Journal of the British Interplanetary Society, Vol. 38, pp. 305-314~1985.

Huntington Beach, California, USA.

A space station, an investment in permanent space occupancy, is justifiably concerned with evolutionary growth, adaptability and interchangeability. The system proposed here for its construction is based on the premise that whatever configuration is first launched will be found less satisfactory than envisaged in ground-based studies. At this point, rearrangement would be preferable to starting all over again.

To assure construction of any shape (or any size) of assembly, the system is a truss work of equal-length bars, habitable modules interchangeable with struts. There are six standard units and no adapters. The key element is a nodal sphere made from 12 identical sub-assemblies, each corresponding to a face of a bulged rhombic dodecahedron. Each port includes a berthing mechanism that allows lateral engagement.

Habitable modules are integrally stiffened shells with standard patterns of attachment fittings for equipment installation, an extension of the concept that transformed a Saturn S-IVB stage into the Orbital Workshop of Skylab.

External appendages (antennae, solar arrays, etc.) plug into ports in the nodes at the assembly's edges, while hangar spaces are multi-cell volumes inherent in the lattice. Clocking and identification of the port interfaces assure correct assembly.

Not an alternate configuration, the construction system could build any of them.


The Space Station will differ from previous space programmes by being not only re-usable but continuously in use. Since it will be revisited regularly and can therefore be frequently up-dated, it does not require advanced and risky technology at the outset. However, it must inherently possess adaptability for new technology as it develops. It is imperative that this capability be achieved without rebuilding the basic framework. This means that initial priority should be given to mounting provisions, distributive systems, standard interfaces and basic geometry.

This paper offers a system designed to meet these specifications. Not intended as an alternative to any Space Station configuration, it is a means to constructing any of them and, if necessary, converting from one to another. Developed initially as a set of modules for long-duration Shuttle Orbiters, the system was expanded to "evolve" into an embryo Space Station. Attention has therefore been focused on general construction rather than specific forms.

Earth-bound activities such as cutting, drilling and trimming produce chips that are swept up and discarded after falling to the floor. In space, they produce floating clouds of dangerous debris. Similarly, loose bolts, nuts and washers, if lost, threaten the reliable operation of machinery. Clearly, standardised, pre-installed attachment provisions and captive fasteners must be emphasised.

Skylab experience offers precedents for such standardisation. The walls and partitions built into an S-IVB hydrogen tank to make an Orbital Workshop were made of open lattice grid, a regular pattern of nodes 107 mm apart at the intersections of bars forming equilateral triangles. The construction was open to drain hydrogen fuel, since it was originally intended to function as a working stage. The triangular form was necessary for carrying shear loads across an open trusswork. Figure 1 shows a typical installation of equipment and the S-IVB that was modified to house the living and laboratory quarters. It is evident that any installation on this standard "pegboard" could be removed and replaced by any other equipment designed to fit the pattern without the cost and delays from negotiating changes in a structure designed without these accommodations.

Fig. 1. The Skylab Orbital Workshop, showing the grid floor.

The whole habitable frame and associated equipment (about 17.,000 kg) was suspended from the waffle-stiffened S-IVB cylinder. Fortuitously, this structure contained about 2,000 attachment opportunities at the intersections of stiffening ribs. These were exploited by installing a stud in a tapped hole at these locations (Fig. 2).

Fig. 2 Standard stud in S-IVB tank waffle node

In this case, the tapping operation was a pre-launch rework; for a space station, such provisions must be designed in at the start. The discovery of inherent accommodations accidentally offered by the structural arrangement was, in Skylab, followed by intentional inclusion of similar capability in the added structure. If Skylab had not fallen out of orbit before it could be re-boosted, it would even now be able to accept new equipment in exchange for the old. For the Space Station, better initial planning should provide even more versatility.

It is obvious that the growth and adaptability so desired in the installed subsystems must also be a prominent and planned feature of the subsystem that supports all of them: the structure. In fact, it should start there. Structure, in this sense, connotes "architecture" and all that implies: space allocation, protection, accessibility and maintainability, along with its more obvious role as a sustainer of loads.

In response to this compelling logic, the system here proposed for space stations:

Is not an alternate configuration: it implements any of them

Can fill all of space - or any part of It

Has elements sized by the Shuttle payload bay

Is structurally efficient, its growth and controllability not being limited by low natural frequency

Uses only six base construction units with standard interfaces and component mounting provisions.
Considers pressurised modules and construction struts interchangeable parts of a common lattice

Permits evolutionary growth, shrinkage and reconstruction in another form after initial operation

Requires little new technology at the start but can accommodate any that develops later

To summarise, this system is not only "technology transparent,' as requested, it can also be called "configuration transparent."


If a large space lattice is to be truly versatile it must be able to assume any shape. To do this, it must consist of identical cells or cell groups endlessly repeatable. If the edges of such cells are equal-length elements (a desirable condition of standardisation), two basic systems are derived:

Cubic: square faces and corners

Tetrahedral: equilateral triangles

Both forms can be modeled three-dimensionally, as shown in Fig. 3.

Fig. 3 Models of cubic and tetrahedral space trusses

To explain space networks in a two-dimensional manner, square and triangular area arrays are depicted in Fig. 4.
Fig. 4 . Approximations of an arbitrary shape with square and triangular networks

The small patches shown can be expanded infinitely, if necessary. Their three-dimensional equivalents are cubic and tetrahedral lattices. The latter is somewhat misnamed because space cannot be filled with tetrahedrons. Bars running in the same six directions as the edges of a tetrahedron divide space into tetrahedrons and octahedrons in the ratio of two to one. However, the nodes are all identical; 12 bars converge at each intersection.

As indicated, an arbitrary shape can be approximated with either array, though the triangular one, running in three directions instead of two, can more closely approximate the slope of the figure's outline.

For any large space structure, the Space Station included, the tetrahedral truss is structurally superior to the cubic form because triangles are inherently rigid and stable. A cubic lattice is made stable by adding diagonal struts or cables, thereby creating triangles; but in this case the lengths are different. So, for a cellular system or a crystalline lattice, the rational choice is tetrahedral, both for large frames and for clusters of habitable modules. Inconsistency of geometric arrangement implies sets of cost-incurring adapters.

The apparent popularity of cubic cells may stem from a universal understanding of the geometry at an intersection. It is also obvious that the polyhedron that presents a face normal to each of the lattice bars is a cube. It has not often been recognised that a pressurised node with no preferential directions of strength (as in an infinite lattice) needs to be a sphere divided into six identical bulged patches, the projections of a cube face on a surrounding sphere. Instead, most airlocks have been cylinders, or intersections of them.

What has rarely been recognised is that for a tetrahedral lattice there is also a polyhedron with a face normal to each incoming bar on which an interface can be centred. Figure 5 shows some polyhedral subdivision of spheres.

Fig. 5. Polyhedral subdivisions of spheres and a spherical node for a tetrahedral lattice.

The cubic version has already been mentioned. The form of interest for the tetrahedral lattice is the rhombic dodecahedron. Not generally known to designers of large space structures, it is well enough known to students of polyhedrons to have been given the name it bears. In a close hexagonal packing of spheres in space, where 12 spheres are in contact with a central one while touching each other, the planes of tangency at the points of contact around the central sphere are the faces of a rhombic dodecahedron. It, like the cube, when packed against others of its kind, completely fills space without voids. This seems to be a necessary characteristic for the node of a universal space-filling lattice. The literature indicates that this polyhedron and the cube are the only two with identical faces that possess this property.

A model of this shape at a tetrahedral nodal intersection is shown in Fig. 6.

Fig. 6. Model of a rhombic dodecahedron at a tetrahedral lattice node.

It is the key element in an endlessly repeating lattice system which, by its ability to fill all of space, can also fill any part of it. Thereby, it permits the construction of any space station configuration or, for that matter, an assembly to carry explorers to the vicinity of Mars.

The foregoing discussion treated nodes in the lattice as a given condition, although the existence of such units has been the subject of debate concerning the question of affordability. Without any formal economic analysis, this uncertainty should be settled by rational discussion. Firstly, a node should be expected to function as an airlock and a space station without airlocks is unthinkable. In addition, even the simplest two-dimensional array (the square one shown in Fig. 4) has intersections where four bars meet. Each bar has two end connections. Therefore, to come out even, the infinite array must have twice as many bars as junctions. The triangular array has three times as many bars as intersections, and a three-dimensional tetrahedral lattice requires six times as many bars (or modules) as nodes. Therefore, for universal adaptibility, nodes cannot be integral with bars because the required numbers of each are different. Growth would be restricted and more nodes (or the equivalent) would be procured than are required.


The Space Station depicted in Fig. 7 is the initial all-up configuration of the so-called "power tower."

Fig. 7. Space Station study reference configuration, the "power tower."

This design is the reference for the definition phase Space Station studies. It features a deployable structural main frame that links appendages such as power-generating arrays, experiments, hangars, antennae and habitable modules. The external size diflferences between initial and growth versions is small. Chiefly, the power-generating systems switch from solar cells to heat-cycle turbines, the number of pressurised modules increases and the main frame is reinforced with parallel beams. Increases in mass and technology state are not matched by size or shape changes in basic architecture. Deployable mainframes do not permit it.

The design's counterpart in the construction scheme proposed here is shown in Fig. 8.

Fig. 8. Representation of a tetrahedral "power tower"

This is a model of the main frame; appendages are not included. While the overall arrangement is essentially the same, there are basic detailed differences:

The core structure is erectable, not deployable, each bar representing a span of 17.76 m.

Some of the bars - any appropriate clustered group - are habitable modules with attendant nodal airlocks.

Appropriately scaled up, the assembly as shown is 134 m long with a span across the stub arms of 50.3 m. The deployable power generating equipment extends beyond these arms, supported on rotary bearings.

Hangars are screen-enclosed octahedral cells of the core, one such opportunity being shown on the model. Each hangar cell has a volume of over 1700 m3, accessible through a triangular door (one removed side) 15 m wide by 13 m high. Probably four or five such cells are usable as hangars in the size of assembly shown, though the core itself can be expanded indefinitely.

An assembly of this size and form is made from 120 bars and 40 nodal joints, both of variable composition. In all, 240 connections must be made after the bars are made up and transported to the assembly point. The connections are built into the standard units with no loose parts and no adapter brackets.

A now-discarded configuration that received considerable early attention is the "delta" space station (Fig. 9).

Fig. 9. "Delta" configuration studied by NASA.

As designed, its three deployable structural panels each contained about 1600 members, almost 5000 altogether including splices. It was designed this way "to minimise EVA" as stated in the final report. While this statement may be true for an identical design, one intended to be erectable would be expected to entail much fewer and larger elements.

Figure 10 shows another model representing the "delta" design as built with the tetrahedral truss system. Fig 10. Tetrahedral version of the "Delta".

As expected, while it is about 47 m high, it uses only 147 bars and 51 nodes. As many as 10 of the bars are standard pressurised modules, interconnected through six nodal balls; the rest consists of standard struts and corresponding nodal fittings. Unoccupied nodal ports are available for EVA, docking and accessory modules. As the manned portion of the station increases in size, bars and nodal connectors can be replaced by habitable modules and nodal airlock balls.

Solar panels on top of the configuration are assumed to be NASA-developed fold-up panels contained in split boxes extended by an "astromast" or the equivalent, stretching the solar blanket between box-halves. In this case, the panels are double blankets 7.9 by 33.5 m in size; there are 16 altogether, providing about 4250 m2 of collection area for 400 kW or more of power. The arrays do not all have to be there at the start, nor does all the structure. Between the solar panels and the occupied modules at the apex of the delta is a hangar space made by removing structurally unnecessary bars from the centre; it is a large octahedron with 33.5 m edges. Its volume is about 10,500 m3. Note that slices at appropriate angles through the lattice uncover square patterns where they may be needed.


The "meccano set" for a universal tetrahedral construction system consists of seven basic units (six, if a habitable module is always made in a single size). At least three of these elements make up each of the assemblies in the configuration models. Figure 11, showing how the standard 16.76 m spacing is filled by standard units, identifies the first three elements, as follows:

1. Nodal Airlock. This is the spherical entity subdivided into 12 identical rhombic subassemblies that is the key to the lattice geometry and provides ports for EVA, docking or accessories.

2. Habitable Module Cylindrical Barrel. This is the standard shell segment for all habitable modules: laboratories, command centres, sleeping quarters or workshops. Each segment's length is the same as the distance across opposing flats of the nodal ball to allow the combinations shown. The assumed diameter is 4.32 m,

3. Habitable Module Cone End. This unit closes out habitable spaces, incorporating at its narrow end the same attachment and functional interface as must be built into each of the 12 faces on the nodal ball. Its 60 conical shape is determined by the equilateral triangular relationship with adjacent units.

Where the structural frame extends beyond the inhabited volume of a space station, the node-to-node spacing is maintained by the remaining three standard elements:

4. Half-Strut. This unit is a tapered structural column with androgynous connectors on both ends. Its length is half that required to reach between structural cluster fittings at the nodes ("hedgehogs").

Fig. 11. The seven construction elements in a hypothetical 2-dimcnsional assembly.

5. Folding Tripod. All ties between pressurised units and the open structural frame are made through this unit. Its length is half the distance between the opposing faces of two nodal balls. It combines with a half-strut when the elements at the nodes are dissimilar.

6. "Hedgehog" Cluster Fitting. This is an assembly with 12 stubs identical to the androgynous ends on assembled strut pairs. The stubs are geometrically related in the manner of faces on airlock nodes.

7. Pressurized Connecting Tunnel.

The system is sized by the assumption that one Shuttle load consists of one airlock ball and one full-sized 13.4 m habitable module as shown in Fig. 12. For the start of a space station, the first launch should probably orbit a power module complete with attitude controls and a small propulsion system. When the following three launches bring up three bay-filling payloads, as shown, a minimum system can become operational. This small system could barely sustain a 2 or 3 man crew between resupply visits but it would be a start.

The 12 port node is the determinant of the construction when its 12 identical sub-units are properly identified. The orientation and labelling of the faces for this purpose are shown in Fig. 13. It starts with a 3-face patch forming the spherical projection of a face of a tetrahedron. There are four such groups to a ball. As illustrated, the groups are 12-3, 5-6-11, 4-7-12 and 8-9-10.

Fig. 13. Nodal port "clocking" and identification.

Note that the "12-o'clock" angular position of each port is pointed toward the centre of the group, most clearly shown in patch 1-2-3. When this procedure is followed for all four patches, the opposing ports on the ball are always oriented in the same direction. Therefore, nodal balls spatially oriented in the same manner always correctly align with cylindrical modules whose ends are also "clocked" alike.
Assembly aligment can be automatically assured when the polybedral faces are numbered like cubical dice (whose opposite face numbers add up to 7), for a 12 faced entity like this, the sum of opposite face numbers is 13. Thus, when a habitable module end is plugged into port 5 of a ball, its opposite end fits port 8 of the next node. The proper assembly of any space station arrangement can be defined unambiguously in this manner.


In a lattice assembly of habitable modules interconnected by spherical nodes, any nodal unit is an airlock. As such, it must be sealable to prevent loss of the breathable atmosphere. This, in turn, leads to a requirement for inward opening doors, one for each penetration, or 12 in the element appropriate for tetrahedral truss geometry. In the airlock photograph shown in Fig. 14, one of the 12 identical subunits is shown removed (at lower right). Accessories shown plugged in include a 4-thruster propulsion unit and a suggested battery pack.

Fig. 14. Nodal airlock ball with standard sub-assembly (lower right) and accessories - thruster group and battery pack.

A spherical shape is indicated for good reasons, not the least of which is that a round door covering a circular opening fits well against the spherical shape near the opening. Internal space is less likely to be cluttered with unstowable hatches. Spheres are, of course, the most efficient form of pressure vessels, while their compound curvature also makes them resistant to compressive forces. Circular holes in spheres also remain round and planar under pressure, simplifying the problem of maintaining an effective seal.

Figure 15 shows in some detail how a rhombic subassembly can be constructed. It is double-walled for a rigid docking interface. In addition, the curved and planar surfaces enclose a space through which system runs (wiring, fluid lines, etc.) can be routed and interconnected.

Fig. 15. Rhombic sub-module assembly.

Since the inner polygon is a machined open grid lattice, it not only protects and supports these system runs but allows access to them. Integral to these panels is a door jamb and sealing surface. Twelve identical units like this form a 3.56 m diameter sealed ball. The diameter selected is the minimum required to stow hatches between 1.02 m openings without encroaching on them. For larger openings, proportionately larger nodal balls are required, this enlargement being limited by the clearance in the Shuttle payload bay.

Although it has not been indicated, a further intention for adaptibility, is some form of socket, stud or similar attachment opportunity at each of the 14 "corners" of the structural ball. At these points, inner and outer structures converge, offering locally high strength. While there is no specific requirement for such provisions, the opportunity, if offered, will almost certainly be used. If passed up, it will be difficult to add later.

Since the lattice intersection must accommodate as many as 12 attached modules, nodes at the edges of any configuration offer many unoccupied ports, usable for docking, EVA

or standard plug-in accessories like the previously-shown propulsion and storage units, or tanks for consumables or antennae. Any accessories, once developed, are directly transferable to another port, another module or a different configuration.

It is obvious that an adaptable nodal shell can be used in many ways. It could be a manned module for a recovery repair and service vehicle. With nearly 11.5 m3 of volume inside the inner polyherdon, it could be fitted for life support of a four man crew for ten days to two weeks. Essentially the same vehicle can serve at a space base as a rescue "lifeboat." Among the accessories that could be developed for this role is a remote manipulator turret.

Figure 16 illustrates a proposed docking/berthing interface for this system. The type developed for Apollo is unsuitable because, projecting beyond the interface plane, it prevents lateral approach; it fills the tunnel, requiring disassembly before use; and the mating faces are different. The single-use Apollo/Soyuz device, although androgynous (both sides identical), also projects awkwardly across the interface plane.

Fig. 16. Docking/berthing interface for lateral assembly.

These precedents and the unique problems of this closed system combine to define the docking requirements:

Androgynous (one system to develop)
Flush faces for lateral engagement
Positive guidance for head-on encounter
Light connection before pull-up
Minimum size and weight (many identical ports)

An added feature proposed for this mechanism is retractability to move out of the way when a port is occupied by an accessory module.

As a consequence the suggested method features inward sloping triangular paddles, motor-driven through a range of positions appropriate to each operational mode. With hooks at their ends, they also perform the latching function by reaching through the opening in the opposing interface and engaging sockets in the tunnel wall. When fully engaged and locked, the paddles from both mating units lie inside the opposing tunnels, leaving a clear passage available for immediate use. ,

For a typical head-on docking the paddles are extended at an angle such as 45. As the interface planes near each other (approximaely 25 mm short of complete closure), a spring-loaded detent hook (not shown) prevents recoil while the paddles further expand to engage. After the hooks enter sockets in the opposing tunnel wall, a pull-up actuator in each paddle rigidises the connection.

As the figure shows, the paddles can be positioned anywhere in a range exceeding 180. For lateral insertion into the structural stack, the paddles are flush with the interface plane or below it. As soon as the mating units are approximately aligned, the paddles extend to centre the connection and continue to the engaged position. When withdrawn completely into their own tunnel, the paddles and latches are out of the way of an installed submodule whose attachment points can be the six spots shown at the hinge ends.

The paddles are driven to any commanded position by worm drive actuators located at the end of the hinges and bussed together by a chain, perforated tape or similar drive girdling the tunnel. Each actuator is indirectly connected to the paddle it drives by a spring and damping system built into the base hinge.

All of the mechanisms described and the dimensions assumed are tentative and subject to revision as a result of further study or such approvals as will be required for an international standard. Whatever develops will become an international standard, one that is long overdue.


The habitable module required for most Space Station functions seems to be close to the same length and diameter as the now-operational Spacelab, but while Spacelab has been funded and developed, its design has been concerned with a single unit (of variable length) compatible with the Shuttle Orbiter payload bay. Little thought has been given to clustering multiple units into space station assemblies; particularly, the end geometry does not lend itself to assembly in any way but tandem. To a considerable extent, the thinking applied to all previous manned vehicles has concentrated on making one unit work. Much of this thinking can, it is hoped, be applied to the module introduced in Fig. 17

Fig. 17. Universal habitable module.

Shaped for the tetrahedral truss assembly method, this module has relatively long conical ends with converging space which, if intelligently used for service functions, should be no disadvantage. The module is envisaged as a stiffened shell with entry only at the ends; any exposure or exi~t to the vacuum of space or to other modules is through airlock nodes at the intersections. Except for circumferential bands with thermal control shutters, it is completely covered with multilayer insulation of any appropriate variety.

Since a leak-tight presurised canister with this niuch volume, once built, should find many useful applications, the structural shell has been designed to maximise adaptability and usable volume. Typical shell construction for both cylindrical and conical sections is shown in Fig. 18.

Fig. 18. Integrally stiffened pressure shell with standard nodal attachment provisions.

The integral external stiffening ribs form an isogrid array of equilateral triangles with intersection nodes spaced about 180mm apart. The panels are machined from large aluminium alloy plates about 20 mm thick, leaving a skin of about 1.5 mm between ribs. Each nodal intersection incorporates an outer socket and inner button for all thc attachments to the shell. A suggested detachable captive collet fastener for all interior attachments is shown in Fig. 19.

Fig. 19. Standard captive collet fastener.

There are no frames in this construction, not even where transportation trunnion fittings are attached. Frames are avoided because these fittings introduce their loads into the shell tangentially at patches with enough fasteners to match the load-carrying capacity of the nodes. No additional shell reinforcement is expected, though it can be provided if necessary by leaving enough extra material in the machined plate. The trunnion fittings, needed only for transport, are removable, being attached by the releasable fastener shown in Fig. 18.

The versatility of this construction is further demonstrated by a standard window construction, shown in Fig. 20. Instead of removing the whole shell area, only the skin between stiffeners is cut out - in as many pockets as viewing requirements dictate (16 in the example shown).

Fig. 20. Standard window, exploiting attachment features.

The outer edge is sealed by an elastomeric molding of appropriate cross section for window retention. Its underside includes sockets that are retained by the ubiquitous internal buttons around thc edge. Each of the buttons at nodes in the middle of the window also retains an elastomeric supporting collar. This mid-panel support allows the clear pane to be thinner.

Reinforcement to replace the skin removed for vision can be an internal bonded doubler or enlargement of the local ribs if the window location is known before the panel is machined. With the generous structural margins expected for the pressure loads, no reinforcement may be necessary.

This module is larger than those shown in most space station design descriptions. However, having been sized by STS payload bay dimensions and compatibility with the nodal ball, it should probably stay as it is. Any extra volume is sure to be found useful after the start of service.


As has been indicated in Fig. 11, there are three combinations for struts bridging the node-to-node spacing. This is caused by the different size of the two elements serving as nodes; pressurised balls are significantly larger than strutconnecting "hedgehog" fittings. Where there are no habitable modules, the nodal span is covered by two identical halfstruts. Between nodal balls, if such a situation is realistic, two tripods are appropriate.

When the gap between a "hedgehog" and a nodal ball is filled, the two elements are combined. This means that the middle joint on a tripod must be the same as that on a half-strut. In the nested strut system proposed every two-element column can be made from elements of the kind developed by the NASA Langley Research Center.

For the centre joint at the large end of the strut, and in the corresponding position on a tripod, the androgynous multi-fingered joint already developed should
be adequate, although it may need an additional locking ring on each member to trap the fingers so they cannot spread and release.

The small end androgynous joint proposed here is again similar to the Langley design but secured by external sleeves instead of an internal spring latch. It is shown in Fig. 21. This is a well-known mechanism, generally secuted by a single outer sleeve. However, since in an androgynous connection both sides must be identical, two threaded sleeves are used.

Fig. 21. Androgynous strut end connector.

This choice offers a bonus: the two can be tightened against each other to minimise play in the joint. Once the ends are brought together by hand, advancement of the first collar holds the assembly firmly enough to simplify operation of the second collar and subsequent tightening.

The foldable tripod can be attached at three of the six available spots on the docking interface. Since the space between two legs is as wide as the access opening diameter, the port remains usable for EVA and the legs themselves should be useful handholds.

Where construction struts meet at a node, there is a "hedgehog" cluster fitting corresponding to an airlock nodal ball. As illustrated in Fig. 22, the stubs project in the same directions as elements emanating from ball nodes. The "12 o'clock" position at the interface converges on the centre of three-stub groups exactly in the same manner as previously described for docking ports. The same numbering strategy can also be employed here.

As with airlocks, a "hedgehog" is made from identical stubs with rhombic bases. These bases, welded along their edges form the same polyhedron. '

As the figure demonstrates, "hedgehog" fittings attached to each other can, by themselves, make a double-faced truss sandwich panel. With appropriate trunnion adapters, the extended assembly should serve well as a cradle for its own transportation to orbit; a pair of these cradles could, between them, support their struts for the ride.


The Space Station is a new and different type of programme; it will be a continuing process and not a one-off event. As such, it will demand the characteristics offered by this construction system:

An efficient structural system which, because it uses the only stable polygon available, the triangle, is inherently rigid; growth is not limited by low natural frequency.

The adaptability that makes it rebuildable in another form.

Adaptability to accept a variety of plug in modules.

The adaptability offered by structure patterns that physically integrate the subsystems without rework and without time-consuming argument and negotiation.

The adaptability inherent in structural frame consistency, the construction struts and pressurised units conforming to the same lattice geometry - no extra adapters.

The economy offered by developing only six interchangeable core construction units.

The economy of identical interfaces and identical
sub-units which, at an early stage, offer the cost benefits of quantity.

While it is desirable to study and understand what a space station must do, and, in the course of such study, to define what the starting configuration should be, this system permits major changes of mind after first service and for an indefinite period thereafter. In short, the development of the elements can proceed without a definite configuration decision.