Professor Dr. Hubert Hahn (Auth.) - Rigid Body Dynamics of Mechanisms - 1 Theoretical Basis-Springer-Verlag Berlin Heidelberg (2002) PDF

HubertHahn
Rigid Body Dynamics of Mechanisms 1

Springer-Verlag Berlin Heidelberg GmbH
ONLINE LIBRARY
http://www.springer.de/engine/
Hubert Hahn
Rigid Body Dynamics

of Mechanisms
1 Theoretical Basis
Springer
Professor Dr. Hubert Hahn
Universität Gh Kassel
Regelungstechnik und Systemdynamik, FB Maschinenbau
Mönchebergstraße 7
D-34109 Kassel
Germany
e-mail: hahn@hrz. uni-kassel.de
ISBN 978-3-642-07617-6 ISBN 978-3-662-04831-3 (eBook)

DOI 10.1007/978-3-662-04831-3
library of Congress Cataloging-in-Publication-Data
Hahn, Hubert:
Rigid body dynamics of mechanisms I Hubert Hahn.- Berlin ; Heidelberg ; New York ; Barcelona ; Hong
Kong ; London ; Milan ; Paris ; Tokyo : Springer
I. Theoretical basis. - 2002
ISBN3-540-42373-7
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concer-
ned, specifically the rights of translation, reprinting, reuse of illustrations, recitations, broadcasting, repro-
duction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts
thereof is permitted only under the provisions of the German copyright Law of September 9, 1965, in its cur-
rent version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for
prosecution under the German Copyright Law.
Springer-Verlag Berlin Heidelberg New York
a member of BertelsmannSpringer Science+ Business Media GmbH
http://www.springer.de
© Springer-Verlag Berlin Heidelberg 2002
The use of general descriptive names, registered names trademarks, etc. in this publication does not imply,
even in the absence of a specific Statement, that such names are exempt from the relevant proteelive laws and
regulations and therefore free for general use.
Typesetting: Data delivered by author
Cover Design: de'blik Konzept & Gestaltung, Berlin
Printedon acid free paper SPIN: 10843612 62/3020/M- 54 3 2 1 0
To
Mechthild and Elke
Preface
The dynamics of mechanical rigid-body mechanisms is a highly developed

discipline. The model equations that apply to the tremendous variety of ap-
plications of rigid-body systems in industrial practice are based on just a
few basic laws of, for example, Newton, Euler, or Lagrange. These basic laws
can be written in an extremely compact, symmetrical, and esthetic form,
simple enough to be easily learned and kept in mind by students and engi-
neers, not only from the area of mechanics but also from other disciplines
such as physics, or mathematics, or even control, hydraulics, or electronics.
This latter aspect is of immense practical importance since mechanisms, ma-
chines, robots, and vehicles in modern industrial practice (sometimes called
mechatronic systems) usually include various subsystems from the areas of
hydraulics, electronics, pneumatics, informatics, and control, and are built
by engineers trained in quite different disciplines.
Conventional methods of modeling rigid-body mechanisms
In contrast to the comparatively simple and easy-to-learn basic laws of rigid-

body systems, the practical application of these laws to the planar or spatial
motions of industrial mechanisms rapidly leads to extremely lengthy and
complex equations of motion, where the form and complexity of the model
equations depends critically on the choice of the model coordinates. Until
recently this had the following consequences:
1. A large variety of specialized techniques have been developed, each suit-

able for efficiently modeling a special-purpose mechanism.
2. These techniques have usually been applied to comparatively simple
mechanisms, as most of them were developed at universities or academic
institutes, where there was no need to model complex realistic industrial
systems, and no pressure to do this within a predetermined time schedule.
3. The overwhelming majority of practicing industrial engineers have not
had the opportunity to learn all these special modeling techniques. They
were usually neither capable of finding a special modeling approach suit-
able to a given mechanism, nor of deriving efficiently and correctly the
realistic model equations, nor of estimating in advance the effort required
to derive those models and to set up a time schedule for the task.
Vlll
As a consequence there has been a large gap between the available basic
laws of mechanics and the ability of practicing industrial engineers to apply
them to large rigid-body systems.
General-purpose rigid-body analysis programs as efficient modeling

tools
In the past two decades the above problems have been overcome by worldwide
intensive research activity. As a result, various general-purpose rigid-body
analysis programs have been developed that:
1. A utomatically set up the equations of motion of rather complex kinematic
and dynamic mechanisms.
2. Provide efficient and accurate computer simulations of most of these sys-
tems.
3. Perform the first analysis steps, such as static analysis, kinematic anal-
ysis, locallinearization, eigenvalue analysis, and sensitivity analysis.
Examples of general-purpose rigid-body analysis programs include ADAMS
([1],[2]), DADS ([3]), NUSTAR ([4], [5]), and various other software packages
discussed in ([6], [7]). Teaching computers to automatically formulating the
equations of motion was equivalent to developing systematic generat methods
for setting up and solving model equations of quite generat mechanisms. Using
these computer programs, practicing industrial engineers can simulate and
analyse complex rigid-body systems:
1. By setting up an engineering model of the mechanism based on their
intuitive practical understanding ofthat system.
2. By handling a rigid-body analysis program without the burden of deriv-
ing complex analytical model equations, developing computer simulation
code, and developing numerical solution algorithms of these equations.
Many of these rigid-body analysis programs have been equipped with
graphical user interfaces that can be easily handled even by engineers who
have a limited understanding both of the underlying mechanics and numer-
ics, and of the problems that may occur in the computer-aided modeling and
solution process. However this latter inexperience may have serious conse-
quences: numerical results may be obtained by these programs that are far
more erroneous than any results obtained in laboratory experiments.
Objectives of this monograph
Volume I of this monograph presents:

1. An introduction into the theoretical background ofrigid-body mechanics.
2. A systematic approach for deriving model equations of mechanisms, as a
first step in symbolic differential-algebraic equations (DAE) form.
lX
Volume II presents:
1. Various exercises to systematically apply this approach to examples of
planar and spatial mechanisms.
2. A symbolic approach for mapping the DAEs in a second step into sym-
bolic differential equations (DEs}, into nonlinear and linear state-space
equations, and sometimes also into transfer function form.
The objectives of both the theoretical discussions (Volume I) and the practical
applications (Volume II) are:
1. To prepare the reader for efficiently handling and application of general-
purpose rigid-body analysis programs to complex mechanisms, and
2. To set up symbolic mathematical models of mechanisms in DAE form for
computer simulations and/or in DE form, as is often required in dynamic
analysis and control design.
From the point of view of these two objectives this monograph can be consid-
ered as an introduction to basic mechanical aspects of mechatronic systems.
Organization of the books (Volumes I and II)
The two volumes of this monograph provide a systematic theoretical approach

for setting up model equations of planar and spatial rigid-body systems in DAE
form (Volume I), and present various applications of the modeling methodol-
ogy to examples of planar and spatial mechanisms (Volume 11).
Volume I includes six chapters and Jour appendices. Chapter 1 gives a
brief introduction to the subject of modeling rigid-body mechanisms, which
is illustrated by several simple examples and by some more complex appli-
cations of mechanisms from industrial practice. Chapter 2 presents a brief
review of vector and matrix algebra and of multivariable calculus for the pla-
nar and spatial cases. Spatial rotations are derived in terms of Bryant angles
together with the associated kinematic DEs. Due to the introductory charac-
ter of this book, quaternions or Euler parameters of spatial rotations are not
considered here (despite the fact that singularities may occur in the kinematic
DEs of Bryant angles). Time derivatives of vector functions together with the
gradient vector and the Jacobian matrix of those functions are introduced.
They will be used extensively for describing constraint relations. Some use-
ful relations of scalar products and cross products of vectors are derived in
Appendix A.1, together with different expressions for the time derivatives of
vectors and orientation matrices of planar and spatial vectors, and with a
brief review of derivatives of vector functions. Relations of planar and spatial
kinematic and active constraints, represented in Cartesian coordinates, are
discussed in Chapter 3 together with the associated velocity and accelera-
tion constraint equations, including formal relationships between constraint
reaction forces and torques, and with a discussion of possible singularities
X
of the constraint equations, illustrated by an example. Kinetic equations of

planar and spatial rigid-body mechanisms are developed in Chapter 4 and
in Appendix A.2. Starting with the concepts of linear momentum and angu-
lar momentum in Beetion 4.1, the Newton-Euler equations of the planar and
spatial motion of a single unconstrained rigid body are derived in Beetion 4.2,
together with the model equations of planar and spatial mechanisms in Bee-
tion 4.3. Abrief discussion of the numerical solution of DAEs is presented in
Beetion 4.4. Parallel to the Newton-Euler approach, the Lagrange formalism
is briefly discussed in Appendix A.2. Basic differences between the theoretical
constituents of planar and spatial mechanisms are collected in Appendix A.3.
In Chapter 5 a systematic approach for deriving the constraint equations of
planar and spatial joints is presented based on suitable representations and
projections of vector and orientation loop equations. The constraint equa-
tions of various joint types in common use are derived there. Theoretical
models of joints of planar mechanisms are presented in Beetion 5.1. Model
equations of joints of spatial mechanisms are derived in Beetion 5.2 and in
Appendix A.4. Constitutive relations of applied forces and torques of planar
and spatial mechanisms are discussed in Chapter 6. Among those, theoretical
models of translational and torsional springs and dampers as well as models
of actuators and motors are briefly presented.
Various simple and some more complex applications of rigid-body mech-
anisms are modeled in symbolic DAE form and in DE form, and for se-
lected mechanisms also in nonlinear and linear state-spaee form and using
the transfer funetion matrix representation in Volume II. They include var-
ious combinations of theoretical models of joints, and of active and passive
force elements. In Chapter 1 of Volume II, the modeling methodology is sum-
marized, and a software package is briefly discussed ([8]) that maps symbolic
model equations from DAE form into DE form (in most cases where this is
feasible). Two applications of planar models of an unconstrained rigid body
are discussed in Chapter 2. Several applications of a planar rigid body un-
der constrained motion are presented in Chapter 3. Various applications of
planar mechanisms that include two rigid bodies under constraints are dis-
cussed in Chapter 4. Applications of a rigid body under unconstrained spatial
motion are collected in Chapter 5, followed by several applications of a con-
strained spatial rigid body in Chapter 6, and by several applications of spatial
mechanisms including between two and thirteen constrained rigid bodies in
Chapter 7.
U se of the text
The text of the books is intended for use and self-study by practicing indus-
trial engineers that have a bachelor's degree, and by students of undergradu-
ate university courses. The contents of the books have been used in lectures
and courses held over many years:
XI
1. In several industrial companies (like BMW and IABG) for practicing

engineers from the areas of mechanics, vibration techniques, vehicle sim-
ulation, control, hydraulics, pneumatics, measurement, testing, electro-
magnetics, and electronics.
2. In the undergraduate courses of several universities (Universities of Mu-
nich, Tübingen, and Kassel) for students from the areas of mechanical
engineering, control engineering, electrical engineering, civil engineering,
physics, and mathematics.
The practicing engineers who attended these courses have influenced both
the contents and the direction of this monograph, resulting in more emphasis
being placed on:
1. A systematic choice of notation (with indices of the variables that
uniquely identify the frames of their representations and time deriva-
tives).
2. An algebraic formulation of all expressions in a form suitable for direct
implementation in a computer.
3. Applying these methods to both simple and complex mechanisms.
The engineers and students that attended these lectures had the opportunity
to apply these methods to practical examples of mechanisms using general-
purpose rigid-body analysis programs like NUSTAR, ADAMS, and DADS.
Spatial mechanics is conceptually more complex and its theoretical mod-
eling provides much lengthier and more unwieldy formal expressions than
planar mechanics. To enable the beginner reader to successfully master his
or her study of rigid-body dynamics and to keep the amount of notation and
formal expressions of the applications presented within acceptable limits, only
planar rigid-body systems are considered in the first parts of Chapters 2, 3,
5 and 6 of Volume I. They present vectors, matrices, kinematics, forces and
torques of planar geometry and planar mechanics. The equations of motion
of rigid bodies under planar motion are collected in Chapter 4 of Volume I.
Various planar mechanisms are discussed in Chapters 2, 3 and 4 of Volume II.
Teaching experience shows that the methodology of modeling rigid-body sys-
tems can be basically understood by considering planar systems only. Having
developed confidence and enough intuition in the basic methods of theoreti-
cal modeling of planar mechanisms, the reader is encouraged to study spatial
mechanisms in the second parts of Chapters 2, 3, 5, 6 and in alt of Chapter
4 of Volume I, and the applications of spatial mechanisms of Chapters 5, 6,
and 7 of Volume II. Basic differences between the model equations of planar
and spatial mechanisms are summarized in Appendix A.3 of Volume I.
Acknowledgements
The author thanks Dr. Roger A. Wehage (TACOM, Warren, USA) for many
stimulating discussions on rigid-body dynamics during common development
Xll
work of the rigid-body analysis program NUSTAR at IABG, and Dipl.-Ing

Wolfgang Raasch (IABG, Ottobrunn, Germany) for various useful discus-
sions on setting up realistic and efficient engineering models of industrial
mechanisms and vehicles. The author is further indebted to Dipl.-lng. Willy
Klier for several useful discussions and to Dipl.-Ing. Axel Dürrbaum and Mr
Ralf Rettberg for preparing the many illustrations and diagrams. Last but
not least, the author thanks Mrs Michaela Görgl for her patience in typing
the lengthy mathematical relations and the manuscript, and Dipl.-lng. Axel
Dürrbaum for preparing, handling, and correcting the process of creating the
lbT'gX document.
Hubert Hahn
Sporke/Westfalen
Germany
April 2001
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vu
1. Introduction.............................................. 1
1.1 Tasks in multibody simulation, analysis, and control . . . . . . . . 1
1.2 Coordinates and frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Formulation of the model equations . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Prototype applications of rigid-body mechanisms . . . . . . . . . . . 7
1.5 General-purpose rigid-body analysis programs. . . . . . . . . . . . . . 21
1.5.1 Design of an engineering model . . . . . . . . . . . . . . . . . . . . 22
1.5.2 Input and output data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6 Purpose of this monograph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2. Planar and spatial vectors, matrices, and vector functions 33

2.1 Planar vectors and matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.1.1 Elementary vector and matrix operations. . . . . . . . . . . . 34
2 .1.1.1 Geometrie vectors . . . . . . . . . . . . . . . . . . . . . . . . 34
2.1.1.2 Algebraic vectors . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.1.2 Time derivatives of displacement vectors and Orienta-
tion matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1.2.1 Velocities and angular velocities . . . . . . . . . . . . 48
2.1.2.2 Accelerations and angular accelerations.. . . . . 50
2.2 Spatial vectors and matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2.1 Displacement vectors, frames, and orientation matrices 54
2.2.1.1 Basistransformation . . . . . . . . . . . . . . . . . . . . . . 56
2.2.1.2 Coordinate transformation . . . . . . . . . . . . . . . . . 59
2.2.1.3 Bryant angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.2.2 Time derivatives of displacement vectors and orienta-
tion matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.2.2.1 Velocities and angular velocities . . . . . . . . . . . . 65
2.2.2.2 Accelerations and angular accelerations . . . . . . 67
2.2.2.3 Kinematic differential equation . . . . . . . . . . . . . 67
XIV Contents
3. Constraint equations and constraint reaction forces ofmecha-

nisms..................................................... 75
3.1 Kinematics of planar and spatial rigid-body systems . . . . . . . . 75
3.1.1 Kinematics of planar mechanisms . . . . . . . . . . . . . . . . . . 75
3.1.1.1 Pure kinematic analysis of planar mechanisms 79
3.1.1.2 Regular and singular planar kinematics . . . . . . 81
3.1.1.2.1 Regular constraint Jacobian matrix . 81
3.1.1.2.2 Singular constraint Jacobian matrix. 82
3.1.1.3 Kinematics in planar dynamic analysis . . . . . . 83
3.1.2 Kinematics of spatial mechanisms . . . . . . . . . . . . . . . . . . 84
3.1.2.1 Pure kinematic analysis of spatial mechanisms 84
3.1.2.2 Kinematics in spatial dynamic analysis . . . . . . 86
3.1.3 Singularity analysis of a planar slider-crank mechanism 87
3.1.3.1 Identification of singularities by direct inspec-
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.1.3.2 Local algebraic singularity analysis ofthe slider-
crank mechanism . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.1.3.2.1 Local analysis of Case 1
(~/Ji(t) =a 1 (t)) ................... 91
3.1.3.2.2 Local analysis of Case 2
(x~ 0 = -a2(t)) .................. 103
3.2 Constraint reaction forces and torques of mechanisms ....... 120
3.2.1 Constraint reaction forces of planar mechanisms ...... 120
3.2.2 Constraint reaction forces of spatial mechanisms ...... 123
4. Dynamics of planar and spatial rigid-body systems ....... 129

4.1 Linear momentum and angular momentum of a rigid body ... 129
4.1.1 Linear momentum ................................ 129
4.1.2 Angular momentum .............................. 131
4.1.3 Properties of the inertia matrix .................... 134
4.1.3.1 Physical interpretation of Jj; ............... 134
4.1.3.2 Time dependence of Jj; and JfJ ............ 135
4.1.3.3 Steiner-Huygens relation .................. 135
4.2 Newton-Euler equations of an unconstrained rigid body ..... 137
4.2.1 Forcemomentsand couples ........................ 137
4.2.2 Newton's law .................................... 140
4.2.3 Euler's law ...................................... 141
4.2.4 Newton-Euler equations of a rigid body under planar
and spatial motion ............................... 143
4.2.4.1 Spatial motion ........................... 143
4.2.4.2 Planar motion ........................... 147
4.3 Equations of motion of planar and spatial rigid-body mecha-
nisms ................................................. 150
Contents xv
4.3.1 Equations of planar motion of unconstrained rigid

bodies in DE form and of constrained rigid-body sys-
tems in DAE form ................................ 151
4.3.1.1 A single unconstrained rigid body ........... 152
4.3.1.2 System of unconstrained rigid bodies ........ 154
4.3.1.3 A single rigid body constrained with respect
to the base ............................... 154
4.3.1.4 System of constrained rigid bodies .......... 156
4.3.2 Equations of spatial motion of unconstrained rigid
bodies in DE form and of constrained rigid-body mech-
anisms in DAE form .............................. 158
4.3.2.1 A single unconstrained rigid body ........... 158
4.3.2.2 System of unconstrained rigid bodies ........ 159
4.3.2.3 A single rigid body constrained with respect
to the base ............................... 159
4.3.2.4 System of constrained rigid bodies .......... 161
4.4 Numerical solution of DAEs- abrief discussion ............ 162
4.4.1 Ideal situation ................................... 163
4.4.1.1 Algebraic aspects ......................... 163
4.4.1.2 Numerical integration step ................. 165
4.4.2 More realistic situations ........................... 166
4.4.2.1 Singularmatrix A ........................ 166
4.4.2.2 Constraint violation ....................... 166
5. Model equations of planar and spatial joints .............. 171

5.1 Theoretical modeling of planar joints ..................... 173
5.1.1 Absolute constraints .............................. 174
5.1.1.1 Position constraints between a body and the
base ..................................... 174
5.1.1.1.1 Partial-position constraint (massless
revolute-translationallink) ........ 174
5.1.1.1.2 Complete-position constraint (revo-
lute joint) ....................... 179
5.1.1.2 Grientation constraint (massless translational
link) .................................... 181
5.1.1.3 Grientation and partial-position constraint (trans-
lational joint) ............................ 181
5.1.1.4 Combined orientationfpartial-position
constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.1.1.5 Constant-distance constraint (massless revolute-
revolute link) ............................ 184
5.1.2 Relative planar joints between two bodies ........... 186
5.1.2.1 Position constraints ....................... 186
5.1.2.1.1 Partial-position constraint (massless
revolute-translationallink) ........ 186
xv1 Contents
5.1.2.1.2 Complete-position constraint (revo-

lute joint) ....................... 190
5.1.2.2 Grientation constraint (massless translational
link) .................................... 192
5.1.2.3 Relative orientation and partial-position con-
straint (translational joint) ................. 193
5.1.2.4 Combined orientationfpartial-position
constraint ................................ 196
5.1.2.5 Constant-distance constraint (massless revolute-
revolute link) ............................. 196
5.1.3 Pseudo-joint and forcejtorque elements ............. 198
5.1.3.1 Example of a translational spring element .... 198
5.1.3.2 Example of a torsional spring .............. 198
5.2 Theoretical modeling of spatial joints ..................... 200
5.2.1 Building blocks of joint models ..................... 200
5.2.1.1 Common-point constraint (BB1; three con-
strained translational DOFs) ............... 201
5.2.1.2 ParaUel-axes constraint (BB2; two constrained
rotational DOFs) ......................... 204
5.2.1.3 Straight-line-point-follower constraint(BB3; two
constrained translational DOFs) ............ 208
5.2.1.4 Rotation-blocker constraint (BB4; one con-
strained rotational DOF) .................. 212
5.2.1.5 Constant-distance constraint (BB5; one con-
strained translational DOF) ................ 218
5.2.2 Theoretical models of common joints ................ 220
5.2.2.1 Spherical joint (BB1; constrains three trans-
lational DOFs) ........................... 220
5.2.2.2 Massless spherical-spherical link (BB5; con-
strains one translational DOF) ............. 222
5.2.2.3 Translationaljoint (BB2, BB4; constrains three
rotational DOFs) ......................... 223
5.2.2.4 Universal joint (BB1, BB4; constrains three
translational and one rotational DOF) ....... 226
5.2.2.5 Revolute joint (BB1, BB2; constrains three
translational and two rotational DOFs) ...... 228
5.2.2.6 Cylindrical joint (BB2, BB3; constrains two
translational and two rotational DOFs) ...... 231
5.2.2.7 Prismatic joint (BB2, BB3, BB4; constrains
three rotational and two translational DOFs). 234
6. Constitutive relations of planar and spatial external forces

and torques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.1 Constitutive relations of planar external forces and torques .. 239
6.1.1 Gravitational force (weight) ....................... 241
Contents xvu
6.1.2 Applied force and moment ......................... 241

6.1.3 Translational force elements between two bodies ...... 243
6.1.3.1 Translational spring ....................... 246
6.1.3.2 Translational damper ..................... 247
6.1.3.3 Actuator ................................ 250
6.1.3.4 Torsional spring and damper ............... 250
6.1.3.5 Torque generated by a motor ............... 250
6.2 Constitutive relations of spatial external forces and torques .. 251
A. Appendix ................................................. 255

A.1 Special vector and matrix Operations used in mechanics ..... 255
A.l.1 Euclidean vector space ............................ 255
A.l.2 Scalar product and cross product of planar vectors ... 258
A.l.3 Cross product of spatial vectors .................... 262
A.l.4 Time derivatives of planar orientation matrices and of
planar vectors in different frames . . . . . . . . . . . . . . . . . . . 266
A.l.5 Time derivatives of spatial orientation matrices and of
spatial vectors in different frames . . . . . . . . . . . . . . . . . . . 273
A.l.6 Derivatives of vector functions ..................... 282
A.2 Lagrange formalism of a rigid body under spatial motion .... 290
A.2.1 Kinetic energy of an unconstrained rigid body ....... 291
A.2.2 Spatial equations of motion of an unconstrained rigid
body for P = C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
A.2.3 Spatial equations of motion of a constrained rigid
body ............................................ 296
A.3 Model equations of planar and spatial mechanisms ......... 298
A.4 Constraint equations of a general universal joint ............ 302
A.4.1 Notation and abbreviations ........................ 303
A.4.2 Computation of constraint equations ................ 305
A.4.2.1 First constraint equation ................... 305
A.4.2.2 Second constraint equation ................. 309
A.4.2.3 Third constraint equation .................. 313
A.4.2.4 Fourth constraint equation ................. 316
A.4.3 Computation of the shortest distance between two ro-
tation axes ...................................... 319
References ................................................. 321
Index ...................................................... 329
List of figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
1. lntrod uction
The mechanical systems discussed in this book (e.g., Figure 1.1) are collec-
tions of rigid-badies connected by translational and torsional spring, damper
and friction elements, and by joints, links, bearings and gear boxes, in which
some or all ofthe bodies can move relative to each other. They may be driven
by external forces or torques to achieve specified performance requirements as
well as desired loading and operation conditions. They are called rigid-body
systems or mechanisms. A rigid body is defined as an assembly of particles
that do not move relative to each other. This means in reality that "deforma-
tions of rigid bodies" have no significant influence on the gross body motion.
Rigid bodies of mechanisms move relative to each other consistent with the
joints that limit their relative motion. Simultaneaus large displacements and
rotations of those bodies lead to nonlinear model equations with geometric
nonlinearities that in most cases must be solved numerically.
1.1 Tasks in multibody simulation, analysis, and control
Multibody systems are commonly investigated under different aspects, de-

pending on the task to be solved. In kinematic analysis, the motion of a
system (positions, orientations, velocities, and accelerations) is considered
without taking into account forces that cause this motion. Usually time his-
tories of some position coordinates (independent variables) of the rigid bodies
are prescribed, and time histories of the remainder position, velocity, and ac-
celeration coordinates (dependent variables) are determined by solving non-
linear algebraic equations for the position, and linear algebraic equations for
the velocities and accelerations. In kinematic analysis, the number of degrees
of freedom of the mechanism must be equal to the number of independent
driver constraint equations. All required model parameters are assumed to
be known in kinematic analysis.
In kinetic ( dynamic) analysis and computer simulation, the motion (de-
pendent variables) of a system is determined from given time histories of
forces and torques (independent variables) applied to the system, by solv-
ing a set of nonlinear differential equations (DEs) or differential-algebraic
equations (DAEs). In dynamic analysis the number of unknown variables in
the constraint equations is larger than the number of constraint equations.
Therefore a unique solution is only obtained by specifying a proper set of
initial conditions. The model parameters are assumed to be known here, too.
2 1. Introduction
spherical joint
Fig. 1.1: Examples of multibody systems

1.2 Coordinates and frames 3
In inverse kinetic ( dynamic) analysis the time histories of the minimal

coordinates of the mechanism are prescribed. Solving the nonlinear algebraic
constraint equations of the kinematics provides the time histories of the po-
sition, velocity, and acceleration of the remainder coordinates. Solving the
dynamic equations as nonlinear algebraic equations with respect to the forces
and torques provides their time histories, associated with the prescribed mo-
tions. Again the model parameters are assumed to be known.
In parameter identification, the time histories of the motion (position,
velocity, acceleration) of each rigid body together with the time histories of
the associated forces and torques are measured. The kinematic and dynamic
model equations are solved as nonlinear algebraic equations with respect to
the unknown model parameters, taking into account measurement errors.
In control synthesis, desired motions (position, velocity and/or acceler-
ation) of selected bodies are chosen. Assuming that the dynamic and kine-
matic model equations of the system together with the model parameters are
exactly or approximately known, dynamic or static control algorithms are
computed by specific design techniques that tend to minimize deviations of
the actual motions from the desired motions, and simultaneously guarantee
the stability and sometimes certain robustness properties of the closed-loop
system.
Each of the preceding investigations and tasks is based on analytical and
numerical models of the dynamics of the mechanism considered, where rigid
body dynamics includes kinematics and kinetics.
1.2 Coordinates and frames
Model equations (equations of motion) of rigid-body systems may be for-

mulated in quite different (moving and/or inertial) frames and coordinates.
As a general result, depending on the coordinates chosen, some model equa-
tions of a system will be more involved than others. Any set of variables
(coordinates) that uniquely specifies the position and orientation of all bod-
ies in a mechanism, that is, the configuration of the mechanism, is referred
to as set of generalized coordinates 1 (p) (coordinates in general, regardless of
their nature). Generalized coordinates may be independent (each free to vary
arbitrarily) or dependent (required to satisfy constraint equations).
Independent generalized coordinates are called minimal coordinates. The
minimal number of independent coordinates required to specify uniquely and
completely the position and orientation of each "component part" of a rigid-
body mechanism is called set of degrees of freedom (DOFs) of the system,
where the term "component part" used in this context refers to any part ofthe
systemsuch as a platform, wheel, motor, disk, or lever, which must be treated
1 This definition of generalized coordinates in rigid-body mechanisms differs from
the traditional definition of generalized coordinates in mechanics.
4 1. Introduction
as a rigid body. Rigid-body configurations may be specified by introducing

an inertial (global, absolute) frame Rand body-fixed (local, relative) reference
frames Lij, with j as index of the local frame and i as index of the body (if
only a single local frame is defined on body i, it will be called Li)· Then each
rigid body may be located by specifying global (absolute, inertial) coordinates
of the position of the origin of a frame Li, and its orientation with respect
to a global frame R (Figure 1.2). The relative location and orientation of a
frame Li fixed on a body i with respect to a frame Lj, fixed on a body j, is
specified by local (relative) coordinates.
global body i
reference
frame R
Fig. 1.2: Global (inertial) and local (relative) vectors and frames
1.3 Formulation of the model equations
A lot of research work into rigid-body dynamics has been devoted to the
selection of system coordinates and DOFs that provide a trade-off between
the generality and efficiency of dynamic formulation and simulation. The
modeling methods of multibody systems may in general be devided into two
main approaches:
In the first approach , a minimum number of relative (local) or joint Coor-
dinates are used to formulate a minimum number of DEs that are expressed
in terms of the system DOFs. In many applications, this approach Ieads to a
complex recursive formulation based on loop closure equations. The incorpo-
ration of general forcing functions, constraint equations (e.g. model equations
of joints, compare Chapter 5) and/or specified trajectories in the recursive
1.3 Formulation of the model equations 5
formulation is difficult. This approach, however, may be desirable in several

applications (e.g. serial robots).
In a second approach, the configuration of the system is identified by
using a set of Cartesian (global) coordinates that describe the location and
orientation of the bodies in the mechanism. This approach leads to model
equations in DAE form. It has the advantage that the dynamic formulation
of the equations that govern the motion of the system is straight forward, and
that it allows easy and flexible addition and removal of rigid bodies, of complex
force functions, and constraint equations. For each spatial rigid body in the
system, six coordinates are sufficient to describe the body configuration.
The second approach to set up model equations in DAE form and absolute
coordinates from the Newton-Euler equations will primarily be used in this
monograph, where:
1. The position and orientation of a rigid body under planar motion will be
specified by planar Cartesian coordinates p := (x, y, 'ljJ)T.
2. The position and orientation of a rigid body under spatial motionwill be
specified by spatial Cartesian Coordinates p := (x, y, z, r.p, e, '1/J)T.
Formulation of model equations and system constraints in global coor-

dinates is flexible with practically no limitation on the type of multibody
system. This easier modeling is obtained at the expense of a larger system of
model equations (maximum number of coordinates). Moreover, various anal-
ysis and control design techniques, developed for systems in minimal Coor-
dinates and in state-space form, are not yet available or may become much
more complicated for system equations written in DAE form ([9] and [10]).
As a consequence, the model equations of the applications of Volume II
will be mapped from symbolic DAE form to symbolic DE form by suit-
able elimination or projection techniques. It should be mentioned that this
symbolic elimination process is in general restricted to mechanisms that are
not too complex, or to mechanisms with a particular structure (like a tree
structure).
Newton's and Euler's laws, together with the concept of virtual work, may
be regarded as a foundation on which all considerations of rigid-body mechan-
ics rest. However, it should be realized that the basic laws of mechanics can
be formulated (mathematically written) in several ways other than that given
by Newton, such as D 'Alembert 's principle, Lagrange 's equations, H amilton's
principle and Hamilton's equations, all of which are basically equivalent to
Newtons's laws and the principle of virtual work.
The "basic laws" of dynamics are merely statements of a wide range of
experience. They cannot be obtained by logic or mathematical manipulations
alone but are founded on careful experimentation. We cannot "explain" why
these laws are valid. We can only say that they represent a compact statement
of past experience regarding the behavior of a wide variety of mechanical
systems.
6 1. Introduction
Here the equations of kinetics (dynamics) of rigid bodies will be mainly

derived from Newton's and Euler's basic laws, and sometimes also from La-
grange's equations (cf. Appendix A.2). Newton's and Euler's basic laws of
unconstrained motions of a rigid body are:
1. "Force (F) equals the product ofthe mass (m) times the acceleration (v)
of a rigid body" (Newton's second axiom};
d
m·v=F or dt (P) = F' (1.1)
with
P=m·v as the linear momentum of the rigid body.
2. "Torque (M) equals the product of the moment of inertia (J) times the
angular acceleration (w) of a rigid body" (Euler's law);
d
J ·W = M or dt(D) = M' (1.2)
with
D=J·w as the angular momentum of the rigid body.
3. "Reaction forces (Fij, Fji) (or torques Mij, Mji) between two bodies
i and j are equal in magnitude and opposite in direction" (Newton's third
axiom};
(1.3)
The above equations (1.1) and (1.2) are only valid in this simple form if
the time derivatives of P, v, and D are measured relative to an "inertial
frame", and if all vectors are represented in "inertial coordinates".
Choosing local (noninertial) coordinates of the accelerations, velocities,
and force and torque vectors may provide quite complex representations of
Newton's and Euler's equations, as will be shown in Chapter 4. As a conse-
quence, the treatment of every theoretical problern in rigid-body mechanics
begins with a consideration of an inertial frame. The concept of "inertial
frame" is of fundamental theoretical and practical importance though it is
only a hypothesis that is never really satisfied in applications; it is a fictious
concept introduced for formal convenience. Due to rotations and other mo-
tions of the earth, a coordinate frame attached to its surface is obviously
noninertial. Nevertheless, the acceleration of this frame is so "low" that for
most technical purposes it may be regarded as inertial.
Lagrange 's equations of the second type of an unconstrained rigid body
are (cf. Appendix A.2)
~
dt
(f)L) + 8L = Q
av op (1.4)
1.4 Prototype applications of rigid-body mechanisms 7
with L as the Lagrange function, p = (rT, TJT)T as vector of a minimum

set of generalized coordinates, v = (rT, wT)T as velocity vector, and Q as
vector of generalized forces, associated top and v.
1.4 Prototype applications of rigid-body mechanisms

Rigid-body systems may range from very simple to very complex mecha-
nisms. Simple rigid-body mechanisms (Figures 1.3 and 1.4) are traditionally
modeled by writing down the Newton-Euler equations by direct inspection
of associated free-body diagrams or by using the Lagrange equations after
having defined the Lagrange function. This modeling approach can still be
applied by engineers with some experience in this field to slightly more com-
plex systems such as those of Figure 1.5. The complexity of a model of a
technical system depends on its purpose. A model for vibration analysis of
the steering mechanism of Figure 1.6 may be obtained by direct application
of the Lagrange equations (compare Figure 1.6d and [11], [12], [13]). Another
model for simultaneously studying both, the spatial kinematics and dynam-
ics of the steering mechanism of Figure 1.6a to 1.6c, may be already quite
complex when it includes large spatial motions of the wheels, the steering
gear, and the steering wheel.
Mathematical models of kinematic and dynamic systems with several
DOFs have traditionally been modeled in terms of "clever formulations"
that take advantage of specific properties of the system considered to obtain
simplified forms of model equations. Ingenious selection of independent posi-
tion and orientation coordinates occasionally may lead to a formulation that
allows manual derivation of the equations of motion. This "clever formulation
approach" is nevertheless limited to relative simple rigid-body systems and
can only be performed by specialists that have quite a deal of experience in
this field.
More complex rigid-body systems like, for example:
1. the press model of Figure 1.7 that includes three rigid-bodies subject
to large spatial motion (ram, pitman, and eccentric drive), additional16
rigid bodies that model small spatial deflections of the frame, and several
revolute and universaljoints as well as various springs and dampers ([14],
[15], [16]); or
2. the model of a loaded roller rig of Figures 1.8, and 1.9, 1.10, 1.11 and 1.12
that includes more than 60 rigid bodies (most subject to small spatial
motion), connected by various revolute, prismatic, and universal joints,
and driven by several servo-hydraulic actuators (compare the technical
drawing of the roller stand of Figure 1.10), and that includes models of
the rolling contact of elastic bodies in the presence of dry friction; ([17],
[18], [19]);
8 1. Introduction
(a) Systems with one translational degree of freedom
(b) Systems with one rotational degree of freedom
(c) Discrete model of a string (d) Coupled pendulum
Fig. 1.3: Examples of simple unconstrained and constrained mechanical systems

body 2
body 1
J
r elast ic rope in lastic rope
shaft 1
)
171
I I
~~- - -- -- -- -- -- -- -- -
.. I I
)
shaft 2
damp r
--,
rubber
~--=~element
'-T-----.----' --,
Fig. 1.4: Examples of simple mechanisms including joints

10 1. Introduction
flyball governor
(spind le driven by engine) steam valve
(a) Watt's steam engirre governor (b) Schematic drawing of a fly ball
governor
bogic 1
(c) Planar modelsofasimple locomotive and of a planartest facility
(d) Planar models of an inverted pendulum and of an excavator
Fig. 1.5: Slightly more complex mechanisms

......
;...
'"0
.....
0
0
'""'"
'<:
'""'"
"0
ct>
(a) Front view
"0
"'
'2.
frontal wheel (r) ('; '
~
c;·
1:1
[/)
0
....,
.....
CJ'q'
~
o"
0
~
s
ct>
&
§
tij'
(c) Horizontal projection (d) Engineeringmodel for vibration analysis s
[/)
Fig. 1.6: Steering mechanism of an automobile

......
......
......
"'
revol ute joint ~

......
eccentric drive ~
spherical joint "'00...
>=
(")
pitman '"'"
::;·
;::
spherical joint
.....
-·· spatial dampers
. . ~heri cal joints
photograph schematic drawing eng m eermg model
Fig. 1.7: Photograph, schematic drawing, and engineering model of a single-point-drive eccentric press (LVWU laboratory, University
of Kassel)
(a) Photograph of the roller rig
(b) Technical drawing of the roller rig
Fig. 1.8: Roller rig ofthe German railway company (DB AG)
14 1. Introduction
(a) Roller rig loaded by an ICE high-speed locomotive
(b) Engineering model of a roller stand loaded by a wheel set
Fig. 1.9: Roller rig of the German railway company (DB AG)
(a) Roller rig with bogie (technical drawing)
(b) Bed plate and roller stands (technical drawing)
Fig. 1.10: Roller rig and bogie

16 1. Introduction
(a) Roller rig with bogie (technical drawing)
(b) Rollerstand (engineering model)
Fig. 1.11 : Roller rig and bogie

(horizontal projection)
(front and side views)

Fig. 1.12: Drawing of elastic and dissipative coupling of a bogie
18 1. Introduction
may be modeled by special-purpose rigid-body programs, obtained for exam-

ple by symbolic computation, where the computer is used to differentiate
the Lagrange function of a mechanism according to the Lagrange formalism,
to substitute variables, and to perform algebraic manipulation. The chance
(probability) of deriving correct model equations of the roller rig (that cover
more than 300 pages) by hand without using symbolic computation tends
towards zero, even for engineers that are well trained in providing "clever
formulations" .
A special purpose simulation program deals with only a single type of
applications. Such a program can be well adapted to the particular structure
of the application, taking into account, for example, specific different motions
of subsets of rigid bodies (like spatial, planar, or single axis motions) and
typical kinematic behavior. Such a tailor-made program for a specific single
application can be made computationally quite efficient. The major drawback
of such a special-purpose program is its lack of flexibility for handling other
types of applications.
Practicing engineers must usually be capable of theoretically modeling
and simulating complex mechanisms of quite different types in a short time:
1. Like serial robots with various degrees offreedom that include many rigid
bodies subject to large spatial motion, various joints and actuators (e.g.
Figure 1.13a, [20], [21], [22], [23], [24]).
2. Like parallel robots
2.1 constructed as multi-axis test facilities including up to 17 rigid bodies
subject to large spatial motion together with 8 universal, 8 spherical,
and 8 prismatic joints (e.g. Figure 1.13b, [25], [26], [27], [28], [29],
[30], [31]); or
2.2 constructed as hexapods including from 1 to 13 rigid bodies subject
to large spatial motion, with 6 universal, 6 spherical, and 6 prismatic
joints (e.g. Figure 1.13c, [32], [33], [34]).
3. Like off-road vehicles including various rigid bodies subject to large spa-
tial motion and a large number of dissipative and elastic connection el-
ements as well as revolute and universal joints. Compare the following
two examples:
3.1 The truck of Figure 1.14 that has been modeled by the general pur-
pose rigid-body analysis program NUSTAR. The model includes 17
rigid bodies subject to large spatial motion, 8 universal joints, 5 rev-
olute joints, 4 tire models, an engine model and more than 32 spa-
tial spring and damper elements. More than 96 spatial frames were
needed to specify the geometry of this vehicle ([35], [36], [37]).
3.2 The tank model of Figure 1.15 that has been constructed from a
large number of rigid bodies subject to large spatial motion and from
a large number of elements connecting these bodies ([38], [39]). It
has been modeled and simulated by the general-purpose rigid-body
analysis program NUSTAR.
lner ia measurement robot devel-

op ed by RTS , Univer. ity of Ka el Serial robot imulated by USTAR
(a) Serial robots
(b) Parallel robots (left-hand side robot built as multi-axis test facility
by RTS, University of Kassel)
(c) Parallel robots built as hexapods

Fig. 1.13: Drawings and photographs of serial and parallel robots
0
"'
~
......
~
"'00...
>=
(")
'"'"
::;·
;::
(c)
Fig. 1.14: Photograph (a) , technical drawing (b) , engineering model (c) , and animation graphics (d) of a truck obtained using the
program NUSTAR at IABG , Ottobrunn
1.5 General-purpose rigid-body analysis programs 21
Fig. 1.15: Animation graphics of the german tank Leopard II, obtained using the
program NUSTAR at IABG , Ottobrunn
Programs that can handle a large variety of different complex rigid-body

mechanisms are called general-purpose rigid-body analysis programs.
1.5 General-purpose rigid-body analysis programs
Rather than relying on "clever formulations" and on special-purpose pro-

grams for simulating mechanisms, general-purpose rigid-body analysis pro-
grams have been developed in the past two decades to automatically set up
model equations of those systems (usually in numerical form) , and solve them
numerically ([1], [3], [4], [5], [6], [7], [40], [41]). While the number and type
of elements and the kinematics of the above applications may differ signif-
icantly, the modeling concept remains the same. General purpose programs
22 1. Introduction
are based on methods for systematically deriving the model equations. As

there are mechanisms for which dependent coordinates cannot be eliminated
symbolically (e.g. for special classes ofmechanisms with loop structure), sym-
bolic computer algebra computations are usually replaced by pure numerical
computations. Due to the efficient and accurate numerical algorithms used
in these programs, they are capable of providing reliable numerical solutions
of the model equations. Due to user-friendly graphical interfaces and efficient
pre- and postprocessors, many of these programs enable users with compar-
atively poor knowledge of theoretical mechanics to rapidely model, simulate
and analyze complex rigid-body systems numerically. As a result, general-
purpose rigid-body analysis programs are very flexible. They can be easily
handled and applied to a large variety of different industrial mechanisms.
A general purpose rigid-body analysis program performs four basic tasks:
1. Accepts the model and control data from the user (preprocessor, input
phase).
2. Generates the model equations (usually in numerical form as DAEs).
3. Solves the model equations.
4. Delivers the desired results to the user (postprocessor, output phase).
Steps 2 and 3 are essentially performed by the program. The user of a general
purpose analysis program has to:
1. Set up an engineering model of the mechanism taking into account the
purpose of the model (pre-input phase).
2. Choose, collect and enter control data of the program, and data of the
engineering model (input phase).
3. Select desired output data (output phase}.
4. Judge their quality (post-output phase}.
1.5.1 Design of an engineering model
An engineering model is a schematic drawing of a mechanism that includes all

components and data needed to fulfill its purpose. Therefore an engineering
model of a mechanism is very dependent on the purpose of the model. Depend-
ing on the intended application of the model, simplifying assumptions of the
mechanism are made to reduce the expenditure of modeling and simulation
effort. Consider, for example, the vehicle of Figure 1.14. If investigations focus
on the driving behavior and driving stability of the vehicle, high-frequency
vibrational motions of the vehicle components play only a minor role; they
will be excluded from the engineering model. Instead, sophisticated tire mod-
els will be included in this engineering model. As a consequence the design
of a suitable engineering model is of crucial importance for the efficient and
successful application of the model. The user has to decide: (1) which com-
ponents of the mechanism must be included in the model, and (2) which
type of component models and characteristics must be chosen, in order to
1.5 General-purpose rigid-body analysis programs 23
efficiently achieve the objectives of the intended use of the model. Choosing
inadequate component models prevents the user from achieving satisfactory
model validation results even when the model parameters are carefully iden-
tified in laboratory experiments. Figure 1.16a shows a comparison between
field experiments and computer simulation results obtained by a truck model
that includes a simple engirre model ([38]). Various model parameter iden-
tification and model validation experiments (that took months of intensive
work) could not provide a satisfactory agreement between the computer Sim-
ulations and the field experiments. Replacing the simplified engine model of
the truck by a more sophisticated engirre model , and introducing a simple
driver model,provided in a single step excellent agreement between the time
histories of all simulated and measured variables of the system (Figure 1.16b).
1.5 """'!"- -
_:~I' ~ FS}f!b
-~: @UJ~A:Jts
..-~l-1....,.....-..,......
2
0.5 ~"!--!
-0. 5 ......_...._
1~ ~~C:Q;;~~;i!:..~
-10 ......_ _ _ ~ .......~---
'~ P ;
-10
; --;;;&]
:t !
:~t~~:
10~
F:==;::E§~S
~:=*::;:;;.;:;:k:=~;
-1~E: ~ ~i
11(5:: ~
:
=~7
- ::: 6::: 7r - ·~ .
6 . . . ;
L,
2 3 4 5 7 8
. 3
sec
'
0 1 2 3 4 5 6 7 8 sec
1. P itch angular veloci y 4. Vert ical mo ion of rear wheels
2. Vertical acceleration 5. Actual vehicle velocity
3. Vertical motion of front wheels 6. Obstad e profile (a ra mp)
Fig. 1.16: Measured and simulated vertical transient motions of a truck crossing an
obstacle (a ramp) : without sophisticated engirre and driver modules (a)
and including sophisticated engirre and driver modules (b)
The success in setting up an engineering model depends crucially on the

practical experience of the user, on his intuitive understanding of the compo-
nents and properties of the mechanism, and on his understanding of possible
24 1. Introduction
critical situations that may occur in the practical behavior of a mechanism

as well as, to a certain extent, in the modeling and numerical solution pro-
cess. In detail, the design of an engineering model includes the following steps
(compare the truck model of Figure 1.14):
1. Simplification of the system by isolating elements or components of pri-
mary importance (for the purpose of the model) and construction of a
schematic drawing of the mechanism. The actual shape or outline of a
body may not be of immediate concern to the modeling process.
2. Choice of rigid-badies as models of components with significant inertial
properties and collection of inertial parameters such as the mass, the Co-
ordinates of the center of mass, and the moments and products of inertia.
3. Choice of an inertial frame R of the mechanism, of a local reference frame
Li for each rigid body i, fixed on that body, and of various other local
(body-jixed) frames Lij used to identify attachment points, locations, and
orientations of connection elements between the bodies, and of forces,
torques, actuators, and sensing elements.
4. Choice of the joint types of the connection elements between the bodies.
5. Choice of the spring, damper, and friction elements together with the
associated characteristics.
6. Choice of the types of the external forcesjtorques acting on the bodies
together with the forcejtorque characteristics, lines of action of forces,
and rotation axes of torques.
7. Refinement of the diagram or network of the engineeringmodelthat in-
cludes all components (e.g. rigid-bodies, connection elements and drivers),
all frames, all displacement vectors, rotation angles, all force and torque
vectors, all lines of action of forces, and all rotation axes of torques of
the mechanism.
1.5.2 Input and output data
Based on the engineering model (network) of a chosen mechanism, all phys-

ical and geometric model data needed to setup the simulation program are
collected and entered into the model data file. They include:
1. Data that control the size and complexity of the mechanism such as the
number (nb) of rigid-bodies, and the number (nc) and type of joints.
2. Data of the reference points and local frames on the bodies with respect
to a chosen inertial frame.
3. Data of the inertia parameters of the rigid bodies such as the mass (m),
the center of mass (rpc), the moments and products of inertia (Jijc).
4. Data that specify the connectivity of the mechanism such as the attach-
ment points of the joints, springs, dampers, actuators and the orientation
of the external forces and torques.
5. Data about the characteristics of the springs, dampers, tires, etc.
1.6 Purpose of this monograph 25
In addition to these model data, control data are entered that select and
control desired options of the intended simulation runs and analysis steps.
Other control data select desired output data and control their representation.
The collection and input of model and control data of theoretical models
of complex mechanisms is a cumberful and tedious task that must be pre-
pared very carefully in order to efficiently achieve correct simulation results,
represented in a form that can be easily interpreted and efficiently judged.
1.6 Purpose of this monograph
General purpose rigid-body analysis programs are widely and successfully

used in industry to automatically set up model equations of quite complex
rigid-body mechanisms in numerical form, solve these equations, and hence
simulate these mechanisms on a computer. As well as usually not being as
computationally efficient as special-purpose programs, general-purpose rigid-
body analysis programs may have two further drawbacks:
1. Due to the user-friendly interface of these programs, users with minor
knowledge and understanding of the underlying laws of mechanics and
of the problems and fallacies that may occur in the modeling and so-
lution process can apply these programs to complex applications, with
the consequence that they may provide simulation results that are much
more erroneous than any results measured in laboratory experiments.
This is an increasingly observed phenomenon in industry (similar obser-
vations are made for laboratory experiments based on digital measure-
ment equipment, performed by engineers with minor knowledge of digital
signal analysis).
2. These programs usually do not provide model equations of rigid-body
mechanisms in symbolic form that are often needed in the design of var-
ious nonlinear control algorithms, nonlinear prefilter algorithms, distur-
bance compensation algorithms, signal- or image-processing algorithms,
and in algorithms used as model hypotheses in model parameter identifi-
cation.
This monograph is addressed to all engineers that whish to model, sim-
ulate, control, or/and experimentally identify rigid-body mechanisms. These
engineers may be from the areas of mechanics, robotics, and mechatronics,
as well as from other areas such as control, electronics, hydraulics, and signal
processing, or even from disciplines such as physics, informatics, and applied
mathematics.
Volume I presents:
1. An introduction into the foundations of rigid-body mechanics of mecha-
nisms.
26 1. Introduction
2. A systematic approach for deriving, as a first step, symbolic model equa-

tions of mechanisms in DAE form.
Volume II presents:
1. Various exercises to systematically apply this modeling approach to ex-
amples of planar and spatial mechanisms.
2. A systematic approach for mapping the DAEs in a second step into sym-
bolic DEs, into nonlinear and linear state-space equations, and sometimes
also into transfer function form.
The objectives of both the theoretical discussions (Volume I), and the
practical applications (Volume II) are (Table 1.1) to overcome some of the
above mentioned drawbacks of general-purpose rigid-body analysis programs,
by:
1. Preparing the reader for efficiently handling and application of general-
purpose computer programs to complex mechanisms:
1.1 To obtain a deeper understanding of the basic mechanical relations
behind the software packages.
1.2 To set up adequate engineering models of mechanisms and to choose
suitable component models, coordinates and frames.
1.3 To become moresensitive and confident with the possibilities, restric-
tions, and fallacies when applying rigid-body programs to practical
applications.
1.4 To find adequate interpretations of the simulation results obtained
by these programs.
1.5 To gain enough intuitive understanding for reasonably and critically
judging and evaluating the modeling approach.
2. Systematically deriving analytical mathematical models of mechanisms in
DAE andjor DE form, as is often needed in:
2.1 The design of sophisticated linear and nonlinear control, disturbance
compensation, and signal-processing algorithms.
2.2 The theoretical analysis of model equations like symbolic lineariza-
tion, eigenvalue analysis, and stability analysis, sensitivity analysis
or frequency response analysis.
2.3 Experimental identification of model parameters and model valida-
tion of mechanisms.
Various simple and more advanced examples of planar and spatial mech-
anisms of Figures 1.17, 1.18, 1.19, and 1.20 will be theoretically modeled in
detail and discussed in Volume II.
Applications
The objectives of these books are to provide: Control and
Computer
Analysis identi-
simulations fication
background understanding of the theoretical

foundations of rigid body mechanisms X X X
the ability of setting up engineering models

that serve certain purposes (choice of X X X
component models, coordinates, and frames)
the ability of systematically deriving model

equations of mechanisms in symbolic X X
DAEform
sufficient sensitivity with respect to

possible singular situations and fallacies X X X
in the model equations
some routine in mapping and in simplifying model

equations from DAE to DE form (linearization,
state-space and frequency response representations) X X
for special applications and purposes
Table 1.1: Objectives of this monograph (Volumes I and II)
28 1. Introduction
~ ~ b:a ~
~ ~ ~
[1~ ~
~1.!_11:/:L
~&~~
....... ~ (Q
Fig. 1.17: Planar mechanisms including a singlerigid body
0
""
,~ ~
......
~ ~
"'00...
>=
("")
® ::;·
"""
;::
Fig. 1.19: Mechanisms including a singlerigid body subject to !arge spatial motion
~~ ~Ü{)
C%C% 00-o \J
......
0'>
'"0
E:;
"0
0
[/)
C1l
0
....,
tf ;:r
"""
tij '
s
0
1:1
0
~
.....
"0
"'
;:r
Fig. 1.20: Mechanisms including several rigid hoclies subject to !arge spatial motion C;:i
......
2. Planar and spatial vectors, matrices, and
vector functions
A treatise of vector algebra and vector analysis is a standard ingredient in ba-

sie engineering education, and the reader will undoubtedly have some knowl-
edge of this topic. However, in this chapter, we shall provide a refresher on
some basic concepts that will be instrumental for the further development in
this book.
Vectors are basic entities of analytic geometry. They are defined by their
geometric properties (invariants). These so-called geometric vectors are enti-
ties in their own right without referring to any special basis of a vector space.
They allow compact formulations of physical laws and theoretical relation-
ships. In particular, geometric vector notations provide an adequate tool for
efficiently formulating the kinematics and dynamics of rigid-body systems.
Alternative representations of vectors (with respect to a basis of an underlying
vector space), referred to as algebraic vectors, are better suited to engineer-
ing applications of rigid-body dynamics including formula manipulation and
computer implementation. Differential calculus of vector functions provides
an approved tool for formulating and analysing kinematic relations. The de-
velopment in this chapter is for planar (Section 2.1) and spatial (Section 2.2)
vectors. Spatial rotations will be expressed in terms of Bryant angles. They
may include singular situations that might be avoided by Euler parameter
or quaternion formulations of rotations. Due to the introductory character of
this book, quaternion formulation will not be discussed here. Multivariable
calculus is written in a form that will be directly used in the formulation of
constraint equations in Beetion 3. Some elementary results from Euclidean
vector spaces, elementary vector algebra, and geometry, together with a dis-
cussion of time derivatives of vectors, orientation matrices, vector functions,
and some results from multivariable calculus are collected in Appendix A.1.
2.1 Planar vectors and matrices
In this section vector representations, Operations and transformations in the

plane will be briefly discussed together with their time derivatives. Vectors
and matrices will be written in bold faced letters.
34 2. Planar and spatial vectors, matrices, and vector functions
2.1.1 Elementary vector and matrix operations
In this section planar geometric and algebraic vectors will be considered to-
gether with elementary vector operations and mappings of vectors by means
of orientation matrices.
2.1.1.1 Geometrie vectors. A vector (displacement, velocity, acceleration,
angular velocity, angular acceleration, force, torque, linear momentum, or an-
gular momentum) is an entity in its own right. Its basic properties (invariants)
are independent of any reference frame and special coordinate representation.
Those properties of vectors are theoretically studied and analysed in math-
ematics (analytical geometry), theoretical physics and analytical mechanics
without referring to special coordinates. These vectors are sometimes called
geometric vectors.
Consider the geometric displacement vector r PO in Figure 2.1, with start
point 0 and end point P. It is defined as the directed straight line from 0 to
P and is represented by an arrow pointing from 0 to P. This vector has the
length Ir PO I and a direction with respect to another vector TQO (from point
0 to point Q), described by an angle '1/JQP, measured from Tpo to TQO· The
sum of two vectors r PO and TQP is defined as the vector TQo from point 0
to point Q, written as (Figure 2.1)
TQO = TQP + T PO . (2.1)
The scalar product (or dot product) of two vectors Tpo and TQO is defined
as product of the magnitudes of the vectors times the cosine of the angle
between them,
Tpo • TQo := lrpol·lrQol· cos'l/JQP E ffi.1 , (2.2a)
where lrQol · cos'l/JQP is the projection of TQo onto Tpo (Figure 2.1). For
r PO =j:. 0 and TQo =j:. 0 the scalar product is only zero if cos '1/JQP = 0. Two
vectors are said to be orthogonal to each other if their scalar product is zero;
i.e.,
Tpo • TQo = 0 (orthogonal vectors Tpo and TQo) . (2.2b)
Since '1/JQP = 2n- '1/JQP, the order of the factors of a scalar product is imma-
terial. For each vector r PO
Tpo • Tpo = lrpol 2 E ffi.1 (square of the length of Tpo) , (2.3)
due to cos'lj!pp = cosO = 1. The vector product (or cross product} of two
vectors Tpo and TQo is defined as the vector (Figure 2.1)
Tc:= Tpo x TQO = (lrpol·lrQol· sin'l/JQP) ·ec, (2.4)

E ffi.l
2.1 Planar vectors and matrices 35
TQo=TQp+r~p . ."AQP
0 rpo
Geometrie displacement vectors r PO Geometrie sum of the vectors r PO and

and TQo TQP
Projection of TQo onto r PO Projection of TQo onto a vector per-

pendieular to r PO
Tc= TpQ X TQO
rpo
Plane spanned by rpo and TQo
Vector product of the vectors r PO and TQo
Fig. 2.1: Geometrie vector, vector length, vector sum, projections, and vector pro-
duct
where ec is a unit vector (vector of length 1) that is orthogonal to the plane

spanned by r PO and TQo, taken in the positive right-hand direction. Since
reversal of the order of vectors Tpo and TQO in (2.4) yields an opposite
direction of ec,
TQP X TpQ = -TpQ X TQP· (2.5)
Comment 2.1.1 (Cross product of planar vectors): The definition of

the cross product of two vectors implies for the planar case (1~_2) that the
vector generated by this cross product is orthogonal to this plane (1~_2). As
a consequence the plane is not closed under cross product operations. This
implies that cross product Operations of vectors in OC2 can only be formulated
in the space OC3 , considered as an extension of OC2 • This will be clone subse-
quently in order to describe planar rotations by expressions that are formal

identical to spatial rotations.
Comment 2.1.2 (Cross product vector): The above vector Tc, intro-
duced as the result of the cross product of the displacement vectors r PO and
TQo, is no longer a displacement vector. It has no start point and no target
point, but only a direction and a length (compare Comment 2.1.3 and the
notion of a moment or torque vector, defined later).
Comment 2.1.3 (Different geometric vectors in mechanics): In rigid-

body dynamics, different types of geometric vectors occur:
1. A displacement vector (TQP) defined by a fixed start point and by a fixed
end point (Figure 2.2a).
Q end point
start point ~
(a) Displacement vector TQP with fixed start point and fixed end point
F amplitude
F= F·ep e F unit vector
ep
line of action unit vector
F on the line of
action of
the force
(b) Force vector F as an element of a set of vectors with common line of
action and equal length
M=M·eM
M amplitude
or ~
M e M unit vector
(c) Moment {torque) vector M as an element of a set ofvectors ofidentical

orientation, direction, and length; these vectors are not restricted to a
common line of action
Fig. 2.2: Differenttypes of geometric vectors in rigid-body dynamics

201 Planar vectors and matrices 37
20 A force vector (F = F 0ep) as an element of a set of geometric vectors on

a common line of action, with equallength IFI and with the amplitude F,
where Fis counted positive if the force acts in the direction of the arrow
F, and negative otherwise (Figure 202b) (vectors (F, M) are drawn by
hold faced arrows) 0
30 A moment (torque) vector (M = M 0 eM) as an element of a set of
geometric vectors of identical orientation, identical direction, equal length
IMI, and with an amplitude Mo They arenot restricted to a common line
of action (Figure 202c)o M is counted positive if the torque acts in the
direction of the arrow M 0
Comment 2.1.4 (Vectors in rigid-body dynamics:):

Torque vectors of
identical orientation, 1
identical direction, and ~
equal length J
have an equal action on a rigid bodyo
Force vectors of
identical orientation, 1I
I
identical direction,
equal length, and ~
I
J
I
placed on the same line of action
have an equal action on a rigid bodyo
Displacement vectors of
fr
identical orientation,
identical direction,
equal length, and
: 0
with an identical start point )
are equalo
These latter statements hold for planar and spatial geometric vectorso
2.1.1.2 Algebraic vectors. Traditional vector analysis and algebra of geo-
metric vectors are not well suited for computer implementation and formula
manipulationo As a consequence, instead of using geometric vectors and vector
products, etco, the equations of motion of rigid-body systems will be written
in terms of algebraic vectors and matrices here, represented with respect to
38 20 Planar and spatial vectors, matrices, and vector functions
suitably chosen local and global frameso This simplifies numerical computa-
tions and provides conceptual clearness (as, for example, cross products of
vectors are not included in the axioms of linear vector spaces and of linear
algebra (Appendix A0101))o
Consider a reference frame R with the origin 0 = 0 R and with the planar
orthogonal basis BR = {exR, eyR} , where eiR is a basis vector in ffi.2 , with
J: {l,i=j (2o6a)
eiR•ejR=Uij:= O,i#j; z,J=x,yo
0 0
Then (Figure 203a) r PO can be represented with respect to frame R as

R
Tpo = Xpo €xR
0
+ YPO
R 0
eyR , (2o6b)
frame R yR
Y~o ----------- P
rpo
(a) Components of a vector Tpo in frame R
rso = Tpo + TQo = Tpo + rsp

Y~o + Y~o +-----+-----:.t•S
R
YQo
rsp
0 x~ 0 x~ 0 x~ 0 + x~ 0
(b) Sum of two vectors
Figo 203: Algebraic vectors

or in abbreviated algebraic form
R R R T Xpo
R )
r PO = (X PO ' YPO) = ( R E ffi.2' (2o6c)
Ypo
where x~ 0 and y~0 are the Cartesian coordinates or components of r PO

with respect to frame Ro Consider two vectors r PO and TQo, represented in
frame R (Figure 2o3b):
and R - (2o6d)
rQo-
Then
R R R R R
rso = rpo
0
0 + rQo = rpo + rsp (2o6e)
is defined as the vector sum of r~ 0 and r~ 0 with
(206f)
Consider two reference frames R and L with a common origin 0 L = 0 R = 0

and with orthogonal basis vectors {e"R, eyR} and {e"L, eyL} (Figure 2.4)0
Then r PO can be written (cfo Equation 2o6b) as
L L
rpo = Xpo 0e"L + YPo 0eyL 0 (207)
The two algebraic representations (2o6b) and (20 7) of r PO are abbreviated as
R -
rpo- ( x~
Ypo
0) representation of r PO by
Cartesian coordinates in R)
(2o8a)
and
L representation of r PO by
rpo = ( xto)
Cartesian coordinates in L)o
(2o8b)
Ypo
These representations of the geometric vector r PO are sometimes called al-

gebraic representations of r PO, or algebraic vectorso Then the following rela-
tions among the coordinates of a vector r PO represented in different frames
R and L with coinciding origins 0 = 0 R = 0 L are obtained by elementary
geometry (Figure 2.4):
YR frame R
x~0 ° cos'lj;
Figo 2.4: Vector r PO represented in frames R and L
R _ L .!, L .!,
y PO
o
X PO - X PO 0 COS 'P - 0 Slll 'P
(2o9a)
R
y PO = X PO
L
0
o
Slll 'P
.!,
+ y PO
L
0 COS 'P
.!,
or
(x! 0 ) = ( c~s '1/J , - sin '1/J )

Ypo sm'lj; , cos'lj;
0 ( xt0 )
Ypo
(2o9b)
=:ARL
or
R = ARL (•!•)
r PO 'P
L = ARL r PO
r PO L 0 0 (2o9c)
with
RL RL ( COS '1/J , - sin '1/J )
A := A ('1/JLR) = 0 (2010)
sm'lj; , cos'lj;
as the planar coordinate transformation matrix or planar orientation matrix
that maps a vector, represented in L, into a vector, represented in R, where
'1/J = '1/JLR is the angle of rotation from R to Lo The relations
A RL 0
( A RL f = ( c~s '1/J , - sin '1/J ) 0
( c~s '1/J , sin 'lj; ) (2olla)
sm '1/J , cos '1/J - sm '1/J , cos '1/J
(
cos 2 'lj; + sin 2 '1/J
sin '1/J cos '1/J - cos '1/J sin '1/J ,
, cos '1/J sin '1/J - sin '1/J cos '1/J )
sin 2 '1/J + cos2 'lj;
( 1
0,1
'0)-
- 12
and
(2ollb)
hold for arbitrary values of 'lj;o As a consequence, the matrix ARL is an or-
thogonal matrixo Due to
(2ollc)
the following relations hold:
RL RL ( COS ( -1/JRL), - sin (-1/JRL))

A (1/JLR) = A ( -1/JRL) =
sin (-1/JRL) , COS ( -1/JRL)
= ( cos'l/JRL , sin'l/JRL) ( RL
= A (1/JRL)
)T = A LR (1/JRL)
-Sill 1/JRL , COS 1/JRL
0
or
(2olld)
As a special example of the above coordinate transformation, the basis vectors

{exL, eyL} of L will be represented in frame R (Figure 205)0 Due to (2o6a),
e"L and eyL, represented in L, are written as
e"L = 1 e"Lo +0 o eyL or

and (2012a)
eyL = 0 ° e"L +1 ° eyL or
Then, in agreement with Figure 205, the following relations hold:
e"L=cos'lj;oexR+sin'lj;oeyR, 'l/J:='l/JLR, or (2012b)
e~L = ( :~:~) = ( :~:~ : - :~:~) o ( ~) = ARL e~L

o
(where e~L is identical to the first column vector of ARL), and
eyL =- sin 'ljJ 0e"R + cos'lj; 0eyR, or (2012c)

R _ ( - sin 'ljJ ) _ ( cos 'ljJ , - sin 'ljJ ) 0( 0 ) _ RL 0 L
eL- - -A eL
Y cos 'ljJ sin 'ljJ , cos 'ljJ 1 Y
(where e:L is identical to the second column vector of ARL)o By analogy with
(2012b) and (2012c), the following relations hold:
yR frame R
frame L
1
I
I ''
I
I
I
I
I
1/JLR = 1/J
I
I
I
-1 ~
I
I I
I I
\ I
\ I
I
I
I
G)
I
'
...... ____ ... ''
eyR = G~~ ~) = A RL 0
= first column of ARL

yR frame R
frame L
/ '
/
I
I
I
I
I
I
I
I
-sin'ljJ
/
1
ef}L = (- sin 1jJ) 0
exR + cos 1/J 0
eyR,
/ eyR = ( -~~:~) = ARL 0 G)

= second column of ARL
Figo 205: Representation of the basis vectors exL, eyL of frame L in frame R
L _ LR R _ ( cos'ljJ) or exR = cos 1/J exL - sin 1/J eyL

exR - A exR - 0 0
- sm 1jJ
0
o
and (2013)
sin 1jJ)
eL _ ALR
yR-
0
eR _
yR- ( or eyR = sin 1/J 0exL + cos 1/J 0eyL 0
cos 1jJ
Compared to the previously considered geometrical situation with coinciding

origins OR and OL of frames Rand L (Figure 2.4), a slightly new situation
occurs when these origins do not coincide (Figure 206)0 Then the vector r PO
from 0 R = 0 to the point P can be written as (Figure 206)
or in coordinates of R
R R R
rpo = roLo + rpoL (2.14a)
with
roLo as a translation (displacement vector) from OR = 0 to OL.
Taking into account the above rotation relation (2.9c) of L with respect to
R yields
(2.14b)
with r~ 0 L as a representation of r POL in frame L. Combining the above

two transformations implies that each coordinate transformation can be in-
terpreted as a Superposition of a translation
R R R
rpo = roLo + rpoL (2.15a)
and of a rotation
(2.15b)
Then
(2.15c)
Frame R
roLo = displacement vector of OL with respect to 0

'1/JLR = rotationangle of L with respect to R
Fig. 2.6: Translation and rotation of frame L with respect to frame R

is a combined translation rotation transformationo Written in coordinate rep-

resentation we have, with 'l/JLR as the rotationangle of frame L from frame
R
or (2016)
The inverse relation of (2o15c) is
(2017a)
with the special cases:
pure rotation roLo =0 (for OL =0), and

(2017b)
pure translation ARL = l2 (for 'l/JLR = O)o
The relations (2ollb) and (2015c) serve as a basis for deriving kinematic
relations and inverse kinematic relations of planar rigid-body systems (based
on vector loop and orientation loop equations in Beetion 3)0
The scalar product of two vectors r PO and TQo, introduced in (202), may
be algebraically written as (Appendix Aolo2)
R • rQo
rpo R 0=
0 ( R
rpo )T 0 (rQo
R ) (2018)
R R XQO
= (X PO ' YPO) 0 (
R )
R
YQo
= xpo XQo 0
+ YPo 0
YQo E lR1 o
The scalar product of two vectors is independent of their representations:
(2019)
The length of a vector r PO is
Ir pol= ( (r~o) T r~o )1/2 = ( x~o 2 + Y~o 2)1/2

0 (2020)
Due to (2o2b), the scalar product of two orthogonal vectors r~ 0 and r~ 0 is

zeroo Then
R
( T PO )T TQO
0
R
=
( R
XPO ' YPO
R ) 0
( x~o ) = x~o x~o + Y~o Y~o = 0
0 0
YQo
(2021a)
or
R R R R
Xpo 0xQo = -YPo 0YQo 0 (2021b)
Consider the vector (r~ 0 ) j_ generated from r~0 by the relation
with (2022a)
Ro- (0' -1)

1 ' 0
Then the relation
(r~of (r~o)j_ = (x~0 , Y~o)

0 0
( -y~o)
Xpo
(2022b)
= -x~o Y~o 0
+ x~o Y~o 0
= 0
j_
holds, (r~ 0 ) is orthogonal to r~ 0 , and the matrix R satisfies the orthog-
onality relation
RT OR= (1,0) (2022c)

0 ' 1
and the relation
ROR= (-1 0) '

0' -1
= -I2 (2022d)
holdso In addition
In summary, the matrix R maps a vector r~ 0 E JR2 into a vector (r~ 0 ) that j_
is orthogonal to r~ 0 and has equallengtho Then R rotates r~ 0 by 90 degrees
around an axis perpendicular to the x-y plane. As a consequence, R will be

called orthogonal rotation matrix in the x--y plane. This is in agreement with
the relation
0 , -1 ) ( cos 90° , - sin 90° ) RL 1 o

R -- ( 1 - . goo , cos goo -- A (·'P•LR -- 90 ) ' (2.22e)
, 0 - sm
where the frame L is obtained from the frame R by a rotation of '1/JLR = 90°.
The vector product of two vectors r PO and rqo in ffi.3 , and represented
in frame R, is (Appendices A.1.2 and A.1.3)
(2.23a)
+ ( Zpo.
R Xqo-
R R zqo
Xpo. R) . eyR
R R R R )
+ ( Xpo·Yqo-Ypo·Xqo ·ezR
or
Ypo
R · Zqo-
R Zpo
R · YqoR )
r POR x rQOR = ( z}J0 · x~ 0 - x}J0 · z~0 (2.23b)
R · Yqo-
Xpo R R · Xqo
Ypo R
Introducing the skew-symmetric matrix
0 , -zpo
R
, Ypo
R )
-R ,_ R R
rpo .- ( z~o ' ~ ' -xpo
-Ypo Xpo ' 0
constructed from r}J 0 = (x}J 0 , y}J0 , z}J0 )T, shows that the product
_R R
rpo ·rqo =
( ~
Zpo'
' - zfJo '
0 '-xpo
) Y~o .
( x~o
Yqo
)
(2.24)
R
-Ypo R '
Xpo 0 R
Zqo
YfJo · z~o - zfJo · Y~o )
R xqo-
( zpo. R R zqo
Xpo. R
R · Yqo-
Xpo R R · Xqo
Ypo R
yields the same vector as the cross product (2.23b). As a consequence, the
vector product (2.23) can be replaced by the product of a matrix r}J 0 times
an algebraic vector r~ 0 (2.24):
R R -R R TID3
(2.25)
r PO X r QO = r PO ' r QO E JN..
Given two vectors r~ 0 and r~ 0 , located inside the x-y plane in ffi.3 ,
( Y~oR)
Xpo xqo
R )
r~o := and r~o := ( Y~o · (2.26a)
Then the vector
Y~o · z~o - zfJo · Y~o )

-R
Tpo. rqo =
R
( R R R
Zpo. Xqo- Xpo. Zqo
R 0 )
(
R R R R R R O R R
Xpo · Yqo- Ypo · xqo Xpo · Yqo- Ypo · xqo
(2.26b)
is perpendicular (orthogonal) to the x-y plane. As a consequence, the x-y
plane is not closed under vector-product operations. In the planar case the
above cross product operation will be abbreviated in the form
-L L ' L
wLR. Tpo = R. '1/JLR. Tpo (2.27a)
with
L R ( .
WLR=WLR= 0,0,'1/JLR
)T =(0,0,1) T·'1/JLR
. (2.27b)
as the angular velocity vector around the z-axis perpendicular to the x-

y plane, with ~LR as the time derivative of the rotation angle '1/JLR, with
R as the orthogonal rotation matrix in the x-y plane (Equation 2.22e and
Section 2.1.2.1), and with
0) ~LRl
wfR := ( wfR restricted to the x-y plane in ffi.3 ) (2.27c)
-1
0 :0 · ( 0,-1)
1 ' 0 °
0
'1/JLR =R 0
0
'1/JLR,
0 '0
or (2.27d)
WLR = R · ~LR · ezR = R · ~LR · ezL.
The above notation will be used in the planar case to obtain identical formal
relations for rotations of planar and spatial mechanisms.
2.1.2 Time derivatives of displacement vectors and orientation

matrices
Time derivatives of displacement vectors and orientation matrices provide ve-

locity vectors, angular velocity vectors, acceleration vectors, and angular ac-
celeration vectors as basic ingredients of rigid-body dynamics. In this section,
time derivatives of planar vectors and orientation matrices will be considered.
2.1.2.1 Velocities and angular velocities. Consider a planar vector

roLo from point 0 of frame R to point OL (Figure 206), where roLo is
assumed to change its length and orientation smoothly in time; ioeo,
(2o28a)
The time derivative of roLo(t)
can be measured with respect to frame R while roLo(t) may be represented

in frame L, and vice versao Consider the vector roLo(t), represented in frame
R,
or (2o28b)
Introducing the time derivative operator with respect to frame R "Rd/ dt"
and applying this operator to roLo(t) yields:
Rd Rd Rd
dt (roLo(t)) = dt (x~Lo(t) 0exR) + dt (y{t 0 (t) 0eyR) 0 (2028c)
The basis vectors eiR (i = x , y) of frame R are, by definition, constant in Ro

This yields together with Rd eiR
dt
= 0 (i = x , y) the relation
R~t (roLo(t)) = X~Lo(t) 0

€xR + iJ!to(t) 0
eyR , (2028d)
or written in abbreviated form
Rr~L 0 (t) := ( x~L 0 (t), y{JL 0 (t)) T (2o28e)
with unambiguous real numbers x~L 0 (t), y{JL 0 (t)o The same result holds for
the time derivative of another vector r~ 0 L (t) with respect to frame R:
(2028f)
On the other hand, representing the vector r~ 0 L (t) in the form
(2029)
yields (Aol.llf)
Rd ( R ) Rd ( RL L ) 0 RL L Ld (
dt rpoL = dt A 0rpoL = A 0rpoL + A RL0 dt L )
rpoL ' (2o30a)
or written more compactly
RoR oRL L
rpoL =A 0 rpoL + A RL 0
LoL
rpoL' (2o30b)
or (Aol.llf)
Rro R _ ARL
POL- 0 ( - L
WLR
L
0TpQL + L TpQL
L0
)
=
ARL 0R TpQL'
L0
(2o3la)
with (Aol.lle)
R L Rd ( L ) - L L L L
(2o3lb)
dt TpQL
o o
TpQL:= =WLROTpoL+ TpQL'
with (Aol.llb)
ALROA_RL=Ro~LR= 0 , -~LR) -L (2o3lc)

=:WLR'
0
(
'1/JLR, 0
or (Aol.llc)
RL RL -L RL
A =A (2o3ld)
0 0
OWLR=A ORO'l/JLR,
with the orthogonal planar rotation matrix R, with the angular velocity vec-
tor (Aol.lld)
(2o3le)
and with w};R as defined in (2027c)o Subsequently the following abbreviations

will be used:
oL LoL
rpoL := rpoL =
Ld ( L )
dt rpoL =
(oL oL
XpoL' YPoL
)T (2o32a)
and
(2032b)
Then
(2033)
or
and finally
2.1.2.2 Accelerations and angular accelerations. Consider the previ-

ously derived velocity vector (2033)
R oR
Vpo := Tpo :=
d ( Tpo
dt R ) (2034)
=
0
R
( R T0L0 + ARL 0WLR TpoL +
- L 0 L ARL 0L L
TpoL
) 0
Rd
The associated acceleration vector a~ 0 := - - (v~ 0 ) can be written as
dt
(2o35a)
or as
R RooR RL -L L 0
RL L :.L L
apo= ToLo+A oWLROTpoL+A 0 WLROTpoL
+ A RL OWLRO
- L L L
TpQL
+ Ao RL
o
0
L oL
TpQL
+ ARL 0
L L
oo
TpQLo (2o35b)
Inserting the relation (2o3ld)
Ao RL-
-
ARL -L
OWLR (Aol.3)
into (2o35b) yields the acceleration relation
+2 0
A RL OWLRO
- L L L
TpQL
o + ARL 0
L L
TpQLo
oo
(2036)
Assuming that the vector r~ 0 L is constant in the frame L ( rigid-body property

in the case that L is a frame fixed on a rigid body) provides
LToL -0 and L .. L _ 0 (2037)

POL= TpoL = '
and yields the relations

R R R R
o
TpQ =
o
T0L0 + ARL 0
- L
WLR 0
L
TpQL (2o38a)
and
(2o38b)
Written in components, the acceleration relation (2036) is
or
-sin'I/JLR, -cos'I/JLR) (x~oL) ..

+ ( 0
o'I/JLR (2039)
cos'I/JLR , -sin'I/JLR Y~oL
+ ( x~ 0 L
° cos '1/JLR - Y~o L 0 sin '1/JLR )
"L
xp 0 L 0 sm 'f'LR
o .!,
+ "L
YpoL 0 cos 'f'LR
.!, '
and finally
ooR = xooR L - X L
Xpo 0 0 poL 0
(•i•'f'LR ) 2
0
cos .!,
'f'LR + YPoL
L 0
(•i•'f'LR ) 2
0
sm .!,
o
'f'LR
L 00 L 00
-X poL 0
'l/JLR 0
sin 'l/JLR- YpoL 0
'l/JLR 0
cos'l/JLR
-
2 0
0 L
X poL 0
.i. .!.
'f'LR sm 'f'LR-
0
2 0
0
YPoL
0 L 0
.i. .!.
'f'LR cos 'f'LR0
+ X~oL ° COS 'l/J LR - Y~OL 0

Sill 'l/J LR (2.40a)
and
ii~o = ii~Lo- x~oL 0

(-J;LRf 0
sin'l/JLR- Y~oL 0
(-J;LRf 0
cos'l/JLR
L 00 L 00
+X poL 0
'l/JLR 0
cos'l/JLR- YpoL 0
'l/JLR 0
sin 'l/JLR
+2 ° x~ 0 L 0
-J;LR 0
cos 'l/JLR- 2 ° y~ 0 L 0
-J;LR 0
sin 'l/JLR
+x~ 0 L 0
sin'l/JLR+ii~oL 0
cos'l/JLRo (2.40b)
In addition, the following relations hold:
1 R R ARL L
0
0
WLR = 0
0
WLR with (2.41a)
-R _ ARL
W w-L ALR (2o4lb)
20
LR- 0
LR 0
'
30 (2o4lc)
Proof of (2.41a):
The relation
R oR Rd ( RL L ) RL L RL L 0
o
WLR=dt A OWLR =A OWLR+A OWLR
yields together with
Ao RL-
-
ARL -L
OWLR
the relation
RoR ARL -L L ARL oL
WLR= OWLROWLR+ OWLR'
and with
the relation (2.41a)o D

2.2 Spatial vectors and matrices 53
Proof of (2.41b):
The relation
-R
WLR'T R = ARL . (-L
WLR'T L)
= A RL 0 w±R 0 ( A LR 0 A RL) 0 rL = A RL 0 w±R 0 A LR 0 rR
implies
-R _ ARL w-L
W ALR D
LR- ' LR' ·
Proof of (2.41c):
R.:,R - Rd (ARL -L ALR)

WLR- - - ·WLR'
dt
-
-
A' RL - L
·WLR'
ALR + ARL ' L WLR'
.:, L ALR + ARL - L
·WLR'
A' LR
provides, together with

A. RL-
-
ARL -L
·WLR
and A. LR- ALR -R -
- ·WRL--
ALR -R
·WLR'
the relation
R .:, R ARL - L - L ALR
W LR = . W LR . W LR .
+ A RL ' Lw.:,LLR' ALR + ARL ' w-LLR' ALR ' w-RRL·

Using
A RL ·WLR'
-L ALR -R
·WRL
-_ -ARL . w-L
LR . ALR . w-R
LR . ARL . ALR -_ -ARL . w-L
LR . w-L
LR . ALR
-L
:WLR
yields
R.:, R ARL L .:, L ALR
WLR = . WLR. . D
2.2 Spatial vectors and matrices

In this chapter, vector representations, operations and transformations in
the space OC3 will be discussed together with their time derivatives. The basic
differences from the planar case OC2 are:
1. That the space OC3 is closed under vector-product operations.
2. That the angular velocity vector in OC3 is in general no longer computed
as time derivative of an angle as in case of the OC2 .
In the spatial case angles and angular velocities are related by means of
the kinematic differential equation. This equation may include singulari-
ties that can be avoided by formulating rotations by means of quaternions

or by Euler parameters ([42], [43], [44], [45]) instead of Bryant angleso
Due to the introductory character of this book, quaternions and Euler
parameters will not be considered hereo
2.2.1 Displacement vectors, frames, and orientation matrices
The basic definitions and statements of geometric vectors in the plane pre-
sented in Section 201.1.1 also hold in the spatial case, with slight modifica-
tionso As in the planar case, the spatial vectors and matrices will also be
written in bold faced letterso
Consider a reference frame R with origin 0 = 0 R and with the orthogonal
basis BR = { exR' eyR' ezR} ' eiR E OC3 ' defined by the relation
( R
eiR
)T 0
R
ejR =
{ 1 , i =j
0 ' i -1- j i,j=x,y,z, (2.42a)
with efk as the basis vector eiR represented in Ro Consider a second frame
L with origin OL and with orthogonal basis BL = {exL, eyL, ezL} fixed
to a rigid body (Figure 207a)o Let roLoR be the (geometrical) displacement
vector from 0 to OLo Then roLoR = roLo can be represented with respect
to Ras
(2.42b)
or in algebraic form
(2.42c)
with x~Lo, y~Lo, z{5L 0 as Cartesian coordinates of roLo with respect to

frame Ro Consider an arbitrary point P on a rigid body (Figure 207a) and
the displacement vectors r POL from 0 L to P and r PO from 0 to P 0 Then
the following translation relation holds:
rpo = rpoL + roLo,

or, representing r PO in the frame R, (2.43)
Consider the vector r PO in Figure 20 7b and two orthogonal frames R

and L with a common origin 0 and with basis vectors { e"R, eyR, ezR} and
{exL, eyL, ezL}, respectivelyo Then rpo can be written as
R
rigid body
/
/
/
/
/
/
/
/
/
"'--------
/
/
/
/
/
/
/
/
/
(a) Frames and displacement vectors
ezR
ezL z~o ..... R
..... .....
..... ..... p
L_..
Y~o
eyL
z~o
Y~o
..... ..... eyR
..... /
_ _ _ .....
/
_ ':...V
exL
(b) Representation of a vector r PO in different frames R and L
Fig. 2.7: Vector representations

Tpo
R 0exR + YPO
= Xpo R 0eyR + Zpo
R 0ezR, (2.44a)
or
rpo =
L 0exL
Xpo
L
+ YPo 0eyL L 0ezL,
+ Zpo (2.44b)
or abbreviated in algebraic form
R = ( Xpo'
Tpo
R YPO' R )T
R Zpo (representation of r PO in R) (2.44c)
and
L = ( Xpo'
Tpo
L YPO' L )T
L Zpo (representation of r PO in L) 0 (2.44d)
2.2.1.1 Basis transformation. Given two frames R and L (Figure 207b)

with orthogonal basis vectors {exR, eyR, ezR} of Rand {exL, eyL, ezL} of
Lo Then
exR = 1 oexR + 0 oeyR + 0 oezR,
eyR = 0 exR + 1 eyR + 0 ezR,
° ° ° (2.45a)
ezR = 0 oexR + 0 oeyR + 1 oezR,
or
e~R~ 0). e:n~ G} R --

and ezR
0).
and
exL = 1 oexL + 0 oeyL + 0 oezL,
eyL = 0 ° exL + 1 ° eyL + 0 ° ezL,
ezL = 0 oexL + 0 oeyL + 1 oezL,
or
e~L~ 0). e~L~ 0} L

and ezL =
(D
Consider the projections of exL, eyL and ezL onto the basis vectors of R;
(2.45b)
ioeo, the expressions
(e:Rf 0 (e~L) =: fx , (e:R)T 0 (e~L) =: fy , (e:R)T 0 (e~L) =: fz
(e:Rf 0 (e~L) =: mx, (e:R) T 0 (e~L) =: my, (e:R) T 0 (e~L) =: mz
(e~)To (e~L)=:nx, (e~)To (e~L)=:ny, (e~)To (e~L)=:nz

(2.45c)
202 Spatial vectors and matrices 57
with numbers Ci, mi, ni, (i = x, y, z), called direction cosines due to the
relation
with O:x as the angle between the basis vectors exR and exLo Then
exL= ((e~R)T (e~L)) oexR+ ((e~)T (e~L)) oeyR+ ((e~)T (e~L)) oezR,
eyL= ((e~R)T (e~L)) oexR+ ((e~f (e~L)) oeyR+ ((e~)T (e~L)) oezR,
€zL= ((e~R)T (e~L)) oexR+ ((e~Rf (e~L)) oeyR+ ((e~R)T (e~L)) oezR,
(2.46a)
or
exL = fx exR + mx eyR + nx €zR,

0 0 0
eyL = Cy exR + my eyR + ny ezR,

0 0 0 (2o46b)
€zL = fz €xR + mz €yR + nz €zR,
0 0 0
and
(2.46c)
By analogy to (2.45c) and (2.46b), the following basis transformation relations

hold:
exR = fx exL + Cy eyL + fz €zL,

0 0 0
eyR = mx oexL + my oeyL + mz oezL, (2o47a)

€zR = nx exL + ny eyL + nz €zL,
0 0 0
or
(2.47b)
Inserting (2.47a) into (2.46a) yields, together with (2o45c),

exL = lx (Cx exL + Cy eyL + lz ezL)
0 0 0 0 (2.48a)
+ mx o (mx oexL + my oeyL + mz oezL)
+ nx o (nx oexL + ny oeyL + nz oezL),
eyL = Cy o (Cx oexL + Cy oeyL + lz oezL)
+ my o (mx oexL + my oeyL + mz oezL)
+ ny o (nx oexL + ny oeyL + nz oezL),
ezL = lz (Cx exL + Cy eyL + lz ezL)
0 0 0 0
+ mz o (mx oexL + my oeyL + mz oezL)

+ nz o (nx oexL + ny oeyL + nz oezL),
or
exL = ( c; + m; + n;) oexL + (lx ° Cy + mx 0
my + nx 0
ny) oeyL (2.48b)
=: 1 =: 0
=: 0
eyL = (Cy fx o + my 0
mx + ny 0
nx) oexL + (C~ + m~ + nn oeyL
=: 0 =: 1
=: 0
=: 0 =: 0
+ (c; + m; + n;) oezL
=: 1
together with the six relations
(Cx oCy + mx omy + nx ony) = 0,
(Cx olz + mx omz + nx onz) = 0,
(2o48c)
(Cy olz + my omz + ny onz) = 0,
(f-T + m; + nT) = 1 , (i = X , y , Z) o
Introducing the direction cosine matrix
fx , f!y , fz )
( (2.48d)
mx 'my 'mz
nx ' ny ' nz
the relations (2 .48b) and (2 .48c) prove the orthogonality relations
ARL 0 (ARL) T = ARL 0 ALR = 13

or (2o49a)
( A RL)T = ( A RL)-l =A LR 0
Then, due, to (2.46b) and (2.46c),
R
exL =
(~) =ARL 0
0) = ARL 0eLxL' (2o49b)
R -
eyL -
(~) =ARL 0
(D = ARL 0eLyL'
R -
ezL - (::) = A"' G) = A"' e~L'
and
L
exR = ALR 0exR'
R L
eyR = ALR 0eYR'
R eL
zR
= ALR 0eRzR
with (2050)
ALR = (ARL)T 0
2.2.1.2 Coordinate transformation. Given two orthogonal frames Rand
L with a common origin 0 (Figure 207b) and a vector rpo from 0 toP with
representations in R and L,
R
rpo = Xpo 0exR + Ypo
R 0eyR + Zpo
R 0ezR (2o5la)
and
L L L
rpo = Xpo 0exL + Ypo 0eyL + Zpo 0ezL, (2o5lb)
or
( R)
Xpo Xpo
R _ R L _ L )
rpo- Y~o and rpo- ( Y~o 0
Zpo Zpo
Inserting the basis transformation relations (2o47a) into (2o5la) yields
rpo= x~ 0 ° (fxoexL+fyoeyL+fzoezL)
=: exR
+ Y~o o (mx oexL + my oeyL + mz oezL)
=: eyR
+ Z~o o (nx oexL + ny oeyL + nz oezL),
=: €zR
r PO = ( x~o 0Rx+ Y~o 0mx + z~o 0nx) 0exL

+ (x~ 0 Cy + y~0 my + z~0 ny) eyL
° ° ° 0
+ (X~o 0Rz + Y~o 0mz + z};o 0nz) 0€zL

or
+ mx y};o
( :t~)
0
+ my y};o 0 (2052)
z~o + mz y.J;o 0
..._"_.._..,
= rL
0
0PO
and finally
L
rpo = ALR rpo
R
0 (2o53a)
with the transformation matrix
Cx, mx, nx)

A LR = ( Cy , my , ny = ( A RL)T = ( A RL)-l (2o53b)
Rz , mz , nz
that maps the coordinates of r PO represented in frame R into coordinates

represented in frame L ( coordinate transformation) 0
2.2.1.3 Bryant angles. Combining the translation and rotation relations

(2.43) and (2.53a) provides the following representations of the vector r PO
in Figure 2. 7a:
TpO = TOLO + TpOL (2.54a)

and
R R R
rpo = roLo + rpoL ( r POL represented in frame R) (2.54b)
or
(r POL represented in frame L). (2.54c)
Thematrices A RL and A LR of (2.53a) include nine elements, called direction

cosines. Due to the orthogonality relation (2.49a), these coordinates are re-
stricted by six additional algebraic relations (2.48c). As a consequence, only
three of these nine numbers are independent. In practical applications, rep-
resentations of ARL usually only include three (independent) coordinates.
Among those coordinate representations, the Cardan- or Bryant-angle repre-
sentation is often used in mechatronic systems, due to the fact that the matrix
H(p) ofthe associated kinematic differential equation (Section 2.2.2.3) has no
singularity for small Cardan angles (contrary to the Euler angle representa-
tion). The Cardan angle representation of ALR is constructed by interpreting
the spatial rotation of a rigid body with a body fixed frame L with respect
to frame R as result of three successive rotations. Before the first rotation,
frames Land R coincide (Figure 2.8). The first rotation is carried about the
e"waxis through an angle </J := rPL"R· This results in the auxiliary frame L"
with basis vectors { exL" , eyL" , ezL"} and exL" = exR· The second rotation
of L" through an angle e := eL'L" about the eyL"-axis provides a second aux-
iliary frame L' with basis vectors { e"L' , eyL' , ezL'} and eyL' = eyL"· The
third rotation of L' through an angle 'lj; := '1/JLL' about the ezL'-axis provides
the desired final orientation of the body and frame L. The transformation
matrices of these different planar rotations are:
( 1, 0, 0)
z L" L" y L"
AL" R = 0 ' c <P ' s <P (2.55a)
0 , -s </J , c </J
(first rotation of frame R

about the e"R-axis into L",
</J := rPL" R = -</JRL" ),
X
L"
X
L' L' z L'
cO , 0, -sO)
AL'L" = ( 0 1 0 (2.55b)
' '
sO , 0 , cO
(second rotation of frame L" z L"

about the eyL"-axis into L',
eyL" = eyL' L"
0 := OL'L" = -OL''L'),
y L'
c '1/J , s '1/J , 0 ) YL
L'L"
A = ( -s'lj; , c'lj; , 0 (2.55c)
0 ' 0 ' 1
(third rotation offrame L' X
L'
about the ezL"-axis into L,
€zL' = €zL L'
'1/J := '1/JLL' = -'1/JL'L) ,
with
scjJ := sinc/J and ccjJ := coscjJ. (2.55d)
The complete rotation matrix A LR that describes the rotation from R to L

is
A LR =A LL' ·A L'L" ·A L"R
or
cO·c'I/J , cc/J·s'I/J+sc/J·sO·c'I/J, sc/J·s'lj;-cc/J·sO·c'I/J)

ALR= ( -cO·s'I/J, cc/J·c'lj;-sc/J·sO·s'I/J, sc/J·c'I/J+cc/J·sO·s'I/J .
sO , -sc/J·cO , cc/J·cO
(2.56)
This matrix describes the transformation of coordinates of a vector repre-

sented in frame R into coordinates represented in frame L in terms of Bryant
angles.
For sufficiently small rotation angles cjJ , 0, and '1jJ, the linear approxima-
tions
I
I
'
\
I
I
I
eyL' = eyL"
I
/ I
\
exR = exL" ' /
/
I
exL
(a) Diagram of three successive rotations (around exR by c(J, around eyL" by
(} and around ezL' by 'lj;)
ezL"
(1, 0, 0)
-sinc/J
AL"R= 0, cos c(J, sin cfJ
0, - sin c(J, cos cfJ
(b) First rotation around exR by cfJ = cfJ L" R = -cp RL"
Fig. 2.8: Bryant-angle transformation
(2o57a)
provide the matrix
1' '1/J' -{})

A~:an ~ ( -'1/J , 1 , <P (2o57b)
() ' -<P ' 1
with the skew-symmetric matrix
0 , -'1/J , +B) ( </J )

j, := ( +'1/J ' 0 ' -<P =: () 0
(2o57c)
-{} ' +<P ' 0 '1/J
that is constructed from the vector (</J , () , 'lj;) T Then
0
(2o57d)
This implies that within linear approximations small rotation angles can be
added like vectorso Due to the relation
(2o57e)
it makes no difference whether this "small angle vector" is represented in

frame R or in frame Lo As a consequence, the sequence of small rotations is
arbitraryo
A general transformation of a vector can be interpreted as a Superposition
of a translation (see Equation 2.43)
R R R
rpo = roLo + rpoL (2o58a)
and a rotation
R ARL L
TpoL = 0 TpoL 0
(2o58b)
This yields the combined ( translation-rotation) transformation

R R ARL L (2o59a)
rpo=roLo+ orpoL
and its inverse
r POL
L - ALR
- 0
( R
r PO -
R
r 0 L0
)
'
with the special cases
pure rotation and

(2o59c)
pure translation
These latter two relations are the basis for deriving kinematic relations of
rigid-body systems (vector loops and orientation loops) under spatial motion
(see Beetions 3 and 5)0
2.2.2 Time derivatives of displacement vectors and orientation

matrices
In this section, time derivatives of spatial displacement vectors and orien-

tation matrices will be considered together with the kinematic differential
equation that relates angular velocities to the time derivatives of angleso
2.2.2.1 Velocities and angular velocities. Consider a vector from the
point 0 offrame R to a point OL (Figure 207a), where roLoL is assumed to
move smoothly in time,
(2o60a)
Represented in frame R, the vector roLo(t) is written as
roLo(t) = x~Lo(t) 0exR + Y~Lo(t) 0eyR + z{!;Lo(t) 0ezRo (2o60b)
Taking the time derivative of r~L 0 (t) with respect to frame R yields
RoR Rd Rd ( R )
roLo(t) := dt (roLo(t)) := dt x 0 L0 (t) 0
exR (2o60c)
Rd Rd
+dt (yf5Lo(t) 0eyR) + dt (zf5Lo(t) 0ezR) 0
Rd
By the definition of the operator " - " (see Equations 2028a to 2028f), the
dt
basis vectors eiR (i = x, y, z) of frame R do not depend on to This yields,
together with eiR := 0, the relation
or
(2o60e)
with unambiguous time derivatives x~L 0 (t), y{!;L 0 (t), z{!;L 0 (t) of the scalar
functions x~L 0 (t), y{!;L 0 (t) and z{!;L 0 (t)o The same result holds for the time
derivative of the vector r POL (t) with respect to frame R:
(2o60f)
Assurne that the vector r POL (t), fixed on a moving body, is represented in a
body-fixed frame L, but differentiated with respect to time in frame Ro Then
(see Equation Aol.31f):
(2061)
or
R R RL L
= A + A RL L L
(2062)
0
o o
rpoL 0 rpoL 0 rpoLO
Together with (Aol.31b) and (Aol.39a)
A LR 0
Ao RL- -L
-WLR or Ao RL-
-
ARL -L
OWLR or Ao RL 0
ALR- -R
-WLR'
and with (Aol.39b)
and with the formal vector wfR = ( w~LR, w~LR, w~LR) T, this provides the
expressions (see Equation Aol.31d)
(2063)
and (see Equation Aol.31c)
R oL Rd ( L ) - L L oL A LR oR
TpoL = dt TpoL = WLR 0
TpoL + L TpoL = 0
R
TpoLO (2064)
The relation
together with
LToL -0 (rigid-body condition) (2o65b)
POL=
implies
R RoR
Vpo := T0L0 + ARL 0
-L
WLR 0
L
TpoLO (2o65c)
2.2.2.2 Accelerations and angular accelerations. The acceleration vec-

Rd
tor a~ 0 := dt (v~ 0 ) can be written as
R Rd ( R ) RoR
apo := dt Vpo = Vpo (2o66a)
=
Rddt (RoR
ToLO +
ARL -L L
0WLR 0TpoL +
ARL LoL )
0 TpoL
or as
(2o66b)
Inserting the relation (Aol.39a)
Ao RL-
-
ARL -L
OWLR
into (2o66b) yields the acceleration relation
+ ( 2 0A RL OWLRO L
- L L TpoL+ ARL 0L TpoL
o .. L ) 0 (2067)
Assuming that the vector r~ 0 L is constant in frame L ( rigid-body property

in the case that L is a frame fixed on a rigid body) provides
LToL =0
POL- and L .. L
TpQL
_ 0
= (2068)
and yields the relations

R R
TpQ
o
= R T0L0
o R
+ ARL 0
-L L
WLR TpQL 0 (2o69a)
and
(2o69b)
2.2.2.3 Kinematic differential equation. In this section the formal rela-

tion (Ao1o39a) will be derived and physically interpreted for the special case
that ARL is represented in terms of Bryant angles (see Equation 2056)
ALR = (ARL) T
c(}oc'l/J , ccpos'l/J+scpos(}oc'l/J , scpos'lj;-ccposßoc'lj;)

(
= -c (} 0s 'ljJ , c cp 0c 'ljJ - s cp 0s (} 0s 'ljJ , scpoc'l/J+ccpos(}os'l/J 0
sB -scpoc(} ccpoc(}
This will achieve a physical interpretation of the vector as an angular wiR

velocity vectoro The time derivative of A LR is (using the abbreviations 2o55d)
oLR
A (20 70)
(-sßoc'lj; 0 8-cßos'lj;o~) 0
(
,
= (sßos'lj;oß-cßoc'lj;o'lj;)
cßo8
-s rjJ s 'ljJ ~ + c rjJ c 'ljJ ~

0 0 0 0
+ c rjJ s (} c 'ljJ 0 0 0
~ + s rjJ c (} c 'ljJ 0 0 0
8 - s rjJ 0
s (} s 'ljJ ~ 0 0
,
-s rjJ c 'ljJ ~
0 0
- c rjJ s 'ljJ ~
0 0
- c rjJ s (} s 'ljJ ~
0 0 0
- s rjJ c (} s 'ljJ
0 0 0
8 - s rjJ 0
s (} c 'ljJ ~ 0 0
,
-c rjJ c (} ~ 0 0
+ s rjJ 0
s (} 0
8
c rP s 'ljJ ~ + s rP c 'ljJ ~ + s rP s (} c 'ljJ ~- c rP c (} c 'ljJ ~ + c rP s (} s 'ljJ ~)

0 0 0 0 0 0 0 0 0 0 0 0 0
c rP c 'ljJ rP - s rP s 'ljJ 'ljJ - s rP s (} s 'ljJ rP + c rP c (} s 'ljJ + c rP s (} c 'ljJ 'ljJ

0 0 0 0 0 0 0 0 0 0
(}
0 0 0
-s rjJ c (} ~- c rjJ s (}
0 0 0 0
8
or
ÄLR= :t (ALR(ry))
0 , -s cp s 'ljJ + c cp s () c 'ljJ , c cp s 'ljJ + s cp s () c
0 0 0 0 0 0
'ljJ)
= ( 0 , -s cp c 'ljJ - c cp s () s 'ljJ , c cp c 'ljJ - s cp s () s 'ljJ
0 0 0 0 0 0 0
~
0, -ccp cB , -scp cB
0 0
-sBoc'lj;, scpoc()oc'lj;, -ccpoc()oc'lj;)

+ ( s () s 'ljJ , -s cp c () s 'ljJ , c cp c () s 'ljJ
0 0
iJ 0 0 0 0
cB scpos() -ccpos()
-c () s 'ljJ , c cp c 'ljJ - s cp s () s 'ljJ , s cp c 'ljJ + c cp
0 0 0 0 0 0
s () s 'ljJ ) 0
+ ( -c () c 'ljJ , -c cp s 'ljJ - s cp s () c 'ljJ , -s cp s 'ljJ + c cp

0 0 0 0 0 0
s () c 'ljJ 0 0
~
0 ' 0 ' 0
On the other hand
RL LR
A A = ('yii) i =
0
0
1, 2, 3 with (20 72)
j=1,2,3
/'11 = - s '1/J c () 0 0
( 8 ° s '1/J s () - -J; c '1/J c ())
0 0 0
+ c '1/J c () 0 0
( -8 ° c '1/J s () - -J; s '1/J c ()) + 8 ° c () s (),
0 0 0 0
/'12 =c'l/Joc()o ( --J;os</Jos'l/Jos()+~oc</Joc'l/Jos()

+ 8 s <P c '1/J c () _ ~ s <P s '1/J + -J; c <P c '1/J)
0 0 0 0 0 0 0
-s'lj!oc()o ( -~oc</Jos'l/Jos(}--J;os</Joc'l/Jos()
_ 8 ° s <P s '1/J c () _ -J; c <P s '1/J _ ~ s <P c '1/J)
0 0 0 0 0 0
+s()o (8os</Jos(}-~oc</Joc()),
/'13 = - s 'ljJ c () 0 0 ( - ~ s <P s 'ljJ s () + -J; c <P c 'ljJ s ()

0 0 0 0 0 0
+ 8 c <P s '1/J c () _ -J; s <P s '1/J + ~ c <P c '1/J)

0 0 0 0 0 0 0
+ c'ljJ c() (-J; c</J s'ljJ

0 0 0 0 0 8 () + ~ s</J c'ljJ 0 0 0 8 ()
_ 8 c <P c '1/J c () + ~ c <P s '1/J + -J; s <P c '1/J)

0 0 0 0 0 0 0
+ s () 0
( -8 ° c <P s () - ~ s <P c ()) '
0 0 0
/'21 = ( 8 ° s '1/J s () - -J; c '1/J c ()) 0 0 0 0

( c <P c '1/J - s <P s '1/J s B)
0 0 0
+ ( -8 ° c 'ljJ s () - -J; s 'ljJ c ()) 0 0 0 0

( s <P c 'ljJ s () + c <P s 'ljJ)
0 0 0
-8os</J(cB) 2 ,
')'22 = (s </J c 'ljJ s () + c </J s 'ljJ)
0 0 0 0
( - -J; s </J s 'ljJ s ()
0 0 0
+ ~oc</Joc'ljJos()+8os</Joc'ljJoc(}-~os</Jos'ljJ+-J;oc</Joc'l/J)
+ ( -~oc</Jos'l/Jos(}--J;os</Joc'l/Jos()-8os</Jos'l/Joc()
- -J; c <P s 'ljJ - ~ s <P c 'ljJ)
0 0 0 0 0 ( c <P c 'ljJ - s <P s 'ljJ s ())
0 0 0
-s</Joc()o (8os</Jos(}-~oc</Joc()),
/'23 = ( c <P c 'ljJ - s <P s 'ljJ s ())

0 0 0 0
( - ~ s <P s 'ljJ s ()
0 0 0
+ -J;oc</Joc'ljJos()+8oc</Jos'ljJoc(}--J;os</Jos'ljJ+~oc</Joc'ljJ)
+ (s q; c 'ljJ s () + c q; s 'ljJ)
0 0 0 0 ( -J; c q; s 'ljJ s ()
0 0 0
+ ~os</Joc'ljJos(}-8oc</Joc'ljJoc()+~oc</Jos'ljJ+-J;os</Joc'ljJ)
- s <P c () 0 0
( -8 ° c <P s () - ~ s <P c ()) '
0 0 0
1'31 = ( iJ 0s '1/J 0s e - ~ 0c '1/J 0c e) 0(c q; 0s '1/J 0s e+ s q; 0c '1/J)

+ ( -iJ 0c '1/J 0s e - ~ 0s '1/J 0c e) 0(s q; 0s '1/J - c q; 0c '1/J 0s e)
+ iJ 0
c </J 0
( c B) 2 ,
1'32 = (s </J 0s 'ljJ - c </J 0c 'ljJ 0s 8) 0 ( - ~ 0s </J 0s 'ljJ 0s e

+ ~oc</Joc'l/Josß+Öos</Joc'l/Jocß-~os</Jos'l/J+~oc</Joc'l/J)
+(c</Jos'l/Josß+s</Joc'l/J) 0 ( -~oc</Jos'l/Josß
- ~os</Joc'l/Josß-Öos</Jos'l/Jocß-~oc</Jos'l/J-~os</Joc'l/J)
e
+ c <P 0c 0 ( iJ 0s <P 0s e - ~ 0c <P 0c e) ' and
e
1'33 = ( c q; 0s '1/J 0s + s q; 0c '1/J) 0 ( - ~ 0s q; 0s '1/J 0s e
+ ~oc</Joc'lj!osß+Öoc</Jos'lj!ocß-~os</Jos'l/J+~oc</Joc'l/J)
+ (s </J 0s 'ljJ - c </J 0c 'ljJ 0s B) 0 ( ~ 0c </J 0s 'ljJ 0s e
+ ~os</Joc'ljJosß-Öoc</Joc'ljJocß+~oc</Jos'ljJ+~os</Joc'ljJ)
e
+ c q; 0c 0 ( -iJ 0c q; 0s e - ~ 0s q; 0c e) 0
This provides (after some trigonometric manipulations) the skew-symmetric
matrix
AR~ ALR (2073)
0 , -s<f;oÖ-c<f;ocßo~, c<f; Ö 0-s<f;oc~o~)

0 0
0
= ( e
s q; 0 + c q; e 0c 0'1/J , o , -<P- s 0'1/J 0 e
-c q; 0iJ + s q; 0c e 0~ , +~ + s e 0~ o
Consider, on the other hand, the angular velocity vectors (Section 20201.2)
-C'~'") crR)
(2o74a)
R (rotation araund the x-axis
WL"R
from R to L", represented in R),
(2o74b)
( OL:L'') ( w,~L'')
L" (rotation araund the y"-axis
WL'L" :=
from L" to L', represented in L"),
c:J c:J
(2o74c)
L' (rotation araund the z'-axis
WLL'
from L' to L, represented in L')o
The resulting angular velocity vector of these three relations, represented in

frame R, is
Using the abbreviations

OR OL" OL'
'1/J :='1/JLL' (2o74e)
0 0
<P := <PL''R , e := eL'L"

0
and the relations (see Equation 2055)
( 1, 0, 0)
A RL" owL'L"
L" = 0, -s</J c</J,
O,s</J, c</J
and
A RL" 0 A L"L' 0w L'

1,
LL' = ( 0 , c </J , -s </J
o, o) ( ce,o,se) (o) 0 0 , 1 , 0 ° 0
0 ' s <P ' c <P -s e ' 0 ' c e ~
( 1, 0, 0)
0 , c </J , -s </J 0
(sßo~)
0
0, s</J, c</J cßo~
sßo~ )
( -s</J 0
ce ~ 0
c</Jocßo'ljJ
yields the resulting vector wrR:
wrR = ( ~) + ( 0) oiJ + (
0 C </J -S
sß )
</J oC ß o~ =
( ~+sßo~
- s </J oc ß o~ )
c </J o~
0 s</J c</Jocß s</Joß+c</Jocßo'ljJ
(20 75)
and the associated skew-symmetric matrix

-R (20 76)
WLR =
0 o 0 , -s <P iJ- c <fJ c e ~ , c <P iJ 0- s <P 0 c ~ 0 ~)

0 0 0 0
( s</Joß+c</Jocßo'ljJ , 0 , -</J-sßo'ljJ 0
-c <P 0iJ + s</J 0 ce 0~ ' ~ + s e 0~ 0

Comparing (20 73) and (20 76) yields
A RL
o
0
ALR- -R
-WLR or A RL-
o -R
-WLRO
ARL
or (20 77)
A RL-
o
-
ARL -L
OWLRO
This proves that the formal vector

L ALR R
WLR = OWLR'
introduced in (Aol.38), can be physically interpreted as the angular velocity

of frame L, with respect to frame R, represented in Lo The relation (20 75)
can be written as
1, 0 , sO ) ( ~)
( 0 ' c cjJ ' -s cjJ c 0
0
° ~
O,scjJ, ccjJocO 'ljJ
..._"_,__..,
=: H-1(17) =: il
(2o78a)
and finally as
R
WLR=
H-1( 11 ) °11 o
or (2o78b)
with the Bryant angles c/J, 0, and 'lj;, and with
1 , 0 , sO )
and H- 1 (17) = ( 0, ccjJ, -sc/JocO 0 (2o78c)
O,scjJ, ccjJocO
This yields the well-known kinematic differential equation
(2o79a)
with
co , s cfJ 0so , -c cfJ 0so)

H (11) = c10 °
( o , c c/J 0c o , s c/J 0c o , (2o79b)
0 , -sc/J , ccjJ
and (after some trigonometric manipulations)

cos'lj; - sin'lj;
,0
cos () cos ()
sin 'lj; cos'lj; , 0 (2o79c)
sin () 0 sin ()
-cos'lj; 0 - - sm'lj; 0 - - 1
cos () ' cos () '
for Bryant angles cp, (), and '1/Jo The matrix H(17) is singular for
() = (n + 1) 0
(7r/2) n E ZU {O}o
Taking into account
p = (rT' 17T)T = (x, y, z, cp, (), '1/J)T, (2o80a)

r = (x, y, z)T, 11 = (cp, (), 'lj;)T
and
o (oT
p= r ,11oT)T (2o80b)
provides, tagether with
the kinematic differential equation
v = T- 1 (17) o jJ (2o80c)
with
(2o80d)
Conversely,
p= (:) =T(17) 0
(:;R) =T(17)ov (2o81a)
with
(2081b)
3. Constraint equations and constraint
reaction forces of mechanisms
The constraint equations of mechanisms come in many forms. They may

be kinematic constraint relations due to passive joints, or active constraint
relations that stem from reactions due to active force elements. Constraint
equations are used to solve quite different tasks in mechanisms and mecha-
tronic systems.
Kinematic constraint equations may serve:
1. As theoretical models and numerical algorithms of joints.
2. As projection operators of forces.
3. As nonlinear and linear prefilters and kinematic decoupling controllers of
mechanisms and mechatronic systems.
4. As algorithms used for transformation of signals and for image processing.
Active constraint equations may be used:
1. As theoretical and numerical models of active drives.
2. As algorithms for computing desired forces acting on a mechanism.
3. As models of disturbances acting on a mechanism (compare Examples 3. 7
and 7.1 of Volume 11).
4. As algorithms for disturbance rejection in control systems.
In this chapter general representations of position, velocity and accelera-
tion equations of planar and spatial kinematic and active constraints will be
presented together with a discussion of the singular behavior of constraint
equations ( lock-ups and bifurcation points of a mechanism, Beetions 3.1.1.2
and 3.1. 3). Reaction forces and torques generated by these constraints will
be discussed in Beetion 3.2.
3.1 Kinematics of planar and spatial rigid-body systems

In this section basic relations of holonomic kinematic and active constraints
will be briefly discussed both for planar mechanisms (Bection 3.1.1), and for
spatial mechanisms (Bection 3.1.2). The singular behavior of constraints will
be extensively discussed by means of an example in Beetion 3.1.3.
3.1.1 Kinematics of planar mechanisms
Kinematics is the study of motion (position, velocity, acceleration) of mecha-

nisms without considering the forces and torques that generate the motion. A
76 3. Constraint equations and constraint reaction forces of mechanisms
mechanism or configuration is considered as an assembly of rigid bodies and

links that are arranged and connected by joints to allow absolute and rela-
tive motion. Joints impose constraints on the absolute and relative motion
of the bodies. They will be theoretically modeled by nc kinematic constraint
equations
(3.1)
that reflect the geometry of the joint. In the planar case, p (p E ffi.3 nb) is the
vector of generalized coordinates of the mechanism. Each scalar constraint
relation reduces the number of DOFs of a mechanism by one. Kinematic con-
straint relations of the form (3.1) do not depend explicitly on time but only
on the generalized coordinates. They are called holonomic constraint equa-
tions. Holonomic constraint equations may also depend explicitly on time and
on the velocity vector p if p is integrable in closed form. If the nc kinematic
constraint relations (3.1) are consistent and independent, a system of nb rigid
bodies under planar motion has 3nb- nc DOFs. Then either (3nb- nc) addi-
tional active ( driving) constraint relations are needed ( kinematic analysis),
or external forces together with initial conditions ( dynamic analysis) are
used to uniquely determine the motion p(t) of a mechanism.
Constraints imposed on a system by a drive are described by active or
driving constraint relations. Active constraint relations model the time his-
tories of some coordinates produced by ideal actuators. They usually depend
explicitly on time and on some generalized coordinates. They are written in
the form
ga (p(t), t) := 0
More general constraint relations that include inequalities or velocity com-

ponents that are not integrable in closed form are called nonholonomic con-
straint relations.
Here holonomic kinematic constraint relations will be considered for bod-
ies: (1) that undergo motions in a single plane or in parallel planes, and
(2) that are connected by joints imposing constraints on the motion of these
bodies.
These constraint elements may be joints, massless links, spur gears, or
driving links that connect two different bodies to each other, or a singe body
to its base. To specify a configuration (the position and orientation or location
of each rigid body of the configuration), an inertial (global) frame R with
global Cartesian coordinates xR and yR, and a body-fixed (local) frame Li
for each rigid body i with local Cartesian coordinates xL; , yL; and with
'l/JL;R =: 'l/Ji as the angle of rotation of Li with respect to R are introduced
(Figure 3.1).
The location of the body i in the plane is then specified by the global
coordinates of the body-fixed origin Üi of the frame Li with respect to the
frame R:
301 Kinematics of planar and spatial rigid-body systems 77
frame Li
fixed on body i
inertial
frame R
eyR
OR=O
--~~~~----------------------------------~
Figo 301: Global and local frames in the plane
R
roio =
( R R
xaio' Yoio
)T 0
(3o2a)
A point Pi fixed on this body is specified by constant local coordinates with

respect to the frame Li:
rLi
PiOi
= ( x~:oi)
Li
(constant local vector)o (3o2b)
Ypioi
Alternatively, the point Pi can also be specified by global coordinates as
(3o3a)
or written in expanded form
or as
R L· Li •, Li ,
Xpio = xa:o + Xpioi 0
1
cos 'Pi- Ypioi 0
o 1
sm 'Pi
•
(3o3b)
R L·
Ypio = Yo:o + Li
Xpioi 0 sm
o .!, Li
'Pi+ Ypioi 0
.!,
cos 'Pi
with
'1/Ji := '1/JLiR
as the rotation ofthe body (frame Li) with respect to frame R. The geometric
vector roio together with the angle '1/Ji that specify the location of the body
i in the inertial frame R will henceforth be written in the form of an algebraic
vector of Cartesian coordinates
R R
Pi := ( Pxi, Pyi, '1/Ji
)T := (roio
R T
, '1/JLiR
)T = (Xoio,
R R
Yoio, '1/Ji
)T · (3.4a)
This vector includes two position coordinates and a rotation coordinate. It

uniquely describes the position and orientation (location) of a rigid body in
the xR-yR plane.
Then a planar mechanism including nb rigid bodies is specified by a global
algebraic vector of 3nb generalized Cartesian coordinates
(3.4b)
Based on these notations, holonomic constraint relations of planar configu-

rations that include nb bodies are written in the form:
gk : JR3nb ---+ ]Rnc )

I
U)
U) ~ nc kinematic constraint equations (3.5a)
p f-----t gk (p) = 0 J
and
ga : JR3nb X JRl ---+ JR3nb -nc )
I
3nb - nc active or driving
~ (3.5b)
U) U)
constraint relations.
(p t) f-----t ga (p, t) = 0 J
'
In pure kinematic analysis of a mechanism with (3nb - nc) DOFs exactly
(3nb - nc) active constraint relations are needed to completely specify the
motion of the system. In dynamic analysis, in general, there are no driver
equations to be specified. For 3nb > nc there are (3nb - nc) more unknowns
in the constraint equations then there are equations. As a consequence, there
is no unique solution to these equations. A unique solution is obtained here
by specifying a proper set of initial conditions and by solving the kinematic
constraint equations simultaneously with the kinetic equations, either as a
system of differential algebraic equations (DAEs), or as a system of DEs, after
having eliminated nc dependent coordinates and the Lagrange multipliers. In
dynamic analysis, the motion of the system is completely specified by the
external forces and torques, by the nc kinematic constraint equations, and by
the initial conditions of p.
3ol Kinematics of planar and spatial rigid-body systems 79
3.1.1.1 Pure kinematic analysis of planar mechanisms. In kinematic

analysis, position constraint equations and the associated velocity and ac-
celeration equations are analyzed and solvedo They will be written in the
following form (Appendix A0106):
Kinematic constraint equations
Constraint position equation
(3o6a)
Constraint velocity equation
(3o6b)
Constraint acceleration equation
or
g; (p) p = - ( g; (p) jJ) p jJ =: ß~ (p ' jJ)
0 0 0 0 (3o6c)
Active constraint equations
Constraint position equation
(3o7a)
Constraint velocity equation
ga = g~(p) ojJ + g~(p) := Oo (3o7b)
Constraint acceleration equation
or
g~(p) p = - (g~(p) iJ)p jJ- 2 g~t(P) jJ- g~t(P)

0 0
0
° 0
=: ß~ (p, p, t) 0
(3o7c)
Complete set of kinematic and active constraint equations
Combining the two sets of kinematic and active constraint equations yields
the total constraint position equation
80 30 Constraint equations and constraint reaction forces of mechanisms
t - ( gk (p) ) - 0 g(p, t) E ffi.3nb, (3o8a)

g(p, ) - ga(p, t) - '
and, together with
iJ(p t) = ( gk(p) ) = ( g~(p) ) 0 p+ ( 0 ) = 0

' iJa(p, t) g~(p, t) g~(p, t) '
the constraint velocity equation
Yp(P, t) 0
P = -gt(P, t) =: ac(P, t) (3o8b)
with
t g~(p) ) and Yt(P, t) := ( 0 ) =: -ac(P, t),

g~(p' t)
o- (
Yp(P, ) o-
g~(p' t)
and from
jj(p' t) =
+ ( 0 ) + ( 0 ) ::::0
2og~t(p,t)op Y~t(p,t)
the constraint acceleration equation
Yp(P, t) 0 P = - [Yp(P, t) 0 P]p 0 P- 2 Ypt(P, t) 0 P- Yu(P, t)

° (3o8c)
=: ßc(P, p, t)
with
Ypt(P, t) := ( Y~t(:' t)) , Yu(P, t) := ( Y~t(:, t)) '

and
( Yp ( p, t) 0p
0) 0
P 0 p :=
[( g~ (p)
g~(p' t)
) 0]
0p
0
p 0 po
In the previous relations the following notations have been used:

3.1 Kinematics of planar and spatial rigid-body systems 81
3.1.1.2 Regular and singular planar kinematics. Consider the nc kine-

matic constraint position equations
(3.9a)
Consider the partitioning of p into
(3.9b)
with
u E ocnc as a vector of dependent Coordinates, and (3.9c)
w E oc3 nb -nc as a vector of independent Coordinates, (3.9d)
sometimes called the driving coordinates in kinematic analysis. The objec-

tive of kinematic position analysis is to solve the nc kinematic constraint
equations
g(p) = g(u, w) = 0 (3.10)
with respect to the vector u for a given vector w of driving coordinates. This
system of algebraic equations is usually highly nonlinear in the coordinates
of u. As a consequence, usually it must be solved numerically.
3.1.1.2.1 Regular constraint Jacobian matrix. A regular (nonsingular) con-
straint Jacobian matrix
Yu(uo' Wo) E ocnc,nc (3.11a)

with
det(gu(uo, wo)) -:j:. 0 or rank(gu(uo, wo)) = nc (3.11b)
guarantees, due to the implicit function theorem ([46], [47]):

1. The existence of a unique local solution,
u = u(w(t)), (3.12a)
of the constraint position equation (3.9a) in (u 0 , w 0 ).

20 The existence of a unique local solution,
(3ol2b)
of the constraint velocity equation
Yu U + Yw W = Oo
0 0 (3ol2c)
30 The existence of a unique local solution,
ü = (gu)- 1 0 [- (gw 0w)- (gu 0u)w 0w- (gw 0w)u 0u (3ol2d)
- (g w 0w) w 0w - (g u 0u) u 0u] '
of the constraint acceleration equation
Yu 0ü = - (gu 0u)u 0u- (gu 0u)w 0w- (gw 0w)u 0u

- (Yw 0w)w 0w- Yw 0w
or (3ol2e)
Yu ü =ßc(u, it, w, w, w)o
0
This result already indicates that the constraint Jacobian matrix Yu(u) in-
cludes useful information about the theoretical and numerical (local) solv-
ability of the model equations (cfo Beetions 30103 and 404)0
301010202 Singular constraint Jacobian matrixo Constraints with a singular

constraint Jacobian matrix with
det(gu(uo, wo)) = 0 (3ol3a)
describe singular configurationso They may provide pathological kinematic be-

havior of the mechanism and may lead to computational trouble in kinematic
and dynamic analysiso In those situations the constraint position equations
no Ionger satisfy the implicit function theorem locally, or even globallyo Then
the constraint position equations are not independent but redundant, or even
not compatible with each other. Consider, for example, the constraint accel-
eration equation
Yu(uo' wo) ü = ßc(uo' Uo' Wo' Wo' wo)

0 (3ol2e)
as a linear equation in Üo Then a singular constraint Jacobian Yu(u 0 , w 0 )

may have two consequences:
1. For (3ol3a) and

rank(gu(uo, wo)) 'I rank( Yu(uo, wo), (3ol3b)
ßc(uo, Üo, (wo, Wo, (~ 0 )),

(3.12e) does not have a local or global solution ü(t) (loss of existence of
solutions). This behavior of a mechanism is often called lock-up. In such
a singular point (u 0 , w 0 ) the mechanism can either not be moved by a
chosen driving mechanism, or it cannot even be assembled in this form.
2. For (3.13a) and for

rank(gu(uo, wo))= rank( Yu(uo, wo), (3.13c)
ßc(uo' Uo' Wo' Wo' wo))< nc

(3.12e) does not have a unique local solution ü(t) (loss of uniqueness of
solutions). Then (3.12e) may have a finite number of different solutions
(which may branch or bifurcate, trifurcate etc., from a chosen solution),
called bifurcation solutions, or it may even have an infinite number of
different solutions.
A refined analysis of the singular behavior of the constraint equations may be
obtained from the Weierstraß preparation theorem for analytical constraint
equations ([46], [47]), from the Malgrange preparation theorem for C00 con-
straint equations ([48]), and from Fredholm 's alternative theorem for con-
straint equations of quite generalform ([46]). Bifurcation properties are often
discussed in the framework of Newton diagram techniques ([49]), ([50]) or by
an approach of Ljapunow-Schmidt and others ([51], [52]). Parametrie classi-
fication schemes of singularities have been derived in an approach of Thom
([53], [54]). To convince the reader that singularities of constraint equations
are not only of theoretical interest and do not only appear in pathological
mechanisms, an example of the singular behavior of a simple and often dis-
cussed mechanism will be analyzed in Beetion 3.1.3 (cf. [40], [41],[55]).
3.1.1.3 Kinematics in planar dynamic analysis. Consider nc kinematic
and active constraint equations in the dynamic analysis of a planar mecha-
nism. Then
(3.14a)
A regular constraint situation, necessary for providing a unique solution of
the dynamic model equations, is obtained for
rank (gp(p 0 )) = nc (gp(p0 )has maximum rank). (3.14b)
Singular behavior of the mechanism is obtained for
rank (gp(p 0 )) < nc. (3.14c)
The discussion of the singular behavior in pure kinematic analysis of Sec-

tion 3.1.1.2 still holds for this situation by assuming 3nb- nc (independent)
coordinates to be dynamically predetermined. In order to achieve unique so-
lutions of the dynamic model equations in all 3nb coordinates, (3.14b) must
be satisfied (cf. Chapter 4-4), and proper initial conditions must be chosen.
3.1.2 Kinematics of spatial mechanisms
Spatial mechanisms including nb rigid bodies are specified by a Cartesian

coordinate vector
p = (PlT' 000 'PnbT)T Wlth

o
Pi=
( R R
Xoio' Yoio' Zoio'
R
'Pi' ei' '1/Ji
)T
or (3ol5a)
If this mechanism includes constraints modeled by nc consistent and inde-

pendent holonomic constraint equations
nc kinematic constraint equations, (3ol5b)
it has (6nb - nc) DOFso Then either (6nb - nc) additional active ( driving)
conditions
ga : ffi.6nb X ffi.l --+ ffi_6nb -nc 1
~ ~ ~ 6nb - nc active or driving
constraint equations,
(3ol5c)
(p , t) f---t ga(p, t) = 0 J
are needed (in kinematic analysis) or external forces and initial conditions of
(6nb- nc) coordinates are used (in dynamic analysis) to uniquely determine
the motion p(t) of the mechanismo
3.1.2.1 Pure kinematic analysis of spatial mechanisms. The kine-

matic and active constraint equations (3010), (3ol2c) and (3ol2e) of pure
kinematic analysis of planar mechanisms similarly hold for spatial mecha-
nisms with modified dimensions of the functions and variables, and with jJ
replaced by v according to the kinematic differential equation (2o8la):
(p ' t) f---t g(p' t) = 0,
_ (9It, ooo, 96nbt )T E llll6nb

Yt - lß ,
_
9it -
0 10r
c z - 1 , ooo , nc,
0
_
(3ol5d)
Ytt -_ (91tt, ooo, 96nbtt )T E llll6nb 9itt -_ 0 10r

c z = 1 , ooo , nc,
0
lß ,
and
In order to describe the complete set of model equations of a mechanism

in consistent coordinates, the kinematic differential equation (cfo Equation
2o8la)
p=T(p)ov (3ol6a)
must be included in the constraint equations of spatial mechanisms with
p = (PlT '0 00' Pnb

T )T
'Pi=
T T)T
( roio,'11i ' (3ol6b)
'Pi, Bi, '1/Ji as Bryant angles of the body i, measured from frame R to frame
Li,
po -_ (poT oT ) T E ~6nb
1 , ooo, Pnb
o
Pi=
( oT
rQiO' '11i
oT)T
E ~ '
6
(3ol6c)
oR ( oR oR
rQio = Xoio' Yoio' zoio
o )T R (
'11i = 'Pi , ei , '1/Ji
0 0 o o )T ,
Vi = ( (r~io)T, (wf:R)T) T Li _
W LiR -
( Li Li Li
WxLiR' WYLiR' WzLiR
)T
'
with (20 79a)

R
= H i ('11i ) ARL. ( '11i ) wL;R,
L.
0
'11i 0
'
0
with (cfo Equation 20 79b)
cB·'t ' s"'·osB·

'P'l -c"'·osB·)
'P'l 'l.
'l, '
( 0 ,c</Ji 0
cei' s</Ji 0
cei ' (3ol6d)
0 ' - s rPi ' c rPi
with (2056)
ARLi(1Ji) :=
cei 0c'I/Ji ' -cei 0s'I/Ji ' sei )

( c cl>i 0s '1/Ji + s cl>i 0s (}i 0c '1/Ji , c cl>i 0c '1/Ji - s cl>i 0s (}i 0s '1/Ji , -s cl>i 0c (}i
s cl>i 0s '1/Ji - c cl>i 0s (}i 0c '1/Ji ' s cl>i 0c '1/Ji + c cl>i 0s (}i 0s '1/Ji ' c cl>i 0c (}i
(3ol6e)
with (2o8la)
(3ol6f)
and
(3ol6g)
This yields the constraint position equations
g(p, t) = 0, (3ol7a)
the constraint velocity equations
Yp(P, t) 0P = Yp(P, t) 0T(p) 0v = -gt(P, t) =: a.c(P, t) (3ol7b)
and the constraint acceleration equations
gp(P, t) 0T(p) 0v =- [Yp(P, t) 0T(p) 0v(p)] p 0T(p) 0v (3ol7c)
- 2 ° Ypt(P, t) T(p)
0 0
V- Yu(P, t) =: ßc(P, V, t)
(3ol7d)
with Tt(P) := Oo
3.1.2.2 Kinematics in spatial dynamic analysis. In the dynamic anal-

ysis of mechanisms including nc constraints, it is assumed that
g : ]R6nb -----+ ]Rnc nc < 6nb, (3ol8a)
and
Yp(Po, to) T(po) E
0 JRnc ,Bnb o (3ol8b)
Then the mechanism is called regular if
(3ol8c)
Based on the previous discussion, kinematic constraint equations of planar

and spatial joints will be derived in Beetions 501 and 502, and in various
examples in Volume II.
3.1.3 Singularity analysis of a planar slider-crank mechanism
The slider-crank mechanism of Figure 302 is important from both theoretical

and practical point of viewo It is included in a great variety of mechanisms
from industrial practice and serves in various publications as a simple example
for discussing the pathological or singular behavior of mechanisms ([40], [41],
[55]) 0 In this section the singular behavior of a planar two body slider-erank
mechanism of Figure 302 will be systematically analyzed by means of matrix
calculus, in order to study and illustrate pitfalls that may arise in kinematic
and dynamic analysis even of simple but often-used mechanismso This will
be clone for two different driving funetions of the mechanism (Figure 302):
Gase 1: gf(p, t) = '1/Ji- a1(t) = Oo (3ol9a)
Gase 2: gg(p, t) = x~ 0 + a2(t) = x~Jo + (lj/2) 0 cos'lj;j + a2(t) = 00

(3ol9b)
In agreement with Beetion 301, the planar mechanism including the two
bodies i (crank) and j (slider) is characterized by the generalized coordinate
vector (Figure 302)
"'= i' jo
(3020)
3.1.3.1 Identification of singularities by direct inspection. Most sin-

gularities of the mechanism are seen by direet inspeetion of Figure 3020 The
mechanism driven by a 1 (t) (Figure 3o2b) has
1. For li < lj, no loek-up and no bifureation pointo
20 For li > lj, loek-ups for '1/Jj ± ~7r 0

~
S; l 1 ;2 ~
0 --------------------------------;----- ----
(a) Disassembled form
(b) Ca e 1: Mechanism with driving condition 'f/;; = a1(t)
(c) Case 2: Mechanism wit h driving condition x~ 0 = - a2(t)
Fig. 3.2: Slider- crank mechanism driven by two different active constraints
3. For li = lj , bifurcation points
for ('1/Ji = 1rj2 and '1/Jj = ~1r) and
The mechanism driven by a 2 (t) (Figure 3.2c) has

1. For li < lj, lock-ups
for ('1/Ji = 0 and '1/Ji = 27r) and
for ('1/Ji = 1r and '1/Ji = 21r).

2. For li > lj, lock-ups
for ('1/Ji = 0 and '1/Ji = 21r), and
for ('1/Ji = 0 and '1/Jj = 1r).

3. For li = lj, lock-ups
for ('1/Ji = 0 and '1/Jj = 27r) and
for ('1/Ji = 1r and '1/Jj = 1r).
A more refined singularity analysis of the mechanism can be easily obtained

by a geometrical analysis using the minimal coordinates of the mechanism
([40]). These two approaches can hardly be applied to complex spatial mech-
anisms and cannot automatically be performed by computers. The above
objective may be reached by a formal singularity analysis of the mechanism
based on matrix algebra.
3.1.3.2 Local algebraic singularity analysis of the slider- crank me-

chanism. In a first step the kinematic constraint equations of the mechanism
will be derived. In a second step the complete set of constraint equations of
the mechanism will be analyzed for the two different driver functions selected
(Cases 1 and 2).
Kinematic constraint equations. The geometry of the slider- crank mech-
anism of Figure 3.2 implies the following kinematic constraint equations:
lo The poillt Si Oll body i coillcides with the poillt 0 of frame R (Figure
3o2b)o This implies the commoll-poillt collstraillt equatioll (vector loop
equatioll)
TR - A RL; 0 TL; - 0
PiO P;O- '
alld, together with the geometry relatiolls
RL 0 ( cos '1/Ji , - sill '1/Ji )

A '=
sill '1/Ji , cos '1/Ji
the collstraillt equatiolls
R [i
xP;O- 2 ° cos'l/Ji =0 (3o2la)
alld
R
YPio-
[i
2 0 sm'r'i
.!,o
= O0 (3o2lb)
20 The poillt Sj Oll body j lies Oll the xR-axis (P Sj)o This implies,
together with the vector loop equatioll
R
-TP;O
+ Tpo
R + ARL; 0
L;
TP;P =
0
alld the geometry relatiolls
R _ (X~o)
Tpo-
R _ (
'Tp.p-
-lj/2)
0
A RL.
J=
( cos '1/Ji , - sill '1/Ji )
0 J '
sm .!.
0
'f/j , cos .!.
'f/j '
the relatiolls
or
x~;o- x~ 0 + (lj/2) 0 cos'l/Ji = 0 (3022a)

Y~;O + 0 + (lj/2) 0 sill'l/Jj = Oo (3o22b)
30 The poillt Qi Oll the crallk coillcides with the poillt Qj Oll the slider
(Q = Qi = Qj)o Together with the vector loop equatioll
0
R
TP;O + TQP;
R
+ TP;Q-
R R
TP;O =
alld the geometry relatiolls
alld
this implies the relations
or
x~,o + (li/2) 0 cos'l/Ji + (lj/2) 0 cos'l/Jj- x~;o = 0 (3o23a)
Y~,o + (li/2) 0 sin '1/Ji + (lj /2) 0 sin '1/Jj - Y~;O = Oo (3o23b)
Kinematic constraint position equations. The assembled mechanism

has a single DOFo Collecting (3021), (3022), and (3023) provides the following
five kinematic constraint position equations:
(o, o, o, o, o? =gk(p)
x~,o + 0 - (li/2) 0 cos '1/Ji + 0 + 0 + 0 (3021a)
0 + Y~,o - (li/2) 0 sin '1/Ji + 0 + 0 + 0 (3o21b)
x~,o + 0 + (li/2) 0 cos '1/Ji - x~;O + 0 + (li /2) 0 cos '1/Ji (3o23a)
0 + Y~,o + (li/2) 0 sin '1/Ji + 0 - Y~;O + (li /2) 0 sin '1/Ji (3o23b)
0 + 0 + 0 + 0 +y~.3 0 +(li/2)osin'l/Ji (3022b)
and the following relation for the relative coordinate x~ 0 :
(3o22a)
The complete constraint equations including the kinematic and active con-
straints of Cases 1 and 2 (3019) are
gk(p) )
g,.(p, t) = ( g~(p, t) =
(
0
0) K = 1, 2 o (3024)
Local algebraic analysis of the constraint equations.
Subsequently the constraint position, velocity, and acceleration equations of

the slider-crank mechanism will be analyzed algebraically in two steps: for
the driving function a 1 (t) (Gase 1) and for the driving function a 2 (t) (Gase
2)0
301030201 Local analysis of Gase 1 ('l/Ji(t) = a 1 (t))o Written in components,
the complete constraint position equations are
(0, 0, 0, 0, 0, 0) T = 9t (p, t) (3o25a)

x~,o + 0 - (li/2) cos'l/Ji +
0
0 + 0 + 0 + 0
0 + y~ 0 - (li/2) sin '1/Ji + 0 + 0
0 + 0 + 0
x~,o + 0 + (li/2) cos'l/Ji - x~;O + 0
0 + (lj/2) cos'l/Ji + 0
0
0 + y~ 0 + (li/2) sin'l/Ji + 0
0 - y~;O + (li/2) sin'l/Jj + 0
0
0 + 0 + 0 + 0 +y~J 0 +(li/2)osin'l/Ji+ 0
0 + 0 + '1/Ji + 0 + 0 + 0 - a1(t)
with the constraint Jacobian matrix

1,0, (li/2) sin 'l/Ji , 0 0 0 0
0 ' 1 ' - (li/2) cos 'l/Ji ' 0 0 0 0
1,0, - (li/2) sin'l/Ji , -1 0 - (lj /2) sin '1/Ji
' '
0 0
9tp(P) =
0,1, (li/2) 0
cos 'l/Ji ' 0 -1 (lj /2) cos '1/Jj
0
0,0, 0 0 1 (lj /2) cos '1/Jj

0
0,0, 1 0+ 0+ 0
(3o25b)
the constraint velocity equations
9 1 p(P, t) op= -glt(p, t) with 9tt(P, t) = -(0, 0, 0, 0, 0, ät(t))T

(3o25c)
and the constraint acceleration equations
9tp(P, t) 0
P = - (9tp(P, t) 0
P) p P- 2
0
° 9tpt(P, t) 0
P- 9ttt(P, t) (3o25d)
=: ßc1 (p' P' t)

with
-gltt := +(0, 0, 0, 0, 0, ät(t))T 9tpt(P, t) =0 E ffi.6 ,
9tp(P, t) 0
P= (3o25e)
oR
xp,o + 'l/Ji (li/2) sin'l/Ji
0
0 0
OR
Ypio - 'l/Ji (li/2) cos'l/Ji
0
0 0
oR R
Xpio - 'l/Ji (li/2) sin'l/Ji - 'l/Ji (li/2) sin'l/Ji
0 0
0 0
xP;O - 0 0
y~ 0 + -J;i 0
(li/2) 0
cos '1/Ji - Y~;o + -J;i 0
(li /2) cos '1/Ji
0
oR
+ 'l/Jj (lj/2) COS'ljJj
0
YPOOJ
0
°
-J;i
OR
0 ' 0 ' (li/2) 0~i 0 cos '1/Ji ' 0 ' 0 ' 0 xp,o
OR
0 ,0 , (li/2) 0~i 0 sin '1/Ji , 0 , 0 , 0 YP;O
0, 0, -(li/2)o~io cos'I/Ji, 0, 0, -(lj/2)o~jo cos'lj;j ~i

oR '
0 , 0 , - (li/2) 0~i 0 sin '1/Ji , 0 , 0 , - (lj /2) 0~j 0 sin '1/Ji XP;O
OR
0 ,0 , 0 , 0 , 0 , - (lj /2) 0~i 0 sin '1/Ji YP;O
0,0, 0 ,0,0, 0 ~j
and
02
- (li/2) 0'1/Ji 0 cos '1/Ji
- (li/2) 0~T 0 sin '1/Ji
2
+ (li/2) 0'1/Ji 0 cos '1/Ji + (lj /2) 0'1°2
/Jj 0 cos '1/Jj
0
0 (3o25g)
+ (li/2) 0~T 0 sin '1/Ji + (lj/2) 0~J 0 sin '1/Ji
02
+ (lj /2) 0'1/Jj 0 sin '1/Ji
+ äl (t)
Local singularity analysis of the constraint Jacobian matrix. From
(3o25b) it follows directly that
(3o26a)
For lj > 0 singular situations (Figure 303) occur for
det(g 1p(P)) = 0 +------+ cos'lj;j = 0 +------+ '1/Ji = (2n + 1) OK/2 =: '1/Jj,

(3o26b)
n E ZU {O}o
Then
rank(g 1p(P)) I =5° (3o26c)

'1/Jj = (2n + 1) 1r /2
0
Starting from the initial condition
the following situations occur (compare the different cases of Figure 303):
Gase 101: li < lj

Gase 102: li > lj (Gase 102a)
and
'1/Jj
5
= 27r ' '1/Ji = -arcsin t[o
0
(Gase 102b)
3
Gase 103: '1/Jj = 27r '1/Ji=n/2 (Gase 103a)
and
5 3
'1/Jj = 27r ' '1/Ji = 27r 0
(Gase 1o3b)
The question that remains to be answered is which of the above identified

singular situations are associated with lock-ups, and which with bifurcation
points of motions of the slider-crank mechanismo Answers to this question
will be obtained by the subsequent analysis of the constraint velocity and
constraint acceleration equationso
Analysis of the constraint velocity equations. The constraint velocity
equation (3o25c) is a linear equation in jJ:
oR
1 ' 0 ' (l;/2) sin '1/J; , 0 0
0 0 xp,o 0
OR
0 ' 1 ' - (l;/2) cos '1/J; ' 0
0 0 0 Yp,o 0
1 ' 0 ' - (l;/2) sin'lj;; , -1 0
'
0
'
- {lj /2) sin '1/Ji 0
"j;i 0
OR
0' 1' (l;/2) cos '1/J; '
0
0 -1 (lj /2) cos '1/Jj 0
XP;O 0
oR
0,0, 0 0 1 (lj /2) cos '1/Jj 0
YP;O 0
0,0, 1 0+ 0+ 0 "j;j a1 (t)
(3o26d)
In the regular case (det(g 1p(P)) =I 0 or for cos'lj;j =I 0), (3o26d) has the
unique finite solution
al (t) 0
li sin '1/Ji
0
OR 2
XP;O a1 (t) zi cos '1/Ji 0 0
OR
YP;O 2
-a1(t)
?j;i
p= OR al (t) 0
li 0
( cos '1/Ji sin '1/J j - 2 ° sin '1/Ji
0 0
cos '1/J j) 0 (3o26e)
XP;O
2° COS'lj;j
OR
YP;O a1 (t) zi cos '1/Ji 0 0
2
a1 (t) zi cos '1/Ji
0 0
lj cos'l/Jj
0
This solution tends to infinity for cos '1/Ji = 0 (singular case) unless additional
conditions are satisfiedo These singular situations will be discussed belowo
Case 1.1: 0 < l; < lj
no singular si uation
occurs for ~ 7r < 1jJ1 < ~ 7r and
for arbitrary values of 1/J; = a1 (t )
Ca. e 1.2: l; > 11 >0

singular situations occur for
Case 1.2a: 1/Ji = ~7r a.nd 1/J.; = arcsin ~
a.nd
Case 1.2b: 1/J) = ~ 7r and 1/J; =- arcsin f
Case 1.3: I; = li >0

1.3a singular ituations occur for
Case 1.3a: 1/Ji = ~7r and 1/J; = ~

and
Case 1.3b: 1/Ji = ~ 7r and 1/J.; = ~ 7r
Fig. 3.3: Singularsituations of Case 1 (1/Ji (t) = a1(t) for 1/Ji = ~7r or 1/Ji = ~1r)
In the singular case (det(g 1 p(O)) = 0 or cos'lj!j = 0) , (3.26d) does not have a
finite solution p( t) for
rank( (g 1 P, -glt) I ) >rank( (g 1p) I ) = 5. (3.26f)

7/>j =(2n+1)·7r/2 7/>j =(2n+1) · 7r/2
On the other hand, the relation
(3.26g)
with the augmented matrix

(Ylp' -glt) I = (3o26h)

7/Jj
1 , 0 , (li/2) 0
sin 7/Ji , 0 , 0 , 0 0
0 ' 1 ' - (li/2) 0
cos '1/Ji ' 0 ' 0 ' 0 0
1 , 0 , - (li/2) 0
sin '1/Ji , -1 , 0 , - (lj /2) 0
sin '1/Jj , 0
0 , 1 , (li/2) 0
cos'l/Ji , 0 , -1 , (lj/2) 0
cos'l/Jj , 0
0' 0 ' 0 ' 0 ' 1 ' (lj /2) cos '1/Jj ' 0
0
0' 0 ' 1 '0, 0, 0 ,il!(t)

7/Jj = 7/Jj
1 , 0 , (li/2) 0
sin '1/Ji , 0 , 0 , 0 0
0 ' 1 ' - (li/2) 0
cos '1/Ji ' 0 ' 0 ' 0 0
1 , 0 , - (li/2) 0
sin '1/Ji , -1 , 0 , - (lj /2) 0
sin '1/Jj , 0
0 ' 1 ' (li/2) 0

cos '1/Ji ' 0' -1' 0 0
0 ' 0 ' 0, 0, 1 ' 0 0
0 ' 0 ' 1 0, 0, 0 ' äl (t)

' 7/;j = 7/Jj
with
rank((g 1P, -glt)) =rank( (g 1P, -glt) (g 1P, -glt)T), 0 (3o26i)
and with
provides the relation
det ( (g 1P , -glt) 0
( _:r,)) I (3o26j)
7/Jj
- [(4oäi(t)olTolJ- cos'l/Jio sin'l/Jio cos'l/Jjo sin'l/Jj
+ (4oai(t) ofT olJ- cos'l/Ji+ (-5oai(t) olT-4oai(t) -4) oz;) 0

cos 2 '1/Jj
+ (-4oai(t) ofT oz; -4oai(t) ozT) 0

cos 2 '1/Ji)/4] I
7/Jj
= ~ 0
ai lT (ZJ + 4) cos 2 '1/Ji = Oo
0 0 0
Then (3o26d) has ( at least) two finite solutions jJ for

~ 0z; 0(ZJ + 4) 0 cos 2 '1/Ji 0ai = 0 , li >0 , li >0 and for cos '1/Jj = 0 0
(3o26k)
This equation is either satisfied for
and cos '1/Ji = 0 (ioeo, for '1/Ji = +1r /2 or '1/Ji = -1f /2) (30261)
or for
a1(t) =0 and cos'l/Ji-::/:- 0 (3o26m)
or for
a1 (t) =0 tagether with cos'l/Ji = 00 (3o26n)
Comment 3.1.1 ( Constraint velocity analysis): As will be shown in
the subsequent constraint acceleration analysis, the relation (30261) is asso-
ciated with Gases 103 and with the bifurcation points, whereas (3o26m) is
associated with Gases 102 and with lock-up points, and sometimes also with
some bifurcations (Figure 3.4)0
Analysis of the constraint acceleration equations. In the regular case
(for cos '1/Jj -::/:- 0), the constraint acceleration equation
Ylp 0p = ßcl with p = (x~io' ii~io' ;j;i' x~;O' ii~;O' ;fj) T (3o27a)
has the unique finite solution

p= (3o27b)
- ~ 0 ( ä 1 ( t) 0 sin '1/Ji + 1}; 0 cos '1/Ji)

- ~ 0 (1}; 0 sin'l/Ji -ä 1 (t) 0 cos'l/Ji)
äl (t)
- 2 1 'ljJ {-
0 cos j
(1}7 olj 0 sin'l/Ji- äl(t) oli 0 cos'l/Ji) 0 sin'l/Ji
+ (2 äl(t) oli 0 sin'l/Ji + 2 °1}; olj 0 cos'l/Ji) cos'l/Jj + 1}; olj}
°
- il· 0 ('1/Jt 0 sin'l/Ji- ä (t) 0 cos'l/Ji

0
1
)
1}; olj 0 sin'l/Ji + li 0 (1}7 ° sin'l/Ji- ä 1 (t) 0 cos'l/Ji)

lj o COS'l/Jj
In the singular case (for cos '1/Jj = 0), this solution tends to infinity unless
additional conditions are satisfiedo For
rank((g 1P, ßcJ) I > rank(g 1p) I = 5 (3o28a)

'lj;j 'lj;j
or for
(3o28b)
with the augmented matrix
(glp' ßc,) =
1,0, (l;/2) · sin 1/Ji , 0 , 0, 0 , - (l;/2). ,J;;. cos'lj;i
0 , 1 , - (l;/2) cos 1/Ji , 0 , 0
0, 0 , - (l;/2) · .f'f · sin 1/Ji
1, 0,- (l;/2) · sin'lj;i, -1, 0,- (!1/2) · sin'lj;1 , + (l;/2) · .f'f · cos'lj;i + (!1/2) · ,P] · cos'lj;1
0,1, (l;/2) · cos 1/Ji , 0 , -1 , (l1 /2) · cos 'lj;1 , + (l;/2) · .f'f · sin 1/Ji + (l1/2) · ,P] · sin 'lj;1
0,0, 0 , 0, 1, (l1 /2) · cos 'lj;1 , + (l1/2) · ,PJ- sin 'lj;1
0,0, 1 , 0, 0, 0 ä,(t)
(3.27a} does not have a finite solution. For
det ( (glp, ßcJ · ( ;! )) I =
1/JJ
- [((((4·?j;t-4·äi(t)) ·li·l}- cos'I/Ji· sin'I/Ji (3.29)
-8. ä 1 (t). ?j;i. zT ·lJ. cos 2 '1/Ji + 4. ä 1 (t). ?j;i. zT ·l;). cos'lj;j
+ ( 2 . 'ljJ i . 'ljJ j . li . l j + 8 . 'ljJ i . 'ljJ j . li . l j . sin '1/Ji
'2 '2 3 '2 '2 )
+ (- 2 ät(t) ?j;J-li ·l]- 8 ät(t) ?j;J-li

° 0
° 0
.zn ° cos'I/Ji) 0
sin'lj;j
+ (8 ät (t) ?j;; z; z; cos '1/Ji sin '1/Ji + (4
° 0 0 0 0 0
° ?j;t - 4 ° äi (t)) z; z;0 0 0
cos 2 '1/Ji
+ ( (?j;t + 5 · äi(t)) ·lT- 4 · ?j;j + 4 · äi(t) + 4) ·l;) · cos2 '1/Ji

+ ( 4 · ä1 (t) · 'lj;j'2 · li · lj3 · sin '1/Ji + 4 · 'lj;i'2 · 'lj;j'2 · li · lj3 · cos '1/Ji ) · cos '1/Ji
+ ( - 2 ät (t) ?j;i z; z; - 8 ät (t) ?j;T- l;) cos '1/Ji sin '1/Ji
° 0 0 0
° 0 0 0
+ ( (äi (t) - ?j;t) . z; . z; + (4 . äi (t) - 4 . ?j;t) . z;) . cos 2 '1/Ji + ?j;j . lj
+ (
'1/Ji4 li2
0
0
+4 ° '1/Ji '1/Jj lj2
° 4 ° 4) 0
+4 ° '1/Ji
°4 0
li2) 4 I ]I = 0,
1/JJ
the constraint acceleration equation (3.27a) has at least two finite solutions
Pi· This will now be analyzed for the different singular situations of Figure
3.4.
In Gase 1.3a, (3.29) implies, together with
li = lj ' 'lj;*J = ~7r

2
and for
7r
'1/Ji = - and a1 -:j:. 0, (3.261)
2
the relation
det ( (Ylp, ßc,) · (Ylp , ßc,) T) I
1Pj = ~Ir
= (tPj. + tPi· )2 . (tPj. - tPi. )2 . lj2. (2lj + 4)/ 4
= (tPj. + a1. )2 · (tPj. - a1. )2 · lj2· (lj2+ 4) I 4 = 0
iff
(Gase 1. 3a1) (3.30a)

or ( velocity analysis criteria)
(Gase 1.3a2). (3.30b)
From this singular configuration (Gase 1.3a), four different motions may bi-
furcate (bifurcation point):
and (3.31)
tpj
tpj
=
=
-(h(t) > 0,
-a 1 (t) < O. (Gase 1.3a2)
In Gase 1.3b, (3.29) implies, together with

_ASt ~-
and for
3
al -: /:- 0
7r
tPi = -Ir = -- and (3.261)
2 2
the relation
iff
~j (t) = 1h(t) f:. 0 (Gase 1.3b1} (3.32a)
or (veZocity anaZysis criteria)
~j (t) = - ä 1 (t) f:. 0 (Gase 1. 3b2}. (3.32b)
From this singuZar conjiguration (Gase 1. 3b}, again four different motions
may bifurcate (bifurcation point):
~j = äl(t) > 0,
~j = ä1 (t) < O, (Gase 1. 3b1}
and (3.33)
~j = - äl(t) > 0,
~j = - ä 1 (t) < O. (Gase 1. 3b2}
In Gase 1. 2a, (3.29) implies, together with
Zi > Zj , '1/Jj = ~7r , sin'I/Ji = Zj/Zi and cos'I/Ji = Jl- (lj/Zi)2 ,

the relation
det ( (glp , ßcJ · (glp , ßc, ?) I (3.34a)
'1/J~J = ~7r
2
(z; + 4) .[2. ä1 ( t) . zj . Jz; - z; . (~; - ai)
'--v-" "-..-'
>0 >0
+ (~; - aif .z; + äi(t). uJ 14 = o.
>0
The resulting equation,
··2 )
al ( t +
2ZjJ(lr -
(Zr - ZJ)
m. ( "2
'1/Jj - al
•2 )
+
z; ( "2 •2 )
(Zr - Z}) . '1/Jj - al = 0,
is quadratic in äi(t) and has two identical roots

or
..
a l± = -
lj J(l[- ZJ) ( .2
(lT - l}) . 1/Jj - al
·2)
= (lT - m .(•2
ljJ(l[ -lJ)
al
.2)
- 1/Jj . (3.34b)
Together with
ä1 = 0 {velocity analysis) (3.26m)
this implies
.. ljJ(l[ -ZJ) .i,2 0

al = - (lT - l}) . '+'j < . (3.35)
>0
This relation proves that the situation äl >0
~- (3.36)
cannot occur - no such motion is possible. This is a lock-up situation.

On the other hand, the situation
(3.37)
allows two possible motions. They may bifurcate from the singular configura-
tion of Gase 1. 2a:
1/Ji> O , ä1(t)=O, ä1<0,

(Gase 1.2a1) (3.38)
and
1/;j < o , a1 (t) = o , ä1 < o,

{Gase 1.2a2) (3.39)
For Gase 1. 2b, (3.29) implies, together with
li > lj ' 1/Jj = ~Jr ' . ·<pi

Sln '· = -z;l . , and
J cos1j;i = Jl- (l j/ li )
2,
the relation
det ( (Ylp, ßc,) · (Ylp, ßc,) T) I
1/Jj = ~Jr
=- (z; + 4) . [2. ä1 (t) . zj . Jzr -z; . (~; - ai)

'-v--' '-..-'
>0 >0
- (~;-an .z; + äi(t). u; -m J14 = o.
2
'-...-" '-v--'
>0 >0
The resulting quadratic equation,
~
··2() 2l·Jy·i . ( z2J ) ("2 ·2) 2
a1 t + (li - ZJ)
- .j
· ( '1/Jj2 - ·2) ..
a 1 · a1 + Z7 - l] · '1/Jj - a1 ,
has two identical roots
or
.
al± =
ljJ(l7 - I]) ( "2 •2)
u; -l]) . '1/Jj - al .
Together with
(velocity analysis) (3.26m)
this implies
..
zjJz;-z12 •
. ;,2 0
al = + z2 - z2 "'Pj > . (3.40)
t J
'-...-"
>0
This relation proves that the situation
~-
äl < 0 (3.41)
cannot occur - no such motion is possible. This is a lock-up situation.
On the other hand, the situation
(3.42)
allows two possible motions. They may bifurcate from the singular configura-
tion of Gase 1.2b:
'lj;j>O, ä 1(t)=O,ä1>0, (Gase1.2b1) (3.43)
and
'lj;j < 0 , ä 1 (t) = O, ä 1 > 0. (Gase 1.2b2)
äl > 0
~
The above results of the singular situations of the slider- crank mechanism
(3.44)
under the drive a 1 ( t) are collected in Figure 3.4.

3.1. 3.2.2 Local analysis of Gase 2 (x~ 0 = -a 2 (t)). Written in components,
the constraint position equations are
(0, 0, 0, 0, 0, O)T =g2(p, t) = (3.45a)

x~,o + 0 - (l;/2) · cos'lj;i + 0 ++ 0 0 + 0
0 + Y~o - (l;/2) · sin1/Ji + 0 ++ 0 0 + 0
x~,o + 0 + (l;/2) · cos'lj;i - x~;O +
0 + (lj/2) · cos1/;j + 0
0 + y~ 0 + (l;/2) · sin1/Ji + 0 - Y~;O + (lj/2) · sin1/;j + 0
0 + 0 + 0 + 0 + y~ 0 + J
(lj/2) · sin1/;j + 0
0 + 0 + 0 +x~ 0 + 0 +
J
(lj/2)cos 1/;j +a2(t)
1 , 0 , (li/2) · sin 'lj;i , 0 , 0 , 0

0,1,-(li/2)·cos 'lj;i, 0 , 0, 0
1 ' 0 ' - (li/2) · sin 'lj;i , -1 , 0 , - (lj /2) · sin 'lj;j
Y2v(p) = . (3.45b)
0 ' 1 ' (li/2) · cos'lj;i , 0' -1 ' (lj/2) · cos'lj;j
0 '0' 0 0, 1 ' (lj/2) · cos'lj;j
0 '0 ' 0 1 ' 0 , - (lj /2) · sin 'lj;j
The constraint velocity equations are
(3.45c)
with
g2t(p, t) = (0, 0, 0, 0, 0, a2(t))T. (3.45d)
The constraint acceleration equations are
(3.45e)
......
0
singularity li = lj
li > lj """
analysis
(Case 1.2) (Case 1.3)
(Case 1: 1/J; = a 1) <;;:>
singular 1/Jj = !1T 1./Jj = ~ 1r 1/;j = ?1T 'ifli = %7i Q

0
constraint (Case 1.2a) (Case 1.2b) (Case 1.3a) (Case l.3b) 1:1
[/)
J acobian: ...."""
1/;1 = 1/Jj, cos'!f;j = 0 §_'
"'
C1l
'W; = (2n + 1) · ~
[4] [SE] rn rn >§
constraint ~
c;·
velocity 1:1
[/)
a1= o a1 = o 'ljJ·,-
- "2 1/;; = ~1T
analysis:
§
a1 = o 0..
or
cos'if;; f 0 cos 'l/Ji f 0 a1 f. o a1f. o ("')
0
cos'if;; = 0 1:1
[/)
constraint ,j,1 = a1(Gases I.3al) ·0; = a, (Gases 1.3bl) ....""'

l; . <Ll- t]Jl .2 .. - I; (l?- !J) ~ .2 ~
acceleration äl = J?- J2 • .P; a,- j2- F . .P, ~
' 1 ' 1 ....C1l
analysis:
for for ~
Gases 1.3 ä, >0 ä, <0 """
c;·
,P; = al lock-up lock-up § ~~ ,j,1 = a, > o ,j,1 = a, < o
or
OOGU
·01 = a1> o ..P; = a1 < o
or
1:1
and 8'
for
~ for ,P; = -a1 (Gases 1.3a2) ..P; = -a 1 (Gases 1.3b2) ....
("')
-.P; = -a1 C1l
ä1< o,,P; > o ä1 > o,..j,; > o [/)
Gases 1.2a (Gase 1.2al) (Gase 1.2b l ) 0

....,
.. (1/-1') ..j,2 0 or for or for
a1 · j <
= - I, (r;-rj
ä• < o,..j,; < o
~ ät>O,..j,;< O
s
C1l
Gases 1.2b
~ ~ ~ ~~
(Gase 1.2a2) (Gase 1.2b2) a1 = -..j,1 > o a1 = -..j,1 < o ä, = -..j,; > o a, = -..j,; < o &
~(1/;'i> bifurcations
§
äI = + 11 (1;-1 )
.,j,12 > 0 bifurcations ~ bifurcations ~ bifurcations tn•
s
[/)
Fig. 3.4: Singularsitua tions of the slider-cr a nk mech a nism of C a se 1 (drive a1(t))
with the right hand side
ßc 2(P, p, t) = -g2tt(P, t)- 2 Y2pt(p, t) 0P- (g2p(P, t) 0P)p 0p,

° (3.45f)
with
-g2tt(P, t) = (0, 0, 0, 0, 0, ä2(t))T , -g 2pt(P, t) = 0 E lR6 , (3.45g)
Y2p(p, t) P=0 (3.45h)

OR
xp,o + '1/Ji 0 (li/2) 0 sin '1/Ji
0
OR
Yp,o - '1/Ji 0 (li/2) 0 cos'l/Ji
0
x~,O - ~i o (li/2) o sin '1/Ji - x~;O - ~j o (lj/2) o sin '1/Jj

Y~o + ~i 0 (li/2) 0 cos '1/Ji - Y~;o + ~i 0 (li /2) 0 cos '1/Ji
R
+ '1/Ji 0 (lj/2) 0 cos'l/Ji
0
YP·O
'
±~. 0 - ~j 0 (lj /2) 0 sin '1/Ji
'
(Y2p(p,t)op)pop= (3o45i)
OR
0 ' 0 ' (li/2) 0;pi 0 cos '1/Ji ' 0 ' 0 ' 0 xp,o
OR
0 ,0 , (li/2) 0,(;;i 0 sin '1/Ji , 0 , 0 , 0 Yp,o
0 ' 0 ' - (li/2) 0;pi 0 cos '1/Ji ' 0 ' 0 ' - (lj /2) 0~j 0 cos '1/Jj ;pi
OR '
0 , 0 , - (li/2) 0,(pi 0 sin '1/Ji , 0 , 0 , - (lj /2) 0~j 0 sin '1/Ji xP;O
oR
0,0, 0 ,0,0,-(lj/2)o~josin'l/Ji YP;O
0 '0' 0 0 0 - (l·/2) 0.i.. 0 cos·'·· ;pj

' ' ' J 'f'J 'f'J
and
- (li/2) 0~; 0 cos '1/Ji

02
- (li/2) 0'lj!i 0 sin '1/Ji
+ (li/2) 0~; 0 cos '1/Ji + (lj /2) 0~J 0 cos '1/Jj
ßc2(p,p,t)= (3.45j)
+ 2
(li/2) 0'1/Ji 0 sin'l/Ji + °2
(lj/2) 0'lj!j 0 sin'l/Jj
0
+ (lj /2) 0~J 0 sin '1/Ji

- ä2(t) + (lj/2) o~J o COS'lj!j
Local singularity analysis of the constraint Jacobian matrix. From

(3.45b), it follows directly that
(3.45k)
For li 0lj > 0 singular situations occur for

det(g 2 p(P)) = 0 f------+ 7/Ji - 7/Jj = 7/J7 - 7/Jj = n ·Jr , n E ZU {0}. (3.451)
This implies
rank(g 2p(P)) I = 5 .
7/J7 - 7/Jj
Starting from the initial conditions
(3.45m)
the following singular situations may occur for (0 < E < ~):
(3.45n)
0 . 1 . 2
Gase 2. 1: li < lj : '1/;j = 21f 7/Ji = 0 (Gase 2.1a) ~
' a2
1 i 0 2
7/Jj = 2Jr 7/Ji = 1f (Gase 2.1b) ~ e--
j a2
0 . 1 . 2
Gase 2.2: li > lj : '1/;j = 2Jr 7/Ji = 0 (Gase 2.2a) ~
a2
0 i 2 1
'1/;j = 1f '1/Ji = 0 (Gase 2.2b) ___@[ <Ea./''
Gase 2.3: li = lj : 0 . 1 J. 2
2
~
li-
(Gase 2.3a)
i j
0, 2 · a2
(Gase 2.3b)
5
'1/;j = -Jr
2
{Gase 2.3c)
2 j 1 i 0
{Gase 2.3d) ~
i
~ a2
tPj = 27r tPi = 7f (Gase 2.3g} J
'
3 7f
tPj = 27f- c:, tPi = - - c
2
(Gase 2.3h} ~ 2
5 3
tPi = 21r+ c:, tPi = - ] f + c (Gase 2.3i} ~.
e
2
tPi = 7f tPi = 0. (Gase 2.3j)

'
Analysis of the constraint velocity equations. Written in components,
the constraint velocity equations (3.45c) are
·R
1 , 0 , (l;/2) · sin tPi , 0 , 0 , 0 Xpio
·R
0 ' 1 ' - (l;/2) . cos tPi ' 0 ' 0 ' 0 Y pi o
1 ' 0 ' - (li/ 2) · sin 'lj!i, -1 , 0,- (lj/2) · sin 'lj!j ~i

(lj /2) . cos tPi ·R
0 ' 1 ' (l;/2) . cos tPi ' 0 ' -1' Xpjo
0 0 , 1 ' (lj /2) . cos tPi ·R
0 '0 ' Y PjO
0 '0' 0 1 ' 0,- (lj/2) · sin 'lj!j ~j

(3.46)
In the regular case
det (g p(P))
2 = li ·lj · sin('lj!i - tPi ) =f. 0 (3.47)
(3.46) has the unique finite solution
·
P = ( ·R ·R .i, · R ·R .i, ) T
Xpio, YPio , '1-'i, Xpjo, YPjO , '~-'i (3.48)
= [a2(t) . sin('lj!j + tPi )- a2(t). sin('lj!j - tPi )
4·sin('ljJj + tPi ) '
a2(t) · cos('lj!j + tPi ) + a2(t) · cos('lj!j - tPi )
4 · sin('lj!j - 'lj!;)
a2(t) · COS'lj!j a2(t) · sin('lj!j + tPi )- 3 · a2(t) · sin('lj!j - tPi )
li · sin('lj!j - tPi )' 4 · sin('lj!j - tPi )
a2(t) · cos('lj!j + tPi ) + a2(t) · cos('lj!j - tPi ) a2(t) · cos'ljJ" JT
4·sin('ljJj - 'ljJ;) ' lj·sin('ljJJ - 'IjJ" )
For
sin( 1f)i - 1f)j) = 0, (3.49)

this solution tends to infinity (singular case) unless additional conditions are
satisfied. These conditions will now be derived.
Elimination of the coordinates x~;O and x~io from rows 1, 3, and 6 of the
constraint velocity equations (3.46) provides the following expression for the
driving velocity, ä2(t),
ä2(t) = lj · sin 1f)j · -J;j + li · sin 1f)i · -J;i (3.50a)
and its time derivative
ä2 (t) = (lj · cos1f)i ·1f}i'2 + li · cos1f]i ·1f}i'2) (3.50b)
+ (ti · sin 1f)j · {/Jj + li · sin 1f)i · {/Ji) .

These two equations will now be used in the discussion of the singular situ-
ations of Gase 2.
In the singular case (det(g 2p(P)) = 0), (3.46) does not have a finite solution
for
(3.51a)
On the other hand, the relation
rank(g 2P, -g 2 t) I = rank(g 2P) = 5 (3.51b)

'1/Jt-'1/Jj '1/Jt-'1/Jj
implies
det ( (g 2P, -g 2 t) · ( -~~)) I = 0= (3.51c)

1f)7 - 1f)j
- [ ( (4 · ä~(t) + s) ·lJ ·lJ · cos1f]i · sin1f]i · cos1f)i · sin1f]j
+ ( (4. ä~ (t) + 8) . z; .t] . cos 2 1f)i + (( - 5 . ä~ (t) - 4) . z; - 4. ä~ (t) . t]))

. cos 2 1f]j + ( (- ä~(t)- 4) ·lJ ·l]- 4· ä~(t) ·l;) · cos 2 1f]i) / 4] I·
1f)7 - 1f)j
This implies for Gases 2.1a and 2.2a (1f]j = 27r, 1f)i = 0, li < lj, and li > lj)
that
>0
(3.52a)
which enables at least two finite solutions jJ of (3.46)0 For a2 (t) f::- 0, no
motion is possible (lock-up)o For Gase 201b ('lj;j = 2n, '1/Ji = n, and li < lj),
this implies again (3o52a) and enables at least two finite solutions jJ of (3.46)0
For a2 (t) f::- 0, again no motion is possible (lock-up)o For Gase 202b ('1/Jj =
n, '1/Ji = 0, and li > lj), this implies again (3o52a), which enables at least two
finite solutions jJ of (3.46)0 For a2 (t) f::- 0, no motion is possible (lock-up)o For
Gase 203a ('lj;j = 2n, '1/Ji = 0, and li = lj), this implies
det ( (g 2 p, -g 2 t) 0 ( _:;,)) = a~(t) 0 z; 0 (z; + 4) j2 = o (3o52b)

>0
iff a2(t) = o,
motion is possible (lock-up)o For Gase 2o3b ('lj;j = ~1r, '1/Ji = ~' and li = lj),
this implies
det ((g -g
2P, 2 t) 0 ( _:;,)) = 0 (3o52c)
for arbitrary driving velocity a2 (t), which enables at least two finite solutions
jJ of (3046) (bifurcation)o For Gase 203c ('lj;j = ~1r, '1/Ji = ~1r, and li = lj), this
implies
(3o52d)
for arbitrary driving velocity a2 (t), which enables at least two finite solutions
jJ of (3046)0 For Gase 2o3d ('lj;j = n, '1/Ji = n, and li = lj) this implies
det ( (g 2P, -g 2 t) 0 ( _:;,)) = a~(t) 0 z; 0 (z; + 4) j2 (3o52e)

>0
iff a2(t) = o,
motion is possible (lock-up) 0
The subsequent singular Gases 2o3e to 203j can be obtained by additional

control inputs (in addition to the driving function a 2 (t)):
For Gase 203e ('lj;j = ~7r + c:, '1/Ji = ~ + c:, 0 < c: < ~' and li = lj), we obtain
det ( (g 2P, -g 2 t) 0 ( _:;,)) = a~(t) 0 sin 2 c: 0 lJ- (z; + 4) /2 = o

>0
(3o52f)
This enables at least two solutions jJ of (3.46)0 For ä2(t) -::/:- 0, no motion is
possible (lock-up)o For Gase 203/ ('1/Jj = ~7r- c, '1/Ji = ~7r- c, 0 < c < ~' and
li = lj), we obtain
det ( (g 2P, -g 2 t) 0 ( -:~)) = ä~(t) 0 sin 2

c0 tJ- (t] + 4) /2 = 0
>0
(3o52g)
iff ä2(t) = 00
possible (lock-up)o For Gase 203g ('1/Jj = 21r, '1/Ji = 1r, and li = lj), we obtain
det ((g 2P, -g 2 t) 0 ( -:~)) = ä~(t) 0 lJ 0 (t] +4) /2 = 0 (3o52h)

>0
iff ä2(t) = Oo
possible (lock-up)o For Gase 2o3h ('1/Jj = ~7r- c, '1/Ji = ~- c, and li = lj), we
obtain
det ( (g 2P, -g 2 t) 0 ( -:~)) = ä~(t) 0 sin 2 c0 tJ- (t] + 4) /2 = 0

>0
(3o52i)
iff ä2(t) = 00
possible (lock-up)o For Gase 2o3i ('1/Jj = ~7r+c, '1/Ji = ~7r+c, and li = lj), we
obtain
det ( (g 2P, -g 2 t) 0 ( -:~)) = ä~(t) 0 sin 2 c0 tJ- (t] + 4) /2 = 0

>0
(3o52j)
iff ä2(t) = Oo
possible (lock-up)o For Gase 203j ('1/Jj = 1r, '1/Ji = 0, and li = lj), we obtain
det ( (g 2P, -g 2 t) 0 ( -:~)) = ä~(t) 0 t] 0 (t] + 4) /2 = 0 (3o52k)

>0
iff ä2(t) = 00
possible (lock-up) 0
Analysis of the constraint acceleration equations. In the regular case

(for sin( 1/Ji -1/!j) -::/:- 0) the constraint acceleration equation
(3o53a)
has a unique finite solution
oo (ooR ooR ooR ooR oo)T

1/!i, 1/!j
00
P = xpio, YPio, xpjo, YPjO, = (3o53b)

ä 2(t) 0
( sin('lj;j + 'lj;;)- sin('lj;j- 'lj;;))- 2 ° (~tol; sin'lj;j + ~J-lj 0 0
sin1/J;)
4 ° sin('lj;j- 1/J;)
ä 2(t) 0
( cos('lj;j + 'lj;;) + cos('lj;j- 'lj;;))- 2 ° (~tol; 0
cos'lj;j + ~J olj 0
cos1/J;)
4 ° sin('lj;j- 1/J;)
~[ol; 0
cos('lj;j- 1/J;)- ä2(t) cos'lj;j + ~J olj 0
l; sin( 1/Jj - 1/J;)

0
ä 2(t) o ( sin('lj;j + 'lj;;) - 3 o sin( 'lj;j - 'lj;;)) - 2 o ( ~[- l; o sin 1/Ji + ~J olj o sin 'lj;;)
4 ° sin('lj;j- 1/J;)
ä 2( t) 0
( cos( 1/Ji + 'lj;;) + cos( 1/Ji - 'lj;;)) - 2 ° ( ~t l; 0 0
cos 1/Ji + ~J lj 0 0
cos 'lj;;)
4 ° sin('lj;j- 1/J;)
2 2
'lj;j olj cos('lj;j- 1/J;)- ä 2(t) cos'lj;; + 1/J; ol;
0 0
0 0
lj 0
sin( 1/Jj - 1/J;)
In the singular case (for sin( 1/Ji -1/!j) = 0) this solution tends to infinity unless
additional conditions are satisfiedo Then the constraint acceleration equation
(3o53a) does not have a finite solution p(t) for
det ((92p, ßc 2]
0
(~~)) I =
1/Jj
- [ ( ( ( ( 8 °~i + 8 °~t - 4 °ä§ (t) - 8) 0
q 0
lJ 0
cos 1/J;
+(6 °ä2(t) ~; 0 0
zr + 8 °ä2(t) ~; ol;) olJ) 0 0
sinl/J; 0
cos1/;j
+((8oä2(t) o~J olT olj -2oä2(t) o~J olT olJ) 0
cos'lj;;
+ 4 °1/J;2 'lj;j
°2 l;3 + 8 °1/J;
°2 'lj;j
°2 l; ) olj3 + 8 °1/J;
°2 'lj;j
°2 ol;3 lj ) sin 'lj;; ) sin 'lj;j
0
( 0 0 0 0 0 0 0 0
+ 8 °'lj;j4 + 8 °1/J;-
°4 4 °ä 22(t)- 8 ) ol;2 olj2 cos'lj;;
0
(( 0
+(6 °ä2(t) ~; 0 0
zr + 8 °ä2(t) ~; ol;) olJ 0 0
cos'lj;;
+(( -4o~[-4°~[+5oä§(t)+4) ol7+4oä§(t)) olJ) 0
cos'lj;;
+((8oä2(t) o~J olT olj -2oä2(t) o~J olT olJ) 0
cos 2 'lj;;
+ 4 °1/J;2 'lj;j
°2 l;3 + 8 °1/J;
°2 'lj;j
°2 l; ) olj3 + 8 °1/J;
°2 'lj;j
°2 ol;3 lj ) cos 1/J;
0
(( 0 0 0 0 0 0 0
+(6 oä2(t) o~J oz; + 8 oä2(t) o~;) olJ) o COS1/;j

+(( -4o~j-4o~t+ä2(t)+4) olTolj+4oä§(t)olT) 0
cos 2 1/;;
+ (8 ä2 (t) ~;
° 0 0
q - 2 ä2 (t)° 0
~; 0
q lj)0 0
cos 1/;; + (2 ~j oz; + 4 ~j) lj
° ° 0
+ (2 0 ~t 0
lt + (4 ° ~j + 4 ~t) ° 0
ll) lj + 4
0
° ~t 0
lt) I 4] I # Oo
1/Jj
For
det ( (Y2p' ßc 2 ) ' (Y2p' ßc 2 )T) = 0, (3054)

1/Ji -1/Jj
(3o53a) has at least two finite solutionso This situation will now be analyzed
for the different singular caseso For Gases 201 a and 202a (1/;j = 27r, 1/;; =
0; li < lj, and l; > lj), this implies
det ((Y2p, ßc 2 ) 0
(~t)) = (3o55a)
(~; ozj + ~; ozi- ä2(t)) 2 0

( (lr + 2) oz; + 2 ozn ;2 = o
>0
for
(3o55b)
(bifurcation for ~j > 0, and ~i < 0, and for ~j < 0 and ~i > O)o For ä2 (t) < 0,
no motion is possible (lock-up)o For Gase 201b (1/;j = 27r, 1/Ji = 1r, and li < lj),
this implies
det ((Y2p, ßc 2 ) 0
(~t)) = (3o55c)
(~; ozj- ~; ozi- ä2(t)) 2 0

( (lr + 2) oz; + 2 ozn ;2 = o
>0
for
a2(t) = lj 01/Jj2 -li 01/Ji
°2
(3o55d)
0
-
Due to a2 (t) = 0 and (3o50a),
( cos'I/J; =
cos'I/Jj = -1
-1)
(3o55e)
Inserting (3o55e) into (3o55d) yields
or
(3o55f)
(bifurcation for ~j < 0 and ~i < 0, and for ~j > 0 and ~i > 0)0 For ä 2(t) > 0,
no motion is possible (lock-up)o For Gase 202b ('1/Jj = 1r, '1/Ji = 0, and li > lj),
this implies
det ((Y2p, ßcJ o (~!)) = (3o55g)
(~; ozj- ~7 ozi + ä2(t)) 2 ((Zr+ 2) oz; + 2 ozT) ;2 = o

0
>0
for
00 °2 °2
a2 (t) = li 0'lj; i - l j 0'lj; j 0 (3o55h)
Due to a2 (t) = 0 and (3.48a),
( cos'lj;i = 1)
cos'lj;j = -1
(3o55i)
Inserting (3o55i) into (3o56h) yields
or
(3o55j)
(bifurcation solutions for ~j > 0 and ~i < 0, and for ~j < 0 and ~i > O)o For
ä2 (t)> 0, no motion is possible (lock-up)o For Gase 203a ('lj;j = 21r, '1/Ji = 0,
and li = lj), this implies
det ((Y2p, ßc 2 ) 0
(~!)) = (3o56a)
z; 0
(~; ozj +~7 ozj -ä2(t)) 2 (z;
0
+4) j2 = o
for
(3o56b)
(bifurcations occur for ~j > 0 and ~i < 0, and for ~j < 0 and ~i > O)o For
ä2 (t)> 0, no motion is possible (lock-up)o For Gase 2o3b ('1/Jj = ~1r, '1/Ji = ~'
and li = lj), this implies
det ((Y2p, ßc 2 )
0
(~!)) = (3o56c)
(~j-~ifo(~j+~ifozjo (l]+4)/2=0
>0
iff
~j = ~i a2(t) = li (-~j +~i) = o)

'
0
(3o50a) :I
(3o56d)
and I
I
I
I
ä2(t) = 0, I
or
~j = -~i
'
a2(t) = 2olio~i ~
I
(bifurcation )o
(3o50a) I
I
I
and I
I (3o56e)
I
ä2(t) = 2 ° li ;j;i 0
J
For ä2 (t) > 0, no motion is possible for (3o56d) (lock-up)o For Gase 2o3c
('1/Jj = ~1r, '1/Ji = ~1r, and li = lj), this implies
det ((Y2p, ßc 2 )
0
(~!)) = (3o56f)
( ~j _ ~i) 2 0 ( ~j + ~i) 2 0
zj z; + 4) j 2= o
0
(
>0
iff
~j '
(3o56g)
or (bifurcation )o
(3o56h)
For ä2 (t) > 0, no motion is possible for (3o56g) (lock-up)o For Gase 2o3d
('1/Jj = 1r, '1/Ji = 1r, and li = lj), this implies
det ( (g2p, ßc 2 ) o ( ~!)) = (3o56i)
(3o56j)
iff
(3o56k)
(bifurcation solutions for ~i > 0 and ~j < 0, and for ~i < 0 and ~j > O)o For
ä2 > 0, no motion is possibleo
Additional singular cases are obtained by a particular control strategyo For
Gase 203e ('1/Jj = ~7r + c:, '1/Ji = ~ + c:, 0 < c: < ~' and li = lj), we obtain
det ((Y2p, ßc 2 ) 0
(~!)) = (3o57a)
z; 0
( ~; 0
zj - ~: 0
zj - ä2 (t) 0
sin(c:)) 2 0
(z; + 4) 12 = o
~
iff >0
ä2(t) = zj 0
(~;- ~n 1 sinc: = o, (3o57b)
where due to (3o50a), (3o52f), and '1/Ji = '1/Ji + 1r,

~j = ~io (3o57c)
For ä2 > 0, no motion is possibleo For Gase 203/ ('1/Jj = ~7r- c:, '1/Ji = ~7r
c:, 0 < c: < ~' and li = lj), we obtain
det ((Y2p, ßc 2 ) 0
(~!)) = (3o57d)
z; 0
( ~; 0
zj - ~; 0
zj - ä2 (t) 0
sin(c:) f 0
(z; + 4) 12 = o
~
iff >0
(3o57e)
where due to (3o50a), (3o52g), and '1/Ji = '1/Ji + 1r,

~j = ~io (3o57f)
For ä2 > 0 no motion is possibleo For Gase 203g ('1/Jj = 21r, '1/Ji = 1r, and
li = lj), we obtain
det ((Y2p, ßc 2 ) 0 (~!)) = (3o57g)
z; 0 ( ~; 0zj - ~7 ° zj - ä 2 ( t)) 2 0 ( z; + 4) j 2 = o
iff
(3o57h)
where due to (3o50a), (3o52h), and '1/Ji = '1/Ji + 1r,

~i = ~jo (3o57i)
For ä2 > 0, no motion is possibleo For Gase 2o3h ('1/Jj = ~7r- c:, '1/Ji = ~- c:,
and li = lj), we obtain
det ((g2p, ßc 2 ) 0 (~!)) = (3o57j)
z; 0 (~; ozj- ~7 ozj + ä2(t) 0 sin(s)) 2 0 (z; + 4) j2 = o

..______.......
iff >0
(3o57k)
where due to (3o50a), (3o52i), and '1/Ji = 1r + '1/Ji,

~i = ~jo (30571)
For ä 2 > 0, no motion is possibleo For Gase 2o3i ('1/Jj = ~7r + c:, '1/Ji = ~7r + c:,
det ((Y2p, ßc 2 ) 0 (~!)) = (3o57m)
z; 0 (~; ozj- ~7 ozj + ä2(t) 0 sin(s)) 2 0 (z; + 4) j2 = o

..______.......
>0
iff
(3o57n)
where due to (3o50a), (3o52j), and '1/Ji = 1r + '1/Ji,

~i = ~jo (3o57o)
For ä 2 > 0, no motion is possible (lock-up)o For Gase 203j ('1/Jj = 1r, '1/Ji = 0,
det ( (g2p, ßc 2 ) o ( ~!)) = (3o57p)
z; (~; ozj -~7 ozj +ä2(t)) 2 (z; +4) j2 = o

0 0
...._"..__...
>0
iff
(3o57q)
where due to (3o50a), (3o52k), and '1/Jj- '1/Ji = 1r,
~i = ~jo (3o57r)
For ä 2 > 0, no motion is possible (lock-up)o
The above results of singular situations of the slider-crank mechanism under

drive a 2 (t) are collected in Figures 305 and 3060
Comment 3.1.2 (Singularity analysis of the constraint equations):

The above singularity analysis of the slider-crank mechanism shows:
1. Pathological behavior is not restricted to strange or peculiar mechanisms
invented by theoreticianso Even simple often used mechanisms may have
various singularitieso
20 Computational singularity checks are required in general purpose rigid
body programs to identify both singularities of mechanisms and incon-
sistent constraint equationso
30 A complete singularity analysis, even of comparatively simple mecha-
nisms, may need comprehensive computational steps and may be quite
cumbersomeo
40 The constraint Jacobian matrix gP is of fundamental importance, both
for finding singular behavior and for solving the constraint equationso
50 The occurrence of certain types of singularities can be numerically
checked by computing determinants of certain matrices [gp , (gp, -gt) 0
(gp, -gt)T, (gp, ßc) 0(gp, ßc)T] or by checking the rank of some matri-
ceso These computations may be clone numerically and sometimes even
symbolically as in the above simple slider-crank mechanismo
Besides checking algebraic criteria that include the constraint Jacobian ma-
trix, lock-up situations or bifurcation points may also be empirically identified
by observing a rapid growth of some velocity and/or acceleration coordinates
of a mechanismo
......
singularity analysis ......
I;< lj l; > lj l; = lj IX!
(Casc 2: x~ 0 = -a.(l)) (Case 2. 1) (Case 2.2) (Case 2.3)

<;;:>
'1/Jj = 2'11 , ·1/J; = 0 1/lj = 2rr , 1/l; = rr 1/lj = 2'11, 1/J; = 0 1/lj = 'II, ·1/J; =0 '1/Jj = 2'11 , 1/>; = 0 1/lj = ~'II , 1/>; = ~
Q
(Casc 2.la) (Case 2. lb) (Casc 2.2a) (Casc 2.2b) (Case 2.3a) (Case 2.3b) 0
1:1
[/)
0 i l j 2 l i 0 2 0
l j 2 i 0 i 2 l 0 ; l j 2 ; j
..• :& @ .,_. ....@!: .....
""
~ ~ 'i!o=<l, ,., ~ 0,2 .
a, j
"• a,
"•
11-l-2
tPj = ~'II, tPi = ~;r 1/lj = '11,1/l; = rr
"'§.'
C1l
(Case 2.3c) (Case 2.3d) >§
singular const.raint 2 j 1 i 0
:[ha• ~
Jacobian: . ~ c;·
1:1
[/)
"'j = ~1r+<, !/J; = ~ +< 'Pi= ~1T' - f., 'l/J; = ~7f - (
sin(t/>;- tPi) = 0 (Case 2.3e) (Casc 2.31) §
or 0..
r , l -2 ("")
(1/>;- tPj) = n · 11 ~ a~ ] 0
1:1
[/)
tPj = 2rr, 1/>; = 11 1/Ji = ~1r - f.)f/;j = ~- f
.....
'""
(Case 2.3g) (Case 2.3h) ~
~
.....
C1l
]
~
~"·
'1/Ji = ~1t
+ ~rr + E, 1/J; = l 1/lj = 11,1/l; = 0 ~
M-
(Case 2.3i) (Case 2.3j) o'
1:1
~·. 8'
.....
("")
constrai nt a2= 0 : bifurcations for Cases

e C1l
[/)
velocit.y analysis ä2 = 0 ä2 = 0 ä2 = 0 ä2 = 0 2.3a, 2.3d, 2.3e, 2.3f, 0

....,
2.3g, 2.3h, 2.3i, 2.3j
än~ o: Iock-ups fo r these cases,
s
C1l
~~0 : lock-up ~ is free for one of the branching &

solutions of Cases 2.3b, 2.3c §
tij'
s
[/)
Fig. 3.5: Singular solutions of the slider-crank mechanism of Case 2 (drive a2(t))
constraint äz > 0 ä2 < 0 ä.2 >0 äz < 0 ·;;r
~
acceleratian
tagether with tagether with tagether with tagether with
analysis
·,j >0 ·,j >0 ~j > 0 ~j > 0 Gase 2.3b: ä~ arbitmry l
and and and and ä2 = 0 or. ~ a2
··\ < 0; i > 0; .. i,_i
< 0; ~i < 0; ä2=2·l;'I/Ji ~ or~
bifurcation ~
ar with ar with ar with ar with ......
Gase 2.3c: ä2 arbitrary
:;.:::
'j <0 •j <0 ~j < 0 ~j < 0 ä~ = 0 or
and and and and s·
('!)
a~ = 2 ·l;<b;
i > 0. i < 0. > 0. 'i ~i > 0. s
~ U2 ~
~ '"
~--------~
__.---:~' a~
~--<1(---- (=;'
~~ [fJ
Gf· ·-· ·--- •• "(ij.· ·-----~ ä~ < 0 bifurcation 0
....,
bifurcatian bifurcation bifurcation bifurcation ä~ > 0 lock-up "0
Case 2.3e: ä2 = 0 lock-up §"
äz < 0 äz > 0 ä.z <0 ä.z >0 ~ ~
lack-up lack-up lack-up lack-up lock-up ~p_.
[fJ
"0
ä2 = 0 lock-up
~
E
lock-up .....
CJ'q'
lock-up
~
o"
0
~
Iock-up [fJ
'<
[fJ
('!)
s""
[fJ
Fig. 3.6: Constraint acceleration analysis of the slider- crank mechanism of Case 2 (drive a2(t))
......
......
<D
3.2 Constraint reaction forces and torques of

mechanisms
In this section, basic relations of reaction forces and torques will be briefly
discussed for planar and spatial mechanismso By the definition of kinematic
analysis, forces and torques are only considered in the dynamic analysis of
mechanisms and not in pure kinematic analysis; this section is concerned with
certain aspects of dynamic analysiso
As a consequence andin agreement with Beetions 301.1.3 and 301.202, the
inequality
(planar case} (3o58a)

or
(spatial case) (3o58b)
will be assumedo
3.2.1 Constraint reaction forces of planar mechanisms
Constraints connected to rigid bodies produce constraint reaction forces (and

torques) that act on these bodieso Constraint reaction forces clearly depend on
the associated constraint equations, and so it should be possible to express
them in terms of these constraint equationso This can be achieved if the
reactions of the constraint forces are expressed with respect to the same
coordinates (frames) as the vectors of the generalized coordinates
(3o59a)
and the associated velocities
V:= P = (xp,o, YP2 0, -J;l; 000 ; Xpn•O, YPn•O, -J;n.) T =: (rT, "j;T) T (3o59b)
with
Wzi := "j;io (3o59c)
It will be assumed that the nc kinematic constraint equations are consistent

and independento Then, due to
(3o60a)
the 3nc constraint equations
g(p) =0 (3o60b)
302 Constraint reaction forces and torques of mechanisms 121
can be uniquely solved with respect to the 3nc (dependent) coordinates of

p, v = jJ andv = p, for given independent coordinates of these vectorso
The work done by a force F providing a displacement p of a system is
(3o6la)
A virtual displacement 8r p of a system is defined as a small (infinites-

imal) variation in the location r p of a system at a jixed time instant t,
consistent with the constraints and forces acting on the systemo Expressing
Tp = rp(p(t), t) in terms of the generalized coordinates p, small physical
displacements of Tp are
8rp 8rp
drp = - odp+- odt (3o6lb)
ap at '
whereas a virtual displacement 8r p may be considered as a partial derivative
of r p with only the physical coordinates varied and the timet kept constant:
8rp
8rp = ßp 8po0 (3o6lc)
The virtual work done by constraint reaction forces and torques cf is zero:
cw = cfT 8p = 0
0 (virtual work principle)o (3062)
The Taylor series expansion of the constraint equations with respect to p

provides, to a first-order approximation,
g(p + 8p) = g(p) + Yp(P) 0

8p,
and together with the constraint equation
g(p) =0,
and by the definition of the virtual displacement 8p, ioeo, for
g(p+8p) =0
the relation
Yp(P) 8p
0
= Oo (3063)
The coordinate vector p E ffi.3 nb may be (conceptually) partitioned into nc

dependent coordinates u E ffi.nc and (3nb- nc) independent coordinates w E
ffi.3nb -nc 0 Then
8g
8p =: Yp = (gu' Yw)'
(3064)
with a regular matrix 8g I au = g u E ]Rnc ,nc' and
Using these notations, (3062) and (3063) can be written as
(ifT 8u = --,;!T 8w
0 0
(3o65a)
and
Yu o8u = -gw o8w, (3o65b)
or as
(3o65c)
and
As the last nc rows of the matrix ( ,';::uT) E JRnc+l,nc are, by assumption,
linearly independent, the first row ,';:fT of this matrix can be written as
a linear combination of the last nc
rows with the coefficient vector A =
(.A1, ooo, .AnJT -:j:. Oo Then
or (3066)
The coefficient vector A is known as the vector of the Lagrange multiplierso

Substituting (3066) into (3o65a) yields
AT 0g u 08u = - wcfT 08w ' (3o67a)
and tagether with (3o65b) yields
(3o67b)
The relation (3o67b) holds for arbitrary variations 8w of the ( independent)

coordinates w 0 Then
or (3068)
Appending (3068) to (3066) yields
or (3069)
3.2 Constraint reaction forces and torques of mechanisms 123
as the constraint reaction forces, expressed in terms of the constraint Jacobian

of the constraint equations and of a vector of Lagrange multipliers.
The Lagrange multipliers A may be computed from the model equations in
DAE form, or eliminated from the model equations to produce model equa-
tions in DE form.
In any case, the signs of the components Ai of A are not known before the
model equations are solved. As a consequence, the directions of the constraint
reaction forces and torques cf are also not known in advance. This implies
that the arrows of the constraint reaction forces drawn in a mechanical net-
work of a system do not tell anything about their true directions. On the other
hand, due to Newton's law that "action is equal to reaction", the constraint
reaction forces and torques, produced by a joint that connects two bodies i
and j, are equal in magnitude and act in opposite directions on these bodies
(Figure 3.7).
3.2.2 Constraint reaction forces of spatial mechanisms

Spatial mechanisms including nb rigid bodies are specified by a Cartesian
coordinate vector
(3.70a)
or
R T T)T 6
Pi= ( rpio , '11i E lR
with
T~iO = (x~iO' Y~iO' ZpiQ) T ' 'IJi = (r.pi' ßi' '1/Ji)T,
and with the velocity vector
·- ( v 1
v .- T, ... , vnb
T )T E lR6nb (3.70b)
with
·
Pi= (rpio) T ''11i
( ·R ·T)T
v = T(p). jJ = T('IJ). jJ, (3.70c)
T(p) := diag (T1(P1), ·· ·, Tnb(Pnb)),
Ti(Pi) = Ti('11i) = diag (r3, Hi2('11i) · ARLi),

(3.70d)
revolute joint
(a) Two bodies connected by a revolute joint
cpR = _cpR cpR = _cpR

ZX )X ' zy JY
(b) Free-body diagram for the constraint reaction forces
Figo 307: Revolute joint and the associated constraint reaction forces
Ci2Ci3 , Ci1Si3 + Si1Si2Ci3 , Si1Si3- Ci1Si2Ci3)

(
-c i2s i3 ' c i1 c i3 - s i1 s i2 s i3 ' s i1 c i3 + c i1 s i2 s i3 '
s i2 - S il C i2 C il Ci2
(3o70e)
1 , 0 , Si2 )
Hi 1 (11i) := ( 0 'Cil ' -Sil ° Ci2 (30 70f)

0 ' s il ' c il 0
c i2
and
s il := sin 'PiLR , s i2 := sin ()iLR s i3 := sin '1/JiLR , (3o70g)

C il := COS 'PiLR , C i2 := COS ()iLR , and C i3 := COS '1/JiLR ,
for Bryant angles 'PiLR, ()iLR, '1/JiLRo
Then the principle of virtual work
(3o71a)
introduced in Section 30201 with
Cf- -- (Cf- T oooCf- Tnb ) T

E
ID6nb
(3o71b)
1 , .Ii\\. ,
Cf- (cfR cfR cfR CM CM CM ) T oc6 (3o71c)

i = xi ' yi ' zi ' xi ' yi ' zi E ,
and with the coordinate vector p (30 70a) can be written as

nb
cw = L (cf~i 08x~;O + cf~ 08y~o + cf! 08z~o (3o71d)
i=l
+ cMxi 08r.pi + cMyi 08()i + cMzi 08'1/Ji)

with
cMji , j = x , y , z , and 8r.pi , 8()i , 8'1/Ji , i = 1 , 000 , nb (3o7le)
associated with the chosen rotation axeso Consider nc consistent and inde-
pendent constraint position equations
U! U! (3o72a)
p f---t g(p) =0
with
rank g p (p) = n c, gp E lTilnc
.Ii\\.
,6nbo (3o72b)
The Taylor series expansion of the constraint equations with respect to p

provides, to a first order approximation (compare the derivation of Equation
3063), the relation
Yp(P) 8p := Oo0 (3o73a)
The coordinate vector p E OC6 nb may be (conceptually) partitioned into nc

dependent Coordinates U E 0Cnc and (6nb- nc) independent Coordinates W E
OC6 nb -nc 0 Then
with the regular matrix Yu E ocnc ,nc, and with
(3o74b)
Using these notations, provides (compare the derivation of Equation 3069)

the equation
or (3075)
as the constraint reaction forces expressed in terms of the constraint J acobian

matrix and of the vector of the Lagrange multiplierso
Divergent from the planar case, spatial equations of motion are written in
the velocity coordinate v instead of po This must be taken into account in the
formulation of the constraint reaction forceso Starting from the virtual power
represented in the coordinates jJ and v yields, for a body i, the relations
with
öwf~R = ALiR(11i) Hi1(11i) MJi.
0 0 (3o76b)
Inserting (30 76b) into (30 76a) yields
(3o76c)
(
c5<iJi , Mh , 8'1/Ji
o o )T 0
As this equation holds for arbitrary variations c5iJi, we have
(cMxi' cMyi' cMzi)T = (ALiR(11i) 0Hi1(11i)f 0 (cM~i' cM~i' cM~if 0

(3o76d)
Combining the constraint reaction forces and torques for the virtual transla-
tions and rotations yields, tagether with (3o7ld) and (3o76d), the relation
(3o76e)
cMxi cML!
XZ
CMyi cML!
yz
cMzi cML!
...._"_,__... .......___.....
zz
=: cJi =: cfi
or
or
- (
cfo = T.T ('17·) ) -l
0 cf. = g.T (p) Ai,
0 (3o76f)
' ' ' ' (30 75) •P
and finally the relation
(3077)
4. Dynami es of planar and spatial rigid-body
systems
In this chapter the equations of spatial and planar motion of uneonstrained

rigid bodies will be derived based on the laws of Newton and Euler (such
as in standard textbooks like [44], [56], [57], [58], [59], [60], [61], [62], [63],
[64], [65]). The equations of motionwill be written with respect to a general
body-fixed reference point Pi (Pi f::- Ci , Ci is the center of mass) of a body i.
In Beetion 4.1 the notions of linear momentum and angular momentum of a
rigid body will be introduced and rewritten in a form suitable for representing
the Newton-Euler equations in a desired form. Here the notions ofthe eenter
of mass and the inertia matrix of a rigid body will also be introduced, and
some properties of this matrix will be briefly discussed. In Beetion 4.2 the
Newton-Euler equations will be derived for planar and spatial motion of a
rigid body represented with respect to the reference points Pi and Ci. (In
Appendix A.2 the Lagrange formalism, applied to a rigid body under spatial
motion, will be briefly discussed). The equations of motion of unconstrained
and constrained planar and spatial rigid-body mechanisms will be collected in
Beetion 4.3, combining the kinematie eonstraint equations of Beetion 3.1 and
the rigid-body equations of Beetion 4.2.4. This provides model equations in DE
form for uneonstrained rigid hoclies and in DAE form for eonstrained rigid
bodies and rigid-body meehanisms. A few aspects concerning the numerieal
solution of DAEs will be briefly discussed in Beetion 4.4.
4.1 Linear momentum and angular momentum of a

rigid body
This section considers a rigid body of mass m and volume V, a point P fixed
on the body, and a vector r PO from the origin 0 of the inertial frame R to
P (cf. Figure 4.1).
4.1.1 Linear momentum
Consider a mass element of the rigid body of mass dm located at an arbitrary

point Q of the rigid body, specified by the vector
T := TQO· (4.1)
130 40 Dynamics of planar and spatial rigid-body systems
mass element dm
R rigid body of mass m
and volume V
inertial frame R
R
Figo 401: Vector diagram used in the definitions of linear momentum and angular
momentum
Then the expression
P~ :=I RrRdm (402)
is called linear momentum of the rigid body with respect to the origin 0 of
frame R, represented in R, and
L~ := IrR 0
R7-R dm (403)
is called angular momentum of the rigid body with respect to 0 of R and

represented in R, where the velocity vector RrR in (402) and (403) is measured
in R and represented in Ro Consider a second frame L with origin 0 L = P,
and fixed on the bodyo Then the vector r can be written as (Figure 401)
rR := r~o + XR = r~o + A RL 0 XL 0 (4.4a)
The velocity vector RrR is
RrR = Rr~o + Rx_R (4o4b)
or
4.1 Linear momentum and angular momentum of a rigid body 131
Due to the rigid-body property
(4.5)
and to the relation

A. RL-
-
ARL -L
·WLR'
the resulting velocity vector is
(4.6)
Then the linear momentum P~ of the rigid body can be written as
or as
(4.7a)
with C as the center of mass of the body, with
L
rcp 1
:= m. I X L dm (4.8)
as the vector from the origin P of L to C, and with
m :=I dm (4.9)
as the mass ofthe body. For the special case P = C, where the reference point
Pischosen as the center of mass C (rcp = 0), a simplified representation of
P~ is obtained:
(4.7b)
4.1.2 Angular momentum
The angular momentum of the rigid body
(4.3)
can be written tagether with (4.4a) and (4.4b) as

or as
L~ = I r~o Rr~o dm + I
0 XR R.X.R dm
0 (4010)
+ I XR Rr~o dm + I r~o Rx_R dmo

0 0
The different terms of (4010) can be rewritten as follows:
10 1 T-R RoR -R RoRd

PO 0 TPO m = TPO 0 TPO 0
ld -R
m = m 0T RoR
PO 0 TPO 0 (4011)
2° I XR 0RT~o dm =(I xRdm) 0RT~o =(I xRdm) 0RT~o'

which yields, together with the vector
R 0= m1 0 I X Rd m= ARL orcp,
Tcp 0 L (4012)
the relation
-R RoR d ( 1 I Rd ) RoR -R RoR

I X 0 TPO m = m 0 m 0 X m 0 TPO = m 0TCP 0 TPO
or
-R R R d ARL 0T- CP
L 0ALR 0RTPO
R 0
I X 0 TPO m = m 0 (4013)
0 0
30 I r~o 0Rx_Rdm = r~o 0 (I Rx_Rdm)'

which yields together with the rigid-body property (Lx_L = 0)
and the relation
Rx_R = .ARL 0LxL + ARL 0Lx_L = ARL 0wfR 0xL, (4014)
the relation
I r~ 0 °Rx_R dm (4015)
-R
=rpoo I ARL owLROX
-L L dm=rpoo
- R ARL owLRO
- L I X L dm
- R 0ARL 0WLR
= m 0Tpo - L 0Tcp
L = -m 0Tpo
- R 0ARL 0Tcp
- L 0WLRO
L
401 Linear momentum and angular momentum of a rigid body 133
40 f XR 0RXR dm = f XR 0A RL 0wfR 0'XL dm (4016)
= - A RL 0 (/XL 0XL dm) 0wfR
or
-oJL
-o p
and finally
f XRORXRdm =:J~owfR=ARLoJ~oALRowfR
= A RL 0
JL
p 0
L
WLR
with
J~ :=- J xL oxLdm
and (4017)
J~ = ARL 0 J~ 0 ALR
as the inertia tensor (inertia matrix) of the rigid body with respect to the
point P, and represented in the local (body-fixed) frame L and in the global
frame R, respectivelyo
Together with (4011), (4013), (4015), and (4016) the angular momentum L~
of the body can be written as
L 0R = JR
p
R
0W LR + m 0T PO 0
-R RoR
T PO (4o18a)
+mo A RL OTcpo
-L ALR
0R Tpo-moTpoo
R 0
-R ARL -L L
OTcpoWLR
or as
(4o18b)
Together with
RoR RoR
Tco = Tpo + ARL -L L
OWLR OTcp (4019)
this implies
L o=
R ARL
0
JL L -R R R
powLR+morpoo rco
o
(4020)
+m 0 ARL 0r~p 0ALR 0Rr~oo
=
For the special case P = C, L K, rco = rpo, and rcp
momentum L~ (4020) is written as
= 0, the angular
R
L o= ARK JK K -R R
cowKR+morcoorco
0
0
(4021)
with the (constant) inertia matrix of the body with respect to point C
(4o22a)
and with
Jffxx := j ((P:J) 2
+ (P~) 2 ) dm as a moment of inertia (4o22b)
and
Jffxv := J p{f 0pl_J dm as a product of inertia
of the rigid body with respect of its center of mass C, and with
PK = ( PxK , PvK , PzK) (4o22c)
as the vector from point C to the location of the mass element dm (Fig-
ure 401)0
4.1.3 Properties of the inertia matrix
4.1.3.1 Physical interpretation of Ji;,. Using the relation
xL 0xL = - [ (xL) T0xL 013 - xL 0 (xL) T] = (xL 0xL) T (4023)
im plies, together wi th (4 01 7)
J~ = f [(xL) T 0XL 013- XL 0 (xL) T] dm, (4o24a)
or written in components with respect to frame L,
(4o24b)
with
J~xx := J[(x~ f + (x~ f] dm, etco, as the moments of inertia and
JX~ X~
(4o24c)
J~xy := 0 dm, etco, as the products of inertia
401 Linear momentum and angular momentum of a rigid body 135
of the rigid body with respect to a body-fixed point P, and with
XL = ( L
Xx, L
Xy, XzL)T
as the vector from point P to the mass element dm located at the point Q
(Figure 401)0
4.1.3.2 Time dependence of Jf; and Jf!. For rigid bodies of constant
mass m, the inertia matrix J~ is a constant matrix when represented in
frame L, whereas the matrix
(4025)
represented in frame R, is in general a time-dependent matrix, due to A RL =

A RL ( 7](t) )o
4.1.3.3 Steiner-Huygens relation. Changing the reference point of J
from point C (origin of frame K) to point P (origin of frame L) provides
the following transformation of J:
JP
L = ALK 0JK
e 0AKL L )T 0Tep
+ m 0 ((Tep L 0I 3 - Tep
L L )T) 0
0( Tep (4o26a)
Or, written in components,
Jf5xx '
(
J~ = -J~xy' (4o26b)
-Jexz '
-oJL
-o e
+mo L 2 L 2
'Xep +zep '
L L
-ZepYeP
with
L
Tep =
( L L
Xep' YeP' Zep
L )T (4o26c)
Planar case (moment ofinertia): In the planar case only rotations about
the z-axis perpendicular to the x-y plane occuro Then
J}J = J~ = hzz =: Jp E lR1 , (4o27a)
and the Steiner-Huygens relation is
L
Jp =Je+ m
L 0
( L
Xep
2+ YeP 2) =Je+ m
L L 0 (4o27b)
Proof of the Steiner-Huygens relation (4.26):
Inserting the relation
X= rcp +p (4028)
(Figure 401) into (4017) provides the relation
J~=- jxLoxLdm=- j(r~p+PL)o(r~p+PL)dm, (4029)
and, together with
(a + b) = a + b,
the relation
J Lp=- f-p L op- L dm-r- 0L porcpom

-L
- r~p 0 ( / pL dm) - ( / pL dm) 0r~po
Together with
j PL dm = 0 , pL := r~ 0 , and (4030)
J pL dm = 0 (due to the definition of C),
this yields the transformation relation
JL
P =
JL -L
c - rcp -L
rcp m 0 0 (4031)
f f(
with
J~ := - PL 0PL dm = (pL)T 0PL 013- PL 0(pL)T) dmo (4032)
Due to
J~ =ALK 0 J{§ AKL

0
(4033)
and to
r~p 0r~p =- ( (r~p)T 0r~p 013 - r~p 0 (r~p)T) ,
the following relation holds
J Lp= ALK 0JK L )T orcpo

co AKL + (( rcp L I 3-rcpo
L L )T) omo (4034)
(rcp
This proves (4o26a)o D

402 Newton-Euler equations of an unconstrained rigid body 137
4.2 Newton-Euler equations of an unconstrained rigid

body
Based on the previously introduced linear momentum P~ and angular mo-
mentum L~, Newton's and Euler's laws can be compactly and precisely
stated in symmetric form, when applied to an unconstrained rigid body, as
Rd pR = '""'F!l (Newton's law) (4035)

dt 0 ~ •
•
(Euler's law), (4036)
with 2:i Ff as the resultant external force acting on the body and 2:i M{6
as the resultant external force moment (torque) with respect to the point
0 = OR, acting on the body, with R as an inertial frame, and with Rr~ 0
included in both, P~ and L~, where r~ 0 is differentiated with respect to
time in the inertial frame R (Rdjdt; compare Equation 2028c)o
4.2.1 Forcemomentsand couples
A torque M may occur as a force moment, generated by a force, or as a

couple moment (generated by two collinear forces of equal absolute values
and of opposite signs)o
Consider a force Fi, acting on a body (as shown in Figure 4o2a), and a
point P not located on the (dashed) line of action of Fio Consider two points
Pi and Qi on the line of action of Fio Then the force moment (moment,
torque) M iF of F i acting on the body with respect to the reference point P
is defined as
R -R
M iPF:=rP;Po FR -R
i =rQ;Po
FR
io (4037)
Subsequently the point Pi will be used in MiPF as the point on the line of
action of the force Fio The torque Mi produced by two collinear forces Fj
and -Fj of equal absolute values and opposite signs (a "couple of forces")
is known as couple momento Due to the relations (Figure 4o2b)
M R -R
jP=rP;Po FR -R
j -rQ;Po FR R - R ) FR
j = (-rP;P-rQ;P j 0
and
M R
js= (-R -R ) FR
rp;s-rQ;S i'0
tagether with
-R R R R R
8 P;Q; = rP;P- rQ;P = rP;S- rQ;S'
138 4. Dynamics of planar and spatial rigid-body systems
F;
external force
rigid body
I
I
I
1--.....
line of action of F;
I
I
(a) Torque generated by a force
rigid body
(b) Torque produced by a couple of forces
(c) Torque with respect to the origin 0
Fig. 4.2: Forces and torques

the couple moment
(4038)
is independent of the reference points P and Qo Thus the moment of a couple

of forces is the same about arbitrary reference pointso Therefore couple mo-
ments are called free vectors, whereas force moments are called sliding vectors
in the sense that they are moments about an axis through a reference pointo
Consider the vector diagram of Figure 4o2c, in which the torque MiF
generated by a force F i acting on the rigid body can be taken with respect
to the origin 0 of a reference frame Ro This implies
R -R
M iOF=rpioo FR -R
i =rpoo
FR -R
i +rpipo
FR -R
i =rpoo
FR
i + MRiPF'
(4039)
where
R
M iPF -R FR
:= Tpip i 0
is the moment of Fi with respect to point P on the rigid body, and MiOF
is the moment of Fi with respect to point 0 of the frame Ro Then the total
moment acting on a body and measured with respect to point 0 can be
written in the form
M~=L Mfh
i ~
+L -R
rpio
~
0
FR
i 0
(4040)
couple moment of
moment the force Fi
Tagether with
R R R
Tpio = Tpo + Tpip, (4.41)
this yields
R "MR
M o = ~ i + '""'-R
~ r PiP 0
FR
i
-R
+r PO 0
"FR
~ i
i i i
-oMR
-o p
or
M~ = M~ +r~ 0 LF~ ° (4o42a)
with
M~ = Z:M~+ z=r~ip oF~ (4.42b)
as the total moment on the body measured with respect to the point P 0
Planar case: The product rR ·FR with the vectors rR = (xR, yR, zR)T =
(XR 'yR' o)T and FR = (FR FR FR)T = (FR FR o)T located inside
X'Y'Z X'Y'
the xR-yR plane of 0C provides the torque vector in 0C3 :
3
MR=fR.RFR=( ~R (4.43)
-yR
or
M = ( -yR · F: + F: · xR) · e 2 R,
that acts in the e 2 R-direction, perpendicular to the xR-yR plane.
4.2.2 Newton's law
Assuming that the mass m of the body is constant, and inserting (4.7a) into
(4.35) yields, together with
R·R Rd (R. R
p o = m . dt r PO + A RL . w- LLR . r CP
L )
' (4.44a)
the relation
Rp~ = m. (Rr~ 0 + ARL. wiR. wiR. r~p- ARL. r~p. wiR)= L Ff
i
(4.44b)
that can be written as
( m . 13 , - m . A RL . r~ P) . ( R.rfo ) (4.44c)
WLR
= "FR
L.....J i - m. ARL -L WLR.
. WLR. -L Tcp·
L
i
=
The special case P = C, where the reference point P is chosen as the
center of mass C of the body (r PC 0), yields the simplified version of
Newton's law as pure translational equation of motion:
m. R rco
.. R = "FR
L.....J i ' (4.44d)
that describes the acceleration of the center of mass C of the body with re-
spect to the inertial frame R, measured in R. This does not explicitly depend
on its angular velocity WLR·
4.2 Newton-Euler equations of an unconstrained rigid body 141
4.2.3 Euler's law
Contrary to the linear momentum relation (4.44a), where the constant mass
has been extracted from the bracket of P~ before the time differentiation,
the time-dependent inertia matrix J~ = ARL · J~ · ALR cannot be treated
identically. This provides a more complex expression of Euler's law. lnserting
(4.20) into (4.36) yields, tagether with
·L
Jp :=0, (4.45)
the relation
Rd ( RL
R · R L L -R R ·R
La=-- A ·Jp·WLR+m·Tpo· Tco
dt
+m. A RL ·Top.
-L ALR . R Tpo
R ) 0
or
Rd ( RL
R · R
L0 =dt A
L L -R R ·R -R R ·R )
·Jp·wLR+m·Tpo· Tco+m·Tcp· Tpo
= A
·RL
0
L
J p 0
L
w LR +A RL
0
·L
J p 0
L
w LR + A RL 0
L
J p 0
·L
w LR
+m. R:.R
Tpo.
R·R -R R··R
Tco +m ·Tpo. Tco
R:.R R·R -R R··R
+m . T CP . T PO +m . T CP . T PO'
or
Rt~ = ARL wfR J~ w~L 0 0 0 + ARL 0 J~ wfR

0 (4.46)
+ -R R··R
m . T PO . T CO + R:.R R·R
m . T PO . T CO (4.47)
+m .
R:.R
Tc P . T PO + m
R·R -R R··R
. T CP . T PO'
or
l
R · R
La = A
RL -L L L
. w LR . J P . w RL + A RL . J L
P . w LR
· L
R··R (R:.R R:.R ) R. R -R R··R

+ m. -R
[ Tpo. Tco + Top- Tco . Tpo +Top. Tpo .
-· R:.R
Tpo
=0
Taking into account
R·R R·R R·R
Top- Tco = - Tpo (4.48)
and
R:..R RoR -Q
rpo 0 rpo = (4.49)
provides
R R
=
RL -L
0w LR 0J P 0w RL
L L
+A RL L
0J P 0w LR
L
(4050)
0
o
La A
-R
( rpoo R .. R -R R .. R )
+mo rco+rcpo r p o '
and finally yields Euler's equation in the form

RLo R (-R RooR
= m 0 r PO 0 r CO + r-RCP 0RooR )
r PO +
ARL 0JL 0 L (4051)
0
0 P W LR
+A RL -L 0JL 0 L
0WLR "'MR
P WRL = ~ iOO
i
Inserting (4.42a) into (4051) yields

m 0 (r~o 0RT~o + r~p 0RT~o) + ARL 0J~ 0wfR
RL -L JL L ""MR ""-R FR
+A OWLRO poWLR=~ iP+~rpoo i (4052)
with
R MR
M iP = i + r-RP;P 0
FR
i o
Taking into account (4.44d) yields

m 0 (r~o 0RT~o + r~p 0RT~o) + ARL 0J~ 0wfR
+ARL owfR 0J~ owfR = LM~ +m 0r~o 0RT~o' (4053)
and after multiplication of (4053) by ALR, the relation

LR -R ARL ALR RooR JL L -L JL L "'ML
mo A orcpo Tpo+ poWLR+WLRO poWLR= ~ iP
0
0 0
i
-L
=: rcp
or
-L ALR RooR JL L "'ML -L JL L (4o53a)
morcpo Tpo+ poWLR= ~ iP-WLRO poWLR'
0
with
L M f p = ALR 0 LM~ (4o53b)
and
(4.42b)
Written in vector form, Euler's equation of motion is (for P =j:. C)
-L
( morcpoA LR ,Jp
L) 0
( RT~o
oL ) " " ML
=~ - L
iP-WLRO JL L
poWLRO (4o54a)
WLR
4.2 Newton-Euler equations of an unconstrained rigid body 143
Special case (P = C): Choosing the reference point P as the center of

mass C of the body yields, together with rcp 0 and L = K, the simptified =
version of Euter's taw as the equation of motion of pure rotations of the body
with respect to C:
J CK ° K
. W KR
"MK
= L.....J
iC - W KR
- K . JK
C
K
. W KR'
i
with (4.54b)
K
M iC MK -K FK
= i + r Pie · i ·
4.2.4 Newton-Euler equations of a rigid body under planar and

spatial motion
In this section the Newton-Euter equations of a rigid body under planar and
spatial motion will be collected.
4.2.4.1 Spatial motion. Combining the previous results (4.44c), (4.54a),
(4.42b), and (4.26a) yields the Newton-Euter equations of a rigid body under
spatiat motion for P -:j:. C:
r
-m . ( ~~~~ . A LR : .._J_~_+_m_·_(_(r_~_:_)_~_·_.:_~" RP_L_· -~-~-~-T-~-p-·(_r_~_P_)_T_,)1
-·JL
- . p
~ ~ ~
=:M
R··R
rpo )
( (4.55a)
·L
WLR
l
=:f
[ m·ARL·wfR·wfR·T~p
- wiR· [J~+m· ((r~p)T·T~p-13-T~p·(T~p)T)] ·wiR .
=: Qa
They are coupted in the coordinates of Rr~ and wfR· The matrices and
vectors of (4.55a) have (for a body i with P =Pi -:j:. Ci= C) the form:
.....
""'""'""
t:l
'<:
~
m,,O,O) ~
r;·
( 0 , m,, 0 "'g,
0 , 0 , m, "0
M '·= (4.55c) g
Li
0 -ZciPi ' ' c il s i3 + s il s i2 c i3 , Si1Si3-Ci1Si2Ci3) ~
L·
Ycl:i ) .(~ i2 C i3
( Li §
ffii" Zcip~ 0 -XciPi Cz2 Sz3 c il c i3 - s il s i2 s i3 Sil Ci3 + Cil 8;2 8;3 0...
L· Li C il Ci2
-yci"'Pi XciPi 0 Si2 -s il c i2 "0
"'
~
[
....,
( e;,o;, , -c;";' , <n ) z~;Pi ' -yci"'Pi cJq"
' L )
ffii" c >1 s ,3 + s >1 s i2 c i3 , c il c i3 - s il s ,2 s ,3 , -s il c i2 · ( 0 x~fPi ~
-z~.,; o"
' 0
s d s '3 - c d s '2 c '3 ' s d c '3 +cd s '2 s '3 ' c d c '2 Y~iiPi ' -x~fPi ' 0 ~
Li Li '<:
"'
L·
-Jci."XY -Jc"'."xz
L ) ( ( Yc"'"p"'
L -Xe iPi Yc iPi ) ~
J{'f" (!)
2+ ZclPi
L· 2) -x~fp, z~;Pi
( L· L· L." Li Li ( L· L· 2) s
-Jci"'yx Jci"'Y -Jc",yz + ffiz" -yciPi XciPi ' XclPi Zci"Pi ' -y~fPi z~iiPi
2+
L· 2 2
"'
J.L." L"' Li
-J!:iiZX , -Jc/zy ' c"'z -ZC 1 Pt XciPi -z~fPi Y~iiPi , (XeL·lPi + YcL·lPi )
c i2 c i3 , -c i2 s i3 , s i2 ) ~
I-:>
0
m; ( c ;1 s i3 + s ;1 s ;2 c i3 , c ;1 c i3 - s ;1 s ;2 s i3 , -s ;1 c ;2
z
s '1 s '3 - c >1 s '2 c '3 ' s >1 c '3 + c >1 s '2 s '3 ' c '1 c '2
~
0
L L L· Lt Li Lt
0 0 Li Li Li )
- ( WzitR 2) . Xc;pi WxLiR · WzLiR · ZciPi
wxL,R WYLiR YciPi
2+ WY_LtR + +
(
~
.,::
Lt Lt Li L· L· L· Lt Li Li
0 0
WxLtR 'WYLtR 'XciPi ZciPi
( WxLiR 2+wzLiR 2) OYc:Pi + wyL'/.R WZLiR w
("1)
L, L, Li + WLi ·WLi . Li L· 2 L· 2 L·
WxLtR. WzLtR. XciPi yLiR zLiR YciPi (wxLiR +wYLiR) ozc;Pi >§
~
50
Qai= 1:1
Li rn
0 -WzLiR ' WYiiR Jcitx L·
' -Jctxy L·
' -Jctxz ) 0
....,
L· L· ) [( L·
( Li TLi
WziiR 0 -WXLiR . -..Jciyx ' J{:iiY ' -J!:iiYZ g
Li Li .,::
-WYLiR WXLiR 0 -J{:fzx ' -J!:iiZY ' J!:iiZ 1:1
(")
0
1:1
L· 2) Li Li ~
( YcL·i."Pi Zc i'/.Pi -XciPi YciPi
2+
-xf;Pi zffPi )
L· L· 2) w~~iR) "'1:1~0
-y~fPi x~:Pi ( Xc;Pi
2+zc;Pi -yfiiPi zfiiPi ("1)
+m; ( WYLiR
0..
-zfiiPi x~;Pi - zfiiPi yfiiPi
L· 2
(Xe;Pi +
L· 2
Yc :Pi )
l( Li
WZLiR "'
c§:
0..
(4055d) o"
0
~
......
Q1
""'"
with s il := sin c/JiLR s i2 := sin ()iLR , s i3 := sin '1/JiLR,

(4o55e)
C il := COS c/JiLR C i2 := COS ()iLR and C i3 := COS '1/JiLR,
and
For P =C the Newton-Euler equations are no langer coupled:
( m
0
l3 , 03: )
' Je
0 ( ~rJo
WKR
) (4o56a)
)
03,3
-(~~
-
7' "'MK
i 7' Tpic
+ "'-K 0
FK
i -
-K
WKR °
JK
C 0
K
WKR
0
The matrices and vector functions of (4o56a} are:
(
mi, 0, 0)
0 , mi , 0
0,0,0)
( 0,0,0
0 , 0 , mi 0,0,0
(4o56b)
c,O,O)
Mi=
( J''Ci~ ' -
JL
Ci~Y
JL
, -J:;~,)
JL,
0 '0 '0 '
-Jttx Ciy ' - c~yz
-JLi JLO JL,

0 '0 '0 CiZX ' - Ci~Y ' c,z
2:
(M ijx
Li)
Mb~
J Li
Mijz
and
J L·
Ci~X ' - J L,
CiXY ' - JLi
CiXZ )
( L; J CiY
L; JL'
0
-Jc;yx ' ' - CiYZ
J CiZX
L, J L, JLi
' - CiZY ' CiZ
4.2.4.2 Planar motion. Using, in the planar case for (P f::. C), the vectors
L
Tp,p = ( L L
Xp,p' YP;P' 0 )T L
Fi =
( L L
Fix ' Fiy ' 0
)T '
L
rcp = ( L L
Xcp, Ycp, 0 )T , (4o57a)
)T oL oo)T
L
W LR = (
0, 0, L
WzLR = (0, 0, '1/Jo)T , WLR = (
0, 0, '1/J '
and the rotationangle 'lj; := '1/JLR from R to L around the ezR-axis, provides
the expressions
A RL -L L
OTcpoWLR
o
c~s 'lj; , - sin 'lj; , 0) ( 0 , 0 , y~p ) ( 0)

( sm'lj;, cos'lj;,O 0
0 0 ,-x0 p 0
0
0 0 , 1 -y0L p , x 0L p , 0 'lj;-
0 0 , cos 'lj; y§ P
0
+ sin 'lj; x§ P ) 0
( 0)
( 0 ' 0 , sin 'lj; 0 y§ P - cos 'lj; 0 x§ P 0 0
-y§p , x§p , 0 ;(;
(cos 'lj; 0 y§ P + sin 'lj; 0 x§ P) 0;(; )
( L L 00
(sin'I/JoYcp-COS'Ij;oXcp) o'I/J
0) (
0
c~s 'lj; , - sin '1/J , Y~p )

( sm'lj; , cos'lj;, 0 ° -x 0 p (4o57b)
0 0 ' 1 0
0 ' -?j; ' 0) ( 0 ' -1 ' 0)

wfR = ( ?j; , o, o = 1 , o, o 0
?j;, (4o57c)
0, 0,0 0, 0,0
A RL -L -L L
oWLROWLROTcp
2
cos 'ljJ , - sin 'ljJ , 0 ) ( 0 , -1 , 0 )
( sin '1/J , cos '1/J , 0 1 , 0 , 0
0 0 ,1 0, 0,0
cos '1/J , - sin '1/J , 0 ) ( -1 , 0 , 0) ( x§p )

( sin'I/J, cos'ljJ,O 0
0,-1,0 ° y§p o?j; 2
0 0 ,1 0, 0,0 0
- cos ./,
'P x 0
L p 0
+ sm .!,'P
o 0
YcP
L )
(
- sin 'ljJ x§ P - cos 'ljJ y§ P
0 0
cos '1/J , - sin '1/J , 0) ( x§p )

= - ( sin '1/J , cos 'ljJ , 0 ° y§ P 0 ?j; 2 , (4o57d)
0 ' 0 '0 0
0 , 0 , Y~iP ) ( Fi~ )
-L
Tpip 0
FL
i = ( 0 0 , -x~iP ° Fk
-y~,P , x~iP , 0 0
or
( -F;~ Y~,P: x~,P)· Ff; ·
(4o57e)
or
(Dl (
L L 2 L 2 00
Jcz+m(xcp +YcP) o'l/J
00 )
'
(4o57f)
o
JpoWLR= [ Jcz+mo
L L +YcP
( Xcp L 2 2)] o'ljJoezR,
oo
and
-(~ '-~ '0)

- 'ljJ' 0 '0
0
(
L
00
L 2 L 2 00
)
0 ' 0 '0 Jcz +m 0

Xcp + YcP ) '1/J 0
l
(
This yields the planar equations of motion of an unconstrained rigid body

for P-::/:- C:
( m ' 0) , _ m 0 ( c~s '1/J , - sin '1/J ) 0 ( Yfp )

[ 0 ' m Sln • 1•
'+" '
COS • 1•
'+"
-XL
cp
l
L L cos 'ljJ , sin 'ljJ
- ffio Y
(cp'
-X
cp) 0
(-sin'ljJ,cos'ljJ)'
L
Jcz+mo (xcp +Ycp)
L 2 L 2
Xpo
R ) rE ( pR
pR )
~
1
ooR
00
_ i
Ypo -
(
•Y
y Mtz + (-y~P ' x;;p)

0 [
;j; 0
(Fi~ ' Fitf
m 0 ( c~s '1/J' - sin '1/J ) 0 ( ) 0 x~ ~2 ]
+ [ sm 'ljJ , cos '1/J Ycp , (4o58a)
0
or, written in expanded form for P -::/:- C,
ooR
m 0 Xpo +m ( - cos .J, L .J, L ) .i:
'P 0YcP- sm 'P 0 x 0 p 0 'P
o
(4o58b)
.. R
m Ypo
0
+m (
- sm .!,
o L
'P Ycp 0
+ cos .!,'P 0
L )
x0 p 0
.i:'P (4o58c)
= "L.....J FiyR + m ( sm '1/J '1/J 2 L p

x0 + cos '1/J L ) ,
'1/J 2 YcP
0 0
o 0 0 0 0
and
L
m [( - YcP 0
cos .!, L
'P- x 0 p 0
sm .1,)
o
'P
.. R
Xpo 0
+ ( -YcP
L
0 sm 'P
o .!, + Xcp
L
0 cos 'P .!,) YPO
0
.. R ]
(4o58d)
For P = C, the planar Newton-Euter equations of motion have the decoupled

form
and the expanded form (P = C, L = K)
.. R
m Xco = L.....J ix'
0
"FR (4o59b)
.. R
m Yco 0 = "FR
L.....J iy' an d
4.3 Equations of motion of planar and spatial rigid-body

mechanisms
In this section the model equations of unconstrained and constrained planar

and spatial rigid hoclies and rigid-body mechanisms will be collectedo In the
mostgeneralform they include: (1) DEs ofthe kinematic and kinetic behavior
403 Equations of motion of planar and spatial rigid-body mechanisms 151
of the mechanism, (2) algebraic equations of the kinematic and active con-
straints of the mechanism, and (3) expressions of the associated constraint
reaction forceso
The complete set of these equations and expressions is called the system of
DAEso These DAEs may be written in different forms; eogo, as
jJ = T(p) v o (kinematic DEs) (4o60a)
M(p)ov=f+cf+qa (kinetic DEs) (4o60b)
g(p) =0 ( constraint equations) (4o60c)
Cf= TT(p) g~(p) A 0 0

( constraint reaction forces), (4o60d)
or as
(4o61a)
(4o61b)
(4o61c)
with
Yp(P) T(p)0 0
v= ßc
as the constraint acceleration equation, with
(4o61d)
as the constraint reaction forces and torques, with f as the applied forces and
torques, and q 0 as the centrifugal forces and gyroscopic termso Theoretical
analysis of the DAEs (Equations 4060 and 4061) shows that these two types
of model equations have slightly different analytical and numerical properties
([66], [67], [68], [69])0 The analytical properlies of the DAEs will not be
discussed hereo A few aspects of the numerical solution of DAEs will be
discussed in Beetion 4040 Subsequently the DAEs will be written in the form
(4061)0
4.3.1 Equations of planar motion of unconstrained rigid hoclies in

DE form and of constrained rigid-body systems in DAE
form
Based on the results of Sections 301.1 and 30201 (planar kinematics) and of
Section 40204 (dynamics), the equations of motion of rigid-body mechanisms
under planar motion will be collected in the following order: (1) single un-
constrained planar rigid body, (2) system of unconstrained planar rigid bodies,
(3) single constrained planar rigid body, and (4) system of constrained planar
rigid bodieso
4.3.1.1 A single unconstrained rigid body. Consider a rigid body i

under planar motion, connected to other rigid bodies or to the ground only
by force elements (eogo, springs, dampers, and actuators) and not by joints
(Figure 403)0 Let fi be the vector of all external forces and torques acting on
the bodyo Then the translational and rotational equations of motion of this
unconstrained planar rigid body can be compactly written as follows:For a
reference point Pi f::- Ci on body i (cfo Equation 4058),
(4o62a)
(4o62b)
(4o62c)
(4o62d)
( mi, 0) , _ mi c~s 1/Ji

, - sin 1/Ji ) y~:P, )
0 , mi
>, ,
0 ( 0 (
sml/Ji, cosl/Ji -x0

-xL'
c,P,
) 0 ( cosl/Ji, sinl/Ji)
- sin 1/Ji , cos 1/Ji ,
JL
c,~ + mi 0
( L, 2 L, 2)
xc,P, + Yc,P,
(4o62e)
and
(4o62f)
For a reference point Pi = Ci on body i, the model equations are (cfo Equation
4059):
(4o63a)
with
vi :=Pi =
o .. ("R .. R .i: )T ,
xc;o , Yc;o , 'f'i (4o63b)
4.3 Equations of motion of planar and spatial rigid-body mechanisms 153
...... ---- --
'
'
'
-' '
...
,
I
body j I
'
' '
'I
I
yR
t ranslat ional
translational ground X
damper
F ;3 := F ci2
F ;1 := F dil F ;2 := F eil
Fig. 4.3: Unconstrained planar rigid body i together with its free-body diagram
r(~, , ~J (~) ] ' 0 ' 0 )

0 ,
l
, mi
(0 ' 0) J~~ ' 0 'J~~
(4.63c)
and
q0 i = (0 , 0 , 0) T . (4.63d)
4.3.1.2 System of unconstrained rigid bodies. Consider a system of

nb unconstrained planar rigid bodies connected by force elements (Figure
4.4)0 Taking into account (4o62a) and (4o62b), the equations of motion of this
system are
jJ (4o64a)
Mov (4o64b)
with
p o- P1T , 000 , Pnb
o- ( T )T E lR3nb
Pi =
R
( Xp;Q' R
YP;O' '1/Ji
)T E lR3'
v o- v T
o- T )T E lR3nb
1 , 000 , vnb
( T (xP;O,
oR
vi :=
oR
YP;O, '1/Ji
)T E lR3, 0
q 0 o- q T
o-
0 1
(
, 000 , q T
0 nb
)T E lR3nb ,
and with Vi, Mi, fi, and q 0 ; as defined in (4o62c) to (4o62f), where f contains
all external forces and torques acting on these bodieso
Figo 4.4: System of unconstrained planar rigid bodies
4.3.1.3 A singlerigid body constrained with respect to the base. A

single unconstrained planar rigid body i has three DOFs (3nb = 3)0 It still
can move if it is constrained with respect to the base (Figure 405) by less
than three consistent and independent constraint position equations (nc <
3nb = 3)
(4o65a)
revolute
joint
tran oia-
tional
damper
body i
• +cp iy
~0
• ( p cp
\. I IX
M cl'i
Figo 405: A single constrained rigid body and its free-body diagram
The kinetic equations of planar motion of a constrained rigid body i are
Mi Vi = f i
0
+ cf i + Qc i (4065b)
with
Vi =Pi (planar motion), (4065c)
with external forces and torques fi, with (cfo Equation 3069)
(4065d)
as the vector of all constraint reaction forces and torques acting on the body,
and with
(4o65e)
as the constraint Jacobian matrix of the constraint position equationso Then

the equations of motion of the body are
Mi Pi- g{pi Ai
0 0
= fi + Qci
together with the constraint position equation
(4.66a)
Taking into account the constraint acceleration equation
Yipi (Pi) · Vi = ßci (4.66b)
of (4.66a) yields, tagether with (4.61), the DAEs
(4.67a)
( ~i , g{pi) ( _ Vi) = ( fi ) + ( Qai) , (4.67b)

Yzpi ' 0 A, ßci 0
with ßci as defined in (3.8c). Thesemodel equations are DAEs in Pi, Vi, and
Ai·
4.3.1.4 System of constrained rigid bodies. Consider a system of nb
rigid bodies that are constrained by joints and modeled by nc (nc < 3nb)
algebraic constraint equations (Figure 4.6). For Pi =j:. Ci, and in agreement
with the above introduced notations, its model equations are DAEs in p, p,
and .X. The kinetic equations of motion are
(4.68a)
with
massless revolute
translational revolute link
Fig. 4.6: System of constrained planar rigid bodies

qGi defined in (4o62f),
f := (fJ, ooo , l!JT , fi defined in (4o62e), (4o68b)
with
Cf:= g~ 0A ' Cf:= (eil' 000 'cfnb)T E ocnc and A. E ocnc,
and with the constraint relations

g(p) := (gl (p), 00 o , YnJP))T = 0,
and
(4o68c)
with
Yp := [ (ogifopf, ... '

and
ßc := (ß~1 , 00 0
, ß~nJ T E ocnc o
The DAEs are compactly written in matrixform (cfo Equation 4061)
p= v (4o69a)
( M(p) 'g~(p)) 0 ( iJ) = ( f(p,~)) + (qo(p,p)) 0 (4o69b)

Yp(P) , 0 -.X ßc(P, p) 0
For the special reference point Pi= Ci (4069) yields, together with
P = (P1T, 000 ' PnbT)T ' Pi =

(Xcio
R R
' Ycio ' '1/Ji
)T '
mi, 0 , 0 )
and Mi = ( 0 , mi , ~ 0 ,
0 l 0 l JCi~
and with
T
f = (f1 'oo o'f n
T)T (4070)
and
the system of (6nb + nc) DAEs in the 6nb + nc unknown variables p, v, and
.X, (p E ffi.3nb, ,.x E ffi.nc)
p=v (4o71a)
( M(p), g~(p)) 0
( iJ) = ( f(p, v)) 0
(40 71b)
gp(p) , 0 -A ßc(P, v)
4.3.2 Equations of spatial motion of unconstrained rigid hoclies in

DE form and of constrained rigid-body mechanisms in DAE
form
The model equations of spatial mechanisms of this section will be written in

complete analogy to the planar case of Section 40301.
4.3.2.1 A single unconstrained rigid body. Fora reference point Pi =j:.
Ci on a body i, the following model equations (in DE form) hold (cfo Equation
4055):
Pi= Ti(Pi) Vi 0 (kinematic DEs} (4o72a)

MioVi=fi+qa; (kinetic DEs) (4o72b)
with
Pi= ((r~;o)T' '17T)T E ffi.6 R

rp;o =
( R
Xp;o' Ypio' zp;o
R R )T ' (4o73a)
o ( oR T oT)T (4o73b)
Pi= (rP;o) ''11i '
_ (( oR )T L; T) T (4o73c)
Vi- Tp;Q 'WL;R '
C i2 C i3 , Cil S i3 + Sil S i2 C i3 , S il S i3 - C il S i2 C i3)
A L;R = ( -c i2 s i3 , c il c i3 - s il s i2 s i3 , s il c i3 + c il s i2 s i3 , (4o73d)
s i2 ' - s il c i2 ' c il c i2
(4o73e)
with
Ti(Pi) = Ti('11i) = diag(1 3 , Hi('11i) oARL;), (40 73f)
with c ij := cos llij and s ij := sin llij for llil = 'Pi' lli2 = ei' and lli3 = '1/Ji
for Bryant angles 'Pi, Bi, and '1/Ji of the body i, and with fi, Mi, and Qa; as
defined in (4o55c), (4o55d) and (4o55e)o For Pi= Ci (4o55b) yields
Pi= Ti(Pi) 0Vi (4o74a)

Mi 0Vi = fi(Pi, vi) + q 0 i(Pi, vi), (4o74b)
with Hi and Ti as defined in (4o73e) and (4o73f), and with Mi, fi and q 0 i
as defined in (4o55c), (4o55d) and (4o55f)o
4.3.2.2 System of unconstrained rigid bodies. Consider a system of nb
unconstrained spatial rigid bodies connected by force elementso Taking into
account (4o72a) and (4o72b), the equations of motion of this system are
jJ = T(p) v
o (4o75a)
Mov = f(p, v) +qa(p, v) (4o75b)
with
(4o76a)
and Vi defined in (4o73c), with
M -_ dolag (M 1 ' 000 ' M nb

) E llll6nb,6nb
Jß ' (4o76b)
and Mi defined in (4o55c), with
(40 76c)
and Ti(Pi) defined in (4o73f), with
(4o76d)
and fi defined in (4o55f), and with
(40 76e)
and q 0 ; defined in (4o55d)o

4.3.2.3 A single rigid body constrained with respect to the base.
An unconstrained spatial rigid body i has six DOFs (nb = 1, 6nb = 6)0 It
can still move if it is constrained to the base by less than six independent
constraints (nc < 6nb) with the constraint position equations
with
R R R
Pi= ( Xpio' Ypio' Zpio' 'Pi' ei' '1/Ji
)T = ((rpio)
R T''11iT)T E ~6' (4o77a)
R ( R R R
rP;O = xP;O' YP;O' zP;O
)T (4o77b)
o ( oR T oT)T oR )T L; T) T
Pi= (rp;o) ''11i Vi = (( TpiQ 'WLiR ' (4o77c)
(40 77d)
c i2 c i3 ' c il s i3 + s il s i2 c i3 ' s il s i3 - c il s i2 c i3 )
A L;R = ( -c i2 s i3 , c il c i3 - s i1 s i2 s i3 , s il c i3 + c il s i2 s i3 ,
s i2 ' - s il c i2 ' c il c i2
(4o77e)
( <Pi)
il=
:: '
(40 77f)
and
Ti(Pi) = Ti('11i) = diag(I 3 , Hi('11i) 0ARL;), (4o77g)
where Cij := COSctij, Sij := sinaij for ail ='Pi, cti2 = (}i, and ai3 = '1/Ji for
the Bryant angles 'Pi, Bi, '1/Ji The equations of spatial motion of a constrained
rigid body i for Pi =I Ci (written as DA Es) are (Equation 4061):
(4o77h)
(40 77i)
with
( constraint Jacobian matrix of gi),
( vector of the Lagrange multipliers),
and (cfo Equation 4055)
(40 77j)
(4o77k)
and with
cfi = T'f (Pi) g'fr,, (Pi)
0 0
Ai (constraint reaction forces and torques},
(4o77m)
introduced in (30 76f) and (3077), ßc; (Pi, vi) (right hand side ofthe constraint
acceleration equation) defined in (3o8c)o
4.3.2.4 System of constrained rigid bodies. Consider a mechanism that
includes nb rigid bodies that are constrained by joints modeled by nc (nc <
6nb) algebraic constraint equationso In agreement with the above-introduced
notation, the kinetic model equations and constraint position equations of
this mechanism are:
M 0v = f + cf + q 0 and g(p) = 0, (4o78a)
with
v -_ ( v T T )T E ffi:6nb
1 , 000, vnb , p -_ (p 1T , ooo, Pnb
T )T E ffi:6nb , (4o78b)
R
Tp;o = ( R R R
Xp;o' YP;O' Zp;o
)T ' and
'11i=(rpi,Bi,'l/Ji)T, Pi= ((r~;o)T,i7r)T, vi= ((r~;o)T,wf:RT)T,

(40 78c)
with
Lo ( Lo Lo Lo)T LOR -1 o
wL;R = Wxl' Wyl' Wzl = A • 0Hi 0(17i) 0'11i, (4o78d)
Hi 1(11i) =
1 , 0 ,
( 0 , eil , -sil
Si2
0
ci2
)
( <Pi) (40 78e)
i7 = :: '
0 ' s il ' c il c i2
0
(4o78f)
M := diag(M1, ooo, Mnb) E ffi.6nb, 6nb, with Mi defined in (4o77j),

H(17) := diag (H1('171)' 000 'Hnb(11nJ) E ffi.3nb,3nb,
q 0 := ( q~ 1 , 000, q~nJ T E lR6 nb , with q 0 ; defined in (40771), (4o78g)
f := (!T, 000, l!b) TE JR6 nb, with fi defined in (4o55b) , (40 78h)
cf = TT 0
g~ .X0
(constraint reaction forces and torques),
c (c T c T )T 6nb (4o78i)
f o- f 000' f nc E ffi: '
0 _
1 '
A := (..X1, 000, AnJT E JRnc (Lagrange multipliers), (40 78j)
g~ E ]R6 nb ,nc (constraint Jacobian matrix), (40 78k)
g P (p) 0
v = ßc (p , v) ( constraint acceleration equation), (40781)
and
ßc(P, v) = (ßc 1 , ooo, ßcnJ T E JRnc (right-hand side of Equation 40781)0
Compactly written in DAE form, these model equations are, for Pi =j:. Ci,
jJ = T(p) v o (4o79a)
(
M(p) , TT(p) g~(p))o
0
( v) (4o79b)
Yp(P) T(p) , 0
0 -.X
= ( f(p, v)) + (qa(P, v)) 0
ßc(P, v) 0
4.4 Numerical solution of DAEs- abrief discussion
In this section a few aspects of common numerical DAE solvers will be dis-
cussed brieflyo These may be divided into algebraic and analytical aspectso
They will firstly be discussed for ideal situations (Section 4.401), and then
for more realistic situations (Section 4.402) 0 Before starting this discussion
it should be mentioned that basic properties of DAEs may not be preserved
when mapping time-continuous DAEs into time-discrete DAEs that are used
in the numerical solution processo Among these, their controllability or even
their causality may be lost (eogo, [70]) 0
4.4 Numerical solution of DAEs- abrief discussion 163
4.4.1 Ideal situation

Here a few algebraic and analytical aspects underlying common DAE solvers
will be briefly discussed, considering first an ideal situation which is charac-
terized by the relations (4.81a), (4.81b), (4.81c), (4.81d), and (4.81e).
4.4.1.1 Algebraic aspects. Consider the DAEs written in the form (4.61)
~--------------v-------------~ ~
=:x
(4.80a)
=: b
They may, as a first step, be treated as a "linear" algebraic equation
A·x =b (4.80b)
in the unknown variables jJ, v, and .X, or in x, with
(4.80c)
and
Consider (4.80b) for p and v from a regime, where

det(M(p)) =J 0 (regular mass matrix) , (4.81a)
det(T(p)) =J 0 (regular kinematic matrix), (4.81b)
rank(gp(P)) = nc (independent and consistent (4.81c)
constraint equations),
and for vectors p and v that are consistent with the constraint position
equations
g(p) = 0, (4o8ld)
and with the constraint velocity equations
gp(P) T(p)0 0 V= Oo (4o8le)
Then
det(A(p)) =J 0, (4o82a)
and (4o80a) and (4o80b) have a unique solution
X= A- 1 (p) bo 0
(4o82b)
Proof of ( 4.82a) and ( 4.82b):
The relation (4o82a) holds iff
(4o83a)
This relation is satisfied iff the homogeneaus linear equation
( _::) = ( ~) (4o83b)
has the unique solution
(4o83c)
Multiplication of the first row of (4o83b) by vr yields

oT
vh 0M 0vh- vh
oT 0TT 0YpT 0 Ah
0
' = 00 (4o83d)
The second row of (4o83b) is
Yp 0T 0vh
0
= 0 or oT
vh 0TT 0YpT = oT 0 (4o83e)
Inserting (4o83e) into (4o83d) yields the relation

oT
vh 0
M 0 vh
0
= 0
which, due to (4o8la), implies
(4o83f)
Inserting (4.83f) into the first row of (4.83b) yields
0 + TT. g~. Ah = 0 (4.83g)
and, together with (4.81 b), the relation
g~ · >.h =0
or
nc
L_g'fi, ·Aih = 0 g'Ji, ith column of g~. (4.83h)
i=l
Due to the assumed linear independence of the nc columns g'Ji, of g~ (Equa-

tion 4.81c), (4.83h) only holds if the coefficient vector is zero; i.e., for
>.h = 0. (4.83i)
This proves that the homogeneaus linear equation (4.83b) only has the trivial
solution (4.83c), and that the relation (4.83a) holds. D
The algebraic equation (4.80a) can then be uniquely solved with respect to
x. This provides the local solution
jJ = T(p) · v
(4.83j)
(
f(p, v) + Ya(P, v))
ßc(P, v)
of (4.80a) for a-priori-given vectors v and p, that are consistent with the
constraint equations (4.81d) and (4.81e).
4.4.1.2 Numerical integration step. After having solved the linear al-
gebraic equations (4.80a) as a first step with respect to jJ, v, and ).. for
consistent vectors jJ and v, all taken at a time instant t, a numerical inte-
gration algorithm is applied in a second step to compute (pT, v T)T at the
time (t + 1) from (PT, vT)T at the time t. There is an important difference
between the integration of initial value problems including DEs or DAEs:
the initial values (p;j, v;j)T of a DE can be specified arbitrarily, whereas
the initial values of a DAE must satisfy the constraint position and velocity
equations (Equations 4.81d and 4.81e).
Clearly, the above approach is a simple and crude method for solving DAEs
that does not take into account any modeling error or any error accumulation
in the numerical solution process that may invalidade one or more of the above
assumptions (4.81a), (4.81b), (4.81c), (4.81d), and (4.81e) caused by:
1. An inconvenient engineering model of the mechanism.

2. A local singularity (bifurcation, lock-up) in the constraint equations.

3. A local kinematic singularity in an operation point of the mechanism.
4. An inappropriate choice of the initial conditions of the mechanism that
are inconsistent with (4.81d) or (4.81e).
5. Numerical errors in the discretization process of the time continuous
equations.
6. Numerical solutions of the equations only providing approximations to
the theoretical solutions.
4.4.2 More realistic situations
In this section, two critical situationswill be briefly discussed: (1) a singular

matrix A of (4.80a), and (2) the phenomenon of constraint violation.
4.4.2.1 Singularmatrix A. The matrix A of (4.80a) may be singular due
to an inappropriate engineering model of a mechanism, due to a local sin-
gular kinematic matrix T(p), a local or global singularity in the constraint
equations, or to numerical errors in the algebraic solution process of (4.80a).
There exist well-approved and efficient numerical methods from linear algebra
that provide reliable numerical solutions of the linear equations (4.80a) and
check the regularity or singularity of A. Among those, matrix factorization
techniques like standard Gaussian elimination with full pivoting, LU factor-
ization, singular value decomposition, QR decomposition, or Gram-Schmidt
procedure offer well-approved numerical algorithms to perform this task. As
the matrix A (4.80c) of a rigid-body mechanism usually includes many nul-
lelements (i.e., it is sparse), sparse matrix algorithms have been developed
for handling those systems.
Two such well-tested and documented computational algorithms are part of
the United Kingdom Atomic Energy Authority Harwell science and mathe-
matics library of FORTRAN subroutines ([71]). Theseprograms provide the
capability of finding a trade-off between optimal pivoting and preserving the
sparsity of A in each solution step by selecting suitable values of control pa-
rameters. This algorithm has been implemented in the rigid-body program
NUSTAR ([4]).
4.4.2.2 Constraint violation. On account of the numerical integration
errors, solutions may be obtained that violate the constraint position and
constraint velocity equations (4.81d) and (4.81e). Two main methods for cir-
cumventing this problem, called the coordinate partitioning method and the
constraint stabilization method, will be briefly discussed. These methods use
quite different ideas to control the accumulation of the numerical errors. In
addition, algorithms have been developed that include both of the above
approaches ([40]).
Coordinate partitioning
The coordinate partitioning method was first developed for planar motion
in the computer program DADS-2D ([72]). Following the lines of [72], the
idea behind this method may be briefly summarized as follows (for planar
mechanisms with np = 3nb):
The coordinate vector p E ocnp of a planar mechanism may be partitioned as
(4.84a)
with
as the vector of dependent coordinates and

w E ocnp -nc as the as vector of independent coordinates, associated with u.
Let
(4.84b)
be the integration arrays, defined in terms of the independent coordinates w,

the independent velocities w and the independent accelerations w.
Then the constraint position and velocity equations (Equations 4.81d and
4.81e) can be written as (Sections 3.2.1 and 3.2.2)
g(u, w) = 0 (4.84c)
and
Yu · U = -gw · W, (4.84d)
where (4.84c) and (4.84d) are independent nonlinear algebraic equations in

u and it, respectively. Then, at each integration step YtH is computed from
Yt· Inserting YtH and Yt+l in (4.84c) and (4.84d) provides the remainder
coordinates ut+ 1 and itt+l· In principle, this guarantees that the obtained
solution of the DAEs is consistent with the constraint equations.
As is well known, reliable numerical solutions of the nonlinear equations
(Equations 4.84c and 4.84d) - for instance by means of iterative methods
like a Newton-Raphson algorithm - are critically dependent on a "good"
estimate of the start value of u and it at each solution step.
In addition, the numerical error accumulation severely depends on a
proper partitioning of p into u and w. It may even be necessary to switch
from a chosen set of independent coordinates Wlf to a different set w 2t+ 1
during the integration process. For example, this may be necessary if the
number of iterations in the integration steps of the Newton-Raphson process
keep increasing from one time step to another.
Based on these criteria, automatic techniques for partitioning the coordinate

vector p into u and w have been developed. They again use matrix factoriza-
tion techniques, like those mentioned in Section 4.4.2.1. Then the selection of
the dependent coordinates may depend critically on ([41]): the physical unit
system chosen in the theoretical model equations, the type of pivoting, and
the method of matrix factorization.
Further information on the numerical solution of DAEs can be found in [73],
[74], [75], [76], [77], [78], [79], [80], and [81].
Constraint stabilization
The constraint stabilization method is another method to avoid numerical
"solutions" of DAEs that do not satisfy the Equations (4.81d) and (4.81e).
This method was introduced by Baumgarte in [82].
The idea behind this method may be described as follows: The DAE
model (4.80a) of a rigid-body mechanism includes the constraint acceleration
equation
gP · T(p) · v := 0 or jj = 0. (4.85a)
This relation is a linear DE of second order in g with 2nc eigenvalues that
are located at the origin of the complex plane. Stability theory tells that
such a system is extremely sensitive with respect to all types of disturbances
and to the accumulation errors in the numerical solution process. In order
to improve this situation, Baumgarte replaced the constraint relation (4.85a)
by the extended relation
jj + lil g + li2
0
° g =0 (4.85b)
with free coefficients li 1 , li 2 E JR1 . The relation (4.85b) is a linear DE of
second order in g. Its 2nc eigenvalues can be placed at arbitrary positions in
the complex plane by apropriate choices of ti 1 and ti 2 . This provides stable
solutions g(t) of (4.85b) that converge asymptotically towards the desired
=
equilibrium solution g 0. As a consequence, (4.85b) is called the stabilized
constraint acceleration equation. This equation may be written in the form
Yp · T · v = -lil · g- li2 · g + ßc
or (4.85c)
Yp. T. v= -lil. Yp. T. v- li2. g + ßc(P' v).
=: ßc(P, V, lil, li2)
Combining the kinematic and kinetic DEs with (4.85c) provides the modified
DAEs
16nb '
( 06nb,6nb ,
06nb,6nb
M(p)
' 06nb,nc
, TT(p) · g~(p)
)
·
( jJ)
v
Onc,6nb , Yp(P) 'T(p) , Onc,nc -A
T(p) ·v )
= ( f(p,v}+Ya(p,v), (4.85d)
ßc(P, v)
with
ßc(P, V, "'1, "'2) := -1\:,1 · Yp(P) · T(p) ·V- "'2 · g(p) + ßc(P, v).
(4.85e)
These differ from the normal DAEs (4.80a) by a modified right-hand side of
the constraint acceleration equations.
Comment 4.4.1 (Constraint stabilization method): The constraint

stabilization method has been successfully implemented in the rigid-body
program NUSTAR ([4]). Various applications of this program to rather com-
plicated rigid-body mechanisms like off-road vehicles (trucks and tanks) un-
der extreme driving maneuvers have provided stable and reliable numerical
results, when the stabilization parameters "' 1 and "' 2 were chosen carefully.
The constraint stabilization method is an intuitive appealing ad hoc
method. Though its functioning seems to be obvious, its theoretical justi-
fication is not trivial. A description of a basic problern behind this procedure
was the follows.
The stabilized constraint relations are cross-coupled to the remainder
kinematic and kinetic DEs in a closed loop. It must then be expected that
the artificially introduced and modified eigenvalues of a subsystem (constraint
equations) in the closed loop will also shift the eigenvalues of the remainder
subsystems (kinematic and kinetic model equations). This would imply that
the numerically motivated eigenvalues - that do not have any physical mean-
ing - would shift the eigenvalues of the subsystems with a clear-cut physical
meaning. If this would happen, the constrained stabilization method would
not be reliable.
A theoretical analysis of the local dynamic behavior of the DAEs (4.85d)
shows that here a separation principle holds that (to a first approximation)
guarantees that the numerical and physical eigenvalues of the DAEs do not
influence each other ([83]). This result provides a formal justification of the
constraint stabilization method.
This situation is formally equivalent to a linear controlloop that includes
a linear observer. As a consequence, the stabilization parameters "' 1 and "'2
and the resulting eigenvalues can be chosen by direct analogy to the choice
of the eigenvalues of an observer in a control loop. The eigenvalues of the
stabilized constraint relations shall provide well-damped transients of g(t).
The absolute values of these eigenvalues should be chosen to be roughly three
times larger than the absolute value of the largest significant eigenvalue of
the kinetic and kinematic subsystems of the mechanism.
5. Model equations of planar and spatial joints
Joints prevent a body from either moving along one or several axes, or from
rotating araund one or several axes, or both. As a consequence, joints reduce
the number of DOFs of a body or of a mechanism and thereby constrain
their motion. Therefore mathematical models of joints are called constraint
equations ( constraint position, velocity, or acceleration equations). Theoret-
ical models of joints (between a body and the ground) that constrain the
motion of a body with respect to the ground are called absolute constraint
equations, whereas models of joints (between two bodies) that constrain the
relative motion of two bodies, are called relative constraint equations.
Joints are described by geometric relations between absolute (global} Coordi-
nates and/or relative (local} coordinates and body-fixed vectors and orien-
tations. These geometric relations are included in the model equations of a
mechanism by means of suitably chosen projections and representations of
vector loop equations and/or orientation loop equations.
Vector and orientation loop equations may be written using quite different
notations, some of which are suited to model specific joint types. It is of
common practice to differ between absolute and relative joint model equations.
The constraint equations of absolute joints are always modeled by absolute
generalized coordinates. The constraint equations of relative joints may in-
clude relative coordinates:
1. If these relative coordinates are unconstrained and appear in models of
external forces (springs, dampers, actuators), they are calculated accord-
ing to some compliant relations.
2. If the relative coordinates appearing in vector loop equations are con-
strained by geometric joints, they are either eliminated (by applying
special projection operators to the constraint equations), or isolated (by
applying alternative projection operators to the constraint equations) if
they are needed for monitaring or control purposes.
To illustrate the above loop equations, consider the simple example of a vector
loop equation and of an orientation loop equation between two bodies i and
j in Figure 5.1. The Vector loop equation (represented in frame R) is:
172 5o Model equations of planar and spatial joints
Figo 501: Geometrical representation of a vector loop including two bodies
R
0= rpilo + ....__,_....
ARLil 0 rLi1
...._,_,__..,
Pi2Pi1
(5ola)
'-v-"
absolute free constant
free vector orientation vector
R A RLjl rLjl _ A RLj 1 ALj1Lj2 rLj2

,____.......
0
rpjlo
0
..._"_,__.., Pj 2 Pj 1 ...._"_,__.... Pj3Pj2

~
free ,____....... fixed
absolute orientation constant
vector orientation constant
vector
free vector
A Lj2Lj3 0 rLj3
...._"_,__.... Pj4Pj3
fixed ,____.......
orientation constant
vector
The orientation loop equation (with respect to frame R) is:
0ALj4Lj3 0 ALj3Lj2 0 ALj2Ljl 0 ALjlR
....__,_....
ARLil 0 ALi1Li2
"____.......
0 ALi2Lj4
"____....... ..._"_,__..,
free fixed free relative fixed free
orientation orientation orientation orientation orientation
= 130 (5olb)
In the spatial case, each unconstrained rigid body i has six DOFs: three
rotational and three translationalo As a consequence:
1. A vector loop equation including three independent scalar equations may
constrain one, two, or three translational DOFs of a body or of a mech-
anism (this may be achieved by suitable projections of the vector loop
equation)o
20 An orientation loop equation may constrain one, two, or three rotational
degrees of freedom (this may again be achieved by suitable projections of
the orientation loop equation) 0
5.1 Theoretical modeling of planar joints 173
If, for example, a single translational DOF is to be constrained, then either

of the following my be used: a suitable projection of the vector loop equation,
or a scalar product of a vector inside the vector loop with itself, or a function
depending on the above equations of the vector loop, or a scalar function
depending on one or both of the above scalar equations of the vector loop
equations and of the orientation loop equation. The above loop equations
provide a basis for deriving systematically algebraic constraint equations for
a broad dass of different joints. In particular, they suggest the use of scalar
loop equations and scalar functions built from several of those as building
bloeks, from which a large number of common joint models can be set up.
This has been done, to a certain extend, in [4], [40], and [41] (compare also
the spatial case of Beetion 5.2.1).
Besides standard joint models, models of so called massless links (eompos-
ite joints) play an important role in rigid-body dynamics. Various complex
mechanisms from industrial practice include rigid bodies or groups of rigid
bodies that are characterized by two properties:
1. Their only function within the mechanism is to suitably connect other
bodies.
2. Their mass as well as their moments and products of inertia are much
smaller than the inertia parameters of the adjacent rigid bodies.
These eouplers of adjacent bodies are often considered as eomposite joints
without inertia properlies and are theoretically modeled by kinematic con-
straint equations. In Beetion 5.1, constraint equations of planar joints will
be derived using the above-discussed approach. Among those, several abso-
lute and relative planar joints will be discussed together with some technical
realizations. In Beetion 5.2, constraint equations of several building bloeks
of spatial joints will be derived as a first step. In a second step, composi-
tions of these building blocks will be set up that provide constraint equations
and relative coordinates of several spatial joints that are commonly used in
rigid-body systems.
5.1 Theoretical modeling of planar joints
In the planar ease the number of different joint types is limited. To illustrate
the previous discussion, eonstraint position equations (and for several joints,
the associated eonstraint veloeity and aeeeleration equations) will now be
derived for a number of eommon planar joints. Among these two classes
of kinematic constraints are of common interest: absolute planar eonstraints
between a body and the base (Beetion 5.1.1) and relative planar eonstraints
between two bodies (Beetion 5.1.2).
5.1.1 Absolute constraints

A rigid body moving in a plane may have up to three DOFso Absolute planar
constraints of a body are those that occur between a body and the base
(inertial frame R) 0 They constrain absolute coordinates of the body with
respect to frame Ro Among various absolute constraint situations of a body
in a plane, six commonly used configurations will be discussed (cfo Figure
502):
1. Absolute orientation constraints ( massless translational links - two re-
maining translational DOFS)o
20 Absolute partial-position constraints, ( massless revolute-translational
joint, or pin and slot mechanism - one remairring translational and one
rotational DOF)o
30 Absolute complete-position constraints ( revolute joint - one remairring
rotational DOF)),
40 Absolute orientation and partial-position constraints ( translational joint
- one remairring translational DOF),
50 An absolute constant-distance constraint (massless revolute-revolute link
- two remairring DOFs)o
60 A combined absolute orientationjpartial-position constraint (ball rolling
on the ground or pulley- a single remainder DOF)o
5.1.1.1 Position constraints between a body and the base. Absolute
X-position andjor y-position constraints require that the difference of the X-
coordinates andjor of the y-coordinates between a point Pi on body i and a
point Qi on the base are kept constanto Then (Figure 503)
R
Xpio- R - Cx-
xQio- - cons t ant (x-position constraint} (5o2a)
and/or
R
YPiO R
- YQiO -- cy -- cons t ant (y-position constraint}o (5o2b)
501010101 Partial-position constraint ( massless revolute-translational link) 0
An absolute-position constraint that only includes one of (5o2a) and (5o2b) is

called (Figure 5.4) an absolute partial-position constraint (massless revolute-
translationallink}o A mathematical model of this joint is obtained as followso
Consider the vector diagram of a body i connected to the ground by a mass-
less revolute-translational link as in Figure 5.4ao The line of translation of
the body is defined by the vector ti between two noncoinciding points Qi and
Si on the body io The revolute joint has a constant distance c from the line
of translation of the body io The constraint position equation of this link is
computed from the scalar product
(5o3a)
R
d oi R
= rQoQi
Typ of (absolute)
Jointtyp 'D chnical realization
planar con traint
airplane ~
no joint
no constraint
pseudo-joint
~~
Lh
test facility
partial-
position
constraint
translational
link;
pin and slot
eh
massless revolu e-
vertical
V hi I
mod I
u~~?!
complete- revolu e
position joint
constraint
I
- pendulum unbalance
orthogonal • • mechanism ~
massless • • fürpure
••
ori ntation
orthogonal
LI
on t rai nt t ranslational
links • • tran~l ation al :
• • motiOns
orientation and
t ran lational translational
partial-posit ion me hani al
.0:: bfL
constraint joint
o illator
combined r':l'c~ and ~

ori ntation/ ptmon;
:a:
partial-po it ion rolling • -
on traint wheel ..
rack and om1on
constant- massless
distance revolut spe ial
wh 1
constraint revolute
susp n IOn
link
Fig. 5.2: Common absolute joints

176 5. Model equations of planar and spatial joints
Fig. 5.3: Drawing associated with (5 .2a) and (5 .2b)
and
R = ( 0 ' - 1) (orthogonal rotation matrix).

1 ' 0
Together with the vector loop equation
R R
rQoo-r pio -
ARL L·
' ·rQ:Pi -
dR
oi = 0' (5.3b)
the constraint position equation is
(5.3c)
or
(5.3d)
or, together with
A RLi ·RT ·A LiR -_ (cos 7/J, -sin 7/J ) · ( 0, 1) · ( cos 7/J , sin 7/J )
sin 7/J , cos 7/J -1 , 0 - sin 7/J , cos 7/J
= ( 0' 1)
-1 , 0
= RT
'
(5.3e)
massless revolute-
translational link
~~~--;;;:.1---- line of tran lation
of body i
(a) Vector diagram of an absolute massless revolute- translational link
pin and slot
(b) Technical realizations of mechanisms that include a massless revolute-

translational link
Fig. 5.4: Vector diagram and technical realizations of mechanisms that include an
absolute massless revolute- translational joint
it is
1 0
= II R T
R
T ( R R ARL L
=
0 0
)
g llrsiQi 0 (rsiqJ 0 0 rQoO- Tpio- '0 rQ:Pi - c 0
(5o3f)
The constraint velocity equation associated with (5o3f) is
1 ( R )T RT ( R Ao RLi Li ) _ O
o _
g - llrsiQi II 0 rsiQi 0 -rpio-

o
0 Tqipi - 0
It may be written, together with (Aol.llc)
Ao RLi == ARLO
~-wL:R==
-Lo ARLO
t.
R "«.p
.i. (5o3g)
as
(5o3h)
R
Yp ( P ) -- ( - ( rsiQi )T RT 0 ' -
(rsiQi
R )T RT 0 0
ARLi 0
R 0
Li )
rQiPi 0
(5o3i)
The constraint acceleration equation is obtained from
jj=O
as (5o3j)
0= (rRs.oq,o)T.
. . RT. ( -rpio-
.. R A. RLi · R L .i. ARLO'·
·rq:pi·'P- R Lo
·rq:pi''P .i:)
or as
(5.3k)
=:ßc
(cf. Examples 3.3 and 3.4 of Volume 11).
501 Theoretical modeling of planar joints 179
501010102 Complete-position constraint (revolute joint)o An absolute position

constraint that is modeled by both (5o2a) and (5o2b) is called an absolute
complete-position constraint (revolute joint)o This joint is modeled in more
detail by a vector loop equation
(5.4a)
(Figure 5o5a), that together with the geometrical relations
A
RL
' = ( cos 1/J; , - sin 7/J; ) L
r P;P
' = ( TJx )
= constant ' and rp,o
R
= ('"'"'/yYx)
sm .!,
o
'l'i , cos •'l'i
1, '
-T]y
(5.4b)
yields the constraint position equations
x~ 0 + cos '1/J; o'f/x + sin '1/J; o'f}y - 'f'x = 0

and (5.4c)
Y~o + sin '1/J; 0'f/x - cos '1/J; 0'f}y - '/'y = 00
The constraint velocity equation, associated with (5.4a) and (5.4c), is
Together with
(5.4d)
this provides the equation
(12, +ARL, oRor~:P) o (r~o) = Oo (5o4e)
=: Yp(P)
The associated constraint acceleration equation is

R 0 RL· L 0 RL L 0
r ~0 +A ' 0•'I'1• + A
' 0R 0r ~p ' 0R 0r ~p
' 0.1.
'I' = 0
or
(1 2 , ARL, oR o r~:P) o ( r~o) = -ARL, oR 2 o r~:P o~ 2 0 (5.4f)
=: Yp(P) =: ßc
Absolute complete-position constraints are technically realized by revolute

joints (Figures 5o5b and 502, and Example 304 of Volume II)o
(a) Vector diagram of an absolute complete-position constraint (revolute joint)
general rotor special rotor
P;
stiff rod pivotecl in poi nt Q;
cx ·cntcr Ioad
pcnd ulum
(b) Technical realizations of mechanism that include revolute joints
Fig. 5.5: Drawings of (absolute) revolute joints

5.1.1.2 Orientation constraint (massless translational link). An ab-

solute orientation constraint requires that the orientation tPi = tPL;R of a
body i is kept constant with respect to the base (Figures 5.6a and 5.2). The
associated constraint equation is
tPi = c.p = constant. (5.5)
Technical realizations of this constraint are combined massless translational
links (Figure 5.6b and Example 3.1 of Volume II).
(a) Absolute orientation constraint of a body i
(b) Technical realization of an absolute orientation constraint
Fig. 5.6: Drawings of an absolute orientation constraint
5.1.1.3 Orientation and partial-position constraint (translational

joint). A simultaneaus absolute orientation and partial-position constraint
is technically realized by a translational joint between a body and the groundo

This constraint requires that the orientation of body i with respect to the
base is kept constant,
'1/Ji = c,p = constant, (5o6a)
and that, for example, the x-coordinate x~;O of this body is constant,
R
xP;O = Cx = constanto (5o6b)
M echanisms that include translational joints are shown in Figures 50 7 and
502, and in Example 302 of Volume II.
A moregeneraland more refined constraint equation than (5o6b) is obtained
as followso Consider Figure 5o7a with Q 0 , S0 and Qi, Si as noncoincident
points on the base and on the body i, respectivelyo These four points are
located on a common straight line that is chosen as the direction of the
translation of the body i with respect to the baseo Then the vectors r SoQo
and rs;Q; must remain collinear, as well as the vectors
diO := r siQo and to := r SoQo,
where diO is assumed to be non zeroo The above requirement can be written
as
(t~)T diO 0
=0 (orthogonality relation) (50 7a)
or as
(R t 0R)T0
°
R
diO = 0 (5o7b)
with
( toR) j_ := R tRo 0
an d R = ( 01 ,' -1)
0 0
Taking into account the vector loop equation (Figure 5o7a)

R
0 =rQao+ dR
iO-
ARL '0 ( rs;Q;+rQ:P;
L0 L 0 ) -rP;O
R 0
yields (5o7c)
R = ARL '0 ( rs;Q;
d iO L0 0
L0 )
+ rQ:P; R
+ rpio- R
rQoOO
This provides the constraint position equation
0 -_ ( (taR) T0RT) 0 [ARL; 0 ( rsiQi

L; + rQiPi
Li ) + rpio-
R R ]
rQo (5o8a)
and
'1/Ji = c,p = constant (5o8b)
of an (absolute) translational jointo The constraint velocity and acceleration
equations associated with (508) are obtained by analogy to the previous cases
(see also the various planar mechanisms of Volume II)o
(a) Vector diagram of a translational joint
--- t
(b) Technical realizations including an absolute translational joint
Fig. 5.7: Drawings of translational joints
5.1.1.4 Combined orientationjpartial-position constraint. A wheel

rolling on an inertial plane without slippage is an example of an absolute
combined orientationjpartial-position constraint (Figure 5.8) . This constraint
is modeled by the constraint velocity equation
·R .
x 00 +R· 'l/J = 0 (5.9a)
that holds at the contact point A between the wheel and the base or by the
associated constraint position equation
X~o + R · 1/J - Cx = 0 Cx = constant (5.9b)
with R as radius of the wheel. Further technical situations that include this
type of constraint are shown in Figures 5.8b and 5.8c, (see also Example 3.3
of Volume II).
(a) Wheel rolling on an inertial plane
(b) Rack and pinion mechanism (c) Mass-spring pulley
Fig. 5.8: Mechanisms that include a combined absolute orientationjpartial-position

constraint
5.1.1.5 Constant-distance constraint (massless revolute-revolute

link) . An absolute distance constraint between point Pi on a body i and
point Qi on the ground is modeled by the constraint equation
T
r Pi Qi · r Pi Q i =
d2 = constant, (5.10a)
with das the constant distance between points Pi and Qi (Figure 5.9). Taking
into account the vector loop equation
R
Tpo + ARLi . T Lpii p R R -
- T pi Q i - ToQ i - 0 (5.10b)
with
(X~o) cos tPi , - sin tPi ) '

TR
p o -_ R A RL '. =
(
Yp o sin tPi , cos tPi
Li p
rpi = (x~:P
0
) = constant, and R
rQiO = ({x ) = constant
/y
yields
{x + X Rp o + COS .lf'!,i Li
· Xpi p -
_ R
X pi Q i
and
R + · . !,
+ Ypo Li _ R
/ y sm 'l'i · X pi p - Y Pi Q i'
and the relation
massless revolut
revolute link of
length d
Fig. 5.9: Vector diagram of a mechanism with a constant-distance constraint (mass-

less revolute- revolute link)
and finally the constraint position equation

2] l / 2
_ [(
g- R Li
"fx+Xpo+XpipoCOS'lj;i ) 2
+ ( "(y+Ypo+Xpiposm'l/Ji
R Li o ) _
-d-0
(5o10c)
which is a function of the generalized coordinates x~E' yjt0 , and '1/Ji of the
body, and ofthe system constants "fx, "(y, x~;p, and Ypiipo Massless revolute-
revolute links are technical realizations of constant-distance constraints (see
also Example 304 of Volume II)o
5.1.2 Relative planar joints between two borlies
Relative constraints between two bodies i and j constrain relative motion of

these bodieso Two rigid bodies (i and j) moving under relative constraints in
a plane may tagether have between one and five degrees of freedom (DOFs)o
By analogy to the previous absolute constraints of a planar rigid body, the
following relative constraints of two rigid bodies moving in a plane or in
parallel planes are in common use:
1. Partial relative position constraint ( massless revolute-translational link
between two bodies) 0
20 Complete relativeposition constraint (revolutejoint between two bodies)o
30 Relative orientation and partial-position constraint ( translational joint
between two bodies) 0
40 Relative orientation constraint (combined translationallinks between two
bodies)o
50 Relative constant-distance constraint ( massless revolute-revolute link be-
tween two bodies) 0
60 Combined relative orientationjpartial-position constraint ( rack and pin-
ion between two bodies) 0
5.1.2.1 Position constraints. A relative X-position (or y-position) con-
straint requires that the difference between the X-coordinates (or y-coordi-
nates) of a point Qi on the body i and of a point Qj on the body j is equal
to a given constant c" (or cy); ioeo,
Lo - xj'
xi' Lo = c" ( re l ative X-position constraint) (5olla)
or
yfi - yfi = cy (relative y-position constraint)o (5ollb)
5010 20101 Partial-position constraint ( massless revolute-translational link) 0
A single of these two constraint equations provides a relative partial-position

constrainto It is technically built by a massless revolute-translational link,
as shown in Figures 5010, 5011 and 50120 More refined mathematical model
equations of a massless revolute-translationallink are obtained as followso
Type of (relative)
planar con traint Joint type Technical realization
no join IJ\Q~oad
no constrain ~able
p. eudo-join
G body l
partial-
position
con traint
'/6)'
ma: I revolut
tran. lational link roll r rig
~~
cart loaded ~
complete- by a
po ition pendulum .. -
con. traint
body 2
r volute joint excavator ~
orientation
constraint
m~~
with two r lative
,t:2
translational links
orientation and
partial-position
~~ody2
constraint
t ra:atiw . 1 t freedom mechani m
0 y.
j~ b~dy
~
combined
1
orientation/ CJi}ody 2
partial-position
con. t raint
rack and pinion cam foliower
constant-
di tance
con traint massless revolute-
r volutelink
Fig. 5.10: Common relative joints

constant
mas Je s revolu te-

translational link
(a) General massless revolute-translational link
(b) Special massless revolute-translational link
Fig. 5.11: Two types of massless revolute- translational links

L;
L
XQ;Q, = constant
body 3
bo ly 1
Fig. 5.12: Mechanisms (wheel Suspensions) including relative partial-position con-

straints (massless revolute- translational links)
Consider the vector diagram of Figure 5.11a showing a massless link between
two bodies with a revolute joint on body i and a translational joint on body
j. The line of translation of body j is defined by two noncoinciding points Qj
and Sj on this body. The revolute axis is located in point Qi on body i. The
revolute joint is a constant distance c from the line of translation. Then the
constraint position equation of this link can be written as the scalar product
(d~) T · ARL j · t~ij_ / 11ti 11 = c (c = constant relative distance) (5.12a)
with
t iR = r sjQj
R t Ljj_ - R. t Lj - R. Lj
i - i - r sjQj ,
R = (0 ' - 1) ( or·thogonal mtation matr-ix) , and R · A RL j = A RL; · R

1' 0
This provides, together with the vector loop equation
the constraint position equation

(5ol2h)
501020102 Complete-position constraint (revolute joint) 0 The ahove two con-

straint equations (5olla) ancl (5ollh) together clefine a complete relative-
position constrainto This constraint is technically huilt hy a revolute joint
shown in Figure 5ol3a, with technical realizations shown in Figures 5ol3h
ancl 50100
More refinecl constraint position equations for a revolute joint at a common
point Qi = Qj of two hoclies i ancl j are ohtainecl from the vector loop equation
(Figure 5ol3a)
g = TR
PiO
+ A RLi 0 TLi - A RL; TL;
QiPi
TR - 0
Q;P; - P;O -
0
0
(5ol3a)
Together with the geometry relations
ARL" = ( cos'lj;", , - sin 'lj;",) (5ol3h)

sin 'lj;",
cos'lj;",
Ii = i' j,
ancl
Lo
rt
QiPi
=( Ax)
O
Lo'
TQOPO- - ( -TJx)
' (5ol3c)
' ' 'T]y
this provicles the constraint position equations of the revolute joint hetween
the hoclies i ancl j representecl in coorclinate form:
x~i o - x~; o + Ax o cos 'I/J1 + TJx o cos 'I/J2 + 'T]y o sin 'I/J2 = 0
ancl (5ol3cl)
Y~o- Y~;O + Ax 0 sin'lj;1 + TJx 0 sin'lj;2- 'T]y 0 cos'lj;2 = Oo
The constraint velocity equations associatecl with (5ol3a) are
oR Ao RLi Li Ao RL; L; oR 0
Tpio + 0
TQiPi + 0
TQ;P;- TP;O =
or (5ol3e)
Tpio
oR)
(I 2 '
A RLi 0R 0TLi
QiPi '
- I
2 '
A RL; 0R 0TL; ) 0
Q;P;
( R
TP;O
o
'1/Ji =00
=: Yp(P) ~j
The associatecl constraint acceleration equations are

(a) Vector diagram of a (relative) revolute joint
double pendulum coupled roto ·
cart loaded by a pendulum

(b) Technical realizations of planar mechanisms that include revolute joints
Fig. 5.13: Drawings of revolute joints between two bodies

(5ol3f)
=:ßc
5.1.2.2 Grientation constraint (massless translationallink). A rela-
tive orientation constraint between two bodies i and j implies the relation
(Figure 5ol4a)
'1/Ji- '1/Ji = c'l/J = constant, (5ol4a)
with
'1/Ji := '1/JLiR (5ol4b)

and with
'1/Jij := '1/Ji- '1/Jj (5ol4c)
as the relative angle between the two bodieso It is related to the orientation
loop equation
(5ol4d)
or
ALiLj = ALiR ARLj' 0
(5ol4e)
with
or
cos'l/Jij := cos'I/Ji 0 cos'lj;j + sin'I/Ji 0 sin'lj;j = cos ('1/Ji- '1/Ji), (5ol4f)
which implies the relation

'1/Jij='l/Ji-'1/Jjo (5ol4g)
A more general relative orientation constraint is obtained by mechanisms

that include special gear sets (Figure 5015)0 The constraint equation of this
mechanism is obtained by the velocity balance for rotations without slippage:
r1 ° ~1 = -r2 ° ~2 or r1 ° ~1 + r2 ° ~2 = 0
or
with 'lj; 10 and 'lj; 20 as initial orientationso

/
/
/J,-'1/J,
/
= '1/J,; 1-j ll
- - '1/J,
---------
A~~~"......::.:--
,
(a) Orientation loop associated with a relative orientation con-

straint equation (5 .14d)
(b) Technical realization of a relative orientation

constraint
Fig. 5.14: Drawings of a relative orientation constraint
5.1.2.3 Relative orientation and partial-position constraint (trans-

lational joint). A relative orientation and partial position constraint is tech-
nically realized by a translational Joint between two bodies i and J (Figures
5.16 and 5.10 ). The constraint position equations of the translational Joint
are obtained from the requirements that this joint does not allow relative
rotations of the two bodies i and J; i.e.,
Figo 5015: Mechanism including a gear set
'1/Jij := '1/Ji - '1/Ji = c'l/J = constant

or (5o15a)
L L . ( cos '1/Jij , - sin '1/Jij )
A ' = = constant
0
sin '1/Jij , cos '1/Jij '
but only relative translations along a common axis, where the latter require-
ment can be mathematically modeled as followso Consider the geometrical
situation of Figure 5o16a with noncoincident points Si, Qi, and Sj, Qj on
the hoclies i and j, respectively, located on a common straight line that de-
fines the direction of the relative translation of the bodieso Then the vectors
ti := rsiQi and rs;Q; must remain collinear as well as the vector dji := rs;Qi
where the vector rs;Qi is assumed tobe nonzeroo The above requirement can
be written in terms of the orthogonality relation
or (5o15b)
(R 0 tLi)T
•
0 dl:.i
]2
= 0
with
tfij_ := R tfi0
and R := ( 01 : -01 ) 0 (5o15c)
Taking into account the vector loop equation (Figure 5o16a)

0
R
Tpio + TQiPi
R
+ dRj i - Ts;Q;
R
-
R
TQ;P; -
R
TP;O =
yields
(a) Vector diagram associated with the constraint

equations of a translational joint
revolutc
torsional spring
and damper
body 2
(b) Mechanism including a translational joint be-

tween two bodies
Fig. 5.16: Drawings of (relative) translational joints

This provides, together with (5o15c) and (5o15a), the constraint position equa-
tions
(5o16a)
and '1/Ji- '1/Jj = c,p = constanto (5o16b)

5.1.2.4 Combined orientation/partial-position constraint. Combined
relative orientationfpartial-position constraints are theoretical models of
rack-and-pinion mechanisms and cam followers (Figure 5017) 0 For the rack-
and-pinion mechanism of Figure 5o17a, the noslippage condition guarantees
identical velocities of bodies 1 and 2 at their contact point Ao Then
(5o17a)
with R as the radius of the piniono This provides the constraint position
equation
Y~1o- Y~1 o(O) = -R ('lj;2- 'I/J2(0)) (5o17b)
with y~10 (0) and 'I/J2(0) as starting conditions of y~10 and 'I/J2o
5.1.2.5 Constant-distance constraint (massless revolute-revolute
link). A constant-distance constraint between two points Pi and P j, located
on bodies i and j, respectively, is modeled by the constraint equation
- T
g-TpipjoTpipj- d2-o
-' (5o18a)
with das the constant distance between the points Pi and Pj (Figure 5018)0
Using the vector loop equation
0 = TR
P10
- TR
P20
+ A RLi 0TLi
PiP1
+ A RLi 0TLi
PjPi
+ A RLj 0TLj
P2Pj
and the geometry relations
A RL ~ = ( cos'lj;"' , - sin'lj;"')
Ii = i' j, (5o18b)
sin 'lj;"' cos'lj;"'
and
yields
L0 0 .!, L0 0 .!,
Yp;pi 0 sm 'Pi- xp;pi 0 cos .!,
'Pi =
R R
Xp1 o- Xp2 o + cos .!,'Pi 0fix+ sm 'Pi 0fiy
+ cos'lj;j orJx + sin'lj;j o'T]y

ground
(a) Rack-and-pinion mechanism
(b) Cam foliower
Fig. 5.17: Technical realizations of joints, modeled by a combined relative

orientationjpartial-position constraint
and
- x~>i · sin 7)!i - Y~>i · cos 1)!i = Y~1 o - Y~2 o + sin 7)!i · "'x - cos 1)!i · "'v
+ sin 7)!j · 1Jx - cos 1)!j · 1Jy. ( 5.18c)
This provides the constraint equation

ma I rcvolutc- rcvolutc link
Fig. 5.18: Vector diagram of a mechanism with a constant-distance constraint
(5.18d)
with x~>; and y~i,P; obtained from (5.18c). Massless revolute- revolute links
are technical realizations of constant-distance constraints.
Comment 5.1.1 (Examples ofplanar joints): The above-discussed mod-

els of joints between two bodies under planar motionwill be extensively used
in the various examples of Section 4 in Volume II.
5.1.3 Pseudo-joint and forcejtorque elements
The various forces and torques applied to rigid bodies act in specific direc-
tions or around specific axes. Theoretical models of these forces and torques
sometimes implicitly include joints or "pseudo-joints" (cf. Beetion 6).
5.1.3.1 Example of a translational spring element. Translational
springs are assumed to act in the direction of a straight line, connecting
their attachment points to the bodies (Figure 5.19a). Technically this can be
achieved by connecting the attachment points of the spring by a massless
revolute-translational-revolute link as a "pseudo-joint" that does not con-
strain any DOF of the bodies but forces the spring to act in the desired
direction through the attachment points (Figures 5.19a and 5.19b).
5.1.3.2 Example of a torsional spring. Torsional springs are always as-
sumed to act around the axis of a revolute joint connecting two bodies. In
contrast to the "pseudo-joint" of Section 5.1.3.1 , this planar joint constrains
two translational DOFs of adjacent bodies (Figure 5.19c).
points of a
T
line of
action of a
translational translational
spring (revolute joints) spring
(a) Line of action of a translational spring
translationallink
revolute
joint
"pseudo-joint"
(b) Model of a translational spring including a spring force element and
a "pseudo-joint", which does not constrain any DOF of the adjacent
bodies but prescribes the line of action of the spring
torsional spring
(c) Model of a torsional spring acting around the axis of a revolute joint
that constrains two relative translational DOFs of the adjacent bodies
under planar motion
Fig. 5.19: Combinations of planar translational and torsional springs with joints
Comment 5.1.2 (Spatial torsional spring): Spatial as well as planar

torsional springs are always assumed to act around the axis of a revolute joint
connecting two bodies. In contrast to the "pseudo-joint" of Section 5.1.3.1,
this joint constrains in the spatial case three translational DOFs and two
rotational DOFs of the adjacent bodies.
5.2 Theoretical modeling of spatial joints
In this chapter mathematical models of spatial joints will be derived. As al-

ready shown in Section 5.1 for planar joints, models of spatial joints will be
derived from geometry relations and from suitable representations and pro-
jections of vector loop equations and/or orientation loop equations that may
serve as constraint positionjorientation relations. Associated relative Coordi-
nates that may be used for measurement or control purposes are isolated
by suitable (local or global) representations and projections of these vector
or / and orientation loop equations.
As there exists a great variety of different spatial joints and massless links,
only a few joints of common use in industrial applications will be discussed
here. As a first step (Section 5.2.1} several building blocks (BBs} of joint mod-
els will be derived. Among those, constraint relations of a common-point con-
straint (BB1}, paraltel-axes constraint (BB2}, straight-line-point foliower con-
straint (BB3}, rotation-blocker constraint (BB4}, and constant-distance con-
straint (BB5) will be considered. From these building blocks, constraint rela-
tions and relative coordinates of the following standard spatial joints will be
derived in Beetion 5.2.2 (Table 5.1): spherical joint (BB1}, massless spherical-
spherical link (BB5}, translational joint (BB2, BB4), universal joint (BB1,
BB4}, revolute joint (BB1,BB2}, cylindrical joint (BB2, BB3}, and prismatic
joint (BB2, BB3, BB4). A more general universal joint with nonintersecting
and nonorthogonal rotation axes will be derived in Appendix A.4.
5.2.1 BuHding blocks of joint models
The discussion in Section 5.1 shows that combining different absolute and
relative constraints of two rigid bodies under planar motion already enables
the setting up of a large number of different planar mechanisms. Increasing
the number of bodies allows the construction of a large variety of different
planar kinematic and dynamic mechanisms and machines. It is easy to imag-
ine that a tremendous variety of different spatial mechanisms can be built
from only a small number of rigid bodies that move in space (ffi.3 ). In this
section certain geometrical situations associated with spatial joints will be
theoretically modeled by constraint position, velocity, and acceleration equa-
tions. In Section 5.2.2 they serve as building blocks for modeling different
types of spatial joints ([4], [40], [41]).
5.2 Theoretical modeling of spatial joints 201
5.2.1.1 Common-point constraint (BBl; three constrained transla-

tional DOFs). A common-point constraint forces two points Qi and Qj of
two rigid bodies i and j to a common position Q that may move in space (Fig-
ure 5.20). This constraint does not allow relative translational motions of the
bodies i and j in the points Qi and Qj , but only relative rotations. Let R be
an inertial frame and L" be a local frame with origin P" ("' = i , j) on a body
"'· Let ARLK be the orientation matrix of R with respect to L" ("' = i , j).
This geometrical situation is described by the vector loop relation (see Fig-
ure 5.20)
This provides the common-point constraint position equation
(5.19a)
the associated constraint velocity equation
or
or
incrtial
framc I ody i
Fig. 5.20: Vector diagram of a common-point constraint

RoR
Tpio
WLi
(I 3 ' - ARLi 0T-LiQPi '
-I
3 '
+ ARL; 0T-L;
QP;
) 0
LiR
RoR = 0,
Tp;o
L;
WL;R
..._____"_,_._
=: (v{, vi)T
(5ol9b)
and the associated constraint acceleration equation

RooR ARLi -Li -Li Li ARLi -Li Li RooR
TpiQ- 0WLiR 0Tqpi 0WLiR- 0Tqpi 0WLiR- 0
TP;O
+ ARL; 0wf;R 0rLQ~o 0wf;R

J J 3
+ ARL; 0rLQ~o 3
0wf;R
3
=o
or
R"R
Tpio
Lo
o
wL:R
R .. R
rP;O
=: Yp(P) 0
T(p) WL;R
o L;
- +ARL, 0WL, 0".L, 0WL, - ARL, -L, 0 -L; 0 L;

0WL;R (5ol9c)
- L,R QP, L,R TQP; WL;R
tagether with the constraint Jacobian matrix
[Jp(P) : = gp(P) T(p)

0
_ (I 3 , _ ARLi
-
0 -Li
Tqpi ,
_I 3 , ARL; 0 ".L; ) E oc3,12
QP; o
(5ol9d)
The constraint acceleration equation (5ol9c) will be used as a building block

(BB1)o It provides a theoretical model of a spherical joint connecting the
hoclies i and j (see Section 5020201)0
Comment 5.2.1 (BBl): In the case that relative rotation angles of the
joint are needed, either for sensing or for control purposes, this building
block must be reformulatedo
Comment 5.2.2 (Alternative representation of the constraint Ja-

cobian of BBl): The time derivative of the constraint position equation
(5ol9a)
Rd ßg
dtg(p) = 8p 0p = gp(P) 0T(p) 0V
0
(5ol9e)
may tagether with
p= R )T
(( Tpio T (Tp;o
,TJi' R )T
,TJjT)T R ( R R R
Tpio = Xpio' YPio' Zpio
)T
502 Theoretical modeling of spatial joints 203
and
be written in the form
R~tg(p) = ( a:I,o ' ::i ' a:~o ' ::i) 0
P (5ol9f)
oR
l
rpio
8 ( -ARL; (17j) o r~~;) r,i
a17j J oR
rP;O
r,j
Together with
8(ARLi( rL;) 8(ARLi( rL;)

17, 17, 0Ho( 0) 0 ARL; ( 0) 0 L;
0 0
0 0 0-
0 ) 0 )
QP; QP;
a11i 17,- ß
11i
• 17, 17, WL;R'
this implies
l
(5ol9g)
8 ( -ARL; ( 17 j) o r~~;)
' -13'
a17j
13 ' 03,3
03,3 ' Hi(11i) 0
ARL; (11i)
oR
rP;O
13 ' 03,3
L;
03,3 ' Hj(17j) 0 ARL; (17j) WL;R
Comparing the block matrices of (5ol9c) and (5ol9g) yields the relation
8 (ARL; ( 0) 0rL; )
- A RLi( 170 ) 0rQPO
-Li -_ 17, QPi 0H i (170 ) 0ARLi(170 ) (5ol9h)
" , a11i " "
or
8 (ARL; ( 0 ) o rL; )
-Li = -AL;R( o)O 17, QP; oHo( 0) 0 ARL;( 0) (5ol9i)
rQP; 17, ß " 17, 17,
11i
0
Multiplying both sides of (5ol9i) by wf:R yields

-Li
T QPi
0 Li
w LiR -
_ -ALiR( 0) 0
17,
° (ARLi( 17 i)
017i
0r~'p.) 0 Ho( 0) 0ARLi( 0) 0 Li
• 17, 17, w LiR 0
= TJi
(5o19j)
Together with
Ao RLi 0TQPO
Li _
-
_!!___ (ARLi (
17 ) Li ) 0 o
0TQPO 17 (5o19k)
, Ö17i ' , '
0 0
this yields
0 (ARLi ( TLi )
17,
0 ) 0
-Li Li - ALiR( ) QPi

TQPi 0 WLiR-- 11i 0 017i
- -ALiR( ) Ao RLio Li (50191)
- 11i TQPi
0
or
rLi 0WLi _ -ALiR 0 (ARLi ('l'lo) 0[;;Li ) 0TLi (5.19m)

QPi LiR - "H LiR QPi'
and finally the identity
(5.19n)
This proves the equivalence of the representations of the constraint Jacobian

matrix in (5o19b) and (5o19g)o
5.2.1.2 ParaUel-axes constraint (BB2; two constrained rotational

DOFs). This constraint allows two bodies i and j to perform rotations
around two parallel axes only. Assurne that these parallel axes are the x-axes
of the frame LQi fixed on the body i (exQJ, and of the frame LQ; fixed on
the body j ( exQ;). This geometrical situation will be compactly formulated
by the orientation loop equation (Figure 5021)
or (5o20a)
ALQ;LQi = ALQ;L; OAL;R ARLi ALiLQi = ALQ;L; OAL;Li ALiLQi
.........__.... 0 0
-----..,....... .........__.... -----..,.......

constant constant constant constant
with
AL; Li = AL;R 0 ARLi.
Its time derivative is

0 Ao L;R
Ao Lo;Loi -_ AL 0 30L 3. •
ARLi •
ALiLQo'!. + AL 0 0L 0 3 3.
AL·R
3 •
Ao RLi •
ALiL 0 '!..
Parallel rotation axes e o:Q;, e o:Q; wit h the relative rotation angle O!j i
inertial
~ a= CYQi Q; = CYji
frame
,,
/-A L;R
,,
,,
,,
,
Fig. 5.21 : Grientation diagram of a paraUel-axes constraint
or
ALQ L; A L;R r:;;R . ARL; . A L;LQ;
~. . RL; '-...--'
constant constant
+ ALQ ; L ; ·AL;R. ARL i . r:;;L i . AL i LQ ;

"-..-" L; R '-...--'
constant constant
This yields, tagether with
A. RL-
-
ARL -L
. W LR
and A. LR -
-
ALR -R -
. W RL -
-L
-W LR .
ALR
>
the relation
- A LQ;L; ·WL; . AL;R . ARL; . AL ; L Q;
"-..-" L; R '-...--'
constant constant
(5.20b)
constant constant
The second time derivative of ALQ; L Q; is

A"Lq.Lq.
J ,
(5o20c)
Due to the geometrical assumptions of Figure 5021, the orientation matrix
c
ALq;LQi has the form
0
, 0 )
ALq;LQi = 0 cosa , sina (5o21a)
a := 'PQ;Qi = 'Pji = Cl!jio
0 -sina , cosa
This implies
C'
0
ALq;LQi , 0 )
= 0 -sina cosa oa (5021b)
0 -cosa -sina
and
C' G'
0 0
, 0 ) , 0 )
ÄLq;LQi = ~ -cosa -sina 00:2 + -sina c~sa 0ö: 0
sina -cosa -cosa -sma
(5o21c)
As this building block only constrains two (relative rotational) DOFs of the
hoclies i and j: (1) only two independent equations of the acceleration equa-
tion (5021c) are needed as constraint acceleration equations, and (2) a single
equation is used for computing the relative coordinate ao
Due to common zeros in rows 2 and 3 and column 1 of the matrices
ALq;LQi, .ALq;LQi, and ÄLq;LQi, andin columns 2 and 3 and row 1 ofthese
matrices, the relative coordinate a can be eliminated from (5o20a), (5o20b),
and (5o20c) by projecting these equations from the left by means of
PrT (y, z) := ( eJ ) = ( 0 , 1 , 0 )
T
ez 0 ,0 , 1
and from the right by
Pr(x)=ex=(1,o,ofo
This implies the following constraint orientation, velocity, and acceleration

equations of this building block (BB2):
P? (y, z) ( ALo;L; . AL;R · ARLi · ALiLQi )Pr(x) = 0, (5.22a)
pT(y z)(-ALo;L; ·WL; ·AL;R,ARLi ·ALiLQi

r ' L;R
+ ALo;L; 0 AL;R ARLi wi:R ALiLQi )Pr(x)
0 0 0 = 0 (5.22b)
and
[02,3 ' P? (y' z){- ALo;L; . AL;R. ARLi . [ALiLQi Pr(x)] }, ooR
rpio
. L·
wL:R
02,3 ' P? (y' z){ ALo;L; . [AL;R. ARLi . ALiLQi Pr(x)]}] . ooR
rP;O
. L;
=: Yp(P) · T(p) WL;R
+ 2 .pT(y
r '
z)(ALo;L; ·WL; ·AL;R,ARLi .[;;Li ·ALiLQi)P (x)
L;R LiR r
_pT(y z)(ALo;L; ·AL;R,ARLi .[;;Li .[;;Li ·ALiLQi)P (x)
r ' LiR LiR r
(5.22c)
Yp. T(p) = [o2,3' - P?(y' z){ ALo;L;. AL;R. ARLi. [ALiLOiPr(x)] },

02,3 ' + P?(y' z){ ALo;L; [AL;R. ARLi . ALiLQi Pr(x)]}].
(5.22d)
The relative coordinate a := aQ;Qi = aji of this building block and its first
and second time derivatives are isolated from (5.20a), (5.20b), and (5.20c) by
the projections
sina=-P?(z)(.6.)Pr(Y) , cosa= P?(y)(.6.)Pr(Y) (5.23a)
as
with
.6. := ALQ;L; AL;R ARLi ALiLQi'
0 0 0
P?(z) := (0, 0, l)T P?(y) := (0, 1, O)T,

( 1
I - - - · {P?(y)(D)Pr(Y)}
I sina
ä = ~ or (5.23b)
I 1
- - - · {P?(z)(D)Pr(Y)}
l cosa
with
(5.23c)
and
( 1
I
I
---·
sina
{ [P?(y) ( 1>4) Pr(Y)] + cosa · ä 2 }
ä = ~ or (5.23d)
I 1
- - - · {[P?(z)(1>4)Pr(Y)] -sina·ä 2 }
l cosa
with
5.2.1.3 Straight-line-point-follower constraint(BB3; two constrain-

ed translational DOFs). A straight-line-point-follower constraint allows-
besides arbitrary rotations - translations of two bodies i and j only along
a straight line. Assurne that this straight line is oriented in the direction of
the x-axis exQi of the frarne LQi fixed on body i with origin Qi. Assurne
further that point Qj on the body j can only rnove on this straight line. This
geometrical situation is forrnulated by the vector loop equation (Figure 5.22)
or, written in the local frame LQ.,
(5.24a)
Fig. 5.22: Vector loop of a straight-line-point-follower constraint
with
(x~~Q; is the relative displacement).
Differentiation of (5.24a) with respect to the time yields
or
_ - A Lq ' L; . e;LL;R
g~ - ;
.
A LiR (r R
. PiO -
rR )
P;O
+A Lq;Li . A L;R. (r~io - r~;o) +A Lq;Li . wf : R. A LiR . ARL; . r~>j

- AL Q; L i . AL; R. ARLJ . WL J . rL ;
L1 R Q ;P;
+ rLq i
Q;Q;
= 0 (5.24b)
with
( X. L q.' is the relative velocity) (5.24c)

Q;Qi
or
L q.d
' g
-- (A)-
-- AL q.• L ; AL; .
·w AL.
iR [( r R R )
P;O - r P;O -
ARL · L;
J . r Q;P;
]
dt L;R
+ ALq;Li 0 ALiR 0 [(rRP;O _ rRPjO ) _ ARLj 0 wLj

LjR
0 rLj
QjPj
]
+ (x~~Q;, o, of = (o, o, o)To (5o24d)
Differentiation of (5o24d) with respect to the time yields
(5o24e)
The purpose of the straight-line-point-follower is to constrain two transla-

tional DOFs of two hoclies and to provide the relative coordinate x~~Q; of
the remainder unconstrained translational DOFo Due to (5o24b) and (5o24c),
the constraint position, velocity and acceleration equations are obtained as
the projections
g = P? (y, z) g (elimination of x~~QJ,

g= PT(
0
r y,z ) g~ (elimination of x~~QJ, (5o25a)
and
g.. =PrT( y, Z
) g:: (ehmmat10n of xQOQO), wlth Pr (y,
o o 0 .. Lq 0 0 T
z) :=
( 0,1,0) 0
' ' 0 ' 0 ' 1
The relative Coordinates x~~Q;' x~~Qi and x~~Q; are obtained as the projec-
tions
P rT( X ) g- A- Lqi
-xQJQi (isolation of x~~QJ,
P T( ) ~
r X g =
oLqo
-xQJQi (isolation of x~~QJ, (5o25b)
and
P r T( X ) g-
:: _ .. Lqi (isolation of x~~QJ, with P? (x) := (1, 0, O)o
-XQJQi
This provides the constraint position equation
g = P?(y, z) ALqJ; OAL;R

[ '-v--" 0 (rR
P;O
_ rR )
PjO
+ '-v--"
ALqiL; 0 rLi
Q;P;
constant constant c~t
- ALqJ; OAL;R
'-v--"
0 ARLj rLj
Q i Pj
J= (o o)T
' '
(5o25c)
constant ~t
cons ant
the constraint velocity equation
g. = pT(y
r '
z) { _ ALq;Li. c;;L; . AL;R. [ (rR
L;R P;O
_ rR ) _ ARL;. rL;
P;O Q;P;
J
+ ALqiL;. AL;R. [ (r~;o- r~;o) - ARL;. wf~R. r~>;] } = (o, o)T,
(5.25d)
and the constraint acceleration equation
-P?(y' z) (ALq,Li AL'R) ' P?(y' z) (ALq,Li AL,R ARL; r~~PJ]

..,
0 0 0 0
=: gp(P) · T(p) (constraint Jacobian matrix)
.
.. R )T ' ('Li
[(Tp,o )T ("R )T ('L; )T]T -_
WL,R ' TP;O ' WL;R
(5.25e)
The relative coordinate x~~Q; and its first and second time derivatives are
computed by the relations
xLqi
Q;Qi
= _pT(x)
r
[ ALqJ;. AL;R. ( R R )
Tpio-rP;O
+ ALq.L;
'
L;
·rQiPi
(5.26a)
and
r J
Lq; =-PT(x){ALq;Li. [AL;R, ( R
X.. Q;Qi r
R )
rP;O- rP;O -
ARL; L; J ·L;
'rQ;P; 'WL;R
+ ALq;L;. c;;Li
L;R
. c;;Li . AL;R. (rR
L;R P;O
_ rR
P;O
_ ARL;. rL;
Q;P;
)
(5o26c)
5.2.1.4 Rotation-blocker constraint (BB4; one constrained rota-

tional DOF). The constraint of this building block prevents two bodies
i and j from rotating around a selected axis (axis C in Figure 5023)0
Consider two orthogonal rotation axes (dashed lines in Figure 5023) that keep
their relative orientation while moving in the spaceo Consider the body i and
the body-fixed frames Li and LQi with origins Pi and Qi, respectivelyo Let
the x-axis (e"Qi) of frame LQi be oriented in the direction of the first rotation
axis, and assume that the body i may rotate around this axis by an angle
'Px =: r.p 1 o Consider the second body j and the body-fixed frames Lj and LQi
with origins P j and Qj, respectivelyo
Let the y-axis (eyQi) of frame LQi be oriented in the direction of the second
rotation axis, and assume that body j may rotate around this axis by an
angle 'Py =: r.p 2 0
Consider the straight line h in Figure 5023 that is parallel to rotation axis 2
and intersects rotation axis 1 at the point Ko This point is the origin of a
frame LK with the x-axis e"K = e"LK oriented in the direction of exQ;, and
with the y-axis eyK = eyLK oriented in the direction of eyQi (Figure 5023)0
Let
'Pl := 'Px := 'PQ;K (5o27a)
be the rotationangle from LK to LQ; (rotation angle of LQ; with respect to
LK) around the common x-axes exQ; = e"K of frames LQ; and LK, and
'/)2 := 'Py := 'PQjK (5o27b)

be the rotationangle from LK to LQi (rotation angle of LQi with respect to
LK) around the common y-axes eyQi = eyK of frames LQi and LK 0 Then
the straight line C (dotted line in Figure 5023) is oriented in the ezK-axiso
Introducing the abbreviations
AQ~>;K := ALQ~>;LK (~>; = i' j)
the geometrical situation of Figure 5023 provides the orientation matrices
0
1' - si~r.py)
AQ;K = ( 0 cosr.p" ,' sinr.p"
0 )
0
, - sinr.p" , COS 'Px cos 'Py
and
AQ;Qj = AQ;K 0 AKQj = AQ;K 0 (AQjK) T
first rotation axis parallel straight lines
~ ~ second rotation axis
<{Jx = <{JQ;I< = <{)1
body i : y <py = <{)Qjl< = <{)2
'' J
N xQ; I
h....}.
·'
A LKLQ; = AL.Q;~' CJl
Li "'
~
;:r
r · ('!)
l g
('!)
/(
(';
"""'
/ ' A LKLQ; 2:..
s
0
/ A LKLQ, """ 0...
~
L Q, LQ; s·
(l'q
orthogonal
0
....,
axes t A LQ, Li tA LQjLj
~
~
0 Li Li ~
o·
§_'
[/)
'<L;R~;R
R
Fig. 5.23: Geometrical situation of a rotation-blocker constraint I-'
"'
""'
or
1' 0 ,0 , sin<py)
COSi.py
( 0 , ' 0 ) (
COSi.px , sin 'Px 0 ' 1 ' 0
0 , - sin<px , COS i.px - sin<py ,0 ' cos i.py
(5o28a)
s <p := sin<p
with
c <p := cos <p,
together with the associated time derivatives
-s <py , 0 , c <py )
Q;Qj
A
0
= ( -S<px oCi.py , 0 , -S<px oSi.py o<{;y

-c i.px oc i.py , 0 , -s i.py oc i.px
0 ,0' 0)
+ ( -c i.px o S i.py , -s i.px , C i.px o C i.py (5o28b)
S 'Px o S i.py , -c 'Px , -s 'Px o C i.py
and
-S<py , 0 ,
Ci.py )
ÄQiQj ( -s i.px c i.py
o , 0 -s i.px s i.py o
-c 'Px c i.pyo , 0 , -s i.py c 'Px o
0 ' 0 ' 0 )
+ ( -Ci.px Si.py , -s 'Px ,
0
C 'Px o C i.py
S 'Px 0
S i.py , -c 'Px , -s i.px c i.py o
(
-C<py , 0, -e<p, )
+ S 'Px S i.py , 0 , -s i.px oc i.py 'Py
2
0
0
0
C 'Px 0
S i.py , 0 , -c i.py oc 'Px
+(
0 0
, 0 ) 2
, -c 'Px , -s 'Px oc i.py
0
S 'Px 0
S i.py 0 'Px
C 'Px 0
S i.py , S 'Px , -C<px ° C i.py
'0'
+20 '0' (5o28c)
'0'
Then the following orientation loop equation holds:

ARL; 0 AL;Q; 0 AQ,Q; 0 AQ;L; 0 AL;R = 13
or (5o29a)
This provides the orientation relation

AQiLi
---------------
...............
AL,R
..___,_,_..,
0 0 ARL; 0 AL;Q; - AQ,Q;
..___,_,_..,
=0 (5o30a)
constant global constant local
together with its time derivative

AQ;L; OÄL;R 0 ARL; 0 AL;Q; + AQ;L; OAL;R 0 ÄRL; 0 AL;Q; -ÄQ;Q; 0 =
"-v-" ............... "-v-" ............... -
constant constant constant constant
(5o30b)
and its second time derivative

A Q;L; 0
A" L,R
0
ARLO
' 0
ALOQO
' ' +2 0
AQ,L;
0
Ao L;R
0
Ao RL;
0
ALOQO
' '
+ A Q,L; 0
AL;R
0
A" RL;
0
ALOQO
' ' -
A" Q,Q;-
-
0 (5o30c)
Inserting
Ao L;R-
-
AL,R -R - -L;
OWRL,- -WL,R 0
AL,R
'
.ARL; = ARL, 0 wL'

L,R
and
ÄRL;- ARL; 0 -L; 0 -L; + ARL; 0 _:.L;
- WL;R WL;R WL;R'
A.. L,R - _:. L, AL,R - L, - L; AL;R

- -WL,R 0
+wL,R OWL,R 0
into (5o30b) and (5o30c) yields the relations
0 -- AQ,L, ( -L;
-W L,R 0
AL;R
0
ARLO
' + AL,R 0
ARLO
' 0
-L; )
W L; R 0
ALOQO
' ' -
Ao Q,Q;
(5o3la)
and
A Q;L; 0
(
-
-=.Li
WL,R 0
AL;R
0
ARLO
'+ WL,R
-L,
0
-L,
WL,R 0
AL,R
0
ARLO
'
In the next steps the constraint equations and relative coordinates of this
building block will be derived:
1. As a first step, the constraint equations will be obtained by suitable

projections of (5o30a), (5o31a), and (5o31b) that simultaneously eliminate
the relative Coordinates 'Px, tPx, 'Px, r.py, cpy, and rpy 0
2 As a second step, the relative Coordinates 'Px, r.py, tPx, cpy, 'Px, and rpy will
0
be isolated by alternative projections of (5o30a), (5o31a) and (5o31b)o
Due to (5o28a), (5o28b), and (5o28c), the matrices AQiQ;, AQiQ;, and ÄQiQ;
have a common zero element in the first row and second columno Both rela-
tive coordinates 'Px and 'Pv can be eliminated by projecting (5o30a), (5o31a),
and (5o31b) from the left by means of P? (x) := (1, 0, 0) and from the
right by means of Pr(Y) := (0, 1, O)To This provides the following constraint
orientation, velocity, and acceleration relations of this building block:
P?(x)( AQiLi ALiR ARL; AL;Q;)Pr(Y) = 0,

0 0 0
(5o32a)
P r T( X ) [ AQOLO ( -wL:Ro
-Lo
, , 0
ALOR ARLO3+ ALOR ARLO3oWL;R
-L; ) ' 0 ' 0 (5o32b)
0 AL;Q;]Pr(Y) = 0, (5o32c)
and
[o,,, , P." (x) { AQ;L; · rr AL;R · ARL; · AL;Q; )P,(y;]} , O, ,, , (5.32d)
-P?(x) { AQiLi ALiR ARL; 0 0 0 (AL;Q;pr(Y))}]
=: gp(P) 0
T(p) =constraint Jacobian matrix
[( .. R )T (oLo )T ("R )T (oL; )T]T

r~o ' WiR ' r~o ' w~R
= _pT(x) [AQiLi
r
0 (wLi
LiR
c;;Li
LiR
0 ALiR ARL; 0 0 + ALiR 0 ARL; wL;
0
L;R
In agreement with matrices AQiQ;, AQiQ;, and ÄQiQ; the following isola-
tion relations of the relative coordinates hold: suitable projections of (5o30a)
together with (5o28a) imply
(5o33a)
and
sin(r.py) = P?(x) (AQiLi 0 ALiR ARL; AL;Q;) Pr(z)
0 0
(5o33b)
as well as
cos(r.px) = P?(y) (AQiLi ALiR ARL; AL;Q;) Pr(Y)
0 0 0
(5o33c)
and
(5.33d)
This implies
and
P?(x) (AQ;L; · AL;R · ARL; · AL;Q;) Pr(z)}
r.p = arctan { (5.34b)
y P?(x) (AQ;Li ·ALiR.ARL; ·AL;Q;)Pr(x)
for -1r /2 < 'Px < 1r /2 and -1r /2 < 'Pv < 1r /2. By analogy with the above
expressions, we obtain from (5.28b) together with (5.31a) the relations
( +-1 -·P?(x)(~)Pr(z)
cos 'Py
l
0 )
'Pv = or (5.35a)
l
1 --.-1 -·P?(x)(~)Pr(x)
sm 'Pv
and
( 1 T
1 - - - · P r (z)(~)Pr(Y)
• ) COS 'Px
'Px = l or (5.35b)
ll --._1_. P?(y) ( ~) Pr(y),
Sllli.{Jx
with
uA ·.= AQiLi 0 ( -L;
-WL;R. ALiR . ARL·3 + AL;R •
ARL·3 • WL;R
-L; ) . AL·Q·
3 3,
(5.35c)
and finally from (5.28c) and (5.31b) the relations
( 1 T sinr.py 2
I - - · P r (x)(M)Pr(z)+--·<P
I cos 'Py cos 'Py y
<Pv = ~ or (5.36a)
1 --.-
1 - · P?(x) ( M) Pr(x)- _c~_s_r.p_y. <(; 2
l Sllli.{Jy Sllli.{Jy y
and
(1 - - -1· P rT ( Z ) (
M
)
Pr ( y ) +- 'Px · 'Px
sin - .2
.. ~ COS 'Px COS 'Px
'Px = or (5.36b)
1 PrT ( Y) ( M ) Pr ( Y ) - -
COS
.-'Px- . r.p. 2 '
l
I - -.-- •
Sllli.{Jx Sllli.{Jx x
with
~= AQiLi 0(- i;/LiR
i
0ALiR 0ARL; + c;;Li
LiR
0c;;Li 0ALiR 0ARL;
LiR
(5o36c)
5.2.1.5 Constant-distance constraint (BB5; one constrained trans-

lational DOF). This building block constrains the distance (C) between two
points Qi and Qj located on two bodies i and j 0 The associated constraint
relation will be derived from the vector loop equation (Figure 5024)
rR
PiO
- rR
P;O
+ ARLi 0 rLi
QiPi
- ARL; 0 rL;
Q;P;
+ rRQ30Q,. = 0
or (5o37a)
- R R ARLO L; ARLi Li
TQ;Qi- rP;O- Tpio + 3 0 rQ;P;- 0 rQiPi'
with
)T 0 (rQ;Qi )] 1/2
R
[( rQ;Qi R
= lrQ;QJ
R
= lrQ:QJ
Lo
= C = constant > 00
(5o37b)
This provides the constraint position equation of BB5:
0= -C +{ [rR
P;O
- rR
PiO
+ ARL; 0rL;
Q;P;
- ARLi 0rLi
QiPi
JT (5o38a)
0 [rR
P;O
- rR
PiO
+ ARL; 0 rL;
Q;P;
-ARLi 0 rLi
QiPi
J }1/20
............... '-v-'
constant constant
Differentiation of (5o38a) with respect to the time yields
0= ~ 02 (r~;Q.)
° T 0 (r~;Qi) j [(r~;Qi) T 0 (r~;Qi) J112
or approximately (5o38b)
O= (r~;Qi)T (r~;Qi)
0
je , C>Oo
This provides the following approximation of the constraint velocity equation:
(5o38c)
inert ial body i irQ1Q,i = l = constant

pherical- pherical link
Fig. 5.24: Vector loop associated with a massless spherical- spherical link
The time derivative of (5.38b) is
(5.38d)
with
+ A RL ; . r:;;LL;; R . r:;;L ; . rL ;
L; R Q ;P;
_ A RL ; . rL;
Q ;P;
. wL ; _
L; R -
0·
This provides the constraint acceleration equation
[- (r~;Q; f, (r~;Q; f ·ARL ; · r~: P; ' (r~;Q; f, - (r~;Q.)T · ARL; - r~>; ]

' ~ ~
=: gp(P) · T(p) = constraint Jacobian matrix
.
··R )T ' (WL
[( rP;O · L ;R
; )T
'
( ··R )T ( · L ; )T]T _
rP;O ' WL;R -
( T Q;Q
R
;
)T (5.38e)
or written explicitly:
=: Yp(P) · T(p)
[( .. R )T ('Li )T ("R )T ('L; )T]T (5.39)
. r~o ' w~R ' r~o ' w~R
5.2.2 Theoretical models of common joints

In this section the constraint relations of several spatial joints in common
use will be set up from the building blocks of Section 5.2.1. The spatial joints
considered are collected in Table 5.1.
5.2.2.1 Spherical joint (BBl; constrains three translational DOFs).
A spherical joint is completely modeled by the building block BB1 (cf. Bee-
tion 5.2.1.1). The constraint position, velocity, and acceleration equations of
a spherical joint (Figure 5.25) are as follows.
Constraint position equations of a spherical joint
rR
PiO
_ rR
P;O
+ A RLi . rLi
QPi
_ A RL; . rL; _
QP; -
0· (5.40a)
Constraint velocity equations of a spherical joint
Rr-R
PiO
- ARLi . ".Li
QPi
. wLi
LiR
- Rr-R
P;O
+ ARL; . rL;
QP;
. wLL;R
,
= o. (5.40b)
Constraint acceleration equations of a spherical joint

R .. R
rpio
. L·
[I3 ' - ARLi . r-LiQPi '

-I
3 '
+ ARL; . r-L;
QP;
J.
wL:R
R .. R
rP;O
=: Yp(P) · T(p) . L;
WL;R
- + ARL, . wLL,R. rLq'p . wLL,R- ARL, . wLL,,R. rqL,p·,.. wLL,;R· (5.40c)

- ' ' '
Spatial joints
massless
spherical translational universal revolute cylindrical prismatic
spher.-spher.
joint link joint joint joint joint joint
1jT
~ 0(s ~ ~ ~ J!b cAl \..._j I
BB 1
common point
X X X
_.-... Q1
U>
CO BB 2 "'>-,3
CO
-..._..-
pamllel O$eS :::-
('!)
U>
,..;.:: X X X X 0
...,
<..> ('!)
0 .,...
::0 (i '
b.O
BB 3 e:..
1:: str-aight line
:§ 8
0
point follower· X X 0..
·:; ~
.D
...., s·
()'q
1:: BB 4
-~ 0
.....,
,_.. r·otation blockeT'
..., X X X
rg {l
0 ;;.
0 ~
BB 5 c;·
constant distance X s·
.,...
"'
Table 5.1: Construction of spatial joint models from building blocks "'"'......
inertial
fram
body j
Fig. 5.25: Spherical joint
5.2.2.2 Massless spherical- sphericallink (BB5; constrains one

translational DOF). A massless spherical- spherical link is completely
modeled by the building block BB5 (cf. Section 5.2.1.5). Its constraint equa-
tions are as follows.
Constraint position equations of a massless spherical- spherical link
(5.41a)
R R
. [ r pjo - rpi o + ARLj r Li
Q iPi
]}1/2 -f=O.
'-v-"
constant
Constraint velocity equations of a massless spherical- spherical link
[TR
PjO
-TR
PiO
+ARLj ·TLj
Q jPj
-ARL i ·TLi
QiPi
]T (5.41b)
Constraint acceleration equations of a massless spherical-spherical link

=: gp(P) · T(p)
··R )T (·L )T (··R )T ("Lj )T]T

[( rpio (5.41c)
' wL ;R ' rpjo ' wLjR
body i
Fig. 5.26: Massless spherical- spherical link
5.2.2.3 Translational joint (BB2, BB4; constrains three rotational

DOFs). A joint that allows two rigid bodies i and j to perform relative
translational motions in three orthogonal directions and no relative rotation
to each other (Figure 5.27) is called spatial translational joint. It is modeled
by a combination of building block BB4 and of a modified version of building
block BB2, where the modified BB2 constrains relative rotations around the
common body-fixed x- and y-axes, and BB4 constrains rotations around the
remaining z-axis. As the modified BB2 allows relative rotations around the
parallel z-axes e zQ; and e zQj of bodies i and j, the matrix A Lq j Lq i of (5.21a)
is now modified to the form
c~sa , sina , 0)
( - sm a , cos a , 0 (5.42a)
0 ' 0 ' 1
The constraint position equation of the modified BB2 is obtained by the
following projections of the orientation loop equation (5.20a):
(5.42b)
with
(5.42c)
Fig. 5.27: Spatial translational joint

The constraint position, velocity, and acceleration equations of a spatial trans-

lational joint are as followso
Constraint position equations of a spatial translational joint
Constraint velocity equations of a spatial translational joint
P?(x' y) (- ALQ;L; 0wf~R 0AL;R 0 ARLi 0ALiLQi

+ALQ;L; 0AL;R 0ARLiwf~R 0ALiLQi )Pr(z)
[ 002]0 (5.43b)
P?(x) [AQiLi 0 (- wf~R 0 ALiR 0 ARL;
+ALiR 0ARL; 0wf~R) 0AL;Q;] Pr(Y)
Constraint acceleration equations of a spatial translational joint
[ 02,3 ' P?(x' y){ - ALQ;L; 0AL;R 0ARLi 0 [ALiLQi Pr(z)]} '
[o1,3 ' P?(x){AQiLi 0 [(ALiROARL; OAL;Q;)Pr(Y)J}
02,3 ' P?(x, y){ALQ;L; 0 [AL;ROARLi OALiLQiPr(z)J}]
01,3 ' -P?(x){ AQiLi 0ALiR 0ARL; 0 [AL;Q;pr(Y)J} l

=: gp(P) T(p) = constraint Jacobian matrix
0
.. R )T (oLo )T ("R )T (oL; )T]T

[( r~o (5.43c)
' WiR ' r~o ' w0R
5.2.2.4 Universal joint (BBl, BB4; constrains three translational

and one rotational DOF). Consider the mechanism of Figure 5028 com-
prising two rigid hoclies i and j connected hy a universal jointo Let Q he a
point that is common to the hoclies i and j Assurne that a frame LQi is fixed
0
on hody i with dorigin Q, and with the x-axis exQi placed in the direction of
the first rotation axis of the universal joint, and that a frame LQ; is fixed on
hody j with origin Q, and with the y-axis eyQ;, placed in the direction ofthe
second rotation axis of the universal joint, perpendicular to the first rotation
axiso Consider a third frame Lk with origin Q fixed to the two (massless)
rotation axes with unit vectors
exK = exQi and eyK = eyQ; 0

(5.44a)
Let
'Pl = 'Px := 'PQiK (5.44h)
he the angle of rotation of hody i around the x-axis (exK) of frameLk and
'P2 = I.{Jy := 'PQ;K (5.44c)

he the angle of rotation of hody j around the y-axis (eyK) offrame L k 0
This so-called universal joint hetween hoclies i and j constrains three trans-
lational DOFs and a single rotational DOF of the two hodieso The three
translational DOFs are eliminated hy a vector loop equation providing a
common-point constraint relation of BB1. The rotational DOF is eliminated
hy a suitahle projection of an orientation loop equation summarized in BB40
The relative angles r.p 1 and r.p 2 around the rotation axes of the universal joint
are isolated for measurement or control purposes hy suitahle representations
and projections of the orientation loop equationo
Collecting the relations included in the BB1 and BB4 provides the fol-
lowing constraint position, velocity, and acceleration relations and relative
rotation angles of the two hoclies i and j connected hy a universal joint (Fig-
ure 5028)0
Constraint position equations of a universal joint
(5.45a)
Constraint velocity equations of a universal joint
P?(x)
RoR
rpiQ-
ARLi 0 -Li 0 Li RoR
rQPi WLiR- rP;O
[AQiLi 0 ( -wf:R 0 ALiR 0 ARL;

+ARL; 0 -L; 0 L;
rQP; WL;R
+ ALiR 0ARL; 0wf~R) 0AL;Q;] Pr(Y)

l
(5.45h)
body i
LJ<
/ ' A LKLqi
T f10 / A LKLQ; ""'
LQ, LQj
t A Lq, L; t A Lqj Li
L; Li
0 ~L;R / { ri R
R
(a) Geometrical configuration of a universal joint
(b) Computer drawing of a universal (c) Technical realization of a univer-

joint sal joint (built at RTS, University
of Kassel)
Fig. 5.28: Drawings of a universal joint

Constraint acceleration equations of a universal joint
.....,..
=: gp(P) T(p) = constraint Jacobian matrix of the universal joint E JR4 •12
0
.. R )T (oLo )T ("R )T (oL; )T]

[( r~o (5o45c)
' WiR ' r~o ' w~R
A RL, 0w-L, r-L, wL, _ ARL, 0 -L, 0-L; 0 L;
L,R 0 QP, 0 L,R WL;R rQP; WL;R
_pT(x)
r
[ AQ,L, 0(wL' wL'
L,R 0 L,R 0
AL,R 0ARL; + AL,R 0ARL; 0
-L; -L;
WL;RoWL;R- -L,
2 oWL,Ro AL,R ARL·'oWL;R
-L; ) 0 0
AL Q]pr (Y)
' '
Relative coordinates 'Px and <py of a universal joint
(5.46a)
and
<py = arctan [P?(x) (f).)Pr(z)jP?(x) (f).)Pr(x)] (5.46b)
for -1f /2 < 'Px < 1f /2 and -1f /2 < <py < 1f /2 , with
/). := AQ,L, 0AL,R 0ARL; 0AL;Q; (5.46c)
and
Pr(x) := (1, 0, O)T , Pr(Y) := (0, 1, O)T , Pr(z) := (0, 0, 1)T o
(5.46d)
5.2.2.5 Revolute joint (BBl, BB2; constrains three translational
and two rotational DOFs). Arevolutejoint between the rigid bodies i and
j constrains three translational DOFs (BB1) and two rotational DOFs (BB2)
of the bodies (Figure 5029)0 The three translational DOFs are eliminated
by a vector loop equation defining the common-point constraint relations of
BB1. The two rotational DOFs are eliminated by a suitable projection of an
orientation loop equation of BB20 The relative angle around the rotation axis
of the revolute joint is isolated by suitable projections of the orientation loop
equationo
O = Oj;= Qp; P; 0
rota.tion ~nglc
'joint axis e x Lq = e x Lq
' j
(a) Vector diagram
(b) Revolute joint

Fig. 5.29: Vector loop and orientation loop of a revolute joint
Collecting the relations included in BB1 and BB2 provides the following
l
joint relationso
Constraint position equations of a revolute joint
rR
PiO
_ rR
PjO
+ A RLi 0 rLi _ A RLj
QPi
0 rLj
QPj
[ 0032]
[ (5.4 7a)
0
P? (y, z) ( ALQjLi 0 ALiR 0 ARLi 0 ALiLQi )Pr(x)
Constraint velocity equations of a revolute joint
P?(y, z)( -ALQjLj owf~R OALjROARLi OALiLQi
+ALQjLj 0 ALjR 0 ARLi 0 wf:R 0 ALiLQi )Pr(x)

(5.47b)
Constraint acceleration equations of a revolute joint

-ARLi of-Li
QPi
-13
o,'' '
' +ARLj
P,T (y, z) { A
0 rLi
QPO
~Q;L; , [ A L;R , ARL; , AL;Lq, P,(x) l}

l
~ ~
:= Yp(P) T(p) 0
(5.47c)
Relative coordinate of a revolute joint

with (5o48a)
~ := ALQ;L; 0AL;R 0ARL; 0AL;LQ;'
( 1
I - - - o [P?(y)(D)Pr(Y)]
I sina
ä = ~ or (5.48h)
I 1
- - - o [P?(z)(D)Pr(Y)]
l cosa
with
D := (-ALQ;L; 0WL; 0AL;R 0ARL; 0AL;LQi (5.48c)
L;R
+ALo;L; 0AL;R 0ARL; 0e;Li 0AL;Lo;)
L;R '
ancl
( 1
I
I
- - - o { [P?(y) ( ~) Pr(Y)]
sina
+ cosa 0ä 2 }
ä = ~ or (5.48cl)
1
I
- - - o { [P?(z) ( ~) Pr(Y)]- sina 0ä 2 }
l cosa
with
))Cl := ( + ALQ;L; 0AL;R 0ARL; 0~~:R 0AL;LQ;
- A LooL;
3 ow.:L;
L;R oALOR
3 oARL; 0AL;Loo,
+ ALQ;L3 0WL3
L 3R
0WL;
L;R
0AL;R 0ARLi 0ALiLQ;
-2 0ALo3L3 0wL3 0AL;R 0ARL; 0e;Li 0AL;Lo;
L 3R L;R
+A L 0 3. L;
0AL; R oARL; ow- L; w- Li AL;L0 •.) 0
L;R o L;R o
5.2.2.6 Cylindrical joint (BB2, BB3; constrains two translational

and two rotational DOFs). A cylinclrical joint hetween hoclies i ancl j
constrains two rotational DOFs ancl two translational DOFs, where the com-
mon rotation axes of the two hoclies point into the clirection of the common
translational motiono The common rotation axis of the two hoclies are chosen
in the exLo; -clirection ancl in the exLo; -clirection of frames LQi ancl LQ; fixecl
on the hoclies i ancl j, respectively (Figure 5030)0
Two translational DOFs are eliminated - ancl the relative coorclinate x~~Q;
is isolated - hy means of the moclel equations of BB3 (straight-line-point-
follower constraint in the x-clirection)o
Two rotational DOFs are eliminated - ancl the relative coorclinate aji is
isolated - hy means of the moclel equations of BB2 (parallel-axes constraint
in the x-clirection) 0
This provicles the following moclel equations of a cylindrical joint with a
relative motion in ancl arouncl a common x-axis:o
L, - ( L,
T Q;QI - - XQ,QI ,
0 , O)T
body j
Fig. 5.30: Vector diagram of a cylindrical joint
Constraint position equations of a cylindrical joint
P?(y' z) ( A LqiLi . A LiR . ARLi . A LiLqi )Pr(x)
pT(y z) [ALq i L 1 _AL 1R . (rR -rR) (5.49a)

r ' P 10 PjO
+A Lq 1L1 . rL i _ A Lq 1L1 . A L1R. ARLj. rLj J
~~ ~~
Constraint velocity equations of a cylindrical joint
p T(y z) (A LqiLi ·WLi ·ALiR .ARLi ·ALiLq 1

r ' LjR
+ALq j Li. ALjR. ARL1 . r:;;L1 . AL 1Lq i )P
L1R r
(x)
= (02T' 02T)T . (5.49b)

Constraint acceleration equations of a cylindrical joint
-P?(y, z)(ALQ;L; OAL;R)

' P? (y'
' r '
z){ ALQ;L; 0 [AL;R 0 ARL; 0 AL;LQ; Pr(x)
pT(y z)(ALQ;L; 0 AL;R 0 ARL; 0rL;
QjPj
)
l} l
"' ......... ~
=: Yp(P) T(p) 0
(5.49c)
-P?(y, z) ( ALQ;L; owf~R owf~R oAL;R oARL; oAL;LQ; )Pr(x)

+2 oP?(y, z) ( ALQ;L; owf~R oAL;R oARL; owf:R oAL;LQ, )Pr(x)
_pT(y
r '
z)(ALQ;L;OAL;ROARL;owL'
L;R owL'
L;R OAL;LQ,)P(x)
r
Relative coordinates of a cylindrical joint
a= Ctji = CtQJQ, = -arctan { [P? (z) ( .6.) Pr(Y)] j [P? (y) ( .6.) Pr(Y)J},
with (5o50a)
.6. := ALQjLj 0 ALjR 0 ARL; 0 AL;LQ,'
and
x~~Q, = -P?(x) ( ALQ,L, 0

AL,R 0
(r~,o- r~Jo)
+ ALQoL,
, orL'
Q;P; -
ALQoL,
, oAL,R oARLj orLj
QjPj
) 0
Relative velocities of a cylindrical joint
( 1
I - - - o [P?(y)(D)Pr(Y)]
I sina
ä = ~ or (5o50b)
I 1
- - - o [P?(z)(D)Pr(Y)],
l cosa
with
and
oLQo
XQ):J; =
PT(
r X
) [ALQOL -Lo
' ' 0W L:R 0
0
ALOR ( R
' 0 rP;O- R
Tpjo-
ARLO
3 0
Lj
rQjPj
)
Relative accelerations of a cylindrical joint
( 1
I
I
- - - o { [P?(y) ( M) Pr(Y)]
sina
+ cosa 00: 2 }
ä = ~ or (5o50c)
I 1
- - - o {[P?(z)(M)Pr(Y)] -sinaoä 2 } ,
l cosa
with
.... o-+ALQoLj
""o- 3 0ALJR 0ARL; 0w:.Li
LiR 0AL;LQo,
- A LQoLj
3 :.,Lj oALjR oARLi oAL;LQo,
oWLjR
+ ALQ3L3 0WL3
L R
0WLj 0ALjR 0ARL; 0AL;LQ;
LjR 3
-2 0ALQ3L3 0WL3 0ALjR 0ARL; 0c;;L; 0AL;LQi
LjR L;R
+ ALQjLj 0ALjR 0ARL; 0c;;L;
L;R
0c;;L;
L;R
0AL;LQ; 0
5.2.2.7 Prismatic joint (BB2, BB3, BB4; constrains three rota-

tional and two translational DOFs). A prismatic joint between bodies
i and j constrains three rotational DOFs and two translational DOFso Two
rotational DOFs around the x- and y-axes are constrained by the modified
BB2 (cfo Equation 5.42b)o The remaining rotational DOF around the z-axis
is constrained by BB40 The two translational DOFs in the y- and z-directions
are constrained by BB3 (Figure 5031)0
L,
T Q;Q; = ( -XQJQI, 0 ' O)T
bocly j
Fig. 5.31 : Vector diagram of a translational (prismatic) joint
This provieles the following moclel equations of a prismatic joint that may
also be consiclered as a combination of the model equations of a translational
joint (modified BB2, BB4) with the model equations of BB3.
Constraint position equations of a prismatic Joint
P? (x, y) ( ALQ; L; . A L ; R . ARLi . ALi LQ 1 )Pr(z)
P?(x) (AQiLi. AL iR . ARL; . AL;Q;) Pr(Y)
(5.51a)
Constraint velocity equations of a prismatic Joint

pT(x
r '
y)( -ALQ;L; owL;L;R OAL;ROARL; OAL;Lq,
+ALQ;L; 0 AL;R 0 ARL; 0 wf~R 0 AL;Lq, )Pr(z)
P?(x) [AQ;L; 0 ( - wf:R 0 AL;R 0 ARL; + AL;R 0 ARL; 0 wf~R) 0 AL;Q; ]Pr(Y)
= (02T'02T'0 )T 0
(5o5lb)
Constraint acceleration equations of a prismatic joint
' P?(x' y){- ALQ;L; AL;R ARL, [AL,Lq, Pr(z)]} 0 0 0
'P?(x){AQ,L, 0 r(AL,ROARL; OAL;Q;)Pr(Y;]}
'P?(x, y){ALq;L; [AL;ROARL, oAL,Lq,Pr(z)]} 0
01,3 '-P?(x){ AQ;L; AL;R ARL; [AL;Q;pr(Y)]} 0 0 0
-P?(y, z)(ALq,L; OAL'R) pT(y z)(ALq,L'OAL,ROARL;ofL; )

' r ' QjPj
" =: Yp(P) T(p) = constraint Jacobian matrix
0
(5o5lc)
- P r T( X ) [AQ;L; 0
(-L; -L;
WL;RoWL;Ro
AL;R 0
ARLO
'+ AL;R 0
ARLO -L; -L;
'oWL;RoWL;R
-
2 0
-L;
WL;R 0
AL;R
0
ALOR
' 0
-L; )
WL;R 0
ALOQO]
' '
p r (y )
Relative coordinate of a prismatic joint
Lq; - pT( X ) [
XQ;Q;-- r ALq;Li 0 ALiR 0 (r~;O- r~;O) (5o52a)
+ ALqoL;
, orL;
Q;P; -
ALqoL;
, oAL;R oARL; orL;
Q;P;
J 0
Relative velocity of a prismatic joint
R
( rp;o- R ARL; L; )
rp;o- 0
rQ;P;
Relative acceleration of a prismatic joint

r J
xLQ;
Q;Q;
=- pT(x){ALq;Li 0 [AL;R 0 (rR
r P;O
- rR ) - ARL; 0 rL;
P;O Q;P;
J 0 wLi
L;R
+ ALqJ; 0 wLi 0 wLi 0 AL;R 0 (rR

L;R L;R P;O
- rR
P;O
- ARL; 0 rL;
Q;P;
)
- 2 0A
LqoL; -L;
' 0 WL;R 0
AL;R oR
0 TpiQ- rP;O-
(oR
ARL· -L; L;
J 0 WL;R 0 rQ;P;
)
+ ALq.L;'
0
AL;R 0
("R ooR
Tpio-rP;O-
ARL; - L; - L; L;
oWL;RoWL;RorQ;P;
)
+ ALqJ; 0 AL;R 0 ARL; 0 rL;

Q;P;
0 WL; }0
L;R
(5o52c)
6. Constitutive relations of planar and spatial
external forces and torques
In this chapter constitutive relations of some commonly used (passive and

active) force and torque elements will be discussed. They usually act due
to relative motion of system components. An important class of forces is
associated with compliant elements such as coil springs, leaf springs, dampers,
tyres, shock absorbers. and other deformable media that generate reaction
forces and torques between - most times two - bodies. They may be a function
of the relative position and velocity of the components. Sometimes they may
also depend on their relative acceleration. Another class of forces and torques
are those that act on a single body like, for example, the gravitational force.
In this chapter constitutive relations of the gravitational force, translational
springs and dampers, torsional springs and dampers, and actuators and mo-
tors will be briefly discussed. In Beetion 6.1 the model equations offorces and
torques acting in a plane or in parallel planes will be considered. In Beetion 6.2
the forces and torques acting in a space will be discussed. These constitutive
relations will be extensively used in the applications of Volume II.
6.1 Constitutive relations of planar external forces and

torques
An external force or torque applied to a rigid body is commonly represented

(Figure 6.1 and Section 2.1.1.1) geometrically by an arrow (F, M) and for-
mally by the relations
F = F · eF and M = M · eM (6.1)
with the symbols of the amplitudes F of F and M of M possibly changing

their signs, and with the unit vectors eF (in the direction of F) and eM (in
the direction of M) .
A sign convention states that F (or M) is counted positive, F > 0 (or M > 0),
if it acts in the direction of the arrow, and negative otherwise. This implies
that a force or torque is geometrically represented as an element of a set of
geometrical vectors: (1) that are placed on a common line of action (in the
240 6. Constitutive relations of planar and spatial external forces and torques
ep, unit ve tor e M, unit vector in frame L M

\ in frame L p
/
~-- Grientation
of M
M =M ·eM
rigid body
Fig. 6.1: Geometrical representation of a force F and a torque M
case of F) and may have equal or opposite directions, or (2) that have common
orientation (in the case of M) and may have equal or opposite directions.
The associated algebraic vectors, represented in frame L, are
F L = ALL F . FLF ALL F = C'~F

sin '1/JF
0
, - sin 1/JF
cos'ljip ,0
,0)
0 ' 1
FL F = (F , 0, O)T 1/JF := 1/JLFL, (6.2a)
and
ML = ALL M . ML M ML M = (0 0 M)T (6.2b)
' ' '
with Fand M as the amplitudes of the force and torque, eF and eM as unit
vectors in the direction of the force and torque arrows, IFI and IMI as the
absolute values of Fand M or as the length of F and M, and ALLF or A LLM
as mappings from frame L F with the basis vectors {exF , eyF, e zF } with
exF := eF, or from frame L M with the basis vectors {exM , ey M, e zM } with
e zL := eM , respectively, into frame L with the basis vectors (e xL , eyL , e zL)-
Then the force (or torque) vector points into the direction of the arrow F
(or M) for F > 0 (or M > 0), andin the opposite direction of the arrow for
F < 0 (or M < 0).
6.1 Constitutive relations of planar external forces and torques 241
Comment 6.1.1 (Characterization offorces and torques): The action

of a force on a body is uniquely characterized by the amplitude, direction, and
line of action of the force (force arrow). Attachment points of forces arenot
needed. The action of a torque on a body is uniquely characterized by the
amplitude and direction of the torque (torque arrow). Lines of action and
attachment points of torques are not needed.
6.1.1 Gravitational force (weight)
In this context the gravitational field will be assumed to act in the negative
eyR direction (Figure 6.2). Then the weight Fwi of a body i is
Fwi = -mi · g · eyR or Fw

R = ( 0 ) = constant, (6.3a)
' -mi ·g
with g as the gravitational constant. Since the line of action of Fwi meets
the center of mass Ci of body i , the force Fwi does not generate a moment
with respect to C;; i.e. ,
-o ·
M cL iiwi = (6.3b)
1 ody-fix d
frame L;
iner ial
frame R
rigid body i
(-g. € yR
0
€ xR
Fig. 6.2: Gravitational field acting on a body in the ( -e yR) direction
6.1.2 Applied force and moment
Consider a force represented by an arrow F i of length Fi through the point

P; on body i (Figure 6.3). Then
(6.4a)
with the unit vector eF, = exF of frame LF and with
as the associated algebraic force vector, or

FR = A RL F . FL F = ( c~s 'l/J F , - sin 'ljJ F ) . ( Fi ) , (6.4b)
" 2 Sill 'ljJ F , COS 'l/J F Ü
and finally
and
with
Fi~ := Fi · cos'l/JF
The torque Mi of Fi with respect to the point OLi , and represented in frame
Li is
iner ial
frarn R
J.,i ~
XL'
,PL;R
1
bocly i
e zR 1
Fig. 6.3: A body acted upon by a constant force F ; and moment M ;

6ol Constitutive relations of planar external forces and torques 243
, Y~;oL,) (
L,
' -XP;OL; 0 -
cosV;;
o .J,
Slll 'Pi
, sin V;;
' cos V;;
,0)~ (F//;)]
~~
0
: 0 '
' 0 0
or
r( L
-xp'o Lt sm o.p·-
0
o .!,
~
Yp'o Lt cos o.p·t. L 0
.!,
'
,
L;
°
L;
Xp. 0 Li cos 'Pi- Yp..!,
0
0
0 Li sm .!,
'Pi °
o
t t. t t t
or
M; = [ ( - x~:oL; sin V;; - 0

y~:oL," cos V;;) F;~ ° (6.4c)
+ (x~:oL; 0 cos V;; - y~:oL; 0sin V;;) F;~ J 0 ezL; °
= F; 0 [ (- x~:oL," sin V;; 0 cos V;F + x~:oL," cos V;; 0 sin V;F)
+ (- y~:oL," cos V;; 0 cos V;p - y~:oL," sin V;; 0 sin V;F) J 0 ezLn
and finally
M; = F; 0
[ - x~:oL, sin (V;; - V;F) - y~:oL," cos ( V;F -V;;) J ezk
0
(6.4d)
6.1.3 Translational force elements between two borlies
Consider a massless translational force element that exerts a force with an

amplitude F along a line of action through the point Pi on a body i and
point Pj on a body j (Figures 6.4a and 6.4b) without imposing any kinematic
constraint on these bodies (cfo Section 501.3)0 The kinematics of this force
element has been (symbolically) modeled by a massless revolute-revolute-
translational link ( ''pseudo-joint") that does not constrain any DOF of the
bodies i and j (Figure 6.4b) 0
body i
tran lational
force element
with a force F
revol ute joint
(a) Translational force element between two bodies i and j
translationallink
revolute
joint
"pseudo-joint"
(b) "Kinematics" of the massless revolute-revolute-translational link ( "pseudo-
joint") as part of a model of a translational force element between two bodies
i and j
?-\F,
\_7
a fo rce element pulls two bodies
Fi ~
~
a for ce element pushes two bodies
(c) Sign convention of translational forces between two bodies i and j
Fig. 6.4: Forces and vector diagram of a translational force element

6ol Constitutive relations of planar external forces and torques 245
=
Due to Newton's third axiom ( action reaction) the forces Fi and Fj exerted
from this element on the bodies i and j, respectively, have equal absolute
values and opposite signs; ioeo,
The vector dji from point Pi to point P j is computed from the vector loop
equation (Figure 605)
or represented in frame R as
dR.
1•
0= rR
ojo
_ rR
oio
+ ARLj 08 Lj _ ARLi 08 Li
, (606)
0
1 0
The distance fij = Rji between points Pi and Pj (length ofthe force element)
is
(607)
Then a unit vector

d··
e J" (608)
ldjil
o-
ji o-
will be introduced (Figure 605)0 By a common sign convention, the pair of

forces (F i , F j) in Figure 6 .4c is defined as:
1. Positive if the force element pulls the bodies (tends to draw the bodies
together and to tension the force element) 0
20 Negative if the force element pushes the bodies (tends to increase the
distance between the bodies and to compress the force element) (Figure
6o4c)o
Then, in the case of "pull", Fi (acting on the body i) has the same
direction as eji, and Fj (acting on the body j) has the opposite direction of
eji; ioeo,
Fi= Foeji,
and (pull situation) (6o9a)
In the case of ''push", Fi has the opposite direction as eji; ioeo,
Fi = -F eji,0
and (push situation) (6o9b)

frame R
yR
revolute
join
0 =0
Fig. 6.5: Vector diagram of a translational force element
The above translational force element can be technically built by units such
as hydraulic or pneumatic actuators or bearings, electromagnetic actuators
or bearings, spring elements, and damper elements, that are connected to the
bodies by revolute joints attached to each end of the force element, where
a single or several of the above components may be included in a single
translational force element.
6.1.3.1 Translational spring. Translational (point-to-point) springs are
often used as (passive) force elements in rigid-body systems. The force of the
spring of Figure 6.6 is defined as
and may be written as

Fi =Ci· [1 - /!jiO / (d; · dji) 112 ] · dji {linear spring)
or as
Fi :=Ci (l!ji - l!jiO) · eji , Ci: lR1 ---+ lR1 {nonlinear spring), (6.10b)
wi th Ci > 0 as the stiffness coefficient, Ci ( ·) as the stiffness characteristic,

l!ji as the deformed length and l!jiO as the undeformed length of the spring,
and dji = r P; p i as the vector from the attachment point Pi of the spring on
body i to the attachment point P j on body j. These force relations are in
agreement with the above sign convention:
1. For (l!ji - l!jiO) > 0 the two bodies are pulled and the spring is under
tension (Figure 6.6). Then F i acts in the direction of eji and Fj acts in
the direction of ( - eji)-
frame R
yR
(l;; > l;iO, pu l! situation)
Fig. 6.6: Translational spring between points P i and P j of bodies i and j
2. For (R.ji - R.jiO) < 0 the two bodies are pushed and the spring is under
compression. Then Fi acts in the ( -eji) direction, and Fj acts in the
eji direction.
The deformed length .eji of the spring and the associated unit vector eji are
computed by (6.6) , (6.7) , and (6.8).
6.1.3.2 Translational damper. Given a translational (point-to-point)
damper between the bodies i and j , as shown in Figure 6.7. The damper force
on body i is defined as
Fi = di · fji · eji , di >0 (linear damper) (6.11a)
or as
Fi = di (iji) · eji, di: IE.1 ---+ IE.1 (nonlinear damper), (6.11b)
with
with di > 0 as damping coefficient, di(-) as damper characteristic, and with
(6.12a)
and dn as defined in (6.6). Computation of the time derivative

revolute joints
frame R
yR fram L1
frame L;
(Lji > 0, pull Situation)

'!f;; := 'if;L;Fl , '!f;; := 'lj;LjR
Fi = - F;
€ zR
Fig. 6.7: Translational damper between points P ; and P j of bodies i and j
~
dt
(tJ•
2 ) = 2 · fji · fji = ~ [(dL; )T · (dj i)Li]
(6.12a) dt J'
(6.12b)
= 2. (d L! )T . (Li(l') = 2. (ALiR. dR_) T .

JZ JZ JZ
~
dt
(ALiR. dR_)
)t
= (2. A LiR . dR.)T.

JZ
(A.L iR. dR_
JZ
+ A LiR .R(zR_)
JZ
__ 2. (dR)T
ji . ARLi . [A LiR . (Rd"ji
R+W~ RLi ji T
R . dR)]
=2· (df;) T· (R(z~) -2· (df;) T· (ii~) ·W~Li

provides, together with
R T ~R
(dji) . dji =0 (6.12c)
the relation
~t (fJi) = 2 · fji · fji =2· (dJ/ )T· (Li {z~n = 2 (df;) T. (Rd~) , (6.12d)
and finally the relation
(6.13)
or
(dfii f . (Li (z~n
( (dJ/ )T. (dJii ) f/2
Equations (6.11) and (6.12d) yield the following expressions for the force of
a linear translational damper:
(6.14a)
or
(6.14b)
These force expressions are usually much briefer for model equations written
in generalized coordinates that include the relative coordinate dji· For model
equations written in absolute generalized coordinates, the following relations
hold. Consider the vector
(6.15a)
or represented in frame R,
dR.
1•
= rR
o;o
- rR
oio
+ ARL; . sL;- ARLi. sLi
1 " '
(6.15b)
with
A RL; = ( cos '1/Jj , - sin '1/Jj ) ' A RLi = ( cos '1/Ji , - sin '1/Ji )
sin '1/Jj cos '1/Jj sin '1/Ji cos '1/Ji
and with
'1/Jj := '1/JL;R,
The time derivative of d~ with respect to the frame R is

.R R .R R .R . RL; L;
dji := ro;o- roio +A . sj
+ A RL·' . L·'sj-L; -
A. RLi Li
. sj -
ARLi Li ·Li
. si .
Using
A. RLi == ARL·
~
-L·
. wL:R ARL;- ARL; 0 -L; (6.16a)
- WL;R'
and the rigid-body property
(6.16b)
yields, together with

·R R·R
roio := roio and (6.16c)
the time derivative of dn:

0
R - d. R - R R
0
ARL L;
0
L; ARLi -Li Li
0
Tp;pi- ji-rO;o-roio+ '·WL;R'Sj- ·WLiR.Si.

-
(6.17)
Then the above damper force relations (6.11a) and (6.11b) are in agreement
with the previous sign definition. Since the damper opposes the relative mo-
tion of the two bodies when they move away from each other (Rji > 0), the
damper forces exhibit a pull on the bodies. Then Fi acts in the direction of
eji and Fj acts in the direction ( -eji)· For (Rji < 0) the two bodies move
towards each other and the damper forces exhibit a push on the bodies. Then
Fi acts in ( -eji) direction and Fj acts in the direction of eji·
6.1.3.3 Actuator. Forces between two bodies generated by hydraulic, pneu-

matic, or magnetic actuators are defined by complete analogy to the above
sign conventions. They may be written in the form
(6.18)
6.1.3.4 Torsional spring and damper. Torsional (rotational) springs and

dampers between two bodies are always assumed to act around the axis of
a revolute joint that connects these bodies. Consider two rigid bodies i and
j connected at a point P by a revolute joint (Figure 6.8). Assurne that a
torsional spring and a torsional damper act around the rotation axis of the
joint, and that they are attached to the arrow Tj fixed on body j and arrow ri
fixed on body i. This spring damper element exerts torques of equal magnitude
=
but opposite orientation on the bodies i and j (action reaction}. Let '1/Jji be
the difference between the rotation angles of the bodies j and i (measured as
the angle from r i to r j), and let '1/JjiO be the angle of the undeformed spring.
As a torque Mi is positive if it acts (for a compressed spring ('1/Jji > '1/JjiO))
counter-clockwise on the body i, and clockwise on the body j, it is formally
described by the relation
Mi:= [er· ('1/Jji- '1/JjiO) + dr · ~ji] · ezR, (6.19)
with Cr as the stiffness of the torsional spring and dr as the damping coeffi-
cient of the torsional damper.
6.1.3.5 Torque generated by a motor. Torques between two bodies that
are generated by electric, hydraulic, or pneumatic motors are defined by anal-
ogy to the above sign convention of torsional spring-dampers. Those torques
are usually written as
(6.20)
6.2 Constitutive relations of spatial external forces and torques 251
straight line r; fixed on body i

'1/Jii
·.----....... r;
inertial
frame R
yR
Fig. 6.8: Torsional spring and damper
6.2 Constitutive relations of spatial external forces and

torques
The planar forcejtorque relations of Section 6.1 can be easily extended to

spatial forcejtorque relations. The general forcejtorque relations (6.2a) and
(6.2b) hold for planar and spatial vectors.
Gravitational force ( weight)
Assuming that the gravitational field acts in the negative eyR-direction. Then
the gravitational force on a body i of mass mi is
constant. (6.21)
Applied force and torque
Consider an applied force represented by an arrow F i of the amplitude Fi

and acting on the body i along a line of action and through a point P Fi
(Figure 6.9). Let epi be the unit vector in the direction of F i on the line of
action of F i. Then
•
I
body i
~
~ Vf j \,
I '
·- ·- ·- ·- · ·-·--
,
/
Fig. 6.9: Constant force F i acting on a body i
or (6.22a)
Fi E IE.1 .
The torque of F i with respect to the point Pi represented in frame Li is
· (
c i2 c i3
- c i2 s i3
'
,
c il s i3
c il c i3
+ s i l s i2 c i3
-
' s i l s i3
s i l s i2 s i3 ' s i l c i3
- c i l s i2 c i3
+ c il s i2 s i3
) l
s i2 - s il c i2 ' c il c i2
with the Bryant angles 1>L i R, eL i R, and '1/JL i R and t he abbreviations

Sil := sincf>L iR S i2 := Sill BL i R s i3 := sin '1/JL i R,
eil:= coscf>L iR Ci2 := COS0LiR
Translational springs, dampers , and actuators between two hoclies

The relations (6.5) to (6.9b) of planar forces also hold for spatial forces,
taking into account spatial vectors d ji, rpj p i, Sj , Si , etc., and the spatial
6.2 Constitutive relations of spatial external forces and torques 253
transformation matrices:
(6.23a)
The same holds for the relations (6.10a) to (6.18) of translational springs,
dampers, and actuators.
Torsional springs, dampers, and motors
Torsional springs, dampers, and motors are in the planar case, as well as in
the spatial case, assumed to act araund the axis of a revolute joint. Then the
torque relations (6.19) and (6.20) of planar torsional elements also hold for
the spatial case.
A. Appendix
This appendix includes four sections. In Beetion A.1 special veetor and ma-
trix operations used in rigid-body dynamics will be discussed. The Lagrange
equations of a rigid body under spatial motion are briefly discussed in Bee-
tion A.2 as an alternative to the Newton-EuZer equations. In Beetion A.3
vector and matrix notations together with the model equations of planar and
spatial meehanisms are compared with each other, followed by the eonstraint
equations of a generat universal joint, derived in Beetion A.4.
A.l Special vector and matrix Operations used m

mechanics
In this section some special vector and matrix Operations that are often used
in rigid-body dynamics will be briefly discussed. Starting with some basic
definitions and properties of Euclidean veetor spaees in Section A.1.1, alge-
braie and geometrie properlies of the sealar produet and of the veetor produet
(or eross produet) of planar vectors will be discussed in Section A.1.2. In Bee-
tion A.1.3, eross produet operations of spatial veetors will be proved. Time
derivatives of matriees and veetors represented in different frames will be
discussed in Beetions A.1.4 and A.1.5 for planar and spatial situations, re-
spectively, followed by a brief review of the derivatives of the vector functions
(gradient, Jaeobian matrix) in Beetion A.1.6.
A.l.l Euclidean vector space

In the sense of linear algebra, a real Euclidean veetor spaee is defined as a
quintuple
(V, K, +, ·, •), (A.l.l)
which includes the following entities: a field of real numbers (scalars) K, a
eommutative group (V , +) of vectors V, and the mappings
VxV ~V ( addition of veetors)

\.1) \.1) \.1)
(x, y) f-----tZ :=X+ y y +x ( eommutative law)

256 Ao Appendix
VxK~V (multiplication of a vector by a scalar)

\.1) \.1) \.1)
X, A.) f---tZ := A o X,
that satisfy the properties
(A. M) x = A. (M x)
0 0 0 0 A.,J-tEK; xEV ( associative law)
A.o (x+y) =A.ox+A.oy ; x, y E V

(distributive laws)
(.X+ M) 0 X = A. 0 X+ f.j 0 X A.,J-tEK ;xEV
and
lox = x 1EK x EV (existence of a unity)o
Then
(V, K, +, o)
is called linear vector spaceo A mapping
(Aol.2)
between two linear vector spaces vl and v2 is called linear iff
cp(x1 + x2) = cp(xl) + cp(x2) X1, X2 E V1

cp(A. xl) = A. cp(xl)
0
0
x E v1 , .x E Ko
A linear mapping cp (Aol.2) can be represented with respect to two bases L
of vl and R of v2 by means of a matrix A RL:
A RL : v1 -----+ v2
\.1) \.1)
rL f---trR := A RL 0 rL
with rL and rR as algebraic vectors, represented in L and R, and with A RL 0
rL as the product of the matrix A RL with the algebraic vector rL, where
the dot in the product A RL rL is sometimes omittedo Introducing a bilinear 0
mapping
ß: V xV ~K
\.1) \.1) \.1)
(x, y) f---tA. = ß(x, y) =: x • y
which is assumed tobe linear in both arguments (x and y); ioeo,

Aol Special vector and matrix Operations used in mechanics 257
ß(xl + x2, y) = ß(xl + y) + ß(x2 + y),

ß(x, Y1 + Y2) = ß(x + Y1) + ß(x + Y2)
and
ß(>. x, y) =).. ß(x, y) = ß(x, ).. y),
0 0 0
and postulating the additional properties
ß(x, y) =ß(y, x) xEV yEV (symmetry)

and
ß(X , X) > Ü for X -:j:. 0 ( definiteness)
provides a scalar product "•" on V and associates with the linear vector space
(V , K , + , a Euclidean space (V , K , + ,
0
)
0
, •) 0
The scalar product enables to introduce a norm into a vector spaceo This
allows defininition of the length of a vector and relative angles between two
vectors as well as the concept of orthogonalityo This introduces topological
properties into a vector space V and enables an analysis in V 0
Comment A.l.l (Vector product): As is easily seen from the above def-
initions (Aol.l), the vector-space concept of linear algebra does not explicitly
include the notion of a vector product or cross product; ioeo, no mapping
x: VxV ~V
(y, w) f---tz =: (y x w)o
On the other hand, it is common in rigid-body dynamics, to define a

vector product yR x wR of algebraic vectors yR and wR of V with respect to
a common basis R as the result of a matrix multiplication (linear mapping)
fJR :V -----+ V (Aol.3)

U! U!
with the skew-symmetric matrix fJR constructed from a given algebraic vector
yR E V that provides for each wR E V the same result as the standard
vector product (compare Equations Aol.5e and Aol. 7b of Sections Aol.2 and
Aol.3)o This notation provides both conceptual clearness ofthe vector product
defined as a linear mapping between vector spaces, and the possibility to
compute a vector product in the framework of standard matrix multiplicationo
258 A. Appendix
A.1.2 Scalar product and cross product of planar vectors
Scalar product
Given two planar geometrical vectors TQo and r PO· The scalar product of
these geometrical vectors is defined as the scalar (Figure A.l.l)
(A.l.4a)
Algebraic computation of the scalar product
The scalar product (A.l.4a) of two algebraic vectors is computed as
(A.l.4b)
R R R R L L L L
= xQo · Xpo + YQo · YPo = xQo · Xpo + YQo · YPo
from algebraic vectors represented in frames R or L.
Proof:
Representing TQO and r PO in frame R yields
lrQol·lrpol·
R R COS!.{JQP = Ix~o
R I · Ix~o
R I · COS!.{JQP (A.l.4c)
YQo YPo
= lx~o ·exR+Y~o ·eyRI·Ix~o ·exR+Y~o ·eyRI· cosr.pQP·
Then (cf. Figure A.l.l)

R R R R
. YQo YQo xQo xQo
Slll!.{JQR = -R1I= -R1I' COS!.{JQR=-1
R 1=-
R 1I'
rQo rQo rQo rQo
(A.l.4d)
R R R R
. YPo YPo Xpo Xpo
Slll!.{JpR=-1
R 1=-
R 1I' COS!.{JpR=-1
R 1=-
R 1I'
rpo rpo rpo rpo
and
cos('PQR- 'PPR) = cos(r.pQP) (A.1.4e)
= COS!.{JQR · COS!.{JPR + sin!.{JQR · sin!.{JPR·
Furthermore,
A.l Special vector and matrix Operations used in mechanics 259
Fig. A.l.l: Vector diagram of a scalar product
R R R R
( ) xQo Xpo YQo Ypo
cos 'PQP = -R1I -R1I + -R1I -R1I (A.l.4f)
TQo Tpo TQo Tpo
0 0
or
R R
xQo · Xpo + YQo
R R
· Ypo
cos(r.pQp)= IR 1·1 R I
TQO Tpo
Inserting (A.l.4f) into (A.l.4c) yields
R I I R I
ITQo. Tpo. cosr.pQp= TQo. Tpo.
I R I I R I x~o ·IR
x~o 1·1
+ Y~o
R I
· Yfto (A.l.4g)
TQO Tpo
R R R R
=XQo·Xpo+YQo·Ypo= ( R R )
XQo"YQO. ( X~O
R ) =: ( TQo
R )T R
·Tpo·
Ypo
D
Comment A.1.2 (Scalar product and orthogonality ofvectors): Due

to (A.1.4a), two vectors T1 and T 2 areorthogonal to each other iff their scalar
product is zero.
Vector product
The vector product of two geometrical vectors TQO and T PO from the x-y
plane (Figure A.1.2), is defined as the vector
Tc:= TQo X Tpo = ITQoi·ITPol · sinr.pQP · ezR = ITcl· ezR, (A.l.5a)
with ezR as the unit vector perpendicular to the exR-eyR plane.
Algebraic computation of the vector product
The vector product Tc is computed from the vectors

260 Ao Appendix
€zR x~ 0
r~ = r~ 0 x r~0
Figo Aol.2: Vector diagram of a vector product
R -
rqo- (x~
Yqo
0) (Aol.5b)
and
rpo = Xpo
R 0
exR + YPo
R 0
eyR or R -
Tpo- (x~0 )
Ypo
as
R -R R
Tc = Tqo Tpo, 0
(Aol.5c)
with the algebraic vectors r~ 0 and r~ 0 extended to the OC3 ; ioeo,
Tpo = Xpo
R 0
exR + YPO
R 0
eyR +O 0
€zR = ( Xpo,
R R
YPO, o)T
and
R
Tqo R
= xq 0 ° exR + Yqo
R 0
eyR +O 0
ezR = ( Xqo,
R R
Yqo, O)T ,
and with the skew-symmetric spatial matrix
_R ( 0 , 0 , Y~o)
rqo = 0 0 , -xq 0 (Aol.5d)
-y~o , x~o , 0
built from the extended spatial vector r~ 0 Then 0
Y~o)
( Y~oR)
Xpo
0 , 0 ,
R -R R
Tc = T QO 0 T PO = ( 0 0 -xq 0 0
-y~o 'x~o ' 0

(Aol.5e)
or
Tc-
R YPO-
_ ( XQo R R Xpo
YQO R ) €zR
0 0 0
-o
-o Zc 0
€zRo
Proof:
By the definition of the vector product, the relation
r~ = lr~ 0 I lr~ 0 I sin 'PQP

0 0 (Aol.5a)
holdso Then
Together with the scalar product
(r~ 0 ) T (r~o) = lr~ololr~olo

0
cosr.pQP
and
this yields
lr~l 2 = {lr~ol 2 olr~ol 2

0
[1r~ 0 1 2 lr~0 1 2 - ( (r~ 0 ) T (r~o))
0 0
f} / (1r~ol 2 olr~ol 2 ),
r
or
lr~l 2 = lr~ol 2 olr~ol 2 - [ (r~o f (r~o) 0
= [(x~o) 2 + (Y~o) 2 ] 0
[ ( x~ 0 ) 2 + (Y~o) 2 ]
R OXpo
- ( XQO R +YQO
R OYPO
R )2
= (x~of (x~of + (Y~of (Y~of

0 0
+ (x~o) 2 Y~o) 2 + (x~o) 2 Y~o) 2 - ( x~ 0 ) 2

0
(
0
(
0
( x~ 0 ) 2
R )2
- (YQo 0
R )2
( YPo R R R R
-2oxQooXpooYQooYpo,
and finally
262 Ao Appendix
2
I rcR 1
2
= ( zcR) = ( R R R R )2
XqooYPo-YqooXpo
or
Zc = , (R
A. Xqo
0 0
R
Ypo- R
Yqo R) ezR
Xpo 0 0 A. = +1 or - 1. (Aol.6a)
On the other hand
R R -R R 0 , 0 , Y~o ) ( x~o )
Tqo X Tpo= Tqo Tpo = ( 0 0 0 -xq 0 ° Ypo
-y~o , x~ 0 , 0 0
(Aol.6b)
00 ) -(R R
- Yqo Ypo- R
Yqo R) ezRo
Xpo
( 0 0 0
R R R R
xqo Ypo- Yqo Xpo
0 0
Comparing (Aol.6a) and (Aol.6b) proves (Aol.5e) for A. = +1. D
A.1.3 Cross product of spatial vectors
The vector product of two spatial vectors r PO and rqo, represented in R, is

in agreement with Appendix Aolo2:
(Aol.7a)
+ ( Zpo
R 0
R
Xqo- R
Xpo R) eyR
zqo 0 0
+ R R R R ) oezR
( XpooYqo-YPooXqo
or
Ypo
R zqo-
R zpo
R Yqo
R ) 0 0
R R ( R R R R (Aol.7b)
T PO X T QO = ZPO Xqo - XPO Zqo 0 0
o
R R R R
Xpo Yqo- Ypo Xqo 0 0
Introducing the skew-symmetric matrix
0 , -zpo
R
, Ypo
R )
-R ( R R (Aol.7c)
z~o ' ~ ' -Xpo
o-
Tpo o-
-yp0 xp0 , 0
constructed from the vector r~ 0 , the product

A.l Special vector and matrix Operations used in mechanics 263
-R R
rpo ·rQo =
(
Zpo'
~
' - Z~o ' Y~o )
0 '-Xpo .
( X~o )
YQo (A.l.7d)
-Ypo
R R
Xpo '
0 R
zQo
Y~o · z~0 - z~0 · Y~o )

( R R R R
Zpo. xQo- Xpo. zQo
R R R R
Xpo · YQo- Ypo · xQo
yields the same vector as the cross product (A.l.7a). As a consequence, the
vector product (A.l.7b) can be replaced by the matrix relation (A.l.7d)
R R -R R TID3
r PO X r QO = r PO . r QO E ~ . (A.l.8)
The vector product of two vectors (in JR3 ), represented in frames R or L,

satisfies the relation
-R R ARL -L ALR ARL L ARL -L L )
rpo·rQo= ·rpo~·rQo= . ( rpo·rQo
= 13
(A.l.9a)
with
r-R _ ARL r-L
PO- . PO.
ALR
· (A.l.9b)
Furthermore, the following relations hold:
1.
-R
( rpo )T = -R
-rpo = rop·
-R
(A.l.9c)
2. (r~OL + T~LO) = T~OL + T~LO· (A.l.9d)

3. -R
TpoL. R
TpoL = 0. (A.1.9e)
4. -R
TpoL. -R
TpoL = (-R
TpoL )T . (-R
TpoL )T = (-R -R
TpoL. TpoL )T . (A.l.9f)
5. r~0L r~0L [(r~oJ T·r~oL r~oL (r~oJ Tl

· =- ·l3- · (A.l.9g)
scalar product dyadic product
This latter relation will be used in representations of the inertia matrix of a

rigid body in Chapter 4.
Proof of A.1.9c:
By definition,
264 A. Appendix
-R )T := -zf5o , Y~o) T 0 , zf5o , -y~o )

( Tpo 0 , -xp 0 ( -z~0 , 0 xp 0
x~ 0 , 0 Ypo , -x~o , 0
, -zf5o , Y~o)
( z!
-YPo
0
0
x~ 0 ,
-xp 0
0
= -R -R
-rpo =Top·
This proves (A.1.9c). D
Proof of A.1.9d:
The skew-symmetric matrix associated to the vector sum
is
On the other hand
-R
TpQL (
0
ZpQL
R
'-ZpQL
R
0
, Y:o,)
-XpQL '
R 0
, Y~,o)
-yf5oL XpQL '
R
'-zoLo
-R ( R0
0
roLo = zoLo -xoLo '
R
-yfto xaLo' 0
and
Relation (A.l.lOa) together with (A.l.lOb) implies
D
Proof of (A.1.9e):
The relation
proves (Ao1o9e}o D
Proof of (A.1.9f):
The expressions
and
imply that the relation
holds This proves (A 010 9/)

0 0 D
Proof of (A.1.9g):
266 Ao Appendix
1,0,0)
( 0,1,0
0 ' 0 ' 1
due to the proof of (Aol.9f)o This proves (Ao1o9g)o D
A.1.4 Time derivatives of planar orientation matrices and of

planar vectors in different frames
In this section the following relations will be proved:
R oL o- Rd ( L ) _ L oL LR RL L
(Aol.lla)
+A
0
TpQL o- dt TpQL - TpQL OA OTpQL'
LR RL -L
A (Aol.llb)
0 0
OA =WLR=Ro'l/JLR,
or
.ARL = ARL 0 wfR = ARL 0 R 0 ?j;LR, R= (~ , -~) (Aol.llc)

with
WLR = ?j;LR 0
€zL, (Aol.lld)
RoL LoL -L L
TpoL = TpoL + WLR 0 TpoL' (Aol.lle)
and
Proof of (A.l.ll):
Assurne that the vector r POL (t) with P fixed on a moving frame L (see Figure
206) is represented in L, but differentiated with respect to the time in frame
Ro Then
Rr POL := R~t ( X~oL (t) 0

exL(t) + Y~oL (t) 0
eyL(t)) (Aol.l2a)
with basis vectors eiL(t) that are not constant with respect to Ro Applying
the product rule of differentiation to (Aol.l2a) provides the relation
RTPOL(t) = (R~tX~oL(t)) oflxL +x~0 L(t) 0 R~t (exL(t)) (Aol.l2b)
+ ( R~tY~oL (t)) 0flyL + Y~oL (t) 0 R~t (eyL(t)) 0
By definition of the product rule of differentiation the basis vectors eiL,

marked by underlined letters in (Aol.l2b) are considered to be constanto The
time derivatives of the scalar functions x~ 0 L and y~ 0 L are
and (Aol.l3)
This implies
x~oL (t) 0
exL(t) + x~ 0 L (t) 0
Rd (exL(t))
dt
(Aol.l4)
+ il~oL (t) eyL(t) + y~ 0 L (t)

0 0
R~t (eyL(t)),
where the overlined basis vectors eiL (i = x, y) in (Aol.l4) may change with
respect to frame Ro They are no langer constant in time - they rotate with
respect to frame Ro Representation of the basis vectors eiL of L (i = x , y)
in frame R according to (2ol2b) and (2ol2c) yields, together with
exL(t) = cos'lj; 0exR + sin'lj; 0eYR' eyL(t) = - sin'lj; 0exR + cos'lj; 0eYR'
(Aol.l5)
268 Ao Appendix
the relations
Rd Rd
dt (exL(t)) = dt (cos'lj; 0
exR + sin'lj; eyR) 0 (Aol.l6a)
and
Rd Rd Rd
- (eyL(t)) = - (- sin 'ljJ exR) + - (cos'lj; eyR) 0 0
(Aol.l6b)
dt dt dt
0 0
~
= -sm 'f' 0 exR - (sm
o ·'·)
'f'
o
0 R exR + -------;;/;
cos 'f' 0 eyR + cos .!,
'f'
o
0 R eyRo
Since the basis vectors eiR are constant in frame R, the relations
(Aol.l7)
imply
Rd
dt (exL(t))
0
= [(- sin '1/J) exR + cos'lj; eyR]o 'ljJ

0 0
and
Rd
dt (eyL(t))
0
= [(- cos'lj;) 0 exR- sin 'ljJ 0 eyR] 0 'lj;o
Inserting these relations into (Aol.l4) yields the relation

Rd
dt (rpoL (t)) = (x~oL exL + Y~oL eyL)
0 0 (Aol.l8)
L
+X poL sin '1/J) exR + (cos'lj;) eyR]o '1/J
0
0
[(-
L
+ YPoL cos'lj;) exR- (sin '1/J) eyR] 'ljJ
0
0 0
[(-
or
R~t(rpoL(t))= (x~ 0LoexL+Y~oLoeyL) (Aol.l9a)
+ [(x~ 0 L 0
(-sin'lj;)+y~ 0 L 0
(-cos'I/J)) oexR
+(x~ 0 L 0
cos'lj;+y~ 0 L 0
(-sin'I/J)) oeyRJ ?j; 0
0
Representing the basis vectors exR, eyR according to (2013) in frame L yields
R~t (r POL (t)) = ( x~0 L 0

exL + Y~oL 0
eyL) (Aol.l9b)
+ [(x~ 0 L 0 (-sin'lj;)+y~ 0 L 0 (-cos'I/J)) 0 (cos'lj;oexL-sin'lj;oeyL)] o?j;

+ [(x~ 0 L 0 cos'lj;+y~ 0 L 0 (-sin'I/J)) 0 (sin'lj;oexL+cos'lj;oeyL)] o?j;
or
R~t (rpoL(t)) = (x~0 L oexL +y~0 L oeyL) (Aol.l9c)
+ [(- sin "p) 0 (cos "p) 0x~ 0 L + (- cos "P) 0 (cos "P) 0y~ 0 L
+ (cos "P) 0 (sin "P) 0x~ 0 L + (- sin "P) 0 (sin "P) 0y~ 0 J 0exL 0~
+ [(-sin"P) 0 (-sin"P) ox~ 0 L + (-cos"P) 0 (-sin"P) oy~ 0 L

+ (COS 'lj1) o (COS 'lj1) oX POL
L + (- sin 'lj1) o (COS 'lj1) oy POL
L ] oey L o'lj1 00
In order to obtain the formal expression (Aol.lla), several terms of (Aol.l9c)

that cancel will not be dropped in the next stepso Then (Aol.l9c) may be
written as
R~t (r~OL (t)) = ( ~~OL ) (Aol.20)

YpoL
[(cos "p) 0(- sin "p) + (cos "P 0 sin "P)] 0x~ 0 L
+~0 [
[(- sin "p) 0(- sin "p) + (cos "P) 0(cos "P)] 0x~ 0 L
+ [cos "P 0(- cos "P) + (sin "P) 0(- sin "P)] 0Y~o L ]
+ [(- sin "P) 0(- cos "P) + (cos "P) 0(- sin "P)] 0y~ 0 L
or
Rd
dt
(rLPOL (t)) - (
-
j;~OL)
oL +
YpoL
cos 'ljJ 0(- sin 'ljJ) + cos "P 0 sin "P , cos 'ljJ 0(- cos "P) + sin 'ljJ 0(- sin "P) )
(
(- sin "P) 0(- sin 'ljJ) + cos 'ljJ 0 cos "P , (- sin 'ljJ) 0(- sin 'ljJ) + cos 'ljJ 0(- sin "P)
or as
(Aol.21a)
and together with (cfo Equation 2032a)
(Aol.21b)
270 Ao Appendix
as
(Aol.22)
and finally as
R L
= L TpoL
L LR RL L
(Aol.23)
0
TpoL +A OA orpoL
o o
together with
A RL 0Rro POL
L _ ARL 0Lro L + Ao RL 0rL 0 (Aol.24)
- POL POL
This proves (Aolo11a}o D
Using
ARL = Rd ( cos 'ljJ , - sin 'ljJ ) (

- sin 'ljJ , - cos 'ljJ ) 0 0
dt sin 'ljJ , cos 'ljJ cos 'f' , - sm 'f'
·'· 0 ·'·'1/J,
provides
A LR 0ARL = ( cos '1/J , sin 'ljJ ) 0 ( cos 'ljJ , - sin 'ljJ )

(Aol.25a)
- sin 'ljJ , cos 'ljJ sin 'ljJ , cos 'ljJ
= ( c~s 'ljJ , sin 'ljJ ) 0 ( - sin 'ljJ , - c~s 'ljJ ) 0~

- sm 'ljJ , cos 'ljJ cos 'ljJ , - sm 'ljJ
= ( 0 ' -1 ) 0~ = ( ~ ' -~)

1 ' 0 'ljJ ' 0
...__"._..,
=R
proves (A01011b}, and yields (Aol.llc) directly,with the planar orthogonal
rotation matrix R defined in (2022a), and (cfo Comment Aol.3) with
WL
LRO=
0
Of WLR = ~LR ezL 0
as the angular velocity around the ezL-axis0 This proves (A01011d}o D

Together with (Aol.llb), (Aol.lla) yields
R oL
TpoL
oL
= L TpoL -L
+ WLR L
0TpoL oL
= L TpoL +R L
0TpoL 0.i,
'f'LR
or, written in components,

(Aol.25b)
or
R oL L oL L .i, (Aol.25c)
X poL = X poL- YPoL 0
'f'LR
and
This proves (A01011e}o D
Comment A.1.3 (Formal notation wfR): The above notation using w LR

is motivated by the intention to write the equations of motions of planar rigid-
body systems in a form that is identical to the equations of motion of spatial
mechanisms, discussed in Section Aol.5 (compare Section Aol.5 and Equation
2027)0
In the spatial case,
L )L
WxLR
( W~LR (Aol.26a)
wfR :=
WzLR
is the angular velocity vector of frame L with respect to frame R, represented
in Lo Then wfR is defined as the skew-symmetric matrix
L L L
-L ( , -WzLR , Wy~R )
0
WLR := W~LR ' 0 ' -WxLR (Aol.26b)
-w{;LR ' W~LR ' 0
In the planar case, only vectors r = xL exL + yL eyL or rL = (xL, yL)T
0 0
with basis vectors { exL , eyL} of the reference frame L inside the x-y plane
occur, where frame L can only rotate around the ezL-axis perpendicular to
this plane by an angle '1/JLRo Then the angular velocity vector wfR associated
with this rotation around ezL may be written in an extended form in ffi.3 as
with
-L
WLR I =
JRß
272 Ao Appendix
Projecting wiR I into the x-y plane yields, in agreement with (Aol.llb),
JR3
-L
WLR := WLR
-L I
lR 2 =<emL, eyL >
= (0,-1)
1 , 0
0
0
'l/JLR =
(Aol.25a)
R 0
0
'l/JLR, (Aol.28)
with Ras the planar orthogonal rotation matrix (2022e)o

Mapping the vector R.;.~ 0 L by means of ARL into frame R yields, together
with (Aol.lla) and (Aol.llb), the relation
A RL 0
R L
TpoL=
o ARL (L L - L L
TpoL+WLROTpoL
o )
0
(Aol.29a)
On the other hand
or
Together with
Rd ( L ) oL
an d dt XpoL = XpoL' (Aol.29b)
this yields
+ (
cos .!, oL
'f'LR 0X poL - sm .!,
o oL
'f'LR 0YpoL
)
'
.!. L
sm 'f'LR 0xp 0 L
0 0
+ cos .!.'f'LR 0YpoL

oL
or
- sill'l/JLR , -
( ~~0)
cos'l/JLR)
(
Ypo COS '1/JLR , - sin '1/JLR
and finally
(Aol.29c)
The relations (Aol.29a), (Aol.llb), and (Aol.29c) imply
R oR Rd ( R ) RL R oL RL ( LR RL L oL )
dt rpoL = A A A A + L rpoL
0
rpoL = 0 rpoL = 0 0 0 rpoL
(Aol.29d)
This proves (Aolo11f}o D
On the other hand (Aol.29d) also implies
(Aol.30)
D
A.1.5 Time derivatives of spatial orientation matrices and of

spatial vectors in different frames
In this section the following spatial relations will be proved:
R L Rd ( L ) L L
+ A LR RL L
(Aol.31a)
0
:= dt rpoL = A
o o
rpoL rpoL 0 0 rpoL'
A LR 0
Ao RL- -L
-WLR or Ao RL-
-
ARL -L
OWLR' (Aol.31b)
RoL LoL -L L
TpQL = TpQL + WLR 0
TpQL' (Aol.31c)
Rro R _ ARL
POL- 0
( - L
WLR 0
L
TpoL + L TpoL
L
0
)
=
ARL 0
R 0
L
TpoL' (Aol.31d)
274 Ao Appendix
R 0L - A LR 0R 0R (Aol.31e)
TpQL- TpQL'
and
(Aol.31f)
except
R.;.~ 0 L -1- ARL 0Lr~ 0 L for wfR -1- Oo (Aol.31g)
The formal vector wfR introduced here does not yet have any physical mean-
ingo A physical interpretation of wfR is introduced in Section 20202030
Proof of (A.1.31):
Consider the time derivative of a vector r~ 0 L represented in frame L and

differentiated with respect to time t in frame Ro Then
with the basis vectors eiL of L (i = x , y , z) that are not constant in time
when measured in frame Ro They may rotate with respect to frame Ro Ap-
plying the product rule of differentiation to (Aol.32a) yields
RTpoL(t)= (R~tX~oL(t)) o~xL+X~oL(t)o R~texL(t)

+ ( R~tY~oL (t)) 0 ~yL + Y~oL (t) R~t eyL(t)
0
+ ( R~/~OL (t)) 0
~zL + Z~OL (t) R~t ezL(t)o
0 (Aol.32b)
The basis vectors eiL of L, marked by underlined letters in (Aol.32b), are

kept constant with respect to frame R by definition of the product rule of
differentiationo The time derivatives of the scalar functions XpoL, YPOL, and
ZpoL are
This implies
Rr POL (t) = x~ 0 L (t) 0

exL + Y~oL (t) 0
eyL + z~ 0 L (t)
(Aol.32d) 0
ezL
Rd Rd Rd
+ x~ 0 L (t) dt (exL(t))
0
+ y~ 0 L (t) dt (eyL(t)) + z~ 0 L (t) dt (ezL(t)),
0 0
where the overlined basis vectors in (Aol.32d) may change with respect to
frame R and are therefore no langer treated as constantso Representation of
a basis vector eiL of L in frame R according to (2.46) yields, together with
eiL(t) =Ci exR 0

+ mi 0
eYR + ni 0
ezR or efi(t) =(Ci, mi, ni)T (Aol.32e)
and
Ci = IeiL I IexR I cos a = cos a , a = angle between eiL and eiR,
0 0
(Aol.32f)
the relations
Rd Rd
-(eoL(t)) = -(C·oe R+m· Oe R+n·oe R)
dt ' dt ' X ' y ' z
and
(Aol.32g)
Rd Rd Rd
+ C·' 0
- (e R)
dt X
+ m· '
0
- (eR)+ n·
dt y '
(eR)
dt z
0
- 0
As the basic vectors eiR of (Aol.32g) are constant in frame R, the relations
Rd
dtejR = 0, for j = x, y, z (Aol.32h)
imply
R~t (eiL(t)) =Ci 0

exR + mi 0
eyR + ni 0
ezR, i =X, Y, Z, (Aol.32i)
with the time derivatives of the scalar functions Ci , mi , ni defined as

0 Rd Rd Rd
Ci := dt (Ci) = dt (Ci) = dtci, etco (Aol.32j)
Then (Aol.32b) can be written as
(Aol.33a)
276 Ao Appendix
This implies, together with (Aol.32i), the relation
Rd
dt (rpoL (t)) = ( x~oL 0 exL + Y~oL 0 eyL + i~oL 0 ezL) (Aol.33b)
+ ( XpoL
L 0
L
ofx+YPOL L L ofz
ofy+zp 0
0 0
)
oexR
+ (x~oL 0 rhx+ Y~oL 0 + z~oL rhz) eyR

rhy 0 0
+ (X~oL 0 nx + Y~oL 0 ny + Z~oL nz) ezRo 0 0
Representing the basis vectors exR, eyR, and ezR by means of (2o47a) in
frame L,
exR = L Ci eiL
0
eyR = L mi 0
eiL ezR = L ni eiL
0
i=x,y,z i=x,y,z i=x,y,z
yields the relations
(Aol.34a)
or
or
(Aol.34c)
(~x of!x + rhx 0

mx + nx 0
nx) 0
X~OL
( (f!x of!y + mx ffiy + nx ny)
0 0 0 X~OL
(Rx of!z + rhx mz + nx nz) X~oL

0 0 0
+ (Cy of!x + rhy mx + ny nx) Y~oL

0 0 0
+ (fy l!y + rhy my + ny ny) Y~o L

0 0 0 0
+ (Ry of!z + rhy mz + ny nz) YpoL

L
zfo,)
0 0
0
+ (Rz of!x + rhz mx + nz nx) 0 0
+ (Rz of!y + mz my + nz ny) ZpQL

0 0
0
+ (Rz of!z + rhz mz + nz nz) ZpQL

L 0 0
0
~x oRx + ~x omx + ~x onx ,

( Rx ol!y + mx omy + nx ony ,
Rx Rz
0
+ rhx mz + nx nz
0 0
'
fy Rx + rhy mx + ny nx ,
0 0 0
fy l!y + rhy my + ny ny '
c)
0 0 0
n,)
Ry Rz 0
+ rhy 0
mz + ny nz ' 0
fz fx
0
+ rhz 0
mx H, XpQL
fz f!y 0
+ rhz o my + nzony 0
YfoL '
fz of!z + rhz 0
mz + nz nz 0 ZpQL
or, together with (2o53a),

Rd
dt (r~ 0 L (t)) (Aol.34d)
278 Ao Appendix
RL
=:A
o
and finally with
(Aol.34e)
the relation
R L Rd ( L ) Ld ( L ) LR RL L
TpoL := dt TpoL(t) = dt TpoL(t) +A OA OTpoL
0
o
or (Aol.35)
R L L L LR RL L
TpoL = TpoL +A OA OTpQLo
0
o o
Differentiation of the orthogonality relation
(Aol.36)
with respect to time yields the relations
or
(Aol.37)
or
This proves that the matrices
( ALR 0 ARL) and
are skew-symmetric; ioeo, that
LR RL
Each skew-symmetric matrix A A can be generated by a formal vector
0
L
WLR = ( L L L
WxLR' WyLR' WzLR
)T ' (Aol.38)
represented in frame L according to the relation
A LR 0
Ao RL- -L
-WLR or Ao RL-
-
ARL -L
OWLR (Aol.39a)
with the skew-symmetric matrix
-L ( 0 '
L
-W~LR'
0
W~LR) -L
(Aol.39b)
WLR := W~LR' -wxLR =: -WRLO
L
-WyLR' wxLR' 0
This proves (Aolo31b}o D
Inserting (Aol.39a) into (Aol.35) yields the relation

RoL
TpoL = LoL -L
TpoL + WLR 0
L
TpoL 0
(Aol.40a)
This proves (Aolo31c}o D
Interchanging the indices R and L in (Aol.40a) yields

LoR RoR -R R RoR ARL -L L
TpoL= TpoL+WRLOTpoL= TpoL- oWLROTpQLo (Aol.40b)
U sing the similarity transformation of matrices
ALR 0 BR (ALR) -1 = BL B, ALR E oc3,3 (Aol.41a)
and the orthogonality property
(ALRr 1= (ALR) T= ARL
yields (compare also Equation 2.41c of Section 20202)

BL = A LR 0 BR 0 A RL and -R _ ARL
W LR- 0
w-L
LR 0
ALR
o
(Aol.41b)
Interchanging the indices R and L in (Aol.39a) yields

RL LR -R -R Ao LR - ALR - R - ALR - R
or
0
A OA =WRL=-WLR - 0 W RL - - 0 W LR
together with (Aol.41b),

Ao LR-
-
ALR -R
OWRLO
ARL ALR- -L
-WRLO
ALR- -L
--WLRO
ALR 0
0 (Aol.41c)
Multiplication of (Aol.40a) from the the left-hand side by ARL yields
(Aol.42)
On the other hand,

280 Ao Appendix
Taking into account Bryant angles yields, together with (cfo Equation 2056)
c8oc'lj; , Cipos'lj;+sipos(}oc'lj;, Sipos'lj;-cipos(}oc'lj;)

ALR = ( -ce s'lj; ' Cip c'lj;- Sip se s'lj; ' Sip c'lj; + Cip se s'lj;
0 0
' 0 0 0 0 0
s(} , -SipoC(} , CipoC(}

(Aol.43a)
and (cfo Equation 2070)
ÄLR= (Aol.43b)
-s8oc'lj; 0 0 Ö-c8os'lj;o~ ,
( se 0 s'lj; e- ce c'lj; o'lj;
0 0
ceoiJ
scpos'lj;o~+ccpoc'lj;o?j;+ccpos(}oc'lj;o~+scpoc(}oc'lj;oÖ-scpos(}os'lj;o?j; ,
scpoc'lj;o~-ccpos'lj;o?j;-ccpos(}os'lj;o~-scpoc(}os'lj;oÖ-scpos(}oc'lj;o?j; ,
-c ifJ c e ~ + s ifJ s e iJ
0 0 0 0
+c ifJ s 1/J ~ + s ifJ c 1/J ~ + s ifJ s e c 1/J ~ c ifJ c e c 1/J ~ + c ifJ s e s 1/J ~
0 0 0 0 0 0 0
-
0 0 0 0 0 0
)
+c ifJ c 1/J ifJ - s ifJ s 1/J 1/J - s ifJ s e s 1/J ifJ + c ifJ c e s 1/J e + c ifJ s e c 1/J 1/J
0 0 0 0 0 0 0 0 0 0 0 0 0
,
-s ifJ c e ~
0 0
- c ifJ s e
0 0
e
the relation
Ld ( Ld ( LR
dt TpoL
L )
=: dt A 0
R
rp 0 L
)
= (Aol.43c)
c8oc'lj; , ccpos'lj;+scpos(}oc'lj;, scpos'lj;-ccpos(}oc'lj;)
L~t [ ( -cBos'lj;, ccpoc'lj;-scpos(}os'lj;, scpoc'lj;+ccpos(}os'lj;
sO , -scpoc(} , ccpoc(}
G::Jl
-sOoc'lj;ooÖ-cOos'lj;o~,
( sOos'lj;o()-c()oc'lj;o'l/J (Aol.43d)
cOOÖ
scpos'lj;o~+ccpoc'lj;o?j;+cc/Jos(Joc'lj;o~+scpocßoc'lj;oÖ-scpos(}os'lj;o?j; ,
scp c'lj; ~- ccp s'lj;
0 0 0 0
-J;- ccp 0
sß s'lj; ~- scp cß s'lj;
0 0 0 0 0
e- scp 0
sß c'lj;
0 0
-J; '
-c cp c (} ~ + s cp s (}
0 0 0 0 e
+c rjJ s 'ljJ ~ + s rjJ c 'ljJ ~ + s rjJ s () c 'ljJ ~

0 0 0 0 0 0 0
- c rjJ c () c 'ljJ ~ + c rjJ s () s 'ljJ ~
0 0 0 0 0 0
)
+c rjJ c 'ljJ rjJ - s rjJ s 'ljJ 'ljJ - s rjJ s () s 'ljJ rjJ + c rjJ c () s 'ljJ
0 0 0 0 0 0 0 0 0 0 () + c rjJ s () c 'ljJ 'ljJ
0 0 0
-s rjJ c () 0 0 (p - c rjJ s () iJ 0 0
' ~ ~
= _ÄLR
c B c 'ljJ , c rjJ s 'ljJ + s rjJ s B c 'ljJ , s rjJ s 'ljJ - c rjJ s B c 'ljJ)

0 0 0 0 0 0 0
( x~oL )
+ ( -C (} 0 S '1/J ' C rjJ 0 C '1/J- S rjJ 0 S (} 0 S '1/J ' S rjJ 0 C '1/J + C rjJ 0 S (} 0 S '1/J 0 ~!OL
sB , -srjJocB , crjJoc() ZpoL
'---------------------v---------------------J
= ALR
~
Rr~oL
and the relation
L L Ld ( L )
o
TpQL = dt TpQL
LR R LR Rd ( R LR R
+A )
+ A LR R R
0 0
= dt =
o
A 0 TpoL 0 TpoL A 0 TpoL 0 TpoL
or
(Aol.44)
and finally, together with (Aol.40b), the relations
(Aol.45)
and
R L - A LR R R RoR - ARL RoL
0
TpQL-
0
0
TpQL or TpoL- 0
TpQLo (Aol.46)
This proves (A01031d} and (A01031e}o D
Inserting (Aol.31c) into (Aol.31d) yields
Rr-RPO = ARL 0 Rr-LPO = ARL 0 Lr-LPO + ARL 0 WLLR 0 TLpQ (Aol.47)

L L (Aol.31c) L L
or
R oR
TpoL = ARL
0
( - L
WLR 0
L
TpoL + L TpoL
L ) o
' (Aol.48)
282 A. Appendix
and, together with (A.l.39a), the relation

R·R ·RL L
rpoL = A . rpoL + A RL . L·L
rpoL· (A.l.49)
This proves (A.1.31f). D
Then
Rd Ld ( L ( )) w.fR-::/:- 0.
t -::/:- A RL . dt
R
d t r POL
( )
r POL t = A RL . L r· POL
L ( )
t for
(A.l.31g)
This proves (A.1.31g). D
A.1.6 Derivatives of vector functions
In the kinematics and dynamics of mechanical systems, vectors representing

the positions of points on bodies, or equations describing the geometry of
the dynamics of the motion are often functions of several variables. When
analyzing these equations, partial derivatives of vectors and functions with
respect to these variables occur. These situations can be efficiently handled
by vector and matrix calculus notation.
Gradient of a mapping
Let
and
g E C1 (I~n) (g is a differentiable function)
\.1) \.1)
Then gP := 8gj8p is a row vector (called the gradient of g):
Yp = (8gj8p1, ... , 8gj8pn).
Example A.l.l (Gradient of a mapping): Let
(A.l.50a)
-og) = ( -og , -og) = (2 · x 1 + 3 · x2 , 3 · xl) .

8p2 OX! OX2
(A.l.50b)
The gradient of the special functions
91 : ocn X ocn ---+ ocl (Aol.51a)

U! U! U!
(y x) f---t 91(y, x) :=yT OAOx A=AT,

'
and
92 : ocn X ocn ---+ ocl (Aol.51b)
U! U! U!
(x x) f---t 92(x' x) := xT A x A=AT

'
0 0
is
(Aol.52)
and
Y2x = [:x (xT OAOx)] = 2° (Aox) = 2oxT Ao 0 (Aol.53)
Proof of (A.1.52):
Then
gJx= [:x (xTOAOy)r= [:x (yTOAOx)r=AToy=A yo 0

(Aol.54a)
D
284 Ao Appendix
Proof of (A.1.53):
Y2x = :X (X T 0 A 0 X)
[ 0~ 1 (ot aij Xi Xj) , Ö~n (ot aij Xi Xj)]
t
= 0 0
00 0
,
0 0
<,J=l <,J=l
= [t ail 0
Xi + ~ a1j 0
Xj , 0 0 0 , ain ° Xi + ~ anj 0
Xj]
= [ (au, 0 0 0 , an!) x 0
+ (au, 0 0 0 , aln) x ,0
0 0 0 ,
(aln, ooo , ann) o X+ (anl, ooo , ann) ox]
= xT 0
( a~1 ,
0
;
0
, a~n) + xT 0
( a~1 ,
0
;
0
, a~1)
anl ' 0 0 0 ' ann aln ' 0 0 0 ' ann
or
and
(Aol.54b)
Jacobian matrix of a mapping
Let
p E ]Rn
and
(Aol.55a)
P f-t g(p) = (gl(P), 00 o, 9m(P))T
= (gl (Pl , 00 o , Pn) , 00 o , 9m (Pl , 00 o , Pn)) T o
Then
8gl/8pl , ooo, 8gl/8Pn)

8g
_- (
g p o -8p-
o-
(Aol.55b)
8gmf8pl , ooo , 8gmf8pn
is called the Jacobian (matrix) of g(p)o
Example A.1.2 (Jacobian matrix of a mapping): Let
P = (X 1, X 2, X 3 ) T' g 00 llll3
Jß
----' llll2,
---, Jß
with
91(P) = xr + 2x1X3 + X2X~ + X~ and

(Aol.56a)
g2(p) = 4 + xix~ + x{x3 + x2x~o
Then
8g , 2xl + 2x2x3 + 4x~ )

gp o--- E JR2,3
o- 8p- 0
, xi + 3x2x~
(Aol.56b)
The following relations are easily proved by direct calculationso Let
p E ]Rn and a , b , c : ]Rn -+ ]Rn (Aol.57)
Then:
1.
a
(a • b)p := f)p (a • b) = b • ap + a • bp E JR1 ,m 0 (Aol.58a)
20 (a • b)p -/:. bp • a + a • bp (a not well-defined right-hand side)o

(Aol.58b)
(Aol.58c)
Proof of (A.1.58a):
Select an arbitrary basis of R in JRm Then for p = (p 1, 0 0 0 0 ,pn)T

286 Ao Appendix
Example A.1.3 (Example related to Equation A.1.58a): Let
(Aol.59a)
and
o[(aR)T bRJ
0
(Aol.59b)
op
2 ox1 oX~ ox 3 + 2 ox1 oX~ + 4 ox1 oX~ + 3 oxi ox 3 T
+2 0 Xl 0 X2 0 X3 +8 0 xr 0 X~ 0 X3
2 ox2 oxi ox 3 + 3 oX~ oxi + 8 oX~ oxi + xi ox 3

+4 ox 2 oXf ox 3
xi oX~ + xr + xi oX2 + 2 oXf oX~
l
Xt (2 X~ X3 + 2 X~ + 4 X~ + 3 Xt X3 + 2 X2 X3 + 8 XI X~ X3)
0
° 0
° ° ° 0
° ° ° 0 0
T
[ XI o (2 ox 2 ox 3 + 3 oX~ + 8 oX~ + x 3 + 4 ox 2 oxi ox 3 ) 0
xi (x~ + x 1 + x 2 + 2 ° xi x~)
0 0
X~ , 2 Xt X2 , 0 1
r
o o
(1 +2 ° X2) , 2 ° Xt , 0 ,
(1 +4 ° x 1 x~) , (1
° +4 ° xi x 2 )
0
, 0
(Aol.59c)
, Xt1
' 0 '
'xi
(Aol.59d)
Xl o (x~ ox 3 +X~ o(1 + 2 ox 2 ) + Xt X3 (1 + 4 Xt X~)) 0

° ° 0
1
T
r xi o (2 ox 2 ox 3 + 2 oX~+ x 3 o(1 + 4 oxi ox 2 ))

0
(Aol.59e)
Xt oX~ ox 3 + Xt o(1 + 2 ox 2) oX~ + (x 1 + x 2 + 2 oxi ox~) o2 oXt ox 3 ] T
[ Xt o(1 + 2 ox 2 ) o3 oX~ oXt

xl2 x22 + ( Xt + x2 + 2 xl2 x22) xl2
0 0 0 0
and finally
288 Ao Appendix
(bR)T o(aR)0 + (aR)T 0 o(bR) = (Aol.59f)

8p 8p
o (x~ ox 3 +X~ o (1 + 2 ox 2) + Xt ox 3 o(1 + 4 oXt ox~)
T
Xt
+x~ ox 3 +X~ o(1 + 2 ox 2) + 2 ox 3 o (x 1 + x§ o2 oXt ox 3 ))
l
xi o (2 ox 2 ox 3 + 2 oX~+ x 3 o(1 + 4 oxi ox 2) + 3 ox§ o(1 + 2 oxt))
xi (x~ 0
+ x 1 + x2 + 2 ° xi x§)
0
Xt o (2 ox§ ox3 + 2 oX~ o(1

+ 2 ox 2 ) + 3 oXt ox3 + 2 X2 X3 + 8 X~ x§ X3) ° ° ° 0 0
T
= xi o (2 ox 2 ox 3 + 3 ox§ + 8 oX~ + x 3 + 4 oxi ox 2 ox 3)

[
o
xi x§ + Xt + x2 + 2 xi x§)
:h) , ~
0
( ° 0
Proof of A.1.58co Let äR '~ ( äf ith row vcctor of iiR. Thcn
This proves (Ao1o58c}o D
Relations used in constraint equations

The subsequent relations will be extensively used in the kinematic constraint
relations andin the relations of the constraint reaction forces (see Beetion 3)0
Given a vector p(t) = (p 1 (t), 0 0 0 ,Pn(t)) E C2 (lRn) and a vector function
g: ]Rn X JR1 --+ JRm g E C2 (JRn+l) (Aol.60a)
\.1)
w w ( Yt (p, t) )
(p(t) ' t) 1---t g(p(t)' t) = : '
Ym(p,t)
then
iJ(p(t), t) := ~t (g(p(t), t)) = Yp(p(t), t) p(t) 0

+ Yt(p(t), t),
and
Furthermore,
jj(p(t), t) := ~; 2 (g(p(t), t)) = ~t (gp(p(t), t) 0

jJ(t) + Yt(p(t), t))
= (gp(p(t)' t) jJ(t)) p jJ(t) + gp(p(t)' t) p(t)
0 0 0
+ Ypt(p(t), t) jJ(t) + Ytp(p(t), t) jJ(t) + Yu(p(t), t)

0 0
or (Aol.60c)
jj(p(t)' t) := gp(p(t)' t) p(t) 0
+ (gp(p(t)' t) jJ(t) )p jJ(t)
0 0
+2 ° Ytp(p(t), t) jJ(t) + Yu(p(t), t)o

0
Then the equation
jj(p(t)' t) = 0 (Aol.61a)
may be written as
gp(p(t), t) jj(t) = ßc(p(t), jJ(t), t),

o (Aol.61b)
with
ßc :=- (gp(p(t), t) op(t) + (gp(p(t), t) ojJ(t)) P ojJ(t)

+ 2 Ytp(p(t), t) jJ(t) + Yu(p(t), t))
° 0 (Aol.61c)
and with gP as the Jacobian matrix of go

290 Ao Appendix
A.2 Lagrange formalism of a rigid body under spatial

motion
The Lagrange formalism is an alternative approach for deriving the equations

of motion of rigid-body systemso It is equivalent to the Newton-Euter ap-
proacho Due to its "integral character", the application of Lagrange formalism
to examples sometimes appears tobe easier compared to the Newton-Euter
approach, which is characterized by its "differential character"o It may hap-
pen if a mechanism includes passive components of different energy species
(mechanical, electrical, magnetic, or thermodynamic; ([56], [65]))0
The Lagrange equations of a mechanism may be written in the form
=0
~ (f)L)T _(f)L)T =F ~ (oL/_ (8L)T =
dt 8v 8p 9 dtmJ 0
a>.. o, (Ao2ola)
with
T T )T ( T T)T
P = ( P1 ' ooo ' Pnb ' Pi = Ti ' '11i '
as the vector of the generalized coordinates of the mechanism,
v = (v T
1 ,
T )T
000 , vnb , Vi = T-1
i
('11i ) 0Pi,
o
as the vector of the velocities, with
T(p) = diag(T1, 000 , T nb) Ti= (

13
0
' 03 3
Hoo_ARL,
)
3,3 ' "
with F 9 as the vector of the generalized forces and torques associated with
p and v, with cf = TT (p) gp(P) >.. as the vector of the constraint reaction
0 0
forces, and with L as the Lagrange function, where usually
(Ao2olb)
with Tk(v, p) as the kinetic energy of the mechanism, U(p) as the potential
energy of the mechanism, 0 = g(p) E JRnc as the constraint equations of
holonomic constraints, and >.. E JRnc as the vector of the Lagrange multiplierso
Comment A.2.1 (Lagrange equations): In case that the generalized Co-

ordinates p are minimal coordinates, the term g T (p) >.. does not appear in 0
Lo The basic idea behind this approach is defining the generalized forces F 9
associated to the generalized coordinates and velocitieso
Ao2 Lagrange formalism of a rigid body under spatial motion 291
A.2.1 Kinetic energy of an unconstrained rigid body
Consider a "point mass" with the mass m located at the point Q of an inertial
frame R (Figure Ao2ola)o The kinetic energy of the point mass with respect
to R is
Tk = ~ ( R~t (r~ 0 )) T 0
( R~t (r~ 0 )) 0
m (Ao2o2)
= 1 (R r QO
2 R )To
0
(R R
r QO ) m E
o
0
lllll
.Ii\\. 0
Consider an unconstrained rigid body of mass m, with center of mass C and

volume V, with a reference point P fixed on the body, and a vector r PO from
the origin 0 of R to the point P (Figure Ao2olb)o
Consider a mass element of mass dm of the body at an arbitrary point Q
on the body, specified by the vector
T := TQOo (Ao2o3)
The kinetic energy of the body is
(Ao2.4)
where the velocity vector R,;.R := Rr~ 0 is measured and represented in Ro

Consider a second frame L with origin 0 L = P fixed on the bodyo Then the
vector r = TQO can be written as (Figure Ao2olb)
r R o-
o-
rR
QO-
_ rR
PO
+ vR _
A -
rR
PO
+ ARL 0 vL
A o
(Ao2o5a)
The velocity vector Rj-R is

RoR RoR +RoR
r = rpo X (Ao2o5b)
or
(Ao2o5c)
Due to the rigid-body property
LXL := 0 (Ao2o6)
and to the relation

Ao RL-
-
ARL -L
OWLR (Ao2o7a)
the resulting velocity vector is
(Ao2o7b)
292 A. Appendix
R point mass m
'point Q
(a) Point mass in space
mas. element dm
R rigid body of mas m
and volume
in rtial frame R
R
O=On
(b) Unconstrained rigid body in space
Fig. A.2.1: Vector diagram used in the definition of the kinetic energy of a rigid
body
Inserting (A.2.7b) into (A.2.4) yields the following expressions for the kinetic
energy of the rigid body, written with respect to the reference point P:
A.2 Lagrange formalism of a rigid body under spatial motion 293
TkP =~I (RrR) T. (RrR) dm (A.2.8)
=2 r PO + A RL · w-LLR · X L)T · (R r PO
11(R·R R + A RL · w L
LR · X L) dm
= ~ . ( Rr-~ 0 ) T . ( Rr-~ 0 ) . I dm
11((X L)T .
+2 (-L
W LR
)T ·-.._:,._...
ALR ARL · W-LLR · X L)d m
= 13
+ (RT~o) T 0 A RL wi;R
0 0 I XL dm.
Together with
T~p :=!I xLdm (A.2.9a)
as the vector from the origin P of L to the center of mass C of the body,
with
(A.2.9b)
I
and
dm=m (A.2.9c)
this yields
1 ·m·
Tkp= 2 I RTpo
R
0
11 2 +m· (R Tpo
R ) T ·A RL ·wLR·Tcp
0
-L L
+~I xLr ·(wiR) T · (wiR)· xL dm
or
Tkp=
1
2 ·m· I RTpo
R
0
11 2 +m· (R Tpo
R ) T ·wLR·A
- R
0
RL ·Top
L (A.2.10)
+~I llwiR · xLII 2 dm.

Using the vector relation
IIw · xll 2 = (w x x)T · (w x x) = (w x x)T · d = wT · (x x d)

..._"_,__..,
=:d
=wT· [xx(wxx)]=wT· [(xT·x)·w-(x·xT)·w]
=wT· [(xT·x)·I3-x·xT]·w
yields
294 A. Appendix
=(wiR) T. [(xL)T. XL. 13- XL. (xL)T] . wiR·

(A.2.11a)
Together with
~I llwiR ·xLII 2
dm
=21 (w LLR )T . { I [ (X L ) T ·X L · l3 -X L · (X L ) T] (A.2.11b)
and with the abbreviations (4.17), (4.23) and (4.24a)
J~ :=-I XL. XL dm = I [(xL)T. XL. 13- XL. (xL)T] dm,

(A.2.11c)
this yields, the following equivalent expressions for the kinetic energy of a
rigid body under spatial motion, written with respect to an arbitrary reference
point P:
Tkp = ~2 . m . (R . R ) T R . R
Tpo . Tpo (A.2.12a)
1 ( L )T
+2 W LR
L
.Jp
L
. W LR +m .
1 . m . (R T·R
= 2 PO
)T . R T·RPO
+~ (wfRf. ALR. ARL. J~. ALR. ARL. (wiR)
+m· ( R Tpo
·R )T - R
·WLR'A
RL L
·Top
1 .m .
=2 (R ·R
T PO
)T . R T·RPO
+ ~ (wfR f ·J ~ · ( R ·R )T
wfR) - m · ( Tpo ·A
RL - L L
·Top·WLR
1 . m . (R T·R
= 2 PO
)T . R T·RPO
+ 21 (W RLR )T . J Rp . (W R
R·R
LR ) - m . ( Tpo
)T. ARL . Top.
-L ALR . WLR
R
1 .m .
=2 (R ·R
T PO
)T. R T·RPO
+2 LR )T . J R
1 (W R +m R ) T ARL - L ALR R R
. (W R
LR ) . ( WLR ·Top· . Tpo·
0
p .
A.2.2 Spatial equations of motion of an unconstrained rigid body

for P = C
The kinetic energy of a rigid body with respect to the center of mass C is
obtained from (A.2.12a) for P = C and for Tpo = 0 as
1
Tkc=2·m·ll Tcoll 2+21 ( WLR
R. R R )T R
·Je·
( R )
WLR
or (A.2.12b)
Tkc
1
=2 · m ·II R.rcoll
R 2+ 21 ( wLR
L )T L
·Je·
( L )
wLR ·
Choosing
p:= (x~o,Y~o,z!Jo,c/J,B,~)T = ((r~o)T, 11T)T (A.2.12c)
as the vector of the generalized coordinates together with the associated

velocity vector
provides the following expressions of the Lagrange equations for P = C.

Equation (A.l.53) of Appendix A.1.6
(A.2.12d)
applied to (A.2.12b) yields
OTkc )T _ . R. R ( OTkc )
T
R R
( m rco
a ( ·R )
rco
- owfR =Je. WLR'
Rd ( {)Tkc )T = m . R:;.R (A.2.12e)

dt a (r~ 0 ) 00 '
and
Rd (
dt
oTkc
owfR
)T = Rd ( R R ) Rd ( RL
dt J c w LR =: dt A
0
L
Jc 0 0 A
LR
0
R )
w LR
= A
' RL
0 Jc
L
0
L
w LR + A RL 0 Jc
L
0
' LR
A 0
R
w LR + A RL 0
L
Jc 0 A
LR
0
R · R
w LR
= A
RL ( - L JL - L JL - L ALR
'WLR' c·WLR- c·WLR·~+ c ·
R JL ALR 'WLR
R R )
·
0
L
=WLR
=0
Together with
-L
WLR ·WLR
L
= O
and
296 A. Appendix
R · R Rd ( R ) Rd ( RL L ) ' LR L RL L · L
WLR=dt WLR = dt A ·WLR =A ·WLR+A . WLR
_ ARL w- L wL +ARL Lw· L _ ARL Lw· L

- ' LR ' LR ' LR - ' LR'
...._".__....
=0
this yields the relation
Rd (
dt
ÖTke ) T _ ARL . (JL
öwfR
-
. L
e ' W LR + W- LLR ' JLe R )
' W LR · (A.2.12f)
Together with
öLe öU öLe
and (A.2.12g)
ör~ 0 ÖTJ
this provides the equations of motion of an uneonstrained rigid body for P =

C (cf. Equation 4.56a of Beetion 4.2.4.1):
( m · l3 , 03,3 ) . ( ~~o ) = (A.2.13a)

03,3 ' J~ WLR
( 0 )
-L L L
-WLR ' J e . WLR
A.2.3 Spatial equations of motion of a constrained rigid body
The equations of motion of a rigid body under eonstrained spatial motion for
P = C in DAE form are (cf. Equation 4.77i of Beetion 4.3.2.3)
(ögjör~ 0 )T
J~ ALR(TJ) 0 0 I
HT (TJ) (ög ÖTJ)T
-(öu /ör~ 0 f
-(öu jöTJ) T
(A.2.13b)
They include the eonstraint aeeeleration equations of the holonomic con-

straints
(A.2.13c)
and the constraint reaction forces and torques
(A.2.13d)
298 Ao Appendix
A.3 Model equations of planar and spatial mechanisms
In this section relevant differences between vector notations and equations

of the motion of planar and spatial mechanisms will be briefly summarized:
10 The dimension of geometrical displacement, velocity, acceleration, and
force vectors is increased from two (in the planar case) to three (in the
spatial case) 0
20 Angular velocity and torque vectors of planar mechanisms have a fixed

direction perpendicular to the plane (OC2 ) Angular velocity and torque 0
vectors in spatial mechanisms may have arbitrary directions in OC3 0
30 Contrary to OC2 (considered as a subspace of OC3 ), OC3 is closed und er

"vector product operations "; ioeo,
for (Ao3o1)
for
40 All matrices (including orientation matrices) are extended from elements

of OC2 , 2 in the planar case to elements of OC3 , 3 in the spatial caseo
50 The transformation matrix
(Ao3o2)
of the planar case is extended to a product of three elementary transfor-

mation matrices,
(Ao3o3a)
with the rotation angles r.p, (), 'lj; chosentobe around suitable axes in the
spatial caseo In the special case of Bryant angles, this yields
T
cBoc'ljJ ,cr.pos'ljJ+sr.posBoc'ljJ,sr.pos'ljJ-cr.posBoc'ljJ)
A RL = ( -c B s 'ljJ, c r.p c 'ljJ - s r.p s B s 'ljJ, s r.p c 'ljJ
0 0 0 0 0 + c r.p 0 s B s 'ljJ
0 0
sB , -s r.p c B
0
, c r.p c B
0
(Ao3o3b)
60 A basic difference between planar rotations and spatial rotations is that:
in the planar case, WzLR is the time derivative of a rotationangle '1/JLR
around the z-axis perpendicular to the x-y plane
L T
= '1/JLR = '1/JLR = (0, 0, 1) (Ao3o4)
0 o o
WzLR 0
€zL 0
€zR Or WzLR 0
'1/JLR,
whereas in the spatial case, wfR is related to the time derivatives of the
angles r.p, (), 'lj; by the (nonlinear) kinematic differential equation
Ao3 Model equations of planar and spatial mechanisms 299
with H (r.p , () , 'ljJ) depending on the special choice of A RL 0 For Bryant

angles this yields
1 , 0 , sB )
H- 1 (r.p, (), 'lj;) = ( 0, cr.p, -sr.poc() (Ao3o6)
0, sr.p, cr.poc()
with sa := sina and ca := cosao
70 The kinematic differential equation (Equation Ao3o5) of the spatial case

must be solved simultaneously with the kinetic differential equations of
the mechanismo
80 The vector of Cartesian coordinates of a body under planar motion
R R
P = ( Xpo, YPo, 'l/JLR
)T E lR3 (Ao3o7a)
is extended to the vector
P = ( x~o , Y~o , zj;o , 'P , () , 'l/J f E lR6 (Ao3o7b)
in the case of spatial motionso
90 The velocity equation
P =
o ( oR
XPO ' Ypo '
oR .i,)T =: V
'f' (Ao3o8a)
of the planar case is extended to the velocity equation
jJ = T(p) v o (Ao3o8b)
of the spatial case, with
(Ao3o8c)
(Ao3o8d)
and
- ( 13 ' 03,3 )
T( p
) - RL (Ao3o8e)
H (r.p , () , 'ljJ) A (r.p , () , 'ljJ)
0
03,3 , 0
100 The time derivative of an orientation matrix

A.RL = ARL wfR E JR3 0
(Ao3o9a)
300 A. Appendix
reduces, in the planar case, to
(A.3.9b)
11. The Steiner-Huygens formula is extended from
(A.3.10a)
in the planar case (moment of inertia) to
J~ = [J~ + m. [(r~pf. r~p. 13 - r~p. (r~pf]] E m.3 •3

(A.3.10b)
in the spatial case (inertia matrix}, where J~ includes three moments of

inertia, J~;;, and three products of inertia, Je;; , i =j:. j.
12. The vector of centrifugal forces and gyroscopic terms of the spatial case
_ ( -m· A RL ·WLR'WLR'Tcp
-L -L L )
Qa(P, v)- -L JL L (A.3.11a)
-WLR. P ·WLR
reduces, in the planar case, to
m A RL . R 2 . rLCP' ~2LR
Qa(p,v)= ( - .
)
(A.3.11b)
0
( +m. ARL ~ r~p. ~LR)
(no gyroscopic terms}.
13. Holonomic kinematic constraint position, velocity and acceleration equa-

tions of a planar mechanism have the form
g(p) = 0 ( constraint position equation), (A.3.12a)

Yp(P) ·V = 0 ( constraint velocity equation), (A.3.12b)
'-v-"
constraint
Jacobian
and
Yp(P) ·V= - (gp(P) ·V\· V=: ßc(P, P) (A.3.12c)
( constraint acceleration equation).
For spatial mechanisms these equations have the form

A.3 Model equations of planar and spatial mechanisms 301
g(p) = 0 ( constraint position equation), (A.3.13a)

Yp(P) · T(p) ·V = 0 ( constraint velocity equation), (A.3.13b)
-----..........
constraint
Jacobian
and
gp(P) · T(p) · v =- (gp(P) · T(p) · v) P • T(p) · v =: ßc(P, v) (A.3.13c)
( constraint acceleration equation).
14. In agreement with the above constraint equations, the constraint reaction
forces and torques are computed in the planar case by the relation
(A.3.14a)
from the Lagrange multipliers .A, and in the spatial case by the relation
Cf= (gp(P) T(p)) T A = TT(p) g~(p) A E JR6

0 0 0 0
0 (A.3.14b)
15. The DAEs of a planar mechanism are
jJ = v (A.3.15a)
( M(p) ,g~(p)) . ( v) = ( f(p, v) ) + ( Qa(P, v)) .

Yp(P), 0 .A ßc(P, v) 0
For spatial mechanisms, they have the form
jJ = T(p) V (A.3.15b)
(v) (q 0 (p,v))·
0
( M(p) ,TT(p)·gJ(p)). = ( f(p,v)) +

gp(p)·T(p), 0 A ßc(p,v) 0
16. The number of different joint types is much higher in the spatial case than
in the planar case. In the spatial case there exist many more combinations
of projections and representations of vector loop equations and Orienta-
tion loop equations. The reader should be aware that formal constraint
relations with identical names may theoretically model quite different
planar and spatial joints. For instance, the common-point relation mod-
els a revolute joint in the planar case and a spherical joint in the spatial
case.
17. Planar mechanisms have

np = 3nb- nc (A.3.16a)
DOFs whereas spatial mechanisms have
(A.3.16b)
DOFs, where nb is the number of rigid bodies, and nc is the number of
(independent) constraint equations of the mechanism.
302 A. Appendix
A.4 Constraint equations of a general universal joint
In this appendix model equations of a general universal Joint with nonin-

tersecting and nonorthogonal rotation axes will be derived. This joint model
provides the approximate constraint equations of mechanisms that include a
chain of three rigid bodies that are coupled by two revolute joints according
to the drawing of Figure A.4.1. A precise theoretical model of this 8 DOF
mechanism is a system of 18 kinematic DEs, 18 kinetic DEs, and 10 constraint
equations.
L;, L,.. arc fixcd on body i

( _ { ) 'PI = 'PL,L, L j, L>. are fixed on body j
L 1, L 2 are fLxed on the massless link
r;
rotation axis 2
0 0
Fig. A.4.1: Drawing of a general universal joint (massless revolute- revolute link)
connecting two rigid bodies
In such mechanisms of machirres and vehicles the central body connecting the
two other bodies has negligible inertia compared to the adjacent two bodies.
This rigid body can therefore be treated as a massless revolute- revolute link
A.4 Constraint equations of a general universal joint 303
(Figure A.4.1). This 8 DOF rnechanisrn will be approximately modeled by 12

kinematic DEs, 12 kinetic DEs, and 4 constraint equations.
The Jour constraint equations together with the two relative rotation angles
of this rnechanisrn will now be derived for the general case of rotation axes of
the joint (rnassless revolute-revolute link) that neither intersect each other
nor are orthogonal to each other.
A.4.1 Notation and abbreviations
Two rigid bodies i and j have reference points Oi (origin of the local frarne
Li on body i) and Üj (origin ofthe local frarne Lj on body j), (Figures A.4.1
and A.4.2). Let R be an inertial frarne with origin 0. Let
(A.4.1a)
be the shortest vector between the two rotation axes, with P located on
rotation axis 1 and Q located on rotation axis 2 (this distance will be corn-
puted in Section A.4.3). Consider two frarnes L"' and L 1 with origins in P,
and two frarnes L>-. and L 2 with origins in Q. (To obtain a clearer picture
of this situation, the two frarnes L"' and L 1 are drawn with different ori-
gins Pi and P, where the distance s 1 := rppi is set to zero (Pi = P). The
sarne representation is chosen for (L>-., L 2 ) with origins (Q and Qj), and
s 3 := TQiQ = 0 (Q = Qj).) Assurne that the exL 1 -axis of L 1 is oriented in
the direction of rotation axis 1, and that its ezL 1 -axis is oriented in the direc-
tion of TQP (then L 1 is fixed to the rnassless link). Let L"' be fixed to body i
with its exL" -axis oriented parallel to exL 1 • The rotation angle of L"' relative
to L 1 around their cornrnon x-axes is called r.p 1 := 'PL"L 1 • Assurne that the
ezL 2 -axis of L 2 is oriented in the direction of TPQ (i.e., ezL 2 = -ezLJ, and
that its eyL 2 -axis is oriented in the direction of rotation axis 2 (then L 2 is
fixed on the rnassless link). Let L>-. be fixed on body j with its eyL>. -axis ori-
ented parallel to eyL 2 • The rotation angle of L >-. relative to L 2 around their
cornrnon y-axes is called fh := fh>.L 2 • Let
Ti:= roio and Tj := rojo (A.4.1b)
be the displacernent vector of Oe frorn 0 (C = i, j), and let

· L L L·
S ii · - Li · - rLi
.- 8 i . - PiOi
·.-- TLp,0
· .·
~ and s 11. := s1. j := r QiOi
' ·=
·
r QOi
'
(A.4.1c)
be vectors fixed on bodies i and j, respectively. The vector s 2 := TQP is

represented as s~ := s~ 1 = s~Q = (0, 0, Sz 2 )T. The rotation angles r.p 1 and
ß2 are time dependent, whereas the relative rotation angle 'l/J21 := 'l/JL 2 L 1 of
L 2 with respect to L 1 around the cornrnon z-axes is a constant angle. The
different orientation rnatrices introduced in Figure A.4.2 are
304 A. Appendix
vector loop
0;
orientation loop
0
Fig. A.4.2: Diagrams for deriving the vector loop and orientation loop equations of
the mechanism
AiO := ALiR (variable),

AjO := ALjR (variable),
Aij := AiO 0 AOj (variable),
(constant),
A>.j := AL>.L; (constant),
A21 := AL2Ll('lj;21) (constant),
All<(cpl) := ALlL~(cpl) (variable), and
A2>.(02) := AL 2 L>, (02) (variable). (A.4.ld)
The following abbreviations will be used for angular velocities,
and (A.4.1e)
and projection operators,
P?(x)=(l,O,O), P?(y)=(O,l,O), and P?(z)=(O,O,l).

(A.4.1f)
A.4.2 Computation of constraint equations
The constraint equations
(A.4.2a)
of this mechanism include Jour scalar constraint equations
9i = 0 (i=1,2,3,4), (A.4.2b)
that will now be derived together with the two relative coordinates cp 1 and 02
from suitable representations and projections of vector loop and orientation
loop equations, obtained from the geometrical situations discussed above and
shown in Figures A.4.1 and A.4.2.
A.4.2.1 First constraint equation. The first constraint equation that
eliminates a rotational DOF between the two bodies i and j is obtained
by the orientation loop equation (Figure A.4.2)
(A.4.3a)
or
(A.4.3b)
Together with
(A.4.3c)
this yields
ih = A>.i. Aii. Ai,.. A,.l(cpl)- A>.2(02). A21('1j;21) = 0. (A.4.4a)

306 Ao Appendix
Its first and second time derivatives are, respectively,

t11 = A>-i 0 Aii 0 wL 0 Ait< A~< 1 (cpl)
0
+ A>.i 0 Aii Ai~< Ä~<l('Pl)- A>.2(82)

0 0 0 A21 = 0, (A.4.4b)
and
!h = A >.i 0
Aii 0
(-i -i
wij wij 0 0
:.i)
wij 0
Ai~< A~<l 0
>.0 t<l
+2 A
°
J AJ' A (cpl)
00 0 0
0 0 0 w~i 0 A'~< 0
+ A>.i 0 Aii Ai~< 0 0 Ä~< 1 (cp 1 )- Ä>. 2 (8 2 ) A 210 = Oo (A.4.4c)

Using the abbreviations c cp := cos cp and s cp := sin cp yields, together with
cp 1 :='PL~L 1 as the rotation angle around exL 1 = exL~, the relation (cfo
Equation 2055)
A~<l =
1,
( 0 '
0'0)
c 'Pl ' s 'Pl
0 ' - s 'Pl ' c 'Pl
t<l
0' 0 ' 0 )
A = ( 0 ' - s 'Pl c 'Pl cpl' and (A.4o5a)
0
0 ' - c 'Pl ' - s 'Pl
0' 0 ' 0 ) 0' 0 ' 0 )

( 0 ' - s 'Pl ' c 'Pl 0
tPl + ( 0 ' - c 'Pl ' - s 'Pl 0
o2
'Pl'
0 ' - c 'Pl ' - s 'Pl 0 ' s 'Pl ' - c 'Pl
and, with 82 = 8L).L 2 as the rotation angle around eyL 2 = eyL)., the relation
- s 82 , o, - c 82 )
>.2
A = ( ih, (A.4o5b)
0
0 '0' 0 0
and
c82 ' 0 ' - s82
- c 82 , o, s 82 )
2 ( - s 82 , o , - c 82 )
+ o ,o, o
0
Ä,\2 = ( 82
0 '0' 0 0
- s 82 , o , - c 82 c82 ' 0 ' - s 82

Using 'l/J21 := 'l/JL 2 L 1 as the constant rotation angle of L2 with respect to L1
about the common ezL 1 = ezL 2 axis yields the relations
A 21 = ( _ : ~:: ~:: ~)
: : : and
• 21
A = 0. (A.4.5c)
0 ' 0 ' 1
'0)
,0 (A.4.6a)
' 1
and
A A~< · A ~< 1 (cpl) = (

an , a12 , a13 )
a21 , a22 , a23 ( 1, 0' 0)
0 ' c 'P1 ' s 'P1 '
a31 , a32 , a33 0 ' - s 'P1 ' c 'P1
(A.4.6b)
respectively. Multiplication of the left-hand side of (A.4.6a) and (A.4.6b) by

P?(y) and right-hand side by Pr(x) yields the projections
(A.4. 7a)
and
(A.4.7b)
that are independent of cp 1. This eliminates the yet unknown relative Coor-
dinates cp 1 and 02 from (A.4.4a) and provides the first constraint orientation
equation
0 := 91 := P?(y) AA~< · A~< 1 (cpi) Pr(x)- P? AA 2 (82) · A 21 (1f]21) Pr(x)
or (A.4.8a)
91 :=P?(y)AA~<Pr(x)+sin1f]21 :=0
or (A.4.8b)
91=P?(y) (AAi.Aii.Ai~<) Pr(x)+sin1f]21=0 , 1f]21=constant,
and the associated constraint velocity and acceleration equations
(A.4.9a)
and
ih=P?(y) [AAi.Aii. (w~i·w~i+~~i)·Ai~<] Pr(x)=O. (A.4.9b)
Together with
308 Ao Appendix
A ij = AiO AOj 0
i -
W ij -
i
W iO -
i -
W jO -
i
W iO -
Aij 0 W jO'
j (Ao4o10a)
i _
0
i
Wji-WiO- Aii oWjioWjO-
-i 0
i Aii oWjO
i 0
-
_
0
i
W iO -
Aii 0
( W- ijO - W- iiO ) 0 W jO
i - Aii 0 W jO
0
i
-
_
0
i
W iO + Aii 0
-i
W iO
i - Aii0 W jO 0
-i
W jO 0 W jO
i - Aii 0 W jO
0
i
-
_
W iO + W iO
0
i - i
0
Aii W ijO - Aii
0 0 W jO'
0
i (A.4ol0b)
(Ao4o10c)
and
w=w, (A.4ol0d)
the constraint acceleration equation is
0 = fh = P?(y) (A~i Aii w!i w!i Ai") Pr(x)
l
0 0 0 0
+ P? (y) [ AA.i Aii ( 0

~~0 - Aii w; 0
0 + w!0 ° Aii 0
w; 0) 0
Ai" Pr(x)
or
ih = P?(y) ( AA.i Aii 0 0

~~ 0 ° Ai") Pr(x)
- P?(y) ( AA.i Aii Aii 0 0 0 w; 0 ° Ai") Pr(x)
+ P?(y) ( AA.i 0
Aii w!i w!i Ai") Pr(x)
0 0 0
Taking into account the relations
:. i
wiO Ai" p r (X )
o =- (Ai" p r (X )) wiO
i o
0
and (A.4o10e)
( Aii 0
w; 0) 0
Ai" Pr(x) =- (Ai" Pr(x)) 0
Aii 0
w; 0
yields the final form of the constraint acceleration equation

(A.4.11)
= -P?(y) { A-\j · Aji · [w:j ·w:j + (w: 0 · Aij · w~ 0 )]} Pr(x),

=: ßcl
where
(A.4.12)
i _ i Aij j
wij - wiO- . w]o'
. i _ . i Aij · j Aij - j j
W ij - W iO - • W jO - . W ji . W jO'
Comment A.4.1 (Computation of cp 1 and fh): The relative coordinates

cp 1 ( t) and (} 2 ( t) may be isolated by suitable projections of the preceding ori-
entation loop equations. Here these relative rotation angles will be computed
later by means of proper projections of suitable vector loop equations of the
mechanism.
A.4.2.2 Second constraint equation. The second constraint equation

that eliminates a translational DOF of the mechanism is computed from
a suitable projection of the vector loop equation (Figure A.4.2)
(A.4.13)
Projecting this equation into frame L 1 yields, together with
S~ = (0, 0, Sz2), Sz2 > 0, (A.4.14a)
the relation
(A.4.14b)
or
(A.4.14c)
Together with
310 Ao Appendix
C~J
1' 0
A"l s~ = ( ' 0 )
0
0 ' c 'Pl ' s 'Pl (
Sz2 o :incp1) ,
0 ' - s 'Pl ' c 'Pl Sz2 ° COS 'Pl
(A.4ol4d)
this yields the relation
(A.4ol5a)
with
[h : = + A 0 i os~- A 0 i osn
A" 0 o [ (r?- r~)
= A"i Aio (r?- r~) + A"i s~- A"i Aio Ai 0 s;o
0 0 0 0 0 0
(A.4ol5b)
These three scalar equations only include the relative coordinate cp 1 They will 0
be used to derive the second constraint equation and to isolate the relative
coordinate cp 1 o The projection
g 2 + P?(x) [A"i o Ai0 o (r?- r~) + A"i os~- A"i o Ai0 oAi0 osn := 0
(Ao4ol6a)
of (A.4ol6a) is chosen as the second constraint position equationo It no Ionger
includes the relative coordinate cp 1 (t)o The associated constraint velocity equa-
tion is
!J2 = P?(x) { A"i [.Aio (r?- 0 0

r~) + Aio 0
(r?- r~)
- AiO A Dj s; - A io Ä Oj s;J }
0 0 0 = 0
or
g2 (t) = P?(x) { A"i 0

[Ai0 0
w8i 0
(r?- r~) + Ai (r?- r~) 0 0
-Ai0 o (w8i oA 0i os; + A 0 i ow; 0 os;)]} = 0,
and finally
!J2(t) = P?(x) { A"io [-w!0 oAio o (r?- r~) + Aio o (r?- r~)
-Wio
•
AiO AOj 0 0 0 s~- AiO
J
0 AOj wjJÜ
0 0 sj] } =
J
0
0
(A.4ol6b)
+ W- iiO . Aio . w- oOi . Aoj . sjj + w- iiO . Aio . Aoj . w- jO

j . sjj
- AiO. w8i. A 0 j. W~o. s~- AiO. A 0 j. W~o. W~o. s~
- A iO . A Oj . ~~0 . sn =0
or
ih= P?(x)A~<i.{ Ai 0 ·r?+Ai 0 ·[(r?-r~)-A 0 j·s~]·w~ 0
- A iO . f~ + ( A iO . A Oj . sn .W~o }
+ PrT (x) A~<i . { - i - i Aio · ( ri-rj-

wiO-wiO· o o Aoj ·sjj)
- 2. Wio.
• AiO. (r •0 - r J0 ) - AiO. A 0 j. w~JO . wjJO . sjJ
+ 2. Wio.
•
AiO. A 0 j. wj
JO
. sj} = 0
J '
and finally
[P?(x) (A~<i · Ai0 ), P?(x) { A~<i · [Aio · (r?- r~)- A 0 j · s;]},

-P? (x) ( At<i. Ai0 ) ' P? (x) ( At<i. AiO. A 0 j . s;)]
=: Y2p(p) · T(p)
··0
Ti
. i
wiO
··0
-P?(x) { A~<i · [w~ 0 · w~ 0 · Ai0 · (r?- r~- A 0 j · s;)
rj
-AiO. A 0 j. w~ . wj . sj + 2. Wio. AiO. A 0 j. w~ . s~
. j
wjo
JO JO J • JO J
=: ßc2
(A.4.17)
312 Ao Appendix
Isolation of the relative coordinate r.p 1 (t)
The relative coordinate c.p 1(t) is isolated by projecting the vector loop equa-
tion (A.4o15a) to its second and third componento This yields the relations
1
sinc.p1 = - - 0P?(y) (!h) for Bz2-::/:- 0 (Ao4o18a)
Sz2
and
1 T
cosc.p1 = - - 0Pr (z) (ih) for Sz2-::/:- 0, (A.4o18b)
Sz2
with
!h = A"i oAi0 o [(r?- r~) + A 0 i os~- A 0 i os;] 0 (Ao4o18c)
This yields the relative rotation angle
1n =tn =arctan[P?(y)(!h)] for -7r<_,n 1 <_7ro (A.4o19)

rl r~<l P?(z)(i'h) r
The time derivatives of (A.4o18a) and (A.4o18b),
1
'Pl 0 cosc.p1 =--Pr
o T( Y) (92) (Ao4o20a)
Sz2
and
o o
- p r T( z ) (-'-Y2 ) ,
'Pl o smc.p1 = 1 (A.4o20b)
Sz2
finally yield the relations
- -1- - 0P T (y) (g
o
1
)
for 0 < I sin 'Pli < c:
Sz2 ° COS 'Pl r
(A.4021)
10 ° P?(z) (9!) for 0 <I cosc.p1l < C:o
Sz2 osm 'Pl
The time derivatives of (Ao4o20a) and (A.4o20b),
o2 o sm
'Pl o cos 'Pl - 'P1
00 o 'Pl = - - r Y ) ("
1 0pT( Y2 )
Sz2
and
o 'Pl
'Pl o sm
oo
+ c.po21 o cos 'Pl 1 opT()(")
= -- r z Y2 ,
Sz2
yield the relation
(
c~s 'Pl
(Ao4o22a)
sm 'Pl
and with
(A.4.22b)
its solution
and the final result
(A.4.23)
A.4.2.3 Third constraint equation. The third constraint equation that

eliminates another translational DOF of the mechanism is computed from
another representation and projection of the vector loop equation (A.4.13)
(A.4.13)
Representation of this equation in L 2 yields
(A.4.24a)
with the constant vector
(A.4.24b)
the rotation matrix
(A.4.24c)
.
and the relation
+ Sz2 • ~infh)
(
- Sz2 • cosfh
(A.4.24d)
This provides together with
(A.4.25)
the relation
314 Ao Appendix
( ~O) ( +- Sz2 o ~in(h) (A.4o26a)

Sz2 ° COS ß2
with
!h: = A'>.o o [(r?- r~) + A 0 i os1- A 0 J o sn (A.4o26b)

= A>.j 0 AJ 0 0 (r~-
•
r J0 ) - A>.j sjJ 0 + A>.j 0 AJ 0 A 0 i s~•
0 0
These three scalar equations only include the relative coordinate B2 (t)o They
will be used to derive the third constraint equation and to isolate the relative
coordinate B2 (t)o The projection
g 3 := P?(y) [A>.j AJ 0 0 0 (r?- r~)- A>.j 0 s~ + A>.j 0 AJ 0 A 0 i s1] = 0

0 0
(Ao4o27a)
of (Ao4o26a) is chosen as the third constraint equation of the mechanismo It

no langer includes the relative coordinate 82 (t)o The associated constraint
velocity equation is
g3 (t) = P?(y) { A>.j o [Aj 0 ow8j o (r?- r~) + AJ0 o (r?- r~)
+ AJ 0 w8j A Oi s1 + AJ 0 A Oi w1o s1 -
0 0 0 0 0 0
0] }= 0
or
g3 = P?(y) { A>.j o [- w~ 0 oAJ0 o (r?- r~) + AJ0 o (r?- r~)

(A.4o27b)
- r y ) A>.j
.. -PT(
93 0 [
- W:_jjO Ajo 0 0 j + W-jjO
( Tio - To) 0
-jjO
W 0
Ajo 0
(r?o- rJoo)
o
_ w- j 0 0
AjO AOi w- i 0
0 0 0 0 8i 0 + AjO 0
AOi w- i 0 w- i 0
0 0 0 0 0 8i 0
JO ' ' ' ' '
-w- Jj 0 °A1oo OA Oi w- •i 0 °s•i +A1oo OA Oi w:_ •i 0 °s•i] =0

0 0 0 0 0 0 0
or
ih = P?(y) { A>..j · Ajo, -A>-.j. Ajo. A 0 i. s~, -A>-.j. Ai 0 ,
A>-.i. [Aio.Aoi·s1+Aio. (rf-rf)]}·

[(r:?) T, (w~o) T, (r:J) T, (w~o) Tf
+P?(y) { A>-.i.w~ 0 ·w~ 0 ·Ai 0 · (r?-r~)
-2·A>-.i.wi ·Aio.
JÜ
(ro-ro)
" J
+A>-.j . wio. w~o. Ajo. Aoi. si

J J "
-2 . A>-.j . w-ijO . Ai 0 . A 0 i . w- iiO . sii
+A )..j . AjO . AOi -i -i

. wiO . wiO . sii } = 0,
and finally
[ P? (y) A>-.j . Ajo, -P? (y) A>-.j. Ajo. A 0 i. s~, -P? (y) A>-.j. Ajo,
P?(y)A>-.i. {Ai 0 . [(r?-r~)A 0 i·s1]}]
..,
=: Y3p · T(p)
[( r··O)T
· i
(. i )T , (••O)T
, w iO ri , (.w jjO )T]T
= -P?(y) { A>-.i · w~ 0 · w~ 0 · Ai 0 · [(r?- r~) + A 0 i · s1]

+A )..j . AjO . AOi . wiO
-i - i
. wiO . sii
-2·A>-.i.wi ·Ai0 . [(r 0 -rq)+A 0 i·w~ 0 -si]}
JÜ " J " " •
=: ßc3
(A.4.27c)
Isolation of the relative coordinate 0 2 (t)
The relative coordinate fh(t) is isolated by projecting the vector loop equation
(A.4.26a) to its first and third component. This yields the relations
sinfh = _J_ · P?(x) (i'h) (A.4.28a)

Sz2
and
1 T -
cosB2 =-·Pr (z) (g 3 ). (A.4.28b)
Sz2
This yields the relative rotation angle

316 Ao Appendix
() = () = -arctan { P? (y) Uis)} for - 1r < B2 < 7fo (A.4o29)

2 >-.2 P?(z)(?h)
The time derivatives of (A.4o28a) and (A.4o28b),
1 T _,_
B2 ° cosB2 = - - Pr (x) (93)
o
0 (Ao4o30a)
Sz2
and
(A.4o30b)
finally yield the relations
( -1 T
Pr (x) (ih) for 0 < sinB2 <
o
1 () 0
E:
1 Sz2 ° COS 2
= )
0
()2 (A.4o31)
l -1 T
(ih) < cosB2 < E:o
o
l
1 0 ()
0
Pr (z) for 0
Sz2 ° Sill 2
The time derivatives of (Ao4o30a) and (A.4o30b) yield
and, together, the relation
cos B2 , - sin B2 )
( 0
sin B2 , cos B2
with the solution
and finally
Ö2 = - 1 0
[cosB 2 ° P?(x) (g 3 ) + sinB2 P?(z) (g 2)]
° 0 (A.4o32)
Sz2
A.4.2.4 Fourth constraint equation. The fourth constraint equation is

again computed from the vector loop equation (A.4o13)
or, represented in frame R,

(A.4.33)
Taking into account the constant distance of the mechanism between the
points P and Q, and introducing this distance as
(A.4.34)
the following constant-distance constraint relation is introduced as the fourth

constraint position equation of the mechanism (compare with the vector loop
of Figure A.4.1 and BB5 of Chapter 5.2):
0 T 0 ) 1/ 2 -
94 := ( (rqp) · rqp -Co= 0
Its time derivative is
(A.4.36a)
or
(A.4.36b)
Assuming only small deviations of lrqpl from C0 , the substitution
(A.4.36c)
is introduced into (A.4.36b), which provides the following approximations of

the constraint velocity and acceleration equations:
(A.4.37)
and
(A.4.38)
This implies the constraint position equation
(A.4.39)
318 Ao Appendix
the associated constraint velocity equation
il4 := (ij4) ° Co= 0 + [r~- r? + A 0j s;- A 0i s~r 0 0
0 [r J0 -r 0• +A 0iow~JÜ osi-Aoiowi0 osi]

J • •
=0
or
94 = 0 + [-(r~p)T' (r~p)T 0
A 0i 0
s1' (r~p)T' -(r~p)T A 0i 0 0
s;]
=: Y4p(P) T(p)0
0
[( roO)T
i
( i )T ' (roO)T
' w iO j
( j )T] T -_ O
' w jO (A.4o40)
and, together with
(A.4o4l)
the associated constraint acceleration equation
94 =
00
[ ( 0 )T ' (rQP
- rQP 0 )T 0
AOi
0
-i (rQP
si'
0 )T ' - (rQP
0 )T 0
AOj
0
-j]
sj
=: Y4p(P) T(p) 0
.. O)T , (wo,i0 )T , ("O)T

[( r, r J , (woJD
j )T] T
0 )T
= ( rQP 0
(AOi oWiOoWiOoSi-
-i -i i AOj oWjooWjooSj
-j -j j) - (oO
rQP )T 0
(oO
rQP ) o
' ~ ~
=: ßc4
(A.4.42)
Comment A.4.2 (Complete set of model equations of the above

joint): The complete set of model equations of the above general universal
joint is built from (Ao4o8b), (Ao4o9a), (A.4oll), (Ao4ol6a), (A.4ol6b), (Ao4ol7),
(A.4o27a), (A.4o27b), (A.4o27c), (A.4o39), (A.4.40), and (A.4.42)o The com-
plete constraint Jacobian matrix is (cfo Equations A.4oll, A.4ol7, A.4o27c,
and A.4.42)
Yp(P) T(p) = (gJp(P), Yip(P), gjp(P), gJp(P)) T T(p),

0
o (Ao4.43a)
with
(A.4.43b)
and
(A.4.43c)
The complete right-hand side of the total constraint acceleration equation is

(according to Equations A.4.11, A.4.17, A.4.27c, and A.4.42)
(A.4.43d)
A.4.3 Computation of the shortest distance between two rotation

axes
Consider the following description of the two rotation axes in Figure A.4.1:
y1= T · a1 + b1 (rotation axis 1) (A.4.44a)
and
y 2 = u · a2 + b2 (rotation axis 2), (A.4.44b)
with a 1 , b1 and a 2 , b2 as known constant vectors and with T and u as free

scalar parameters (cf. Figure A.4.3). Let lrQPI = ls2l = Sz2 (cf. Equation
A.4.24b) be the shortest distance between the rotation axes 1 and 2 with
TQP as the vector from point P on axis 1 to point Q on axis 2. Then
r PO = Ta · a1 + b1 (A.4.45a)
and
(A.4.45b)
with the parameter values
Ta=
and
[ (b1 - b2) T · a2 J · [a[ · a1] - [ (b1 - b2) T · a1 J · [ai · a1]
[ai · a2] · [a[ · a1] - [ai · a1] 2
(A.4.46b)
This provides the shortest distance between axes 1 and 2 as
or
(A.4.47)
The parameter values Ta and ua are computed as the solution of the linear
equation
320 A. Appendix
axis2
Fig. A.4.3: Shortest distance between two rotation axes
(A.4.48)
which follows from the geometrical condition that vectors a 1 and a 2 are both
orthogonal to vector r QP = (r QO - r PO).
References
[1] R. R. Ryan, ADAMS - Multibody System Analysis Software, pp. 361-

402, Multibody Systems Handbook, Ed. W. Schiehlen, Springer, Berlin
Heidelberg New York, 1990.
[2] N. Orlandea, M. A. Chace, D. A. Calahan, A Sparsity-Oriented Ap-
proach to the Dynamic Analysis and Design of Mechanical Systems,
ASME Journal of Engineering for Industry, Vol. 99, pp. 773-779, 1977.
[3] R. C. Smith, E. J. Haug, DADS- Dynamic Analysis and Design System,
pp. 161-179, Multibody Systems Handbook, Ed. W. Schiehlen, Springer,
Berlin Heidelberg New York, 1990.
[4] H. Hahn, R. Wehage, Theoretical and Computational Aspects of Program
NUSTAR, IABG Bericht B-TF 2070, 1986.
[5] H. Hahn, R. Wehage, Dynamic Simulation of Terrain Vehicles, Proceed-
ings of the ninth International Conference on Terrain Vehicles, Septemer
1987, Barcelona, pp. 491-503, 1987.
[6] W. Schiehlen, Multibody Systems Handbook, Springer, Berlin Heidelberg
New York, 1990.
[7] W. Schiehlen, Advanced Multibody System Dynamics, Simulation, and
Software Tools, Kluwer Academic, Boston, 1993.
[8] W. Klier, A. Dürrbaum, Computer Aided Mapping of Symbolic Model
Equations of Spatial Multi-Body Systems from DAE-Form toDE-Form,
RTS Bericht 37, University of Kassel, 2000.
[9] X. Liu, Input-Output Decoupling for Nonlinear Singular Control Sys-
tems, Preprints of Papers of the 12th IFAC World Congress, July 18-23,
1993, Sydney, Australia, Vol. 6, pp. 395-398, 1993.
[10] X. Liu, S. Celikowsky, Feedback control of affine nonlinear singular con-
trol systems, International Journal of Control, Vol. 68, pp. 753-774, 1997.
[11] H. Hahn, Simulationsmodelle für Zahnstangenlenkungen, BMW Bericht
EG-14/16-474, 1978.
322 References
[12] H. Hahn, A. Röttger, Rechnersimulationsmodell zur Analyse von Hinter-

achsaufhängungen, BMW Bericht EG-514/2651, 1978.
[13] H. Hahn, Dreidimensionales Modell der elastischen Lagerung eines Mo-
tors im Fahrzeug, BMW Bericht EG-142/16-474, 1979.
[14] M. Neumann, H. Hahn, Computer simulation and dynamic analysis of
a mechanical press based on different engineer models, Mathematics and
Computers in Simulation, Vol. 46, pp. 559-574, 1998.
[15] H. Hahn, H. W. Wagener, M. Neumann, T. Freiherr, Rechnergestützte
Analyse des dynamischen Verhaltens mechanischer Pressen, EFB For-
schungsbericht Nr. 95, 1997.
[16] M. Neumann, H. Hahn, K.-D. Leimbach, Mathematical modeling, com-
puter simulation and dynamic analysis of a single-point-drive eccentric
press, Proceedings of the second Mathmod Vienna (IMACS Symposium
on Mathematical Modeling), February 5-7, 1997, pp. 291-299, Ed. I.
Troch and F. Breitenegger, 1997.
[17] H. Hahn and I. Aavatsmark, Mathematisches Modell des Rollprüfstands
mit Zweiachslaufwerk als Prüfling, IABG Bericht B- TF-1191, Vols. I, II,
III, and IV, 1983.
[18] H. Hahn, G. Siegl, Numerical Simulation and Experimental Verification
of the German Roller Rig for Rail Vehicles, Vehicle System Dynamics,
Vol. 15, pp. 303-334, 1986.
[19] H. Hahn, G. Siegl, Simulation und Experimentelle Verification des Rad/
Schiene Rollprüfstands, Archive of Applied Mechanics, Vol. 57, pp. 329-
348, 1987.
[20] J. J. Craig, Introduction to Robotics, Mechanics and Control, Addison-
Wesley, Reading, Mass., 1986.
[21] H. Hahn, Numerische Modellierung eines seriellen Sechsfreiheitsgrad-
roboters mit dem Starrkörperprogramm NUSTAR, Interner Bericht des
Fachgebiets Regelungstechnik, Universität Kassel, 1988.
[22] R. P. Paul, Robot Manipulators: Mathematics, Programming and Con-
trol, MIT Press, Cambridge, Mass., 1982.
[23] H. Hahn, M. Niebergall, Development of a Measurement Robot for Iden-
tifying the ten Inertia Parameters of a Rigid Body in a Single Ex-
periment, IEEE Transactions on Control Systems Technology, Vol. 9,
pp. 416-423, 2001.
[24] M. Niebergall, H. Hahn, Identijication of the Ten Inertia Parameters of
a Rigid Body, Nonlinear Dynamics, Vol. 13, pp. 361-372, 1997.
References 323
[25] H. Hahn, K. Keller, W. Raasch, Multi-Axis Servo-Hydraulic Test Facility

Mas 83 (Simulation of Actual Tests), IABG Bericht B-TF-1713, 1984.
[26] H. Hahn, K. Keller, W. Raasch, Multi-Axis Servo-Hydraulic Test Facility
Mas 83 (Eigenvalue Analysis of the Test Facility with Test Item), IABG
Bericht B-TF-1683, 1984.
[27] H. Hahn, W. Raasch, Multi-Axis Vibration Test on Spacecraft Using
Hydraulic Exciters, AGARD Conference Proceedings No. 397, Septemer
9-13, 1985, Oberammergau, pp. 23-1-23-23, 1986.
[28] H. Hahn, K.-D. Leimbach, Simulation of a spatial servo-hydraulic test fa-
cility for space structures, Proceedings of the 18th Space Simulation Con-
ference, October 31-November 3, 1994, Baltimore, IES-NASA, pp. 427-
439, 1994.
[29] H. Hahn, W. Klier, Nonlinear control of a multi axis servopneumatic
test facility with redundant actuators, Proceedings of the International
Conference on Systems, Signals, Control and Computers, September 22-
24, 1998, Ed. V. Bajic, Vol. II, pp. 461-465, Durban, South Africa, 1998.
[30] W. Klier, H. Hahn, Mathematisch-physikalische Modellgleichungen und
Regelstrategien räumlicher servopneumatischer Parallelroboter, VDI Be-
richt Nr. 1552 der VDI-GMA Fachtagung Robotik 2000, pp. 71-76, 2000.
[31] M. Neumann, H. Hahn, Entwicklung und Realisierung eines pneuma-
tischen Mehrachsenprüfstands, VDI Bericht Nr. 1552 der VDI-GMA
Fachtagung Robotik 2000, pp. 59-64, 2000.
[32] D. Stewart, A platform with six degrees of freedom, Proc. ofthe Institute
of mechanical engineers, Vol. 180, pp. 371-386, 1965.
[33] Y. K. Byun, H. S. Cho, Analysis of a novel six-dof, 3 PPSP parallel
manipulator, Int. Journal of Robotics Research, Vol. 16, pp. 859-872,
1997.
[34] H. Hahn, W. Klier, D. Fürst, Mathematical modeling of spatial servo-
pneumatic parallel robots, RTS Bericht 33, University of Kassel, 2000.
[35] H. Hahn, Erstellung eines Ingenieursmodells zum Fahrzeug UNIMOG
DB U1300L (Type 435), IABG Bericht B-TF-2231, 1987.
[36] H. Hahn, G. Siegl, R. Wehage, Datenerhebung zum Fahrzeug UNIMOG
DB U1300L (Type 435), IABG Bericht B-TF-2233, 1987.
[37] H. Hahn, Eingabedatenfile zum Rechnersimulationsmodel eines Unimogs,
Interner Bericht des Fachgebiets Regelungstechnik und Systemdnamik
(RTS), 1989.
324 References
[38] J. Eysenbach, H. Hahn, Überprüfung des Rechenmodells FANAL anhand

von Verifikationsfahrten, IABG Bericht B-TF-1470, 1984.
[39] J. Eysenbach, NUSTAR Version 32, Anwenderhandbuch, Version 34.0,
IABG Bericht B-TA-3640, 1998.
[40] E. J. Haug, Computer Aided Kinematics and Dynamics of Mechanical
Systems, Allyn and Bacon, Newton, Mass., 1989.
[41] P. E. Nikravesh, Computer Aided Analysis of Mechanical Systems, Pren-
tice Hall, Upper Saddle River, N.J., 1988.
[42] J. Wittenburg, Dynamics of Systems of Rigid Bodies, Teubner, Stuttgart,
1977.
[43] R. Wehage, Quaternions and Euler Parameters - A brief Exposition,
Computer Aided Analysis and Optimization of Mechanical System Dy-
namics, Ed. E. G. Haug, pp. 147-180, Springer, New York, 1984.
[44] R. Wehage, Application of Quaternions to Rigid Body Kinematics and
Dynamics, Unpublished personal communication, 1984.
[45] R. Wehage, H. Hahn, Quaternion and Vector Algebra: Application to
rigid Body Kinematics and Dynamics, IABG Bericht B-TF-2071, 1986.
[46] E. Hille, Analytic Funtion Theory, Vol. II, Ginn, Boston, 1972.
[47] Raghavan Narasimhan, Introduction to the Theory of Analytic Spaces,
Springer Lecture Notes in Mathematics, Vol. 25, Springer, Berlin Hei-
delberg New York, 1966.
[48] J. Mather, Stability of C 00 -mappings, Publications Mathematics, IHES,
Vol. 35, pp. 127-156, 1968.
[49] D. T. Whiteside, The Mathematical Papers of Isaac Newton, Vol. III,
Garnbridge University Press, Cambridge, 1969.
[50] H. Hahn, Higher Order Root Locus techniques with Applications in Con-
trol System Design, Advances in Control Systems and Signal Processing,
Ed. I. Hartmann, Vieweg, Wiesbaden, 1981.
[51] D. H. Sattinger, Topics in Stability and Bifurcation Theory, Springer
Lecture Notes in Mathematics, Vol. 309, Springer, Berlin Heidelberg
New York, 1973.
[52] M. M. Vainberg, V. A. Trenogin, Theory of branching of solutions of
nonlinear equations, Noordhoof, Groningen, 1974.
[53] R. Thom, Stabilite structurelle et morphogenese, Benjamin, New York,
1972.
References 325
[54] G. Wassermann, Stability of Unfoldings, Springer Lecture Notes in Math-

ematics, Vol. 393, Springer, Berlin Heidelberg New York, 1975.
[55] R. Grammel, Kinetisch unbestimmte Systeme, Zeitschift angewandte
Mathematik und Mechanik, Vol. 24, pp. 215-223, 1944.
[56] A. Sommerfeld, Analytische Mechanik, Diederich'sche Verlagsbuchhand-
lung, Wiesbaden, 1949.
[57] H. Goldstein, Klassische Mechanik, Akademische Verlagsgesellschaft,
Frankfurt am Main, 1963.
[58] K. H. Hunt, Kinematic Geometry of Mechanisms, Charleston, Oxford,
1978.
[59] C. Hibbeler, Engineering Mechanics, MacMillan, New York, 1978.
[60] L.A. Pars, A Treatise on Analytical Dynamics, Oxbow, Woodbridge,
Con., 1979.
[61] R. E. Robertson and R. Schwertassek, Dynamics of Multibody Systems,
Springer, Berlin Heidelberg New York, 1988.
[62] R. I. Huston, Multibody Dynamics, Butterworth-Heinemann, Boston,
1990.
[63] K. Popp, W. Schiehlen, Fahrzeugdynamik, Teubner, Stuttgart, 1993.
[64] D. Bestle, Analyse und Optimierung von Mehrkörpersystemen, Springer,
Berlin Heidelberg New York, 1991.
[65] D. A. Wells, Theory and Problems of Lagrangean Dynamics, Schaum's
Outline Series, McGraw-Hill, New York, 1967.
[66] C. W. Gear, Differential-Algebraic Equation Index Transformation, SIAM
Journal on Scientific Statistical Computing, Vol. 9, pp. 39-47, 1988.
[67] C. Führer, Differential-algebraische Gleichungssysteme in mechanischen
Mehrkörpersystemen, Theorie numerischer Ansätze und Anwendungen,
Dissertation, Mathematisches Institut, TU München, 1988.
[68] P. C. Müller, Analysis and Control of Linear Descriptor Systems, Pro-
ceedings of the International Conference on Systems, Signals, Control
and Computers, September 22-24, 1998, Ed. V. Bajic, Vol. I, pp. 11-17,
Durban, South Africa, 1998.
[69] P. C. Müller, Linear-Quadratic Optimal Control of Mechanical Descrip-
tor Systems, Systems and Networks: Mathematical Theory and Appli-
cations, Proceedings of the International Symposium MTNS-93, August
1993, Regensburg, Eds. U. Helmke, R. Mennicken, J. Sauer, Vol. 11,
pp. 361-366. Akademie, Berlin, 1994.
326 References
[70] H. Hahn, H.-J. Sommer, On the Causality of Linear Time Invari-

ant Descriptor Systems, Zeitschrift für Augewandte Mathematik und
Mechanik, Vol. 81, pp. 753-768, 2001.
[71] I. S. Duff, J. K. Reid, Some Design Features of a Sparse Matrix Code,
ACM Transactions on Mathematical Software, Vol. 5, pp. 18-35, 1979.
[72] R. A. Wehage, E. J. Haug, Generalized Coordinate Partitioning for Di-
mension Reduction in Analysis of Constrained Multibody Systems, Jour-
nal of Mechanical Design, Vol. 104, pp. 247-255, 1982.
[73] S. L. Campel, Singular Systems of Differential Equations II, Pitman,
London, 1982.
[74] W. C. Reinboldt, On the existence and uniqueness of solutions of nonlin-
ear implicit differential-algebraic equations, Nonlinear Analysis, Theory,
Methods and Applications, Vol. 16, pp. 642-661, 1991.
[75] A. Brasey, Half-explicit methods for semi-explicit differential algebraic
equations of index 2, Dissertation, University of Geneva, 1994.
[76] E. Hairer, C. Lubich, and M. Roche, The Numerical Solution of Differen-
tial-Algebraic Systems by Runge-Kutta Methods, Spinger Lecture Notes
in Mathematics, Vol. 1409, Springer, Berlin Heidelberg New York, 1989.
[77] E. Eich, Projezierte Mehrschrittverfahren zur numerischen Lösung der
Bewegungsgleichungen von Mehrkörpersystemen mit Zwangsbedingungen
und Unstetigkeiten, Fortschritts-Berichte VDI, Reihe 18, VDI, Düsseldorf,
1992.
[78] C. Lubich, U. Novak, C. Engstler, MEXX - Numerical Software for
the Integration of Constrained Mechanical Multibody Systems, Technical
report SC92-12, Konrad-Zuse-Zentrum für Informationstechnik, Berlin,
1992.
[79] A. Brasey, A half-explicit Runge-Kutta method of order Jive for solving
constrained mechanical systems, Computer Methods in Applied Mechan-
ics and Engineering, Vol. 48, pp. 191-201, 1992.
[80] B. Simeon, Numerische Integration mechanischer Mehrkörpersysteme:
Projezierte Deskiptorformen, Dissertation, Fakultät für Mathematik, TU
München, Fortschritts-Berichte VDI, Reihe 20, VDI, Düsseldorf, 1994.
[81] E. Eich-Soellner, F. Führer, Numerical Methods in Multibody Dynamics,
Teubner, Stuttgart, 1998.
[82] J. Baumgarte, Stabilization of contraints and integrals of motion in dy-
namic systems, Computer Methods in Applied Mechanics and Engineer-
ing, Vol. I, pp. 1-16, 1972.
References 327
[83] H. Hahn, B. Simeon, Separation Principle of Mechanical System Models

Including Stabilized Constraint Relations, Archive of Applied Mechanics
(Ingenieur-Archiv), Vol. 64, pp. 147-153, 1994.
Index
acceleration, 3, 6, 23, 34, 47, 140, 239 singular, 82

angular, 6, 34, 47 stabilization, 166, 168, 169
analysis control
constraint synthesis, 3
acceleration, 97 coordinate
velocity, 97 Cartesian, ix, 5, 39, 54, 76, 78, 299
dynamic, ix, 1, 76, 78, 82-84, 86, 87, driving, 81
120 generalized, 3, 7, 76, 120, 121, 171,
eigenvalue, viii, 26 186, 249, 290, 295
kinematic, viii, 1, 76, 79, 81, 83, 84, global, 5, 76, 77
120 local, 77
position, 81 minimal, 3, 5, 290
sensitivity, viii, 26 relative, 171, 173, 186, 200, 210, 215,
singularity, 89, 117 216, 228, 233, 305, 30~ 309
static, viii
damper
bifurcation, 109, 112, 113
torsional, 250
Bryant angle, ix, 33, 54, 62, 67, 72, 73,
translational, 249
85, 124, 158, 160, 252, 280, 298, 299
damping coefficient, 247, 250
building block, 173, 200, 202, 206, 207,
degree of freedom, 91, 198
212, 215, 216, 218, 220, 222-224, 226,
derivative, ix, 121, 255, 282
230, 231, 317
time, ix, xi, 6, 33, 47, 48, 53, 65, 68,
consistent, 1, 76, 84, 85, 120 108, 202, 204, 205, 207, 211, 214,
constraint 215, 219, 247, 249, 250, 255, 267,
absolute 274, 275, 298, 299, 306, 312, 316,
distance, 184 317
orientation, 181 direction cosine, 59
orientation/partial-position, 174
position, 174, 179 eigenvalue, 168, 169
common-point, 90, 200, 201, 228 numerical, 169
constant-distance, 174, 186, 196, 200, physical, 169
317 energy
equation, 121, 181, 182, 184, 192, kinetic, 290-292, 294
196, 197, 305, 309, 310, 313, 314, potential, 290
316 engineering model, viii, 22-24, 165, 166
holonomic, 76, 78, 84 equation
point follower, 200 constraint, 121, 181, 182, 184, 192,
relative 196, 197, 305, 309, 310, 313, 314,
orientation/partial position, 186 316
position, 186, 190 differential (DE), ix, 1, 4, 26, 53, 61,
rotation blocker, 200 65, 72, 73, 78, 84, 85, 150, 151,
330 Index
158, 165, 168, 169, 298, 299, 302, joint

303 cylindrical, 200, 231, 233
differential-algebraic (DAE), viii-x, prismatic, 200, 234-237
1, 5, 22, 26, 78, 123, 129, 151, revolute, 174, 179, 186, 189, 190, 198,
156-158, 162, 163, 165, 167-169, 200, 228, 250, 253, 301
296, 301 spherical, 200, 202, 220, 301
model, vii-xi, 1, 3, 5, 7, 18, 21, translational, 174, 182, 186, 189, 193,
22, 25, 26, 82, 83, 85, 123, 129, 200, 223, 225, 235
150-152, 156, 158, 161, 162, 168, universal, 200, 226, 228
169, 171, 186, 231, 235, 239, 249, general, 200, 255, 302, 318
255, 302, 318
motion, vii, viii, xi, 3, 7, 37, 126, 129, kinematics, xi, 3, 21, 33, 75, 243, 282
140, 143, 149, 151, 152, 154-156, planar, 151
159, 271, 290, 296, 298 spatial, 7
Euler, 142
Newton-Euler, 129, 143, 146, 150, Lagrange
255 formalism, x, 18, 129, 290
orientation loop, 171-173, 192, 204, function, 7, 18, 290
224, 226, 228, 305 multiplier, 78, 122, 123, 126, 160,
state-space, ix, 26 162, 290, 301
vector loop, 90, 171-173, 176, 179, link, 1, 76
182, 185, 189, 190, 194, 196, 208, massless, 189, 303
218, 226, 228, 245, 309, 312, 313, revolute-translational, 174, 186
315, 316 spherical-spherical, 186, 200, 222,
Euler's law, 6, 137, 141, 143 302, 303
lock-up, 87, 89, 101, 102, 109, 110,
force, x, 5, 6, 24, 34, 37, 121, 137, 152, 112-114, 116, 117
154 lock-up points, 97
applied, 251
centrifugal, 300 mapping
damper, 247, 250 bilinear, 256
external, 137, 239 linear, 256, 257
gravitational, 239, 251 matrix
translational, 243, 246 kinematic, 166
weight, 241 regular, 122, 125
frame skew-symmetric, 46, 64, 70, 71, 257,
body-fixed, 61, 66, 76, 133, 212 262, 264, 271, 278
global, 4, 133 sparse, 166
inertial, 6, 24, 78, 129, 137, 140, 174, mechanism
201, 291, 303 planar, 78, 83, 87, 167, 300, 301
local, 201, 208, 303 rack-and-pinion, 196
reference, 24, 34, 38, 54, 139, 271 slider-crank, 87, 89, 91, 94, 103, 117
function spatial, vii-ix, 1, 3, 5, 47
driving, 91, 109 steering, 7
transfer, ix, x, 26 model
vector, ix, 33, 146, 255, 288 engineering, viii, 22-24, 165, 166
equation, vii-xi, 1, 3, 5, 7, 18, 21,
inertia 22, 25, 26, 82, 83, 85, 123, 129,
matrix, 129, 133-135, 141, 263, 300 150-152, 156, 158, 161, 162, 168,
moment of, 6, 134, 135, 300 169, 171, 186, 231, 235, 239, 249,
parameter, 24, 173 255, 302, 318
product of, 134 symbolic, 26
initial condition, 93 mathematical, 174
Index 331
press, 7 situation, 33, 93, 94, 98, 103, 105,

roller rig, 7, 18 106, 108, 117
symbolic, 26 singularity, 61, 166
tank, 18 situation
truck, 18, 23, 24 pull, 245
vehicle push, 245
off-road, 18, 169 singular, 33, 93, 94, 98, 103, 105,
moment 106, 108, 117
couple, 137, 139 solution
force, 137 bifurcation, 115
momentum existence of, 83
angular, x, 6, 34, 129-131, 133, 134, uniqueness of, 83, 326
137 space
linear, x, 6, 34, 129-131, 137, 141 Euclidean, 257
vector, 33, 255-257
Newton's law, 123, 137, 140 spring, 1, 18, 24, 198, 246
torsional, 198, 200, 250
point translational, 198
attachment, 246 state-space
reference, 24, 129, 131, 135, 137, 139, equation, ix, 26
140, 143, 152, 157, 158, 291, 292, form, x, 5
294, 303 Steiner-Huygens, 135, 136
product symbolic computation, 18
dyadic, 263
scalar, 34, 44, 173, 174, 189, 255, Taylor series
257-259, 261, 263 expansion, 121, 125
vector, 34, 35, 46, 47, 53, 255, 257, term
259, 261-263, 298 gyroscopic, 300
program theorem
general-purpose, xi, 22 Fredholm's alternative, 83
rigid-body, viii, xii, 18, 22, 166, 169 implicit function, 81, 82
special-purpose, 18 Malgrange preparation, 83
Weierstraß preparation, 83
rigid body transformation
constrained, 155, 160 basis, 57, 60
unconstrained, x, 6, 137, 149, 172, coordinate, 40, 41, 43, 60
291, 296
rigid-body vector
condition, 66 algebraic, 33, 37, 39, 240, 256-258,
mechanism, vii, ix, x, 3, 166, 168 260
system, vii-ix, xi, 1, 3, 7, 22, 33, 37, basis, 38, 54, 275
44, 65, 75, 151, 173, 246, 271, 290 displacement, 34, 36, 43, 54, 303
rotation force, 37, 242
orthogonal, 46, 47, 176, 189, 212, geometric, 33, 34, 36, 37, 54
270, 272 gradient, ix
planar, 270, 272 sliding, 139
spatial, 61 spatial, 260
torque, 36, 140
Separation principle, 169 velocity, 1, 6, 120, 167, 196, 233, 290
shortest distance, 319, 320 angular, 23, 34, 47, 49, 53, 68, 70-72,
singular 140, 270, 271, 298
behavior, 75, 83, 87, 117 virtual
case, 94, 95, 97, 108, 111, 112, 115 displacement, 121
configuration, 99-102 work, 5, 121, 124
List of figures
1.1 Examples of multibody systems.......................... 2

1.2 Global (inertial) and local (relative) vectors and frames . . . . . 4
1.3 Examples of simple unconstrained and constrained mechan-
ical systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Examples of simple mechanisms including joints . . . . . . . . . . . 9
1.5 Slightly more complex mechanisms....................... 10
1.6 Steering mechanism of an automobile . . . . . . . . . . . . . . . . . . . . . 11
1. 7 Photograph, schematic drawing, and engineering model of a
single-point-drive eccentric press (LVWU laboratory, Univer-
sity of Kassel) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8 Roller rig oftheGerman railway company (DB AG)........ 13
1.9 Roller rig oftheGerman railway company (DB AG)........ 14
1.10 Roller rig and bogie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.11 Roller rig and bogie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.12 Drawing of elastic and dissipative coupling of a bogie . . . . . . . 17
1.13 Drawings and photographs of serial and parallel robots . . . . . 19
1.14 Photograph (a), technical drawing (b), engineering model
(c), and animation graphics (d) of a truck obtained using
the program NUSTAR at IABG, Ottobrunn............... 20
1.15 Animation graphics ofthegerman tank Leopard 11, obtained
using the program NUSTAR at IABG, Ottobrunn.......... 21
1.16 Measured and simulated vertical transient motions of a truck
crossing an obstacle (a ramp): without sophisticated engine
and driver modules (a) and including sophisticated engine
and driver modules (b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.17 Planar mechanisms including a singlerigid body . . . . . . . . . . . 28
1.18 Planar mechanisms including two rigid-bodies . . . . . . . . . . . . . 29
1.19 Mechanisms including a single rigid body subject to large
spatial motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.20 Mechanisms including several rigid bodies subject to large
spatial motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1 Geometrie vector, vector length, vector sum, projections, and

vector product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Different types of geometric vectors in rigid-body dynamics 36
334 List of figures
2.3 Algebraic vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.4 Vector r PO represented in frames R and L . . . . . . . . . . . . . . . . 40
2.5 Representation of the basis vectors exL , eyL of frame L in
frame R .............................................. 42
2.6 Translation and rotation of frame L with respect to frame R 43
2. 7 Vector representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.8 Bryant-angle transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.1 Global and local frames in the plane. . . . . . . . . . . . . . . . . . . . . . 77

3.2 Slider-crank mechanism driven by two different active con-
straints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.3 Singularsituations of Case 1 ('1/Ji(t) = a 1 (t) for '1/Jj = ~7r or
'1/Jj = ~7r) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.4 Singular situations of the slider-crank mechanism of Case 1
(drive a 1 (t)) ........................................... 104
3.5 Singular solutions of the slider-crank mechanism of Case 2
(drive a 2 (t)) ........................................... 118
3.6 Constraint acceleration analysis of the slider-crank mecha-
nism of Case 2 (drive a 2 (t)) ............................. 119
3. 7 Revolute joint and the associated constraint reaction forces .. 124
4.1 Vector diagram used in the definitions of linear momentum

and angular momentum ................................. 130
4.2 Forces and torques ..................................... 138
4.3 Unconstrained planar rigid body i together with its free-body
diagram .............................................. 153
4.4 System of unconstrained planar rigid bodies ............... 154
4.5 A single constrained rigid body and its free-body diagram ... 155
4.6 System of constrained planar rigid bodies ................. 156
5.1 Geometrical representation of a vector loop including two

bodies ............................................... 172
5.2 Common absolute joints ................................ 175
5.3 Drawing associated with (5.2a) and (5.2b) ................ 176
5.4 Vector diagram and technical realizations of mechanisms that
include an absolute massless revolute-translational joint .... 177
5.5 Drawings of (absolute) revolute joints .................... 180
5.6 Drawings of an absolute orientation constraint ............ 181
5. 7 Drawings of translational joints .......................... 183
5.8 Mechanisms that include a combined absolute orientation/partial-
position constraint ..................................... 184
5.9 Vector diagram of a mechanism with a constant-distance con-
straint (massless revolute-revolute link) ................... 185
5.10 Common relative joints ................................. 187
5.11 Two types of massless revolute-translationallinks .......... 188
List of figures 335
5.12 Mechanisms (wheel suspensions) including relative partial-

position constraints (massless revolute-translationallinks) .. 189
5.13 Drawings of revolute joints between two bodies ............ 191
5.14 Drawings of a relative orientation constraint .............. 193
5.15 Mechanism including a gear set .......................... 194
5.16 Drawings of (relative) translational joints ................. 195
5.17 Technical realizations of joints, modeled by a combined rel-
ative orientationfpartial-position constraint ............... 197
5.18 Vector diagram of a mechanism with a constant-distance con-
straint ................................................ 198
5.19 Combinations of planar translational and torsional springs
with joints ............................................ 199
5.20 Vector diagram of a common-point constraint ............. 201
5.21 Grientation diagram of a paraUel-axes constraint . . . . . . . . . . . 205
5.22 Vector loop of a straight-line-point-follower constraint ...... 209
5.23 Geometrical situation of a rotation-blocker constraint ....... 213
5.24 Vector loop associated with a massless spherical-spherical
link .................................................. 219
5.25 Spherical joint ......................................... 222
5.26 Massless spherical-spherical link . . . . . . . . . . . . . . . . . . . . . . . . . 223
5.27 Spatial translational joint ............................... 224
5.28 Drawings of a universal joint ............................ 227
5.29 Vector loop and orientation loop of a revolute joint ......... 229
5.30 Vector diagram of a cylindrical joint ...................... 232
5.31 Vector diagram of a translational (prismatic) joint ......... 235
6.1 Geometrical representation of a force F and a torque M . . . . 240

6.2 Gravitational field acting on a body in the ( -eyR) direction 241
6.3 A body acted upon by a constant force Fi and moment Mi. 242
6.4 Forces and vector diagram of a translational force element . . 244
6.5 Vector diagram of a translational force element ............ 246
6.6 Translational spring between points Pi and P j of bodies i
and j ................................................. 247
6.7 Translational damper between points Pi and P j of bodies i
and j ................................................. 248
6.8 Torsional spring and damper ............................ 251
6.9 Constant force F i acting on a body i . . . . . . . . . . . . . . . . . . . . . 252
A.l.1 Vector diagram of a scalar product ....................... 259

A.l.2 Vector diagram of a vector product ....................... 260
A.2.1 Vector diagram used in the definition of the kinetic energy
of a rigid body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
A.4.1 Drawing of a general universal joint (massless revolute-
revolute link) connecting two rigid bodies ................. 302
336 List of figures
A.4.2 Diagrams for deriving the vector loop and orientation loop
equations of the mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
A.4.3 Shortest distance between two rotation axes ............... 320

Professor Dr. Hubert Hahn (Auth.) - Rigid Body Dynamics of Mechanisms - 1 Theoretical Basis-Springer-Verlag Berlin Heidelberg (2002) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Professor Dr. Hubert Hahn (Auth.) - Rigid Body Dynamics of Mechanisms - 1 Theoretical Basis-Springer-Verlag Berlin Heidelberg (2002) PDF

Uploaded by

Copyright:

Available Formats

HubertHahn

Rigid Body Dynamics of Mechanisms 1

Rigid Body Dynamics

ISBN 978-3-642-07617-6 ISBN 978-3-662-04831-3 (eBook)

library of Congress Cataloging-in-Publication-Data

The dynamics of mechanical rigid-body mechanisms is a highly developed

Conventional methods of modeling rigid-body mechanisms

In contrast to the comparatively simple and easy-to-learn basic laws of rigid-

1. A large variety of specialized techniques have been developed, each suit-

General-purpose rigid-body analysis programs as efficient modeling

Objectives of this monograph

Volume I of this monograph presents:

Organization of the books (Volumes I and II)

The two volumes of this monograph provide a systematic theoretical approach

of the constraint equations, illustrated by an example. Kinetic equations of

1. In several industrial companies (like BMW and IABG) for practicing

work of the rigid-body analysis program NUSTAR at IABG, and Dipl.-Ing

2. Planar and spatial vectors, matrices, and vector functions 33

3. Constraint equations and constraint reaction forces ofmecha-

4. Dynamics of planar and spatial rigid-body systems ....... 129

4.3.1 Equations of planar motion of unconstrained rigid

5. Model equations of planar and spatial joints .............. 171

5.1.2.1.2 Complete-position constraint (revo-

6. Constitutive relations of planar and spatial external forces

6.1.2 Applied force and moment ......................... 241

A. Appendix ................................................. 255

1.1 Tasks in multibody simulation, analysis, and control

Multibody systems are commonly investigated under different aspects, de-

Fig. 1.1: Examples of multibody systems

In inverse kinetic ( dynamic) analysis the time histories of the minimal

1.2 Coordinates and frames

Model equations (equations of motion) of rigid-body systems may be for-

as a rigid body. Rigid-body configurations may be specified by introducing

1.3 Formulation of the model equations

formulation is difficult. This approach, however, may be desirable in several

Formulation of model equations and system constraints in global coor-

Here the equations of kinetics (dynamics) of rigid bodies will be mainly

with L as the Lagrange function, p = (rT, TJT)T as vector of a minimum

1.4 Prototype applications of rigid-body mechanisms

(a) Systems with one translational degree of freedom

(b) Systems with one rotational degree of freedom

(c) Discrete model of a string (d) Coupled pendulum

Fig. 1.3: Examples of simple unconstrained and constrained mechanical systems

Fig. 1.4: Examples of simple mechanisms including joints

(c) Planar modelsofasimple locomotive and of a planartest facility

(d) Planar models of an inverted pendulum and of an excavator

Fig. 1.5: Slightly more complex mechanisms

Fig. 1.6: Steering mechanism of an automobile

revol ute joint ~

(a) Photograph of the roller rig

(b) Technical drawing of the roller rig

(a) Roller rig loaded by an ICE high-speed locomotive

(b) Engineering model of a roller stand loaded by a wheel set

(a) Roller rig with bogie (technical drawing)

(b) Bed plate and roller stands (technical drawing)

Fig. 1.10: Roller rig and bogie

(a) Roller rig with bogie (technical drawing)

(b) Rollerstand (engineering model)

Fig. 1.11 : Roller rig and bogie

(front and side views)

may be modeled by special-purpose rigid-body programs, obtained for exam-

lner ia measurement robot devel-

(a) Serial robots