Euler Chas Les 4
Euler Chas Les 4
Euler Chas Les 4
Kumar
Euler's Theorem
We have seen that a spherical displacement or a pure rotation is described by a 3×3 rotation
matrix. According to Euler's theorem,
"Any displacement of a rigid body such that a point on the rigid body, say O, remains
fixed, is equivalent to a rotation about a fixed axis through the point O."
We will start with a general spherical displacement and show Euler’s theorem is valid.
Consider a body fixed frame that is displaced from {F} to {M}. The displacement of any point
on the rigid body can be written as:
P = Rp
where p and P are the position vectors before and after displacement respectively, and we are
deleting the superscripts and subscripts associated with FRM. Let us find the eigenvalues and
eigenvectors of R by writing:
Rp = λp
axis of
rotation
u eigenspace spanned
_
by x and x
φ Rv2
φ
v2
v1
Rv1
Figure 1 The real eigenvector, u, and the orthogonal eigenspace for a rotation matrix
If the elements of R are denoted by Rij, the above equation can be written as:
We see that λ = 1 is an eigenvalue of R. In other words, there exists a real vector, u, such that
all points on the line, p = αu remain fixed (invariant) under the transformation R. Thus R is a
rotation that leaves this line or axis fixed.
Define φ so that
cos φ = 1
2
((R11 + R22 + R33 ) − 1) (1)
λ 1 = eiφ , p1 = x , say
λ 2 = e − iφ , p 2 = x (2)
λ 3 = 1, p 3 = u .
-2-
where x is the complex conjugate of x.
Since R is orthogonal, x and x are perpendicular to u and span the plane perpendicular to
the axis of rotation. A real basis for this plane can be constructed using the vectors:
1 i
v1 = (x + x ) , v2 = (x − x ) (3)
2 2
1 i
Rv1 = (λ1x + λ 2 x ) Rv 2 = (λ1x − λ 2 x )
2 2
1
(
= e iφ x + e −iφ x
2
) i
(
= e iφ x − e −iφ x
2
)
= v1 cos φ + v 2 sin φ = − v1 sin φ + v 2 cos φ
As shown in Figure 1, the effect of the transformation, R, is to rotate vectors in the plane
spanned by v1 and v2 through an angle φ about u, while vectors along u are invariant. Thus the
transformation R rotates the rigid body about u through an angle φ. This proves Euler's
theorem.
R = QΛ Q T (4)
where,
cos φ − sin φ 0
Q = v1 v2 u and Λ = sin φ cos φ 0
0 0 1
This is essentially a similarity transformation. If we view the rotation from a new frame, {F’},
whose orientation is given by the coordinate transformation, FRF’ = Q, it is clear that in this
new frame, the displacement is a rotation about the z axis through an angle φ.
1 Note that we could also have chosen to construct our canonical representation of a rotation by viewing it as a
rotation about the x or the y axis.
-3-
4.2. Representations for finite rotations
We have seen that any rotation or spherical displacement can be described by a rotation
matrix. However, because the rotation matrix is orthogonal, its nine elements are not
independent of each other. Specifically, they satisfy six independent equations:
Therefore there are only three independent parameters that completely describe a given rotation
and the rotation matrix is a very redundant representation. In order to develop more compact
representations, we pursue two types of representations. The first relies on decomposing a
given rotation into three finite successive rotations and these three rotation angles, called Euler
angles, completely describe the given rotation. The second is derived from the results of the
previous section and explicitly identifies the axis of rotation and the angle of rotation. The
Euler angles are defined in standard kinematics and dynamics texts and are not discussed here.
Instead, we discuss axis-angle representations in some detail. In this section we define Euler
parameters which provide another compact representation of the axis and angle of rotation. We
also develop a simple result that allow us to obtain the rotation matrix from the axis and angle
of rotation.
Consider the rotation of a rigid body about u through an angle φ as shown in Figure 2.
Consider the triangle QMN in the figure. We can write:
P − p = a1 + a 2
or,
Rp = p + a 1 + a 2
-4-
axis of
rotation
(a)
Q N
φ
M
P-p
p P
(b) (c)
Q
φ
P+p
a2
N
P-p
a1
P-p
M
p P
-5-
From the geometry,
QN = QM = p sin θ = u × p
a 2 = QN sin φ = u × p sin φ
R p = p + a1 + a2
0 − u3 u2
If we let uˆ = u3 0 − u1 , this equation yields:
− u 2 u1 0
This is called the Rodrigues’ formula, a result that should actually be attributed to Euler,
Lexell, and Rodrigues. It describes the rotation matrix in terms of the angle and axis of the
rotation.
To obtain the angle, φ, and axis of rotation, u, from the rotation matrix, we can easily
derive the following formulas:
uˆ =
1
2 sin φ
(
R − RT ) (7)
There are many solutions for the angle of rotation because the inverse cosine function is
multivalued. If we find φ from (6) restricting the range of the inverse cosine function to the
interval [0,π], we can find the axis of rotation from (7) provided φ is not either 0 or π. If R = I,
-6-
φ=0 and the axis of rotation is not well defined. If the trace of the rotation matrix is -1, φ=π,
and from the Euler-Lexell formula:
R = 2uu T − I
For every rotation of angle φ about the axis u, we can also obtain an equivalent axis-angle
representation with a rotation -φ about the axis -u. Also for every solution (u, φ), we have other
solutions (u, φ+2kπ) for all integer values of k.
Euler parameters
We now turn to a representation based on half the rotation angle and the axis of rotation.
Define the four Euler parameters:
φ φ φ φ
c0 = cos , c1 = ux sin , c2 = u y sin , c3 = uz sin (8)
2 2 2 2
Note c0 2 + c12 + c2 2 + c32 = 1 . This is the equation of a unit sphere (or hypersphere) in R4, also
called the projective three sphere. Any point on this sphere corresponds to a unique axis and
angle of rotation and therefore a unique rotation matrix. However, there are two points
(diametrically opposite) on the sphere that give the same axis and angle. c and -c denote the
same rotation. For example, a rotation of 90o degrees about the x-axis yields
c= ( 1
2
, 1
2
, 0, 0 )
T
c= ( −1
2
, −1
2
, 0, 0 )
T
If c is partitioned as follows:
c 0 c 0
c
c = 1 =
c 2 d
c3
-7-
it can be easily seen that Rodrigues formula can be written as:
( )
R = I 2c02 − 1 + 2dd T + 2c0 dˆ (9)
Chasles' Theorem
One of the most fundamental results in spatial kinematics is a theorem that is usually
attributed to Chasles (1830), although Mozzi and Cauchy are credited with earlier results that
are similar. The theorem states:
"The most general rigid body displacement can be produced by a translation along a
line followed (or preceded) by a rotation about that line."
For a planar displacement, one can find a fixed point of the displacement. This is a point
that is left unchanged by the displacement. Consider a general 3×3 homogeneous transfer
matrix:
If R is not the identity matrix, there is one fixed point on the rigid body for any displacement
(in contrast to spatial displacements) called the pole (or the instantaneous center) of the
displacement. If c is the position vector of the pole,
c = Rc+d
c = (I-R)-1 d (10)
-8-
This point corresponds to the eigenvector of the matrix A for a unit eigenvalue.
Thus any planar displacement can be considered as either a pure rotation about a point
called the pole (if (R-I) is not singular) or a pure translation in the direction given by d (if
R=I). Since pure rotations and pure translations are special cases of a screw motion, Chasles
theorem is proved for the planar case.
If we consider a general 4×4 homogeneous transfer matrix, A, it has four eigenvalues two
of which are equal to 1. (The other two are complex conjugate with a magnitude of 1).
However there are no real eigenvectors corresponding to λ=1. This implies that a general
displacement has no fixed points.
To prove Chasles’ theorem for spatial displacements, consider the following similarity
transformation of the matrix A:
Q T − Q T c R d Q c
Λ=
0 1 0 1 0 1
Q T RQ Q T Rc − Q T c + Q T d
=
0 1
Now choose Q as we did for the canonical representation of the rotation matrix in Equation (4)
so that it consists of the eigenvectors of R:
Q = v1 v2 u (11)
The 3×3 orthogonal matrix part of Λ reduces to the rotation matrix corresponding to a rotation
about the z axis:
cos φ − sin φ 0
QR Q = sin φ cos φ 0
T
0 0 1
-9-
( )
Q T Rc − Q T c + Q T d = Q T RQ − I Q T c + Q T d
cos φ − 1 − sin φ 0 c ′x d ′x
= sin φ cos φ − 1 0 c ′y + d ′y
0 0 0 c ′z d ′z
where c'=QTc and d'=QTd. If the top 2×2 submatrix of (QTRQ-I) is nonsingular, we can always
find c' such that
cos φ − 1 − sin φ c ′x d ′x
sin φ = (12)
cos φ − 1 c ′y d ′y
(QTRQ-I) c'=-d'
for the first two components of c' as we did in Equation (10) and let the third component be
zero. In this case, Λ has the form:
cos φ − sin φ 0 0
sin φ cos φ 0 0
Λ= (13)
0 0 1 k
0 0 0 1
where k is given by the third component of the vector QTd. In this case, the displacement can
be described by a rotation about the z axis through an angle φ and a concurrent translation
along the z axis through a distance k. This is a screw displacement and proves Chasles’
theorem for a general displacement.
If the top 2×2 submatrix of (QTRQ-I) is singular, then QTRQ=I. In this special case, Λ is a
pure translation given by the vector d. This is a special case of the screw displacement. Thus
we have proved Chasles’ theorem for this special case as well.
G1 Q c
A E1 =
0 1
-10-
represents the coordinate transformation from {E1} to {G1}. In the reference frame {E1}, the
displacement is a pure rotation about the z axis through an angle φ, acommpanied by a
concurrent translation along the same axis of k=hφ. Thus,
cos φ − sin φ 0 0
sin φ cos φ 0 0
E1
A E2 =Λ= .
0 0 1 k
0 0 0 1
Given the rotation matrix, R, we can find u, a unit vector along the screw axis, and φ, the
angle of rotation about the screw axis from Equations (6-7). In order to completely specify the
screw displacement, we want to find at the position vector of at least one point on the screw
axis and the translation of a point on the rigid body along the screw axis.
dp = d - (d . u) u (15)
Q = v w u
(16)
where,
-11-
{G2}
{E2}
Q2
φ c
P2
hφ
{E1}
{F}
P1
O
{G1} c u
Q1
Figure 3 The displacement of a body fixed frame from {G1} to {G2}, the canonical
representation of this displacement from {E1} to {E2}, and the corresponding screw axis.
dp
v= , w = u×v
dp
and dp is the magnitude of dp or the norm d p . Now Equation (12) simplifies to:
−1
c ′x cos φ − 1 − sin φ
c ′ = −
y sin φ cos φ − 1
−1
cos φ − 1 − sin φ v ⋅d
= −
sin φ cos φ − 1 w ⋅ d
dp (1 − cos φ)
=
2(1 − cos φ) sin φ
Note that we have used the fact d..w = 0. Thus, we get the position vector of a point on the
screw axis:
-12-
1
d p sin φ dp sin φ
c′ = , c = Q c ′=
v + w (17)
2 (1− cos φ ) 2 (1 − cos φ)
0
The displacement of a point on the axis is given by the third component of d':
k=u.d (18)
k
h= (19)
φ
4. References
[1] Bottema, O. and Roth, B., Theoretical Kinematics. Dover Publications, 1990.
[2] Brand, L., Vector and Tensor Analysis, John Wiley, 1947.
[3] Hunt, K.H., Kinematic Geometry of Mechanisms, Clarendon Press, Oxford, 1978.
[4] McCarthy. J.M., Introduction to Theoretical Kinematics, M.I.T. Press, 1990.
[5] Paul, R., Robot Manipulators, Mathematics, Programming and Control, The MIT Press,
Cambridge, 1981.
-13-