Politecnico di Milano

Spacecraft Attitude Dynamics

and Control
Academic Year 2020/2021

Final Project Report

Attitude Control of a CUBESAT

Supervisor:
Author:
Prof. Franco Bernelli
Enrico Bassissi 945090
Zazzera
Contents

1 Introduction 1
1.1 Structure Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Orbit Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Mission Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Model Description 2
2.1 Dynamics and Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.1 Orbital Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.2 Attitude Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.3 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Disturbance Torques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 Aerodynamic Drag Torque . . . . . . . . . . . . . . . . . . . . . . 4
2.2.2 Electromagnetic Radiation Torque . . . . . . . . . . . . . . . . . 4
2.2.3 Gravity Gradient Torque . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.4 Magnetic Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 ADCS Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.1 Earth Horizon Sensor . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.2 Sun Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.3 Gyroscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.4 Magnetometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.5 Magnetorquer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.6 Reaction Wheels . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Hardware Architecture Datasheets 9

3.1 Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1 Gyroscope Datasheet . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.2 Sun Sensor Datasheet . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.3 Earth Horizon Sensor Datasheet . . . . . . . . . . . . . . . . . . . 10
3.1.4 Magnetometer Datasheet . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Actuators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.1 Magnetorquers Datasheet . . . . . . . . . . . . . . . . . . . . . . 10
3.2.2 Reaction Wheels Datasheet . . . . . . . . . . . . . . . . . . . . . 10

4 Attitude Determination and Control 11

4.1 Attitude Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Detumbling Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.2.1 Magnetorquers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 Sun Tracking Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3.1 Reaction Wheels . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 Results 14
5.1 Detumbling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.1.1 Dynamic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.1.2 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.1.3 Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2 Sun Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
 5.2.1 Dynamic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.2.2 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.2.3 Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Robustness Analysis 19

7 Conclusions and Further Developments 20

1 Introduction
1.1 Structure Properties
The spacecraft under study is a 6U Cubesat, structured as shown in Fig. 1.

Figure 1: Model of the 6U Cubesat with solar panel on the back,

center of mass and principal reference frame

Each unit is a cube with side of dimension 0.1 m resulting in a volume occupied of 1
litre each. This leads to a total dimensions of 0.1 x 0.2 m x 0.3 m, along the x, y and
z axes respectively [1]. Within the total surface of the satellite normal to the positive x
direction is included a solar panel. The distribution of mass in the volume is assumed
to be non-uniform, which results in a displacement of the centre of mass with respect
to the geometric centre of 0.018 m, along all the three axes, in compliance with the
misalignment limit of the Cubesat guidelines [1].
From literature and previous mission, it is assumed the maximum weight for a cubesat
unit is 1.3 kg [1], the mass of the solar panel adopted is 0.3 kg [2], and considering all the
sensors and actuators adopted by the ADCS exposed afterwards in the report, the total
weight is 9.004 kg. The inertia matrix is then computed, obtaining the results shown in
Eq. (1). » fi
0.0975 0 0
I“ – 0 0.0750 0 fl kg m2 (1)
0 0 0.0375

1.2 Orbit Description

The initial orbital parameters are reported in Table 1. It is assumed that the spacecraft is
at the pericentre of its orbit at time t “ 0 s. The orbital period is 8095 s ” 134.92 min ”
2.249 h. The initial angular velocities with respect to the rx, y, zs axes are ω 0 “ r5, 10, ´5s
rad/s.

1
 Table 1: Orbital Parameters

a [km] e [-] i [deg] Ω [deg] ω [deg] θ [deg]

8713.75 0.2 45 90 0 0

1.3 Mission Overview

The overall mission requires to detumble the spacecraft and then stabilise the 6U Cubesat
to Sun pointing. The task of the spacecraft’s ADCS is to estimate the current attitude
of the spacecraft by combining the information gathered from its sensors and control the
attitude of the spacecraft by giving commands to its actuators in a closed-loop control
system, as schematically depicted in Fig. 2.

Environment
Disturbances
𝑇𝑑 Model of Celestial Bodies
v

Attitude
Dynamics Sensors Reference Attitude
Kinematics Determination
+ • Euler Equations • Information to • Tracking direction
𝑇 ω • Quaternions α s • Estimated
• Orbital Mechanics the spacecraft and velocity
orientation

𝛼𝑚 -
Control + 𝛼𝑑𝑒𝑠
Actuators
Algorithm +
• Reaction Wheels
𝑇𝑒𝑓𝑓 𝑇𝑐 • Detumbling 𝛼𝑒 𝛼𝑇
• Magnetorquers
• Tracking

Figure 2: ADCS Flowchart

2 Model Description
2.1 Dynamics and Kinematics
2.1.1 Orbital Mechanics
In the analysis of a spacecraft’s dynamics it is usual to decouple orbital mechanics and
attitude dynamics, due to their very different timescales. A simplified orbital mechanic
model related to the propagation only of the true anomaly has been implemented, justi-
fied by the fact that it’s been simulated actually a small time of the mission, in the order
of a single orbit, so the longer term effect on the other keplerian parameter is considered
negligible. The orbit propagation is obtained by integrating the true anomaly over time
in the Earth Centred Inertial (ECI) reference frame, as shown in Eq. (2).
2
c
np1 ` e cos θq µC
θ9 “ ? , n“ (2)
1´e 2 a3

Then, these values are converted to the Perifocal reference frame, as highlighted in

2
Eq. (3). $ , $ ,
&cos θ. c ´ sin θ .
µC &
rP F “ r ¨ sin θ , vP F “ ¨ e ` cos θ (3)
p %
0 0
% - -

After that, rP F and v P F are converted again in the ECI reference frame, according
to Eq. (4).

rECI “ R3 rωs R1 ris R3 rΩs rP F , v ECI “ R3 rωs R1 ris R3 rΩs v P F (4)

The disturbances are not taken into account while computing the orbit dynamics.

2.1.2 Attitude Dynamics

As a first step, it is important to clarify the reference systems acting in this report’s
framework and the associated notation: the body frame, indicated with the subscript B,
and the fixed inertial frame, centred in the Earth, indicated with the subscript N, with
the relative angular velocity between the two frames ω B{N . Since the inertial frame is
fixed ω B{N ” ω B and it will be used as simpler notation ω “ rωx , ωy , ωz sT .
The Euler equation in Eq. (5) is the analytical form describing the attitude motion
of a rigid body, stating the evolution of the angular velocity dynamics subjected also to
disturbance torques and control actions.
dω
J “ Jω ˆ ω ` d ` u (5)
dt
In this case, from the choice of the body frame, dynamics refers to the principal
inertia axes and therefore Eq. (5) can be simplified to Eq. (6), where ri, j, ks represent
recursively the principal axes rx, y, zs.

Ii ω9 i “ pIj ´ Ik qωj ωk ` ui ` di (6)

2.1.3 Kinematics
In addition to the study of the dynamics, the attitude kinematics of the spacecraft has
been studied based on a quaternion representation. They’re not a minimum represen-
tation set for the kinematic but they are easy to manipulate and have no singularities.
The kinematic is recovered by integrating Eq. (7):
» fi
0 ωz ´ωy ωx
dq 1 —´ωz 0 ωx ωy ffi
“ — ffi q (7)
dt 2 – ωy ´ωx 0 ωz fl
´ωx ´ωy ´ωz 0

From that, q “ rq1 , q2 , q3 , q4 sT is the vector representing the attitude. The first three
components of q are known as the vector part of the quaternions, while the last element is
the scalar part used for normalisation. From quaternions it is easy to shift back to direct
cosine matrix representation, which is more intuitive and physically representative.

2.2 Disturbance Torques

During the mission, a satellite orbiting the Earth is subject to several environmental
disturbances, the four most important of which will be covered in this report.

3
2.2.1 Aerodynamic Drag Torque
The interaction between satellite and upper atmosphere generates aerodynamic forces,
that in turn can generates a torque with respect to the centre of mass. The mathematical
model of these forces is that air particles hit the external surface of the satellite and their
kinetic energy is totally transferred to the satellite. The aerodynamic force acting on the
elementary area dA, defined by its perpendicular direction N̂ , is function of the density
ρ and the relative velocity v. The aerodynamic torque is then obtained by integrating
through all surfaces for which N̂ ¨ v̂ ą 0, meaning that they are exposed to aerodynamic
flow, related to the distance between the center of mass and the point of application of
the aerodynamic force itself.
1
dF “ CD ρ|v|2 pN̂ ¨ v̂qv̂ dA , T drag “ ∫ ri ^ dF i (8)
2
The drag coefficient is taken as constant CD “ 2.2 [10]. To make an expression for
the air density, compatible with the experimental data, it can be expressed depending
upon the altitude and some coefficients taking into account other parameters such as
temperature, solar activity, magnetic effects. For the purpose of this report, the CIRA-
72 Atmospheric model was implemented, a piecewise-continuos exponential formulation
of the density [11], [12] as in Eq. (9) where H is the scale height and ρ0 is the reference
value of the air density at the considered range of altitude.
ˆ ˙
h ´ h0
ρ “ ρ0 exp ´ (9)
H

2.2.2 Electromagnetic Radiation Torque

Like aerodynamic drag, the electromagnetic pressure is a non conservative perturbation,
but has opposite behaviour becoming more important at higher altitudes. The electro-
magnetic radiation is modeled as a sum of three contributions: Solar Radiation Pressure,
the direct radiation from the Sun; albedo, the reflected solar radiation by the Earth; the
Earth own infrared radiation.
One of the most difficult aspects of analysing solar radiation is accurately modeling
and predicting the solar cycles variations and to determine the shadowing effect of the
Earth on the spacecraft itself. In this report, the simulation time is small enough to
consider the solar radiation intensity as constant during all the mission and also to
neglect the shadowing effect.
The values of these contributions for this mission are reported in Table 2. Another
applied approximation is that the altitude of 600 km is assumed constant along the whole
mission in order to easily evaluate the intensity coefficients for albedo and IR emission,
even if the orbit of the satellite is not strictly circular.

Table 2: Radiation Power Values at 600 Km

SRP [W ] Earth Albedo [W ] Earth IR Emission [W ]

1358 580 143.4

The radiation power, related to the velocity of the incoming photon (speed of light)
determines the force acting on the spacecraft per each unit surface, a pressure. Inserting

4
in the equation the scattering coefficient ρs , the diffraction coefficient ρd (in Table 3) and
the exposed area to the sun, it’s possible to obtain the force acting on the ith face of the
satellite, which in turn generates a torque around the center of gravity of the body.

P@ “ Fe,@ {c , PC “ Fe,C {c (10)

F @,i “ ´P@ Ai pSˆB ¨ N̂i qpp1 ´ ρs,i qSˆB ` p2ρs,i pSˆB ¨ N̂i q ` 2{3ρd,i qN̂i q , SˆB ¨ N̂i ą 0 (11)
F C,i “ ´PC Ai pEˆB ¨ N̂i qpp1 ´ ρs,i qEˆB ` p2ρs,i pEˆB ¨ N̂i q ` 2{3ρd,i qN̂i q , EˆB ¨ N̂i ą 0 (12)
T @,i “ ri ^ F @,i , T C,i “ ri ^ F C,i (13)

Table 3: Scattering, Diffraction, Absorption Coefficients for

Spacecraft’s Faces and for Solar Panel’s Surfaces

ρs SC [-] ρd SC [-] ρa SC [-] ρs SP [-] ρd SP [-] ρa SP [-]

0.5 0.1 0.4 0.8 0.1 0.1

The Earth’s orbit around the Sun is assumed to be circular and the position vector
is reported in Eq. (14), where n@ is the angular velocity of earth’s orbit around the Sun
and ε@ is the angle between equatorial and ecliptic planes.

S B “ ABN S N (14)
“ ‰T
S N “ AU cospn@ tq, sinpn@ tq cospε@ q, sinpn@ tq sinpε@ q (15)
AU “ 1.496 ¨ 108 km , n@ “ 1.9924 ¨ 10´7 rad{s , ε@ “ 23.45˝ (16)

2.2.3 Gravity Gradient Torque

The gravity field is not uniform over the volume of the spacecraft, therefore there could be
a torque acting on the satellite. Its effect along the three principal axes can be obtained
by integrating the gravitational force over the body, obtaining the formulas shown in
Eq. (17), where ri, j, ks represent recursively the principal axes rx, y, zs and AB,LV LH is
the attitude matrix between the body fixed frame and the Local Vertical Local Horizon
(LVLH) frame.
3µC
Tgg,i “ rpIk ´ Ij qck cj (17)
R3
c “ rcx , cy , cz sT “ AB,LV LH r1, 0, 0sT (18)

2.2.4 Magnetic Torque

The Earth’s magnetic field is conceptually similar to the one generated by a dipole
inclined with respect to the Earth’s axis by 11.5˝ . It is not a constant uniform field,
it slowly rotates, it decreases its amplitude in time, it’s more intense at the poles and
decreases with increasing distance from the ground.
The magnetic field can be considered similar to a dipole field only from 7000 km from
ground to « 10RC . Beyond that limit the magnetic field is deformed by the effect of
the solar radiation, instead at low altitude an higher order expansion is most probably
required to correctly represent it.

5
 Since the orbit reported here is in LEO, the complete International Geomagnetic
Reference Field (IGRF) 13th order model have been implemented, whose result is de-
picted in Fig. 3 for a layer at 600 km altitude. Firstly, the magnetic field is measured in
the Earth Rotating Frame, a frame fixed to the Earth and that rotates with the Earth
bER “ rbr , bθ , bϕ sT and it is the gradient of a scalar potential V, bER “ ´∇V , normally
modeled as a series expansion of spherical harmonics:
n
k ˆ ˙n`1 ÿ
ÿ R
V pr, θ, ϕq “ R pgnm cospmϕq ` hm m
n sinpmϕqqPn pθq (19)
n“1
r m“0

where r, θ and ϕ are the spherical coordinates of the position of the satellite, respectively
the distance from the center of the Earth, colatitude and longitude. The coefficients gnm
and hm n are Gaussian coefficients put forth by the IAGA [3] for the IGRF, and Pn pθq
m

represents the Schmidt quasi-normalised associated Legendre functions of order n and

degree m.
The complication in constructing a more accurate model of the Earth’s magnetic field
is in the evaluation of the Legendre polynomials which is undertaken using an iterative
approach. The relationship between the Gaussian normalised associated Legendre poly-
nomials and the Schmidt quasi-normalised ones is shown in Eq. (20) and the recursive
formulas [13] are applied to the Gaussian normalised associated Legendre polynomials
P n,m and to the Schmidt quasi-normalisation factors Sn,m .

Pnm “ Sn,m P n,m (20)

In a geocentric inertial reference frame, the components of the magnetic field are then
bN “ rbx , by , bz sT made through Eq. (21):

bx “ pbr cospδq ` bθ sinpδqq cospαq ´ bϕ sinpαq (21)

by “ pbr cospδq ` bθ sinpδqq sinpαq ` bϕ cospαq (22)
bz “ pbr sinpδq ` bθ cospδqq (23)

where δ “ π{2 ´ θ is the satellite declination and α is the right ascension.

To obtain the magnetic field as measured in the body frame then we use bB “
ABN bN . The disturbance implication of the magnetic field for attitude determination
and control is the interaction between local magnetic field and the residual parasitic
magnetic moment of the spacecraft induced by the flow of electrical currents. This
induction is a difficult quantity to model in simulation and here it is taken as an average
constant value which may represents a worst case scenario, that produces a torque given
by Eq. (24).
T magn “ m ^ bB , m “ r0.1, 0.1, 0.1sT A{m2 (24)

2.3 ADCS Components

The spacecraft ADCS is provided with a Sun sensor, an Earth Horizon Sensor, a Magne-
tometer and a Gyroscope, and a set of actuators made by a 3-axis Magnetorquer and a
set of 3 Reaction Wheels. As for the sensors, the proper complete modeling of those com-
ponents will not be discussed and implemented in this projects, but their characteristics
and performances will be taken into account to produce the measure required.

6
 49000 49000
49000 49000
49000 49000

52500

52500
49000
49000 49000
49000 45500
49000 45500 49000 45500
45500 42000
45500 42000
42000 45500 38500 42000
38500
35000 38500 42000
38500 42000 31500 35000 38500 35000
35000 38500 28000 31500 35000 31500
31500 35000 24500 28000 31500 28000
28000 31500 28000
0
24500 28000 2450 21000
2450 24500 0
0 50 28000
21000 24 0
28000 00 24500 2800 31500
2450 210 3150
0 35000
31500 0 28000 0 38500 38500
35000 3500 42000
2800 2450
0
31500 42000
0 0 45500 45500
4200 3150 2800 00
0 0
35000 385 00
49000
455 385
3500
0 31500 490
00 00

5250
49 35000 0

0
00 00

5250
0 420 0 42
00 3850

0
45500
38500

49000
49000
00
455 45500
45500
00 42000
420
42000
42000 42000

Figure 3: Magnetic Field Intensity, IGRF order 13th at 600 km

2.3.1 Earth Horizon Sensor

The static Earth Horizon device contains a number of sensors capable of measuring
Earth’s surface IR radiation, through the heating of a material with a temperature-
dependent electrical resistance, with a field of view slightly larger than the Earth, so
that when the image of the Earth is not centred, the detectors have different outputs
and taking appropriate differences it’s possible to have direct information of the mis-
alignment attitude angles. Given that reading, having determined the orbit, it’s possible
to reconstruct the direction vector of the Earth with respect to the spacecraft measured
in the inertial frame ÊN and therefore the ÊB “ ABN ÊN .

2.3.2 Sun Sensor

Sun sensors give as output the relative position vector between the spacecraft and the
Sun measured in the body fixed frame. The Sun radiation hits the photocell material
surface, it changes the angle of travel caused by changing the medium because of the
Snell’s Law. At this point the sensitive material below reads the position of the incident
tilted radiation beam and, knowing the refraction index of the layer, it’s possible to
compute the original angle of arrival of the beam itself, direction in body frame.

2.3.3 Gyroscope
Gyroscopes are used to measure angular velocities. One scheme of angular velocity sensor
is based on the use of piezoelectric materials. Two forks made of piezoelectric materials
are used. If subject to an electric field, the fork is deformed, while if deformed it produces
an electric signal. If the sensor rotates with velocity ω, and the two prongs are electrically
driven in phase opposition, the Coriolis force, perpendicular to ω and the tip velocity,
generates a torque, as in Eq. (25). The torque is transmitted to the base prongs that,
due to their deformation, generate a second electric signal that can be measured.

F c “ ´2ω ^ v r (25)

Also this sensor measures the angular velocity along a fixed direction, and has no
moving parts, so it’s considered a more suitable solution with respect to a classical

7
mechanical gyroscope. However it is extremely sensitive to temperature oscillations,
which cause a thermal deformation that has an effect on the scale factor of the sensor.
These components can be miniaturised, but the precision is not so high and small
errors will add up over time. Gyros are electro-mechanical devices, their intrinsic noise
can be model as:

ωM
i “ ωi ` n ` b , n “ σn ζ n , b9 “ σb ζ b (26)

where σn ζ n is white Gaussian noise with zero mean and standard deviation σn , known as
angular random walk (ARW) attributed to thermo-mechanical noise of the system; σb ζ b
is also a white Gaussian noise with zero mean and standard deviation σb , known as rate
random walk (RRW) attributed to electronic noise. These are the main contributions of
noise in a gyro and their values are usually provided by manufacturers.

2.3.4 Magnetometer
Using three pairs of ferromagnetic material nuclei it’s possible to measure the components
of the magnetic field along three orthogonal directions bB , taken in a body axes frame.
The fluxgate magnetometer works by assessing the difference that the external mag-
netic field apply on the baseline magnetic field produced by the sensor itself. As long as
no external field is present, the same and opposite magnetic field is applied to the two
nuclei; as soon as external field is applied, the rate of change is read by exploiting the
saturation property of the sensor by producing voltage pulses spaced in proportion to
the external magnetic field measurement.
They are the least accurate attitude reference sensors as there is a lot of uncertainty
in the modelling and measurement of the magnetic field, they are also subject to noisy
measurements and interference from on-board electronics.

2.3.5 Magnetorquer
Magnetic actuators generate a torque by inducing a magnetic dipole in a coil that is
surrounded by the Earth’s magnetic field. The magnetic dipole generated can be modeled
as in Eq. (27), where µ is the magnetic permeability, n the number of windings, S the
area of the coil and I the current intensity; the torque is then given by the vector product
of this dipole moment and the external magnetic field.

D “ µnSI , T M T “ D ^ bB (27)

The effectiveness of these actuators varies with the orbit height, due to the change in the
external magnetic field, providing a suitable control action for the Cubesat under study
being in LEO at an altitude of 600 km.
Once the torque is known, it’s possible to invert the problem to find the magnetic
dipole D to generate in order to control the spacecraft. The saturation of this component
has been taken into account, by imposing a limit on the maximum/minimum magnetic
moment that it can generate.

2.3.6 Reaction Wheels

These momentum exchange actuators are based on acceleration and deceleration of spin-
ning rotors, with nominal condition as zero angular velocity. They can only exchange an-
gular momentum with the satellite but not produce a net torque. If the torque required is

8
 the combination of a periodic component and a secular component, it will be inevitable
to reach the saturation speed, condition for which the RW cannot exchange anymore
momentum and need de-saturation mechanisms to be applied, like using thrusters or
magnetorquers. A further development to improve the model could be to implement one
of these techniques in order o guarantee the effective actuation for longer time.
RW can be modeled by writing the appropriate set of Euler equations, including the
effects of the actuators in the expression of the angular momentum Eq. (28), where A
is the distribution matrix whose columns represent the direction of the axis of rotation
of the wheels. Then it’s required to derive it by the time as in Eq. (29) and add the
equations of the relative dynamic of the rotors, the torques applied by the electrical
motor to the rotors to activate them:

h “ Iω ` Ahr (28)
#
I ω9 ` ω ^ Iω ` Ah9 r ` ω ^ Ahr “ Text
(29)
Ah9 r “ M r

In the previous derivation the terms related to I9 and A9 have been neglected because
in RW there is not the possibility to change the axis of rotation, the configuration of the
RW themselves.
Once known the torque required by the control it’s possible to assess the amount of
angular momentum rate of change to apply to the reaction wheel in order to supply the
requested action.
As requirement for this project, 3 RWs have been implemented, aligned with the
three principal inertia axes. In general 3 rotors are enough to solve exactly the control
problem, in practice a fourth rotor is usually added for redundancy with a geometry that
allows a full control capacity even with one rotor fails. This means that the 4 rotors have
distinct rotation axes, and any combination of 3 rotors would allow the invertibility of
the rotors A matrix.

3 Hardware Architecture Datasheets

3.1 Sensors
3.1.1 Gyroscope Datasheet
Gyroscope - STIM300 - Sensonor [4]

Table 4: Gyroscope Datasheet

?
ARW [deg{ h] RRW [deg{h] Sampl. rate [Hz] Power [W ] Mass [g] Dimensions [mm]
0.15 0.3 2000 1.5 ˜ 2 55 35.9 x 38.6 x 21.5

3.1.2 Sun Sensor Datasheet

Sun Sensor - NFSS-411 - New Space Systems [5]

9
 Table 5: Sun Sensor Datasheet

Accuracy [deg] FoV [deg] Sampl. rate [Hz] Power [mW ] Mass [g] Dimensions [mm]
0.1˝ ˘120˝ 5 15 ˜ 150 35 34 x 40 x 20

3.1.3 Earth Horizon Sensor Datasheet

Earth Horizon Sensor - MAI-SES - Adcole Maryland Aerospace [6]

Table 6: Earth Horizon Sensor Datasheet

Accuracy [deg] FoV [deg] Power [mW ] Mass [g] Dimensions [mm]
0.25˝ 7˝ 132 33 43.3 x 31.8 x 20.7

3.1.4 Magnetometer Datasheet

Magnetometer Sensor - NNSS-AMR - NewSpace Systems [7]

Table 7: Magnetometer Sensor Datasheet

Noise [nT rms{Hz] Update Rate [Hz] Power [mW ] Mass [g] Dimensions [cm]
16 @ 1 Hz 18 750 85 96 x 43 x 17

3.2 Actuators
Two sets of actuators are used in this Cubesat, related to the different specific task to
which they are assigned. The first manoeuvre that a spacecraft usually has to perform as
soon as released from the fairing is the detumbling, where the residual angular velocity
of the satellite may be still high, limiting the possibility of using most of the sensors
therefore most of the actuators. So 3 axis magnetorquers from GOM Space were used to
control this phase of the mission, composed by two torque rods and one air core torquer,
aligned with the three principal axes in the body frame.
As for the tracking part of the mission, once the angular velocity of the spacecraft
is considerably reduced, the attitude determination process can take place and more
sophisticated actuators can be applied. For this reason, 3 reaction wheels NanoTorque
GSW-600 from GOM Space have been selected.

3.2.1 Magnetorquers Datasheet

Magnetorquer - NanoTorque GST-600 - GOM Space [8]

3.2.2 Reaction Wheels Datasheet

Reaction Wheels - NanoTorque GSW-600 - GOM Space [9]

10
 Table 8: Magnetorquers Datasheet

Dipole Moment (x,y) [Am2 ] Dipole Moment (z) [Am2 ] Mass [g] Dimensions [mm]
0.31 0.34 156 90.5 x 96.9 x 17.2

Table 9: Reaction Wheels Datasheet

T Max [mN m] h Max [mN ms] Power [W ] Mass [g] Dim. (One Wheel) [mm]
4.5 57 0.3 ˜ 2 540 44 x 44 x 27

4 Attitude Determination and Control

4.1 Attitude Determination
The attitude determination problem consists in finding the matrix that lets overlap the
measure vectors and the inertia reference frame. To obtain the attitude matrix three
methods have been tested, in order to find the one that fits better with the specific case
mission.
The first one is the TRIAD method: given the knowledge of two non parallel vectors
in the reference and body coordinates of a satellite, the TRIAD algorithm obtains the
direction cosine matrix relating both frames. This derivation is algebraic, the ABN
matrix found is considered exact, even thought it is affected by measurement errors. The
derivation is strongly dependent on the measurements: the starting measure p should be
associated with the highest precision sensor.
The other two algorithms, the q-method and the SVD method, fall under the sta-
tistical approaches, for which the errors in the measurement are taken into account
intrinsically, for which the objective is not to find the exact rotation matrix, but the one
that minimise the residual error. This formulation is called the Wahba’s Problem, that
proceed by the minimisation of a cost function, where the measurements are relatively
weighted by the precision of the instrument αi .
N N
1ÿ ÿ
T
argminpJpAqq “ αi |SB,i ´ ABN vN,i |2 “ 1 ´ αi pSB,i ABN vN,i q (30)
2 i“1 i“1

The SVD method finds a solution to the problem following the single value decomposi-
tion, solving an eigenproblem in order to compute the rotation matrix. The q-method
proceeds using quaternions notation and applies the eigenvalues analysis on a tailor-built
matrix to find the optimal quaternion attitude from a statistical point of view.
All the analysed algorithms performed well, with attitude determination errors in the
order of 10´5 rad, and the most accurate resulted to be the TRIAD method, with no
singularity points detected. Starting from the sensor readings in the body frame B of Sun
and Earth and comparing with the model vectors in the inertial frame N , the attitude
matrix ABN,sens is constructed, as reported in algorithm 1. The most accurate sensor in

11
this mission is the Sun sensor, hence the first vectors of the algorithm are related to it.
Algorithm 1: TRIAD Attitude Determination Procedure
Measure vector in body frame
S B,sens S B,sens ^ rB,sens S B,sens ^ b2
b1 “ ; b2 “ ; b3 “
|S B,sens | |S B,sens ^ rB,sens | |S B,sens ^ b2 |
B “ rb1 , b2 , b3 s
Measure vector in inertial frame
S S ^ rN S ^ n2
n1 “ N ; n2 “ N ; n3 “ N
|S N | |S N ^ rN | |S N ^ n2 |
N “ rn1 , n2 , n3 s
Attitude Matrix measured
ABN,sens “ BN T

4.2 Detumbling Control

The main role of the ADCS is to control the attitude of the spacecraft. The first critical
operation that a spacecraft must face, as soon as it is released from the launcher’s fairing,
is to detumble, to stabilise the satellite so that its velocity in the inertial frame is zero.
The initial condition of this mission phase in angular velocity and attitude has been
chosen as:
ω 0 “ r5, 10, ´5s deg/s , q 0 “ r0, 0, 0, 1sT
This is done with a proportional control law written and implemented as in Eq. (31),
with the gain chosen in order to provide convergence in a suitable time and to not exceed
considerably the saturation of magnetometers for a long period of time, proportional to
the magnitude of inertia moments (highest kdet for highest inertia). The perturbations
d have been subtracted for disturbances attenuation.

uid “ ´k det ω ´ d , k det “ r3.48, 2.42, 1.48sT ¨ 106 (31)

4.2.1 Magnetorquers
To actuate this control, considering the limitations and positive factors explained also
in section 3.2 that magnetorquers do not require any precise instrument measure nor
attitude determination, the uid will be mapped into the magnetorquers system in order
to achieve the objective.
After waiting 30 s, to let the gyros have a stable reading of the angular velocity in
order to perform the control on reliable variables, the magnetometers are activated to
generate the ideal required dipole moment. If this mid is higher than the admissible
limit, the saturation block in Eq. (32) takes action and bounds the maximum/minimum
applied magnetic moment of Table 8.
Now the generated magnetic dipole actually interacts with the real magnetic field bB
producing an effective torque on the spacecraft T M T .
mmax
mid “ bB ^ uid ÝÑ m“ ´mmax mid (32)

T M T “ m ^ bB (33)

12
4.3 Sun Tracking Control
Once the satellite is stabilised, meaning that the angular velocity in the body frame is
varying in a small region around zero, tracking manoeuvre can be performed to point and
follow the Sun. This phase of the mission is thought to be subsequent to the detumbling
one, hence the initial attitude, and thus the initial quaternion vector, is the same as the
one obtained at the end of the previous simulation:

q BN, 0, tracking
“ q BN, end, detumbling

The tracking control requires a target direction and a subsequent angular velocity to
maintain. The desired attitude matrix Ad is constructed as expressed in algorithm 2,
where n@ and ε@ are the same Sun quantity as in Eq. (14) of section 2.2.2:
Algorithm 2: Target Attitude and Velocity Assembly Procedure
Velocity of the Sun with respect to the Earth
v S,N “ AU n@ r´ sinpn@ tq, cospn@ tq cospε@ q, cospn@ tq sinpε@ qsT
Relative position Spacecraft/Sun
rS “ S N ´ rN
Build the target attitude matrix:
r rS ^ v S,N z^x
x“ S , z“ , y“
|rS | |rS ^ v S,N | |z ^ x|
T
Ad “ rx, y, zs
Build the target angular velocity
W “ ´A9 d ATd
ωd,x “ ´W p2, 3q; ωd,y “ W p1, 3q; ωd,z “ ´W p1, 2q
ω d “ rωd,x , ωd,y , ωd,z sT
The attitude error matrix Aerr and the angular velocity error ω err can be computed
through the desired attitude matrix, the one of the spacecraft and its angular velocity,
all measured by the system’s sensors, the one obtained after the attitude determination
process, as in Eq. (34):

Aerr “ ABN ATd , ω err “ ω ´ Aerr ω d (34)

The ideal control law used for this part of the mission is expressed in Eq. (35), where
the operator p ¨ qV maps a skew-symmetric matrix into a vector:
d
uid “ ´kt1 Iω e ´ kt2 pATerr ´ Aerr qV ` ω ^ Iω ` I pAerr ω d q ´ d (35)
dt
kt1 “ 0.008; kt2 “ 0.01;
Gains are once again chosen in order to provide convergence and pointing in a suitable
time and to not exceed the limit imposed by the maximum available control torque (so
that they are not brought to saturation).

4.3.1 Reaction Wheels

Tracking manoeuvre cannot be performed with the only use of the three magnetic ac-
tuators, because convergence can be guaranteed only in two axes as the third may be
aligned with the magnetic field, not producing a torque. This is why the actuation of
this manoeuvre has been performed through the use of three reaction wheels.

13
 Once the ideal required control has been established from section 4.3, the second
part of the actuation process is the control distribution, which allows to allocate the
continuous force obtained from the ideal control to the actual rate of change of angular
momentum. This is to be applied to each individual reaction wheel, that is the effect
that actually apply the momentum exchange as in Eq. (36), where A˚RW is the pseudo
inverse of the rotor configuration (simple inverse in this case because there are only three
reaction wheels).
h9 RW “ ´A˚RW puid ` pω ^ ARW hRW qq (36)
Once integrated and saturation checked with the limits imposed by Table 9, the real
control obtained is found and applied to the system, as in Eq. (37):

T RW “ ´pω ^ ARW hRW q ´ ARW h9 RW (37)

5 Results
5.1 Detumbling
5.1.1 Dynamic
The following Fig. 4 represents the variation of the vector ω with respect to time. The
quantity showed here is the one measured by the gyro, the one that the spacecraft’s
systems can actually have access to. The error with respect to the "real" spacecraft
angular velocity is always bounded in the 10´3 rad/s order of magnitude.
As requested by this control, all the three components go near zero. As it can be
seen, this manoeuvre is not very fast due to the small action that the magnetic actuators
can provide. The oscillations around zero velocity become permanently small (in the
order of 10´3 rad/s) after about 1{4 orbit period. The manoeuvre continues for another
quarter of an orbit until apogee is reached, at the end of which the first phase will be
considered completed and the tracking manoeuvre can begin.

Figure 4: Detumbling Angular Velocity

14
5.1.2 Control
In the first 1000 s the angular velocity is still high and this reflect in the saturation of the
magnetorquer as showed in Fig. 5. The noise in the magnetic dipole is most probably
caused by the inevitable noise of the magnetic field measurement.
The control torque applied to slow down the satellite is showed in Fig. 6, which has to
both slow down the Cubesat and counteract the effect of the environmental disturbances.
It can be observed indeed how the higher effort is done during the first half of the
mission phase, afterwards the action of the magnetorquers is one order of magnitude
lower (« 10´6 N) only to keep the satellite at an almost null angular velocity.

Figure 5: Detumbling Magnetic Dipole with Saturation Levels

Figure 6: Detumbling Control Torque

5.1.3 Perturbations
In Fig. 7 the disturbances acting on the satellite during this phase are depicted. The
plot is in logarithmic scale in order to compare their different orders of magnitude.

15
 As it’s possible to see, the drag perturbation is limited in magnitude compared to
the others, with a lowest peak at the end of the mission phase as expected because
the spacecraft is at apogee. The other perturbing effect are all more significant to the
spacecraft dynamic, with a stronger emphasis on the magnetic torque.

Figure 7: Detumbling Disturbances Torques

5.2 Sun Tracking

5.2.1 Dynamic
The Sun Tracking is simulated for one complete revolution of the spacecraft around
the Earth. Once the angular velocity has been reduced sufficiently, the reaction wheels
control can engage in the system, which helps in stabilise further more the angular
velocity to an order of magnitude of 10´5 rad/s, as seen in Fig. 8.

Figure 8: Tracking Angular Velocity

It is important to assess the precision with which the system is determining the
attitude of the satellite. The attitude determination error has been computed as in

16
Eq. (38) and it’s plot is shown in Fig. 9, which oscillates in the order of magnitude of
10´5 rad, a reasonably acceptable error.

erratt “ tracepI ´ ABN ATBN,sens q (38)

Figure 9: Tracking Attitude Determination Error

The most important parameter to check when performing a tracking manoeuvre is

the pointing error for each axis of the attitude matrix as in algorithm 3.
Algorithm 3: Pointing Error Assessment
for Each Column of the Actual Attitude and Desired Attitude Matrix do
With i “ rx, y, zs and n “ r1, 2, 3s
ABN,sens p:, nq
psens,i “
|ABN,sens p:, nq|
Adesired p:, nq
pdes,i “
|Adesired p:, nq|
Angle of pointing error, between each axis of desired and attitude matrix
errpoint “ prppsens,i ¨ pdes,i qs1 q
In Fig. 10 it can be seen how even starting from a high initial pointing angle, the
satellite achieves a correct pointing in the first 1000 s of operations. A special attention
is reserved to the x-axis pointing, the one containing the solar panel and to be correctly
positioned in space with this manoeuvre.

5.2.2 Control
As already done for the detumbling, it is important to check whether the saturation
levels of the actuators are reached or not. For this manoeuvre that involves reaction
wheels, it is particularly important to avoid reaching such values since they would need
to be de-saturated. As can be seen in Fig. 11 and Fig. 12 the saturation never occurred.
In Fig. 13 it’s possible to see the actual torque applied to the system by this actuators.

17
 Figure 10: Tracking ADCS Pointing Error

Figure 11: RW’s Angular Momentum with Saturation Levels

Figure 12: RW’s Angular Momentum Rate of Change with Saturation Levels

18
 Figure 13: Tracking Disturbances Torques

5.2.3 Perturbations
It can be interesting to observe the disturbances acting on the satellite during the tracking
mission in Fig. 14. It can be noted how the solar radiation torque is almost constant
since the satellite is pointing the sun always with the same face.

Figure 14: Tracking Disturbances Torques

6 Robustness Analysis
The results from this simulation are relative to the chosen initial conditions.
An analysis on multiple conditions of detachment from the main launcher body can
be investigated to study the limits of applicability of the control system and the respon-
siveness of the system itself.
To do so, the same Cubesat has been simulated with different initial angular velocity,
increasing from 5 deg/s up to the limit of non-compliance, to assess the response of
the system and whether the detumbling is effectively obtained or not, choosing as a

19
benchmark of success the achievement by the end of a full orbit of a residual angular
velocity of less than 0.5 deg/s.
The results are collected in Table 10, from which it is possible to conclude that
the system, according to the above specifications, is robust to some changes in initial
conditions up to « 25 deg/s.

Table 10: Statistical Analysis Over Different ICS

ICS [deg/s] [5,5,5] [15,15,15] [25,25,25] [35,35,35]

Controlled? Y Y Y N
Time to Reach
Benchmark [s] 5994 6728 8004 -

7 Conclusions and Further Developments

The Cubesat works properly under the made assumptions and the system is capable
both to detumble and to track the Sun. However, several aspects have been neglected
that should be taken into account for a more accurate analysis.
First of all a mass budget and a power budget should be performed, in order to
verify the actual feasibility of the mission. For example, the solar panel mounted on one
face may be too small for the power requested from the spacecraft, and in this case a
different configuration of solar panel should be adopted, like arranging multiple layers of
solar panels on the same face and then deploy them once in orbit.
A strong assumption made is to consider the attitude dynamics uncoupled from the
orbital one, which is not always true; for example drag causes also the deceleration of
the spacecraft, a circularisation of its orbit and a loss of altitude, ultimately changing
the orbit itself and changing also the pointing of the spacecraft.
Providing the satellite with more sensors could be useful to achieve a better atti-
tude determination, as well as enhancing their working precision. In particular, some
noisy results due to some sensors (like the magnetometers) might be avoided. However,
working with more sensors would require to change the algorithm used for the attitude
determination.
Another limit is that the analysis have been conducted under the assumption of struc-
tural rigid body. Considering that torques, even if small, are applied to the spacecraft
in a quite short amount of time, thus flexibility should be taken into account, together
with the acceleration limits of the structure and the components.
Moreover there is further space to analyse the communications capabilities of the
satellite with the ground station; uncertainty over the masses or inertia variation due to
moving component inside the spacecraft; other off nominal condition that may occurs;
but they are beyond the scope of this report and are to be intended as a starting point
for the continuation of an in-depth study in the future.

20
References
[1] S. Williams, J. Puig-Suari, R. Coelho, J. Carnahan. California Polytech-
nic State University. CubeSat Design Specification. 2014, ESA, Phi Lab.
https://blogs.esa.int/philab/files/2019/11/RD-02_CubeSat_Design_
Specification_Rev._13_The.pdf

[2] ISIS, Solar Panels. Delft, Netherlands. https://www.cubesatshop.com/

wp-content/uploads/2016/06/ISIS-Solar-Panels-Brochure-v1.pdf

[3] E. Thébault, C. C. Finlay, C. D. Beggan, et al. International Association of Geo-

magnetism and Aeronomy (IAGA), 13th Generation International Geomagnetic Ref-
erence Field Model and Coefficients. 19 November 2019. https://www.ngdc.noaa.
gov/IAGA/vmod/igrf.html

[4] Sensonor, STIM300 Gyroscope. Horten, Norway. https://www.sensonor.com/

products/inertial-measurement-units/stim300/

[5] New Space Systems, NFSS-411 Sun Sensor. South Africa. UK. https:
//www.cubesatshop.com/wp-content/uploads/2020/11/NewSpace-Sun-Sensor_
2020-10a.pdf

[6] Adcole Maryland Aerospace, MAI-SES Earth Horizon Sensor. United States. https:
//satcatalog.com/component/ai-ses-ir-earth-sensor/

[7] NewSpace Systems, 3 Axis Magnetometer, NMRM-Bn25o485. South

Africa. UK. https://www.cubesatshop.com/wp-content/uploads/2020/11/
NewSpace-Magnetometer_11.2020_10a.pdf

[8] GOM Space, 3 Axis Magnetorquer, NanoTorque GST-600. Denmark.

https://gomspace.com/shop/subsystems/attitude-orbit-control-systems/
nanotorque-gst-600.aspx

[9] GOM Space, Reaction wheel for nanosatellites, NanoTorque GSW-600.

Denmark. https://gomspace.com/UserFiles/Subsystems/datasheet/
gs-ds-nanotorque-gsw-600-20.pdf

[10] D. Mostaza Prieto, B. P. Graziano, P. C. E. Roberts. Spacecraft drag

modelling. 2016, Space Research Centre, Cranfield University, Pag: 3, DOI:
10.1016/j.paerosci.2013.09.001 https://www.research.manchester.ac.uk/portal/
files/52641369/SCDragModPreprint.pdf

[11] D. A. Vallado. Fundamentals of astrodynamics and applications. 1997, McGraw Hill,

Pag: 510.

[12] J. R. Wertz. Spacecraft Attitude Determination and Control. 1978, Ed. J. R. Wertz,
Pag: 820.

[13] J. Davis. Mathematical Modeling of Earth’s Magnetic Field. Virginia Tech, Blacks-
burg, VA 24061 May 12, 2004, Pag: 8.