05 Instantaneous Axis of Rotation
05 Instantaneous Axis of Rotation
05 Instantaneous Axis of Rotation
Charudatt Kadolkar
Contents
1 Review of Kinematics 1
1 Review of Kinematics
Motion of a rigid body is described using a body fixed coordinate system S 0 . The first three generalized
coordinates (out of 6) are coordinates of the origin of S 0 and the remaining three are the parameters
of the rotation matrix R relating S 0 with space fixed coordinate system S.
The parameters of rotation matrix can be given in many ways
Axis-Angle: The rotation R can be expressed as a rotation by an angle θ about an axis of rotation
n̂ as h i
R = R (n̂, θ) = exp θn̂ · J~
1
Euler Angles: The rotatin R can be expressed as a product of three rotation matrices
R = R (ẑ, φ) R (x̂, θ) R (ẑ, ψ)
and three generalized coordinates are φ, θ and ψ.
Pitch-Roll-Yaw: Same as Euler angles, except the three rotations are about x̂, ŷ and ẑ respectively
(or all other permutations of these)
R = R (x̂, φ) R (ŷ, θ) R (ẑ, ψ)
Note:
We will not worry about the translation in this note. Assume that, one point of a rigid
body is fixed and we will use that as the common origin of the two coordinate systems
(S and S 0 ). In this case, we will establish vector identities which are independent of the
coordinate frames. The components of the vectors wrt to S and S 0 are written as a column
matrix and denoted by [·]s and [·]s0 respectively.
2
Theorem 2. Let P be any point of the rigid body with position vector rp . Let vp be the velocity of P
with respect to S, then at each instant t, thre exists a vector ω(t)
vp = ω × rp
for some vector ω. The vector ω is called as the instantaneous angular velocity and ω̂ is called the
instanteous axis of rotation.
Proof. Let S be space fixed coordinate system and S 0 be body fixed coordinate system. Since rp =
Ro + r0p ,
d d
[vp ]s = rp = R [r]S 0 .
dt dt
= Sw R [r]s0
= Sw [r]s = [ω × r]s
where ω = (− (Sw )23 , (Sw )13 , − (Sw )12 ). Remember that the action of any skew symmetric on a vector
is same as a cross-product.
Theorem 3. The instantaneous angular velocity ω is independent of the choice of body fixed coordinate
system.
Example 4. Consider a wheel lying in xy plane with its center at the origin. The wheel is spinning
about the ẑ with angular speed ω. Thus the transformation matrix R = R (ẑ, θ(t)). Here the angular
speed may not be uniform, but the axis of rotation is constant. Then,
cos θ − sin θ 0 cos θ sin θ 0
d T d
Sω = −R R = − sin θ cos θ 0 − sin θ cos θ 0
dt dt
0 0 1 0 0 1
cos θ − sin θ 0 −θ̇ sin θ θ̇ cos θ 0 0 −1 0
= − sin θ cos θ 0 −θ̇ cos θ −θ̇ sin θ 0 = θ̇ 1 0 0
0 0 1 0 0 0 0 0 0
d T
Sω = −R R
dt
d
= − I + (1 − cos θ) Sn2 + sin θSn I + (1 − cos θ) Sn2 − sin θSn
dt
= − I + (1 − cos θ) Sn + sin θSn sin θ θ̇Sn2 − cos θθ̇Sn
2
= θ̇Sn .
Example 6. Consider motion of a gyroscope (See the figure). This is called the two-frame gyroscope
(as opposed to three-frame). The rotor is the massive disk spinning about CD axis, and CD axis is part
of gimbal which rotates about AB axis. AB axis is fixed in the space. Our interest is in the motion of
the rotor.
3
Let S be space fixed coordinate system. The xy plane of the S frame is horizontal with x̂ along OA
and ẑ in upward direction. And ŷ = ẑ × x̂. The x’y’ plane of the body fixed coordinate system is
same as the plane of the rotor disc with x̂0 along OE and ẑ0 along OC. Let us consider another frame
attached to the gimbal and call it S 00 . The x̂00 = x̂ and ẑ 00 = ẑ0 along OC. Thus the transformation
between S 00 and S 0 is a rotation about common Z axis (that is along OC), that is
and
[r]S = R (x̂, ω1 t) [r]S 00 = R1 [r]S 00
Finally, we get
Sω = Sω1 + R1 Sω2 R1T
Show that
ω = ω 1 + R1 ω 2 = ω1 x̂ + ω2 (cos (ω1 t) ŷ + sin (ω1 t) ẑ) .
This saves the day for us. The net angular velocity is just the sum of ω1 (about axis OA) and R1 ω 2
(Note that R1 ω 1 is exactly along OC, just as expected).
Now in the next three theorems, we will formalize the result of the last example.
Theorem 7. If A and R (n̂, θ) are two rotation matrices then AR (n̂, φ) A−1 = R(An̂, φ).
Proof. First, ARA−1 is a rotation matrix and Tr ARA−1 = Tr (R). This means that the angle of
That is An̂ is eigenvector of ARA−1 with eigenvalue +1. Thus ARA−1 is a rotation by angle φ about
axis An̂.
R (n̂ · J) RT = (Rn̂) · J
4
Theorem 9. If n̂1 and n̂2 are two constant vectors and if the orientation of a rigid body is given by
R = R (n̂1 , θ1 ) R (n̂2 , θ2 ), then, the instantaneous angular velocity is given by
ω = ω1 + ω2
where ω1 and ω2 are velocity vectors corresponding to R1 and R2 respectively, that is,
dθ1 dθ2
ω 1 = n̂1 ; ω 2 = R(n̂1 , θ1 )n̂2 .
dt dt
d
Proof. Let R1 = R (n1 , θ1 ). We have already seen that dt (R1 ) = Sω1 R1 = θ̇1 n̂1 · J R1 . Then
d
(R1 R2 ) = θ̇1 n̂1 · J R1 R2 + R1 θ̇2 n̂2 · J R2
dt
= θ̇1 n̂1 · J + R1 θ̇2 n̂2 · J R1T R1 R2
= θ̇1 n̂1 + θ̇2 (R1 n̂2 ) · JR
The last theorem can be applied to the Euler angles representation: Since
ω = ωφ + ωθ + ωψ
= x̂ sin θ sin φψ̇ + cos φθ̇ + ŷ − sin θ cos φψ̇ + sin φθ̇ + ẑ φ̇ + cos θψ̇ (1)
5
This motivates us to define the quantity I as the moment of inertia tensor. We will represent I in
matrix form such that jk the element is
X
mi ri2 δjk − xij xik
Ijk =
i
and X
I= mi ri2 I3 − [ri ] [ri ]T
i
where [·] means vector written in matrix form. For continous bodies, we can replace the summation
over particles by integration Z
Ijk = ρ r2 δjk − xj xk dv.
Example 11. Consider a uniform cube of side L. It is placed in the first quadrant as shown in the
figure.
Then
3
L3
Z
L 2
y 2 + z 2 dxdydz = ρ · L · = M L2 .
Ixx =ρ ·L+L·
3 3 3
1. The matrix is symmetric and hence diagonalizable with real eigenvalues. The eigenvectors can
always be chosen to be mutually orthonormal.
2. Under the coordinate transformation, [r]s = A [r]s0 , the moment of inertia transforms as
[I]s = A [I]s0 AT .
Since this is a defining property of a second rank tensor, the moment of inertia is also a second
rank tensor.
6
If we know the inertia tensor wrt the space-fixed frame then, we can find the principal axes by diago-
nalizing the MI tensor.
Example 13. Consider the previous example of the cube. The eigenvalues are λ1 = λ2 = 11/12 and
λ3 = 1/6. The corresponding eigenvectors are
1 1 1
1 1 1
√ 1 √ −1 , √ 1
6 −2 2 0 3 1
Now, the great advantage of the principal moment is in the expressions of angular momentum and
angular kinetic energy, that is,
Li = Iii ωi i = x, y, z
and
1X
T = Iii ωi2 .
2
i
Thus the axes of inertial ellipsoid are also principal axes of the body.
Example 14. Repeating the same example of the cube, the inertial ellipsoid is shown in the figure.
Note that the axes of the elliposid are indeed the principal axes.
7
Note that the motion of √ the inertial ellipsoid is always same as the motion of the rigid body. Now,
consider the point ρ = n̂/ In , which
√ is intersection of the instaneous angular velocity ω = ω n̂ and the
inertial ellipsoid. Then ρ = ω/ ω In . Clearly, the kinetic energy of the rigid body at this moment
is
√
1 In ω 2
T = ω·I·ω = ρ·L
2
√ 2
=⇒ ρ · L = 2T
Also,
1
∇ρ (ρ · I · ρ) = 2I · ρ = √ L.
2T
Thus, the angular momentum vector L is always perpendicular to the inertial ellipsoid at ρ!
where
dL
both
dL
angular momentum L and the torque N are computed about the origin. But since,
the
dt s = dt s0 + ω × L,
dL
+ω×L=N (3)
dt s0
And since in body-fixed frame, moment of intertia tensor I is diagonal, we can write the components
of the equations as
I1 ω̇1 − (I2 − I3 ) ω2 ω3 = N1
I2 ω̇2 − (I3 − I1 ) ω3 ω1 = N2
I3 ω̇3 − (I1 − I2 ) ω1 ω2 = N3
These set of equations are called as Euler’s equations. These are not very straightforward equations.
Remember the body-fixed frame is moving with the body, motioin of which we dont know a priori.
Then computing components is impossible. However in some cases it is possible to use these equations
to meaningful use.
8
4.2 Examples
Let us consider few simple cases where the axis of rotation is constant,
Example 15. A cuboid of side lengths w, d, and h is placed with its center at the origin. The body-
fixed frame is at t = 0 is aligned with the space-fixed frame as shown in the figure (φ = θ = ψ = 0).
MI tensor is diagonal
I1 0 0
[I (t = 0)]S = [I (t = 0)]S 0 = 0 I2 0
0 0 I3
The inertial ellipsoid is shown on the right side.
Case 1: Suppose that the motion is given by ψ(t) = ωt, θ(t) = φ(t) = 0. clearly, ψ̇ = ω and
θ̇ = φ̇ = 0. First let us calculate in body-fixed f S frame. From Eq 2, we get ω = ωe3 . The
angular momentum
I1 0 0 0 0
[L]s0 = [I]s0 · [ω]s0 = 0 I2 0 0 = 0
0 0 I3 ω I3 ω
= I3 ωe3 .
[I(t)]S = A(t) [I]S 0 AT (t) = (I1 − I2 ) sin(ωt) cos(ωt) I1 sin2 (t) + I2 cos2 (t) 0
0 0 I3
The blue frame is the space-fixed and the red frame is body fixed. The ω vector is
shown in the green.
case 2: Motion is: ψ(t) = ωt, θ(t) = θ0 , φ(t) = 0. Also, clearly, ψ̇ = ω and θ̇ = φ̇ = 0. Since I and ω
are exactly same as in case 1 above, the description in body fixed frame is exactly same!!
9
What changes is the description in space-fixed frame. Now, (cos θ0 = α and sin θ0 = β)
cos ωt − sin ωt 0
A(t) = α sin ωt α cos ωt −β
β sin ωt β cos ωt α
Then, ω = ω (−β ŷ + αẑ). But L and ω are parallel. But we will check this by writing MI in
fixed frame as
1 1
I1 cos2 (tω) + I2 sin2 (tω)
2 α(I1 − I2 ) sin(2tω) 2 β(I1 − I2 ) sin(2tω)
1 α(I1 − I2 ) sin(2tω) α2 I1 sin2 (tω) + I2 cos2 (tω) + β 2I3 αβ I1 sin2 (tω) + I2 cos2 (tω)
− I32
2
1 2 2 2 2 2
2 β(I1 − I2 ) sin(2tω) αβ I1 sin (tω) + I2 cos (tω) − I3 β I1 sin (tω) + I2 cos (tω) + α I3
The blue frame is the space-fixed and the red frame is body fixed. The ω vector is
shown in the green. L vector is parallel to ω.
Case 3: Motion is: φ(t) = ωt, θ(t) = θ0 , ψ(t) = 0. Here the ω = ωẑ. Let us first do this in body-fixed
frame.
0 0
[ω]S 0 = β [L]S 0 = I2 βω
α I3 αω
The transformation matrix is
cos ωt −α sin ωt β sin ωt
A(t) = sin ωt α cos ωt −β cos ωt
0 β α
Then
(−I2 + I3 ) αβ sin ωt
[L]S = A [L]S 0 = ω (I2 − I3 ) αβ cos ωt
I3 α2 + I2 β 2
Clearly, L precesses about ẑ axis. Then the torque is not zero1 and is given by
(−I2 + I3 ) αβ cos ωt
N = ω 2 (−I2 + I3 ) αβ sin ωt
0
1
This is somewhat counter-intitutive since the constant velocity usually means no acceleration and hence no force
for a particle! But, wait the situation here is more like a system with variable mass, in which case the velocity may be
constant but the time-dependent mass can lead to some non-zero force. In, rigid bodies, total mass may be constant but
moment of inertia may depend on time.
10
The blue frame is the space-fixed and the red frame is body fixed. The ω vector is
shown in the green. Now L vector is shown in magneta color, it is not parallel to ω
but is perpendicular to ellipsoid at ρ point.
Points to note in this example:
ω is constant and along ẑ, but not along any of the principal axes.
L is not constant even though ω is! This is because MI is now changing as a function of
time, remember that the inertial ellipsoid is rotating.
Remember L vector is always perpendicular to the inertial ellipsoid at the ρ point on the
ellipsoid.
Case 4: Motion is: φ(t) = φ̇t, θ(t) = θ0 , ψ(t) = ψ̇t. Let us assume that both ψ̇ and φ̇ are constants.
Now,
ω = ω φ + ω ψ = φ̇ẑ + ψ̇e3 .
Now, because ẑ · e3 = cos θ0 = constant, ω precesses about ẑ axis in space-fixed frame and about
e3 in body-fixed frame. Because of this it clear that polhode (curve traced by ω on ellipsoid)
is an ellipse. The remaining calculation is left to you to calculate. In general, this motion will
require torque. The torque-free case happens only in some special cases, one of which is discussed
in the next secton.
The blue frame is the space-fixed and the red frame is body fixed. The ω vector
is shown in the green and moves along the polhode shown on the ellipsoid. Now
L vector is shown in magneta color, it is not parallel to ω but is perpendicular to
ellipsoid at ρ point.
As the inertial ellipsoid rotates about the ẑ axis, the ω traces out a curve on the ellipsoid (called
polhode). The same ω will appear to be precessing about e3 in body-fixed frame (Imagine, you
are sitting on the inertial ellipsoid).
11
5 Torque free motion
Consider a rigid body with principle axes given by {e1 , e2 , e3 } and correspoinding principle momenta
I1 , I2 and I3 . The Euler equations for the torque free motion are given by
I1 ω̇1 = (I2 − I3 ) ω2 ω3
I2 ω̇2 = (I3 − I1 ) ω3 ω1
I3 ω̇3 = (I1 − I2 ) ω1 ω2
We already have two constants of motion. In inertial frame, the kinetic energyT and angular momentum
~ are constants of motion.
vector L
I1 ω̇1 = (I2 − I3 ) ω2 ω3
I2 ω̇2 = (I3 − I1 ) ω3 ω1
I3 ω̇3 = (I1 − I2 ) ω1 ω2
Since I1 = I2 , ω̇3 = 0 and hence ω3 remains constant in time. Let Ωb = (I3 − I1 ) ω3 /I1 . Then,
ω̇1 = −Ωb ω2
ω̇2 = Ωb ω1 .
Combining the two equations we get ω̈1 = −Ω2b ω1 , solution to which is ω1 = ω0 cos (Ωb t) where ω0 is
a constant. and similarly ω2 = ω0 sin (Ωb t).
Thus, in the body frame, we have complete solution:
ω
~ = (ω0 cos (Ωb t) , ω0 sin (Ωb t) , ω3 )b
~ = (I1 ω0 cos (Ωb t) , I1 ω0 sin (Ωb t) , I3 ω3 )
L b
|~
p
ω | = ω02 + ω32 = constant. ω ~ · e3 = ω3 is also constant. Thus, in body-fixed frame ω~ precesses
about the e3 axis with angular speed Ωb . The cone it traces is called the body cone.
L
~ p
~ · e3 = I3 ω3 . Thus L
~ also precesses about e3
= I12 ω02 + I32 ω32 is also constant and so is L
with same angular speed Ωb .
It is also easy to see that ω ~ ×L ~ · e3 = 0 Then all three vectors are coplanar.
Now, we must transform these vectors to the inertial space frame. The features seen by space frame
are:
L
~ is a constant of motion.
12
Figure 1: Space fixed coordinate axes are arrows. The body fixed coordinate axes have small balls at
the tip. The nodal line (shown in black, which is intersection of xy and e1 − e2 plane) is η̂.
In this frame again, ω ~ · e3 = 0 Then all three vectors are coplanar.
~ ×L
To convert to the space frame, we can use the Euler angles. First chooses the z axis of space frame
~ The instaneous axis of rotation is
along L.
ω
~ = φ̇ẑ + θ̇η̂ + ψ̇ê3
To, write this in body-fixed frame, note that ẑ = cos θê3 + sin θ (sin ψe1 + cos ψe2 ) and η̂ = cos ψe1 −
sin ψe2 . Substituting,
~ = θ̇ cos ψ + φ̇ sin θ sin ψ e1 + −θ̇ sin ψ + φ̇ sin θ cos ψ e2 + ψ̇ + φ̇ cos θ e3 .
ω
Now
13
Appendix
A Transfomation Matrix: Euler Angles
14