Professional Documents
Culture Documents
www.elsevier.com/locate/ijsolstr
b,*
, Jeongheon Han
b,*
a
b
Abstract
This paper characterizes the necessary and sucient conditions for tensegrity equilibria. Static models of tensegrity
structures are reduced to linear algebra problems, after rst characterizing the problem in a vector space where direction
cosines are not needed. This is possible by describing the components of all member vectors. While our approach
enlarges (by a factor of 3) the vector space required to describe the problem, the advantage of enlarging the vector space
makes the mathematical structure of the problem amenable to linear algebra treatment. Using the linear algebraic
techniques, many variables are eliminated from the nal existence equations.
2003 Elsevier Ltd. All rights reserved.
Keywords: Tensegrity; Structure; Statics; Equilibrium conditions; Linear algebra
1. Introduction
The tensegrity structures introduced by Snelson (1996) pose a wonderful blend of geometry and mechanics. In addition, they have engineering appeal in problems requiring large changes is structural shape.
Tensegrity structures exist as a prestressed stable connection of bars and strings. Most existing smart
structure methods are limited to small displacements, but the control of tensegrity structures allows very
large shape changes to occur (Skelton and Sultan, 1997; Skelton et al., 2001a; Motro, 1992). Therefore, an
ecient set of analytical tools could be the enabler to a host of new engineering concepts for deployable and
shape controllable structures. This paper characterizes the static equilibria of tensegrity structures in terms
of vectors which describe the elements (bars and strings), thereby eliminating the need to use direction
cosines and the subsequent transcendental functions that follow their use. For a comparison of previous
methods of form-nding in tensegrity structures, see Tibert and Pellegrino (2001), Vassart and Motro
(1999), Motro et al. (1994), Linkwitz (1999), Barnes (1998), and Schek (1974).
6348
It is well known in a variety of mathematical problems that enlarging the domain in which the problem
is posed can often simplify the mathematical treatment. In fact, many nonlinear problems admit solutions by
linear techniques by enlarging the domain of the problem. For example, nonlinear Riccati equations are
known to be solvable by linear algebra in a space that is twice the size of the original problem statement. The
purpose of this paper is to show that the mathematical structure of the equations admits some treatments by
linear algebra methods by enlarging the vector space in which the tensegrity statics problem is characterized.
Our results characterize the equilibria conditions of tensegrity structures in terms of a very small number
of variables since the necessary and sucient conditions of the linear algebra treatment allow the elimination of several of the original variables. These results can be used for ecient algorithms to design and
simulate a large class of tensegrity structures. Tensegrity concepts have been around for 50 years without
ecient design procedures (Kenner, 1976; Pugh, 1976; Connelly, 1982, 1993, 1999; Ingber, 1993, 1997,
1998; Williamson and Skelton, 1998a,b; Motro, 1984, 1990, 2001; Skelton et al., 2001b).
The paper is organized as follows. After the review of mathematical preliminaries in Section 2, Section 3
introduces the network representations of tensegrity structures as an oriented graph in real three dimensional
space. Geometric connectivity, equilibrium, and a coordinate transformation will be introduced. Section 4
introduces the algebraic equilibrium conditions. After we derive necessary and sucient conditions for the
existence of an unloaded tensegrity structure in equilibrium, we write the necessary and sucient conditions
for the externally loaded structure in equilibrium. A couple of examples will illustrate the results.
2. Algebraic preliminaries
We let In dene the n n identity matrix, and 0 dene an n m matrix of zeros. (The dimensions of 0 will
be clear from the context.) We also let qA dene the rank of the matrix A. Let A 2 Rmn and B 2 Rpq ,
then the Kronecker product (Horn and Johnson, 1985) of A and B is dened as
2
3
A1; 1B A1; 2B A1; nB
6 A2; 1B A2; 2B A2; nB 7
6
7
mpnq
AB6
72R
..
..
..
4
5
.
.
.
Am; 1B
Am; 2B
Am; nB
where Ai; j is the i; j element of a matrix A. Then we have the following result.
Lemma 1. The following statements are true.
i(i) Suppose A 2 Rnr , B 2 Rnr , C 2 Rrp . Then
A Im T AT Im
A Im B Im A B Im
A Im C Im AC Im
rankA Im m rankA:
(ii) Suppose A 2 Rnn has eigenvalues fk1 ; k2 ; . . . ; kn g. Then A Im also has eigenvalues fk1 ; k2 ; . . . ; kn g
where each eigenvalue is repeated m times.
The derivations in this paper rely heavily on the singular value decomposition svdA fUA ; RA ; VA g of a
matrix A as expressed in the following result (Horn and Johnson, 1985).
6349
Lemma 2
ii(i) Suppose an n m matrix A has rank rA , then there exists an n n unitary matrix UA , an m m unitary
matrix VA and a positive denite rA rA diagonal matrix R1A such that
R1A 0
T
A UA RA VA ; RA
:
1
0 0
i(ii) If fUA ; VA g are partitioned such that
UA U1A ; U2A ;
V V1A ; V2A
with
U2A 2 Rnn
rA ;
U1A 2 RnrA ;
V1A 2 RmrA ;
V2A 2 Rmm rA
then
UT1A U1A IrA ;
UT1A U2A 0;
UT2A U2A In rA
VT1A V2A 0;
VT2A V2A Im rA
UT2A A
0;
AV2A 0:
where zx 2 Rn
rA is arbitrary.
(iv) Suppose A 2 Rnm and
~ A Ip ;
A
svdA fUA ; RA ; VA g:
~ 2 Rpnpm and
Then A
~ fUA Ip ; RA Ip ; VA Ip g:
svdA
i(v) Suppose A 2 Rnm and
T
~
b ~
bT1 ; ~
bT2 ; . . . ; ~
bTn ;
~
bk bk1
bk2
...
bkp 2 Rp :
166p
b2
...
bn T :
x2
...
xm ;
If
x x1
166p
6350
xj2
...
xjp T 2 Rp
10
is a solution of (6).
In particular, (6) has a solution if and only if UT2A Ip ~b 0. When this condition is satised, then all so~ are of the form
lutions x
T
~
~ V1A R
1
zx
x
1A U1A Ip b V2A Ip ~
For example, if no two bars are in contact, then n~p 0 and np 2nb . Or, if nb
nb0 bars all contact at a
single bar end, then n~p nb
nb0 1.
Thus given a tensegrity structure consisting of np nodes, nb bars and ns strings, the positions of the nodes
are described by the np vectors fp1 ; p2 ; . . . ; pnp g, the positions of the bars are described by the nb vectors
fb1 ; b2 ; . . . ; bnb g, and the positions of the strings are described by the ns vectors fs1 ; s2 ; . . . ; sns g.
Denition 3. The geometry of the tensegrity structure is dened by the tensegrity node vector p 2 R3np , the
tensegrity bar vector b 2 R3nb , and the tensegrity string vector s 2 R3ns where
pT pT1 ; pT2 ; . . . ; pTnp ;
11
12
6351
Force equilibrium. In our study of tensegrity structures, we are concerned with structures in which bars
sustain compressive forces and strings do not. We therefore choose to distinguish between the string (or
tensile) forces ftn g and the bar (or compressive) forces ff m g which are dened in terms of the string and bar
vectors respectively as follows.
Denition 5. Given the tensile force tn in the string characterized by the string vector sn and the compressive
force f n in the bar characterized by the bar vector bn , the tensile force coecient cn > 0 and the compressive
force coecient km > 0 are dened by
tn cn sn ;
f m km bm :
13
The forces of the tensegrity structure are dened by the external force vector w 2 R
vector f 2 R3nb , and the tension vector t 2 R3ns where
h
i
wT wT1 ; wT2 ; . . . ; wTnw ; f T f T1 ; f T2 ; . . . ; f Tnb ; tT tT1 ; tT2 ; . . . ; tTns :
3nw
, the compression
14
f K I3 b
15
where
C diagfc1 ; c2 ; . . . ; cns g;
K diagfk1 ; k2 ; . . . ; knb g:
16
The diagonal matrices fC; Kg shall be referred to as the tensile force coecient matrix and compressive
coecient force matrix, respectively.
Force convention. Suppose each node pk is subjected to compressive vector forces ff mk g, tensile vector
forces ftnk g and external force wk . Then the Law for Static Equilibrium may be stated as follows:
X
X
tnk
f mk
wk 0
17
n
where a positive sign is assigned to a (tensile, compressive and external) force vector leaving a node, and a
negative sign is assigned to a (tensile, compressive and external) force vector entering a node. The negative
sign in (17) is a consequence of the fact that we choose to dene positive force coecients kn and cn .
From the network, it follows that components of the string vector s and the bar vector b in (11) can be
written as a linear combination of components of the node vector p. Also, if branch k is a bar which leaves
node i and enters node j, then bk pj
pi , whereas if branch k is a string which leaves node i and enters
e T p sT ; bT T where the matrix C
e consists only of block matrices
node j, then sk pj
pi . Hence we have C
of the form f0; I3 g. In particular, if we consecutively number the ns nb branches fs1 ; s2 ; . . . ;
e C
e ij is dened
sns ; b1 ; b2 ; . . . ; bnb g as f1; 2; . . . ; ns ; ns 1; . . . ; ns nb g, then the 3np 3ns 3nb matrix C
by
8
if force vector j enters node i
< I3
e ij
I3 if force vector j leaves node i
18
C
:
0
if force vector j is not incident with node i:
e has exactly one block I3 and one block
I3 with all other column blocks 0, and
Also (i) each column of C
e ij I3 . Specically, the bar connectivity matrix B
e
(ii) for any row i there exists a column j such that C
e form the matrix C
e as follows:
and the string connectivity matrix S
T
e
s
eT
19
S
eT p C p
b
B
6352
where
e 2 R3np 3ns ;
S
e 2 R3np 3nb :
B
20
e D
e ij to be the external force incidence matrix dened by
Similarly, we dene D
e ij I3 if external force vector j enters node i
D
0 if external force vector j is not incident with node i:
21
From network analysis, the law for static equilibrium and linear algebra, we have the following result.
Lemma 6. Consider a tensegrity structure as described by the geometric conditions given by (19). Then the
equilibrium force equations for a tensegrity structure under the external load w are
2
3
t
~ 4
f 5 0; A
~, S
e B
e D
e
A
22
w
or equivalently
et B
ef D
ew
S
23
where
e S I3 ;
S
e B I3 ;
B
e D I3
D
24
s3
s1
6353
w1
b2
s2
s4
b2
b1
s4
s3
s2
b1
s2
s4
s1
w2
s1
p1 p3 s 1 ;
p2 p4 s 3 ;
p2
p3 s 2
p1
p4 s 4
p1 p2 b1 ;
p3 p4 b2 :
25
26
Here nb rankB 2 < np 4. Also, in terms of the stated force convention (17), the conditions for static
equilibrium at nodes 14 are
9
t1 t4 f 1 0
>
>
=
t2
t3
f 1 w1 0
:
27
t1
t2 f 2 w2 0 >
>
;
t3
t4
f 2 0
These static equilibrium conditions and the geometric conditions can be written in the form (19)(24), where
3
2
0
1 0
0
1
1 0 0
h i
6
0
1 0 7
1
1 0 1
~ S
e B
e D
e I3 6 0
7 I3 :
A
4 1
1 0
0 0
1 0
1 5
1 0
0
0
0
1
1 0
Example 9. Consider the 4-bar 8-string planar class 2 tensegrity structure illustrated in Fig. 2. Now the bar
connectivity B and the string connectivity S in (24) are given by
6354
1 0
6 0
1
6
6 1
1
S6
6 0
0
6
4 0
0
0
0
0
1
1 0
0
0
1
1
0
0
0
0
0
0
0
0
1 0
0
0
0
1
1
1
3
0
1 7
7
0 7
7;
0 7
7
1 5
0
1
6 1
6
6 0
B6
6 0
6
4 0
0
0
0
1
1
0
0
0
0
1
0
1
0
3
0
1 7
7
0 7
7:
0 7
7
0 5
1
Here
rankB 4 nb < np 6:
Example 10
(a) A 3-dimensional class 1 tensegrity consisting of
strings is illustrated in Fig. 3. The corresponding
2
1 0
0
0
1
0
1 0
6 0
1 0
0
0
1
0
1
6
6 1
0
1 0
0
0
0
1
6
S6
1
0
1 0
0
0
0
6 0
4 0
0
1
0
1 0
0
0
0
0
0
1
0
1 1
0
S6
6 0
6 1
1 0
1
0
0
0
0 7
0
0
1 7
7
6
7
6
4 0
4 0
0
1 0
1
1 5
0
0
0
0
1 5
0
0
1
0
0
0
0
1
1 1
0
0
6355
Here rankB 5 nb
1 < np 6. Note that since there is exactly one loop fb4 ; b5 ; b6 g of bars, it
follows form Lemma 7, rankB nb
1.
(c) Another class 2 tensegrity structure may be derived from the structure in part (b) by replacing the string
s9 by a bar b7 . In this new structure, nb 7; np 6 and ns 5. Since there are now two independent
loops fb4 ; b5 ; b6 g and fb2 ; b7 ; b5 g of bars, we have from Lemma 7 that rankB nb
2 5 < np 6.
We have the following result.
Theorem 11. Suppose the np nb connectivity matrix B of rank qB has the singular value decomposition
RB1 0 VTB1
B UB1 UB2
28
; q B < nb
0 0 VTB2
where UB1 2 Rnp qB , VB1 2 Rnb qB or
R
B UB1 UB2 B1 VTB ; qB nb
0
29
where UB1 2 Rnp nb , VB 2 Rnb nb . Also, given the tensegrity node vector p, dene the coordinate transformation
e Tq
pP
30
where
e P I3 ;
P
PT PT1 ; PT2
qB np
T
;
P1 R
1
B1 UB1 2 R
31
Then
i(i) In terms of the transformed tensegrity node vector q, the tensegrity geometry is given by
T
e Tq S
e T q ; q VT I3 b
q qTd qTe ; s S
d
1 d
2 e
B1
32
6356
e 1 S1 I3 ;
S
where
S1 P1 S 2 RqB ns ;
S2 P2 S 2 Rnp qB ns :
33
34
D1 P1 D;
35
D2 P2 D:
Proof. By Lemma 7, qB < np which means fUB1 ; UB2 g are well dened. Part (i) follows directly from the
e T q s, B
e with S
e T q b with
denition of P
q
q
e T; S
e T ;
e T S
S
q
1
2
e T VB1 I3 ; 0
B
q
36
e1 P
e2 P
e; S
e2 P
e g and f P
e1S
e2S
e2B
e VT I3 ; S
e 0g. Part (ii) follows from the expansion of
e1B
with f S
B1
e
e
e
the transformed equilibrium force equation S q t B q f D q w, where
h
i
e S
eq B
eq D
eq P
e B
e D
e :
S
For notational simplicity, we assume in subsequent expressions that VB1 VB whenever qB nb . h
3.1. Class 1 structures
In a class 1 tensegrity structure, no two bars are connected, and so np 2nb . Without any loss of
generality, we can then label the nodes of the bar bm to be p2m and p2m
1 . Hence for class 1 tensegrity
structures, we have
bm
p2m
1 p2m ;
m 1; 2; . . . ; nb :
37
Lemma 12. Given the tensegrity node vector p with (37), bar vector b, and string vector s, the geometry of the
class 1 tensegrity structure can be described by the algebraic equations
e T p b;
B
e Tp s
S
38
for some reduced np ns connectivity matrix S. The reduced np nb connectivity matrix B is given by
B Ie
Io
39
where odd and even node selection matrix Io , Ie 2 R3np 3np are dened by
ITo , blockdiagf1; 0; 1; 0; . . . ; 1; 0; 1; 0g
ITe , blockdiagf0; 1; 0; 1 . . . ; 0; 1; 0; 1g:
40
Then in (29)
np 2nb ;
qB np ;
VB I:
41
6357
The following two corollaries give two special choices for transformations.
Corollary 13. From Theorem 11
ii(i) Let P1 , P2 2 Rnb np in (31) be given by
PT PT1 PT2 12 B J
42
43
nb np
are dened by
J , Ie Io :
44
qd b
45
3nb
where pcj ,
1
p2j
2
46
p2j 1 .
47
Ie
48
where B and J are given by (39) and (44) respectively, and odd and even selection matrix Io , Ie 2 R
are given by (40).
(ii) The transformed coordinate q is given by
qT qTd ; qTe ;
qd b
3np 3np
49
3nb
50
Example 15. Eqs. (25), (27) of the 2-bar 4-string tensegrity introduced in Example 8 can be written in the
form (22) and (19) where b in (11) is already in the form (37). Then from Corollary 14, we have
1 0 0
1
1 1
1 1
S1
; S2
1 1 0 0
1
1 1
1
51
1 0 0 0
1 1 0 0
P1
; P2
0 0
1 0
0 0 1 1
or from Corollary
1
1
P1
2 0
1 1
S1
2
1
13, we have
1 1 1 0 0
1 0 0
; P2
0
1 1
2 0 0 1 1
1
1 1
1 1
1
1
1
; S2
:
1 1
1
2 1
1 1
1
6358
52
t C I3 s;
53
C , diagfc1 ; c2 ; . . . ; cns g
S2 I3 t D2 I3 w
54
VTB1 I3 f S1 I3 t
D1 I3 w;
f K I3 b;
VTB1 VB1 I
K , diagfk1 ; k2 ; . . . ; knb g
55
56
57
Beyond equilibrium requirements, one might require shape constraints by requiring p to take on a specic
e T q. However, in this paper, our focus is only to characterize possible
set of values p
p, where p P
equilibria, and so the freedom in choosing the nodal vector p will appear as free variables in the vector qe , as
the sequel shows.
As a result of Lemma 2, conditions for the existence of solutions fqd 2 R3qB ; qe 2 R3np
qB g of (52)(56)
are equivalent to conditions for the existence of solutions fq1 2 RqB ; q2 2 Rnp
qB g of the equations:
sr ST1 q1 ST2 q2
58
tr Csr ;
59
C , diagfc1 ; c2 ; . . . ; cns g
S2 tr D2 wr
VTB1 f r S1 tr
D1 wr ;
f r Kbr ;
60
VTB1 VB1 I
K , diagfk1 ; k2 ; . . . ; knb g:
61
62
6359
(60) can be combined to obtain a unique expression for tr in terms of q1 . This is key to the main results of
this paper.
We now establish necessary and sucient conditions for a solution of equations (58)(62) in the absence
of external forces (i.e. wr 0) by examining each of these equations in turn beginning with the solution of
(58) and (60). The next result follows from Lemma 2.
Lemma 17. Suppose
q2 , rankS2 6 minfnp
qB ; ns g
63
U U1 ; U2 ;
R11
R
0
U1 2 Rnp qB q2 ;
64
0
;
0
V V1 ; V2
U2 2 Rnp qB nb q2 ;
V1 2 Rns q2 ;
V2 2 Rns ns q2 :
Then a necessary and sucient condition for (58) to have a solution q2 2 Rnp
qB is given by
VT2 sr
ST1 q1 0:
65
where z2 2 R
nb q2
66
is arbitrary.
M , VT2 C 1 V2
67
in which case
T
T
1 T
1
q2 U1 R
1
11 V1 C V2 M V2
I3 S1 q1 U2 z2
where z2 2 R
nb q2
68
is arbitrary.
Proof. From (64), (65) and Lemma 2, we have tr V2 zt where zt is the free solution of (60). Then from (59)
VT2 sr
ST1 q1 VT2 C
1 V2 zt
ST1 q1 :
Since V2 has full column rank, the matrix M VT2 C
1 V2 is invertible if it exists (that is, if cn > 0,
n 1; . . . ; ns ). Hence (65) is satised when zt M
1 VT2 ST1 q1 , and this gives (67). h
We now consider the solution of (61) and (62) when wr 0.
Lemma 19. When wr 0, a necessary and sucient condition for (61) and (62) to have a solution q1 2 RqB is
given by
X
VTB1 KVB1 q1 0
69
6360
where
X , S1 V2 M
1 S1 V2 T ;
M , VT2 C 1 V2
70
In particular, dene
qY , rankY;
Y , X VTB1 KVB1 :
71
Then
ii(i) when qY qB , q1 0 is the only solution of (69),
i(ii) when qY 0, any q1 2 RqB is a solution of (69), and
(iii) when 0 < qY < qB , all solutions q1 are given by
q1 VY 2 z1
72
qB
qY
where z1 2 R
is free, and where fUY ; RY ; VY g is the singular value decomposition of the matrix
Y 2 RqB qB ; that is
RY 1 0
UY UY 1 ; UY 2 ; RY
; VY VY 1 ; VY 2 ; qY rankY
73
0 0
with UY 1 2 RqB qY , VY 2 2 RqB qB
qY .
Proof. From (67), (62) and (61)
S1 tr
VTB1 f r S1 V2 M
1 VT2 ST1 q1
VTB1 Kbr X
VTB1 KVB1 q1
where X is given in (70). The result then follows from the singular value decomposition of X after writing
(69) in the form Xq1 VTB1 f r , VTB1 f r VTB1 KVB1 q1 . h
There are many choices of fcj ; kk g which guarantee a solution q1 6 0 of (69). One choice is provided in
the following result.
Corollary 20. If C1 cI, and K1 kI, then all solutions of (69) are characterized by the modal data of the
T
(symmetric) matrix S1 V2 S1 V2 . That is, all admissible values of k=c and q1 are eigenvalues and eigenvectors
T
of S1 V2 S1 V2 .
kIq 0 or X
qIq 0,
Proof. From (69) and C1 cI, K1 kI, X cS1 V2 VT2 ST1 cX. Then cX
1
1
where q k=c. Hence q and q1 are the eigenvalues and eigenvectors of X. h
We now make reference to Lemma 2, part (iv) to relate solutions of (58)(62) as provided in Lemma 19
to solutions of (52)(56).
Theorem 21. Consider a tensegrity structure as dened by the geometry and force equations in the absence of
external load as described by the geometric conditions
BT I3 p b;
ST I3 p s;
t C I3 s;
f K I3 b
K diagfk1 ; k2 ; . . . ; knb g:
6361
Then given any tensile force coecients fcm > 0; 1 6 m 6 ns g, there exist compressive force coecients
fkn > 0; 1 6 n 6 nb g which dene an equilibrium structure, if for some q1 6 0, fK; Cg satisfy the condition:
X
VTB1 KVB1 q1 0
T
X , S1 V2 M 1 S1 V2 ;
M , VT2 C 1 V2 ;
74
where VB1 is given by (28) and (29), S1 is given by (33), and V2 is given by (65).
Moreover, for any qd 6 0 which satises
X
VTB1 KVB1 I3 qd 0;
the nodal vector p is of the form
p PQ I3 zTd ; zTe T
75
76
t V2 M 1 VT2 ST1 I3 qd
b BT I3 p;
f K I3 b:
77
78
Suppose that the new equilibrium structure is assumed to be given by node vector p, bar vector b, string
vector s, compressive force vector f, tensile vector t, compressive force coecient matrix K and tensile force
matrix C as described by (52)(56). As a result of Lemma 2, conditions for the existence of solutions
fqd 2 R3qB ; qe 2 R3np
qB g of (52)(56) are equivalent to conditions (58)(62) for the existence of solutions
fq1 2 RqB ; q2 2 Rnp
qB g. Note that all force coecients together with all node geometry will normally
change. We now seek necessary and sucient conditions for the externally loaded structure to be in geometric and force equilibrium. An extension of Lemmas 18 and 19 gives us the following result.
Theorem 22
i(i) All solutions tr of (60) which guarantee (65) are of the form
tr V2 M
1 VT2 ST1 q1 G1 wr ;
M , VT2 C 1 V2
T
G1 Ins
V2 M
1 VT2 C
1 V1 R
1
11 U1 D2 :
79
(ii) A necessary and sucient condition for (61) and (62) to have a solution b 2 Rnb is given by
X
VTB1 KVB1 q1 Gwr
UT2 P2 wr 0
80
6362
81
Proof. Since (58) is not directly dependent on wr , Lemma 17 applies for wr 6 0. Now consider the solution
of (60) for wr 6 0. A necessary condition for the existence of a solution t is UT2 P2 wr 0, and in this case, all
solutions tr are of the form
T
tr V2 zt V1 R
1
11 U1 D2 wr
for any zt 2 Rnb
r where as in (63), r is the rank of S2 . Now in order that condition (65) is satised, zt must
be selected such that
T
T T
VT2 C
1 V2 zt V1 R
1
11 U1 D2 wr
V2 S1 q1 0:
That is
T
zt M
1 VT2 ST1 q1
M
1 VT2 C
1 V1 R
1
11 U1 D2 wr
UT2 P2 wr 0
where zY 2 Rnb
rb is unknown. The possibility of multiple solutions is interesting; either only one solution is
possible and more information is required to determine zY , or many solutions are possible. In the latter
case, the particular equilibrium obtained will then depend on way in which the external load wr is introduced. The structural implications of the null space condition UT2 P2 wr 0 on the external load wr would
then also require a physical interpretation.
The existence of an equilibrium solution however requires the second condition in (80) on the external
force wr to be satised. In this regard, we have the following result.
Lemma 23. For all structures fS; Bg, the nb
r ns matrix product UT2 P2 is of the form
UT2 P2 e1; 1; . . . ; 1
82
wk 0:
83
k1
Proof. It follows from Lemma 2 and svdS2 in Lemma 17 that UT2 S2 0 which from (33) implies
UT2 P2 S 0. Now from Lemma 6, each column of S has exactly 1 and exactly )1 with all other column
elements 0. Furthermore, for every ith row of S, there exists a column j such that the ijth component of S is
1. These properties of S then imply that UT2 P2 is of the form (82) for some vector e, and so
UT2 P2 wr e
np
X
6363
wk :
k1
e T Z1 ; Z2 ; . . . ; Zn it follows from
e 2 is partitioned into the block form U
Now if the full row rank matrix U
2
b
(42) that
e TP
e
U
2 2 Z1 ; Z1 ; Z2 ; Z2 . . . ; Znb ; Znb
e TP
e
which then guarantees that U
e is invert2 2 also has full row rank, and consequently that the matrix ~
ible. h
Condition (83) expresses the requirement that for an externally loaded tensegrity structure to be in force
and geometric equilibrium, it is necessary (but not sucient) that the sum of the external forces is zero.
5. Computational algorithm for equilibria
One procedure for construction of a tensegrity structure in equilibrium is provided as follows.
Step 1. Given the connectivity matrices S and B from the network topology, nd a nonsingular matrix
P PT1 ; PT2 such that BTq BT P VTB1 ; 0nb , and calculate fS1 P1 S; S2 P2 Sg.
Step 2. Choose fcm > 0g and fkn > 0g such that detX
VTB1 KVB1 0.
Step 3a. Select suitable zd and compute qd by
qd VY 2 I3 zd :
Step 3b. When the bar connectivity matrix B has full rank qB nb (i.e. no loops of bar vectors), then one
can select suitable zd and compute b by
b VB1 VY 2 I3 zd :
Step 4. Select ze and compute the node vector p from (75).
Step 5. Compute ft; s; f; bg from
s C
1 I3 t;
b BT I3 p;
f K I3 b
t V2 M VT2 ST1 I3 qd D1 I3 w
T
D1 Ins
V2 M
1 VT2 C
1 V1 R
1
11 U1 P2 :
84
6. Illustrated examples
We now illustrate the construction procedure for a simple tensegrity structure.
Example 24. A general force conguration for the class 1 tensegrity structure in Example 8 with
w1 w2 0 will be investigated in this section. Suppose the force coecient matrices are given by
K diagfk1 ; k2 g and C diagfc1 ; c2 ; c3 ; c4 g.
Step 1. The connectivity matrices S, B and the coordinate transformation P PT1 ; PT2 are given in
Example 8. Since B is of full column rank matrix and V2 spans the null space of S2 , we compute VB1 I2
and
6364
1
61
V2 6
40
0
1
0
1
0
3
1
07
7:
05
1
Step 2. Choose fcm > 0g and fkn > 0g such that detX
VTB1 KVB1 0, where
1
c2 c3 c4 c1
c3 c1 c2 c4
1
X S1 V2 VT2 C
1 V2 VT2 ST1
:
c3 c1 c2 c4
c3 c4 c2 c1
c1 c2 c3 c4
In all choices for fkn g that led to the rank of X
K having rank 1, the 4 2 matrix VY 2 is of the form
VTX 2 AT1 ; AT1 in (72); that is the two bar vectors fb1 ; b2 g are always parallel, so the equilibrium structure
is one dimensional with fp1 p3 ; p2 p4 g. Hence for a two-dimensional structure, X
K must have rank
zero. This requires
c3 c1
c2
c2 c3 c4 c1
k1
c2 c3 c4 c1
c3 c4 c2 c1
k2
c2 c3 c4 c1
c4
85
where c1 , c2 , and c3 are free positive constants. We choose fck 1; k 1; 2; 3g and then C I4 . It follows
that X I2 and K I2 satisfy condition (74) in Theorem 21.
Step 3b. When the bar connectivity matrix B has full rank qB nb (i.e. no loops of bar vectors), then one
can select suitable zd and compute b by b VB1 VY 2 I3 zd . Since we choose c and k such that X
K 0,
T
T
VY 2 I; that is the bar vector is arbitrary. Let us choose b1 2; 0 , b2 0; 2 .
Step 4. Select ze and compute the node vector p from (75).
The nodes fp1
1; 0T ; p2 1; 0T ; p3 0;
1T ; p4 0; 1T g dene an equilibrium solution from (75)
setting ze is zero. When we set ze 1; 1T , we obtain the nodal vector
0:6464
1:3536
0:3536
0:3536
p1
; p2
; p3
; p4
:
0:3536
0:3536
0:6464
1:3536
T
This choices of ze only translate the geometric center of the structure from 0; 0 to 0:35361; 1 , since all
force coecients and b have been specied.
Step 5. Compute ft; s; fg from (84).
Example 25. Consider the 3; 9; 3 class 1 tensegrity structure dened in Example 10. A symmetrical force
conguration will be investigated with equal bar force coecients fk1 k2 k3 kg, equal base string
force coecients fc1 c2 c3 , cb g, equal top string coecients fc4 c5 c6 , ct g, and equal vertical
string coecients fc7 c8 c9 , cv g. Then
1
1
X1
2c2b 8cb ct 6cb cv 6cv ct 2c2t 3c2v c3t c3b c3v
where
2
3
ct cv cb c2v 3cv ct 3cb cv 4cb ct c2b c2t
X1 4
cb cv ct
cb c2t
4c2b ct
3c2b cv
c3b cv c2t 2ct c2v c3v 5
4cb c2t
cb cv ct
c2b ct 2cb c2v c2b cv
3cv c2t
c3t c3v
X2
X3 ;
6365
3
4cb c2t
cb cv ct
c2b ct 2cb c2v c2b cv
3cv c2t
c3t c3v
5:
X3 4
3cb c2t
cb cv ct
cv c2t ct c2v
3c2b ct cb c2v
c2b cv c3v
2
2
2
2
3
3
2
2
6cv ct 5ct cv 10cb cv ct 7cb ct 4cb ct ct cv 3cb cv 2cb cv
Now we need to choose the force coecients such that
detX
VTB1 KVB1 0:
Since bar connectivity matrix is of full column rank, VB1 I and
detX
K
2c2b
k
cv
ksecond 0
8cb ct 6cb cv 6cv ct 2c2t 3c2v
where
ksecond ,
2c3t k 2c2v c2t 2cv c3t 6c3b ct 15c2b c2t 2c2b c2v 6cb c3t 2c3b cv
6c2v ct k
10cv c2t k
16c2b ct k
2c2b k2
2c3b k
10c2b cv k 6cb k2 cv
6cb kc2v 8cb k2 ct
16cb kc2t
22cb kcv ct
2k2 c2t 3k2 c2v 6k2 cv ct 16cb c2t cv 8cb ct c2v 16c2b cv ct :
Since smaller rank of X
K yields more freedom for the choice of b, we choose cv k. Next evaluating
the second term when cv k, we have
ksecond jcv k
6k2 c2t
6c2b k2 6cb c3t 6c3b ct 15c2b c2t
6cb k2 ct 3k4
k2
ct cb
2c2b k2
ct cb
2c2t :
p
p
We conclude k ct cbp
chosen in this case. When ct cb 1 and k cv 3, the structural shape is the prism given in Fig. 3.
7. Conclusion
This paper characterizes the static equilibria of tensegrity structures. Analytical expressions are derived
for the equilibrium condition of a tensegrity structure in terms of member force coecients and string and
bar connectivity information. We use vectors to describe each element (bars and tendons), eliminating the
need to use direction cosines and the subsequent transcendental functions that follow their use. By enlarging the vector space in which we characterize the problem, the mathematical structure of the equations
6366
admit treatment by linear algebra methods, for the most part. This reduces the study of a signicant portion
of the tensegrity equilibria to a series of linear algebra problems. Our results characterize the equilibria
conditions of tensegrity structures in terms of a very small number of variables since the necessary and
sucient conditions of the linear algebra treatment has eliminated several of the original variables. This
formulation oers insight and identies the free parameters that may be used to achieve desired structural
shapes. Since all conditions are necessary and sucient, these results can be used in the design of any
tensegrity structure. Special insightful properties are available in the special case when one designs a
tensegrity structure so that all strings have the same force per unit length (c), and all bars have the same
force per unit length (k). In this case, all admissible values of k=c are the discrete set of eigenvalues of a
matrix given in terms of only the string connectivity matrix. Furthermore the only bar vectors which can be
assigned are eigenvectors of the same matrix. Future papers will integrate these algorithms into software to
make these designs more ecient.
Acknowledgements
Part of this research was conducted by Dynamic Systems Research Inc. under the sponsorship of the Jet
Propulsion Laboratory, technical monitors Dr. Marco Quadrelli and Fred Hadaegh.
References
Barnes, M.R., 1998. Form-nding and analysis of prestressed nets and membranes. Computers & Structures 30 (3), 685695.
Connelly, R., 1982. Rigidity and Energy. Inventiones Mathematicae 66 (1), 1133.
Connelly, R., 1993. Rigidity. In: Gruber, P.M., Wills, J.M. (Eds.), Handbook of Convex Geometry. Elsevier Publishers Ltd,
Amsterdam, pp. 223271.
Connelly, R., 1999. Tensegrity Structures: Why are they stable. In: Thorpe, M.F., Duxbury, P.M. (Eds.), Rigity Theory and
Applications. Plenum Press, New York, pp. 4754.
Desoer, C.A., Kuh, E.S., 1969. Basic Circuit Theory. McGraw-Hill, New York.
Horn, R.A., Johnson, C.R., 1985. Matrix Analysis. Cambridge University Press.
Ingber, D.E., 1993. Cellular tensegrity: dening new rules of biological design that govern the cytoskeleton. Journal of Cell Science 104,
613627.
Ingber, D.E., 1997. Tensegrity: the architectural basis of cellular mechanotransduction. Annual Review of Physiology 59, 575599.
Ingber, D.E., 1998. Architecture of Life. Scientic American. pp. 4857.
Kenner, H., 1976. Geodesic Math and How to Use It. University of California Press, Berkeley, California.
Linkwitz, K., 1999. Form nding by the direct approach and pertinent strategies for the conceptual design of prestressed and hanging
structures. International Journal of Space Structures 14 (2), 7387.
Motro, R., 1984. Forms and forces in tensegrity systems. In: Nooshin, H. (Ed.), Proceedings of Third International Conference on
Space Structures. Elsevier, Amsterdam. pp. 180185.
Motro, R., 1990. Tensegrity systems and geodesic domes. International Journal of Space Structures 5 (34), 341351.
Motro, R., 1992. Tensegrity systems: the state of the art. International Journal Space Structures 7 (2), 7583.
Motro, R., 2001. Foldable Tensegrities. In: Pellegrino, S. (Ed.), Deployable Structures. Springer Verlag, Wien-New York.
Motro, R., Belkacem, S., Vassart, N., 1994. Form nding numerical methods for tensegrity systems. In: Abel, J.F., Leonard, J.W.,
Penalba, C.U. (Eds.), Proceedings of IASS-ASCE International Symposium on Spatial, Lattice and Tension Structures, Atlanta,
USA. ASCE, New York. pp. 707713.
Pugh, A., 1976. An Introduction to Tensegrity. University of California Press, Berkeley, California.
Schek, H.J., 1974. The force density method for form nding and computation of general networks. Computer Methods in Applied
Mechanics and Engineering 3, 115134.
Snelson, K., 1996. Snelson on the tensegrity invention. International Journal of Space Structures 11 (1 and 2), 4348.
Skelton, R.E., Sultan, C. 1997. Controllable tensegrity. In: Proceedings of the SPIE 4th Annual Symposium of Smart Structures and
Materials, San Diego, March.
Skelton, R.E., Helton, W.J., Adhikari, R., Pinaud, J.P., Chan, W., 2001a. An Introduction to the Mechanics of Tensegrity Structures,
Handbook of Mechanical Systems Design. CRC Press, Boca Raton, FL.
6367
Skelton, R.E., Pinaud, J.P., Mingori, D.L., 2001b. Dynamics of the shell class of tensegrity structures. Journal of the Franklin Institute
338 (23), 255320.
Tibert, G., Pellegrino, S., 2001. Review of form-nding methods for tensegrity structures. In: Tensegrity Workshop, Rome, May.
Vassart, N., Motro, R., 1999. Multiparametered form.nding method: application to tensegrity systems. International Journal of Space
Structures 14 (2), 147154.
Williamson, D., Skelton, R.E., 1998a. A general class of tensegrity systems: geometric denition, engineering mechanics for the 21st
century. In: ASCE Conference, La Jolla, California, May.
Williamson, D., Skelton, R.E., 1998b. A general class of tensegrity systems: equilibrium analysis, engineering mechanics for the 21st
century. In: ASCE Conference, La Jolla, California, May.