3.01.

2019

Outline
 6.1 INTRODUCTION
 6.2 ACCELERATION OF A RIGID BODY
 6.3 MASS DISTRIBUTION
 6.4 NEWTON'S EQUATION, EULER'S EQUATION
 6.5 ITERATIVE NEWTON—EULER DYNAMIC FORMULATION
 6.6 ITERATIVE VS. CLOSED FORM
 6.7 AN EXAMPLE OF CLOSED-FORM DYNAMIC EQUATIONS
 6.8 THE STRUCTURE OF A MANIPULATOR'S DYNAMIC EQUATIONS
 6.9 LAGRANGIAN FORMULATION OF MANIPULATOR DYNAMICS
 6.10 FORMULATING MANIPULATOR DYNAMICS IN CARTESIAN SPACE
 6.11 INCLUSION OF NONRIGID BODY EFFECTS
 6.12 DYNAMIC SIMULATION
 6.13 COMPUTATIONAL CONSIDERATIONS

1
 3.01.2019

Introduction
 In this chapter, we consider the equations of motion for a
manipulator—the way in which motion of the manipulator arises
from torques applied by the actuators or from external forces
applied to the manipulator.
 There are two problems related to the dynamics of a manipulator
that we wish to solve. In the first problem, we are given a trajectory
ሶ
point, 𝜃, 𝜃and 𝜃,ሷ and we wish to find the required vector of joint
torques, τ. This formulation of Dynamics is useful for the problem of
controlling the manipulator.
 The second problem is to calculate how the mechanism will move
under application of a set of joint torques. That is, given a torque
ሶ
vector, τ, calculate the resulting motion of the manipulator, , 𝜃, 𝜃and
ሷ
𝜃. This is useful for simulating the manipulator.

6.2 ACCELERATION OF A RIGID

BODY
 At any instant, the linear and angular velocity vectors have
derivatives that are called the linear and angular accelerations,
respectively. That is,

 As with velocities, when the reference frame of the

differentiation is understood to be some universal reference
frame, {U}, we will use the notation

2
 3.01.2019

Linear acceleration
 We start by restating (5.12), an important result from Chapter
5, which describes the velocity of a vector BQ as seen from
frame {A} when the origins are coincident:

 The left-hand side of this equation describes how AQ is

changing in time. So, because origins are coincident, we could
rewrite (6.5) as

 By differentiating (6.5), we can derive expressions for the

acceleration of BQ as viewed from {A} when the origins of {A} and {B}
coincide:

Linear acceleration

 Now we apply (6.6) twice-—once to the first term, and once to

the last term. The right-hand side of equation (6.7) becomes

 To generalize to the case in which the origins are not coincident, we

add one term which gives the linear acceleration of the origin of {B},
resulting in the final general formula:

3
 3.01.2019

Linear acceleration

 A particular case that is worth pointing out is when BQ is

constant, or

 (6.10) simplifies to

 (6.12) -> revolute joints

 (6.10) -> prismatic joints

Angular acceleration
 Consider the case in which {B} is rotating relative to {A} with
AΩ and {C} is rotating relative to {B} with BΩC . To calculate AΩC
B
we sum the vectors in frame {A}:

 differentiating

 applying (6.6) to the last term of (6.14), we get (6.15)

4
 3.01.2019

6.3 MASS DISTRIBUTION

 For a rigid body that is free to move in three dimensions, there are
infinitely many possible rotation axes. In the case of rotation about
an arbitrary axis, we need a complete way of characterizing the
mass distribution of a rigid body.
 Here, we introduce the inertia tensor, which, for our purposes, can
be thought of as a generalization of the scalar moment of inertia of
an object.

6.3 MASS
DISTRIBUTION
 The inertia tensor relative to frame {A} is expressed in the
matrix form as the 3 x 3 matrix

5
 3.01.2019

6.3 MASS
DISTRIBUTION
 The elements Ixx,, Iyy and Izz are called
the mass moments of inertia.
 Note that, in each case, we are integrating
the mass elements, pdv, times the
squares of the perpendicular distances
from the corresponding axis.
 The elements with mixed indices (Ixy, Ixz,
Iyz) are called the mass products of
inertia.
 The axes of the reference frame when so
aligned are called the principal axes and
the corresponding mass moments are the
principal moments of inertia.

EXAMPLE 6.1
 Find the inertia tensor for the rectangular body of
uniform density p with respect to the coordinate system
shown in Fig. 6.2.

6
 3.01.2019

EXAMPLE 6.1

EXAMPLE 6.1

7
 3.01.2019

EXAMPLE 6.1

parallel-axis theorem
 A well-known result, the parallel-axis theorem, is one way of computing
how the inertia tensor changes under translations of the reference
coordinate system.
 The parallel-axis theorem relates the inertia tensor in a frame with origin at
the center of mass to the inertia tensor with respect to another reference
frame. Where {C} is located at the center of mass of the body, and {A} is an
arbitrarily translated frame, the theorem can be stated [1] as

 where Pc = [xc, yc, zc]T locates the center of mass relative to {A}. The
remaining moments and products of inertia are computed from
permutations of x,y, and z in (6.25). The theorem may be stated in vector—
matrix form as

 where I3 is 3x3 identity matrix

8
 3.01.2019

EXAMPLE 6.2
 Find the inertia tensor for the same solid body described
for Example 6.1 when it is described in a coordinate
system with origin at the body's center of mass.

EXAMPLE 6.2

9
 3.01.2019

Some additional facts about inertia

tensors
 1. If two axes of the reference frame form a plane of symmetry for
the mass distribution of the body, the products of inertia having as
an index the coordinate that is normal to the plane of symmetry wifi
be zero.
 2. Moments of inertia must always be positive. Products of inertia
may have either sign.
 3. The sum of the three moments of inertia is invariant under
orientation changes in the reference frame.
 4. The eigenvalues of an inertia tensor are the principal moments for
the body. The associated eigenvectors are the principal axes.
 Most manipulators have links whose geometry and composition are
somewhat complex, so that the application of (6.17) is difficult in
practice. A pragmatic option is actually to measure rather than to
calculate the moment of inertia of each link by using a measuring
device (e.g., an inertia pendulum :D).

6.4 NEWTON'S EQUATION, EULER'S

EQUATION
 In order to move the links, we must accelerate and
decelerate them.
 The forces required for such motion are a function of the
acceleration desired and of the mass distribution of the
links. Newton’s equation, along with its rotational analog,
Euler's equation, describes how forces, inertias, and
accelerations relate.

10
 3.01.2019

Newton's equation

Euler's equation

 the moment N, which must be acting on the body to cause this motion, is
given by Euler's equation

 CI is the inertia tensor of the body written in a frame, {C}, whose origin is
located at the center of mass.

11
 3.01.2019

6.5 ITERATIVE NEWTON-EULER

DYNAMIC FORMULATION
 We now consider the problem of computing the torques
that correspond to a given trajectory of a manipulator.
We assume we know the position, velocity, and
acceleration of the joints.
 With this knowledge, and with knowledge of the
kinematics and the mass-distribution information of the
robot, we can calculate the joint torques required to
cause this motion. The algorithm presented is based upon
the method published by Luh, Walker, and Paul in [2].

Outward iterations to compute

velocities and accelerations
 In order to compute inertial forces acting on the links, it is
necessary to compute the rotational velocity and linear and
rotational acceleration of the center of mass of each link of the
manipulator at any given instant.
 These computations will be done in an iterative way, starting
with link 1 and moving successively, link by link, outward to
link n.
 The "propagation" of rotational velocity from link to link was
discussed in Chapter 5 and is given (for joint i + 1 rotational) by

12
 3.01.2019

Outward iterations to compute

velocities and accelerations

 From (6.15), we obtain the equation for transforming

angular acceleration from one link to the next:

 When joint i + 1 is prismatic, this simplifies to

Outward iterations to compute velocities

and accelerations

 The linear acceleration of each link-frame origin is obtained by

the application of (6.12):

 For prismatic joint i + 1, (6.34) becomes (from (6.10))

13
 3.01.2019

Outward iterations to compute velocities

and accelerations
 We also will need the linear acceleration of the center of mass
of each link, which also can be found by applying (6.12):

 Here, we imagine a frame, {Ci}, attached to each link, having

its origin located at the center of mass of the link and having
the same orientation as the link frame, {i}.
 Equation (6.36) doesn't involve joint motion at all and so is
valid for joint i + 1, regardless of whether it is revolute or
prismatic.
 Note that the application of the equations to link 1 is
especially simple, because 0𝜔0 = 0𝜔ሶ 0 = 0.

The force and torque acting on a

link

 where {Ci} has its origin at the center of mass of the link
and has the same orientation as the link frame, {i}.

14
 3.01.2019

Inward iterations to compute

forces and torques
 Having computed the forces and torques acting on each link, we now need
to calculate the joint torques that will result in these net forces and
torques being applied to each link.
 We can do this by writing a force-balance and moment-balance equation
based on a free-body diagram of a typical link.

Inward iterations to compute

forces and torques
 Each link has forces and torques exerted on it by its neighbors
and in addition experiences an inertial force and torque. In
Chapter 5, we defined special symbols for the force and torque
exerted by a neighbor link, which we repeat here:

15
 3.01.2019

Inward iterations to compute

forces and torques
 By summing the forces acting on link i, we arrive at the force-
balance relationship:

 By summing torques about the center of mass and setting

them equal to zero, we arrive at the torque-balance equation:

 Using the result from the force-balance relation (6.38) and adding a
few rotation matrices, we can write (6.39) as

Inward iterations to compute

forces and torques
 Finally, we can rearrange the force and torque equations so
that they appear as iterative relationships from higher
numbered neighbor to lower numbered neighbor:

 These equations are evaluated link by link, starting from link ii

and working inward toward the base of the robot. These
inward force iterations are analogous to the static force
iterations introduced in Chapter 5, except that inertial forces
and torques are now considered at each link.

16
 3.01.2019

Inward iterations to compute

forces and torques
 As in the static case, the required joint torques are found
by taking the Z component of the torque applied by one
link on its neighbor:

 For joint i prismatic, we use:

 where we have used the symbol r for a linear actuator force.

Note That

17
 3.01.2019

The iterative Newton—Euler

dynamics algorithm
 The complete algorithm for computing joint torques from
the motion of the joints is composed of two parts.
 First, link velocities and accelerations are iteratively
computed from link 1 out to link n and the Newton—
Euler equations are applied to each link.
 Second, forces and torques of interaction and joint
actuator torques are computed recursively from link n
back to link 1.
 The equations are summarized next for the case of all
joints rotational:

18
 3.01.2019

Inclusion of gravity forces in the

dynamics algorithm
 The effect of gravity loading on the links can be included
quite simpiy by setting 0𝑣ሶ 0 = 𝐺, where G has the
magnitude of the gravity vector but points in the opposite
direction.
 This is equivalent to saying that the base of the robot is
accelerating upward with 1 g acceleration.
 This fictitious upward acceleration causes exactly the
same effect on the links as gravity would. So, with no
extra computational expense, the gravity effect is
calculated.

19
 3.01.2019

6.6 ITERATIVE VS. CLOSED FORM

 Equations (6.46) through (6.53) give a computational scheme
whereby, given the joint positions, velocities, and
accelerations, we can compute the required joint torques.
 As with our development of equations to compute the
Jacobian in Chapter 5, these relations can be used in two ways:
as a numerical computational algorithm, or as an algorithm
used analytically to develop symbolic equations.
 Use of the equations as a numerical computational algorithm
is attractive because the equations apply to any robot. Once
the inertia tensors, link masses, Pc, vectors, and 𝑖+1𝑖𝑅 matrices
are specified for a particular manipulator, the equations can
be applied directly to compute the joint torques
corresponding to any motion.

6.6 ITERATIVE VS. CLOSED FORM

 However, we often are interested in obtaining better insight
into the structure of the equations.
 For example, what is the form of the gravity terms?
 How does the magnitude of the gravity effects compare with
the magnitude of the inertial effects?
 To investigate these and other questions, it is often useful to
write closed form dynamic equations. These equations can be
derived by applying the recursive Newton—Euler equations
ሶ
symbolically to 𝜃, 𝜃and 𝜃.ሷ
 This is analogous to what we did in Chapter 5 to derive the
symbolic form of the Jacobian.

20
 3.01.2019

6.7 AN EXAMPLE OF CLOSED-FORM

DYNAMIC EQUATIONS
 Here we compute the closed-form dynamic equations for the two-
link planar manipulator shown in Fig. 6.6. For simplicity, we assume
that the mass distribution is extremely simple: All mass exists as a
point mass at the distal end of each link. These masses are m1 and
m2

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 First, we determine the values of the various
quantities that wifi appear in the recursive
Newton—Euler equations. The vectors that
locate the center of mass for each link are

 Because of the point-mass assumption, the inertia tensor

written at the center of mass for each link is the zero matrix:

21
 3.01.2019

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 There are no forces acting on the end-
effector, so we have

 The base of the robot is not rotating;

hence, we have

 To include gravity forces, we will use

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 The rotation between successive link
frames is given by

22
 3.01.2019

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 We now apply equations (6.46) through
(6.53).
 The outward iterations for link 1 are as
follows:

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 The outward iterations for link 2 are as
follows:

23
 3.01.2019

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 The inward iterations for link 2 are as
follows:

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS
 The inward iterations for link 1 are as
follows:

24
 3.01.2019

6.7 AN EXAMPLE OF CLOSED-

FORM DYNAMIC EQUATIONS

6.8 THE STRUCTURE OF A MANIPULATOR'S

DYNAMIC EQUATIONS - The state-space equation
 When the Newton—Euler equations are evaluated
symbolically for any manipulator, they yield a dynamic
equation that can be written in the form

 where M(θ) is then x n mass matrix of the manipulator,

V(θ, 𝜃)ሶ is an n x 1 vector of centrifugal and Coriolis terms,
and 𝐺(𝜃) is an n x 1 vector of gravity terms.
 We use the term state-space equation because the term
V(θ, 𝜃)ሶ appearing in (6.59), has both position and velocity
dependence [3].

25
 3.01.2019

EXAMPLE 6.3
 Give M(θ), V(θ, 𝜃)ሶ and G(θ) for the manipulator of
Section 6.7.

EXAMPLE 6.3

26
 3.01.2019

The configuration-space equation

The configuration-space equation

 In this form of the dynamic equations, the complexity of the
computation is seen to be in the form of computing various
parameters which are a function of only the manipulator
position, 0.
 This is important in applications (such as computer control of a
manipulator) in which the dynamic equations must be
updated as the manipulator moves. (Equation (6.63) gives a
form in which parameters are a function of joint position only
and can be updated at a rate related to how fast the
manipulator is changing configuration.)
 We will consider this form again with regard to the problem of
manipulator control in Chapter 10.

27
 3.01.2019

EXAMPLE 6.4
 Give B(θ) and C(θ) (from (6.63)) for the manipulator of
Section 6.7.

EXAMPLE 6.4

28
 3.01.2019

6.10 FORMULATING MANIPULATOR

DYNAMICS IN CARTESIAN SPACE
 Our dynamic equations have been developed in terms of
the position and time derivatives of the manipulator joint
angles, or in joint space, with the general form

 We developed this equation in joint space because we could

use the serial-link nature of the mechanism to advantage in
deriving the equations.
 In this section, we discuss the formulation of the dynamic
equations that relate acceleration of the end-effector
expressed in Cartesian space to Cartesian forces and moments
acting at the end-effector.

The Cartesian state-space equation

 it might be desirable to express the Dynamics of a manipulator
with respect to Cartesian variables in the general form

29
 3.01.2019

 We can derive the relationship between the terms of (6.90)

and those of (6.91) in the following way. First, we premultiply
(6.90) by the inverse of the Jacobian transpose to obtain

 Next, we develop a relationship between joint space and Cartesian

acceleration, starting with the definition of the Jacobian,

 and differentiating to obtain

 Solving (6.96) for joint-space acceleration leads to

 Substituting (6.97) into (6.94), we have

 from which we derive the expressions for the terms in the

Cartesian dynamics as

30
 3.01.2019

EXAMPLE 6.6
 Derive the Cartesian-space form of the dynamics for the two-
link planar arm of Section 6.7. Write the dynamics in terms of
a frame attached to the end of the second link.

31
 3.01.2019

6.12 DYNAMIC SIMULATION

 To simulate the motion of a manipulator, we must make use of a model of
the dynamics such as the one we have just developed. Given the dynamics
written in closed form as in (6.59), simulation requires solving the dynamic
equation for acceleration:

 We can then apply any of several known numerical integration techniques

to integrate the acceleration to compute future positions and velocities.
 Given initial conditions on the motion of the manipulator, usually in the
form

32
 3.01.2019

 we integrate (6.115) forward in time numerically by steps of

size There are many methods of performing numerical
integration [11].
 Here, we introduce the simplest integration scheme, called
Euler integration: Starting with t = 0, iteratively compute

 where, for each iteration, (6.115) is computed to calculate

In this way, the position, velocity, and acceleration of the
manipulator caused by a certain input torque function
can be computed numerically.

END

33