Tensegrity

What tensegrity actually is

Walk up to a Kenneth Snelson sculpture and the first reaction is disbelief: a cluster of polished aluminium tubes hangs in mid-air, none of them touching, apparently defying the obvious requirement that a structure needs its solid parts to lean on one another. Look closer and you find thin steel cables running between the tube ends. Those cables are doing all the connecting. Every tube — every strut — is an isolated island of compression suspended inside a single continuous, taut network of cables. That arrangement is tensegrity, a contraction of tensional integrity.

The definition that engineers use is precise. A tensegrity is a structure in which (1) the compression members are discontinuous — no strut touches another strut — while (2) the tension members form a continuous connected network, and (3) the assembly is rigid by virtue of self-stress, an internal state of prestress that exists with the structure standing alone, carrying no external load at all. Snelson called this "continuous tension, discontinuous compression," which is the cleanest one-line description of the mechanics. Remove the prestress and the whole thing collapses into a loose tangle of sticks and string. Add it back and the tangle springs into a stiff, repeatable, load-bearing shape.

That third clause is the surprising one. A bridge truss is rigid because its members are riveted into triangles that cannot change shape. A tensegrity has no rigid triangles of solid members — its struts float free — so by the usual rules it ought to be a mechanism, something that flops between shapes like a four-bar linkage. The prestress is what removes the flop. This is the whole intellectual content of the idea, and it is worth slowing down to see exactly how it works.

Why prestress turns a floppy mechanism rigid

Count the freedoms. The simplest free-standing tensegrity is the three-strut "T-prism": three struts and nine cables (three on the top triangle, three on the bottom, three diagonals). Each strut has two end nodes, so there are six nodes, giving 6 × 3 = 18 nodal coordinates in space. Subtract the 6 rigid-body motions of the whole object and you have 12 internal degrees of freedom. The structure has 12 members. By Maxwell's counting rule a just-rigid frame would need members to exactly cancel the freedoms — but tensegrities are deliberately built to be under-braced: they have both a deficiency of members (leaving an infinitesimal mechanism, a direction in which nodes can move with, to first order, no change in member length) and a redundant load path (a self-stress state, a set of member forces in equilibrium with no external load). Calladine's 1978 extension of Maxwell's rule formalises this: a structure can simultaneously possess s independent self-stress states and m independent mechanisms.

The genius is that the self-stress stiffens the mechanism. Picture the infinitesimal mechanism as a soft sideways motion of one node. To first order it costs no cable stretch. But to second order it does — pushing the node sideways forces the cables to lengthen slightly, and if those cables are already pretensioned to force T, that motion is resisted by a restoring force proportional to T. The stiffness of that soft mode is not the cable's elastic stiffness EA/L at all; it is the geometric stiffness, proportional to the prestress force divided by length, T/L. This is exactly why a slack guitar string offers no resistance to a sideways pluck but a tight one snaps back hard: tension buys transverse stiffness. Pretension every cable enough and every infinitesimal mechanism is "pre-stiffened" into a genuine, if soft, structural mode. The condition that makes a tensegrity stand is therefore not "enough members" but "the right prestress, in equilibrium, large enough to stabilise all mechanisms."

Two consequences fall straight out of this. First, a tensegrity's stiffness scales with how hard you prestress it — you can tune rigidity by turning a turnbuckle, with no change in geometry. Second, the prestress sets a hard load ceiling: the moment an external load drives any cable's tension to zero, that cable goes slack, its geometric stiffness vanishes, the mechanism it was stabilising is released, and the structure can lose rigidity abruptly. Designers therefore size the prestress so that the least-loaded cable stays in tension under the worst load case, with margin.

The governing relations, with numbers

At every node, the sum of member forces plus external load is zero. Writing this for all nodes gives the equilibrium equation in terms of the force-density q = t/L (member tension divided by length) of each member:

A · q = f          (nodal equilibrium, A = equilibrium matrix)
For self-stress:   A · q = 0,  q ≠ 0   →  q lies in the null space of A
Number of self-stress states  s = m_members − rank(A)
Number of mechanisms          m = 3·n_free_nodes − rank(A)
Stability requires the prestress to make the
tangent stiffness  K = K_E + K_G  positive-definite
on every mechanism, where K_G ∝ q.

The force-density method (Schek, 1974) exploits the fact that if you fix the q values, the equilibrium equation becomes linear in the node coordinates — so you can solve directly for a self-stressed shape. This turned tensegrity "form-finding" from trial-and-error into linear algebra and is the workhorse of every modern cable-net and tensegrity design tool.

A concrete sizing example makes the geometric-stiffness point tangible. Take a single horizontal cable of length L = 4 m, cross-section A = 50 mm², steel E = 200 GPa, pretensioned to T = 20 kN (about 400 MPa stress, well below yield). Pull its midpoint sideways. The transverse stiffness it provides is roughly:

Transverse (geometric) stiffness ≈ 4·T / L  = 4·20,000 / 4   = 20,000 N/m
For comparison, axial elastic stiffness EA/L  = (200e9·50e-6)/4 = 2,500,000 N/m

The cable is roughly 125 times stiffer along its axis than across it. That mismatch is the whole story of tensegrity: it is enormously stiff in the cable directions and comparatively soft in the prestress-stabilised mechanism directions. Good designs orient the stiff directions toward the working loads and accept that the soft modes will show up as the structure's lowest natural frequencies and largest deflections.

Classes, modules and how they tile

Tensegrities are organised by how many struts meet at a node. In a Class 1 tensegrity (the "pure" kind Snelson built) no two struts share a node — every strut end connects only to cables, so the struts truly never touch. In a Class 2 tensegrity two struts may share a node, which makes assembly easier and stiffness higher at the cost of the visual "floating" purity. Most deployable masts are Class 2 because the shared nodes give predictable hinge lines for folding.

• Three-strut prism (T3 / simplex). The minimal free-standing module: three struts, nine cables, a single self-stress state and a single twist mechanism. Stacking and rotating these prisms builds a tensegrity mast or tower.
• Six-strut icosahedron (expanded octahedron). Six struts, twenty-four cables, highly symmetric, the canonical "tensegrity sphere." This is the form most often used as a structural cell in biotensegrity models and as the chassis of rolling tensegrity robots.
• Tensegrity mast / tower. A vertical stack of prisms, each rotated relative to the one below. Snelson's 18 m Needle Tower (1968, Hirshhorn Museum) is the iconic example — it tapers upward, every aluminium tube visibly disconnected from every other.
• Cable dome. The large-span structural workhorse: a continuous outer compression ring, radial ridge and diagonal cables, a continuous inner tension hoop, and short vertical compression posts that never touch one another. It is a tensegrity in the strict sense — discontinuous compression posts, continuous tension net — scaled to stadium size.

Where tensegrity is actually built

• Cable-dome stadium roofs. David Geiger's roof for the 1988 Seoul Olympic Gymnastics Arena (120 m diameter) was the first large cable dome; Matthys Levy's triangulated "Tenstar" version roofed the 1996 Atlanta Georgia Dome at roughly 235 m — among the largest column-free spans ever built, at a steel weight far below an equivalent space-frame truss. The La Plata stadium (Argentina) and several arenas worldwide use the same family.
• Deployable space booms. Coilable and tensegrity-derived masts (ABLE/Northrop ADAM, SAILMAST) pack a 60 m boom into a canister under a metre tall, then self-erect on orbit driven by the stored strain energy of the prestressed network. They deploy solar arrays, sails, and instrument booms because a prestressed cable structure has essentially no buckling length until it is extended.
• Tensegrity robots. NASA Ames' Super Ball Bot is a six-strut tensegrity rover concept that can be dropped onto a planetary surface, surviving impact because the load spreads through the whole prestressed net, then rolls and crawls by changing cable lengths with motors. The same principle drives a growing field of soft and "morphing" tensegrity robots.
• Footbridges and towers. The Kurilpa Bridge in Brisbane (2009, ~470 m total) is widely described as the world's largest tensegrity-influenced bridge, using a hybrid of masts, spars, and cables. Pure self-stressed footbridges remain rare because of the stiffness limitation.
• Biotensegrity. Donald Ingber's model treats the living cell's cytoskeleton as a tensegrity: rigid microtubules act as compression struts, the actin and intermediate-filament network as continuous tension, and prestress sets the cell's shape and mechanical response. The same lens is applied to the musculoskeletal system, where bones are the compression islands and the fascia/muscle/tendon network is the continuous tension.
• Art and furniture. Snelson's lifetime body of work, plus the viral "floating table" / "impossible table" tensegrity desk gadgets, which are minimal three-cable Class-1 modules where one cable provides the seemingly impossible suspension and two more provide rotational stability.

Tensegrity versus the alternatives

The honest way to evaluate tensegrity is against the structural systems it competes with for a given job: a braced truss, a suspension/cable-stay system, and a geodesic frame.

Property	Tensegrity	Braced truss	Suspension / cable-stay	Geodesic dome
Compression members touch?	No — discontinuous islands	Yes — shared joints	Towers continuous to ground	Yes — continuous shell of struts
Rigid without prestress?	No — needs self-stress	Yes	Partly (towers yes, cables no)	Yes
Self-weight per span	Very low	High	Low	Low–medium
Stiffness / deflection	Low — large deflections	High	Medium (deck adds bending)	High
Deployable / packable?	Excellent	Poor	Poor	Medium (foldable variants)
Load ceiling set by	Prestress (cable slackening)	Member yield/buckling	Cable strength	Strut buckling
Fabrication complexity	High (form-finding, tolerances)	Low–medium	Medium	Medium
Typical use	Art, deployables, cable domes	Bridges, buildings, towers	Long-span bridges, roofs	Domes, radomes, shelters

Read across the table and the niche is obvious. Tensegrity wins decisively where mass and packability dominate — orbiting booms, planetary landers, very-long-span lightweight roofs — and loses where stiffness, simplicity, or low cost dominate, which is most ordinary building structure. That is why the canonical successes are space hardware, stadium cable domes, and sculpture, rather than office floors and highway bridges.

Failure modes and design trade-offs

• Cable slackening (the dominant failure). When an external load pushes any cable's tension to zero the cable can no longer provide geometric stiffness; the mechanism it stabilised is released and the structure can lose stiffness or snap to a new configuration. Prevention: set prestress so the least-loaded cable retains tension under the worst load combination, with a safety factor on the residual tension.
• Low natural frequency / vibration. The prestress-stabilised soft modes are the lowest modes of the structure. Wind, footfall, or machinery can excite them, and the small inherent damping of cables makes resonance a real problem — pedestrian comfort is a recurring issue for tensegrity footbridges. Tuned mass dampers are often added.
• Prestress loss over time. Cables creep and relax, anchorages bed in, and temperature swings change tension (a 30 K drop can add tens of percent to cable force in a restrained net). The self-stress must be designed with allowance for relaxation and, in long-lived structures, monitored and re-tensioned.
• Strut buckling. The struts are the compression members; in a lightweight tensegrity they are slender and Euler buckling, not yielding, governs their size. Raising prestress (which raises strut force) to gain stiffness simultaneously pushes the struts closer to buckling — the central design tension of the system.
• Fitting fatigue and stress concentration. Every cable terminates in a high-tension swaged or threaded fitting, and these end fittings are the usual fatigue hot-spots under cyclic load. Fatigue detailing of the terminations, not the cable mid-length, sets the cyclic life.
• Sensitivity to fabrication tolerance. Because rigidity comes from a delicately balanced self-stress, small length errors in members redistribute prestress unevenly — one cable ends up over-tensioned, another nearly slack. Tensegrities demand tighter member-length tolerances than ordinary frames, and adjustable turnbuckles to trim the net after assembly.

Common pitfalls when designing a tensegrity

• Confusing "rigid" with "stiff." A correctly prestressed tensegrity is rigid (it has a definite shape) but may still be very flexible (it deflects a lot). Sizing only for strength and forgetting deflection and natural frequency is the classic mistake.
• Designing geometry before form-finding. You cannot pick node positions arbitrarily and expect a valid self-stress to exist. Run force-density or dynamic-relaxation form-finding first; the equilibrium shape is an output, not an input.
• Ignoring the assembly mechanism. Class-1 tensegrities are notoriously hard to assemble because the structure is a floppy mechanism until the last cable is tensioned. Plan a tensioning sequence and temporary supports, or choose a Class-2 layout with shared nodes.
• Under-prestressing. Too little prestress leaves soft modes near-zero stiffness and the structure feels alive under load. Too much over-stresses cables and buckles struts. The prestress level is a design variable to optimise, not a value to guess.
• Treating cables as able to push. Cables carry tension only. Any analysis that lets a "cable" go into compression is silently modelling a strut and will predict a stability the real structure does not have.