Dynamics of The Shell Class of Tensegrity Structures

Journal of the Franklin Institute 338 (2001) 255320
Dynamics of the shell class of tensegrity

structures
Robert E. Skeltona, Jean Paul Pinauda,*, D.L. Mingorib
a
Mechanical and Aerospace Engineering Department, Structural Systems & Control Lab,
University of California San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0411, USA
b
Mechanical and Aerospace Engineering Department, University of California Los Angeles,
Los Angeles, CA 90095-1597, USA
Abstract
A tensegrity structure is a special truss structure in a stable equilibrium, with selected
members designated for only tension loading, and the members in tension form a continuous
network of cables separated by a set of compressive members. This paper develops an explicit
analytical model of the nonlinear dynamics of a large class of tensegrity structures, constructed
of rigid rods connected by a continuous network of elastic cables. The kinematics are
described by positions and velocities of the ends of the rigid rods, hence, the use of angular
velocities of each rod is avoided. The model yields an analytical expression for accelerations of
all rods, making the model ecient for simulation, since the update and inversion of a
nonlinear mass matrix is not required. The model is intended for shape control and design of
deployable structures. Indeed, the explicit analytical expressions are provided herein for the
study of stable equilibria and controllability, but the control issues are not treated in this
paper. # 2001 The Franklin Institute. Published by Elsevier Science Ltd. All rights reserved.
Keywords: Tensegrity; Rigid body; Dynamics; Flexible structures
1. Introduction
The history of structural design can be divided into four eras classified by design
objectives: in the prehistoric era which produced such structures as Stonehenge, the
*Corresponding author. Tel.: +1-858-534-6146; fax: +1-858-822-3107.
E-mail addresses: bobskelton@ucsd.edu (R.E. Skelton), jpinaud@mae.ucsd.edu (J.P. Pinaud),
mingori@seas.ucla.edu (D.L. Mingori).
0016-0032/01/$20.00 # 2001 The Franklin Institute. Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 6 - 0 0 3 2 ( 0 0 ) 0 0 0 7 8 - 8
256
R.E. Skelton et al. / Journal of the Franklin Institute 338 (2001) 255320
objective was simply to oppose gravity, to take the static loads. The classical era,
considered the dynamic response and placed design constraints on the eigenvectors as
well as eigenvalues. In the modern era, design constraints could be so demanding that
the dynamic response objectives require feedback control. In this era the control
discipline followed the classical structure design, where the structure and control
disciplines were ingredients in a multidisciplinary system design, but no interdisciplinary tools were developed to integrate the design of the structure and the control.
Hence, in this modern era, the dynamics of the structure and control were not
cooperating to the fullest extent possible. The post-modern era of structural systems
is identified by attempts to unify the structure and control design to a common
objective.
The ultimate performance capability of many new products and systems cannot be
achieved until mathematical tools exist that can extract that full measure of
cooperation possible between the dynamics of all components (structural components, controls, sensors, actuators, etc.) This requires new research. Control theory
describes how the design of one component (the controller) should be influenced by
the (given) dynamics of all other components. However, in systems design, where
more than one component remains to be designed, there is inadequate theory to
suggest how the dynamics of two or more components should influence each other at
the design stage. In the future, controlled structures will not be conceived merely as
multidisciplinary design steps, where a plate, beam or shell is first designed, followed
by the addition of control actuation. Rather, controlled structures will be conceived
as an interdisciplinary process in which both material architecture and feedback
information architecture will be jointly determined. New paradigms for material and
structure design might be found to help unify the disciplines. Such a search motivates
this work. Preliminary work on the integration of structure and control design
appears in [13].
Bendsoe and others [47] optimize structures by beginning with a solid brick and
deleting finite elements until minimal mass or other objective functions are
extremized. But, a very important factor in determining performance is the
paradigm used for structure design. This paper describes the dynamics of a
structural system composed of axially loaded compression members and tendon
members that easily allow the unification of structure and control functions. Sensing
and actuating functions can sense or control the tension or the length of tension
members. Under the assumption that the axial loads are much smaller than the
buckling loads, we treat the rods as rigid bodies. Since all members experience only
axial loads the mathematical model is more accurate than models of systems with
members in bending. This unidirectional loading of members is a distinct advantage
of our paradigm, since this eliminates many nonlinearities that plague other
controlled structural concepts: hysterisis, friction, deadzones, backlash.
It has been known since the middle of the 20th century that continua cannot
explain the strength of materials. While science can now observe at the nanoscale, to
witness the architecture of materials preferred by nature, we cannot yet design
or manufacture man-made materials that duplicate the incredible structural
eciencies of natural systems. Natures strongest fiber, the spider fiber, arranges
257
simple non-toxic materials (amino acids) into a microstructure that contains a

continuous network of members in tension (amorphous strains) and a discontinuous
set of members in compression (the b-pleated sheets in Fig. 1 [8,9]).
This class of structure, with a continuous network of tension members and
discontinuous network of compression members, will be called a Class 1 tensegrity
structure. The important lessons learned from the tensegrity structure of the spider
fiber is that:
(i) Structural members never reverse their role. The compressive members never
take tension, and of course, tension members never take compression.
(ii) The compressive members do not touch (there are no joints in the structure).
(iii) The tensile strength is largely determined by the local topology of tension and
compressive members.
Another example from nature, with important lessons for our new paradigms is
the carbon nanotube often called the Fullerene (or Buckytube), which is a derivative
of the Buckyball. Imagine a 1-atom thick sheet of a graphene, which has hexagonal
holes, due to the arrangements of material at the atomic level (see Fig. 2). Now
imagine that the flat sheet is closed into a tube by choosing an axis about which the
sheet is closed to form a tube. A specific set of rules must define this closure which
Fig. 1. Natures strongest fiber: the spider fiber [9].
258
Fig. 2. Buckytubes [27].
takes the sheet to a tube, and the electrical and mechanical properties of the resulting
tube depends on the rules of closure (axis of wrap, relative to the local hexagonal
topology) [27]. Smalley won the Nobel Prize in 1996 for these insights into the
Fullerenes. The spider fiber and the Fullerene provide motivation to construct manmade materials whose overall mechanical, thermal, and electrical properties can be
predetermined by choice of the local topology and the rules of closure which
generate the three-dimensional structure from a given local topology. By combining
these motivations from Fullerenes with the tensegrity architecture of the spider fiber,
this paper will derive the static and dynamic models of a shell class of tensegrity
structures. Future papers will exploit the control advantages of such structures. The
existing literature on tensegrity deals mainly with statics [1022], with some
elementary work on the dynamics in [2325].
2. Tensegrity definitions
Kenneth Snelson built the first tensegrity structure in 1948 (Fig. 3) and
Buckminster Fuller coined the word tensegrity. For 50 years tensegrity has
existed as an art form with some architectural appeal, but the engineering use has
been hampered by the lack of models for the dynamics. In fact, the engineering use of
tensegrity has been doubted by the inventor himself, Kenneth Snelson, As I see it,
this type of structure, at least in its purest form, is not likely to prove highly ecient
or utilitarian, [letter to R. Motro, International Journal of Space Structures]. This
259
Fig. 3. Needle Tower of Kenneth Snelson, Class 1 tensegrity. Krueller-Mueller Museum, Otterlo,
Netherlands.
statement might partially explain why no one bothered to develop math models to
convert the artform into an engineering practice. We seek to use science to prove the
artist wrong, that his invention is indeed more valuable than the artistic scope that he
ascribed to it. Mathematical models are essential design tools to make engineered
products. This paper provides a dynamical model of a class of tensegrity structures
that is appropriate for space structures.
This paper derives the nonlinear equations of motion for space structures that can
be deployed or held to precise shape by feedback control, although control is beyond
the scope of this paper. For engineering purposes, more precise definitions of
tensegrity are needed.
One can imagine a truss as a structure whose compressive members are all
connected with ball joints so that no torques can be transmitted. Of course, tension
members connected to compressive members do not transmit torques, so that our
truss is composed of members experiencing no moments. The following definitions
are useful.
Definition 2.1. A given configuration of a structure is in a stable equilibrium if, in the
absence of external forces, an arbitrarily small initial deformation returns to the
given configuration.
Definition 2.2. A tensegrity structure is a stable system of axially loaded-members.
260
Definition 2.3. A stable structure is said to be a Class 1 tensegrity structure if the

members in tension form a continuous network, and the members in compression
form a discontinuous set of members.
Definition 2.4. A stable structure is said to be a Class 2 tensegrity structure if the
members in tension form a continuous set of members, and there are at most two
members in compression connected to each node.
Fig. 4 illustrates a Class 1 and a Class 2 tensegrity structure.
Consider the topology of structural members given in Fig. 5, where thick lines
indicate rigid rods which take compressive loads and the thin lines represent tendons.
This is a Class 1 tensegrity structure.
Definition 2.5. Let the topology of Fig. 5 describe a 3-dimensional structure by
connecting points A to A; B to B; C to C; . . . ; I to I. This constitutes a Class 1
Tensegrity Shell if there exists a set of tensions in all tendons tabg a 1 ! 10; b
1 ! n; g 1 ! m such that the structure is in a stable equilibrium.
2.1. A typical element
The axial members in Fig. 5 illustrate only the pattern of member connections, and
not the actual loaded configuration. The purpose of this section is two-fold: (i) to
define a typical element which can be repeated to generate all elements, (ii) to
define rules of closure that will generate a shell type of structure.
Consider the members which make the typical ij element where i 1; 2; . . . ; n
indexes the element to the left, and j 1; 2; . . . ; m indexes the element up the page in
Fig. 5. We will describe the axial elements by vectors. That is, the vectors describing
the ij element, are t1ij ; t2ij ; . . . ; t10ij and r1ij ; r2ij , where, within the ij element, taij is a
vector whose tail is fixed at the specified end of tendon number a, and the head of the
vector is fixed at the other end of tendon number a as shown in Fig. 6 where a
1; 2; . . . ; 10: The ij element has two compressive members which we will call rods,
shaded in Fig. 6. Within the ij element the vector r1ij lies along the rod r1ij and the
vector r2ij lies along the rod r2ij . The first goal of this paper is to derive the equations
of motion for the dynamics of the two rods in the ij element. The second goal is to
write the dynamics for the entire system composed of nm elements.
Fig. 4. Class 1 and Class 2 tensegrity structures.
261
Fig. 5. Topology of an (8,4) Class 1 tensegrity shell.
Figs. 5 and 7 illustrate these closure rules for the case (n; m 8; 4 and
(n; m 3; 1.
Lemma. Consider the structure of Fig. 5 with elements defined by Fig. 6. A Class 2
tensegrity shell is formed by adding constraints such that for all i 1; 2; . . . ; n; and for
m > j > 1,
$t1ij t4ij 0;
t2ij t3ij 0;
t5ij t6ij 0;
t7ij t8ij 0:
2:1
This closes nodes n2ij and n1i1 j1 to a single node; and closes nodes n3i$1j and
n4i j$1 to a single node with ball joints. The nodes are closed outside the rod; so that
all tension elements are on the exterior of the tensegrity structure and the rods are in
the interior.
The point here is that a Class 2 shell can be obtained as a special case of the Class 1
shell, by imposing constraints (2.1). To create a tensegrity structure, not all tendons
262
Fig. 6. A typical ij element.
in Fig. 5 are necessary. The following definition eliminates tendons t9ij and
t10ij i 1 ! n; j 1 ! m.
Definition 2.6. Consider the shell of Figs. 5 and 6, which may be Class 1 or Class 2
depending on whether constraints (2.1) are applied. In the absence of dotted tendons
(labelled t9 and t10 ), this is called a Primal Tensegrity Shell. When all tendons t9 ; t10
are present in Fig. 5, it is called simply Class 1 or Class 2 tensegrity shell.
The remainder of this paper focuses on the general Class 1 shell of Figs. 5 and 6.
2.2. Rules of closure for the shell class
Each tendon exerts a positive force away from a node and f abg is the force exerted
by tendon tabg and ^f aij denotes the force vector acting on the node naij . All f aij forces
263
Fig. 7. Class 1 shell: n; m 3; 1.
are positive in the direction of the arrows in Fig. 6, where waij is the external applied
force at node naij ; a 1; 2; 3; 4. At the base, the rules of closure, from Figs. 5 and 6,
are
t9i1 $t1i1 ;
i 1; 2; . . . ; n;
2:2
t6i0 0;
2:3
t600 $t2n1 ;
2:4
t901 t9n1 $t1n1 ;
2:5
0 t10i$10 t5i0 t7i0 t7i$10 ;
i 1; 2; . . . ; n:
2:6
At the top, the closure rules are

t10im $t7im ;
2:7
t100m $t70m $t7nm ;
2:8
t2im1 0;
2:9
0 t1im1 t9im1 t3i1m1

t1i1m1 t2i1m1 :
At the closure of the circumference (where i 1):

t90j t9nj ;
t60 j$1 t6n j$1 ;
t80j t8nj ;
t70j t7nj ;
t70 j$1 t7n j$1 ;
t100 j$1 t10n j$1 :
From Figs. 5 and 6, when j 1, then
0 f 7i j$1 f 7i$1 j$1 f 5i j$1 f 10i$1j$1
and for j m where, in Figs. 5 and 6,
0 f 1im1 f 9im1 f 3i1m1 f 1i1m1 :
2:10
2:11
2:12
2:13
2:14
264
Nodes n11j ; n3nj ; n41j for j 1; 2; . . . ; m are involved in the longitudinal zipper that
closes the structure in circumference. The forces at these nodes are written explicitly
to illustrate the closure rules.
In Section 4, the rod dynamics will be expressed in terms of sums and dierences
of the nodal forces, so the forces acting on each node is presented in the following
form, convenient for later use. The definitions of the matrices Bi are found in
Appendix E.
The forces acting on the nodes can be written in vector form:
f Bd f d B0 f 0 W0 w;
2:15
where
2 !d 3
f1
6 d7
6 f2 7
6 7
f d 6 . 7;
6 . 7
4 . 5
f1
6 . 7
7
f6
4 .. 5;
fm
2
f 01
6
6
f 0 6 ...
4
f 0m
f dm
W0 BlockDiag' ' ' ; W1 ; W1 ; ' ' '(;

2
B3
6
6
6 B5
6
6
Bd 6
6 0
6
6 .
6 ..
4
B4
B6
..
!5
B
..
.
'''
..
..
'''
..
.
..
.
B6
!5
B
0
..
.
7
7
7
7
7
7
0 7;
7
7
B4 7
5
B8
3
w1
6 . 7
7
w6
4 .. 5;
wm
2
7
7
7;
5
B1
6
6
6
6
6
B0 6
6
6
6
6
4
B2
0
..
..
..
0
..
.
..
.
B7
..
.
''' '''
'''
..
.
..
.
..
.
0
0
..
.
7
7
7
7
7
7
0 7
7
7
B2 7
5
B7
and
3
f2
6 7
6 f3 7
6 7
6f 7
6 47
6 7
6 f6 7
d
7
f ij 6
6f 7 ;
6 77
6 7
6 f8 7
6 7
6f 7
4 95
f 10 ij
2
f 0ij
"
f5
f1
;
ij
w1
6w 7
6 27
wij 6 7 :
4 w3 5
w4
ij
Now that we have an expression for the forces, let us write the dynamics.
265
3. Dynamics of a 2 rod element

Any discussion of rigid-body dynamics should properly begin with some decision
on how the motion of each body is to be described. A common way to describe rigidbody orientation is to use three successive angular rotations to define the orientation
of three mutually orthogonal axes fixed in the body. The measure numbers of the
angular velocity of the body may then be expressed in terms of these angles and their
time derivatives.
This approach must be reconsidered when the body of interest is idealized as a rod.
The reason is that the concept of body fixed axes becomes ambiguous. Two
dierent sets of axes with a common axis along the rod can be considered equally
body fixed in the sense that all mass particles of the rod have zero velocity in both
sets. This remains true even if relative rotation is allowed along the common axis.
The angular velocity of the rod is also ill defined since the component of angular
velocity along the rod axis is arbitrary. For these reasons, we are motivated to seek a
kinematical description which avoids introducing body fixed reference frames and
angular velocity. This objective may be accomplished by describing the configuration
of the system in terms of vectors which locate only the end points of the rods. In this
case, no angles are used.
We will use the following notational conventions. Lower case, bold faced symbols
with an underline will indicate vector quantities with magnitude and direction in
three-dimensional space. These are the usual vector quantities we are familiar with
from elementary dynamics. The same bold face symbols without an underline will
indicate a matrix whose elements are scalars. Sometimes we will also need to
introduce matrices whose elements are vectors. These quantities will be indicated
with an upper case symbol that is both bold faced and underlined.
As an example of this notation, a position vector pi can be expressed as
2 3
pi1
!
"6 7
pi e1 e2 e3 4 pi2 5 Epi :
pi3
In this expression, pi is a column matrix whose elements are the measure numbers of
pi for the mutually orthogonal inertial unit vectors e1 , e2 , e3 . Similarly, we may
represent a force vector ^f i as
^f i E^f i :
Matrix notation will be used in most of the development to follow.
We now consider a single rod as shown in Fig. 8 with nodal forces ^f 1 and ^f 2
applied to the ends of the rod.
The following theorem will be fundamental to our development:
Theorem 3.1. Given a rigid rod of constant mass m and constant length L; the
governing equations may be described as
"
q Kq H~f;
3:1
266
Fig. 8. A single rigid rod.
where
q
"
q1
q2
"
#
p1 p2
;
p2 $ p1
#
^f 1 ^f 2
~f
;
^f 1 $ ^f 2
"
2
I
24 3
H
m 0
3 2 5;
~
q
L2 2
"
#
0
:
0 L$2 q_ T2 q_ 2 I3
0
The notation ~r denotes the skew symmetric matrix formed from the elements of r:
2 3
2
3
0
$r3 r2
r1
6
7
6
7
~r 4 r3
0
$r1 5; r 4 r2 5
$r2 r1
0
r3
and the square of this matrix is

2 2
r1 r2
$r2 $ r23
6
2
~r 4 r2 r1
$r21 $ r23
r3 r1
r3 r2
r1 r3
7
r2 r3 5:
$r21 $ r22
The matrix elements r1 ; r2 ; r3 ; q1 ; q2 ; q3 , etc. are to be interpreted as the

measure numbers of the corresponding vectors for an orthogonal set of inertially
fixed unit vectors e1 , e2 , e3 . Thus, using the convention introduced earlier,
r Er;
q Eq; etc:
The proof of Theorem 3.1 is given in Appendix A. This theorem will provide the
basis of our dynamic model for the shell class of tensegrity structures.
267
Now consider the dynamics of the 2-rod element of the Class 1 tensegrity shell in
Fig. 5. Here, we assume the lengths of the rods are constant. From Theorem 3.1 and
Appendix A, the motion equations for the ij unit can be described as follows. For
rod 1,
m1ij
"
q ^f 1ij ^f 2ij ;
2 1ij
m1ij
q2ij ) "
q2ij q2ij ) ^f 2ij $ ^f 1ij ;
6
q_ 2ij ' q_ 2ij q2ij ' "
q2ij 0;
3:2
q2ij ' q2ij L21ij ;

and for rod 2,
m2ij
"
q ^f 3ij ^f 4ij ;
2 3ij
m2ij
q4ij ) "
q4ij q4ij ) ^f 4ij $ ^f 3ij ;
6
q_ 4ij ' q_ 4ij q4ij ' "
q4ij 0;
3:3
q4ij ' q4ij L22ij ;
where the mass of the rod aij is maij and jjraij jj Laij . As before, we refer everything
to a common inertial reference frame (E). Hence,
2
3
2
3
2
3
2
3
q11ij
q21ij
q31ij
q41ij
4 6
4 6
4 6
4 6
7
7
7
7
q1ij 4 q12ij 5; q2ij 4 q22ij 5; q3ij 4 q32ij 5; q4ij 4 q42ij 5;
q13ij
q23ij
q33ij
q43ij
h
4
qij qT1ij ;
qT2ij ; qT3ij ; qT4ij
iT
and the force vectors appear in the form

2
3
2
I3
0
I3
2 6
2 6
7
3
;
H
H1ij
4
5
4
2ij
~
q2
m1ij 0
m2ij 0
L21ij 2ij
3
^f 1ij ^f 2ij
7
6
6^
^ 7
4 6 f 1ij $ f 2ij 7
f ij 6
7:
6 ^f 3ij ^f 4ij 7
5
4
^f 3ij $ ^f 4ij
2
0
3 2
~q
L22ij 4ij
7
5;
Hij
"
H1ij
0
#
0
;
H2ij
Using Theorem 3.1, the dynamics for the ij unit can be expressed as follows:
"
qij Xij qij Hij f ij ;
268
where
Xij
"
X1ij
0
#
0
;
X2ij
X1ij
"
#
0
0
;
_ T2ij q_ 2ij I3
0 L$2
1ij q
X2ij
"
0
0
0
_ T4ij q_ 4ij I3
L$2
2ij q
q qT11 ; . . . ; qTn1 ; qT12 ; . . . ; qTn2 ; . . . ; qT1m ; . . . ; qTnm (T :

The shell system dynamics are given by
"
q Kr q Hf;
3:4
where f is defined in (2.15) and

q qT11 ; . . . ; qTn1 ; qT12 ; . . . ; qTn2 ; . . . ; qT1m ; . . . ; qTnm (T ;
Kr BlockDiagX11 ; . . . ; Xn1 ; X12 ; . . . ; Xn2 ; . . . ; X1m ; . . . ; Xnm (;
H BlockDiagH11 ; . . . ; Hn1 ; H12 ; . . . ; Hn2 ; . . . ; H1m ; . . . ; Hnm (:
4. Choice of independent variables and coordinate transformations
The tendon vectors tabg are needed to express the forces. Hence, the
dynamical model will be completed by expressing the tendon forces, f, in terms
îj and qij can be
of variables q. From Figs. 6 and 9, it follows that vectors q
Fig. 9. Choice of independent variables.
269
described by
qij q11
i
X
k1
r1k1 $
i$1
X
k1
t1k1
j
X
k2
t1ik
j$1
X
k1
t5ik $ r1ij ;
4:1
îj qij r1ij t5ij $ r2ij :

q
4:2
To describe the geometry, we choose the independent vectors fr1ij ; r2ij ; t5ij , for
i 1; 2; . . . ; n; j 1; 2; . . . ; mg and fq11 ; t1ij , for i 1; 2; . . . ; n; j 1; 2; . . . ; m; and
i5n when j 1g.
This section writes the relationship between the q variables and the string and rod
vectors tabg and rbij . From Figs. 5 and 6, the position vectors from the origin of the
reference frame, E, to the nodal points, p1ij ; p2ij ; p3ij , and p4ij , can be described as
follows:
p1ij qij ;
p2ij qij r1ij ;
îj ;
p3ij q
îj r2ij :
p4ij q
4:3
We define
4
q1ij p2ij p1ij 2qij r1ij ;

4
q2ij p2ij $ p1ij r1ij ;

4
q1ij r2ij ;
q3ij p4ij p3ij 2^
4
q4ij p4ij $ p3ij r2ij :
4:4
Then,
2
q1
I3
6 7
6 $I
4 6 q2 7
6 3
qij 6 7 6
4 q3 5
4 0
q4
ij
p1
2I3
I3
I3
0
0
I3
$I3
6 7
6 0
07
7 6 p2 7
6
76 7 6
4 0
I 3 5 4 p3 5
I3
32
p4
ij
I3
I3
0
0
2I3
3 2
07
7
7
I3 5
I3
ij
6r 7
6 17
6 7 :
4q
^5
r2
ij
4:5
In shape control, we will later be interested in the p vector to describe all nodal points
of the structure. This relation is
p Pq;
P BlockDiag. . . ; P1 ; . . . ; P1 ; . . .(;
270
I3
6
4 6 $I3
P1 6
4 0
0
3$1
I3
I3
I3
0
0
I3
$I3
07
16
7
6 I3
7 6
24 0
I3 5
I3
$I3
I3
0
0
I3
I3
0 7
7
7:
$I3 5
4:6
I3
The equations of motion will be written in the q coordinates. Substitution of (4.1)

and (4.2) into (4.4) yields the relationship between q and the independent variables
t5 ; t1 ; r1 ; r2 as follows:
"
#
j
j$1
i
i$1
X
X
X
X
q1ij 2 q11
r1k1 $
t1k1
t1ik
t5ik $ r1ij ;
k1
k1
k2
k1
i
X
i$1
X
j
X
j
X
q2ij r1ij ;
"
q3ij 2 q11
k1
r1k1 $
k1
t1k1
k2
t1ik
k1
t5ik $ r2ij ;
q4ij r2ij :
4:7
To put (4.7) in a matrix form, define these matrices:

3
2
2
3
2
3
t1i$11
r1ij
q11
7
6
6r 7
6r 7
6 r1i1 7
6 2ij 7
6 111 7
7;
lij 6
7 for j 2; 3; . . . ; m; l11 6
7; li1 6
7
6 r
4 t5ij 5
4 r211 5
4 2i1 5
t1ij
t511
t5i1
for i 2; . . . ; n
and
l lT11 ; lT21 ; . . . ; lTn1 ; lT12 ; . . . ; lTn2 ; . . . ; lT1m ; . . . ; lTnm (T ;
2I3
I3
6 0
6
A 6
4 2I3
I3
2I3
0
$I3
$I3
6 I
6 3
C6
4 0
0
0
$I3
0
2I3
0
2I3
0 7
7
7;
2I3 5
0
$2I3
6 0
6
B6
4 $2I3
0 7
7
7;
2I3 5
I3
I3
I3
I3
2I3
0
$I3
0 7
7
7;
2I3 5
I3
2I3
2I3
6 0
6
D6
4 2I3
0
2I3
0
0
07
7
7;
05
$2I3
6 0
6
E6
4 $2I3
0
2I3
2I3
0
0
0
07
7
7;
05
2I3
2I3
60
6
J6
40
0
0
0
2I3
0 7
7
7;
2I3 5
2I3
2I3
6 0
6
F6
4 2I3
2I3
0
0
0
0 7
7
7;
2I3 5
$2I3
2I3
2I3
0
2I3
0
0
0 7
7
7:
2I3 5
6 0
6
G6
4 2I3
2I3
271
Then (4.7) can be written simply

q Ql;
4:8
where the 12nm ) 12nm matrix Q is composed of the 12 ) 12 matrices AH as

follows:
Q11
6
6
6 Q21
6
6
6Q
21
Q6
6
6 Q21
6
6Q
6 21
4
..
.
2
Q11
Q22
'''
..
.
Q32
Q22
Q32
Q32
..
.
Q32
Q32
..
.
6
6
6D B
6
6
6
6D E
6
6
6
6D E
6
6 . .
6 .. ..
4
D E
'''
..
.
'''
..
0
..
.
..
.
Q22
Q32
..
.
0
Q22
..
.
''' '''
..
E
..
.
B
..
.
'''
.
..
..
..
7
7
7
7
7
7
7;
7
7
7
7
7
5
39
0 >
>
7>
>
.. 7>
>
>
>
.7
>
7>
>
7
>
.. 7>
>
7
. =
7 n ) n blocks of 12 ) 12 matrices;
.. 7
>
7>
>
. 7>
>
7>
>
7>
>
7
>
0 5>
>
>
>
;
B
4:9
272
Q21
6
6
6D G
6
6
6
6D E
6
6
6
6D E
6
6 . .
6 .. ..
4
D E
''' ''' '''

..
.
..
G
.
.. ..
.
E
.
.. . . . .
.
.
.
E E E
39
>
>
>
7>
7>
>
7>
>
7>
>
>
7>
>
7>
7=
7 12n ) 12n matrix;
7>
7>
>
7>
>
7>
>
7>
>
7
>
0 5>
>
>
>
;
G
0
..
.
..
.
..
.
Q22 BlockDiag. . . ; C; . . . ; C(;

Q32 BlockDiag. . . ; J; . . . ; J(;
where each Qij is 12n ) 12n and there are m row blocks and m column blocks in Q.
Appendix B provides an explicit expression for the inverse matrix Q, which will be
needed later to express the tendon forces in terms of q.
Eq. (4.8) provides the relationship between the selected generalized coordinates and an independent set of the tendon and rod vectors forming l.
All remaining tendon vectors may be written as a linear combination of l. This
relation will now be established. The following equations are written by inspection of
Figs. 57 where
t1n1 qn1 r1n1 $ q11
and for i 1; 2; . . . ; n; j 1; 2; . . . ; m we have,
t2ij qij $ ^
qi j$1 r2i j$1
4:10
j > 1;
î$1j $ qij ;
t3ij q
î$1j ;
t4ij $t3ij r1ij qij r1ij $ q
t6ij qi1 j1 $ ^
qij r2ij j5m;
îj $ qi1 j1
t7ij q
j5m;
îj $r1ij $ t5ij r2ij ;

t8ij qij r1ij $ q
t9ij qi1j $ qij r1ij ;
î1j r2i1j $ q
îj :
t10ij q
For j 1 we replace t2ij with
t2i1 qi1 $ qi11 :
4:11
For j m we replace t6ij and t7ij with

î1m r2i1m $ ^
t6im q
qim r2im ;
îm $ ^
t7im q
qi1m r2i1m :
4
^nj , and i n i. Eq. (4.11) has the matrix form,

^0j q
where q0j qnj ; q
2
6
6
6
6
6
6
6
4
tdij 6
6
6
6
6
6
6
4
$I3
$I3
0
0
0
0
0
0
0
0
0
0
0
0
7
6
7
60
7
6
7
60
7
6
7
6
60
t6 7
7 6
7
60
t7 7
6
7
6
60
t8 7
7
6
7
60
t9 5
4
0
t10 ij
0
0
t2
t3
t4
I3
6
6 $I3
6
6 I
6 3
6
6 0
6
6 0
6
6
6 I3
6
6 $I
4 3
0
2
6
6 0
6
6 0
6
6
6 I3
6
6 $I
6 3
6
6 0
6
6 0
4
0
0
0
0
I3
0
0
$I3
0
I3
$I3
0
0
0
0
$I3
3
7
6
7
60
72 3
6
7 q
60
7
6
76 7
6
76 r1 7
60
76 7
6
74 q
60
7 ^5
6
7
6
7 r2 ij$1 6 0
7
6
7
60
5
4
0
0
0
I3
$I3
0
0
0
0
0
0
0
0
7
6
7
60
72 3 6
7 q
60
7
6
76 7 6
$I3 76 r1 7 6 0
76 7 6
^5 6
0 7
74 q
60
7
6
6
r
0 7
7 2 ij 6 0
7
6I
0 5
4 3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 0
0 I3
0
I3
0
0
I3
$I3
0
0
0
0
0
0
0
0
7
07
72 3
07
7 q
7
r1 7
0 76
7
76
;
6 7
7
^5
0 74 q
7
07
7 r2 i1j1
07
5
0
7
07
72 3
07
7 q
7
r1 7
0 76
7
76
6
7
4
5
^
q
07
7
7
07
7 r2 i$1j
07
5
0
7
7
72 3
7 q
7
76 7
76 r1 7
76 7
74 q
7 ^5
7
7 r2 i1j
7
7
5
273
274
3
2
t2
0
6
6 7
60
6 t3 7
6
6 7
60
6t 7
6
6 47
6
6 7
7
60
6
t
6
4
7 6
tdi1 6
60
6t 7
6
6 77
6
6 7
60
6 t8 7
6
6 7
60
6t 7
4
4 95
0
t10 i1
2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
0
0
0
I3
0
0
$I3
0
0
0
0
0
$I3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
I3
0
3
0 0
0 I3
2
2
t2
0 0
6
6 7
60 0
6 t3 7
6
6 7
60 0
6t 7
6
6 47
6
6 7
7
60 0
6
t
6
4
7 6
tdim 6
60 0
6t 7
6
6 77
6
6 7
60 0
6 t8 7
6
6 7
60 0
6t 7
4
4 95
0 0
t10 im
2
I3
6
6 $I3
6
6 I
6 3
6
6 0
6
6 0
6
6
6 I3
6
6 $I
4 3
0
0
0
I3
0
0
$I3
I3
I3
$I3
$I3
0
$I3
3
2
0
0
7
6
7
6
0
0
72 3
6
60
q
0 7
7
6
7
6
7
60
r
$I3 76
1
7
76
6 7 6
7
60
^5
0 74 q
6
7
6
7
0 7 r2 im 6
60
6I
0 7
5
4 3
0
0
0
0
0
0
I3
0
0
$I3
I3
I3
$I3
0
0 0
7
6
7
6 0
0 0
72 3
6
7 q
6 0
0 0
7
6
76 7
6
6 I3 0 0
0 76 r1 7
76 7
6
7
6 $I 0 0
^5
0 74 q
6 3
7
6
7
0 7 r2 i11 6
0 0
6 0
6 0
0 0
07
5
4
0
0 0
I3
3
2
$I3 $I3
0
7
6
60
0
0 7
72 3
6
60
q
0
0 7
7
6
76 7
6
60
0
0 76 r1 7
76 7
6
7
60
^5
0
0 74 q
6
7
6
7
0
0 7 r2 im$1 6
60
60
0
0 7
5
4
0
0
0
0
0
3
2
0
I3
7
6
6 $I3
07
72 3
6
7
6 I
07 q
6 3
76 7
6
6 0
0 76 r1 7
76 7
6
7
6 0
^5
0 74 q
6
7
6
6 I
r
07
2
i$11 6 3
7
7
6 $I
05
4 3
0
0
$I3
0
$I3
3
0
7
0 7
72 3
0 7
7 q
7
r1 7
$I3 76
7
76
6
7
4
5
^
q
0 7
7
7
0 7
7 r2 i1
0 7
5
0
7
07
72 3
07
7 q
7
r1 7
0 76
7
76
;
6
7
^5
07
74 q
7
07
7 r2 i12
07
5
0
3
0
0
0
7
0 I3 0 7
72 3
0 $I3 0 7
7 q
7
r1 7
0
0
0 76
7
76
6
7
^5
0
0
07
74 q
7
0
0
07
7 r2 i$1m
0
0
07
5
0
0
0
0
0
0
0
0 I3
0 $I3
0
0
0
0
I3
3
0
7
0 7
72 3
0 7
7 q
7
r1 7
I3 7 6
7
76
:
6
7
^5
$I3 7
74 q
7
0 7
7 r2 i1m
0 7
5
I3
4:12
275
Eq. (4.5) yields

3
21
q
2 I3
6r 7
6 0
6 17
6
6 7 6
4q
4 0
^5
2
r2
$12 I3
I3
0
0
1
2 I3
ij
3
0
0 7
7
7qij :
$12 I3 5
4:13
I3
Hence, (4.12) and (4.13) yield

2
6
6
6
6
6
6
6
4
tdij 6
6
6
6
6
6
6
4
t2
$I3
$I3
0
0
0
0
0
0
0
0
0
0
0
0
7
6
7
60
7
6
7
60
7
6
7
6
60
t6 7
1
7 6
7
26
t7 7
60
7
6
7
60
t8 7
6
7
60
t9 5
4
0
t10 ij
0
0
t3
t4
0
0
0
0
7
6
7
60
7
6
7
60
7
6
7
6
7
60
1
7qi j$1 6
7
26
7
60
7
6
7
60
7
6
7
60
5
4
0
0
0
I3
$I3
3
2
0
0
7
6
60
0 7
7
6
60
0 7
7
6
7
6
60
$I3 7
1
7qij 6
7
26
$I3 7
60
7
6
7
60
I3 7
6
7
6I
0 5
4 3
0
I3
I3
6
6 $I3
6
6 I
6 3
6
0
16
6
6
26 0
6
6 I3
6
6 $I
4 3
0
$I3
I3
0
0
I3
0
0
$I3
$I3
0
$I3
0
0
I3
0
0
0
7
07
7
07
7
7
07
7qi1j1 ;
07
7
7
07
7
07
5
0
6
6 0
6
6 0
6
6
1 6 I3
6
26
6 $I3
6
6 0
6
6 0
4
0
0
I3
$I3
I3
$I3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
$I3
0
0
I3
7
$I3 7
7
I3 7
7
7
0 7
7qi$1j
0 7
7
7
0 7
7
0 7
5
0
3
0
7
07
7
07
7
7
07
7qi1j
07
7
7
07
7
07
5
I3
276
3
2
t2
0
6
6 7
60
6 t3 7
6
6 7
60
6t 7
6
6 47
6
6 7
7
60
6
t
6
4
7 16
tdi1 6
6t 7
26
60
6 77
6
6 7
60
6 t8 7
6
6 7
60
6t 7
4
4 95
0
t10 i1
2
$I3
6
6 0
6
6 0
6
6
0
16
6
26
6 0
6
6 0
6
6 I
4 3
0
2
6
6
6
6
6
6
6
4
tdim 6
6
6
6
6
6
6
4
t2
0
0
0
I3
0
0
$I3
0
0
0
0
0
I3
0
0
0
0
0
0
0
0
$I3
0
0
I3
0
7
6
60
t3 7
7
6
60
t4 7
7
6
7
6
60
t6 7
1
7 6
7
26
t7 7
60
7
6
7
60
t8 7
6
7
60
t9 5
4
0
t10 im
I3
6
6 $I3
6
6 I
6 3
6
0
16
6
26
0
6
6
6 I3
6
6 $I
4 3
0
$I3
0
2
I3
$I3
0
I3
$I3
$I3
3
0 0
7
0 07
7
0 07
7
7
0 07
7qi12 ;
0 07
7
7
0 07
7
0 07
5
0 0
0
I3
0
0
$I3
0
0
0
0
0
7
6
60
0 7
7
6
7
60
0 7
6
7
6
60
$I3 7
1
7qim 6
60
2
$I3 7
7
6
7
6
60
I3 7
7
6
7
6I
0 5
4 3
0
I3
0
0
0
I3
0
0
$I3
0
$I3
0
0
0
$I3
7
6
60
0 7
7
6
60
0 7
7
6
7
6
60
0 7
1
7qim$1 6
7
26
0 7
60
7
6
7
60
0 7
6
7
60
0 5
4
0
0
I3
$I3
I3
I3
0
0
$I3
0
$I3
0
0
I3
0
$I3
0
0
$I3
I3
0
0
$I3
0
0
I3
$I3
I3
0
0
0
I3
3
2
0
0
7
6
6 0
07
7
6
6 0
07
7
6
7
6
6 I
07
1
7qi11 6 3
7
6 $I
2
07
6 3
7
6
6 0
07
7
6
7
6 0
05
4
0
0
$I3
3
2
0
I3
7
6
6 $I3
$I3 7
7
6
7
6 I
I3 7
6 3
7
6
6 0
0 7
7qi$11 16
7
26
0 7
6 0
7
6
6 I3
0 7
7
6
7
6 $I
0 5
4 3
0
0
I3
3
0
7
0 7
7
0 7
7
7
$I3 7
7qi1
$I3 7
7
7
I3 7
7
0 7
5
I3
7
$I3 7
7
I3 7
7
7
0 7
7qi$1m
0 7
7
7
0 7
7
0 7
5
0
7
0 7
7
0 7
7
7
I3 7
7qi1m :
$I3 7
7
7
0 7
7
0 7
5
I3
4:14
277
Also, from (4.10) and (4.12)
t1n1 $I3 ; 0;
6r 7
6 17
0; 0 (6 7 I3 ;
4q
^5
r2
t1n1 $12 I3 ;
1
2 I3 ;
E6 ; 0;
r2
11
1
2 I3 ;
3
q11
7
6
6 q21 7
7
6
E7 (6 . 7;
6 .. 7
5
4
qn1
2
. . . ; 0;
6r 7
6 17
0 (6 7 ;
4q
^5
I3 ; 0;
0; 0 (q11 12 I3 ;
E6 q11 E7 qn1
0;
n1
0 (qn1
E6 2 R3)12 ;
E7 2 R3)12 ;
R0 2 R3)12n :
t1n1 R0 q1 R0 ; 0 (q;
4:15
^ 4; E
! 4 ; E5 ,
With the obvious definitions of the 24 ) 12 matrices E1 ; E2 ; E3 ; E4 ; E
equations in (4.14) are written in the form, where q01 qn1 ; qn1j qij ,
^ 4 qi11 E5 qi12 ;
td E2 qi$11 E3 qi1 E
i1
tdij E1 qi j$1 E2 qi$1j E3 qij E4 qi1j E5 qi1 j1 ;

! 4 qi1m :
tdim E1 qim$1 E2 qi$1m E3 qim E
4:16
Now from (4.14) and (4.15), define

dT
dT
ld tdT
1n1 ; t11 ; t21 ; . . . ;
tdT
n1 j
tdT
12 ;
T
tdT
n (
dT
dT
tdT
1n1 ; t1 ; t2 ; . . . ;
. . . ; tdT
n2 j . . . ;
T
tdT
nm (
to get
ld Rq;
R0
6
6^
6 R11
6
6
6R
6 21
6
6
R6 0
6
6
6 ..
6 .
6
6 .
6 .
4 .
2
R 2 R24nm3)12nm ;
R12
'''
..
.
R11
R12
..
R21
..
.
R11
R12
R21
..
.
R11
..
.
'''
R21
'''
'''
q 2 R12nm ;
'''
.
..
..
..
0
..
.
..
.
..
.
ld 2 R24nm3 ;
4:17
7
7
7
7
7
7
7
7
7
7;
7
7
7
0 7
7
7
7
R12 5
! 11
R
Rij 2 R24n)12n ;
R0 2 R3)12n ;
278
R11
E3
6
6
6 E2
6
6
6
6 0
6
6 .
6 .
6 .
6
6
6 0
4
E4
Riik 0
E4
E3
E4
E2
..
.
E3
E2
..
.
'''
if k > 1;
R21 BlockDiag ' ' ' ;

R0 E6 ; 0;
E7
1
2 I3 ;
I3 ;
''';
''' '''
..
.
..
.
E4
E3
..
.
0
E4
..
.
E2
3
2
0
E2
7
6
.. 7
6
6 0
. 7
7
6
7
6 .
.. 7
6 .
6 .
. 7
7; R12 6
7
6 .
7
6 .
0 7
6 .
7
6
7
6
7
6 0
E4 5
4
E3
E5
E5
0
..
0
E5
..
.
..
'''
''' '''
..
.
.. ..
.
.
.. ..
.
.
.. ..
.
.
''' 0
Riki 0 if k > 1;
E1 ;
E1 ; ' ' ' (;
0; E7 (;
0; 0 (:
Ei 2 R24)12 ;
E6 12 $I3 ;
I3 ;
0
..
.
..
.
7
7
7
7
7
7
7
7;
7
7
0 7
7
7
E5 7
5
0
i 1 ! 5;
0; 0 (;
! 11 have the same structure as R11 except E4 is replaced by E

^ 4 and E
! 4,
^ 11 and R
R
respectively. Eq. (4.17) will be needed to express the tendon forces in terms of q.
Eqs. (4.8) and (4.17) yield the dependent vectors t1n1 ; t2 ; t3 ; t4 ; t6 ; t7 ; t9 ; t10 in
terms of the independent vectors t5 ; t1 ; r1 ; r2 Therefore,
ld RQl:
4:18
5. Tendon forces
Let the tendon forces be described by
taij
f aij Faij
:
jjtaij jj
5:1
For tensegrity structures with some slack strings, the magnitude of the force Faij can
be zero, for taut strings Faij > 0. Since tendons cannot compress, Faij cannot be
negative. Hence, the magnitude of the force is
Faij kaij jjtaij jj $ L!aij ;
5:2
where
4
kaij
0
k!aij > 0
if L!aij > jjtaij jj;

if L!aij 4jjtaij jj;
4
L!aij $uaij L0aij 50;
5:3
where L0aij > 0 is the rest length of tendon taij before any control is applied, and the
control is uaij , the change in the rest length. The control shortens or lengthens the
tendon, so uaij can be positive or negative, but L0aij > 0. So uaij must obey constraint
279
(5.3), and
uaij 4L0aij > 0:
5:4
Note that for t1n1 and for a 2; 3; 4; 6; 7; 8; 9; 10 the vectors taij appear in
the vector ld related to q from Eq. (4.7) by ld Rq, and for a 5; 1 the vectors taij
appear in the vector l related to q from Eq. (4.8), by l Q$1 q. Let Raij denote the
selected row of R associated with taij for aij 1n1 and for
a 2; 3; 4; 6; 7; 8; 9; 10. Let Raij also denote the selected row of Q$1 when
a 5; 1. Then,
Raij 2 R3)12nm ;
taij Raij q;
5:5
jjtaij jj2 qT RTaij Raij q:
5:6
From (5.1) and (5.5),

f aij $Kaij qq baij quaij ;
where
Kaij q kaij L0aij qT RTaij Raij q$1=2 $ 1Raij ;
4
baij q kaij qT RTaij Raij q$1=2 Raij q;
Kaij 2 R3)12nm ;
5:7
baij 2 R3)1 :
5:8
Hence,
2
or
6
6
6
6
6
6
6
d
f ij 6
6
6
6
6
6
6
4
f 2ij
f 3ij
f 4ij
f 6ij
f 7ij
f 8ij
f 9ij
f 10ij
7 6
7 6
7 6
7 6
7 6
7 6
7 6
7$6
7 6
7 6
7 6
7 6
7 6
7 6
5 4
K2ij
K3ij
K4ij
K6ij
K7ij
K8ij
K9ij
K10ij
2
b2ij
7
6
7
6
b3ij
7
6
7
6
b4ij
7
6
7
6
7
6
b6ij
7q 6
7
6
b7ij
7
6
7
6
7
6
b8ij
7
6
7
6
b9ij
5
4
f dij $Kdij q Pdij udij
"
f 5ij
f 1ij
"
K5ij
K1ij
b10ij
76
76
76
76
76
76
76
76
76
76
76
76
76
76
54
u2ij
u3ij
u4ij
u6ij
u7ij
u8ij
u9ij
u10ij
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
5:9
and
f 0ij
32
"
b5ij
b1ij
#"
u5ij
u1ij
280
or
f 0ij $K0ij q P0ij u0ij :
5:10
Now substitute (5.9) and (5.10) into
6
6
6
6
!f d 6
1
6
6
6
4
f 1n1
7
6
6
f d11 7
7
6
6
d 7
f 21 7 $6
7
6
6
.. 7
7
6
. 5
4
d
f n1
K1n1
7
6
6
Kd11 7
7
6
6
d 7
K21 7q 6
7
6
6
.. 7
7
6
. 5
4
d
Kn1
d
! dq P
! d!
$K
1
1 u1
2 d 3
2 d 3
2 d
f 12
K12
P12
6 d 7
6 d 7
6
6 f 22 7
6 K22 7
6
7
6 7
6
6
f d2 6 . 7 $6 . 7q 6
6 . 7
6 . 7
6
4 . 5
4 . 5
4
f dn2
32
P1n1
Pd11
Pd21
..
.
Pdn1
Pd22
Kdn2
Hence, in general,
32
ud12
76
76
76
76
76
76
76
76
54
u1n1
7
ud11 7
7
7
ud21 7
7
.. 7
. 7
5
d
un1
76 d 7
76 u22 7
76
7
76 . 7 $Kd2 q2 Pd2 ud2 :
..
76 .. 7
.
54
5
d
d
un2
Pn2
f dj $Kdj q Pdj udj

or by defining
2 !d 3
K1
6 d7
6 K2 7
7
6
Kd 6 . 7;
6 .. 7
5
4
Kdm
2 !d 3
P1
6 d7
6 P2 7
7
6
Pd 6 . 7;
6 .. 7
5
4
5:11
Pdm
f d $Kd q Pd ud :
Likewise for f 0ij forces (5.10),
2
f 011
K011
6 0 7
6 0 7
6
6 f 21 7
6 K21 7
6
7
7
6
6
6
f 01 6 . 7 $6 . 7q 6
6 .. 7
6 .. 7
6
5
4 5
4
4
f 0n1
K0n1
P011
P021
32
u011
76 0 7
76 u21 7
76
7
76 . 7;
..
7
6
. 7
.
54 . 5
u0n1
P0n1
f 01j
K01j
P01j
6 0 7
6 0 7
6
6f 7
6K 7
6
2j 7
2j 7
6
6
6
7 $6
7q 6
f 0j 6
6 .. 7
6 .. 7
6
6 . 7
6 . 7
6
5
4 5
4
4
0
0
Knj
f nj
P02j
281
32 3
u0
76 1j 7
76 u0 7
76 2j 7
76 7;
76 .. 7
..
74 . 5
.
5
0
u0nj
Pnj
f 0j $K0j q P0j u0j ;

f 0 $K0 q P0 u0 :
5:12
Substituting (5.11) and (5.12) into (E.21) yields

f $Bd Kd B0 K0 q Bd Pd ud B0 P0 u0 W0 w;
5:13
which is written simply as

~ Bu
~ W0 w
f $Kq
5:14
by defining,
4
~
Bd Kd B0 K0 ;
K
4
~
Bd Pd ; B0 P0 (;
B
!d
B3 P
1
B4 Pd2
6
6 !d
6 B5 P1
6
6
6
6 0
Bd Pd 6
6 .
6 .
6 .
6
6 .
6 ..
4
0
B6 Pd2
B4 Pd3
! 5 Pd
B
2
..
.
B6 Pd3
'''
! 5 Pd
B
3
..
.
'''
'''
..
.
..
.
..
.
..
.
0
'''
..
..
..
.
d
!
B5 Pm$1
0
..
.
..
.
7
7
7
7
7
7
7
7;
7
7
0 7
7
7
d 7
B4 Pm 5
B8 Pdm
282
B1 P01
6
6
6 0
6
6 .
B0 P0 6
6 ..
6
6 .
6 ..
4
0
B2 P02
B7 P02
..
.
B2 P03
B7 P02
..
.
'''
'''
3
'''
0
.. 7
..
7
. 7
.
7
7
..
7
.
0 7;
7
7
..
0 7
. B2 Pm 5
0
B7 P0m
! d B1 K0 B4 Kd B2 K0
B3 K
1
1
2
2
7
6
! d B6 Kd B4 Kd B7 K0 B2 K0
7
6
B5 K
1
2
3
2
3
7
6
7
6
d
d
d
0
0
! 5 K B6 K B4 K B7 K B2 K
7
6
B
2
3
4
3
4
7
6
7
6
d
d
d
0
0
! 5 K B6 K B4 K B7 K B2 K
d d
0 0
B
~
7; 5:15
6
KB K B K 6
3
4
5
4
5
7
7
6
..
7
6
.
7
6
7
6
6B
! 5 Kd B6 Kd B4 Kd B7 K0 B2 K0 7
m$2
m$1
m
m$1
m5
4
! 5 Kd B6 Kd B7 K0
B
m$1
m
m
2
!
ud1
6 d7
6 u2 7
6 7
6 ud 7
6 37
6 7
6 ud 7
6 47
6 7
6 .. 7
6 . 7
6 7
6 d7
6u 7
~
u 6 m 7;
6 u0 7
6 17
6 07
6 u2 7
6 7
6 u0 7
6 37
6 07
6 u4 7
6 7
6 . 7
6 . 7
4 . 5
u0m
^
ud1
6 d7
6 u2 7
6 7
6 ud 7
6 37
6 7
6 ud 7
6 47
6 7
6 .. 7
6 . 7
6 7
6 d7
6^
u 7
u 6 m 7:
6 u0 7
6 17
6 07
6 u2 7
6 7
6 u0 7
6 37
6 07
6 u4 7
6 7
6 . 7
6 . 7
4 . 5
u0m
5:16
~ in (5.16), u1n1 appears twice (for notational convenience u1n1 appears in

In vector u
!
ud1 and in u01 ). From the rules of closure, t9i1 $t1i1 and t7im $t10im ; i 1; 2; . . . ; n,
but t1i1 ; t7im ; t9i1 ; t10im all appear in (5.16). Hence, the rules of closure leave only
n10m $ 2 tendons in the structure, but (5.16) contains 10nm 1 tendons. To
eliminate the redundant variables in (5.16) define ~u Tu, where u is the independent
set u 2 Rn10m$2 , and ~
u 2 R10nm1 is given by (5.16). We choose to keep t7im in u and
delete t10im by setting t10im $t7im . We choose to keep t1i1 and delete t9i1 by setting
283
t9i1 $t1i1 ; i 1; 2; . . . ; n. This requires new definitions of certain subvectors as

follows in (5.19), (5.20). The vector u is now defined in (5.16). We have reduced the ~u
vector by 2n 1 scalars to u. The T matrix is formed by the following blocks,
0
BT
B 1
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
T B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
@
where
0
S
..
..
S
I8
..
.
I8
T2
..
.
T2
I2
..
I2
..
I6
B
S@
.
I2
06)1
01)6
.
I2
B
T1 B
@ 01)6
T2
T1
2 R10nm1)n10m$2 ;
0
''' 0 1
0
1
5:17
C
C 2 R8)7 ;
A
I7
0 0 $1
1
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
A
06)2 C
8)2
A2R :
0
$1
0
0
0 0
2 R8)7 ;
5:18
284
There are n blocks labelled T1 ; nm $ 2 blocks labelled I8 (for m42 no I8 blocks

needed, see Appendix D), n blocks labelled T2 ; nm blocks labelled I2 blocks, and n
blocks labelled S.
The ud1 block becomes
2
^
ud11
6 d 7
6^
u 7
6 21 7
7
6
4
^
ud 7
^
ud1 6
6 31 7;
6 . 7
6 .. 7
5
4
^
udn1
2
6
6
6
6
6
6
d
î1 6
u
6
6
6
6
4
The udm block becomes
ud1
^
udim
7
u3n1 7
7
u4n1 7
7
7
u6n1 7 2 R7)1 ;
7
u7n1 7
7
7
u8n1 5
u10n1
^
ud1m
6 d 7
6^
u 7
6 2m 7
7
6
4
^
ud 7
^
udm 6
6 3m 7;
6 . 7
6 .. 7
5
4
^
udnm
2
u2n1
u2nm
i 1; 2; 3; . . . n; j 1:
5:19
7
6
6 u3nm 7
7
6
6 u4nm 7
7
6
7
6
6 u6nm 7 2 R7)1 ;
7
6
6 u7nm 7
7
6
7
6
4 u8nm 5
i 1; 2; 3; . . . ; n; j m:
5:20
u9nm
!
ud1
The block is the block with the first element u1n1 removed, since it is included in
u0n1 . From (3.1) and (5.14)
"
_ Kp qq Bqu Dqw;
q Kr q
5:21
where
~
Kp HqKq;
~
B HqBqT;
D HqW0 :
The nodal points of the structure are located by the vector p. Suppose that a
selected set of nodal points are chosen as outputs of interest. Then
yp Cp CPq;
5:22
285
where P is defined by (4.6). The length of tendon vector taij Raij q is given from
(5.6). Therefore, the output vector yl describing all tendon lengths, is
2
3
..
.
6
7
6
7
yl 6 yaij 7; yaij qT RTaij Raij q1=2 :
4
5
..
.
Another output of interest might be tension, so from (5.2) and (5.6)
3
..
6 . 7
6
7
yf 6 Faij 7;
4
5
..
.
2
Faij kaij yaij $ L!aij :
The static equilibria can be studied from the equations

Kp qq Bqu Dqw;
yp CPq:
5:23
Of course, one way to generate equilibria is by simulation from arbitrary initial

conditions and record the steady-state value of q. The exhaustive definitive study of
the stable equilibria follows in a separate paper [26].
Damping strategies for controlled tensegrity structures is a subject of further
research. The example case given in Appendix D was coded in Matlab and
simulated. Artificial critical damping was included in the simulation below. The
simulation does not include external disturbances or control inputs. All nodes of the
structure were placed symmetrically around the surface of a cylinder, as seen in
Fig. 10. Spring constants and natural rest lengths were specified equally for all
Fig. 10. Initial conditions with nodal points on cylinder surface.
286
Fig. 11. Steady-state equilibrium.
Fig. 12. Tendon dynamics.
tendons in the structure. One would expect the structure to collapse in on itself with
this given initial condition. A plot of steady-state equilibrium is given in Fig. 11 and
string lengths in Fig. 12.
6. Conclusion
This paper develops the exact nonlinear equations for a Class 1 tensegrity shell,
having nm rigid rods and n10m $ 2 tendons, subject to the assumption that the
tendons are linear-elastic, the rods are rigid rods of constant length. The equations
are described in terms of 6nm degrees of freedom, and the accelerations are given
explicitly. Hence, no inversion of the mass matrix is required. For large systems this
greatly improves accuracy of simulations.
287
Tensegrity systems of four classes are characterized by these models. Class 2

includes rods that are in contact at nodal points, with a ball joint, transmitting no
torques. In Class 1, the rods do not touch and a stable equilibrium must be achieved
by pretension in the tendons. The Primal Shell Class contains the minimum number
of tendons 8nm for which stability is possible.
Tensegrity structures oer some potential advantages over classical structural
systems composed of continua (such as columns, beams, plates, and shells): the
overall structure can bend but all elements of the structure experience only axial
loads, so no member bending, the absence of bending in the members promises more
precise models (and hopefully more precise control). Prestress allows members to be
uni-directionally loaded, meaning that no member experiences reversal in the
direction of the load carried by the member. This eliminates a host of nonlinear
problems known to create diculties in control (hysterisis, deadzones, friction).
Acknowledgements
The authors recognize the valuable eorts of T. Yamashita in the first draft of this
paper.
Appendix A. Proof of Theorem 3.1

Refer to Fig. 8 and define
q1 p2 p1 ;
q 2 p2 $ p1
using the vectors p1 and p2 which locate the end points of the rod. The rod mass
center is located by the vector,
pc 12 q1 :
A:1
Hence, the translation equation of motion for the mass center of the rod is
m
m"
A:2
pc "
q ^f 1 ^f 2 ;
2 1
where a dot over a vector is a time derivative with respect to the inertial reference
frame. A vector p locating a mass element, dm, along the centerline of the rod
(Fig. 13) is
x
A:3
p p1 rp2 $ p1 12 q1 r $ 12q2 ; 04r41; r
L
and the velocity of the mass dm, v, is
v p_ 12 q_ 1 r $ 12q_ 2 :
The angular momentum for the rod about the mass center, hc , is
Z
hc
p $ pc ) p_ dm;
m
A:4
A:5
288
Fig. 13. A rigid bar of the length L and mass m.
where the mass dm can be described using dx L dr as

%m&
dm
L dr m dr:
L
Hence, (A.2) can be re-written as follows:
Z 1
_ dr;
hc
p $ pc ) pm
0
A:6
A:7
where (A.1) and (A.3) yield

p $ pc r $ 12q2 :
A:8
(A.4)(A.8) yield
$
#
$ (
'
Z 1#
1
1
1
q_ 1 r $ q_ 2 dr
hc m
r $ q2 )
2
2
2
0
( Z
)
$
$
Z 1#
1 #
1
1
1 2
r$
r$
dr
dr q_ 2
mq2 ) q_ 1
2
2
0 2
0
0
1
" #
)
*1
$ 3 #1
1
1
1
1
1
r2 $ r q_ 1
r$
mq2 ) @
q_ 2 A
2 2
2 0
3
2
0
m
q2 ) q_ 2 :
12
The applied torque about the mass center, tc , is
tc 1 q ) ^f 2 $ ^f 1 :
2
Then, substituting hc and tc from (A.9) into Eulers equations, we obtain

h_ c tc
A:9
289
or
m
q2
h_ c q_ 2 ) q_ 2 q2 ) "
12
m
1
q2 q2 ) ^f 2 $ ^f 1 :
q2 ) "
12
2
Hence, (A.2) and (A.10) yield the motion equations for the rod:
m
"
q ^f 1 ^f 2 ;
2 1
m
"2 q2 ) ^f 2 $ ^f 1 :
q ) q
6 2
A:10
A:11
We have assumed that the rod length L is constant. Hence, the following constraints
for q2 hold:
q2 ' q2 L2 ;
d
q ' q q_ 2 ' q2 q2 ' q_ 2 2q2 ' q_ 2 0;
dt 2 2
d
q ' q_ q_ 2 ' q_ 2 q2 ' "
q2 0:
dt 2 2
Collecting (A.11) and the constraint equations we have
m
"
q ^f 1 ^f 2 ;
2 1
m
q ) "
q2 q2 ) ^f 2 $ ^f 1 ;
6 2
q_ 2 ' q_ 2 q2 ' "
q2 0;
q2 ' q2 L2 :
A:12
We now develop the matrix version of (A.12). Recall that

qi Eqi ;
^f i E^f i :
Also note that ET ' E 3 ) 3 identity. After some manipulation, (A.12) can be
written as
m
"
q ^f 1 ^f 2 ;
2 1
m
~
q "
q ~
q2 ^f 2 $ ^f 1 ;
6 2 2
q2 $q_ T2 q_ 2 ;
qT2 "
qT2 q2 L2 :
A:13
Introduce scaled force vectors by dividing the applied forces by m and mL

2
4
g1 ^f 1 ^f 2 ;
m
6
4
g2 ^f 1 $ ^f 2
:
mL2
290
Then, (A.13) can be re-written as

"
q1 g1 ;
~
q2 $g2 L2 ;
q2 "
q2 ~
q2 $q_ T2 q_ 2 ;
qT2 "
qT2 q2 L2 :
A:14
Solving for "

q requires,
" # 2 "
# " #
~
~
q2
0
q2
"
q2
$
g2 L2 :
T
T_
_
$q2 q2
q2
0
A:15
Lemma. For any vector q; such that qT q L2 ;

" #T " #
~
~
q
q
L 2 I3 :
T
q
qT
Proof.
2
6
4 $q3
q2
q3
0
$q1
$q2
q1
0
2
3 0
q1 6
76 q3
q2 56
4 $q2
q3
q1
$q3
0
q1
q2
q2
$q1 7
7
7 L2 I3 :
0 5
q3
Since the coecient of "

q2 in (A.15) has linearly independent columns by virtue of the
lemma, the unique solution for "
q2 is
" # "
# " #
!
~
~
q2
0
q2
2
"
q2
$
A:16
g2 L ;
$q_ T2 q_ 2
qT2
0
where the pseudoinverse is uniquely given by
" # 0" #T " #1$1 " #T
" #T
~
~
~
~q2
~
q2
q2
q2
q2
$2
A
@ T
L
:
T
T
T
q2
q2
q2
q2
qT2
It is easily verified that the existence condition for "q2 in (A.15) is satisfied since
" #" # ! "
# " #
!
~
~
~
q2
q2
0
q2
2
I$ T
$
g2 L 0:
$q_ T2 q_ 2
q2
qT2
0
Hence, (A.16) yields
"
q2 $
q_ T2 q_ 2
q ~
q22 g2 :
L2 2
A:17
Bringing the first equation of (A.14) together with (A.17) leads to

3" # "
" # 20
#" #
0
"
q1
I3 0
g1
q1
5
T_
4
:
_
q
q
2
2 2
"2
q
0 ~
q2
q2
g2
I
0
3
L2
Recalling the definition of g1 and g2 we obtain
3" #
2
3" #
" # 20
0
I3
0
"
q1
q1
f
2
5
5 1 ;
T_
4
4
3
_
q
q
2
2
2
m 0
"
q2
q2
f2
~q
I3
0
L2 2
L2
where we clarify
" # "
#
^f 1 ^f 2
f1
:
^f 1 $ ^f 2
f2
291
A:18
A:19
Eq. (A.19) is identical to (3.1), so this completes the proof of Theorem 3.1.
&
Example 1. In this example we derive the dynamics of a single rod, as shown in Fig. 13.
"
# "
#
p11 p21
q11
q1 p1 p2
;
p12 p22
q12
"
# "
#
p21 $ p11
q21
:
q2 p2 $ p1
p22 $ p12
q22
The generalized forces are now defined as
"
#
"
#
2 ^
2 f11 f21
6 ^
6 f21 $ f11
^
^
f 2 $ f 1
; g2
:
g1 f 1 f 2
m
m f12 f22
mL2
mL2 f22 $ f12
From (A.16),
2
2
3
0
q"11
6
6 q" 7 6 0
6 12 7 6
6
76
4 q"21 5 6 0
4
q"22
0
0
0
0
0
q_ 2 q_ 2
$ 21L2 22
3
2
3 2
q
1
7 11
76 q 7 6 0
76 12 7 6
76
76
0 74 q21 5 4 0
5
q_ 2 q_ 2
q22
0
$ 21 22
0
0
L2
0
0
q222
$q21 q22
3
" #
7 g
0
7 1
:
7
$q21 q22 5 g2
q221
Example 2. Using the formulation developed in Sections 35 we derive the dynamics

of a planar tensegrity. The rules of closure become
t5 $t4 ;
t8 t1 ;
t7 $t2 ;
t6 $t3 :
292
We define the independent vectors l0 and ld :

2 3
2 3
q
t1
6r 7
6 17
6 7
0
d
l 6 7 ; l 4 t2 5 :
4 r2 5
t3
t5
The nodal forces are

3
2
2
3
^f 1 ^f 2
f 3 $ f 2 w1 f 5 $ f 1 w2
7
6
6 ^f 1 $ ^f 2 7 6 f $ f w $ f $ f w 7
2
1
5
1
2 7
7 6 3
!f 6
76
6
7:
6 ^f 3 ^f 4 7 4 f 1 f 2 w3 $f 3 $ f 5 w4 5
5
4
f 1 f 2 w3 $ $f 3 $ f 5 w4
^f 3 $ ^f 4
We can write
2
3
2
I2
$I2
6 $I 7
6 I
27 0
6 2
!f 6
6
7f 6
4 $I2 5
4 I2
I2
I2
where
f 0 f 5 (;
f1
6 7
f d 4 f 2 5;
f3
$I2
$I2
I2
I2
3
2
I2
I2
7
6
I2 7 d 6 I2
7f 6
40
$I2 5
I2
I2
$I2
0
0
I2
I2
3
0
0 7
7
7w;
I2 5
$I2
3
w1
6w 7
6 27
w 6 7:
4 w3 5
2
w4
Or, with the obvious definitions for B0 , Bd , and W1 , in matrix notation:

!f B0 f 0 Bd f d W1 w:
The nodal vectors are defined as follows:
p1 q;
p2 q r 1 ;
^;
p3 q
^ r2
p4 q
and
^ q r1 t 5 $ r2 :
q
We define
4
q1 p2 p1 2q r1 ;
A:20
293
q2 p2 $ p1 r 1 ;
4
q3 p4 p3 2^
q r2 2q r1 t5 $ r2 ;
4
q4 p4 $ p3 r 2 :
The relation between q and p can be written as follows:
2 3 2
32 3
q1
I2 I2
p1
0
0
6 q 7 6 $I I
7
6
0
0 76 p2 7
6 27 6 2 2
7
q6 76
76 7
4 q3 5 4 0
0
I2 I2 54 p3 5
0
$I2
3 2
2I2
q1
6q 7 6 0
6 27 6
q6 76
4 q3 5 4 2I2
I2
I2
0
0
2I2
0
$I2
I2
q4
I2
p4
and
q4
32 3
q
0
6
7
0 7 6 r1 7
7
76 7 Ql0 :
2I2 54 r2 5
0
A:21
t5
We can now write the dependent variables ld in terms of independent variables l0 .

From (4.10) and (4.11):
t1 q r1 $ q r1 t5 $ r2 ;
t2 q $ q r1 t5 $ r2 :
By inspection of Fig. 14, (4.10) and (4.11) reduce to

t1 r2 $ t5 ;
t2 $r1 r2 $ t5 ;
t3 r1 t5
A:22
Fig. 14. A planar tensegrity.
294
or,
2
0
t1
6 7 6
l 4 t2 5 4 0
t3
0
d
I2
$I2
I2
I2
0
Eq. (A.21) yields

2 3
2
q
I2
6 r 7 16 0
6 17
6
l0 6 7 6
4 r2 5 24 0
$I2
$I2
t5
I2
16
d
l 4 I2
2
$I2
I2
$I2
I2
0
0
$I2
I2
$I2
$I2
I2
2I2
0
Hence,
2 3
3 q
$I2 6 7
76 r1 7
$I2 56 7:
4 r2 5
I2
t5
0
32
q1
6 7
0 7
76 q2 7
76 7 Q$1 q:
2I2 54 q3 5
I2
q4
3
I2
7
I2 5q Rq:
I2
We can now write out the tendon forces as follows:

2 3
2 3
2
32 3
K1
b1
u1
f1
6 7
6 7
6
76 7
f d 4 f 2 5 $4 K2 5q 4
b2
5 4 u2 5
f3
K3
b3
u3
or
f d $Kd q Pd ud
and
f 0 f 5 ( $K5 (q b5 (u5 (
or
f 0 $K0 q P0 u0
using the same definitions for K and b as found in (5.7) and (5.8), simply removing
the ij element indices. Substitution into (A.20) yields
!f B0 $K0 q P0 u0 Bd $Kd q Pd ud
$B0 K0 Bd Kd q B0 P0 u0 Bd Pd ud :
With the matrices derived in this section, we can express the dynamics in the form of (3.4)
"
q Kr Kp q Bu Dw;
K r X1 ;
~
Kp HK;
295
~
B HB;
D HW0 H1 W1 ;
where
2
I2
6
0
26
6
H H1 6
m6 0
4
0
2
0
3 ~2
q
L2 2
0
0
I2
0
$2 _ T _
L1 q2 q2 I2
60
6
K r X1 6
40
3
0
7
0 7
7
7;
0 7
5
3 ~2
q
2
4
L
0
0
7
0
7
7;
5
0
$2 _ T _
L2 q4 q4 I2
4
~
K
Bd Kd B0 K0 ;
4
~
B
Bd Pd ; B0 P0 (
and
~
q22
"
$q222
q21 q22
$q221
q21 q22
q_ T2 q_ 2 q_ 221 q_ 222 ;
~
q24
"
$q242
q41 q42
q41 q42
$q241
q_ T4 q_ 4 q_ 241 q_ 242 :
Appendix B. Algebraic inversion of the Q matrix

This appendix
form:
2
Q11
6
6 Q21
6
Q6
6 Q21
6
4 Q21
Q21
will algebraically invert a 5 ) 5 block Q matrix. Given Q in the

0
Q22
Q32
0
Q22
0
0
0
0
Q32
Q32
Q32
Q32
Q22
Q32
we define x and y matrices so that

Qx y;
7
7
7
7;
7
7
0 5
Q22
B:1
B:2
296
where
2
x1
6 7
6 x2 7
6 7
7
x6
6 x3 7;
6 7
4 x4 5
x5
y1
6 7
6 y2 7
6 7
7
y6
6 y3 7:
6 7
4 y4 5
y5
B:3
Solving (B.2) for x we obtain

x Q$1 y:
B:4
Substituting (B.1) and (B.3) into (B.2) and carrying out the matrix operations we
obtain
Q11 x1 y1 ;
Q21 x1 Q22 x2 y2 ;
Q21 x1 Q32 x2 Q22 x3 y3 ;

Q21 x1 Q32 x2 Q32 x3 Q22 x4 y4 ;
Q21 x1 Q32 x2 Q32 x3 Q32 x4 Q22 x5 y5 :
Solving this system of equations for x will give us the desired Q

each equation for x we have
$1
B:5
matrix. Solving
x1 Q$1
11 y1 ;
x2 Q$1
22 $Q21 x1 y2 ;
x3 Q$1
22 $Q21 x1 $ Q32 x2 y3 ;
x4 Q$1
22 $Q21 x1 $ Q32 x2 $ Q32 x3 y4 ;
x5 Q$1
22 $Q21 x1 $ Q32 x2 $ Q32 x3 $ Q32 x4 y5 :
B:6
Elimination of x on the right-hand side of (B.6) by substitution yields

x1 K11 y1 ;
x2 K21 y1 K22 y2 ;
x3 K31 y1 K32 y2 K33 y3 ;

x4 K41 y1 K42 y2 K43 y3 K44 y4 ;
x5 K51 y1 K52 y2 K53 y3 K54 y4 K55 y5 :
Or, in matrix form,

2
K11
6
6 K21
6
Q$1 6
6 K31
6
4 K41
K51
0
K22
0
0
0
0
K32
K42
K33
K43
0
K44
K52
K53
K54
3
0
7
0 7
7
0 7
7;
7
0 5
K55
B:7
B:8
297
where we have used the notation Krow;column with the following definitions:
K11 Q$1
11 ;
$1
$1
K21 $Q$1
22 Q21 Q11 $Q22 Q21 K11 $K22 Q21 K11 ;
$1
$1
$1
$1
K31 Q$1
22 Q32 Q22 Q21 Q11 $ Q22 Q21 Q11 ;
$1
$1
$1
$1
$1
$1
$1
$1
K41 $Q$1
22 Q32 Q22 Q32 Q22 Q21 Q11 2Q22 Q32 Q22 Q21 Q11 $ Q22 Q21 Q11 ;
$1
$1
$1
$1
$1
$1
K51 $ 3Q$1
22 Q32 Q22 Q32 Q22 Q21 Q11 3Q22 Q32 Q22 Q21 Q11
$1
$1
$1
$1
$1
$1
Q$1
22 Q32 Q22 Q32 Q22 Q32 Q22 Q21 Q11 $ Q22 Q21 Q11 ;
K22 K33 K44 K55 Q$1

22 ;
$1
K32 K43 K54 $Q$1
22 Q32 Q22 ;
$1
$1
$1
$1
K42 K53 Q$1
22 Q32 Q22 Q32 Q22 $ Q22 Q32 Q22 ;
$1
$1
$1
$1
$1
$1
$1
$1
K52 $Q$1
22 Q32 Q22 Q32 Q22 Q32 Q22 2Q22 Q32 Q22 Q32 Q22 $ Q22 Q32 Q22 :
Using repeated patterns the inverse may be computed by

3
2
K11
7
6 K
K22
7
6
21
7
6
7
6 HK21
K
K
32
22
7
6
7
6 2
7
6 H K21 HK32
K
K
$1
32
22
Q 6
7
7
6 3
2
K32 K22
7
6 H K21 H K32 HK32
7
6
3
2
7
6 H4 K
H K32 HK32 K32 K22
21 H K32
7
6
5
4
..
..
..
..
..
..
..
.
.
.
.
.
.
.
B:9
using
H I $ K22 Q32 ;
K11 Q$1
11 ;
K21 $K22 Q21 K11 ;
K22 Q$1
22 ;
K32 $K22 Q32 K22 :
Only K11 ; K22 ; K21 ; K32 , and powers of H need to be calculated to obtain Q$1 for any
(n; m). The only matrix inversion that needs to be computed to obtain Q$1 is Q$1
11
and Q$1
22 , substantially reducing computer processing time for computer simulations.
Appendix C. General case for (n; m)=(i;1)
As done in Appendix E, the forces acting on each node is presented,
making special exceptions for the case when j m 1. The exceptions arise for
298
j m 1 because one stage now contains both closure rules for the base and the
top of the structure. In the following synthesis we use ^f 111 and ^f 211 from (E.1) and
^f 311 and ^f 411 from (E.4). At the right, where i 1; j 1:
3
2
^f 111 ^f 211
7
6
6 ^f 111 $ ^f 211 7
7
!f 11 6
7
6
6 ^f 311 ^f 411 7
5
4
^f 311 $ ^f 411
2
$f 211 f 311 f 1n1 f 2n1 w111 f 511 $ f 111 $ f 411 $ f 811 w211
6 $f f f f w $ f $ f $ f $ f w 7
211
311
1n1
2n1
111
511
111
411
811
211 7
6
6
7:
4 f 811 $ f 711 $ f 321 f 421 w311 f 611 $ f 511 $ f 6n1 f 7n1 w411 5
f 811 $ f 711 $ f 321 f 421 w311 $ f 611 $ f 511 $ f 6n1 f 7n1 w411
At the center, where (15i5n; j 1)

3
2
^f 1i1 ^f 2i1
7
6
6 ^f 1i1 $ ^f 2i1 7
7
6
!f i1 6
7
6 ^f 3i1 ^f 4i1 7
5
4
^f 3i1 $ ^f 4i1
2
f 1i$11 f 2i$11 $ f 2i1 f 3i1 w1i1 $f 1i1 $ f 4i1 f 5i1 $ f 8i1 w2i1
6
7
6 f 1i$11 f 2i$11 $ f 2i1 f 3i1 w1i1 $ $f 1i1 $ f 4i1 f 5i1 $ f 8i1 w2i1 7
6
7:
6
7
4 f 8i1 $ f 7i1 $ f 3i11 f 4i11 w3i1 $f 5i1 f 6i1 $ f 6i$1 f 7i$11 w4i1 5
f 8i1 $ f 7i1 $ f 3i11 f 4i11 w3i1 $ $f 5i1 f 6i1 $ f 6i$1 f 7i$11 w4i1
At the left end of the base in Fig. 5, where i n; j 1:
3
^f 1n1 ^f 2n1
7
6
6 ^f 1n1 $ ^f 2n1 7
7
!f n1 6
7
6
6 ^f 3n1 ^f 4n1 7
5
4
^f 3n1 $ ^f 4n1
2
f 1n$11 f 2n$11 f 3n1 $ f 2n1 w1n1 $f 1n1 $ f 4n1 f 5n1 $ f 8n1 w2n1
6
7
6 f 1n$11 f 2n$11 f 3n1 $ f 2n1 w1n1 $ $f 1n1 $ f 4n1 f 5n1 $ f 8n1 w2n1 7
6
7:
6
7
4 $f 7n1 f 8n1 $ f 311 f 411 w3n1 $f 6n$11 $ f 5n1 f 6n1 f 7n$11 w4n1 5
$f 7n1 f 8n1 $ f 311 f 411 w3n1 $ $f 6n$11 $ f 5n1 f 6n1 f 7n$11 w4n1
C:1
299
Using,
2
f 0ij
"
f5
f1
f2
7
f3 7
7
f4 7
7
7
f6 7
7 ;
f7 7
7
7
f8 7
7
f9 7
5
f 10 ij
6
6
6
6
6
6
6
f dij 6
6
6
6
6
6
6
4
;
ij
3
w1
6w 7
6 27
wij 6 7 ;
4 w3 5
2
w4
ij
I3
6I
3
!f 11 6
6
40
0
0
0
0
$I3
I3
I3
$I3
0
0
0
0
07
7d
7f
0 5 n1
$I3
I3
6 $I
6 3
6
4 0
2
60
6
6
40
I3
6I
6 3
6
40
2
I3
0
$I3
I3
0
0
I3
$I3
0
$I3
0
$I3
0
I3
0
0
0
0
0
0
$I3
I3
I3
$I3
0
0
I3
0
I3
I3
$I3
$I3
I3
I3
0
0
I3
3
07
7 d
7f
0 5 11
$I3
0 0
0
2
I3
6 $I
0 07
7d
6 3
7f 21 6
4 $I3
0 05
0
3
I3
0 0
$I3
I3
0
0
I3
6I 7
7
6 37
7 0
7f 11 6 7f 1n1
405
5
0
7
7
7w11 :
5
Or, in matrix notation, with the implied matrix definitions:

!f 11 B
^ d f d Bd f d Bd f d B0 f 0 B1n1 f 1n1 Ww11 ;
n1 n1
11 11
21 21
11 11
I3
0 0
0 0
6I
3
!f i1 6
6
40
0 0
0 0
0
$I3
0
I3
0 0
0 0
$I3
0 0
07
7d
7f
0 5 i$11
0 0
I3
C:2
300
$I3
I3
6 $I
6 3
6
4 0
I3
60
6
6
40
I3
0
0
I3
$I3
$I3
$I3
0 0
0
$I3
0
I3
0 0
0 0
I3 0 0
2
3
$I3
I3
I3
I3
3
0
0
07
7d
7f
0 5 i1
$I3
0
0
I3
0
I3
I3
$I3
0
0
0
0
60 I 7
07
37 0
6
7d
7f
7f i11 6
4
5
0
0 0 5 i$11
I3
6
7
I3 7 0 6 I3
7f 6
0 5 i1 4 0
I3
$I3
$I3
I3
6 $I
6 3
6
4 $I3
2
0
0
$I3
0 I3
I3
7
7
7wi1 :
5
Or, in matrix form,

d d
d d
0 0
0 0
!f i1 B
^ d fd
01 i$11 B11 f i1 B21 f i11 B01 f i$11 B11 f i1 Wwi1 ;
2
I3
6I 0 0
0
0
0
3
!f n1 6
6
4 0 0 0 $I3 I3 0
0 0 0 I3 $I3 0
2
$I3 I3 $I3
0
0
6 $I I
0
0
6 3 3 I3
6
4 0
0
0
I3 $I3
0
0
0
$I3 $I3
2
0
0
0 0 0 0 0
60
0
0 0 0 0 0
6
6
4 0 $I3 I3 0 0 0 0
0 $I3 I3 0 0 0 0
2
2
3
$I3
I3 I3
0
6 I 7
6 I $I
0
3
6 3
6 3 7
6
7f 1n1 6
40
4 0 5
0
I3
0
Or, in matrix form,
I3
07
7d
7f
0 5 n$11
0
0
C:3
$I3
I3
I3
I3
3
0
0
07
7 d
7f
0 5 n1
0
2
0
0
6
7
07 d
60
7f 11 6
40
05
0
0
3
0
0 7
7
7wn1 :
I3 5
I3
I3
6 $I
I3 7
70
6 3
7f n$11 6
4 $I3
05
0
I3
07
70
7f
0 5 n1
0
$I3
d d
d d
0 0
0 0
!f n1 B
^ d fd
!
01 n$11 B11 f n1 B21 f 11 B01 f n$11 Bn1 f n1 B1n1 f 1n1 Wwn1 :
C:4
301
Now, assemble (C.2)(C.4) into the form

f 1 B3!f d1 B1 f 01 W1 w1 ;
C:5
where
3
!f 1j
6! 7
6 f 2j 7
6 7
f j 6 . 7;
6 .. 7
4 5
!f nj
2
6
6
6
6
6
6
6
B3 6
6
6
6
6
6
6
4
B1n1
Bd11
0
..
.
^d
B
01
..
0
0
..
! 1n1
B
Bd21
!f d
1
Bd21
..
.
..
.
..
.
.
0
"
f 1n1
f d1
0
..
..
.
.
..
..
'''
f 01j
6 0 7
6f 7
6 2j 7
0
7
fj 6
6 .. 7;
6 . 7
4 5
f 0nj
^d
B
n1
'''
..
.
..
.
..
.
..
.
0
^d
B
01
..
..
..
7
7
0 7
7
.. 7
7
. 7
7;
7
0 7
7
7
7
d 7
B21 5
Bd11
w1j
B011
6
6 0
6 B01
6
6
B1 6
6 0
6
6 .
6 ..
4
0
..
..
..
7
6
6 w2j 7
7
6
wj 6 . 7;
6 .. 7
5
4
wnj
'''
''' '''
..
.
..
..
.
.
..
. B011
0
B001
0
..
.
..
.
7
7
7
7
7
7;
7
7
7
0 7
5
0
Bn1
W1 BlockDiag. . . ; W; W; . . .(:
Or, simply, (C.5) has the form (E.21), where

f f 1;
f d !f d1 ;
f 0 f 01 ;
w w1 ;
W0 W1 ;
Bd B3 ; B0 B1 :
The next set of necessary exceptions that apply to the model i; 1 arises in the form
of the R matrix that relates the dependent tendons set to the generalized coordinates
ld Rq. For any i; 1 case R takes the form following the same procedure as in
(4.12) and (4.13):
"
#
R0
R
;
C:6
~ 11
R
302
where
2
~ 11
R
E3
6
6
6 E2
6
6
6
6 0
6
6 .
6 .
6 .
6
6
6 0
4
~4
E
6
6
6
6
6
6
16
~
E4 6
26
6
6
6
6
6
4
~4
E
E3
~4
E
E2
..
.
E3
''' '''
..
.
..
~4
.
E
E2
..
.
E3
..
.
~4
E
..
.
'''
E2
$I3
0
0
I3
0
0
0
0
0
0
I3
0
0
0
0
$I3
0
I3
0
$I3
0
0
I3
E2
7
7
7
7
7
7
7
7;
7
7
0 7
7
7
~
E4 7
5
E3
0
..
.
7
7
7
7
7
7
I3 7
7:
$I3 7
7
7
0 7
7
0 7
5
I3
The transformation matrix T that is applied to the control inputs takes the following
form. The only exception to (5.17) is that there are no I8 blocks due to the fact that
there are no stages between the boundary conditions at the base and the top of the
structure. The second set of I2 blocks are also not needed since m 1. Hence, the
appropriate T matrix for ~
u Tu is
2
0
6 0
6 T1
6
6
6
6
6
T6
6
6
6
6
6
6
4
'''
..
01
S
..
..
T01
I2
7
7
7
7
7
7
7
10n1)8n
S7
;
72R
7
7
7
7
7
5
I2
303
where S is defined by (5.18), and

2
1
6
60
6
60
6
6
60
0
T1 6
60
6
6
60
6
60
4
0
0
1
0
0
0
0
0
0
0
0
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0 $1
3
0
7
07
7
07
7
7
07
7 2 R8)6 :
07
7
7
17
7
07
5
0
There are n T01 blocks, n S blocks, n I2 blocks. The control inputs are now defined as
"
#
ud1
:
u01
The ud1 block becomes

2
u0d
11
6 0d 7
6 u21 7
7
6
6 0d 7
0d
u
6
u1 6 31 7
7;
6 . 7
6 .. 7
5
4
0d
un1
u2n1
7
6
6 u3n1 7
7
6
6u 7
6 4n1 7
6)1
0d
ui1 6
72R ;
6 u6n1 7
7
6
6u 7
4 7n1 5
u8n1
i 1; 2; 3; . . . ; n; j m 1:
The next section of the appendix will explicitly show all matrix forms for the specific
example n; m 3; 1.
Appendix D. Example case (n; m)=(3,1)
Given the equation for the dynamics of the shell class of tensegrity structures:
"
_ Kp qq Bqu Dqw:
q Kr q
We explicitly write out the matrices that define the problem:
304
q111
7
6
6 q211 7
7
6
6q 7
6 311 7
7
6
6 q411 7
7
6
7
2
3 6
2
6 q121 7
q11
A
7
6
q221 7
6
7 6
6
7
6
q q1 4 q21 5 6
7 Q11 l1 4 D
6 q321 7
D
q31
7
6
6 q421 7
7
6
6q 7
6 131 7
7
6
6 q231 7
7
6
6q 7
4 331 5
q431
0
B
E
32 3 2
A
0
l11
76 7 6
0 54 l21 5 4 D
l31
0
B
E
X11
_ 6
Kr q
4 0
0
0
X21
0
7
0 5;
X31
which is also a 36 ) 36 matrix.

~ HBd Kd B0 K0 H1 B3 K
! d B0 K0 ;
Kp q HK
1
1 1
0
3
2
3 K131
2
3 2
d
d
d
^
0
0 B B1n1 B11 B21 B
H11
7
n1 6
6
76 Kd11 7
6
7B
d
d
d 76
7
6 0
^
H21
0 5B
Kp 4 0
B
B
B
01
11
21 56
B4
d 7
K
@
5
4
21
! 1n1 Bd B
^ d Bd
0
0
H31
B
21
01
11
Kd31
2
B011
6
4 B001
0
0
B011
B001
32
K011
31
76
7C
0 54 K021 5A:
B0n1
7
6
6 r111 7
7
6
6 r211 7
7
6
7
6
6 t511 7
7
6
7
36
6 t111 7
0 6
7
r121 7
76
7;
6
0 56
7
6 r221 7
B 6
7
6 t521 7
7
6
6t 7
6 121 7
7
6
6 r131 7
7
6
6r 7
4 231 5
t531
where Q11 is 36 ) 36. Futhermore,

2
q11
K031
36 ) 36 36 ) 3636 ) 75 * 75 ) 36 36 ) 18 * 18 ) 36(
305
! d , R is needed. In order to form K0 , Q$1 is needed. Therefore, we

In order to form K
1
1
obtain R as follows:
"
#
"
#
d
R
t
0
131
ld ld1
Rq
q1 (;
~ 11
td
R
1
td131
6 d
6 t11
6
6 td
4 21
td31
E6
7 6
7 6 E3
76
7 6E
5 4 2
~4
E
0
~4
E
3
E7 2
3
7 q11
7
E2 6
74 q21 7
5;
~
E4 7
5
q31
E3
E3
E2
where the matrix dimension is 75 ) 36:

~ HBd Pd ; B0 P0 (T
Bq HBT
2
B1n1
6
d
6
!
B P ( B3 P1 ( 4 0
! 1n1
B
d
Bd11
^d
B
01
Bd21
Bd21
Bd11
^d
B
01
2
3 Pd1n1
d
^
Bn1 6
76 0
6
Bd21 7
56 0
4
Bd11
0
P011
where the dimension are:
Pd11
Pd21
7
0 7
7;
0 7
5
Pd31
36 ) 25 36 ) 75 * 75 ) 25
2
B011
6
B0 P0 ( B1 P01 ( 4 B001
0
0
B011
B001
and the dimensions are:
32
B0n1
P021
76
0 54 0
0
7
0 5;
P031
36 ) 6 36 ) 18 * 18 ) 6
H11
6
B4 0
0
0
H21
0
6 0
6 T1
6
6 0
0
6
7 !d
0 6
0 5B3 P1 ; B1 P1 (6 0
6
6 0
H31
6
6
4 0
3
and the dimensions are:
36 ) 24 36 ) 36 * 36 ) 31 * 31 ) 24
'''
0
T01
0
0
0
0
T01
0
'''
01
S
0
0
S
0
0
0
I2
0
0
I2
7
7
7
7
7
7
S 7:
7
0 7
7
7
0 5
I2
306
The control inputs uaij are defined as
"
ud
u0
"
u0d
1
u01
u0d
11
6 0d 7
6 u21 7
7
6
6 u0d 7
6 31 7
6
7
6 u011 7
7
6
6 0 7
4 u21 5
u031
24 ) 1;
where
u2n1
7
6
6 u3n1 7
7
6
6u 7
6 4n1 7
6)1
0d
ui1 6
72R ;
6 u6n1 7
7
6
6u 7
4 7n1 5
u8n1
i 1; 2; 3; j 1:
Bu 2 R36)1 :
The external forces applied to the nodes arise in the product Dw, where
2
H11
6
0
Dq HW H W1 ( 4 0
0
0
H21
0
32
W 0
0
76
0 54 0 W
0 0
H31
with dimensions 36 ) 36 36 ) 36 * 36 ) 36
2
3
w11
6
7
w w1 4 w21 536 ) 1
w31
so
Dw 2 R36)1 :
3
0
7
0 5;
W
307
Appendix E. Nodal forces

At the base, right end of Fig. 5, where i 1; j 1:
!f 11
3
^f 111 ^f 211
7
6
6 ^f 111 $ ^f 211 7
7
6
6
7
6 ^f 311 ^f 411 7
5
4
^f 311 $ ^f 411
2
3
$f 211 f 311 f 1n1 f 2n1 w111 f 511 $ f 111 $ f 411 $ f 811 w211
6 $f f f f w $ f $ f $ f $ f w 7
211
311
1n1
2n1
111
511
111
411
811
211
6
7
6
7:
4 f 811 $ f 711 f 1011 $ f 321 f 421 w311 f 611 $ f 511 f 212 f 112 w411 5
2
f 811 $ f 711 f 1011 $ f 321 f 421 w311 $ f 611 $ f 511 f 212 f 112 w411
At the center of the base, where (15i5n; j 1):

2
3
^f 1i1 ^f 2i1
6
7
6 ^f 1i1 $ ^f 2i1 7
6
7
!f i1 6
7
6 ^f 3i1 ^f 4i1 7
4
5
^f 3i1 $ ^f 4i1
2
3
f 1i$11 f 2i$11 $ f 2i1 f 3i1 w1i1 $f 1i1 $ f 4i1 f 5i1 $ f 8i1 w2i1
6
7
6
7
f 1i$11 f 2i$11 $ f 2i1 f 3i1 w1i1 $ $f 1i1 $ f 4i1 f 5i1 $ f 8i1 w2i1
7:
6
6 f $ f f
7
7i1
10i1 $ f 3i11 f 4i11 w3i1 $f 5i1 f 6i1 $ f 10i$1 f 1i2 f 2i2 w4i1 5
4 8i1
f 8i1 $ f 7i1 f 10i1 $ f 3i11 f 4i11 w3i1 $ $f 5i1 f 6i1 $ f 10i$1 f 1i2 f 2i2 w4i1
At the left end of the base in Fig. 5, where i n; j 1:

2
3
^f 1n1 ^f 2n1
6
7
6 ^f 1n1 $ ^f 2n1 7
7
!f n1 6
6
7
6 ^f 3n1 ^f 4n1 7
4
5
^f 3n1 $ ^f 4n1
2
3
f 1n$11 f 2n$11 f 3n1 $ f 2n1 w1n1 $f 1n1 $ f 4n1 f 5n1 $ f 8n1 w2n1
6
7
6
7
f 1n$11 f 2n$11 f 3n1 $ f 2n1 w1n1 $ $f 1n1 $ f 4n1 f 5n1 $ f 8n1 w2n1
7:
6
6 $f f f
7
7n1
8n1
10n1 $ f 311 f 411 w3n1 $f 10n$11 $ f 5n1 f 6n1 f 1n2 f 2n2 w4n1 5
4
$f 7n1 f 8n1 f 10n1 $ f 311 f 411 w3n1 $ $f 10n$11 $ f 5n1 f 6n1 f 1n2 f 2n2 w4n1
E:1
308
At the second stage, where (i4i4n; j 2:

2
3
^f 112 ^f 212
6
7
6 ^f 112 $ ^f 212 7
7
!f 12 6
6
7
6 ^f 312 ^f 412 7
4
5
^f 312 $ ^f 412
3
$f 6n1 f 7n1 $ f 9n2 $ f 212 f 312 w112 $f 112 $ f 412 f 512 $ f 812 f 912 w212
6 $f f $ f $ f f w $ $f $ f f $ f f w 7
6n1
7n1
9n2
212
312
112
112
412
512
812
912
212 7
6
6
7;
4 $f 712 f 812 f 1012 $ f 322 f 422 w312 $f 512 f 612 $ f 10n2 f 113 f 213 w412 5
$f 712 f 812 f 1012 $ f 322 f 422 w312 $ $f 512 f 612 $ f 10n2 f 113 f 213 w412
2
3
^f 1i2 ^f 2i2
6
7
6 ^f 1i2 $ ^f 2i2 7
7
!f i2 6
6
7
6 ^f 3i2 ^f 4i2 7
4
5
^f 3i2 $ ^f 4i2
2
$f 6i$11 f 7i$11 $ f 9i$12 $ f 2i2 f 3i2 w1i2 $f 1i2 $ f 4i2 f 5i2 $ f 8i2 f 9i2 w2i2
6
7
6 $f 6i$11 f 7i$11 $ f 9i$12 $ f 2i2 f 3i2 w1i2 $ $f 1i2 $ f 4i2 f 5i2 $ f 8i2 f 9i2 w2i2 7
6
7;
6
7
4$f 7i2 f 8i2 f 10i2 $ f 3i12 f 4i12 w3i2 $f 5i2 f 6i2 $ f 10i$12 f 1i3 f 2i3 w4i2 5
$f 7i2 f 8i2 f 10i2 $ f 3i12 f 4i12 w3i2 $ $f 5i2 f 6i2 $ f 10i$12 f 1i3 f 2i3 w4i2
3
^f 1n2 ^f 2n2
7
6
6 ^f 1n2 $ ^f 2n2 7
7
!f n2 6
7
6
6 ^f 3n2 ^f 4n2 7
5
4
^f 3n2 $ ^f 4n2
2
2
3
$f 6n$11 f 7n$11 $ f 9n$12 $ f 2n2 f 3n2 w1n2 $f 1n2 $ f 4n2 f 5n2 $ f 8n2 f 9n2 w2n2
6
7
6$f 6n$11 f 7n$11 $ f 9n$12 $ f 2n2 f 3n2 w1n2 $$f 1n2 $ f 4n2 f 5n2 $ f 8n2 f 9n2 w2n2 7
7:
6
6 $f f f
7
7n2
8n2
10n2 $ f 312 f 412 w3n2 $f 5n2 f 6n2 $ f 10n$12 f 1n3 f 2n3 w4n2 5
4
$f 7n2 f 8n2 f 10n2 $ f 312 f 412 w3n2 $ $f 5n2 f 6n2 $ f 10n$12 f 1n3 f 2n3 w4n2
At the typical stage (15j5m; 14i4n:

2
^f 11j
6
6 ^f 11j
!f 1j 6
6
6 ^f 31j
4
^f 31j
3
^f 21j
7
$ ^f 21j 7
7
7
^f 41j 7
5
$ ^f 41j
E:2
309
$f 6nj$1 f 7nj$1 $ f 9nj $ f 21j f 31j w11j $f 11j $ f 41j f 51j $ f 81j f 91j w21j
7
6
6 $f 6nj$1 f 7nj$1 $ f 9nj $ f 21j f 31j w11j $ $f 11j $ f 41j f 51j $ f 81j f 91j w21j 7
7;
6
7
6 $f f f
4
71j
81j
101j $ f 32j f 42j w31j $f 51j f 61j $ f 10nj f 11j1 f 21j1 w41j 5
$f 71j f 81j f 101j $ f 32j f 42j w31j $ $f 51j f 61j $ f 10nj f 11j1 f 21j1 w41j
3
^f 1ij ^f 2ij
7
6
6 ^f 1ij $ ^f 2ij 7
7
!f ij 6
7
6
6 ^f 3ij ^f 4ij 7
5
4
^f 3ij $ ^f 4ij
3
2
$f 6i$1j$1 f 7i$1j$1 $ f 9i$1j $ f 2ij f 3ij w1ij $f 1ij $ f 4ij f 5ij $ f 8ij f 9ij w2ij
7
6
6 $f 6i$1j$1 f 7i$1j$1 $ f 9i$1j $ f 2ij f 3ij w1ij $ $f 1ij $ f 4ij f 5ij $ f 8ij f 9ij w2ij 7
7;
6
7
6 $f f f $ f
4
7ij
8ij
10ij
3i1j f 4i1j w3ij $f 5ij f 6ij $ f 10i$1j f 1ij1 f 2ij1 w4ij 5
2
$f 7ij f 8ij f 10ij $ f 3i1j f 4i1j w3ij $ $f 5ij f 6ij $ f 10i$1j f 1ij1 f 2ij1 w4ij
3
^f 1nj ^f 2nj
7
6
6 ^f 1nj $ ^f 2nj 7
7
!f nj 6
7
6
6 ^f 3nj ^f 4nj 7
5
4
^f 3nj $ ^f 4nj
3
2
$f 6n$1j$1 f 7n$1j$1 $ f 9n$1j $ f 2nj f 3nj w1nj $f 1nj $ f 4nj f 5nj $ f 8nj f 9nj w2nj
7
6
6 $f 6n$1j$1 f 7n$1j$1 $ f 9n$1j $ f 2nj f 3nj w1nj $ $f 1nj $ f 4nj f 5nj $ f 8nj f 9nj w2nj 7
7:
6
7
6 $f f f
5
4
7nj
8nj
10nj $ f 31j f 41j w3nj $f 5nj f 6nj $ f 10n$1j f 1nj1 f 2nj1 w4nj
$f 7nj f 8nj f 10nj $ f 31j f 41j w3nj $ $f 5nj f 6nj $ f 10n$1j f 1nj1 f 2nj1 w4nj
2
E:3
14i4n; j m:
3
2
^f 11m ^f 21m
7
6
6 ^f 11m $ ^f 21m 7
7
!f 1m 6
7
6
6 ^f 31m ^f 41m 7
5
4
^f 31m $ ^f 41m
3
2
$f 6nm$1 f 7nm$1 $ f 21m f 31m $ f 9nm w1nm $f 41m f 51m $ f 81m $ f 11m f 91m w2nm
7
6
6 $f 6nm$1 f 7nm$1 $ f 21m f 31m $ f 9nm w1nm $ $f 41m f 51m $ f 81m $ f 11m f 91m w2nm 7
7;
6
7
6
$f 71m f 81m $ f 32m f 42m w31m $f 51m f 61m $ f 6nm f 7nm w41m
5
4
$f 71m f 81m $ f 32m f 42m w31m $ $f 51m f 61m $ f 6nm f 7nm w41m
!f im
3
^f 1im ^f 2im
6
7
6 ^f 1im $ ^f 2im 7
6
7
6
7
6 ^f 3im ^f 4im 7
4
5
^f 3im $ ^f 4im
310
$f 6i$1m$1 f 7i$1m$1 $ f 9i$1m $ f 2im f 3im w1im $f 1im $ f 4im f 5im $ f 8im f 9im w2im
6
7
6 $f 6i$1m$1 f 7i$1m$1 $ f 9i$1m $ f 2im f 3im w1im $ $f 1im $ f 4im f 5im $ f 8im f 9im w2im 7
6
7;
6
7
$f 7im f 8im $ f 3i1m f 4i1m w3im $f 5im $ f 6i$1m f 6im f 7i$1m w4im
4
5
$f 7im f 8im $ f 3i1m f 4i1m w3im $ $f 5im $ f 6i$1m f 6im f 7i$1m w4im
!f nm
3
^f 1nm ^f 2nm
6
7
6 ^f 1nm $ ^f 2nm 7
6
7
6
7
6 ^f 3nm ^f 4nm 7
4
5
^f 3nm $ ^f 4nm
3
$f 6n$1m$1 f 7n$1m$1 $ f 9n$1m $ f 2nm f 3nm w1nm $f 1nm $ f 4nm f 5nm $ f 8nm f 9nm w2nm
6
7
6$f 6n$1m$1 f 7n$1m$1 $ f 9n$1m $ f 2nm f 3nm w1nm $ $f 1nm $ f 4nm f 5nm $ f 8nm f 9nm w2nm 7
7:
6
6
7
$f 7nm f 8nm $ f 31m f 41m w3nm $f 5nm f 6nm $ f 6n$1m f 7n$1m w4nm
4
5
$f 7nm f 8nm $ f 31m f 41m w3nm $ $f 5nm f 6nm $ f 6n$1m f 7n$1m w4nm
E:4
2
f 0ij
"
f5
f1
I3
6I
3
!f 11 6
6
40
0
2
6
6
6
6
6
6
6
d
f ij 6
6
6
6
6
6
6
4
;
ij
0 0 0
0 0 0
0 0 0
0 0 0
f2
7
f3 7
7
f4 7
7
7
f6 7
7 ;
f7 7
7
7
f8 7
7
f9 7
5
f 10 ij
w1
6w 7
6 27
wij 6 7 ;
4 w3 5
w4
ij
2
$I3
7
6 $I
0 0 0 07 d
6 3
7f 6
0 0 0 0 5 n1 4 0
0 0 0 0
0
0 0 0 0
$I3
I3
I3
0
0
0 0 0 0
0 0
6 0
6
6
4 I3
$I3
0
0
0
0 0 0 0
0 0 0 0
0 0 0 0
60
0 07
0
7d
6
7f 12 6
4 0 $I3
0 05
0 $I3
0 0
I3
6 $I
6 3
6
4 $I3
I3
2 3
3
2
3
2
$I3
0
0
I3
I3
6I 7
60
7
6I
I3 7
0
6 37
70
6
70
6 3
7f 6
7f 6 7f 1n1 6
40
0 5 11 4 0 I3 5 12 4 0 5
0
0
0 $I3
0
I3
0
0
0
I3
$I3
0
$I3
$I3
$I3
I3
I3
I3
0 0 0 0
0
I3
I3
0 0 0 0
0 0 0 0
0 0 0 0
07
7d
7f
0 5 21
0
Or, in matrix notation, with the implied matrix definitions:
0
0
I3
I3
0 07
7d
7f
0 I3 5 11
0 I3
I3
$I3
0
0
3
0
0 7
7
7w11 :
I3 5
$I3
311
!f 11 Bd f d Bd f d Bd f d Bd f d B0 f 0 B0 f 0
n1 n1
11 11
12 12
21 21
11 11
12 12
B1n1 f 1n1 Ww11 ;
$I3
I3
0
2
I3
0 0
6I
3
!f i16
6
40
0
0
0 0
0 0
0
0
0
0
0 0
2
0
6 0
6
6
4 I3
0 0
6 $I
0 07
7d
6 3
7f i$116
4 0
0 I3 5
$I3
0
60
6
6
40
2
0 0
0
0
0 0
0 0
0
0
0
0
0 0 0 0
3
2
0 I3
0
I3
I3
7
6
I3 7 0
6 $I3
7f i$11 6
4 $I3
05
I3
Or, in matrix form,
0 0
I3
0
0
7
6
0 07 d 60
7f 6
0 0 5 i2 4 0
3
I3
0
0
I3
$I3
0
$I3
0
$I3
0
I3
0
0
0
0
$I3
$I3
I3
I3
I3
0
0
07
7d
7f
I3 5 i1
I3
0 0
0
0
0 07
7d
7f
0 0 5 i11
$I3 I3 0 0 0 0 0
3
2
3
0
I3 I3
0
0
6
0 7
0 7
7 0 6 I3 $I3 0
7
7f i2 6
7wi1 :
40
I3 5
0
I3 I3 5
0
0 0
2
3
0
$I3
6
7
I3 7 0 6 0
7f 6
0 5 i1 4 0
0
$I3
E:5
$I3
I3
$I3
d d
d d
d d
0 0
0 0
0 0
!f i1 Bd f d
01 i$11 B11 f i1 B12 f i2 B21 f i11 B01 f i$11 B11 f i1 B12 f i2 Wwi1
E:6
!f n1
I3
6I
6 3
6
40
2
0
0
0 0
0 0
0 0 0
0 0 0
0
0
0 0
0 0
0 0 0
0 0 0
3
2
0
$I3
7
6
07d
6 $I3
6
7f
I3 5 n$11 4 0
I3
0 0
0 0
0 0 0
6 0
6
6
4 I3
0 0
0 0
0 0
0 0
0 0
0 0
60
0 0 07
7d
6
7f n2 6
40
0 0 05
$I3
0
60
6
6
40
2
I3
6I
6 3
6
40
0
2
0 0 0
2
3
I3
I3
6 $I
I3 7
6 3
70
6
7f
0 5 n$11 4 $I3
0
I3
$I3
0
0
0
0
I3
I3
Or, in matrix form,
I3
3
0
0 7
7
7wn1 :
I3 5
$I3
3
2
0
0
60
07
70
6
7f 6
0 5 n1 4 0
0
I3
I3
$I3
I3
0
0
0
0
$I3
I3
0
0
0
0
0
0
I3
$I3
$I3
$I3
I3
I3
0
0
0 0 0
0 0
0
$I3
0
I3
0 0 0
0 0 0
I3
0 0 0
0 07
7d
7f
0 0 5 11
$I3
0 0
2
3
3
0
$I3
6 I 7
0 7
6 3 7
70
7f 1n1
7f n2 6
4 0 5
I3 5
0 $I3
3
0
07
7d
7f
I3 5 n1
I3
312
d d
d d
d d
0 0
0 0
0 0
!f n1 Bd f d
01 n$11 B11 f n1 B12 f n2 B21 f 11 B01 f n$11 Bn1 f n1 B12 f n2
! 1n1 f 1n1 Wwn1 ;
B
E:7
!f 12
$I3
6 $I
6 3
6
4 0
I3
I3
0
0
0
I3
$I3
0
0
6 0
0
6
6
4 I3 0
$I3 0
2
0 0 0
60 0 0
6
6
40 0 0
0 0 0
2
I3 I3
6 I $I
3
6 3
6
40
0
2
$I3
I3
0
0
0
0
0
0
0
0
0
0
0
0
0
$I3
I3
I3
0
0
$I3
0
0
0
0
$I3
0 0
0 0
0 0
0 $I3
0
0
0
0
3
0
0
I3
0
0
I3
I3
$I3
2
3
0
0
60
07
6
7d
7f 6
I3 5 12 4 0
I3
0
I3
2
3
0
0 0
60 0
07
6
7d
7f 13 6
40 0
05
0 0
0
2
3
I3
0
6 $I
0 7
6 3
7d
7f 6
$I3 5 n2 4 $I3
I3
I3
0
0
0
0
0 $I3
0 0
I3
$I3
0
0
0
$I3
$I3
0 $I3 I3 0
0 $I3 I3 0
0
0
0 0
0
0
0 0
2
3
0
$I3
60
I3 7
6
70
7f 6
0 5 12 4 0
0
0
0
0
I3
0 0 0
0 0 0
0 0 0
I3
0 0 0
3
0
07
7d
7f
0 5 n1
0
0
0
0
3
0 0
0 07
7d
7f
0 0 5 22
0 0
0 7
70
7f
I3 5 13
$I3
7
7
7w12 :
5
Or, in matrix form,

!f 12 B
! d f d Bd f d Bd f d B
! d f d Bd f d B0 f 0 B0 f 0 Ww12 ;
12 12
21 22
12 13
n1 n1
n2 n2
11 12
12 13
E:8
0 0 0 $I3 I3 0 0
6 0 0 0 $I I 0 0
3
3
!f i2 6
6
40 0 0
0
0 0 0
0 0 0
0
0 0 0
2
0
0 0 0 0 0
6 0
0 0 0 0 0
6
6
4 I3 0 0 0 0 0
2
$I3
0 0
0
0
0
0 0 0 0
0
0
0 0 0
60
0
0 0 0
6
6
4 0 $I3 I3 0 0
0 $I3 I3 0 0
3
2
2
0
0
I3
7
60
6I
0
70 6 3
6
6
7f 6
4 0 I3 5 i3 4 0
0 $I3
0
2
$I3
0
6 $I
07
6 3
7d
6
7f
0 5 i$11 4 0
2
0
0
0
0
I3
$I3
0
0
0
0
0
0
I3
I3
$I3
I3
0
0
0
0
$I3
I3
0
0
0
0
I3
$I3
$I3
$I3
I3
I3
0
3
2
0
0 0 0
6
07
7 d 60 0 0
7f i3 6
40 0 0
05
0
0 0 0 $I3
0 0 0 $I3
0 0 0
0
0 0 0 0 0 0
2
3
0
I3 $I3
6 $I
7
07
7d
6 3 I3 7 0
7f i12 6
7f
4 $I3
05
0 5 i2
I3
0
0
3
0
0
0
0 7
7
7wi2 :
I3 I 3 5
3
I3
$I3
I3
$I3
0
0
3
0
0 7
7d
7f
$I3 5 i$12
I3
3
0
07
7d
7f
I3 5 i2
I3
313
Or, in matrix form,

d d
d d
d d
0 0
!f i2 B
!d fd
!d d
n1 i$11 B12 f i2 B12 f i3 Bn2 f i$12 B21 f i12 B11 f i2
B012 f 0i3 Wwi2 ;
0
0
6
60
0
!f n2 6
6 0 $I
3
4
0 $I3
0 0
0 0
0 0
0 0
I3
0 0
0 0
I3
0 0
0 0
0 0
0 0
0 0
6
60 0
6
60 0
4
0 0
0 0
0 0
0 0
0 0
0 0
0 0
6
6 0
6
6 I
4 3
$I3
I3
6
6 I3
6
60
4
0
E:9
2
0
7
6
60
07 d
7f 6
12
60
07
5
4
0
0
0
$I3
$I3
0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
7
6
60
0 07 d
7f 6
n3
7
60
0 05
4
0 0
0
$I3
I3
I3
0 $I3
I3
0 0
0 0
0 0
7
07 d
7f
n$11
07
5
0
$I3
I3
$I3
I3
I3
$I3
$I3
7
6
6 $I3
0 7d
7f
6
n$12 6
$I3 7
5
4 0
I3
0
0 0
0 0
I3
I3
0 $I3
I3
I3
7
6
6 $I
0 70
7f 6 3
n3
7
6 $I
I3 5
4 3
$I3
I3
$I3
$I3
$I3
I3
I3
I3
I3
$I3
0
0
7
0 7d
7f
n2
I3 7
5
I3
7
I3 7 0
7f
n2
0 7
5
0
7
0 7
7wn2 :
I3 7
5
$I3
Or, in matrix form,

d d
d d
0 0
!f n2 Bd f d B
! d fd
!d d
21 12
n1 n$11 Bn2 f n$12 B12 f n2 B12 f n3 B12 f n3
B011 f 0n2 Wwn2 ;
E:10
314
0 0 0 $I3
I3
$I3
0
0
60 0 0
!f 1j 6
6
40 0 0
0 0 0
2
0 0
60 0
6
6
40 0
6
6
6
4
$I3
$I3
I3
6 $I I
0 0 07
0
0
7d
6 3 3 I3
7f nj$1 6
4 0
0 0 05
0
0
I3 $I3
0 0 0
0
0
0
$I3 $I3
2
3
0 0 $I3
0
0
0 0 0 0 0
0
6
0
0 0 0 0 0
0 0 $I3
0 7
7 d 60
7f nj 6
4 0 $I3 I3 0 0 0 0
0 0
0
$I3 5
0 $I3 I3 0 0 0 0
0 0
0
I3
3
2
3
2
0 0 0 0 0
I3 $I3
0
0
6 $I
7
60
0 0 0 0 07
0
I
3 7 0
7d
6 3
6
7f 1j1 6
7f 1j 6
4 $I3
4 0 I3
0 0 0 0 05
0 5
0 0
0 0
0
0
I3
I3
0
0
0 0
0 0 0 0
0 0 0
0 0
0 0
0 0
$I3 0 0 0 0 0 0 0
2
3
0
0
I3 I3
6 I $I
0
0 7
3
6 3
7
6
7w1j :
40
0
I 3 I3 5
0
0
I3 $I3
I3
$I3
0
0
I3
I3
I3
3
0
07
7d
7f
0 5 2j
07
7d
7f
I3 5 1j
I3
0
3
0 $I3
I3
$I3
7
70
7f 1j1
5
Or, in matrix form,

d d
d d
d d
0 0
!f 1j B
! d fd
!d d
n1 nj$1 B12 f 1j Bn2 f nj B21 f 2j B12 f 1j1 B11 f 1j
B012 f 01j1 Ww1j ;
E:11
$I3
I3
I3
0 0 0
0
0
2
0 0 0 0 0
60 0 0 0 0
6
6
40 0 0 0 0
$I3
6 $I
07
6 3
7d
7f i$1j$1 6
4 0
05
60
!f ij 6
6
40
$I3
0
0
0 7
7d
7f
$I3 5 i$1j
0
2
0
60
6
6
40
$I3
0
$I3
I3
6 0
6
6
4 I3
6 $I
07
7d
6 3
7f ij1 6
4 $I3
05
$I3
I3
6I
6 3
6
40
2
I3
$I3
0 7
7
7wij :
I3 5
I3
I3
Or, in matrix form,
$I3
I3
I3
I3
I3
$I3
I3
$I3
0
$I3
I3
I3
I3
$I3
$I3
I3
I3
$I3
$I3
6
I3 7
7 0 60
7f ij 6
40
0 5
0
$I3
0
3
07
7d
7f
0 0 5 i1j
0
0 7
70
7f
I3 5 ij1
$I3
$I3
0
I3
0
I3
07
7d
7f
I3 5 ij
I3
315
d d
d d
d d
0 0
!f ij B
! d fd
!d d
n1 i$1j$1 B12 f ij Bn2 f i$1j B21 f i1j B12 f ij1 B11 f ij
B012 f 0ij1 Wwij ;
0 0
60 0
!f nj 6
6
40 0
0
0
$I3
$I3
0
E:12
I3
0 0
I3
0
0 0
0 0
07
7 d
7f
0 5 n$1j$1
0 0 0
0
0 0 0 0
2
$I3 I3 $I3
0
0
$I3
6 $I I
I
0
0
I3
3
6 3 3
6
4 0
0
0
I3 $I3 I3
0
$I3
0 0
60
6
6
40
0
0
0
0
0 0
0 0
0
0
0 0
0
6 0
6
6
4 I3
2
$I3
I3
6I
6 3
6
40
2
$I3
$I3
0
0
0 0
0
0
0
0
0 0
0 0
0
0
0
0
0 0
I3
$I3
0
0
I3
0
I3
I3
$I3
$I3
I3
$I3
0
I3
07
7d
7f
I3 5 nj
I3
2
0
0
0
7
6
0
0 7d
60
6
7f
$I3 5 n$1j 4 0 $I3
3
0 0
0
I3
0
0
0
0
0 0
0 0
07
7 d
7f
0 5 1j
0 $I3 I3 0 0 0 0 0
I3
2
3
3
2
3
0
I3 $I3
0
0
6
6 $I
7
07
0 7
70
7d
6 3 I3 7 0 6 0
7f nj1 6
7f nj 6
7f
4
5
4
5
0 I3 5 nj1
0
$I3
0
I3
$I3
7
7
7wnj :
5
Or, in matrix form,

d d
d d
d d
0 0
!f nj B
! d fd
!d d
n1 n$1j$1 B12 f nj Bn2 f n$1j B21 f 1j B12 f nj1 B11 f nj
B012 f 0nj1 Wwnj ;
$I3
6 $I
3
!f 1m 6
6
4 0
0
I3
I3
0
0
$I3
I3
0
0
0
I3
$I3
0
$I3
$I3
E:13
$I3
I3
I3
I3
I3
$I3
0
0
60
07
7d
6
7f 1m 6
40
05
0
0
0 0 0
0 0
0
$I3
$I3
0
I3
I3
0 0 0
0 0 0
0 0 0
0 07
7d
7f
0 0 5 2m
0 0
316
0 0 0
0 0 0
0 0 0 $I3
6 0 0 0 $I
60 0 0
0
0
0 0 07
3
6
6
7d
6
7f nm 6
40 0 0
4 0 0 0 $I3 I3 0 0 0 5
0
0 0 0
0
0 0 0 I3 $I3 0 0 0
2
2
3
3
I3 $I3
I3 I3
0
0
6 $I
6 I $I
7
0
0 7
3
6 3 I3 7 0
6 3
7
6
7f 1m 6
7w1m :
4 $I3
40
0 5
0
I3 I 3 5
I3
0
0
0
I3 $I3
I3
0 0 0
I3
0
0
0 0 07
7d
7f
0 0 0 5 nm$1
0 0 0
Or, in matrix form,

0 0
!f 1m Bd f d Bd f d Bd f d B
!d fd
1m 1m
21 2m
nm nm
n1 nm$1 B11 f 1m Ww1m ;
!f im
2
3
0 0 0 $I3 I3 0 0 0
0 0 0
60 0 0
6 0 0 0 $I I 0 0 0 7
3
3
6
6
7d
6
6
7f
40 0 0
0
0 0 0 0 5 i$1m$1 4 0 0 0
0 0 0
0 0 0
0
0 0 0 0
2
3
2
$I3 I3 $I3
0
0
0
$I3 I3 0
6 $I I
7
60
I
0
0
I
$I
0
3
3
3
6 3 3
7d 6
6
7f 6
4 0
0
0
I3 $I3 I3
0
0 5 im 4 0
0
0
0
$I3 $I3 I3
0
0
0
2
3
2
3
0
0
I3 $I3
I3 I3
6 $I
7
6 I $I
I
0
0 7
3
3
3
3
6
70
6
7
6
7f im 6
7wim :
4 $I3
40
0
I3 I3 5
0 5
I3
0
0
0
I3 $I3
2
0
0
0
0
0
0
$I3
I3
I3
$I3
0
0
0
0
$I3
$I3
0
0
I3
I3
0
0
0
0
0
0
0
0
E:14
$I3
$I3
0
0
3
0
07
7d
7f
0 5 i$1m
0
0
0
0
0
0
0
0
0
3
0
07
7d
7f
0 5 i1m
0
Or, in matrix notation,

d d
d d
d d
0 0
!f im B
! d fd
n1 i$1m$1 Bnm f i$1m B1m f im B21 f i1m B11 f im Wwim ; E:15
2
0 0
0 $I3
I3
0 0
0 0
7
6
6
60 0 0
6 0 0 0 $I3 I3 0 0 0 7 d
!f nm 6
7f
6
n$1m$1
7
60 0 0
60 0 0
0
0 0 0 05
4
4
0 0 0
0 0 0
0
0 0 0 0
3
2
2
$I3 I3 $I3
0
0
$I3 I3 0
0
7
6
6
7
6 $I3 I3 I3
6
0
0
I3 $I3 0 d
0
7f 6
6
nm
6 0
60
0
07
0
0
I3 $I3 I3
5
4
4
2
$I3 $I3 I3
2
I3 I3
0
7
6
6 I $I3 0
I3 7 0
7f 6 3
nm
60
0 7
0
I3
5
4
0
0
0
I3
I3
$I3
6
6 $I3
6
6 $I
4 3
I3
0
3
7
0 7
7wnm :
I3 7
5
$I3
$I3
I3
I3
$I3
0
0 $I3
0 $I3
0
0 0
0 0
$I3
I3
0 0
I3
0 0
0 $I3
7
07 d
7f
n$1m
07
5
0
3
0 0
7
0 07 d
7f
1m
0 07
5
0 0
Or, in matrix notation,

!f nm B
! d fd
n1 n$1m$1
Bdnm f dn$1m Bd1m f dnm Bd21 f d1m B011 f 0nm Wwnm : E:16
Now assemble (E.5)(E.16) into the form

f 1 B3!f d B4 f d B1 f 0 B2 f 0 W1 w1 ;
1
317
E:17
where
!f d
1
"
f 1n1
6
6
6
6
6
6
6
B3 6
6
6
6
6
6
6
4
f d1
f d1j
6 d7
6f 7
6 2j 7
7
f dj 6
6 .. 7;
6 . 7
4 5
f dnj
B1n1
Bd11
0
..
.
Bd01
..
.
0
0
..
! 1n1
B
Bd21
Bd21
..
.
..
.
..
.
..
..
..
..
6 0 7
6f 7
6 2j 7
7
f 0j 6
6 .. 7;
6 . 7
4 5
f 0nj
Bd01
..
..
..
B4 BlockDiag' ' ' ; Bd12 ; Bd12 ; ' ' '(;
w1j
7
6
6 w2j 7
7
6
wj 6 . 7;
6 .. 7
5
4
wnj
Bdn1
'''
..
.
..
.
..
.
..
.
'''
f 01j
3
!f 1j
6! 7
6 f 2j 7
6 7
f j 6 . 7;
6 .. 7
4 5
!f nj
2
7
7
7
7
7
7
7
7;
7
0 7
7
7
7
d 7
B21 5
Bd11
0
..
.
B2 BlockDiag' ' ' ; B012 ; B012 ; ' ' '(;

B011
6
6 0
6 B01
6
6
B1 6
6 0
6
6 .
6 ..
4
0
..
..
..
'''
''' '''
..
.
..
..
.
.
..
. B011
0
B001
0
..
.
..
.
7
7
7
7
7
7;
7
7
7
0 7
5
B0n1
W1 BlockDiag' ' ' ; W; W; ' ' '(:

Also from (E.5)(E.16)
f 2 B5!f d B6 f d B4 f d B7 f 0 B2 f 0 W1 w2 ;
1
where
! 5 (;
B5 0; B
B7 BlockDiag' ' ' ; B011 ; B011 ; ' ' '(;
E:18
318
0
6
6 !d
6 Bn1
6
6
!5 6
B
6 0
6
6 .
6 ..
4
0
0
..
..
..
.
.
.
'''
!d
B
n1
'''
..
.
..
.
..
.
..
!d
B
n1
..
7
7
7
7
7
7;
7
7
7
7
5
0
..
.
!d
B
12
6
6 d
6 Bn2
6
6
6
6 0
B6 6
6 .
6 .
6 .
6
6
6 0
4
Bd21
Bd21
..
.
..
.
..
.
0
..
..
..
..
'''
Bdn2
'''
..
.
..
.
..
.
..
.
0
Bdn2
..
..
..
7
7
0 7
7
.. 7
7
. 7
7:
7
7
0 7
7
7
d 7
B21 5
!d
B
12
Also from (E.5)(E.16)

! 5 f d B6 f d B4 f d B7 f 0 B2 f 0 W1 w3 ;
f3 B
2
3
4
3
4
E:19
d
0
! 5fd
fm B
m$1 B8 f m B7 f m W1 wm ;
E:20
Bd1m
6
6 d
6 Bnm
6
6
6
6 0
B8 6
6 .
6 .
6 .
6
6
6 0
4
Bd21
Bd21
..
.
..
.
..
.
0
..
..
..
..
...
Bdnm
...
..
.
..
.
..
.
..
.
0
Bdnm
..
..
..
7
7
7
7
7
7
7
7:
7
7
0 7
7
7
d 7
B21 5
Bd1m
0
..
.
Or, simply, the vector form of (E.17)(E.19) is

f Bd f d B0 f 0 W0 w;
E:21
where
f1
6 . 7
7
f6
4 .. 5;
2
fm
2 !d 3
f1
6 d7
6 f2 7
6 7
f d 6 . 7;
6 . 7
4 . 5
f dm
3
f 01
6 7
6 7
f 0 6 ... 7;
4 5
f 0m
W0 BlockDiag' ' ' ; W1 ; W1 ; ' ' '(;
3
w1
6 . 7
7
w6
4 .. 5;
2
wm
B3
6
6
6 B5
6
6
Bd 6
6 0
6
6 .
6 ..
4
0
B4
B6
..
!5
B
..
.
'''
..
..
'''
..
.
..
.
B6
!5
B
0
..
.
7
7
7
7
7
7
0 7;
7
7
B4 7
5
B8
6
6
6
6
6
B0 6
6
6
6
6
4
B1
B2
0
..
..
..
0
..
.
..
.
B7
..
.
''' '''
'''
..
.
..
.
..
.
0
0
..
.
319
7
7
7
7
7
7
0 7:
7
7
B2 7
5
B7
References
[1] R.E. Skelton, Systems design, the absentee in system theory, American Automatic Control
Conference, San Diego, CA, June 1999.
[2] R.E. Skelton, A systems approach to integrate structure and control design, SPIE Conference on
Mathematics for Smart Structures, Newport Beach, CA, March 2000.
[3] K.M. Grigoriadis, G. Zhu, R.E. Skelton, Optimal redesign of linear systems, J. Dyn. Systems Meas.
Control 118 (3) (1996) 598605.
[4] M.P. Bendsoe, Optimal shape design as a material distribution problem, Struct. Optim. 1 (1989)
193202.
[5] M.P. Bendsoe, Optimization of Structural Topology, Shape, and Material, Springer, Berlin, 1995.
[6] M.P. Bendsoe, N. Kikuchi, Generating optimal topologies in structural design using a homozenization method, Comput. Methods Appl. Mech. Eng. 71 (1988) 197224.
[7] A.R. Diaz, M.P. Bendsoe, Shape optimization of structures for multiple loading conditions using a
homogenization method, Struct. Optim. 4 (1992) 1722.
[8] A.H. Simmons, C.A. Michal, L.W. Jelinski, Molecular orientation and two-component nature of the
crystalline fraction of dragline silk, Science 271 (1996) 8487.
[9] Y. Termonia, Molecular modeling of spider silk elasticity, Macromolecules 27 (1994) 73787381.
[10] R. Chu, Tensegrity, J. Synergetics 2 (1) (1988).
[11] R.B. Fuller, R. Marks, The Dymaxion World of Buckminister Fuller, Anchor Books Edition, 1973.
[12] R.B. Fuller, Tensile-integrity structures, US Patent 3,063,521, 1962.
[13] R.B. Fuller, Synergetics Explorations in the Geometry of Thinking, Collier Macmillan Publishers,
London, 1975.
[14] H. Furuya, Concept of deployable tensegrity structures in space applications, Int. J. Space Struct.
7 (2) (1992) 143151.
[15] D.E. Ingber, Cellular tensegrity: defining new rules of biological design that govern the cytoskeleton,
J. Cell Sci. 104 (3) (1993) 613627.
[16] D.E. Ingber, Tensegrity: the architectural basis of cellular mechanotransduction, Annu. Rev. Physiol.
59 (1997) 575599.
[17] D.E. Ingber, Architecture of life, Sci. Am. (January 1998) 4857.
[18] H. Kenner, Geodesic Math and How to Use it, University of California Press, Berkeley, CA, 1976.
[19] R. Motro, Tensegrity systems: the state of the art, Int. J. Space Struct. 7 (2) (1992) 7583.
[20] I.J. Oppenheim, W.J. Williams, Tensegrity prisms as adaptive structures, ASME Annual Meeting,
Dallas November 1997.
[21] A. Pugh, An Introduction to Tensegrity, University of California Press, Berkeley, 1976.
[22] R. Adhikari, R.E. Skelton, W.J. Helton, Mechanics of tensegrity beams, UCSD, Structural System
and Control Lab., Report No. 1998-1, 1998.
[23] R.E. Skelton, C. Sultan, Controllable tensegrity, a new class of smart structures, SPIE Conference,
Mathematics and Control in Smart Structures, San Diego, March, 1997.
320
[24] R.E. Skelton, M. He, Smart tensegrity structure for nestor, SPIE Conference, Smart Structures and
Integrated Systems, San Diego, CA, March 1997.
[25] H. Murakami, Y. Nishimura, T.J. Impesullo, R.E. Skelton, A virtual reality environment for
tensegrity structures, Proceedings of the 12th ASCE Engineering Mechanics Conference, La Jolla,
CA, 1998.
[26] D. Williamson, R.E. Skelton, A general class of tensegrity systems: geometric definition. ASCE
Conference Engineering Mechanics for the 21st Century, La Jolla, CA, 1720 May 1998, pp. 164169.
[27] B.I. Yakobson, R.E. Smalley, Fullerene nanotubes: C1,000,000 and beyond, American Scientist,
JulyAugust (1997) 324337.

Dynamics of The Shell Class of Tensegrity Structures

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

Dynamics of The Shell Class of Tensegrity Structures

Uploaded by

Copyright:

Available Formats

Journal of the Franklin Institute 338 (2001) 255320

Dynamics of the shell class of tensegrity

simple non-toxic materials (amino acids) into a microstructure that contains a

Fig. 1. Natures strongest fiber: the spider fiber [9].

Fig. 2. Buckytubes [27].

Definition 2.3. A stable structure is said to be a Class 1 tensegrity structure if the

Fig. 4. Class 1 and Class 2 tensegrity structures.

Fig. 5. Topology of an (8,4) Class 1 tensegrity shell.

Fig. 6. A typical ij element.

Fig. 7. Class 1 shell: n; m 3; 1.

t901 t9n1 $t1n1 ;

0 t10i$10 t5i0 t7i0 t7i$10 ;

At the top, the closure rules are

t100m $t70m $t7nm ;

0 t1im1 t9im1 t3i1m1

At the closure of the circumference (where i 1):

t60 j$1 t6n j$1 ;

t70 j$1 t7n j$1 ;

t100 j$1 t10n j$1 :

From Figs. 5 and 6, when j 1, then

0 f 7i j$1 f 7i$1 j$1 f 5i j$1 f 10i$1j$1

and for j m where, in Figs. 5 and 6,

0 f 1im1 f 9im1 f 3i1m1 f 1i1m1 :

W0 BlockDiag' ' ' ; W1 ; W1 ; ' ' '(;

3. Dynamics of a 2 rod element

Fig. 8. A single rigid rod.

and the square of this matrix is

The matrix elements r1 ; r2 ; r3 ; q1 ; q2 ; q3 , etc. are to be interpreted as the

q2ij ' q2ij L21ij ;

q4ij ' q4ij L22ij ;

qT2ij ; qT3ij ; qT4ij

and the force vectors appear in the form

q qT11 ; . . . ; qTn1 ; qT12 ; . . . ; qTn2 ; . . . ; qT1m ; . . . ; qTnm (T :

where f is defined in (2.15) and

Fig. 9. Choice of independent variables.

^ij qij r1ij t5ij $ r2ij :

q1ij p2ij p1ij 2qij r1ij ;

q2ij p2ij $ p1ij r1ij ;

q4ij p4ij $ p3ij r2ij :

The equations of motion will be written in the q coordinates. Substitution of (4.1)

To put (4.7) in a matrix form, define these matrices:

Then (4.7) can be written simply

where the 12nm ) 12nm matrix Q is composed of the 12 ) 12 matrices AH as

''' ''' '''

Q22 BlockDiag. . . ; C; . . . ; C(;

^ij $r1ij $ t5ij r2ij ;

For j m we replace t6ij and t7ij with

^nj , and i n i. Eq. (4.11) has the matrix form,

Eq. (4.5) yields

Hence, (4.12) and (4.13) yield

Also, from (4.10) and (4.12)

tdij E1 qi j$1 E2 qi$1j E3 qij E4 qi1j E5 qi1 j1 ;

Now from (4.14) and (4.15), define

R21 BlockDiag ' ' ' ;

E1 ; ' ' ' (;

! 11 have the same structure as R11 except E4 is replaced by E

if L!aij > jjtaij jj;

jjtaij jj2 qT RTaij Raij q:

From (5.1) and (5.5),

baij q kaij qT RTaij Raij q$1=2 Raij q;

f dij $Kdij q Pdij udij

Now substitute (5.9) and (5.10) into

f dj $Kdj q Pdj udj