See discussions, stats, and author profiles for this publication at: [Link]

net/publication/283311205

Designing with Space Syntax A configurative approach to architectural layout,

proposing a computational methodology

Conference Paper · September 2013

DOI: 10.52842/[Link].2013.1.357

CITATIONS READS

24 13,094

3 authors:

Pirouz Nourian Samaneh Rezvani

Genesis Lab Demo Consultants
78 PUBLICATIONS 224 CITATIONS 16 PUBLICATIONS 77 CITATIONS

SEE PROFILE SEE PROFILE

Sevil Sariyildiz
Delft University of Technology
217 PUBLICATIONS 1,224 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Architectural Information Map (ArcIMap) View project

Earthy View project

All content following this page was uploaded by Pirouz Nourian on 29 October 2015.

The user has requested enhancement of the downloaded file.

Designing with Space Syntax

A configurative approach to architectural layout, proposing a computa-

tional methodology

Pirouz Nourian1, Samaneh Rezvani2, Sevil Sariyildiz3

1,3
TU Delft, the Netherlands, 2Politecnico di Milano, Italy
1
[Link] 2[Link]
pub/samane-rezvani/28/383/777, 3[Link]
1
[Link]@[Link], [Link]@[Link], [Link]@[Link]

Abstract. This paper introduces a design methodology and a toolkit developed as a

parametric CAD program for configurative design of architectural plan layouts. Using
this toolkit, designers can start plan layout process with sketching the way functional
spaces need to connect to each other. A tool draws an interactive bubble diagram and a
set of tools reveal feasible geometric interpretations of the proposed bubble diagram in
terms of plan layout graphs. Offering real-time Space Syntax analyses at the same time,
the tools provide feedback on the spatial performance, which is translatable into the likely
social performance of the plan layout patterns.
Keywords. Architectural configuration; graph theory; space syntax; spatial performance;
plan layout.

BACKGROUND
Space Syntax theory (Hillier and Hanson, 1984; Hill- ganization’; he specifically mentions that a social
ier, 2007) has established a methodological body role certain space has within a building is very much
of knowledge on spatial qualities of architecture as dependent on its ‘position’ as to the transition from
distinguished from its over-highlighted formal as- public to private.
pects. From an analytical point of view, Space Syn- On the other hand, from a computational de-
tax theory provides a comprehensive and consistent sign perspective, the issue of ‘plan layout’ has been
framework for understanding spatial arrangements mostly addressed from various optimization points
and their likely human effects, which we can term of views (Lobos and Donath, 2010); most of which
as social performance of buildings. From another deem configuration as an order that can be ‘found’
perspective, in a study of building types as social through thousands of trials and errors in putting
constructs, John Habraken (Habraken, 1988) catego- spaces together in different ways in order to maxi-
rizes three major aspects of building types as social mize certain qualities. This approach to plan layout
constructs: spatial organization, physical structure, is in deep contradiction to viewing architectural
and stylistic systems. He suggests that the one most design as an intellectual activity initiated with ‘pro-
intimately related to our behavior is the ‘spatial or- posing’ configurative ideas. “Architectural and urban

Chapter - Computation and Performance - eCAADe 31 | 1

 design, both in their formal and spatial aspects, are computational design methodology (a structured
seen as fundamentally configurational in that the collection of computational methods) embedded
way the parts are put together to form the whole is in a design toolkit in response. It begins with an
more important than any of the parts taken in iso- abstract configurative arrangement of spatial enti-
lation” (Hillier, 2007, p. 1). “Configuration as the way ties by a human designer; follows with provision of
spaces are related to each other in order to serve a interactive bubble diagram; resumes by systematic
functional purpose is the very nature of architec- exploration of feasible geometric interpretations of
ture” (Hillier, 2007, p. 67); and yet we find very little the configurative inputs as plan layout patterns; and
about the way design can be systematically started ends with dimensional specification of them accord-
through dealing with such a matter. ing to the design brief (this feature is still under de-
What is primarily missing in the literature about velopment).
computational layout is a methodological approach A single connectivity graph, as an abstract en-
rooted in consideration of “how designers think” tity, is interpretable to various geometric configura-
(Lawson, 1980); likewise, a comprehensive consid- tions all of which share the same pattern of intercon-
eration of social implications of configurations is nectivity although they may vary in size and shape
absent. Specifically, in the mentioned optimization from one another. Using this methodology, design-
approaches to plan layout, it is often neglected to ers can sketch how the spaces are to interlink, and
relate to design processes as practiced by design- then they can use the toolkit as follows. A tool reads
ers. Designers do not seek to reach an order through these interlinks and interprets them as a graph that
thoughtlessly trying out random arrangements of captures the important spatial properties of a build-
spaces; on the contrary, they usually start with an ing; another tool finds a planar topological embed-
‘idea’ as to how spaces should be put together to ding of this graph; and a set of tools perform Space
function in a certain desired way. Such configura- Syntax analyses such as depth (visualized in justified
tive ideas convey the understanding of architects as graphs), integration, control, choice, and difference
what is ‘socially’ considered as desirable. factor (Hillier and Hanson, 1984; Hanson, 1998).

SUMMARY DESIGN WORKFLOW

The initial idea behind the proposed design meth- The design process put forward by our toolkit is
odology, supported by our toolkit, was to give life to indeed a reflective cycle of see-move-see or “re-
the bubble diagrams conventionally used for spatial flection-in-action” (Schon, 1987), rather than an
arrangement, to allow for communication of con- automated problem solving procedure. The whole
figurative ideas between designers and computers. design workflow proposed by our methodology is
This idea brought about the following questions: about going from an abstract graph description of
How can a computational system interpret configu- spatial connections to a topological planar embed-
rative ideas, put in the form of a bubble diagram, to ding of that graph, analyzing that graph in real-time
plan layout patterns? Does a certain configurative and finding feasible geometric cell configurations
diagram have only a single corresponding layout that admit the proposed graph of connections.
or more? If there are many, how can we systemati- Our design methodology is an innovative fu-
cally find the fundamentally distinct ones? Moreo- sion of what was proposed by Steadman and March
ver, which design qualities result from the proposed (Steadman, 1983, pp. 69-75; March and Steadman,
connectivity patterns; and how can we study them 1974); a Tutte (1963) convex drawing algorithm; our
methodically? innovative force directed algorithm inspired by that
We have addressed these questions from a of Eades (1984); a set of real-time Space Syntax anal-
graph theoretical point of view and proposed a yses (these tools are the first real-time Space Synatx

2 | eCAADe 31 - Computation and Performance - Chapter

Figure 1
Our toolkit: A) graph forma-
tion tools, B) graph reading,
convex drawing and plan
layout tools, C) force-directed
bubble diagram tools, D) our
Space Syntax toolbox. tools integrated with a parametric architectural de- (Figure 1).
sign workflow); and a set of algorithms for finding Technically, the course of actions suggested by
plan layouts inspired by Steadman and (Roth and our proposed design methodology (Figure 2) is as
Hashimshony, 1988). described below.
We have developed a parametric design plugin
in [Link] that is installed as an add-on for Rhinoc- Step 1: Preparing the input
eros® and Grasshopper© ([1], [2]). Our tool suite is Designer starts with making a number of arbitrary
developed as a plugin that is installed on Grasshop- points as for defining the center of functional spac-
per, and it is undergoing final tests before release es, a corresponding list of (rough or exact) area val-

Figure 2
A flowchart describing our
proposed design methodol-
ogy. We could wrap the whole
set of tools in this way, but we
chose to let the curious user
try different tool configura-
tions.

Chapter - Computation and Performance - eCAADe 31 | 3

 Figure 3
Fom left to right, the sketch-
pad, a sample configuration
drawn by a designer, the
unique untangled planar
drawing of the configuration
including and excluding the
nominal North-East-South-
West sides of the configura-
tion. .

ues, and a list of spatial labels (names) for them. A would need the rooms to be adjacent to one an-
tool assigns rainbow colors to the functional spaces other in order to be accessible immediately. This is
to make them more recognizable. To make it easy to say that a set of connectivity requirements can be
for the designer to link the nodes, a “graph reader” thought of as a subset of an adjacency requirements
tool puts circles of sizes specified by the area values set. However, designers usually do not think of ad-
around all center points. The graph reader tool pro- jacencies in advance. It makes more sense to start
vides a sketchpad with the nominal North-South- with a set of required connections, even though it is
East-West sides for the user to draw the connections more difficult to formalize. In our approach, it is easy
(Figure 3). for a designer to add an adjacency link and even dis-
tinguish between adjacency links and connectivity
Step 2: Producing a Connectivity Graph ones. This has important consequences on the ulti-
According to their configurative idea, designer mate floor plans. This point will be clarified in the ex-
draws a line between every pair of points (circles planation of next steps. Designer adds a few links to
representing functional spaces) that they think relate some of the spaces to the nominal Northern,
should be directly linked. These links eventually Southern, Eastern, or Western frontiers of their plan.

Figure 4
our architectural Space Syntax
analysis tool suite in action.

4 | eCAADe 31 - Computation and Performance - Chapter

Figure 5
A set of justified graphs drawn
automatically (user can
change the point of view) by
our Justified graph tool.

• The graph reader interprets the input links and The first thing we need to know about a con-
points and their “label and area and color” at- figuration is how many topological steps a sin-
tributes as a graph (Figure 2 and Figure 4). gle space is away from another one. A distance
• It provides the user with a verbal interpretation measured between two nodes on a graph is
of links between spaces. called the graph theoretical distance between
• It tells the user whether there can ever be a them. We have developed an automated “Jus-
plan in one floor with such connections (corre- tified Graph” drawing tool that visualizes such
sponding to a planar graph). distances on depth levels. In any configuration,
one can choose a point of view to look at their
Step 3: Space Syntax Analyses proposed configuration literally from different
The theory of Space Syntax was initiated as a theory points of views (Figure 5).
of architecture, seeking to explain the meaning of • Integration (Hillier and Hanson, 1984)
spatial configurations as to their social functions. Integration (1) is a measure of centrality that in-
Although it has been mostly used in urban analysis, dicates how likely it is for a space to be private
it is still an architectural theory, and its basic exam- or communal. The more integrated a space, the
ples are architectural. In simple terms, the theory of shallower it is to all other nodes in a configu-
space syntax is focused on how spatial units relate to ration. Integration is calculated by computing
one another in buildings and built environments. In the total depth of a node when the depths of
this context, the terms syntax and morphology are all other nodes are projected on it. It is formal-
used practically in their linguistic senses. We could ized as in (1) in which k denotes the number
consider meanings for spatial arrangements, analo- of nodes, TD is the total depth as explained
gous to the way we do for verbal statements. While above, and Dk, the so-called diamond value, is
studding syntactic issues we look at how spaces re- obtained from (2). It indicates how an individ-
late to each other as a whole. Whereas, from another ual space is private or communal within a con-
point of view, we could look at the individual spaces figuration.
and focus on their morphological aspects, and their
geometrical state of being. This is to say that the for- (1)
mer concerns the topology and the latter concerns
the geometry of built environments. In our tool
suite, we have implemented a few of space syntax (2)
measures including: • Difference Factor (Hanson, 1998)
• Depth (Automatically Visualized in Justified As a measure of spatial articulation for a whole
Graphs) configuration, the difference factor indicates

Chapter - Computation and Performance - eCAADe 31 | 5

 how differentiated the space are within a con- Step 4: Producing a Unique Convex Embed-
figuration. It is calculated according to (3), (4), ding of the Connectivity Graph
(5) and (6). A very important tool in our tool suite is for un-
tangling a connectivity pattern, which still has an
abstract meaning. This tool produces a unique
(3) topological embedding of that on a plane. It is im-
plementing the Tutte algorithm for convex drawing
(Tutte 1963). The valuable point is that once this
(4) (linear-time) algorithm converges into an embed-
ding (usually in a small fraction of a second) we are
certain that it is unique. Therefore, that means that
no matter how we provide the connectivity input,
(5) we always get the one embedding that corresponds
to that single graph of connectivity. A topological
embedding indicates how the vertices of a graph
(6) are connected to one another on a surface. It is usu-
• Control (Hillier and Hanson, 1984; Hillier et al., ally expressed in terms of ‘face’ descriptions. There
1987) is only one convex embedding of a planar graph,
Control value (7) intuitively indicates how which is revealed by Tutte algorithm. The convex
strongly a vertex in a graph (a space in a con- drawing algorithm reveals the unique planar topol-
figuration) is linked to other points in a supe- ogy of the connectivity graph, given that it is linked
rior manner. It is computed by (7) in which Di in a particular way to the nominal “North, East, West,
is the degree of a ‘neighbor’ node, and n is the and South” (NEWS). A topological description is in
number of all neighbor nodes. between an abstract connectivity description and a
concrete geometry. This is exactly the breakthrough
of our computational methodology that it uses a
(7) Tutte embedding for generating geometric graph
• Choice (Originally introduced as Betweenness drawings and plan layout patterns. This tool also
by Freeman (1977)) performs a planarity test and tells the user if a floor
Choice or Betweenness is a measure of impor- plan is admissible for the set of connectivity require-
tance of a node within a configuration. That ments; provides an ordering for automated justified
literally tells how many times a node happens graph drawing; and distinguishes a sub graph of the
to be in the shortest paths between all other whole connectivity graph (excluding NEWS verti-
nodes. It can also be computed for the links ces). This sub graph, its vertices and its attributes will
connecting the nodes in a similar way. It is be used further on (Figure 3). This tool also gener-
computed by (8) in which where σjk (Pi) is the ates error messages when the connectivity graph is
number of shortest paths between node Pj and not planar. The tutte algorithm, however, could de-
Pk which contain node Pi, and σjk the number of liver result with poor geometric resolution in some
all geodesics between Pj and Pk. cases. To overcome this drawback we introduced
our force-directed drawing tool in addition.

(8) Step 5: Force-Directed Graph Drawing

This tool contains our force-directed graph-drawing
algorithm and makes a “kissing disk” drawing of the

6 | eCAADe 31 - Computation and Performance - Chapter

Figure 6
A few samples of interactive
bubble diagrams (user can
change the areas in real-time)
produced by our force-direct-
ed graph drawing algorithm.
The links are colored accord-
ing to their betweenness
importance values.
bubble diagram. This algorithm works by a set of at- residential plans for instance, the various represen-
tractive and repulsive forces (as in (9)) acting recur- tations and measures of Space Syntax show how do-
sively on graph vertices, seeks a ‘relax’ situation for mestic space manifests life styles, social meanings,
a graph, and reaches to a graph drawing. This tool and identities of different sub-groups within society
is quite intuitive and shows in real-time bubble dia- (Hanson, 1998). Using Space Syntax methodology,
grams neatly according to the specified areas and the system interprets spatial arrangement from the
the connectivity graph (Figure 6). very moment it is drawn as a bubble diagram, and
gives qualitative feedback on the implications of
(9) this diagram to the designer as spatial performance
measures. According to the design context, design-
Step 6: Revealing Dimension-Less Plan ers are free to interpret these spatial performance
Layout Patterns measures into the ‘likely’ social performance of their
A convex drawing found in the fourth stage can be ideas. As a result, performance analysis is automated
‘triangulated’ so as to give rise to dual graphs that by the system; but performance evaluation, i.e. judg-
can represent a cell configuration admitting the ing the relative goodness of design alternatives, due
connectivity graph in itself. While triangulating, we to the intellectual complicacy of the matter and es-
may add links that were not proposed as connectiv- pecially because of its contextual essence, is inten-
ity links, but they simply imply adjacencies that may tionally left for human designers using the system.
arise out of compactness and enclosure geometric
constraints. If we confine the triangulations to a par- DISCUSSION
ticular type of triangulations, then we may get rec- It is left for the designers to decide on how they
tangular dual graphs that can be viewed as dimen- want to alter their ideas during the design process,
sion-less plan-layout patterns (Figure 7 and Figure but the tools always provides them with automatic
8). These “dimension-less dissections can be later di- feedback on the properties of what they design;
mensioned by means of two algorithms introduced while showing them their own ideas, literally. It is
in (Steadman, 1983; Roth and Hashimshony, 1988; important to note that these ideas usually evolve
March and Steadman, 1974). We are still developing during the course of design process, as problem for-
the dimensioning process and so far, the tool goes mulations and solutions evolve together (Dorst and-
until delivering dimensionless patterns. Cross, 2007). Viewing a justified graph, designer can
choose from which space the other spaces are seen,
ANALYSIS AND EVALUATION OF SPATIAL say from different points of views, and analyze it in
PERFORMANCE terms of syntactic measures. This helps designers
Space Syntax measures and their distributions are see if what they have proposed in terms of a bubble
qualitatively interpretable into concepts such as pri- diagram actually matches with their initial ideas on
vacy and community (Hillier, 2007, p. 22). In case of privacy/community, spatial articulation and other

Chapter - Computation and Performance - eCAADe 31 | 7

 Figure 7
The course of computational
procedures for triangulating a
connectivity graph by adding
adjacency links, finding a
dual graph and a rectangular
dimension-less plan layout
pattern.

spatial qualities. We argue that through the design feedback on the spatial qualities of their designs
process put forward by this tool and methodology and indications on likely social performance of their
package, designers have full intellectual control over designs. Our design toolkit allows for interactive
the spatial qualities of their designs; they can ben- diagrammatic design by human designers and sug-
efit from computation in seeing their own ideas gests them multiple possible interpretations of their
from different angles; and they receive objective own configurative ideas.

8 | eCAADe 31 - Computation and Performance - Chapter

Figure 8
The 16 feasible plan-layout
patterns of a sample con-
nectivity graph revealed and
enumerated by our tools
exhaustively.

REFERENCES
Dorst, K. and Cross, N., 2007. ‚Co-evolution of Problem and University Press.
Solution Spaces in Creative Design‘. In: Computational Lawson, B., 1980. How designers think. 4th, 2005 ed. Burl-
Models of Creative Design. Sydney: Key Centre of De- ington: Architectural Press, Elsevier.
sign Computing and Cognition. Lobos, D., Donath, D. , 2010. The problem of space layout in
Eades, P., 1984. A Heuristic for Graph Drawing. Congressus architecture:. arquiteturarevista, 6(2), pp. 136-161.
Numerantium , 42(11), p. 149–160. March, L,Steadman, P, 1974. The Geometry of Environment:
Freeman, L., 1977. A set of measures of centrality based An Introduction to Spatial Organization in Design.
upon betweenness.. Sociometry, Volume 40, p. 35–41. s.l.:M.I.T. Press.
Habraken, N. J., 1988. Type as a Social Agreement. s.l., s.n. Roth, J and Hashimshony, R, 1988. Algorithms in graph the-
Hanson, J., 1998. Decoding Homes and Houses. Cambridge: ory and their use for solving problems in architectural
Cambridge University Press. design. computer-aided design, 20(7), pp. 373-381.
Hillier, B, Shinichi, I, 2007. Network and Psychological Ef- Schon, D., 1987. The Reflective Practitioner. Cambridge:
fects in Urban Movement. Melbourne, Springer. Cambridge University Press.
Hillier, B. Hanson, J., 1984. The Social logica of space. Cam- Steadman, P., 1983. Architectural Morphology: An Introduc-
brige University Press. tion to the Geometry of Building Plans. s.l.:Taylor and
Hillier, B., Burdett, R., Peponis, J., Penn, A., 1987, Creating Francis.
Life: Or, Does Architecture Determine Anything?. Archi- Tutte, W. T., 1963. How to draw a graph. s.l., s.n.
tecture et Comportement/Architecture and Behaviour,
3(3), p. 233 – 250. [1] [Link]
Hillier, B., 2007. Space is the Machine. LONDON: Cambridge [2] [Link]

Chapter - Computation and Performance - eCAADe 31 | 9

10 | eCAADe 31 - Computation and Performance - Chapter

View publication stats