Vous êtes sur la page 1sur 5

About Some Tensegrity

Structure s
R. G. Selfridge
Computer and Information Sciences, University of Florida, Gainesville, FL 32611
Selfridg@cise.ufl.edu
(Received 9th May 2000, revised version received 27th March 2001)
Abstract: Tensegrity structures are structures where there are, typically, a few
struts and many links, which are nevertheless stable structures. This paper
considers structures where the platform and base are regular planar n-gons,
with the appropriate connections and derives a very simple formula for their
realizeable/unique construction. The effects of changing a single length can
also be studied.

INTRODUCTION
Tensegrity structures started with studies by
Buckminster Fuller, and are built with struts and links,
with the links usually elastic. A principle
characterization of these tensegrity structures is that
there are a relatively small number of struts with all
other members being (elastic) cables while
maintaining structural rigidity. We consider here only
structures with n-gon platforms and bases. Most
previous studies, especially those of Kenner [2], start
with a prism of a platform and base, with a specified
height. The platform is then rotated, relative to the
base and struts and elastic links inserted as
connections. In this frame the structure then rotates to
an equilibrium position, and the concern is for the strut
and link lengths in that position. It turns out that using
minimum energy, or shortest link length, will result in
the same position as our analysis, but that is incidental
to our study. Our study starts with the final structure
and determines the measurements for unique
realizeability, which in turn provides for links and
struts, or the tensegrity position. Here the parameters
are the edge lengths, strut lengths and link lengths. We
derive a very simple formula as necessary and
sufficient condition for a tensegrity position, while
the number of sides ( 3) is irrelevant to the formula.
The symbol manipulation software also permits
changing single values to show what happens to the
structure, and can permit non-planar platforms and/or
bases.
We consider a structure with a symmetric (planar)
International Journal of Space Structures Vol. 16 No. 4 2001

n-gon platform, of edge length T, and a symmetric


(planar) n-gon base (where n 3) that is inscribed in a
circle of radius 1. Each base vertex is connected to two
platform vertices, as in fig 1 (which is simplified down
to a triangle). There is clearly no loss of generality in
requiring the base to be inscribed in a circle, this
serves only to normalize the various lengths. In the
frame of tensegrity structures the legs split easily into
two groups, one set of which we call R, the other set
we call L. As we go to n-gons with n > 3 there are
quite naturally more and more legs in each group.
Ultimately, as we create the structure one group of
struts will only push, these we identify with R, the
other group will only pull, and are links, identified as
L. In the general case examined in a later section these
legs must remain rigid struts with no push or pull
assumptions. However the initial analysis assumes a
rotational symmetry of the base and platform, and a

Figure 1
231

About Some Tensegrity Structures

single size R, single size L. Such assumptions quite


naturally guarantee that our structure has the base and
platform parallel. In contrast to most earlier studies we
also note that the links are not elastic here, since no
motion or changing of the structure is intended in
reaching our conclusions. Rather the links are shown
to be non-extendable connections (physically
cables), and the struts are shown to be noncompressible (physically just rods).

The symmetry of this structure now allows us to


locate the next vertex of the platform

DERIVATION OF EQUAT IONS

If we now do the required symbolic operations,


with any simplifications from S2 + C2 = 1 we have

We use u 1 as the starting parameter, as in fig 2, noting


that with the proper choice of u 1 we can get any
desired n-gon. Since we will be using many angles we
first introduce the notation

We now can write, for a base vertex

and we have the equation

Which is written as
(1)

(2)
The primary difficulty with (2) is handling C3 and
S 3, so we use a very common tan half angle
substitution
and get

where we assume the origin is the centre of the base


circle.
We can rotate the base through the appropriate
angle u 4 = p 2u 1 with the matrix

where we actually use the proper expressions in S1 and


C1. We also note
(we
use o for the vector/matrix multiplication).
We define u 2 as the angle formed by a base edge, a
link and a rod, as in fig 3, and u 3 as the angle between
that triangle and the base plane.
We can now locate vertex P1 of the platform with

(3)
The symmetry thus generates a biquadratic in x, so
that there is a symmetry across the base plane, as was
to be expected.
We now find the conditions that there be a single
root of this quadratic polynomial by setting the
discriminant1 to zero. There are terms that can be
divided out, and since S 1 only occurs with the
exponent 2, we can replace it by

and get the (simplified) discriminant of the quadratic


(3) to be
(4)

If we now reference fig 3 we have

and if we insert this in (4) we get, as the desired result,


(5)

1The

Figure 2
232

discriminant of ax2+bx+c is b24ac. If the discriminant is


negative there are only complex roots, so varying a, b or c can
drive solutions complex. In this study we usually have a > 0,b < 0
and c > 0.

International Journal of Space Structures Vol. 16 No. 4 2001

R. G. Selfridge

In particular we point out that this result is


independent of u 1, so that the number of sides of the ngon has no effect on the equation (but remember that as
n changes the base changes slightly in the unit radius
circle and the platform grows rapidly). We also add that
if R increases the discriminant goes negative and there
are only complex solutions. Thus R can be a strut that
cannot increase and still have a physically realizeable
structure. Similarly if T or L decrease the structure also
becomes complex and we can replace them with a nonextendable strut (or link). We thus have that (5) is the
necessary and sufficient condition that the structure of
fig 1 can be built with links for all L and struts for all
R. Both the platform and base can also be built with
links, since decreasing T will force all solutions
complex, while if we decrease the base multiplicatively
by (1 e ) the net effect on (5) is as if T were multiplied
by (1 e ), again driving all solutions complex.
While these results are wide ranging it should be
noted that a bad choice of R, T and L can result in a
structure that is realistically quite unstable.

MORE GENERAL CASES


We treat first the case of a triangle platform/base.
Using the notation of fig. 4, and techniques similar to
those above, we can find

We can also allow the base to be hexagonal, planar


or non-planar and continue to find 16 solution sets.
If we now move to, say square platforms and bases
we cannot usually change just one measurement
because the structure is severely over-constrained. The
structure is spatially fixed with any one choice of
triangle pairs so there must be a careful decision about
which lengths change and how in order to avoid
instant collapse.
We will, however, consider the case where we have
a square base in the ground plane, and a square
platform, as in fig 5. If we set R = 3 and T = 1.5 we
have from (5) L = 6. Let us now set R22 = 3 + d and
hold all other lengths fixed, where we let d be positive
or negative and small. Since the base is planar we still
have symmetry across the base plane so all solutions
appear above and below with 8 effectively different
solutions.
We ignore P4 and its connections at first in order to
find a solution, thus using only the edges (B4, B1), (B1,
B2) and (B2, B3) and P1, P2 and P3. We find all 16
solutions but all have significant imaginary part with
d = .01. As d gets smaller the imaginary part of the

where u 4 is the angle between the triangle L1, R1, (B2


B1) and the base plane. Similarly we have
where we use u 2 and u 3 as the angles between the
relevant triangles and the base plane. Using P3 with
P2, and P1 with P3 we get 3 equations. If we now
associate
Figure 3

in tan half-angle substitutions similar to that use for


(3), we have three polynomials P(x, y), Q(y, z), and
R(z, x) each of degree 2 in each variable (and other
relevant lengths). Such a system has 16 roots in
general, though in this problem there is symmetry
across the base plane.
If we keep all the variables symbolic it is possible
to reduce this to a single symbolic polynomial in one
of the variables, though the number of expressions in
this polynomial is large (of order 10K).
In such a framework it becomes possible to change
one length and discover what happens to the structure.
For a large range of choices of R, L and T and triangles
we can show that use of links and struts is still valid.
International Journal of Space Structures Vol. 16 No. 4 2001

Figure 4
233

About Some Tensegrity Structures

solution that holds at the tensegrity point gets smaller


and smaller, with all other solutions having a
significant imaginary part. We provide some numbers
in appendix A.
If, however, we change both R22 and L12, we can
find a realizeable solution. The platform is now nonplanar. It is just that changing only one length
(whether a strut or a link) usually drives all solutions
complex.
Let us now consider the case of a square base, the
same groups of legs, Li = L, and Ri = R, while requiring
the platform to have the same edge lengths, Ti = T but
no further requirement on the platform. As was done for
the triangle case we now have four variables, w, x, y and
z, for the four angles u i. We have four polynomials, each
of degree 2 in two of the variables (and total degree 4),
which can be reduced to 2 equivalent polynomials H (w,
y) and G(w, y). H and G are both degree 4 in w and y,
but total degree 8 for H and 6 for G. We have, further
that G is divisible by w2 y2. As a result the solution set
for this case has 8 roots when w = y, 8 when w = y,
and 16 more roots, for a total of 32 (do not forget the
symmetry across the base). There are 397 expressions
(ie some coefficient and a product of the relevant
symbols) in H, and 124 in G.
If we now tackle the case of a pentagon
platform/base, where we keep the lengths of
links/struts and platform edges all as before, we find
64 solutions total (reflected across the base plane).
The symbolic single polynomial is enormous, though
findable in principle. A numeric solution for specific
lengths can be done but is subject to significant
computational instability. Of course the tensegrity
solution still involves a simple discriminant and is
easily handled.

APPENDIX A
The base is a regular square inscribed in a unit circle,
so the four base vertices are at
.707106, .707106, 0
T = || P1 P2 || = 1.5
Rii = 3
Lij = 6 = 2.44949
With these conditions the only real (realizeable)
solutions are
w = x = y = z = .7404804, as the relevant tan half
angle substitutions.
Note that we do not compute the relative twist of
platform and base, but rather use w, x, y and z as the
angle between an edge plane (for example P2, B1, B2)
and the base plane. If we take these results and locate
the platform in space we get the corners are at
1.06066, 0, 2.3184 and
0, 1.06066, 2.3184
the expected result for the tensegrity solution for
height 2.3184, strut length 3.
If we now set R22 = 3.01, and hold all the others we
have the comparable solutions
x = .7334 + .00215i, y = .74018 + .00216i, z =
.747605 + .00217i and all other solutions (except the
reflection of the above in the base) with imaginary
parts greater than .5.
If we are prepared to let the platform be non-planar
then we can choose
R22 = 3.06168 and L12 = 2.52641
Now we have a realizeable solution where P1, P3
and P4 are in the tensegrity solution positions, and P2
moves up. We have, for this case,
w = x = z = .7404804, y = .74787184
Now let us consider a square base. We set the
platform to have an edge T = 2.75, and choose L =
1.75, R = 3. In this case there is more than one solution
with the platform parallel to the base. We provide the
platform vertices, noting in this case that the platform
is square in both cases. These are not the tensegrity
solutions.
P1 = 1.484
.5726
p2 = .5726 1.848
p3 = 1.484
.5726
p4 = .5726 1.484

Figure 5
234

.5956 | 1.484 .5726


.5956 |
.5726 1.484
.5956 | 1.484
.5726
.5956 | .5726 1.484

1.6264
1.6264
1.6264
1.6264

On the other hand if we use T = 2.25 there are still


two realizeable solutions, one is a square, the other is
saddle shaped (non-planar).

International Journal of Space Structures Vol. 16 No. 4 2001

R. G. Selfridge

P1 =
p2 =
p3 =
p4 =

1.484
1.256
1.484
1.256

1.256
1.484
1.256
1.484

1.662 | 1.484
.4653 .8250
1.662 | .4653 1.484 .825
1.662 | 1.484
.4653 .8250
1.662 |
.4653 1.484 .825

BIBLIOGRAPHY
1

International Journal of Space Structures, v7,#2,1992.

International Journal of Space Structures Vol. 16 No. 4 2001

2
3
4

This is an edition devoted entirely to tensegrity


structures.
Hugh Kenner, Geodesic Math and how to use it, 1976,
Univ Calif Press
B. Roth, W. Whiteley, Tensegrity Frameworks, Trans
Am. Math. Soc v265, #2, 1981 p419446
Anthony Pugh, Introduction to Tensegrity, Univ Calif
Press, 1976

235

Vous aimerez peut-être aussi