Spacecraft Attitude Dynamics and Control

(AE4313-20)

Course Notes No. 2

Course Instructor
Dr.ir. E. van Kampen

Section of Control and Simulation

Faculty of Aerospace Engineering
Office LB0.27
Tel.: 015-2787147
E-mail [email protected]

1
The Attitude Determination and Control System

2
• Previous lecture
– Course Introduction
– Kinematics
• Direction Cosine Matrices
• Euler angles: ordered sequential rotations
• Today:
– Eigenaxis rotations
• Quaternions
• Modified Rodrigues Parameters
– Kinematic differential equations
3
Euler’s eigenaxis rotation theorem

Euler’s eigenaxis rotation states that by rotating a rigid body about an axis
that is fixed to the body and stationary in an inertial reference frame, the
rigid-body attitude can be changed from any given orientation to any other
orientation. Such an axis of orientation, whose orientation relative to both
inertial reference frame and the body remains unchanged throughout the
motion, is called the Euler axis or eigenaxis.

Suppose that unit vectors ai and bi (i = 1, 2, 3) are fixed in reference frames

A and B, respectively. The orientation of B with respect to A is characterised
by a unit vector e along the Euler eigenaxis and the rotation about that axis,
as follows

e  e1a1  e2 a2  e3a3
 e1b1  e2b2  e3b3

where ei are the direction cosines of the Euler eigenaxis relative to both A
and B. The Euler eigenaxis rotation is defined in Fig. 2.2.
4
 a3 e

b3 
b2

a2

a1

b1 Fig. 2.2 Euler eigenaxis rotation

5
Since ei are the direction cosines,

e12  e22  e32  1

This is due to the fact that CC  I and ei occupy one row of the direction
T

cosine matrix C. We can also write:

 e1   e1 

a1 a2 a3 e2   b1 b2 b3 e2  
e3  e3 
 a1   e1   a1   e1 
a   a a a  e   a   b b b
 2 1 2 3  2 
 2 1 2 3  e 
 2
 a3  e3   a3  e3 
 e1   a1   e1  C11 C12 C13   e1 

e   a   b b b e   C 
 2   2  1 2 3  2   21 C22 C23  e2 
e3   a3  e3  C31 C32 C33  e3 
 e1  C11 C12 C13   e1 
e   C C23  e2 
 2   21 C22 Think about why this makes sense using fig 2.2!
e3  C31 C32 C33  e3  6
How can we get convert from an Euler eigenaxis description to a DCM?

To parameterize the direction cosine matrix C in terms of ei and  , a

sequence of rotations is used as follows:

1) Rotate the reference frame A, using a rotation matrix R, to align the a1

axis of A with the chosen direction e . Let A' be the new reference frame after
this rotation and also let A remain the original frame with basis vectors
a1, a2 , a3  . The rotation matrix can then be written as:
 e1 e2 e3 
C A' / A
 R   R21 R22 R23 
 R31 R32 R33 
The reason for this is that
 a1'   a1  C11 C12 C13   a1 
 ' A' / A    C
a
 2  C  2   21 C22
a C23  a2 
 a3'   a3  C31 C32 C33   a3 
 

if a1' is aligned with the Euler eigenaxis e , we have a1'  e1a1  e2 a2  e3a3
7
Same holds for other two axes if they are selected to be the eigenaxes.

 a1'  C11 C12 C13   a1   a1'  C11 C12 C13   a1 

 '   ' 
a2    e1 e2 e3  a2  a2   C21 C22 C23  a2 
 a3'  C31 C32 C33   a3   a3'   e1 e2 e3   a3 
   
This first step is illustrated in Fig. 2.3.

Step 2
Rotate both frames A and A' as a rigid body around direction e through
an angle  . After this eigenaxis rotation, the reference frame A will be aligned
 
with the reference frame B with basis vectors b1 , b2 , b3 , and A' will
''
become another reference frame A via the rotation matrix
1 0 0 
C A'' / A'
 C1    0 cos  sin  
0  sin  cos  

8
 a3

a1'  e

a2

a1

Fig 2.3 Step 1

9
It is important to notice that the relative orientation between A'' and B is the
same as the relative orientation between A' and A, i.e., C A / B  C A / A  R .
'' '

1
3) Rotate A'' through an inverse matrix R  R , then the reference frame
T

A'' will be aligned with B, since C B / A''  R 1 .

These three sequential rotations can be combined as

 b1   a1 
  B/ A  
b
 2  C  a 2
b3   a3 
 

where

C B / A  C B / A C A / A C A / A  C A / A C1  C A / A  RT C1  R
'' '' ' ' ' '

10
Substituting the definitions of these two matrices, we obtain the elements of
B/ A
the total direction cosine matrix C as:


C11  e12  R212  R312 cos  
C12  e1e2  R21 R22  R31 R32  cos   R21 R32  R22 R31 sin 
C13  e1e3  R21 R23  R31 R33  cos   R21 R33  R23 R31 sin 


C33  e32  R232  R332 cos  
In order to parameterise the direction cosine matrix in terms of only e1 , e2 , e3
and the rotation angle  , the properties of the direction cosine matrix have to
be applied. The first property to be applied is the orthonormality of the matrix:

 e1 R21 R31   e1 e2 e3  1 0 0
RT R  e2 R22 R32   R21 R22 R23   I  0 1 0
e3 R23 R33   R31 R32 R33  0 0 1

11
we obtain the following results from this property

e12  R21
2
 R31
2
1
e22  R22
2
 R32
2
1
e32  R23
2
 R33
2
1
e1e2  R21 R22  R31 R32  0
e1e3  R21 R23  R31 R33  0
e2 e1  R22 R21  R32 R31  0
e2 e3  R22 R23  R32 R33  0
e3e1  R23 R21  R33 R31  0
e3e2  R23 R22  R33 R32  0

We also have known that the direction cosine matrix elements can be
expressed using their cofactors:

e1  R22 R33  R23 R32 ; e2  R23 R31  R21R33 ; e3  R21R32  R22 R31

12
Using all these relationships, we have the following results:

 e1 R21 R31  1 0 0   e1 e2 e3 
RT C1  R  e2 R22 R32  0 cos  sin    R21 R22 R23 
e3 R23 R33  0  sin  cos    R31 R32 R33 

 e1 R21 cos   R31 sin  R21 sin   R31 cos    e1 e2 e3 

 e2 R22 cos   R32 sin  R22 sin   R32 cos    R21 R22 R23 
e3 R23 cos   R33 sin  R23 sin   R33 cos    R31 R32 R33 

The results of these matrix multiplications are written in elements as

C11  e12  R21 cos   R31 sin  R21  R21 sin   R31 cos  R31
 
 e12  R212  R312 cos 
 e12  1  e cos 
2
1

 cos   e12 1  cos  

13
C12  e1e2  R21 cos   R31 sin  R22  R21 sin   R31 cos  R32
 e1e2  R21 R22  R31 R32  cos   R21 R32  R22 R31 sin 
 e1e2   e1e2  cos   e3 sin 
 e1e2 1  cos    e3 sin 

C13  e1e3  R21 cos   R31 sin  R23  R21 sin   R31 cos  R33
 e1e3  R21 R23  R31 R33  cos   R21 R33  R23 R31 sin 
 e1e3   e1e3  cos    e2 sin 
 e1e3 1  cos    e2 sin 

C21  e2 e1  R22 cos   R32 sin  R21  R22 sin   R32 cos  R31
 e2 e1  R21 R22  R31 R32  cos   R22 R31  R21 R32 sin 
 e2 e1   e2 e1  cos    e3 sin 
 e2 e1 1  cos    e3 sin 
14
C22  e22  R22 cos   R32 sin  R22  R22 sin   R32 cos  R32
 
 e22  R222  R322 cos 
 e22  1  e cos 
2
2

 e22 1  cos    cos 

C23  e2 e3  R22 cos   R32 sin  R23  R22 sin   R32 cos  R33
 e2 e3  R22 R23  R32 R33  cos   R22 R33  R23 R32 sin 
 e2 e3   e2 e3  cos   e1 sin 
 e2 e3 1  cos    e1 sin 

Students may derive the last row the direction cosine matrix C.

15
 Finally, the direction cosine matrix in terms of e1 , e2 , e3 and  has the form:

 cos  e12 1  cos  e1e2 1  cos   e3 sin  e1e3 1  cos    e2 sin  

 
C  e,    e2e1 1  cos   e3 sin  cos  e22 1  cos  e2e3 1  cos   e1 sin  
e3e1 1  cos   e2 sin  e3e2 1  cos   e1 sin  cos  e32 1  cos   

Note that e1 , e2 , e3 are not independent between each other. They are constrained
by the relationship e1  e2  e3  1 .
2 2 2

By defining
 e1   0  e3 e2 
e  e2  ; E   e3 0  e1 
e3   e2 e1 0 

a more compact expression of the direction cosine matrix can be written as

C  e,   I cos  1  cos  ee  E sin 

T

where I is a 3X3 identity matrix.

16
Now we may come back to

 e1   C11 C12 C13   e1 

 e   C C23  e2 
 2   21 C22
 e3  C31 C32 C33   e3 

Using the result of the direction cosine matrix

C11e1  C12e2  C13e3  cos   e12 1  cos    e1  e1e2 1  cos    e3 sin   e2  e1e3 1  cos    e2 sin   e3
 e1 cos   e1e12 1  cos    e1e22 1  cos    e2e3 sin   e1e32 1  cos    e2e3 sin 
 e1 cos   e1 1  cos    e12  e22  e32   e1

C21e1  C22 e2  C23e3  e2 e1 1  cos    e3 sin   e1  cos   e22 1  cos    e2  e2 e3 1  cos    e1 sin   e3
 e2 e12 1  cos    e1e3 sin   e2 cos   e2 e22 1  cos    e2 e32 1  cos    e1e3 sin 
 e2 cos   e2 1  cos    e12  e22  e32   e2

C31e1  C32e2  C33e3  e3e1 1  cos    e2 sin   e1  e3e2 1  cos    e1 sin   e2  cos   e32 1  cos    e3
 e3e12 1  cos    e1e2 sin   e3e22 1  cos    e1e2 sin   e3 cos   e3e32 1  cos  
 e3 cos   e3 1  cos    e12  e22  e32   e3

17
 So applying the rotation C, defined by going from frame A to frame B,
on the vector e will not change the orientation of e!

a3 e

 e1   C11 C12 C13   e1 
  e   C C23  e2 
b3
 2   21 C22
b2  e3  C31 C32 C33   e3 

a2

a1

b1 18
 We may also obtain

E
1
2 sin 
 
C T  C ;   0,   ,  2 , ... (See Matlab example)

from which, the eigenaxis e can be found as

 e1  C23  C32 
e  e2   C  C 
1 (tip: check if this is true using the
2 sin   31 13 
highlighted equation on slide 16)
e3   C12  C21 

The rotation angle  can be found from:

cos  C11  C22  C33  1  trC  1

1 1
due to 2 2

C11  C22  C33  cos   1  cos  e12  cos   1  cos  e22  cos   1  cos  e32

 3 cos   1  cos   e12  e22  e32 
 3 cos   1  cos  
 2 cos   1
19
Consider two successive eigenrotations to A” from A represented by

C e1 ,1  : A'  A

C e 2 , 2  : A''  A'
where
C e1 ,1   I cos1  1  cos1 e1 e1  E1 sin 1
T

C e 2 , 2   I cos 2  1  cos 2 e 2 e 2  E2 sin  2

T

These two successive Euler eigen axis rotation can also be represented by
an equivalent single eigenaxis rotation to A” from A as follows

C e,  : A''  A
where
C e,   I cos  1  cos ee  E sin 
T

and
C  e,   C  e2 , 2  C  e1 ,1 
20
The equivalent eigenangle θ can be determined as
1
cos   trC  1
2
trC  cos1  cos 2  cos1 cos 2
 1  cos1 1  cos 2  cos 2   2sin 1 sin  2 cos 

where  is the angle between the two eigenaxes e1 , e 2 ; i.e.,cos   e1 e 2 .

T

This can be easily proved by

C  e,   C  e2 , 2  C  e1 ,1 

With the half angle theorem we can obtain the following for convenience

 1 2 1 2
cos  cos cos  sin sin cos 
2 2 2 2 2

21
The equivalent eigenaxis of the rotation e can also be found as

2e sin  e1 sin 1 1  cos 2   sin  2 1  cos1  cos  

 e2 sin  2 1  cos1   sin 1 1  cos 2  cos  
  e1  e2  sin 1 sin  2  1  cos1 1  cos 2  cos  

which can also be written in terms of half-angles

 1 2 2 1 1 2
e sin  e1 sin cos  e2 sin cos   e1  e2  sin sin
2 2 2 2 2 2 2

where

e1  e 2  E1 e 2

These equations for composite rotations of Euler eigenaxis rotations are

not very nice… We need something more elegant! 22
Quaternions (why?  more elegant composite rotations!)

Consider the Euler’s eigenaxis rotation about an arbitrary axis which is fixed in
a body-fixed reference frame B and stationary in an inertial reference frame A.
We define the four Euler parameters or quaternions as

q1  e1 sin  / 2 There are other

q2  e2 sin  / 2 possible
q3  e3 sin  / 2  definitions!!!

q4  cos / 2 

We define a vector q   q1 q3  such that

T
q2
q  e sin  / 2 
where e   e1 e2 e3  . Note that the quaternions are not independent
T

of each other. They are constrained again by the relationship

q q  q42  q12  q22  q32  q42  1

T

due to e1  e2  e3  1 .
2 2 2

23
By substituting the definition of quaternions, direction cosine matrix elements
in terms of quaternions can be found 𝛼
cos𝛼 = 1 − 2sin2
2
Using the half-angle identities: 𝛼
1 − cos𝛼 = 2sin2
2
𝛼 𝛼
sin𝛼 = 2sin cos
2 2

  
C11  cos   e12 1  cos    1  2 sin 2
2
 2e12 sin 2
2
 
 1  2 1  e12 sin 2
2

 
 1  2 e22  e32 sin 2
2

 1  2 q22  q32 
  
C12  e1e2 1  cos    e3 sin   2e1e2 sin 2  2e3 sin cos
2 2 2
    
 2 e1 sin e2 sin  e3 sin cos   2q1q2  q3q4 
 2 2 2 2
  
C13  e1e3 1  cos    e2 sin   2e1e3 sin 2  2e2 sin cos
2 2 2
    
 2 e1 sin e3 sin  e2 sin cos   2q1q3  q2 q4 
 2 2 2 2
24
Note that for deriving these results one may apply these equalities:

   
sin   2 sin   cos 
2 2
2  2  2  2 
cos  cos    sin    2 cos    1  1  2 sin  
2 2 2 2

Finally, we may write down the direction cosine matrix in terms of quaternions:

1  2  q22  q32  2  q1q2  q3q4  2  q1q3  q2q4  

 
CB/ A  

 C q, q4   2  q2 q1  q3q4  1  2  q1  q3  2  q2 q3  q1q4  
2 2


 2  q3q1  q2 q4  2  q3q2  q1q4  1  2  q12  q22  
 

25
Quaternions can also be written in terms of elements of direction cosine matrix

    
C11  C22  C33  1  2 q22  q32  1  2 q12  q32  1  2 q12  q22 
 3  4q2
1   
 q22  q32  3  4 1  q42  4q42  1
1
q4   C11  C22  C33  1
2
C23  C32  2q2 q3  q1q4   2q3q2  q1q4   4q1q4

q1 
1
C23  C32 
4 q4
C31  C13  2q3q1  q2 q4   2q1q3  q2 q4   4q2 q4

q2 
1
C31  C13 
4q4
C12  C21  2q1q2  q3q4   2q2 q1  q3q4   4q3q4

q3 
1
C12  C21 
4q4

26
Using the following definitions

 q1   0 q3 q2 
q   q2  , Q   q3 0 q1 
 q3   q2 q1 0 

the direction cosine matrix in terms of quaternions becomes

 
C  q42  q q I  2qq  2q4Q
T T

We can also write the quaternions in terms of the elements of direction

cosine matrix in a compact form:
1
q4   C11  C22  C33  1
2
C23  C32 
1 
q C31  C13  Singularity if q4 = 0!
4q4  
 C12  C21 
27
Quaternion singularity!

C23  C32 
1 
C31  C13 
1
q q4   C11  C22  C33  1
4q4   2
 C12  C21 

1
q4   C11  C22  C33  1  0
2
C11  C22  C33  1 Use slide 16 for conversion from C to e, 

cos   e12 1  cos    cos   e22 1  cos    cos   e32 1  cos    1

3cos   e12  e22  e32  1  cos    1
3cos   1  cos   1
2 cos   2
cos   1
  180
28
Consider two successive rotations to A” from A represented by

 '
C q , q4' : A'  A

C q , q  : A''  A'
'' ''
4

These successive rotations are also represented by a single rotation to A”

from A as follows

C  q, q4  : A''  A

We have

  
C  q, q4   C q , q4" C q , q4'
" '

The same relationship has been obtained earlier for the case of Euler eigen-
axis rotation

C  e,   C  e2 , 2  C  e1 ,1 

29
With the definition of quaternions:

 q1'   q1'' 
     
q   q2'   e1 sin 1 ; q   q2''   e 2 sin 2
' ''

2 2
 q3'   q3'' 
   
1 2
q4'  cos ; q4''  cos
2 2
We can obtain

q  q q  q q  q q ;
''
4
' '
4
'' ' ''
q4  q q  q
' ''
4 4   ' T
q
''

These can be proved as (use slide 21+22):

 1 2 1 2 1 2 1 2
q4  cos  cos  sin cos   cos  sin
T
cos sin cos sin e1 e 2
2 2 2 2 2 2 2 2 2
qq  q
' "
4 4   ' T
q
"

 1 2 2 1 1 2
q  e sin  e1 sin cos  e 2 sin cos   e1  e 2  sin sin
2 2 2 2 2 2 2
 q4" q  q4' q  q '  q
' " "
30
So the quaternion for composite rotations becomes:

 q1   q4'' q3''  q2'' q1''   q1' 

q   ''  
 2    q3 q4'' q1'' q2''  q2' 
No more trigonometric functions!!!
 q3   q2''  q1'' q4'' q3''   q3'  (compare slide 21/22)
   '' ''   ' 
q4   q1  q2''  q3'' q4  q4 
which is known as the quaternion multiplication rule in matrix form. The 4
by 4 orthonormal matrix is called the quaternion matrix. This relation can
also be written as

 q1   q4' q3' q2' q1'   q1'' 

q   '  
 2    q3 q4' q1' q2'   q2'' 
 q3   q2' q1' q4' q3'   q3'' 
   ' '   '' 
 q4   q1 q2' q3' q4   q4 
This 4 by 4 matrix is also orthonormal and is called the quaternion transmuted
matrix.
31
  q4 q3  q2  q1 
 q q4 q1  q2 
The orthonormality of the matrix 
3
can be simply proved as
 q2  q1 q4  q3 
 
 q1 q2 q3 q4 

 q4 q3  q2  q1   q4  q3 q2 q1 
 q q4 q1  q2   q3 q4  q1 q2 
 3
 q2  q1 q4  q3   q2 q1 q4 q3 
  
 q1 q2 q3 q4    q1  q2  q3 q4 

 q42  q32  q22  q12  q4 q3  q3q4  q2 q1  q1q2 q4 q2  q3q1  q2 q4  q1q3 q4 q1  q3 q2  q2 q3  q1q4 

 
 q q  q4 q3  q1q2  q2 q1 q32  q42  q12  q22  q3 q2  q4 q1  q1q4  q2 q3  q3 q1  q4 q2  q1q3  q2 q4 
 3 4
 q2 q4  q1q3  q4 q2  q3q1  q2 q3  q1q4  q4 q1  q3 q2 q22  q12  q42  q32 q2 q1  q1q2  q4 q3  q3 q4 
 
 q1q4  q2 q3  q3q2  q4 q1  q4 q3  q3q4  q2 q1  q1q2  q4 q3  q3 q4  q2 q1  q1q2 q12  q22  q32  q42 

1 0 0 0
0 1 0 0

0 0 1 0
 
0 0 0 1
32
Relation between quaternions and Euler angles (depends on order!!!)

 1 2 3 1  2 3 
sin cos cos  cos sin sin 
 q1    2 2 2 2 2 2
 3   3 
q  cos sin cos  sin cos sin 
1 2 1 2

 2   2 2 2 2 2 2
 q3   1 2 3 1  2 3 
   cos cos sin  sin sin cos
2 2 2 2 2 2
 q4    2 3 1  2 3 
cos cos cos  sin sin sin 
1
 2 2 2 2 2 2

Proof: In case of C1 1   C2  2   C3 3  rotations

 1   0   0 
 sin 2    2   0 
 0   sin    
  2   sin 3  Each rotation axis is
0 
 0  
   2 considered as an Euler
 1  cos    3 
2
cos 2   2  cos 2  eigenaxis.

33
  q1   q4''' q3'''  q2''' q1'''   q4'' q3''  q2'' q1''   q1' 
Using q   '''   
 2    q3 q4''' q1''' q2'''   q3'' q4'' q1'' q2''  q2' 
 q3   q2'''  q1''' q4''' q3'''   q2''  q1'' q4'' q3''   q3' 
   '''   
q4   q1  q2'''  q3''' q4'''   q1''  q2''  q3'' q4''  q4' 

 1 1    2 
 cos 0 0 sin   cos 2 0  sin 0  0 
 q1   2 2 2 2
1 1  2 2  0 
q   0 cos sin 0  0 cos 0 sin   
 2   2 2  2 2   sin  3 
 q3   1 1  2
   0  sin cos 0   sin 2 0 cos 0  2
2 2  2 2   3 
 q4    1   2  2  cos 2 
 sin 1 0 0 cos   0 sin 0 cos 
 2 2  2 2

 1 2 3 1  2 3 
sin cos cos  cos sin sin
 q1   2 2 2 2 2 2
     
q  cos 1 sin 2 cos 3  sin 1 cos 2 sin 3 
 2   2 2 2 2 2 2
 q3   1 2 3 1  2 3 
   cos cos sin  sin sin cos
2 2 2 2 2 2
 q4    2 3 1  2 3 
 cos 1
cos cos  sin sin sin 
 2 2 2 2 2 2
34
Modified Rodrigues Parameters

It became clear from previous sections that a 3 element attitude parameter

description would be desirable with respect to controllability of the resulting
kinematics relation. Furthermore we found that the quaternion parameters is
constrained to an Euler eigen-axis rotation angle range of 180 ≤  ≤ 180
o o

which constrains the application of continuous attitude tracking. A solution

would be to define a 3 element attitude parameter description in terms of the
quaternion parameters (QPs), an initial solution is defined by the so called
Classical Rodrigues Parameters (CRPs):

Classical Rodrigues Parameters

If the constraint is applied to the 4th scalar quaternion element q4 must

always be defined as:


q4  1  q12  q22  q32  cos
2

35
 Using this fact, the CRPs are defined as follows:

  
 sin
2
 1e

 q1   cos   
   2 e
 1 tan 
 q4     2
 1   sin   

   2  e
q  2   e tan  
      2 2
2 2
q 
 3     cos   
4
 q3  2 
      e3 tan 
 q4   sin  
2
 e3 2
 cos  
 2 

These equations contain the holonomic constraint by definition. But they introduce
a new problem. We clearly see that a singularity occurs when   180o , as the
tangent goes to infinity as q4 approaches zero. To solve this, one can easily modify
the CRPs. This gives us the new parameters referred to as the Modified Rodrigues
Parameters (MRPs).
36
Derivation of the MRPs

The MRPs are defined by adding 1 to the denominator of the definition of the
CRPs. This implies that when q4 approaches zero, there is no singularity.
By doing this, we postpone the singularity to when q4 = −1, which only
occurs at   360o :
qi
i 
1  q4

where i = 1, 2, 3 and using the definition for the quaternions, we get:


qi ei sin
i   2
1  q4 1  cos 
2
From trigonometry:

   
cos   cos 2  sin 2 , sin   2sin cos
2 2 2 2
37
The above equation can be simplified to give:
 
ei 2sin cos
i  4 4
 
1  cos2  sin 2
4 4
 
ei 2sin cos
 4 4
   
sin 2  cos2  cos2  sin 2
4 4 4 4

ei 2sin
 4  e tan 
 i
4
2 cos
4
As such the definition for the MRPs becomes:

 q1   
   e1 tan 
1  q4 
 1   
4

   q2   
   2      e2 tan 
1  q 4
 3   4

 q3    
   e3 tan 4 
1  q4  38
As mentioned, the MRPs experience singularity at   360 . Now one would think,
o

fine we could just modify the Rodrigues parameters again to experience singularity
at   720o and so on. This is possible, defining parameters called Higher-Order
Rodrigues Parameters (HORPs). The problem though is that they reintroduce
singularities in the kinematics rewritten for these HORPs at   360o and the
kinematics are no longer described by simple second-order polynomials further
increasing the computational load. It is therefore decided to stick with the MRPs as
they conveniently only experience singularities when   360o and consists of 3
parameters instead of 4. It should be noted that singularities may be avoided all
together using the MRPs by applying the shadow parameter technique. But let’s
first define the direction cosine matrix in terms of MRPs.

Direction cosine matrix in terms of MRPs

The previously determined direction cosine matrix C was given in quaternions as:

1  2  q22  q32  2  q1q2  q3q4  2  q1q3  q2q4  

 
CB/ A  

 C q, q4   2  q2 q1  q3q4  1  2  q1  q3  2  q2 q3  q1q4  
2 2


 2  q3q1  q2 q4  2  q3q2  q1q4  1  2  q12  q22  
 
39
We will need to redefine the direction cosine matrix in terms of MRPs. Before
we can do so, we must find the relation between quaternions and MRPs. Given
equations in page 37 we define for    1  2  3 T as follows:

q12  q22  q32

        
2 T 2 2 2

1  q4 
1 2 3 2

From the quaternion constraint, solving for q4 :

 
2 1  q42

1  q4 1  q4   1  q4 
1  q4 
2
1  q4 
2
1  q4 
1
2
q4 
1
2

Using the definition of the MRPs, we find:

 1 2  1 2 1 2  2 i

qi   i 1  q4    i  1       
 1  1  1
2 i 2 2


40
Now we may redefine the direction cosine matrix in terms of MRPs by inserting
the relations provided in page 38:

  
 1   2 2  8 2  8 2
2 3 
8 1 2  4 3 1  
2
   
8 1 3  4 2 1  
2

 
C B/ A  C   
1
 8 2 1  4 3 1  2
 1     8  8 8 
2 2 2 2
 4 1    
2

 
2 1 3 2 3 1
1   
2

 8   4 1   2
 3 1 2   8   4 1    1  
3 2 1
2 2 2
  8  8 
2
1
2
2

As an example, the first element of the direction cosine matrix C in terms of MRPs
can be obtained simply with:

 2 2  2  2 
C1,1  1  2  q  q   1  2 
2 2 2
2 
 3
2 

 1     1    
2 3


1     8
2 2 2
2  8 32

1    2 2

41
Kinematic differential
equations

42
Time derivatives of vectors

 b    b 
 1   1

 2 
b    b2
b    b 
 3   3

Since the norm of bi will be fixed to 1, bi will only make a circle due to  .

bi sin  i  
bi  bi sin i 
a3
   bi
m
bi

bi  bi sin i     bi
i
bi
 b sin       b
i i i
a2
bi    bi
a1

43
Transport theorem: motion of a point r, due to a rotation of frame b

r  r1b1  r2b2  r3b3

r  r1b1  r2b2  r3b3  r1b1  r2b2  r3b3
 r1b1  r2b2  r3b3  r1  b1  r2  b2  r3  b3
 r1b1  r2b2  r3b3    r1b1  r2b2  r3b3 

We define now

dr dr
| N  r ; |B  r1b1  r2b2  r3b3
dt dt
We have
dr dr
|N  |B   r
dt dt
44
Kinematic differential equations in terms of direction cosines

Consider two reference frames A and B which are moving relative to each
other. The angular velocity vector of reference frame B with respect to
reference frame A is denoted by    B / A , and it is expressed in terms of
basis vectors of B as follows:

  1b1  2b2  3b3

We have defined the direction cosine matrix C such that

 b1   a1 
  a 
b
 2  C  2
b3   a3 
 
which can also be written as
 a1   b1   b1 
a   C 1 b   C T b 
 2  2  2
 a3  b3  b3 
    45
Taking the time derivative of above relation in A results in:
b 
0   b1   1  b1     b1 
 0   C T b   C T b2   C T    
   b2 
T
   2 b
 2 C
 
0  b3  b3  b3    b3 
     
 
 b1   0 3 2   b1 
    
 C T b2   C T  3 0 1  b2 
b3   2 1 0   b3 
 
where

C11 C12 C13 

 
C  C21 C 22
 
C23 
C 31 C 32 C 33 

46
We obtain
 b1   0 3 2 
 
 
C T  C T  b2   0;    3 0 1 
b3   2 1 0 
 
or

C T  C T   0; C  C  0; T  

This is the kinematic differential equation in terms of direction cosine matrix.

We may write down each element of this matrix differential equation as

C11  3C21  2C31 ; C12  3C22  2C32 ; C13  3C23  2C33

C 21  1C31  3C11 ; C 22  1C32  3C12 ; C   C   C
23 1 33 3 13

C   C   C ;
31 2 11 1 21 C   C   C ;
32 2 12 1 22 C 33  2C13  1C23

47
Kinematic differential equations in terms of Euler angles

Consider the rotation sequence of C1 1   C2  2   C3  3  to B from A, which

can be expressed symbolically as

C3  3  : A'  A
C2  2  : A ''  A '
C1 1  : B  A ''

The time derivatives of Euler angles, called Euler rates, are denoted by
1 ,2 ,3 . These three successive rotations can also be represented as

 A' / A
: A'  A
 A'' / A'
: A''  A'
 B / A''
: B  A ''

48
 and  B / A are aligned with different
''
The angular velocity vectors  , A' / A A'' / A'

basis axes in different reference frames

 A' / A
 3 a3  3 a3'
 A'' / A'
 2 a2'  2 a2''
 B / A''
 1a1''  1b1

The angular velocity vector  B / A becomes

 B/ A
 B / A''
 A'' / A'
 A' / A
 1b1  2 a2''  3a3'
or

1  0 0


 B / A  b1 b2
 

b3  0   a1''  a2''  
a3'' 2   a1' a2' 
a3'  0 
0  0  3 
 
49
 When we write the angular velocity vector 
B/ A
in the B reference frame
1 
 B / A  1b1  2b2  3b3  b1 b2  
b3 2 
3 
We obtain

1  1  0 0

b
1 b2  
b3 2   b1 b2  

b3  0   a1'' a2''  
a3'' 2   a1' a2' 
a3'  0 
3  0  0  3 
 

Left dot multiplying b 

T
1 b2 b3 on both sides of the above equation results in

1  1  0  0  1 0  sin  2  1 

    0   C      C  C    0   0 cos     
 2   1 1  2 1 1 2 2    1 sin 1 cos 2  2
3   0   0  3  0  sin 1 cos1 cos 2  3 
   

50
This matrix equation can be solved by inverting the associated matrix.

1  cos 2 sin 1 sin  2 cos1 sin  2  1 

   0  sin 1 cos 2  2 
1
 
 2  cos  cos1 cos 2
3  2
 0 sin 1 cos1  3 
 

This is the familiar kinematic differential equation in terms of Euler angles.

Note that the inversion has to be performed in the normal way, because the
matrix is not an orthonormal matrix. This can be easily interpreted as the fact
'' '
that b1 , a2 , a3 are not orthogonal between each other.

51
Kinematic differential equations in terms of quaternions

C  C  0;   CC T
 0 3 2   C11 C12 C13   C11 C21 C31 
  
 3 0 1    C21 C22 C23  C12 C22 C32 
 2 1 0  C31 C32 C33  C13 C23 C33 

1   C21C31  C22C32  C23C33 
2   C31C11  C32C12  C33C13 
3   C11C21  C12C22  C13C23 
Substituting the direction cosines matrix elements in terms of quaternions,
(slide 25) the kinematic equations in terms of quaternions can be derived:

1  2q1q4  q 2 q3  q3q2  q 4 q1 
2  2q 2 q4  q3q1  q1q3  q 4 q2 
3  2q3q4  q1q2  q 2 q1  q 4 q3 
52
 q q  q42  q12  q22  q32  q42  1
T
Differentiating the quaternion constraint
0  2q1q1  q 2 q2  q3 q3  q 4 q4 
These four differential equations can be written in the matrix form as

1   q4 q3  q2  q1   q1 
   q q4 q1  q2  q 2 
   2 3
2
3   q2  q1 q4  q3   q3 
    
 
0  q1 q2 q3 q4  q 4 

Since the matrix is orthonormal, we may solve this equation easily as

 q1   q4  q3 q2 q1  1   0 3  2 1   q1 
q   q q4  q1 q2  2  1  3 0 1 2  q2 
   31
2

 q3  2  q2 q1 q4 q3  3  2  2  1 0 3   q3 
       

 4
q   q1  q2  q3 q4   0    1  2  3 0   q4 

A more compact form of the kinematic differential equations in quaternions is

q
1
2
 q4    q  
1
2
 q4  q    ;
1
2
1 T
q4    T q   q 
2 53
An alternative derivation is using the relation between quaternions and
elements of the associated direction cosine matrix

1
q4   C11  C22  C33  1
2
C23  C32 
1 
q C31  C13 
4q4  
 C12  C21 

The time derivatives of the quaternions can be calculated as

1 C11  C22  C33 C  C22  C33

q4     11
4 C11  C22  C33  1 8q4

q3 
C 12 
 C21 q4   C12  C21  q4
4q42

q2 
C 31 
 C13 q4   C31  C13  q4
4q42

q1 
C23 
 C32 q4   C23  C32  q4
4q42 54
Substituting the time derivatives of the direction cosine matrix elements

C11  3C21  2C31 ; C12  3C22  2C32 ; C13  3C23  2C33

C 21  1C31  3C11 ; C 22  1C32  3C12 ; C   C   C
23 1 33 3 13

C   C   C ;
31 2 11 1 21 C   C   C ;
32 2 12 1 22 C 33  2C13  1C23

to the time derivatives of the quaternions, we obtain q4 (positive)

C11  C22  C33

q4 
8q4
3C21  2C31  1C32  3C12  2C13  1C23

8q4
1  C32  C23  1   C13  C31  2   C21  C12  3

2 4q4
1
  q11  q22  q33 
2

55
Similar results can be obtained for other quaternions by using the elements
of direction cosine matrix in terms of quaternions

q3 
C 12 
 C21 q4   C12  C21  q4

C
12  C21 qq
3 4
2
4q4 4q4 q4
1
 3C22  2C32  3C11  1C31  4q3q4 
4q4


1
4q4
 
3 1  2  q12  q32    22  q3q2  q1q4   3 1  2  q22  q32   

1
  21  q3q1  q2 q4   2q3  q11  q22  q33 
4q4 

3 1  q12  q22  q32   2 q1q4  1q2 q4 

1 
 
2q4
1
 3q4  2q1  1q2 
2
1
q2  1q3  2 q4  3q1 
2
1
q1  1q4  2 q3  3q2 
2

The same compact form can be written as

q
1
2
 q4    q    q4  q    ; q4    T q   q 
1
2
1
2
1 T
2
56
Kinematics differential equations in terms of MRPs

The kinematic relations in terms of MRPs may be derived by differentiating the

definitions for the MRPs in page 38 and using the previously determined
kinematic relations for quaternions:

 q1   0 3 2 1   q1 
q    0 1 2   q2 
1
   3
2

 q3  2  2 1 0 3   q3 
    
 4
q  1 2 3 0   q4 

Differentiate the MRPs from the definition

qi
i 
1  q4
we obtain

qi 1  q4   qi q4 qi qi q4 qi q4
i     i
1  q4 
2
1  q4  1  q4  1  q4  1  q4  1  q4 
57
By taking the kinematic differential equations in terms of quaternions in slide 56
and dividing on each side by 1  q4 we are able to solve relations between i and
i :
q4 1 q q q  1
  1 1  2 2  3 3    1 1  2 2  3 3 
1  q4 2  1  q4 1  q4 1  q4  2

Using the result in slide 40

1
2

1 1 1

 
2 2 2
q4 1
q4     1
2
,
1 1  q4 1 1 1
2 2 2 2
2
1
1
2
we obtain

1 q  1
q1 q q
1  q4 1  q4  2 
1
  3 2  2 3  1 4   3 2  2 3  1 1   
1  q4 2  1  q4 2

2 



1 q  1
q2 q q 1 2 
  3 1  1 3  2 4    3 1  1 3  2 1   
1  q4 2  1  q4 1  q4 1  q4  2  2 
 
1 q  1
q3 q q
1  q4 1  q4  2 
1
  2 1  1 2  3 4   2 1  1 2  3 1   
1  q4 2  1  q4 2

2 



58
Now the kinematic differential equations in terms of MRPs in matrix form can
be written as:


 1  1   2
 2  2
2 3  2 1 2 2 2  2 1 3   1 
   1  2  2   
1

  2
      2 3  2 

 2 4  3
2
1 2 1 2 2 2 1 2
 3   2   
 2 2  2  
1 3 2  1  2 
2 3 1   2
 2   3
3 

For simplicity, the kinematic differential equations are often written in the form

1
 N   
4
where

 1   1   2  2 12 2 3  2 1 2 2 2  2 1 3   1 
 
   2  , N     2 3  2 1 2 1    2 22
2
2 1  2 2 3  ,   2 
 3    3 
 2 2  2 1 3 2 1  2 2 3 1    2 32 
2

See Quatview demo on Brightspace for example of kinematic conversions

59
Spacecraft Attitude Dynamics and Control

(AE4313)

Course Notes No. 2

Course Instructor
Dr.ir. E. van Kampen

Section of Control and Simulation

Faculty of Aerospace Engineering

Tel.: 015-2787147
E-mail [email protected]

60