J. Peraire
16.07 Dynamics
Fall 2004
Version 1.1
Lecture D4 - Intrinsic Coordinates
In lecture D2 we introduced the position, velocity and acceleration vectors and referred them to a fixed
cartesian coordinate system. While it is clear that the choice of coordinate system does not affect the final
answer, we shall see that, in practical problems, the choice of a specific system may simplify the calculations
considerably. In previous lectures, all the vectors at all points in the trajectory were expressed in the
same frame. When using intrinsic coordinates (as well as the other coordinate systems presented in the next
lecture), we shall see that the reference frame is a function of the current position of the particle. This means
that, the position, velocity and acceleration vectors at different points are expressed in different frames.
Intrinsic coordinates: Tangential, Normal and Binormal compo-
nents.
From the previous lecture, we know that the velocity vector, v, can be expressed as
v = drdt = vet , (1)
where r(t) is the position vector, v = ˙s is the speed, et is the unit tangent vector to the trajectory, and s is
the path coordinate along the trajectory.
Also, recall that the unit tangent vector can be written as,
et = drds . (2)
The acceleration vector is the derivative of the velocity vector with respect to time, and therefore we can
write, from (1),
a = dvdt = dvdtet + vdetdt . (3)
1
The vector et is the local unit tangent vector to the curve which changes from point to point. Consequently,
the time derivative of et will, in general, be nonzero.
The time derivative of et can be written as,
det
dt =
det
ds
ds
dt =
det
ds v . (4)
In order to calculate the derivative of et, we note that, since the magnitude of et is constant and equal to
one, the only changes that et can have are due to rotation, or swinging.
When we move from s to s + ds, the tangent vector changes from et to et + det. The change in direction
can be related to the angle dβ.
The direction of det, which is perpendicular to et, is called the normal direction. On the other hand, the
magnitude of det will be equal to the length of et (which is one), times dβ. Thus, if en is a unit normal
vector in the direction of det, we can write
det = dβen . (5)
Dividing by ds yields,
det
ds =
dβ
dsen = κen =
1
ρen . (6)
Here, κ = dβ/ds is a a local property of the curve, called the curvature, and ρ = 1/κ is called the radius of
curvature.
Note that in the picture, the sizes of det, ds, and dβ are exaggerated for illustration purposes and actually
represent the changes in the limit as ds (and also dt) approach zero.
Note Curvature and radius of curvature
We consider here two tangent vectors et and e+det, separated by a small ds and having an angle between
them of dβ. If we draw perpendiculars to these two vectors, they will intersect at a point, say, O′. Because
the two lines meeting at O′ are perpendicular to each of the tangent vectors, the angle between them will
be the same as the angle between et and e+det, dβ. The point O′ is called the center of curvature, and the
distance, ρ, between O′ and A is the radius of curvature. Thus, from the sketch, we have that ds = ρdβ, or
dβ/ds = κ = 1/ρ.
2
Intuitively, we can see that each infinitesimal arc, ds, can be represented by a circle segment of radius ρ
having its center at the center of curvature. It is clear that both the radius of curvature and the center of
curvature are functions of s, and consequently they change from point to point.
There are two limiting cases which are of interest. When the trajectory is a circle, the center of curvature
does not change and coincides with the center of the circle, and the radius of curvature is equal to the radius
of the circle. On the other hand, when the trajectory is a straight line, the curvature is zero, and the radius
of curvature is infinite. Note also, that, in this case, the derivative of et is always zero, and the normal
direction is not defined.
Going back to expression (4), we have that
det
dt =
dβ
dsven =
˙βen = v
ρ en . (7)
Finally, we have that the acceleration, can be written as
a = dvdtet + v
2
ρ en = atet + anen . (8)
Here, at = ˙v, is the tangential component of the acceleration, and an = v2/ρ, is the normal component of
the acceleration. Since an is the component of the acceleration pointing towards the center of curvature, it is
sometimes referred to as centripetal acceleration. When at is nonzero, the velocity vector changes magnitude,
3
or stretches. When an is nonzero, the velocity vector changes direction, or swings. The modulus of the total
acceleration can be calculated as a = radicalbiga2t + a2n.
Note Calculation of the radius of curvature for a trajectory
In some situations the trajectory will be known as a curve of the form y = f(x). The radius of curvature in
this case can be computed according to the expression,
ρ = [1 + (dy/dx)
2]3/2
|d2y/dx2| . (9)
This expression is not hard to derive. Try it!
On the other hand, if the trajectory is known in parametric form as a curve of the form r(t), where t can be
time, but also any other parameter, then the radius of curvature can be computed as
ρ = (˙r · ˙r)
3/2
radicalbig(˙r · ˙r)(¨r · ¨r) ?(˙r · ¨r)2 ,
where ˙r = dr/dt, and ¨r = d2r/dt2.
Example
A ball is ejected horizontally from the tube with a speed v0. The only acceleration on the ball is due to
gravity. We want to determine the radius of curvature of the trajectory just after the ball is released.
The simplest way to determine the radius of curvature is to note that, initially, the only nonzero component
of the acceleration will be in the normal direction, i.e. an = g. Thus, from an = v20/ρ, we have that,
ρ = v
20
g .
Alternatively, we can obtain an equation for the trajectory of the form y = f(x) and use expression (9) to
calculate the curvature. The trajectory is given as,
x = v0t
y = ?12gt2 .
Thus, eliminating t, we have
y = ? g2v2
0
x2 .
At x = 0, dy/dx = 0, d2y/dx2 = ?g/v20, and the above expression gives, ρ = v20/g, as expected.
4
Note Relationship between s, v and at
The quantities s, v and at are related in the same manner as the quantities s, v and a for rectilinear motion.
In particular we have that v = ˙s, at = ˙v, and at ds = v dv. This means that if we have a way of knowing at,
we may be able to integrate the tangential component of the motion independently. We will be exploiting
these relations in the future.
The vectors et and en, and their respective coordinates t and n, define two orthogonal directions. The plane
defined by these two directions, is called the osculating plane. This plane changes from point to point, and
can be thought of as the plane that locally contains the trajectory (Note that the tangent is the current
direction of the velocity, and the normal is the direction into which the velocity is changing).
In order to define a right handed set of axes we need to introduce an additional unit vector which is orthogonal
to et and en. This vector is called the binormal, and is defined as eb = et ×en.
At any point in the trajectory, the position vector, the velocity and acceleration can be referred to these
axes. In particular, the velocity and acceleration take very simple forms,
v = vet
a = ˙vet + v
2
ρ en .
The difficulty of working with this reference frame stems from the fact that the orientation of the axis depends
on the trajectory itself. The position vector, r, needs to be found by integrating the relation dr/dt = v as
follows,
r = r0 +
integraldisplay t
0
v dt ,
where r0 = r(0) is given by the initial condition.
We note that, by construction, the component of the acceleration along the binormal is always zero.
When the trajectory is planar, the binormal stays constant (orthogonal to the plane of motion). However,
when the trajectory is a space curve, the binormal changes with s. It can be shown (see note below) that the
derivative of the binormal is always along the direction of the normal. The rate of change of the binormal
5
with s is called the torsion, τ. Thus,
deb
ds = ?τ en or,
deb
dt = ?τven.
We see that whenever the torsion is zero, the trajectory is planar, and whenever the curvature is zero, the
trajectory is linear.
Equations of Motion
Newton’s second law is a vector equation, F = ma, which can now be written in intrinsic coordinates.
In tangent, normal and binormal components, tnb, we write F = Ftet + Fnen and a = atet + anen. We
observe that the positive direction of the normal coordinate is that pointing towards the center of curvature.
Thus, in component form, we have
Ft = mat = m ˙v = m¨s
Fn = man = m v
2
ρ
Note that, by definition, the component of the acceleration along the binormal direction, eb, is always zero,
and consequently the binormal component of the force must also be zero. This may seem surprising, at first,
but recall that the tangent and normal directions are determined by the motion, and, hence, we can say that
the motion “chooses” the binormal direction to be always orthogonal to the applied force. In other words,
if we apply a force to a particle, the particle will experience an acceleration which is parallel to the force.
The normal direction is chosen so that the acceleration vector is always contained in the plane defined by
the tangent and the normal. Thus, the binormal is always orthogonal to the external force.
Intrinsic coordinates are sometimes useful when we are dealing with problems in which the motion is con-
strained, such as a car on a roller coaster. The geometry of the trajectory is known, and, therefore, the
directions of the tangent, normal and binormal vectors are also known. In these cases it may be possible to
integrate the component of the equation of motion along the tangential direction (especially if there is no
friction), and then calculate, a-posteriori, the reaction force using the normal component of the equation of
motion.
Note (optional) Frenet formulae
The Frenet fromulae give us the variations of the unit vectors et, en and eb with respect to the path
coordinate s. The first formula
det
ds =
1
ρen ,
has already been defined. Now, since eb is a unit vector, deb/ds will be orthogonal to eb. Hence, it will be
of the form,
deb
ds = btet + bnen .
6
If we perform the dot product of this expression with et, we obtain
bt = debds ·et = ?detds · eb = ?1ρen ·eb = 0 .
The second equality follows from the fact that the derivative of et · eb is zero, i.e. det · eb + et · deb = 0.
Therefore, only bn is nonzero. Defining bn = ?τ, we obtain the third Frenet formula
deb
ds = ?τen .
Finally, the second formula can be obtained in a similar manner (we leave the details as an exercise) and
gives,
den
ds = ?
1
ρet + τeb ,
or, multiplying by v,
den
dt = ?
v
ρet + τveb . (10)
As we move along s, the osculating plane (and hence eb) may rotate, making the curve non planar. As an
example, an aeroplane may be rolling as it flies along et. The derivative of eb is in the direction opposite to
en if the rotation is in the direction of a right hand screw, and this is taken as the positive direction for the
torsion.
Example “Simplified” Aircraft Kinematics (W. M. Hollister)
The flight of an aircraft through the sky is an example of curvilinear motion. Think of et as the roll axis
aligned with the velocity vector of the aircraft. Think of eb as being the pitch axis. The lift is then directed
along en. The roll rate of the aircraft can be interpreted as τv, and the pitch rate as v/ρ. In order to
turn the aircraft out of the vertical plane, it is necessary to rotate the direction of the lift en so that there
is a component of acceleration out of the vertical plane. Neglecting gravity, the velocity vector along et
determines where the aircraft is going, and the lift along en determines where the velocity vector is going.
The roll rate determines how the lift vector will be rotated out of the osculating plane. As shown by equation
(10), the direction of the lift vector is changed by rolling τv, as well as pitching v/ρ.
Consider the following example. An aircraft follows a spiral path in the sky while doing a barrel roll. The
coordinates are given below, where v0 = 194 ft/s, ω = 0.4 rad/s, and h = 125 ft are constants.
x = v0t
y = hcosωt
z = hsinωt
7
We have,
v = v0i?hωsinωtj + hωcosωtk
v =
radicalBig
v20 + h2ω2 = 200ft/s ≈ v0
a = ?hω2 cosωtj ?hω2 sinωtk
a = hω2 = 20ft/s2
Since ˙v = 0, a = (v2/ρ)en, or,
hω2 = v
2
ρ , ρ =
v2
hω2 = 2000ft
,
et = v0v i? hωv sinωtj + hωv cosωtk
en = ?cosωtj ?sinωtk
eb = hωv i + v0v sinωtj ? v0v cosωtk
Finally,
deb
dt =
v0ω
v cosωtj +
v0ω
v sinωtk
which corresponds to a roll rate of
τv = v0ωv ≈ ω = 0.4rad/s
ADDITIONAL READING
J.L. Meriam and L.G. Kraige, Engineering Mechanics, DYNAMICS, 5th Edition
2/5, 3/5 (normal and tangential coordinates only)
8