revisited. In American Control Con-
ference, pages 1867–1872. IEEE, 2015.
[19] P. Tortora, Y. Oshman, and F. Santono. Space-
craft angular rate estimation from magnetometer
data only using an analytic predictor. Journal of
Guidance, Control, and Dynamics, 27(3):365–373,
IAC-18-C1.3.11 Page 11 of 11
... However, there is a degradation in the B-dot controller performance when the angular velocity is low, which is usually the case when the spacecraft is close to being fully detumbled, or if the initial angular velocity 70 happens to be small. Since the development of the first B-dot control, there have been several variants of it, designed for the detumbling maneuver [27][28][29][30][31], and also for attitude acquisition [32,33]. This paper presents a new variant of the B-dot control. ...
... This paper presents a new variant of the B-dot control. This control is demonstrated to achieve faster detumbling with less power consumption com-75 pared to: (1) the simple form of the B-dot control presented in [18], and to (2) a recent variable gain B-dot control [27]. The key innovation of the proposed control is to introduce an equivalent angular velocity for the spacecraft, based on the measurements of the magnetic field, and use it in the control law to guarantee that the computed magnetic dipole moment vector is orthogonal to 80 the magnetic field vector even when the angular velocity is small. ...
... According to the Nyquist criterion, the maximum frequency of changes in the ambient magnetic field that can be sensed without experiencing aliasing is 1 2Ts [5]. Therefore, the maximum rotational velocity that can be measured from two consecutive ambient magnetic field measurements is π Ts [rad/s], which leads to the following sampling time limitation rule: T s ≤ π |ωmax| , where ω max is the 235 maximum expected angular velocity in any of the spacecraft axes [27]. ...
Full-text available
With using the B-dot control law for magnetic detumbling, there will be no need for angular velocity measurements. The B-dot control law is a torque projection-based controller, however, it is not guaranteed to have the dipole moment orthogonal to the magnetic field, especially as the angular velocity gets smaller during detumbling, resulting in a suboptimal torque vector. This paper presents a new variant of the B-dot control. By computing an equivalent angular velocity, based on the magnetic field data, it is possible to develop a control law that guarantees the magnetic dipole moment to remain in the plane orthogonal to the ambient magnetic field. Using Monte Carlo simulations, the proposed B-dot control is compared to two other variants of B-dot control laws: a simple B-dot control and a recent variable gain B-dot control. The results show that the proposed B-dot control outperforms the other B-dot control laws in terms of the speed of detumbling and the power consumption.
... Yet, they are widely used in satellite attitude control as described in many references such as [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. Magnetic control have been particularly the subject of extensive studies as a primary actuator for detumbling maneuvers [23][24][25][26], and also for attitude acquisition maneuvers [27,28]. For spacecraft that are actuated mainly by reaction wheels, magnetic rods can still be used as secondary actuators for momentum management [1,29]. ...
... Sugimura et al took into account the implementation of the magnetic control torque in a intermittent fashion with duty cycle to avoid interference with the magnetometers. Fonod and Gill [23] presented a new variable gain B-dot control law for fast detumbling maneuvers to dissipate the spacecraft rotational kinetic energy. The magnetic rods in [23] are also activated on a duty cycle to avoid noise measurements of the magnetometers. ...
... Fonod and Gill [23] presented a new variable gain B-dot control law for fast detumbling maneuvers to dissipate the spacecraft rotational kinetic energy. The magnetic rods in [23] are also activated on a duty cycle to avoid noise measurements of the magnetometers. In other studies, however, this issue of the noises on magnetometers due to magnetic rods are not addressed, e.g. ...
Full-text available
A magnetometer is essential in spacecraft magnetic attitude control due to the need for magnetic field information to compute the control command. The measurements of the magnetometer, however, are usually affected by other electric currents in the spacecraft, especially those of the magnetic coils when they are turned on during actuation. As a result, magnetic rods and magnetometers are usually turned on at alternate times, resulting in a reduced duty cycle of the magnetic rods, and hence longer maneuver times. This paper presents a magnetic attitude control system with extended duty cycle and low magnetometer measurements frequency. Instead of measurements, a computed magnetic field strength vector is used for updating the control command during the duty cycle. Using the measured spacecraft rotational motion, and knowing the control torque command, it is possible to compute the magnetic field strength vector. These computations are corrected using magnetometer measurements at a lower rate. The Tikhonov regularization approach is implemented to solve the singular magnetic torque system. This algorithm is demonstrated via Monte Carlo numerical simulations to have faster attitude maneuvers and lower power consumption by the magnetic rods. Real data obtained from the CASSIOPE spacecraft are used for validation of the proposed approach.
... The latter control law is called the B-dot law, which was first introduced in, 28 and since then there are several variants developed such as those in. [29][30][31][32][33][34][35][36] These B-dot laws are based on the fact that the magnetic field rate of change is a good indicator of the rate of change in the spacecraft orientation. However, to the best of the authors' knowledge, no existing work in the literature study the effect of the duty cycle operation during the detumbling maneuver. ...
... Also, the average theory will be used to convert the time-variant system into a time-invariant one. In such case, the Lyapunov function derivativeV 2 is negative definite as shown in Eq. (30), and hence the system would be asymptotically stable under this approximation. ...
Conference Paper
Full-text available
In magnetic attitude maneuvers, magnetic rods and magnetometers usually operate alternatively, to avoid the high noise on magnetometers due to magnetic rods currents. As a result, magnetic rods operate on a duty cycle basis. Most studies neglect the impact of the duty cycle in the stability and steady-state error analyses. It is shown in this paper that this duty cycle does impact the stability and the steady-state error in attitude regulation maneuvers. In addition, this duty cycle causes an increase in the maneuver time and the power consumption, compared to that computed by neglecting the duty cycle effect. For detumbling maneuvers, however, it is shown that asymptomatic stability can be proved. Simulation results confirm the obtained analytical results.
... where the left superscript I on B means "with respect to an inertial frame". Since j is small enough compared to ̇ and the derivative of Earth's magnetic field vector is with respect to a body fixed frame [11,12], We now have ...
... →̇≈ − × (11) Taking all Eqs. (2), (6), and (11) into account, we have ̇( ) as described in Eq. (12). ...
Full-text available
This research proposes an innovative approach to detumble satellites’ triple-axis angular velocities with only one single-axis magnetic torquer. Since magnetic torque is generated perpendicularly to magnetorquers, no intended control torque along the magnetorquer can be produced, which makes systems underactuated. Our paper introduces a control method using Model Predictive Control (MPC) and compares it with B-dot control algorithm. By applying these control laws to Kyushu University Light Curve Inversion (Q-Li) Demonstration Satellite in numerical simulations, we describe the applicability of these control laws to underactuated systems.
... In fact, even though higher gains correspond to higher torques, these are typically limited by the threshold value of the magnetic dipole moments, which can be generated by the attitude control system. Some criteria for an optimal selection of these gains, in the presence of different constraints, have been developed [31][32][33]. ...
... In the selection of the sampling frequency, the Nyquist criterion must be considered to avoid aliasing and, therefore, loss of information. In particular, according to Fonod and Gill [33], the sampling frequency can be related to the maximum angular rate expected for the satellite as ≥ , a condition which is verified for TC1 and = 100 Hz. The MSE for TC1 at 100 Hz sampling frequency is represented in Figure 11, finally showing that the impact of process delay is a major cause of the error and that this delay can be reduced to negligible values in real hardware implementation. ...
Full-text available
Attitude determination represents a fundamental task for spacecrafts. Achieving this task on small satellites, and nanosatellites, in particular, is further challenging, because the limited power and computational resources available o-board, together with the low development budget, set strict constraints on the selection of the sensors and the complexity of the algorithms. Attitude determination is obtained only from the measurements of a three-axis magnetometer and a model of the Geomagnetic field, stored on the on-board computer. First, the angular rates are estimated and processed using a second-order low-pass Butterworth filter, then they are used as an input, along with Geomagnetic field data, to estimate the attitude matrix using an unsymmetrical TRIAD. The computational efficiency is enhanced by arranging complex matrix operations into a form of the Faddeev algorithm, which is implemented using systolic array architecture on the FPGA core of a CubeSat on-board computer. The performance and the robustness of the algorithm are evaluated by means of numerical analyses in MATLAB Simulink, showing pointing and angular rate accuracy below 10° and 0.2°/s. The algorithm implemented on FPGA is verified by Hardware-in-the-loop simulation, confirming the results from numerical analyses and efficiency.
... The Controller block implements algorithms for the two principal modes of operation -detumbling (which uses a modification [15] of the well-known Bdot algorithm) and the nominal pointing regime (which uses a modification of the three-axial magnetic attitude control algorithm proposed in [16]). This block also contains the actuators' model, which is presently limited to the use of magnetorquers (air-core, embedded coil or torquerods), because the requirements to the pointing accuracy have not been very stringent in the considered missions. ...
... where k is the scalar coefficient computed with account of the power consumption constraints according to [15]. It should be taken into account that one cannot measure the geomagnetic field, while the magnetorquers are producing the magnetic moment. ...
Conference Paper
An attitude determination and control system (ADCS) model is presented, which is used for CubeSat preliminary design stages in accordance with the European Space Agency (ESA) concurrent design approach. The model is implemented in Matlab/Simulink and integrated into the Concurrent Design Platform. The model’s performance is exemplified by a mission preliminary design conducted in the Concurrent Engineering Lab at Skoltech.
... The current attitude determination and control design for Delfi-PQ is shown in Fig. 7 and comprises three custom made magnetorquers and two commercially available integrated inertial measurement units. The system was designed to be able to stabilize the satellite from the Fig. 7 Delfi-PQ de-tumbling system engineering model maximum rotational speed of 180°/s [13]. The de-tumbling stops when the measured rotational speed is 5°/s or lower. ...
Full-text available
PocketQubes are a form factor of highly miniaturized satellites with a body of one or more cubic units of 5 cm. In this paper, the characteristics of PocketQubes in terms of their constraints and their (potential) utility are treated. To avoid space debris and limit collision risk, the orbits of PocketQubes need to be constraint. An analysis of orbital decay characteristics has been carried out which, considering existing space regulations and a pro-active attitude, PocketQubes should preferably be launched in low Earth orbits below 400 km altitude. Due to technical constraints, such as form factor, power and attitude control, the domain of applications for single PocketQube missions is limited. Still, they can act as low-cost training and technology demonstration platforms. To make PocketQubes an attractive platform for other types of missions, not only the launch cost, but also the development, production and operations cost should be significantly lower than CubeSats. When the PocketQube platform matures and produced in high numbers, networks of PocketQubes can enable new applications. Applications considered feasible are in the field of (but not limited to) continuous surveillance using optical instruments, gravity field monitoring using precise orbit determination, in-situ measurements of the space environment, low data rate or bandwidth communication services and inexpensive probes around other celestial bodies.
Novel technologies and design techniques of the active attitude control systems (ACS) for microspacecraft are briefly covered in this survey. Electric propulsion (EP) thrusters for satellites orbit correction and interplanetary vehicles acceleration have recently emerged as the focal point and front subject in propulsion fields due to their low mass, low cost and high specific impulse. Therefore, micro propulsion thrusters are explored for small satellites in terms of its performance, lifetime and future prospects. The focus of the review is at the center of research published in the last few years and state-of-the-art design concepts in terms of control, reliability, fault analysis and response to failure has been discussed for reaction wheels and magnetorquers. Finally, novel embedded planar magnetorquers which are integrated inside the internal layers of PCB for providing attitude maneuverability to various small satellites are discussed and simulated for critical tasks such as damping the initial angular velocities after vehicle launch. As the proposed magnetorquers are integrated in the form of copper traces, thermal analyses are done to maintain the temperature gradients.
Magnetic control has been used for decades for spacecraft detumbling, i.e., to bring a spacecraft to a final condition with a sufficiently small angular momentum after separation from the launcher. This task is typically achieved by controllers based on the so-called b-dot principle, which stands out thanks to its simplicity, reliability and ease of on-board implementation. In this paper, we first review existing control methods and study their convergence properties with tools borrowed from general averaging theory, which allows addressing in an accurate manner the time-varying nature of magnetic actuation. Then, some effort is devoted to showing the performance limitations of existing controllers for which increasing the gain too much deteriorates the convergence rate. To overcome this issue, a novel projection-based control law with a state-dependent time-varying gain is presented. By means of Lyapunov arguments for non-autonomous systems, we prove that the proposed controller guarantees that the spacecraft angular momentum converges exponentially fast to zero for all initial conditions, robustly with respect to sufficiently small uncertainties in the inertia matrix, and that the closed-loop solutions are globally uniformly ultimately bounded in the presence of exogenous bounded disturbances. Several numerical simulations have been carried out by referring to realistic detumbling scenarios to show the performance improvement with respect to existing controllers.
Full-text available
Satellite attitude motion subject to the well-known B-dot magnetic control is considered. Unlike the majority of studies the present work focuses on the slowly rotating spacecraft. The attitude and the angular velocity acquired after detumbling the satellite is determined. This task is performed using two relatively simple geomagnetic field models. First the satellite is considered moving in the simplified dipole model. Asymptotically stable rotation around the axis of the maximum moment of inertia is found. This axis direction in the inertial space and the rotation rate are found. This result is then refined using the direct dipole geomagnetic field. Simple stable rotation transforms into the periodical motion, the rotation rate is also refined. Numerical analysis with the gravitational torque and the inclined dipole model verifies the analytical results.
Full-text available
Active magnetic attitude control system providing given inertial attitude is considered. Control algorithm is constructed on the basis of a planar motion model. It decreases attitude discrepancy. Alternative approach is based on the PD-controller design. System behavior is analyzed for specific motion cases and sometimes for specific inertia tensor (axisymmetrical satellite) using averaging technique. Overall satellite angular motion is covered. Necessary attitude is found to be accessible for some control parameters. Stability is proven and optimal algorithm parameters are obtained. Floquet-based analysis is performed to verify and broaden analytical results.
Full-text available
In practical, there exist several occasions that satellitesmay temporarily lose their control over body rate: The high tip-off rate after the separation from launch vehicle, the tumbling caused by avionic reset during daily operation and accidental attitude loss during orbit correction. Unfortunately, uncontrolled body rate and attitude usually prevent on-board optical attitude sensors from locking targets, such as Sun, Earth or stars. This report provides a comprehensive analysis of such a process from initial rate reduction to terminal orientation from the perspective of Lyapunov stability.In addition, with proper coordinate transformation, the B-dot control efficiency is found to be orbit inclination angle dependent. Numerical analysis is performed where the conclusion suggests that the average B-dot control efficiency improves as the inclination angle increases.
Full-text available
We present a satellite attitude control system design using low-cost hardware and software for a 1U CubeSat. The attitude control system architecture is a crucial subsystem for any satellite mission since precise pointing is often required to meet mission objectives. The accuracy and precision requirements are even more challenging for small satellites where limited volume, mass, and power are available for the attitude control system hardware. In this proposed embedded attitude control system design for a 1U CubeSat, pointing is obtained through a two-stage approach involving coarse and fine control modes. Fine control is achieved through the use of three reaction wheels or three magnetorquers and one reaction wheel along the pitch axis. Significant design work has been conducted to realize the proposed architecture. In this paper, we present an overview of the embedded attitude control system design; the verification results from numerical simulation studies to demonstrate the performance of a CubeSat-class nanosatellite; and a series of air-bearing verification tests on nanosatellite attitude control system hardware that compares the performance of the proposed nonlinear controller with a proportional-integral-derivative controller.
In this paper, a control law that detumbles a spacecraft with magnetic actuators only is developed. A rigorous mathematical proof of global asymptotic convergence from arbitrary initial tumbling conditions to zero angular velocity in a time-varying magnetic field is derived. Furthermore, a simple criterion for determining a reasonable value of the control gain is developed. The selected gain results in a quasi-minimum time detumbling for different initial conditions, in the presence of magnetic coil saturation. Performance of the proposed detumbling control law is demonstrated by numerical simulations on a large set of test cases using a Monte Carlo approach.
Conference Paper
Recently, small satellite missions have gained considerable interest due to low cost launch opportunities and technological improvement of micmelectronics. Required pointing accuracy of small, inexpensive satellites is often relatively loose, within a couple of degrees. Application of cheap, lightweight, and power efficient actuators is therefore crucial and viable. Linear attitude control strategies for a low-Earth orbit satellite actuated by a set of mutually perpendicular electromagnetic coils are discussed. The principle is to use the interaction between the Earth's magnetic field and the magnetic field generated by the coils. A key challenge is that the mechanical torque can only be produced in a plane perpendicular to the local geomagnetic field vector; therefore, the satellite is not controllable when considered at fixed time. Availability. of design methods for time-varying systems is limited; nevertheless, a solution of the periodic Riccati equation gives an excellent framework for development and analysis of magnetic attitude control algorithms. An observation that the geomagnetic field changes approximately periodically when a satellite is on a near polar orbit is used. Three types of attitude controllers are proposed: an infinite horizon, a finite horizon, and a constant gain controller Their performance is evaluated and compared in the simulation study of the realistic environment.
This work is concerned with the development of the attitude control algorithms that will be implemented on board of the Delfi-n3xt nanosatellite, which is to be launched in 2013. One of the mission objectives is to demonstrate Sun pointing and three axis stabilization. The attitude control modes and the associated algorithms are described. The control authority is shared between three body-mounted magnetorquers (MTQ) and three orthogonal reaction wheels. The attitude information is retrieved from Sun vector measurements, Earth magnetic field measurements, and gyro measurements. The design of the control is achieved as a trade between simplicity and performance. Stabilization and Sun pointing are achieved via the successive application of the classical Bdot control law and a quaternion feedback control. For the purpose of Sun pointing, a simple quaternion estimation scheme is implemented based on geometric arguments, where the need for a costly optimal filtering algorithm is alleviated, and a single line of sight (LoS) measurement is required - here the Sun vector. Beyond the three-axis Sun pointing mode, spinning Sun pointing modes are also described and used as demonstration modes. The three-axis Sun pointing mode requires reaction wheels and magnetic control while the spinning control modes are implemented with magnetic control only. In addition, a simple scheme for angular rates estimation using Sun vector and Earth magnetic measurements is tested in the case of gyro failures. The various control modes performances are illustrated via extensive simulations over several orbits time spans. The simulated models of the dynamical space environment, of the attitude hardware, and the onboard controller logic are using realistic assumptions. All control modes satisfy the minimal Sun pointing requirements allowed for power generation.
Used increasingly in telecommunications, scientific research, surveillance, and meteorology, satellites rely heavily on complex onboard control systems. This book explains the basic theory of spacecraft dynamics and control and the practical aspects of controlling a satellite. The emphasis is on analyzing and solving real-world engineering problems. Among the topics covered are orbital dynamics, attitude dynamics, gravity gradient stabilization, single and dual spin stabilization, attitude maneuvers, attitude stabilization, and structural dynamics and liquid sloshing.