STUDSAT-2
Divyanshu Sahay, G S Tejaswini, Santosh K, Yuvaraj S, Rakshit
Dr S Sandya, Kannan T, Department of R&D
Nitte Meenakshi Institute of Technology
Bangalore-560064, India
divyanshu.udaysahay@gmail.com
Abstract²This paper describes the attitude control approach Furthermore, the mission aims to de-orbit one of the Twins,
of a nanosatellite in different scenarios of reaction wheel STUDSAT-2B, using a drag sail after mission life and
failure. The approach to allocate the control torques across the image earth with the one left orbiting i.e. STUDSAT-2A.
4 wheels is explained considering the cases of one and two
wheel failures. A Linear Quadratic Regulator is used to find Such an elaborate mission design requires the Attitude
the gains of the Proportional Derivative (PD) controller
Determination and Control Subsystem (ADCS) to be robust
employed to determine the control torques to be generated by
the reaction wheels in order to correct a specific attitude error and failsafe. Either of the twin satellites have different
and damp angular rates. The different analysis done in ADCS configuration. Among sensors, STUDSAT-2A hosts
attempts to stabilize the satellite for all the aforesaid conditions four sun sensors, a magnetometer and a gyro. STUDSAT-
has been presented. The paper also describes the electrical 2B has all of them except Sun Sensors as it does not have
modeling of Brushless DC motor to get the transfer function imaging requirements. Both the satellites have a tetrahedral
and obtain P, I (Integral) and D parameters for optimum Reaction Wheel and three orthogonally placed Magnetic
motor performance. ARM STM32F407 is used as the main Torquer Coils. The paper will discuss in detail about the
microcontroller for the satellite. The operation flow of the different tests and analysis done for the reaction wheel
implementation of control approach is explained. An account
actuator arrangement selected for the mission.
of different interfaces and peripheral used to realize control
system design is described.
TABLE OF CONTENTS
2. REACTION WHEEL TORQUE ALLOCATION
1. INTRODUCTION ....................................................... 1
A control law aims at designing a controller that would take
2. REACTION WHEEL TORQUE ALLOCATION .......... 1 minimum control energy to take the output of a system to a
3. LINEAR QUADRATIC REGULATOR ........................ 2 desired state. Selection of a control approach depends upon
4. SATELLITE MODELING .......................................... 3 the control law used and the actuator selected. Most
5. ATTITUDE CONTROL .............................................. 4 commonly used actuators include reaction wheels, control
6. BLDC MODELING .................................................. 5 momentum gyros, thrusters, torquer etc. In STUDSAT-2,
Tetrahedral Reaction wheels are used primarily because of
7. BLDC TUNING ........................................................ 8
three reasons. First, the wheel configuration can help attain
8. IMPLEMENTATION ................................................ 10 a Zero momentum bias arrangement. Second, the
9. SUMMARY ............................................................. 10 configuration can sustain the failure of one wheel and third,
APPENDIX.................................................................. 10 the arrangement can produce double the amount of
A. TORQUE ALLOCATION MATRICES ..................... 10 maximum torque on an axis than that of single wheel [1].
REFERENCES............................................................. 11 Figure 1, taken from [1] shows the schematic arrangement
of the wheel.
BIOGRAPHY .............................................................. 11
The control approach should use a control law to find the
torque to be produced by the four wheels for a particular
1. INTRODUCTION attitude acquisition. In STUDSAT-2 the Linear Quadratic
Student Satellites provide students and young researchers Regulator is used to determine the gains of a PD control law
the opportunity to understand the complexity and challenges employed to reduce the attitude errors by optimum use of
that satellite development comes with. Low Cost, minimum control torque. This will be explained in detail in following
development time and competency to address wide range of sections. This section is dedicated to explain the process of
application has enabled them to find a place in space torque allocation in the four tetrahedral wheels for a specific
industry.STUDSAT-2 (Student Satellite) is such a mission three axis torque requirement.
by a group of students at Nitte Meenakshi Institute of
Technology, Bangalore, India with a mission objective to A three axis stabilized system has three degrees of freedom
design a Twin-Nanosatellite to establish Inter Satellite along which the torques are generated to reorient the axes to
Communication after performing an In-Orbit Separation. the desired one.
978-5090-1613-6/17//$31.00 ©2017 IEEE
1
Moore-Penrose Pseudo Inverse
The Moore±Penrose (MP) Pseudo inverse method is used to
compute the least square solution to the system of linear
equation without a unique solution.The MP Pseudo inverse
can be used to calculate the allocation matrix, ߬, from the
system of linear equations represented by Eq.5 and Eq.6.
With four wheels at disposal, it becomes an over actuated Where, ܦ᪽ represents the pseudo inverse of ܦ
system control problem.
This paper gives an account of the study made for reaction
For a tetrahedral wheel schematic the reaction wheel torques wheel failure cases, so the allocation matrix ߬ should be
are distributed along the satellite body axes as follows, calculated for one and two wheel failures. In either of the
failure cases the column corresponding to the failed wheel is
ܶ௫ ൌ ܶଵ
ሺͳͻǤͶͳሻ െ ܶଶ
ሺͳͻǤͶͳሻ
ሺͲሻ െ first removed from the distribution matrix and the resulting
ܶଷ
ሺͳͻǤͶͳሻ
ሺͲሻ (1) matrix is evaluated for being invertible. If true, then the
inverse is calculated for the same, if not then the MP Pseudo
ܶ௬ ൌ ܶଶ
ሺͳͻǤͶͳሻ
ሺ͵Ͳሻ െ inverse is used. The column corresponding to the failed
ܶଷ
ሺͳͻǤͶͳሻ
ሺ͵Ͳሻ (2) wheel in the inverted matrix is appended with zeros to give
the final torque allocation matrix߬.
ܶ௭ ൌ െܶଵ ሺͳͻǤͶͳሻ െ ܶଶ ሺͳͻǤͶ ͳሻ െ
ܶଷ
ሺͳͻǤͶͳሻ ܶସ ሻ (3) Normal Operation²As mentioned above, with all the
wheels functioning as expected, the torque allocation matrix
is obtained by finding the MP Pseudo Inverse from Eq.7.
where ܶ௫ , ܶ௬ and ܶ௭ are the torque of the body axes and ܶଵ ,
ܶଶ , ܶଷ and ܶସ are the reaction wheel torques. From this, it
One Wheel Failure²For one wheel failure, the resulting
can be written,
distribution matrix after removing the column
் corresponding to the failed wheel is invertible. So, the
ൣܶ௫ ǡ ܶ௬ǡ ܶ௭ ൧ ൌ ܦሾܶଵ ǡ ܶଶ Ǥ ܶଷǡ ܶସ ሿ் (4) inverse is calculated and the column corresponding to the
failed wheel is appended with zero to find the torque
Eq. 4 can further be generalised as, allocation matrix. Four matrices are obtained for each of the
wheel failure cases.
ܶሺଷൈଵሻ ൌ ܦሺଷൈସሻ ܶோሺସൈଵሻ (5)
Two Wheel Failure²For two wheel failure, the resulting
Where, ܦis the distribution matrix given by, distribution matrix has to be inverted by the MP Pseudo
Inverse method from Eq.8. Six cases are possible for
ሺͳͻǤͶͳሻ െ
ሺͳͻǤͶͳሻ
ሺͲሻ Ȃ
ሺͳͻǤͶͳሻ
ሺͲሻ Ͳ following combination of wheel failures, (1st and 2nd), (1st
ܦൌ Ͳ
ሺͳͻǤ Ͷͳሻ
ሺ͵Ͳሻ െ
ሺͳͻǤͶͳሻ
ሺ͵Ͳሻ Ͳ
െሺͳͻǤͶͳሻ െሺͳͻǤ Ͷͳሻ െሺͳͻǤͶͳሻ ͳ
and 3rd), (1st and 4th), (2nd and 3rd), (2nd and 4th) and (3rd and
4th). The torque allocation matrices obtained is tabulated in
For an actuation problem, the control law gives the control Appendix.
torque, ܶ to be allocated among the reaction wheels. It can
be written as,
3. LINEAR QUADRATIC REGULATOR
ܶோ ൌ ߬ܶ (6)
A Linear Quadratic Regulator (LQR) finds a controller that
where ߬is the torque allocation matrix. It is evident from the would minimize the controlled output, ݖሺݐሻ, of a system at
discussion made above that ߬ will not have a unique the minimum expense of the control signal, ݑሺݐሻ. This is
solution. called optimum regulation.
2
Considering the energy of the controlled output as For such a system the LQR criterion in Eq.9 can be written
in a more general form as
ஶ
න ԡݖሺݐሻԡଶ ݀ݐ ஶ
ܬொோ ൌ ݔሺݐሻ் ܳݔሺݐሻ ݑሺݐሻ் ܴݑሺݐሻ ݀ݐ (11)
and the control input as The optimal state feedback LQR controller is of the form
ஶ
න ԡݑሺݐሻԡଶ ݀ݐ ݑൌ െݔܭ (12)
The feedback gain K is the ݉ ൈ ݊ matrix given by [4]
LQR tries to minimize the criterion given by
ஶ ܭൌ ܴ ିଵ ܲ ்ܤ (13)
ܬൌ ܳԡݖሺݐሻԡଶ ܴԡݑሺݐሻԡଶ ݀ݐ (9)
where ܲ is the solution to the Algebraic Riccati Equation:
where Q and R are weighing parameters in the form of
symmetric positive-definite matrices. By minimizing this ܲ ்ܣ ܲܣ ܳ െ ܲିܴܤଵ ܲ ்ܤൌ Ͳ (14)
criterion, LQR seeks to find a controller that would
minimize both the energies. However, a small control signal The state feedback law results in a closed loop system of the
will lead to a large control output and a small controlled form
output will demand large control signal. Q and R form the
tradeoff between these conflicting goals [2]. ݔሶ ൌ ሺ ܣെ ܭܤሻݔ (15)
A system, to be controlled, should be modeled in order to
A crucial property of LQR controller design is that the
explicitly specify the measurable thus controllable
resulting closed loop system is asymptotically stable. Eq.11
parameters. Consider a system as shown in Figure 2
is referred as cost quadratic function, which represents the
cost of the system and must be reduced to minimum.
ݑሺݐሻʲԹ ݔሶ ൌ ݂ሺݔǡ ݑሻ ݕሺݐሻʲԹ
ݕൌ ݃ሺݔǡ ݑሻ
4. SATELLITE MODELING
Figure 2. System having m inputs and k outputs Satellite Dynamics
)URP WKH (XOHU¶V VDWHOOLWH HTXDWLRQ RI PRWLRQ WKH WRWDO
Where xʲԹ is the state of the system and y is the angular momentum of the satellite is given by
measured output. Assuming function ݂and ݃ to be linear
and thus the system as Linear Time Invariant, the system ݄ሶ௧௧ ൌ ܰ௫௧ െ ߱ ൈ ݄௧௧ (16)
could be written in a general state space form as,
ݔሶ ൌ ݔܣ ݑܤ (10) where ݄௧௧ is the total angular momentum considering all
ݕൌ ݔܥ ݑܦ the parts within the satellite, ܰ௫௧ is the external torque and
߱is the angular velocity of the satellite. So, Eq.16 can be
where ୬ൈ୬ is the system matrix that relates the current state rewritten considering the reaction wheel dynamics as
of the system to the change in the same, ୬ൈ୫ is the control
matrix that describes the effect of control input on the state
߱ ܫ ݄ሶ௪ ൌ ܰ௫௧ െ ߱ ൈ ܫ௦ ߱ െ ߱ ൈ ݄௪
௦
(17)
change, ୩ൈ୬ is the output matrix relating the states of a
system to its output and ୩ൈ୫ is called the feed forward where ܫ௦ is satellite inertia matrix and ݄௪ is the total angular
matrix that allows the control input to effect the output momentum of the wheels. Eq.17 can be rewritten as
directly. More often than not, a LQR tries to control the
state x of the system resulting in a State Feedback LQR ߱ሶ ൌ െܫ௦ ିଵ ሺ߱ ൈ ܫ௦ ߱ሻ െ ܫ௦ ିଵ ሺ߱ ൈ ݄௪ ሻ െ ܫ௦ ିଵ ݄ሶ௪ ܫ௦ ିଵ ܰ௫௧ (18)
represented in Figure 3.
The control torque,ܰ in the body coordinate system is the change
ŽŶƚƌŽůůĞƌ ݔሶ ൌ ݔܣ ݑܤ in the rate of angular momentum of the reaction wheels i.e. ݄ሶ௪ . So,
ݕൌ ݔܥ ݑܦ Eq.18 can be rewritten as
െ
džሺݐሻʲԹ݊ ߱ሶ ൌ െܫ௦ ିଵ ሺ߱ ൈ ܫ௦ ߱ሻ െ ܫ௦ ିଵ ሺ߱ ൈ ݄௪ ሻ െ ܫ௦ ିଵ ܰܿ ܫ௦ ିଵ ܰ௫௧ (20)
3
Satellite Kinematics The quaternion attitude error and the angular velocity of the
satellite comprise the state of the system. The controlled
The satellite kinematics is represented by using the attitude
input is the torque exerted by the reaction wheel. So,
quaternion, ݍൌ ሾݍ௫ ǡ ݍ௬ ǡ ݍ௭ ǡ ݍఏ ሿ. A quaternion consists of
ݔሺݐሻ ൌ ሾݍ௫ ǡ ݍ௬ ǡ ݍ௭ ǡ ߱ ݔǡ ߱ ݕǡ ߱ ݖሿ and ݑሺݐሻ ൌ ሾܶ௫ ǡ ܶ௬ ǡ ܶ௭ ሿ.
vector, ݍሬሬሬሬԦ௩ ൌ ሾݍ௫ ǡ ݍ௬ ǡ ݍ௭ ሿ, that represents the axis about
which the satellite rotation has to be done and the scalar, Eq.20 and Eq.24 is linearized at ݍ ൌ ሺͲǡͲǡͲǡͳሻ and
ݍఏ gives the rotation angle,ߠalong that axis as follows, ߱ ൌ ሺͲǡͲǡͲሻ for computational ease. So, the state space
matrices can be given by [3],
ݍ௫
ݍ௬ ԡԦԡǤ௦ሺఏȀଶሻ ݀ݍሶ ݀ݍሶ
ݍൌ ൦ ݍ൪ ൌ ቂ ቃ ݀ݍሶ
௭ ௦ሺఏȀଶሻ
(21) ۍ ې ۍ ې
݀ݍ ݀ ܶܿ ۑ
ݍఏ ܣൌ ێێ ݀߱ ۑ ܤൌ ێ
݀ ߱ሶ ݀ ߱ሶ ۑ ߱݀ ێሶ ۑ
ێ ۑ
where ԡ݁Ԧԡ represents the normalized rotation axis. The ݍ݀ۏ ݀߱ ےሺ ے ܿܶ ݀ ۏሺ
బ ǡ߱బ ሻ బ ǡ߱బ ሻ
kinematics of the satellite is the rate of change of satellite
attitude, ݍሶ given by, ͳ ڮ Ͳ
ܥൌ ڭ ڰ ڭ൩ ܦൌ ሾͲሿሺൈଷሻ
ݍ௫ Ͳ ߱௭ െ߱௬ ߱ݔ ݍ௫ Ͳ ڮ ͳ ሺൈሻ
ௗ ݍ௬ ଵ െ߱௭ Ͳ ߱߱ ݔ௬ ݍ௬
൦ ݍ൪ ൌ ൦߱ ൪൦ ൪ (22)
ௗ௧ ௭ ଶ ௬ െ߱ݔ Ͳ ߱௭ ݍ௭ So, ܣand ܤcan be written as,
ݍఏ െ߱ݔ െ߱௬ െ߱௭ Ͳ ݍఏ
ሾͲሿሺଷൈଷሻ ͲǤͷሾܫሿሺଷൈଷሻ ሾͲሿሺଷൈଷሻ
ܣൌቈ ܤൌ ൨
ሾͲሿሺଷൈଷሻ െͳ ሺ ሻ െͳ
A satellite rotation is performed in order to correct a specific
attitude quaternion error, ݍ Ǥ The quaternion rotation to where ሺ୵ ሻ represents the cross product. ୱ is the satellite
align the satellite body coordinate system to the target inertia matrix. For STUDSAT-2, ୱ = [0.26, -0.002, 9.6e-4; -
coordinate system or vice-versa is given by [3], 0.002, 0.344, 0.004; 9.6e-4, 0.004, 0.284]
So, considering that quaternion ݍ௦ represents rotation from 5. ATTITUDE CONTROL
inertial to satellite body coordinate system and ݍ from A PD control law is used to correct the satellite attitude
inertial to target coordinate system. We may write given by,
4
ሾܭǡ ܲǡ ܧሿ ൌ ݈ݎݍሺܣǡ ܤǡ ܳǡ ܴሻ (30) Failure Cases
As mentioned before P is the solution to the Riccati The torque allocation matrices are calculated for the
equation and E is the eigen value to the closed loop system aforementioned wheel failure scenarios. The failure cases
represented by ݔሶ ൌ ሺ ܣെ ܭܤሻ[4]. were tested and the responses in each of them were found
different. For explanation, only one of each will be
PD Controller Gain explained in the paper. However, the torque allocation for
all of the cases can be referred from appendix A. In all the
The controlled input for the state space obtained can be cases, system response will be analyzed for change in
written using the inference made in Eq.26. The controller attitude error and angular velocity. The initial attitude error
gain obtained from the LQR is a (3ൈ ሻ matrix which may is assumed to be ݁ݍൌ ሺͳǡͳǤʹǡͳǤ͵ሻ and the initial angular
be represented in general form as, velocity is taken as zero in all directions. The ܳand ܴ
weighing matrices were approximately maintained at
ܭଵଵ ܭଵଶ ܭଵଷ ܭଵସ ܭଵହ ܭଵ
ܳ ൌ ͲǤͳ ݁ݕ݁ כሺሻ and ܴ ൌ ʹͲ ݁ݕ݁ כሺ͵ሻ across all the
ܭൌ ܭଶଵ ܭଶଶ ܭଶଷ ܭଶସ ܭଶହ ܭଶ ൩ simulations. Figure.4, Figure.5, Figure.6, Figure.7, Figure. 8
ܭଷଵ ܭଷଶ ܭଷଷ ܭଷସ ܭଷହ ܭଷ and Figure. 9 shows the simulation results obtained for the
cases of failures and normal operation.
Substituting in Eq.(26)
ܭଵଵ ܭଵଶ ܭଵଷ ܭଵସ ܭଵହ ܭଵ It can be seen that for normal operation and one wheel
ܭ ܭଶଶ ܭଶଷ ܭଶସ ܭଶହ ܭଶ ൩ ݔሺݐሻ failure scenario the responses are almost the same. This
ܶ ൌ െ ଶଵ
validates the argument made in Section 1 for the selection of
ܭଷଵ ܭଷଶ ܭଷଷ ܭଷସ ܭଷହ ܭଷ
tetrahedral reaction wheel. However, for the case of two
This can be written, ZKHHO IDLOXUH WKH DWWLWXGH GRHVQ¶W VHHP WR FRQYHUJH ,Q WKH
case of angular velocity it was observed that even though
ݍ௫ the initial angular velocity was assumed zero the actuation
ݍۍ௬ ې leads to an overshoot and with time goes back to zero.
ܶ௫ ܭଵଵ ܭଵଶ ܭଵଷ ܭଵସ ܭଵହ ܭଵ ۑ ێ
ݍ Similar to attitude error, the angular velocities in the normal
ܶ௬ ൌ െ ܭଶଵ ܭଶଶ ܭଶଷ ܭଶସ ܭଶହ ܭଶ ൩ ߱ ێ௭ ۑ
operation and one wheel failure case is almost same. In the
ܭଷ ۑ ߱ ێ
ݔ
ܶ௭ ܭଷଵ ܭଷଶ ܭଷଷ ܭଷସ ܭଷହ
ۑݕ ێ two wheel failure case, unlike attitude error, the angular
ے ݖ߱ ۏ velocity converges but only requires much longer duration.
ܶ௫ ܭଵଵ ݍ௫ ܭଵଶ ݍ௬ ܭଵଷ ݍ௭ ܭଵସ ߱ ݔ ܭଵହ ߱ ݕ ܭଵ ߱ݖ The study for the failure analysis is still in process and the
ܶ௬ ൌ െ ܭଶଵ ݍ௫ ܭଶଶ ݍ௬ ܭଶଷ ݍ௭ ܭଶସ ߱ ݔ ܭଶହ ߱ ݕ ܭଶ ߱ ݖ authors propose to perform high fidelity testing for the all
ܶ௭ ܭଷଵ ݍ௫ ܭଷଶ ݍ௬ ܭଷଷ ݍ௭ ܭଷସ ߱ ݔ ܭଷହ ߱ ݕ ܭଷ ߱ݖ the failure cases in near future.
ܭଵଵ ܭଵସ
ݍ
ሾܶ௫ ሿ ൌ െ ܭଶଵ ܭଶସ ൩ ቂ ߱௫ ቃ (31)
௫
ܭଷଵ ܭଷସ 6. BLDC MODELING
ܭଵଶ ܭଵହ ݍ A Maxon 3W BLDC motor is used for the reaction wheel.
௬ The selection is done on the basis of power availability, cost
ൣܶ௬ ൧ ൌ െ ܭଶଶ ܭଶହ ൩ ቂ ߱ ቃ (32)
௬ and the feature to work nominally at 3000 rpm which is
ܭଷଶ ܭଷହ
selected as the maximum Reaction wheel speed. This
ܭଵଷ ܭଵ section is dedicated to explain the electrical modeling of the
ݍ
ሾܶ௭ ሿ ൌ െ ܭଶଷ ܭଶ ൩ ቂ ߱௭ ቃ (33) BLDC motor in order to obtain Simulink model of the same.
௭
ܭଷଷ ܭଷ The Simulink model is used to perform the PID tuning of
the motor so as to attain a optimum performance. The tuning
Comparing Eq.(31), Eq.(32) and Eq.(33) and Eq.(27), will be explained in section 7.
Eq.(28) and Eq.(29) we can conclude,
However, [1] presents a simplified modeling approach. In
ܭଵଵ ܭଵଶ ܭଵଷ this paper the detailed modeling is done and an optimum
ܭ௫ ൌ ܭଶଵ ൩ ǡ ܭ௬ ൌ ܭଶଶ ൩ ǡ ܭ௭ ൌ ܭଶଷ ൩ performance is achieved [5].
ܭଷଵ ܭଷଶ ܭଷଷ
Electrical Model
ܭଵସ ܭଵହ ܭଵ
The BLDC motor is a 3-Phase voltage system that is
ܭ௫ ൌ ܭଶସ ൩ǡ ܭ௬ ൌ ܭଶହ ൩ ǡ ܭ௭ ൌ ܭଶ ൩
electronically commuted thus increasing the motors life.
ܭଷସ ܭଷହ ܭଷ
The electrical model is a combination of three R-L Circuits
connected at the same node from one end. Applying
.LUFKKRII¶V voltage law the voltage drop in each arm can be
calculated as follows,
5
Figure. 4. Attitude error in Normal Operation
6
Figure. 7. Angular Velocity in Normal Operation
7
ௗೌ Mechanical Model
ܸܽ ൌ ܴ ݅ ܮ ݁ (34)
ௗ௧
The mechanical component inside the motor can be
ௗ್ modeled as,
ܸܾ ൌ ܴ ݅ ܮ ݁ (35)
ௗ௧
ௗఠ
ௗ ܶ௧௧ െ ܶௗ ൌ ܬ ܤǤ ߱ (45)
ܸܿ ൌ ܴ ݅ ܮ ݁ (36) ௗ௧
ௗ௧
Where ܬis the inertial of the rotor shaft and ܤis the friction
where ܸ , ܸ and ܸ are the phase voltages, ܴ ǡ ܴ and ܴ coefficient.
are the terminal resistance, ݅ ǡ ݅ and ݅ are the phase
currents, ܮ ǡ ܮ and ܮ are the terminal inductance and ݁ ǡ Forming Transfer Functions
݁ and ݁ are the back EMF.
Taking the Laplace Transform of Eq.34,
In BLDC motor the back EMF is a function of rotors
position. As the rotors in the three phases are 120 o apart, the ܸܽሺݏሻ ൌ ܴ ݅ ሺݏሻ ܮ ܵǤ ሺݏሻ ݁ ሺݏሻ
ଶగ
back EMFs generated will have a phase difference of . So
ଷ or,
the back EMFs may be represented as follows,
ܽሺ௦ሻିೌ ሺ௦ሻ
݅ ሺݏሻ ൌ (46)
ோೌ ାೌ ௌǤ
݁ ൌ ܭன ݂ሺߠ ሻ߱ (37)
Similarly Eq.35 and Eq.36 can be written as
ଶగ
݁ ൌ ܭன ݂ ቀߠ െ ቁ ߱ (38)
ଷ
ܾሺ௦ሻି್ ሺ௦ሻ ܿሺ௦ሻି ሺ௦ሻ
ଶగ
݅ ሺݏሻ ൌ ǡ݅ ሺݏሻ ൌ (47)
ோ್ ା್ ௌǤ ோ ା ௌǤ
݁ ൌ ܭன ݂ ቀߠ ቁ ߱ (39)
ଷ
where ܭன is back EMF constant, ߠ is the electrical rotation Taking Laplace Transform of Eq.45
angle and ߱ is the rotor speed.
ܶ௧௧ ሺݏሻ െ ܶௗ ሺݏሻ ൌ ߱ܵܬ ሺݏሻ ܤǤ ߱ ሺݏሻ
The total power delivered by the motor is given by,
ܶ௧௧ െ ܶௗ
ೌ ೌ ା್ ್ ା
ܶ௧௧ െ ܶௗ ൌ (44)
ఠ
8
Filhcclc
Figure 10. BLDC Simulink Block
9
Figure.13 and Figure. 14 shows the tuned response and the APPENDIX
optimum motor performance as obtained in the scope
respectively. The output is in rad/sec, it is converted to rpm A. TORQUE ALLOCATION MATRICES
to have better understanding. As mentioned in Section 6 the
Matrix (߬) Cases of Failure
nominal operation is 3000 rpm, same is obtained with
optimum performance.
Ͳ Ͳ Ͳ
െͳǤͲͲ ͲǤͳʹͶ Ͳ
൦ ൪ R1
8. IMPLEMENTATION െͳǤͲͲ െͲǤͳʹͶ Ͳ
െͲǤͲ Ͳ ͳ
STM32F407 based on ARM Cortex M4 architecture is used
to control the reaction wheel. Maxon motors have hall
sensor. The hall sensors measure the motor speed and are
available at the Motor Driver for access in the form of ͳǤͲͲ െͲǤͳʹͶ Ͳ
analog voltages. The analog signal is fed to the (Analog Ͳ Ͳ Ͳ
൦ ൪ R2
Digital Converter) ADC Channels of STM32F4, the ADC Ͳ െͳǤʹʹ Ͳ
are configured to be 8 bit as it matches the required ͲǤ͵ͷ͵ͷ െͲǤͳʹͶ ͳ
resolution. This ADC value is directly fetched from the
ADC data registers to the memory with the help of the
(Direct Memory Access) DMA. DMA bypasses the
controller and hence the algorithm can be executed fast and ͳǤͲͲ ͲǤͳʹͶ Ͳ
Ͳ ͳǤʹʹͶ Ͳ R3
efficiently. The flow chart shown in Figure 15 explains the ൦ ൪
process. Ͳ Ͳ Ͳ
ͲǤ͵ͷ͵ͷ ͲǤͳʹͶ ͳ
ͲǤͲͳ Ͳ െͳ
െͲǤ͵ͷ͵ ͲǤͳʹͶ െͳ
൦ ൪ R4
െͲǤ͵ͷ͵ െͲǤͳʹͶ ͳ
Ͳ Ͳ ͳ
Ͳ Ͳ Ͳ
Ͳ Ͳ Ͳ
൦ ൪ (R1 & R2)
െͲǤͷ͵Ͳ͵ െͲǤͻͳͺ Ͳ
െͲǤͳͺ െͲǤ͵Ͳʹ ͳ
Ͳ Ͳ Ͳ
െͲǤͷ͵Ͳ͵ ͲǤͻͳͺ Ͳ
൦ ൪ (R1 & R3)
Figure 15. Operation Flow Ͳ Ͳ Ͳ
െͲǤͳͺ ͲǤ͵Ͳʹ ͳ
9. SUMMARY
The paper described the control approach adopted for the Ͳ Ͳ Ͳ
െͲǤͲͳ ͲǤͳʹͶ െͲǤͷͲͲ
Reaction Wheel Control of STUDSAT-2. An account of ൦ ൪ (R1 & R4)
different failure cases is presented and the toque allocation Ͳ Ͳ Ͳ
matrix for each case has been tabulated. An explanation of െͲǤͲͳ െͲǤͳʹͶ െͲǤͷͲͲ
LQR is done specially for state feedback cases and the
process used to find the gains of the PD is presented. One of
HDFK IDLOXUH FDVH LV VLPXODWHG DQG LW¶V HVWDEOLVKHG WKDW WKH
ͳǤͲͲ Ͳ Ͳ
tetrahedral reaction wheel can very well sustain the failure
Ͳ ͲͲ R2 & R3
of one wheel. Furthermore, the BLDC motor used for the ൦ ൪
reaction wheel is modeled and tuned for optimum Ͳ ͲͲ
performance. The paper also touches upon the methodology ͲǤ͵ͷ͵ͷ Ͳ ͳ
of implementation of the control approach planned on the
satellite on-board Controller.
10
Tejaswini G S is a final year
student at Nitte Meenakshi
ͲǤͺͺ͵ͻ െͲǤ͵Ͳʹ െͲǤͷ Institute of Technology,
Ͳ Ͳ Ͳ pursuing M.Tech in Digital
൦ ൪ R2 & R4
െͲǤͳͺ െͲǤͻͳͺ െͲǤͷ Communication & Networking.
Ͳ Ͳ Ͳ She is a core member of
STUDSAT-2 Team. Her
interests are Control Systems
and Satellite Technology.
ͲǤͺͺ͵ͻ ͲǤ͵Ͳʹ െͲǤͷ
െͲǤͳͺ ͲǤͻͳͺ െͲǤͷ
൦ ൪ R3 & R4 Yuvaraj S is a final year student
Ͳ Ͳ Ͳ
Ͳ Ͳ Ͳ at Nitte Meenakshi Institute of
Technology pursuing B.E. in
Electronics and Communication
Branch. He is a Core Member of
STUDSAT-2 Team. His area of
REFERENCES interest are Embedded Systems,
[1] Kumar Saroj, Divyanshu Sahay, Sandesh R. Hegde, S. Programming and Satellite
Sandya, Anik Kumar Jha, T. C. Mahalingesh. "Design Technology.
and development of 3-axis reaction wheel for STUDSAT-
2." 2015 IEEE Aerospace Conference. IEEE, 2015 Santosh Kumar is a final year
student at Nitte Meenakshi
[2] J. P. Hespanha, "Undergraduate lecture notes on Institute of Technology
LQG/LQR controller design", University of California, pursuing B.E. in Electronics &
2007 Communication Branch. He is a
Core Member of STUDSAT-2
>@ %ODQNH 0RJHQV /DUVHQ 0DUWLQ %LUNHOXQG ³6DWHOOLWH
Team. His areas of interest are
Dynamics and Control in a Quaternion Formulation(2nd
Embedded Systems and
Edition), Technical University of Denmark, 2010
Prgramming.
[4] Ibrahim Kሷ N ³&RPSDULVRQ DQG $QDO\VLV RI D &RQWURO
6\VWHPVRID6DWHOOLWH8VLQJ5HDFWLRQ:KHHO$FWXDWRUV´
Luleܽሶ University of Technology, 2012 Kannan T is a Research Associate at Centre for Small
Satellite Research. He has Obtained his M.Tech from Nitte
[5] G.Prasad, N.Shree Ramya,P.V.N.Prasad, G.Tulasi Ram
Meenakshi Institute of Technology. He is the core leader of
Das, ³Modeling and Simulation Analysis of the Brushless
Command and Data Handling Subsystem. His area of
DC Motor by using 0$7/$%´,-,7((
interest includes Embedded System, Power Systems and
[6] Maxon EC motor, May 2012113 edition, EC 20 flat Communication
20mm, brushless 3 Watt, Maxon Flat Motor.
Rakshit is a final year student at Nitte Meenakshi Institute
of Technology pursuing B.E. in Electronics and
BIOGRAPHY
Communication Branch. He is a Core Member of
Divyanshu Sahay obtained STUDSAT-2 Team. His areas of interest are Embedded
bachelor of engineering in Systems, Programming and Satellite Technology.
Electronics & Communication
from Siddaganga Institute of Dr. S. Sandya obtained bachelor in Electrical engineering
Technology, Tumkur, India in the from Bangalore University, India. She is currently the
year 2014. He is currently working Project Director for STUDSAT-2 and Head of the
as a Research student at Nitte Department, Electronics and Communication, Nitte
Meenakshi Institute of Technology, Meenakshi Institute of Technology, Bangalore. She obtained
Bangalore. His areas of interests her PhD from Indian Institute of Science, Bangalore. She
are Physics, Robotics and Control has an Industrial experience of over 20 Years. Her area of
System. He has been working for Team STUDSAT-2 since specialization is Pulsed NQR Instrumentation.
2012. Presently, he is the core leader of Attitude
Determination and Control System for Project STUDSAT-2.
11