The Computation of Design Vocabulary:

Prototyping, Variation, and Composition

by Lee, Hee Won

Master of Architecture in Architectural Design,

University of California at Los Angeles
June, 1992

Submitted to the Department of Architecture

in partial fulfillment of the requirements for the degree of
Master of Science in Architecture Studies
at the Massachusetts Institute of Technology
June, 1997

0 Lee, Hee Won 1997. All rights reserved.

The author hereby grants to M.I.T. permission to reproduce and to distribute publicly paper and
electronic copies of this thesis document in whole or in part.

Signature of the Author

LeeTHee Won
Department of Architecture, May 22, 1997

Certified by
Certified bv Wltiam Mitchell
Thesis Advisor, Professor of Architecture

Certified by
Takehiko Nagakura
Thesis Advisor, Assistant Professor of Architecture

Accepted by
Roy Strickland
Chairman, Departmental Committee on Graduate Students

JUN 2 01997
uB A:ES
Thesis Reader: George Stiny
Title: Professor of Architecture

Thesis Reader: Andrew Scott

Title: Associate Professor of Architecture
 1. ~ * ~

2. V~'
it74 ++

I'

17.

o5ic-
PF
;Cl-11M
I LMJ
O.M. Ungers' sketch for defining 'house types' from Architecture as Theme
Thesis Flowchart

Conclusion:
Design Vocabulary

Bottom-up Method
The Computation of Design Vocabulary:
Prototyping, Variation , and Composition

by Lee, Hee Won

Submitted to the Department of Architecture on May 9, 1997 in partial fulfillment of

the requirements for the degree of Master of Science in Architecture Studies

Abstract

The range of one's vocabulary affects the solution of architectural problems.

The more extensive vocabulary one has, the wider array of solutions one can
provide for the needs of man. Thus when we design a building, we must
know how vocabulary elements are manipulated and composed for the
making of space, structure, and enclosure in response to program and context.

My research interest is in the computation of design vocabulary for

transformations and combinations in the design process as well as the shape
rules involved in the composition of vocabulary elements into an architectural
entity. This thesis investigates how an architectural type can be defined
parametrically to fit into an architectural context, how vocabulary elements
can be assembled and transformed by top-down and bottom-up methods, and
how those forms can be composed by a set of rules. Design experiments
follow the theoretical investigation and demonstrate the proposed theory.

The thesis begins with the study of 'typology' in architectural history which
establishes a theoretical framework for understanding the historical
importance of 'type'. Next follows the analysis of Aldo Rossi's design
methodology to understand how elemental forms are manipulated and
composed by a set of rules, and how the design method can be encoded into
a computer program by shape grammar.

After the historical and theoretical study of type, the computational

background of the thesis is examined for possible application of computers
in the design process. In Computer-Aided-Design, form-assemblage is the
crucial idea for communication of design concepts as well as spatial
exploration. The ability to assemble a form from individual graphic elements
is the most useful feature in CAD. Shape grammar can take advantage of this
CAD capability in classifying basic forms and laying out their
transformations for spatial exploration. To illustrate this idea, a knowledge-
based system is described with graphic examples.

Subsequently, two composition techniques: bottom-up and top-down design

methods are studied through computational design experiments which
implements the idea of typing vocabulary. Furthermore, two sets of building
components are encoded through computer programs to illustrate how
vocabulary elements are manipulated in top-down and bottom-up fashions.

In the end, this thesis concludes that the bottom-up design method is a more
efficient method of exploring form and space than the top-down design
method. It concludes further that the concept of typing vocabulary can play
an important role for using a computer efficiently in the design process with
the bottom-up method.

Thesis Co-Advisor: William Mitchell

Title: Professor of Architecture

Thesis Co-Advisor: Takehiko Nagakura

Title: Assistant Professor of Architecture
Contents

Ideology and Methodology 9

1. Type

1.1 History and Theory 12

1.2 Aldo Rosi: Shape and Rule 16

2. Computation of Design Knowledge

2.1 Shape Grammar 21

2.2 Parametric Design 24
2.3 Knowledge-Based System 26

3. Investigation and Exploration

3.1 Experimental Design

3.1.1 The Summer Academy 32
3.1.2 The Visitor Pavilion 38
3.2 Knowledge Application
3.2.1 Computation of Design Vocabulary 52
3.2.2 Top-down and Bottom-up Methods 56

Reflection 63

Illustration 65

Bibliography 66
Rossi Vocabulary
Ideology and Methodology

My fascination with Aldo Rossi's architecture comes from my personal

obsession with simplicity in life. Architecture is like life; neither is mastered
by man. For this reason, I have consciously and unconsciously fallen in love
with everything simple in all aspects of life - hobby, fashion, car, and people,
to name a few. Architecture is not an exception; I cannot resist
preoccupation with simple architectural forms. This preoccupation has led
me to search for 'simplicity' in the principle of architecture, and finally Aldo
Rossi's architecture provided me with a religious foundation for the
generation of idea and method in design throughout my architectural
education.

Aldo Rossi's design methodology inspired my combined interest in design

and computation. His buildings are created by repeated use of his own
architectural prototypes which he transforms into new design vocabularies in
response to site and program. Rossi's prototypal vocabulary is not transient
with a stylish trend, but they are meant to be reused when he creates new
ideas and new forms. Whenever he designs a building, he uses his existing
vocabulary as a base for form-generation; he transforms them into new
vocabulary in response to new conditions; and he assembles them into an
architectural identity with small shifts. This idea of 'repeatability of
vocabulary' provides great potential for use of computers in the design
process. Once a designer establishes his or her own vocabulary, he or she can
store and retrieve its elements whenever necessary, and manipulate them to
fit into space with the capability of parametric variation. Shape grammar
plays an important role in manipulating these elemental shapes for spatial
composition. It helps to compose basic forms for organizations of space.

There have been numerous studies on the use of computers in the design
process, but few explorations on actual design application. The most of the
research is limited to an analysis or theory of developing software tools
which generally does not demonstrate how the tools are actually going to be
used in design. This lack of demonstration inspired me to experiment the
use of computers in actual design projects. By actual design explorations, I
can learn the advantages and disadvantages of computer use, and thus provide
new and better suggestions for educating designers about use of computers
both in schools and in professional world. Furthermore I hope that the results
of these design explorations will lead to the development of a better software
tool.

There is a well-known method in computer use called 'Parametric Design'

with two techniques: top-down and bottom-up methods. A knowledge-based
system for the top-down design was developed and extensively analyzed for
use in design, but the bottom-up design has not been researched or possibly
avoided for some reason. Two design projects will be explored with both
design methods, particularly with focus on the bottom-design method, in
order to find out their effectiveness in the process of making a building. The
intention of this investigation on top-down and bottom-up methods is to
bring me to a level of confidence where I could lay a theoretical foundation
for a continuous search for an architectural principle. And I hope this thesis
will also show that it is possible to use computers from " scratch to finish".
Design is the computation of shape information that is needed to guide fabrication or
construction of an artifact.. .The process of design takes different forms in different
contexts, but the most usual computational operations are transformations and
combinations of shapes in a two-dimensional drawing or a three-dimensional
geometric model. An initial vocabulary of shapes, together with a repertoire of shape
transformation and combination operators, establishes the shape algebra within which
the computation takes place. The computation terminates successfully when it can be
shown that certain predicates are satisfied by a shape produced by recursively
applying the transformation and combination operators to the initial vocabulary.

William J. Mitchell, The Electronic Design Studio, 1989

 1. Type

By definition, 'type' means the general form, structure, plan, and style
characterizing the members of a class or group ( Webster's New World
Dictionary). But there have been various interpretations about 'type' among
designers and theorists. According to Rafael Moneo in On Typology (1978),
'type' in architecture can be defined as "a concept which describes a group of
objects characterized by the same formal structure." Moneo thinks that
'type' is fundamentally based on the idea of grouping objects by the
similarities of structures. In his article, On the Typology of Architecture
(1963), Giulio Carlo Argan defines 'type' as a common, original form
abstracted from numerous variants. He writes that "in the process of
comparing and superimposingindividualforms so as to determine the 'type',
particularcharacteristicsof each individualbuilding are eliminatedand only
those remain which are common to every unit of the series." Argan believes
that 'type' must have a capability of infinite variations and modification of
'type' itself. U.M. Ungers also talks about a 'type' in another term: a
'model'. He states, in his book City Metaphors (1982), that "a model is
commonly understoodas something that poses as a prototype representingan
idealform." In other words, a model is a basic, ideal structure from which
forms of variations are generated. Ungers implemented this concept in his
various projects. Figure 1.1 shows the thematic models of plan which serve as
a base for all his design. This concept of 'type' has been used and discussed
for a long time in the design and theory of architecture. Designers have
developed ways of 'typing' to produce and reproduce our built world. They
have constructed human places by using certain types of shapes and spaces,
reusing them in a certain context when required.

1.1 History and Theory

Quatremere de Quincy, a theoretician of neo-Classicism, proposed the first

comprehensive, architectural theory of type in his book Encyclopedie
Methodique by the early nineteenth century. He believes that everything must
be generated from an antecedent since nothing comes from nothing. Thus in
his opinion, this concept must apply to all human inventions as well as
architecture. He thinks all buildings must be born of something ideal from the
Figure 1.1 Thematic models of plan past, for example, the Parthenon. The most well-known theory of typology
 -~ AL
I- JQft- 1) -
M-11MI
M
La4k i
. Q 9 0

)
Figure 1.1.1 Classification of round temples from Durand's Recuei, 1801

Figure 1.1.2 Building elements from Durand's PartieGraphique, 1821

was put forth by J.N.L. Durand shortly after Quatremere de Quincy's theory.
Durand developed a highly systematic theory of typology, particularly about
a vocabulary of architectural types derived from existing classical
forms.[Figure 1.1.1] For Durand, to establish the basic types of building
components is the point of departure for architectural design since these
types can be found in any building, regardless of style, i.e., walls, windows,
columns, and roofs. [Figure 1.1.2] Once the basic elements of architecture are
defined, parties of buildings i.e., lobbies, lounges, and courts can be
composed by the combination of the basic elements. Finally the integration of
the parties results in producing a building.

According to Thomas Thiis-Evensen, types are comprised of 3 main parts:

the floor [Figure 1.1.3], the walls [Figure 1.1.4], and the roof [Figure 1.1.5].
He writes that these can be divided into four categories: the elements' major
forms, the construction system (massive or skeletal), the surface treatment of
the major forms, and the openings in the major forms. Thiis-Evensen believes
that on each of these four levels, fundamental formal types must exist to
represent universal solution to problems of various forms that remain the
same anywhere throughout time. To him, creativity depends on how these
basic forms are combined and transformed.

William Mitchell , in The Logic of Architecture (1990), emphasizes the

'type' concept for the use of architectural vocabularies in the description of
buildings. He says that " formation of a type is an act of abstraction; we
identify what is the same about all member of some class of objects and
disregardwhat varies. " What he means here is that 'type' is an object of
class abstracted by the elimination of idiosyncratic variations while
maintaining common characteristics. Mitchell believes that defining 'type'
is necessay for developing an architectural language because it provides "a
strategyfor segmenting architecturalcompositions into parts and a way to
refer to parts by name in order to develop a discourse about some body of
architectural work." Further, he believes that by classifying vocabularies
into different types, a designer can establish the meanings of nouns, such as
column, beam, wall, door, bedroom, and house so that he or she can organize
and incorporate them into design knowledge.

When we develop design vocabularies, it is important to define 'type'

vocabularies which serve as a basis for all later variations and combinations.
These 'type' vocabularies are essential to the composition of an architectural
 b
1 4 1,
C

Figure 1.1.3 Types of floor: Figure 1.1.4 Types of wall: a) horizontal, b) vertical, c) flat, d) convex, Figure 1.1. 5 Types of roof
a) continuous, b) sinking, c) e) concave, f) straight, g) leaning toward, h) leaning away a) done, b) barrel vault, c)
raising, d) framed, e) cen- gable, d) shed, e) flat
tralized, f)equal direction-
nal, g,h) undulating
 order and adaptable to the plurality of the many conditions regardless of time
and place. However, the use of type should not limit a designer's creativity.
'Type' is intended to establish a framework for a designer's process of
thinking within which he or she can do change operations. Here 'change'
does not mean to be arbitrary or transient, but to be contextual and
programmatic. In the transformation process, the designer can create the
subtypes from the type, and overlap different types to produce new ones. He
or she can use a known type or create new types for new inventions
depending on context and program.

1.2 Aldo Rossi: Shape and Rule

The Italian rational architect Aldo Rossi is known for his use of 'typing'
concept in architectural design. His architecture is composed of types of
objects - towers, steps, and corridors - that are used again and again in
different contexts. Rossi always uses these same objects of design
vocabularies in order to find something new for the composition of an
architectural space. He generates new forms by reproducing what is already
there with small modifications. Rossi himself says :

[11 Peter Amell and TeBickford, eds. ALDO Looking at these projects all together, it seems to me that today I could still
ROSSI: Buildings and Projects (New York, 1984), redesign them one by one: perhaps the final result would be different, if only
p. 10. because of small changes and shifts in proportion.[1]

The design vocabularies of Aldo Rossi's architecture are derived from pure,
primitive forms, such as cube, cone, cylinder, and pyramid which are primary
factors for structural and functional elements of his architecture [Figure 1.2.11.
Origins of Prototypes The employment of such simple, stereometric forms resulting from analogies
to historical building forms, plays a central role in the work of Aldo Rossi.
Through reference to history as well as memory from personal experience,
he tries to derive architectural elements that provide a new basis for design.
At the same time, Rossi always endeavors to find "prototypes" for each
particular use. This process of defining architectural prototypes provides him
with a unique freedom of design; this freedom of design allows him to obtain
different results by making the same thing over and over again. Basically he
creates his buildings by repeated use of his own architectural prototypes
Figure 1.2.1
which he transforms into new design vocabularies in response to site and
fa -

Figure 1.2.3
Twin Tower in the Museum
of German History

17
 Prototypes of Vocabularies program. These new vocabularies evolve into a physical work of architecture
through the transformation of scale, volume, and texture. Figure 1.2.2 shows
Rossi's own various prototypes which are found in most of his projects.

A design element that can be analyzed to obtain a better understanding of

Rossi's method of 'typing', is a tower in the Museum of German History.
This tower, a competition-winning project in 1988[Figure 1.2.3], is part of
the twin towers and is comprised of 3 volumes- two octagons and one
cylinder. The two basic shapes(octagon and cylinder) and the tripartite
composition of nesting one element on top of another are found in his design
projects from the early years to the recent. Figure 1.2.4 shows variations of
his tripartite forms in chronological order - how they were transformed into
different shapes in different contexts throughout the years.

Figure 1.2.2

1 2

Transformations and Variations

Thesis: Cultural Center and Theater, 1959

Secondary School, 1979
Floating Theater, 1979
A Landmark for Melbourne, 1979
Entrance Portal, 1980
Tower in Lighthouse Theater, 1987
Yatai, 1989
Urban Monument, 1989
Union Bank of Switzerland, 1991

Figure 1.2.4

Figure 1.2.4 Transformations and Variations

[21 George Stiny states that a shape This phenomenon of shape composition depicts Aldo Rossi's unique
grammar consists of rules and an initial shape. The methodology of designing a building. This methodology is a design knowledge
rules apply to the initial shape and to shapes of comprising various types of architectural elements by shape rules. Figure
produced by previous rule applications to generate 1.2.5 is a diagrammatic analysis of shape composition which is conceptually
designs. He views the generation of designs as the
analogous to Georgy Stiny's shape grammar: rules of addition and
final production resulting from adding and
subtracting shapes. According to his theory, a subtraction.[2] The rule of composition for a tower is the algebra of the sum
design begins with a single shape and a new design of base, shaft, and head. Each component can have two subshapes as shown
can be produced from the current design by adding in example1, 5, 7, and 8. The Base has initial shapes of cube, polycube
a shape to it or subtracting a shape from it. (generally octa-cube), and cylinder. These elements are changed into new
geometries of forms by scale, proportion, and texture to be harmoniously
assembled with shaft and head which also have their own prototypes. By
composing and manipulating these forms with rules of addition and
subtraction (replacement), Aldo Rossi created various types of towers with
small shifts in geometry.

TOWER Initial Vocabulary

HE H H

Head

B B 3B

Shaft H S

B
HL
B J _
HH _ _

Base
B B U-B
4 - 5 6D
6

Algebra Head + Shaft + Base

H AH E fl H

SS S

H: Head
S: Shaft
B: Base

Figure 1.2.5

Rossi generates a typological form of architecture with shape rules of

architectural grammar. He uses types of objects to plan his design projects,
and composes them according to a simple rule of algebra: addition and
subtraction. Even though his design methodology is a far cry from the actual
computer application, it supports the idea that a design can be computed with
shape information through transformation and combination operators. In his
book The Electronic Design Studio, William J. Mitchell talks about a similar
idea of shape computation. He believes that the most usual computational
operations are transformations and combinations of shapes. Further, he
believes that the computation can terminate when a designer recursively
applies those operators to the initial vocabulary until certain predicates are
satisfied by a shape produced by the operators.

Once design vocabularies are defined, we can manipulate them with few but
powerful sets of rules, and combine them into a building with the algebra of
addition and subtraction. But these vocabularies and rules should not limit
the description of design (the creative design) , nor predetermine the final
design. They must be open, flexible, and comprehensive so that they can be
changed anytime when necessary.
 2. Computation of Design Knowledge

Design is the making of a building through knowledge structure. In the

process of making, designers collect, analyze, and implement their
knowledges acquired from direct and indirect experiences. Robert Oxman
[1] McCullough, Malcolm. et al. The Electronic and Rivka Oxman, in the article The Computability of Architectural
Design Studio: Architectural Knowledge and Knowledge [1], explains that design knowledge is a collection of information
Media in the Computer Era. (Cambridge, 1990), and data necessary for integrating object descriptions with the formalization
pp. 187-199. of knowledge in design. The design knowledges are useful when they can be
represented as a form which employs its generic characteristics. Such a form
[2] John Gero calls it 'Routine Design' in the is called 'Prototype'. The concept of prototype is the generalization of
article Structural Design Knowledge as Prototypes. typological knowledge in the process of design. It helps to define proper
For the computation of design, he states that types knowledge of form for solving particular problems. In this process, however,
of design should be differentiated into 3 categories: there is the necessity of acquiring procedural knowledge of the modification
routine design, innovative design, and creative of the prototype - 'Parametric Design' [2] as well as compositional technique
design. for controlling the knowledge of the procedures - 'Shape Grammar'.

2.1 Shape Grammar

The rules of a grammar are not limiting prescriptions, but tools for constructing a
pathfrom the known to the unknown-tools that can be changed if they do not seem to
get you to the rightplace.
William J.Mitchell, The Electronic Design Studio, 1989

Grammars have been used in programming languages to design algorithms.

They facilitate the compilation process by simplifying computational
complexities. In architectural domain, shape grammar is developed to
determine how elements are combined to produce the form of an artifact by a
set of rules. George Stiny [3] explains that shape grammar defines parts of
[3] Stiny, George. "Computing with Form and
shapes with a set of rules and changes the shapes recursively to conform to
Meaning in Architecture," JAE, Fall, 1985, p. 7-19.
given spatial relations. He believes that a shape grammar is made up of an
initial shape and a set of rules; formal computation starts with a rule applied
to the initial shape, and then other rules are applied to a series of shapes
generated by previous rule applications until desired spatial conditions are
satisfied. Figure 2.1.1 shows an example of a shape grammar application.
There are two initial shapes-- a small square and a large square. Two of four
rules are addition rules and the others subtraction rules [4]. They are created
[4] In his article What Designers Do That by spatial relation of nesting one square on another. By applying these rules
Computers Should from the book The Electronic to the initial shape, one can generate different types of interesting nested-
Design Studio, Stiny believes that shapes can be squares.
manipulated by sum and difference with
transformations which are the underling concept of
shape grammar. He explains that the algebra of
sum and difference allow for shapes to be
D: - 3. DE]-KEZ
decomposed in any situation and make it possible 1 [E ]

to replace the traditional design (drawing and

erasing) with the computational design. The
sequence of algebra operations is defined by the
spatia rcuatQmi
following transformation rule: between r1Dap es

C
A -> B
t(A) <= C
(C - t(A)) + t(B)
1 1 2 4

C is an initial shape. A is replaced by B if rule appications

transformed shape A - t(A) is part of C. The Figure 2.1.1 Shape rules of addition and subtraction
final result comes from subtracting t(A) from
Shape grammar has been used to define an architectural language. One of the
C, then adding transformed shape - t(B).
most well-known analyses was the Palladian grammar presented by Stiny and
[5] Stiny, G. and Mitchell, W.J. "The Palladian Mitchell [5]. They investigate Palladio's villa ground plans to define
grammar," Environment and Planning B 5:5-18, Palladio's architectural grammar - a Parametric Shape Grammar. The Villa
1978. The analysis of paper is based on The Four Malcontenta is chosen to describe the parametric shape grammar with a set of
Books of Architecture by the Italian Renaissance composition rules [Figure 2.1.2]. They think that there are six sets of rules to
architect, Andrea Palladio in 1570. generate the villa plan: 1) grid rules , 2) wall pattern rules, 3) room layout
rules, 4) portico and wall inflection rules, 5) column rules, and 6) window
and door rules. It is their belief that the power of the shape rules is in
generating numerous villa plans of Palladian type [Figure 2.1.3]. They
successfully formulated the Palladian grammar as a parametric shape
grammar with illustrations of a step-by-step transformation of the Villa
Malcontenta plan.

There have been numerous studies of architectural styles by shape grammar,

but few actual explorations of design. This may result from not only a lack
of interest in shape grammar among designers, but also negative and passive
attitudes toward computer use in the design process. However, shape grammar
become more acceptable to designers as CAD (Computer-Aided-Design) systems
are more available and easy to use. The nature of the computer requires the logic
 09

Vie ZaTam

VllSeplva

III &nHIMf
rViaineoo
00
..LOWL2'

Figure 2.1.3 Palladian villa plans generated by six

sets of parametric shape grammar

Figure 2.1.2 Six set of rules to generate the plan for the Villa Malcontenta
 of shape grammar. In order to use a computer in design, one must know the
coordinates of point to draw lines, the geometrical relationship between lines
to construct planes, and the spatial inter-dependence between planes to make
volumes. Manipulating these coordinates and relationships does not satisfy
[6] Moneo, Rafael. "On Typology ", Oppositions the needs of man. One must know how to manipulate them for building
13, 1978. Moneo talks about the concept of 'type' humanly-supportive spaces. In this process of constructing lines, planes,
as an important aspect of design in architecture. and volumes, a designer needs a shape rule to place lines, planes , and
He believes that types are intended to play a role of volumes in a CAD environment. Without a defined knowledge of
instrument with a capability of coping with architecture-shape grammar, using computers in design will be no more
diversity of forms. Further he believes that the use
than automatic production by a frozen mechanism [6]. Initial shapes with a
of 'type' does not mean the mass-production by a
frozen mechanism; types should be repeated and set of flexible rules will guide designers to get to the right track of design
transformed to accommodate new conditions of processes. This will enable them to simplify their thinking processes from
program and context. complex design issues, speed up the design process, and come up with a
better design.

2.2 Parametric Design

By definition, a parameter is a constant whose value varies with the

circumstances of its application. The process of changing the parameter's
values is called 'parametric variation'. Parametric design is a computational
method of using parametric variation for creating various sub-shapes from
one shape (prototype) to fit into a certain design environment through the
implementation of a computer program. In the architectural realm, the
parametric design has been studied for form-generation and spatial
exploration. It can be a useful design method, particularly, in the design
process when a shape has to be changed to fit into different spaces in
different situations. By assigning specific values to the parameters of a
shape, it can be transformed into a particular geometry of form for a
particular situation.

[7] McCullough, op. cit, pp. 77-90 There is a consensus among CAD researchers that parametric design has been
successfully applied in many areas including architecture. In Classesof Design
- Classes of Tools (1990) [7] Gerhard Schmitt states that parametric design
assumes the following three conditions: 1) The designproblem is well defined
and client requirements are well understood; 2) There exists a parameterized
prototype for the type of design that is to be developed and a data base of
parametervariations;3) The final design may be derived by refining,but not
fundamentally changing the prototype. He states further that the parametric
design is useful in phases of schematic design, preliminary design, design
development, contract documents, and shop drawings except for program
development and construction drawings.

Parameterized shapes are advantageous for their adaptability to many

different conditions. For example, a parameterized window can fit into any
opening of width and height inside a wall. Once the window is parameterized
with a programming language, such as Autolisp, it can be reused again and
again by simply inputting different values of parameters for width and height
whenever sizes of window openings change. This flexible capability
provides a designer with a wide range of alternative forms in the process of
making a building which ,in the end, helps him or her to come up with a
better architectural solutions.

There are two advanced approaches to parametric design: top-down and

bottom-up design methods. They have a common characteristic in that they
deal with more than two shapes at once and the parameters of one shape are
dependent on the parameters of the other. For instance, if you change a
parameter of one of the shapes, the shape with the changed parameter not
only transforms itself into a new form of geometry, but also affects the other
shape so that the latter changes in relation to the former. On the other hand,
the difference between top-down and bottom-up methods is the structuring
sequence of parametric shapes. In top-down design method, smaller shapes
(sub-shapes) are parameterized to fit within a larger shape so that they can be
correlated properly when any of the shapes changes. Take a classical column
as an example. In general, a classical column consists of a base, a shaft, and
a capital. If a parameter of the base is changed, these sub-shapes have to be
parameterized with correct coordination of dimensions and proportions. The
geometry of the shaft and the capital as well as the base itself, are affected
while the overall figure of the column is maintained.

On the other hand, the sub-shapes do not have to be parameterized within a

larger frame of shape in the bottom-up design method. They are
parameterized independently in the beginning, but they are adapted , when
necessary, to create a desired shape. In this bottom-up fashion, the base,
shaft, and capital are independent elements. They can be used in a
combination of pieces that achieves any desired result. A combination of a base
and a shaft or a base and a capital can be adopted to create different kinds of
 shapes in different contexts. There is a general consensus that top-down and
bottom-up design methods are complimentary. When some shapes are inter-
dependent, they can be parameterized to fit within a geometry resulting from
their combination in top-down fashion. Otherwise sub-shapes are designed
using the bottom-up method to be compatible with any other shapes
parametrically so that their combination can produce a desired shape for a
particular situation.

As design becomes more complex, different conditions at different stages of

design demand different characteristics of architectural shapes. In this
respect, parametric design with top-down and bottom-up design methods can
be an efficient method of manipulating and transforming shapes and provide
design alternatives that can rapidly respond to the changing needs of design
situations.

2.3 Knowledge-Based System

Computer programs, in which human knowledges are represented in

symbolic forms, are called 'knowledge-based systems'. They are based on a
relationship between types of structured knowledge. From the architectural
point of view, the knowledges are encoded as pre-defined structures which
are delivered to the user (designer) through the computer program. The
knowledge-based systems are concerned with the manipulation and modification
of design knowledges , while they provide assistance to designers in the
[8] Mitchell, William J. et al. The Art of Computer process of decision-making.
Graphics Programming:A Structured Introduction
for Architects and Designers. New York: Van
One of the best-known knowledge-based systems is Top-Down developed by
Nostrand Reinhold, 1987. The book explains
extensively the top-down method by PASCL William Mitchell, Robin Liggett, and Milton Tan.[8] They demonstrate that
program. the knowledge system can be useful to a designer when he or she develops a
particular element or subsystem of a building. The system has two graphic
windows - the peek window and the poke window. [Figure 2.3.11 The peek
window displays the design state in use; the poke window shows the result
from the input in the peek window. As can be seen in Figure 2.3.2, one can
choose the desired object in a dialog box to perform a transformation. Actual
parametric variation can be accomplished by using the slide bar in the poke
window. [Figure 2.3.3] Mitchell, Liggett, and Tan use classical columns to
illustrate the operation of Top-Down with several types of capitals, shafts,
 SFPmEdit Ulea SpeciS Color
Peel * ClaOsIe Poke # CiSeti

Figure 2.3.1
Peek, Poke, and taeoke.e-ewukeeastserln
Teesho**W-Welbttlsen
message windows
* flil Optisas
Slew Snimte Ilesm
Peek C oksst Poke e Classic

e.im . t .. cE .
Osis... , PCg

Figure 2.3.2
Transformation
thvatseeteuee-sCieeni

in Poke window
le t11e m o w P 1ierIe elsp
Pke Clae

I
Figure 2.3.3 ll
T.khing
eW1,-ekl.pmm - '"A
las de.y-emew lt4geI*d
|OW
lit t.

Parametric variation Tateletwe-iCeLdoMaswuobdeenme-tetaid

in Poke window
 Figure 2.3.5 Steps in elaboration of an Ionic column

I Il
Figure 2.3.7
Variations of out-of-proportioned columns
I I

Wv

ii
Figure 2.3.6 Figure 2.3.4 A vocabulary of capitals and bases
Variations of proportioned columns
and bases. [Figure 2.3.4] Then they show how a long rectangle can be
transformed into an classical Ionic column as shown in Figure 2.3.5.
Through such an operation, the knowledge system can create various classical
columns [Figure 2.3.6] as well as interesting geometries of objects [Figure
2.3.7].

The Top-down knowledge system demonstrates how an object can be

manipulated to get a desired shape in a top-to-bottom fashion, and how a
knowledge encoded by shape grammar can be implemented to perform a
design operation. But Mitchell, Liggett, and Tan do not explain how the
system can be actually applied to the design process. Furthermore there is a
lack of examples of architectural vocabularies, such as doors, windows, roofs,
and walls. Most of them are two-dimensional and abstract. However, while
there are some limitations to the system, the top-down system seems to
provide great potential for the future development of knowledge-based
systems.

The use of computer requires that a design knowledge be encoded to

represent and symbolize architectural forms. The computation of design
necessitates the formalization of the design knowledge - 'shape grammar' to
construct compositional structures. By generalizing and classifying design
knowledges - 'prototyping knowledges', one can establish a link between
design and computer, thus executing manipulative operations of architectural
forms. The parametric design provides a wide variety of flexible tools for
the transformation of the architectural forms in various programs and
contexts. With the theory of shape grammar and the computation of
parametric design, the design knowledges can be used effectively and
creatively with the talent of a designer.
I I

TO?. 01 TNO14v-
To o

I
Vocabularies of canales and water bins from CONCEPT SOURCEBOOK:
a vocabularyofarchitecturalformsby Edward T. White
 The buildings I want to design are assemblages of forms put together to create a
figure reminiscent of synthetic cubism. They are composed of individualforms joined
together as a single architectural identity by means of shape rules. Architectural
order is created by applying rules of a grammar recursively to these forms until
they are successfully linked to the aspect ofplace and the need of man.

Lee Hee Won, The Computation of Design Vocabulary, 1997

31
 3. Investigation and Exploration

This thesis investigates and explores the idea of using computer in design
process based on Aldo Rossi's 'type' concept of design vocabulary. [1] Two
composition techniques -- top-down and bottom-up methods -- are evaluated
to determine the advantages and disadvantages of both techniques through
actual design explorations in an effort to answer the question of how a
computer can be used efficiently and creatively in design. By the design
exploration, the role of shape grammar is also investigated to test its possible
application in the design process away from the conventional theory and
analysis. Furthermore, the top-down and bottom-up methods are encoded
through a program to demonstrate their capability of parametric variation. Up
to today, a program has been one thing; design has been another. There has
been no inter-dependent relationship between them other than analytical
kinship. For this reason, two program examples of building components are
analyzed from my design experience and illustrated as if they were in actual
design situation. This investigation and exploration is an endeavor to learn
how a designer can use a computer in design process, but not a story to tell
how a computer can support a designer in design process.

3.2 Experimental Design

Based on Aldo Rossi's 'Type' concept of design vocabulary, two design

projects were experimented for using a computer in the design process from
the conceptual idea to the final production. Each design was executed with
the following three questions: 1) Is it possible to use a computer from
"scratch to finish"? 2) Which design method is better in design process--
top-down method or bottom-up method? Or do they compliment each other?
3) Is it possible to lay a computational foundation for an architectural
principle? Due to the experimental nature of this process, mistakes were to
be expected.

3.2.1 The Summer Academy

This project was executed in Antonakakis studio (Fall, 1995). The design
Figure 3.1 approach was carried out in a top-down fashion. This was my first attempt to
Rossi's sketches use a computer in design process from "start to finish".
 -E

[1] Figure 3.1 shows Aldo Rossi's sketches The design goal of the Summer Academy was to promote architecture and its
through which he repeatedly used his vocabulary relation to other arts and cultural activities in a cultural environment. The site
elements to generate ideas and create architectural is located at Chanea, the north-western part of Crete, in Greece, in a recess of
spaces. the Bay of Chanea flanked by the Akrotiri peninsula and the cape of Spathi.
[Figure 3.2.1.1] The building of the Summer Academy was to be built on the
queue of the old harbor. The program includes four main spaces:
administration, teaching Area, performance area, classrooms, and guest
rooms. The total area was about 25, 000 square feet.

The parti resulted from understanding the nature of program and context.
The idea was to create a courtyard scheme to provide a variety of public
spaces by placing the courtyard scheme in the site. The public spaces include
small alleys, a man-made plaza, and open space to be left as it is. The
program required that the museum including theater, exhibition space, and
reception function independently from the school. The solution was to
separate two major programs which are connected by courtyard to promote
the inter-activity between the school and the museum.
Figure 3.2.1.1 Site
Initially my design vocabulary [2] was used to generate the courtyard
scheme, but the scheme had to be given up because no compositional rule
could be developed to manipulate the vocabulary elements due to my lack of
[2] Figure 3.2.1.2 shows my own vocabulary
experience in computational design. Geometric volumes were created in
called 'Cubic Vocabulary' developed for design
experiments. I admire architectures of Richard response to the program requirement. [Figure 3.2.1.4] They were thrown into
Meier, Charles Gwathmey, O.M. Ungers, Tadao the site to create a courtyard scheme with scale to a surrounding environment
Ando, and Aldo Rossi. The conceptual scheme in as can be seen in Scheme 1. Then the physical model in Scheme 2 of Figure
Figure 3.2.1.3 was generated from cubic 3.2.1.5 was made to represent the overall scheme three-dimensionally [3].
vocabulary elements. When I generate the This led me to go back to the computer modeling to articulate each volume in
vocabulary elements, I consider the architectural detail. From this stage, the problem of the top-down design method became
forms of five architects as 'prototypes' of my evident as geometries were complex. It was difficult to understand the inside
vocabulary. Then I try to derive my vocabulary space because of the overlapping of lines. It took a lot of time to see the
elements from them by analyzing their geometry
rendered image which still did not reveal the quality of the interior space.
and sometimes borrowing some of their
vocabularies. Another problem was that a new form had to be redesigned from "scratch"
whenever a new design state was changed. This was not an efficient method.
[3] From the beginning, it was my intention not to The biggest problem was to communicate ideas with the studio instructor.
use a physical model to experiment with the use of The abstract computer model as seen in Scheme 3 did not convince him
computer in design process, but the studio because I was unable to represent design ideas with a three-dimensional
requirement hindered the computational process. model. After Scheme 3, I had to depend heavily on two-dimensional
drawings [Figure 3.2.1.6] to communicate ideas better as well as to satisfy
the studio requirement with a series of superficial physical models. But the
3W aELE 3DOORD MDOORE 3D00RF 3D00RG 3DOORH 4 3EAVEA1 3EAVEB1
3EAVEA

3CHIMD 3CHIMC t

GLS 3GLAS0 3GLASE 3GLASF BAC

3WALLt

3GLASG 3GLASI 3GLASJ

313ALCA

S 3GLASK

3GLASL 3GLASM 3AL

3D00A2

3GRIOD LlD~m 3CHIMA

3GRIIJA 3GRIDC

3LAMPA 3LAMPC - - 3DOORB

,ItAPB13LAPJPD LAMPF

3COLUC

3DOA3EAVEB2 3MONUA 3COLUA

300RA3WALLC 3MONUB

3 FENCA1
3WALLB
3GA
TEA
3 O U

WCHIMS 3COLUB1 3GATEB GTC3LS

Figure 3.2.1.2 Cubic Vocabulary

 Figure 3.2.1.4
Geometric Volumes

Schemes Figure 3.2.1.5 Development of Schemes

Figure 3.2.1.6 Sketch of plans, elevations, and sections

36
process of drawing plans, sections, and elevations as well as making the
physical models was not helpful for understanding and exploring the dynamic
space between objects.

Top-down design is supposed to start with a set of rules to articulate

vocabulary elements, but this rule of thumb was not used in this project due
to my present lack of experience in using computer in the design process.
This seemed to result in the unsatisfactory design process and product. Thus
it might be not fair to say that the top-down design is not an efficient design
method. However, from my limited experience, the method seems not to be
a good design tool for exploring the dynamic, flowing space between three-
dimensional volumes. Furthermore to start from scratch whenever a design
state is changed, seems to hinder the continuous design process with other
design conditions.
 3.2.2 The Visitor Pavilion

This project was executed in the William Mitchell/Andrew Scott studio

(Spring, 1996). The design approach was carried out in a bottom-up fashion.
The intention of the project was to engage in the act of design using a com-
puter as a medium of exploration. The primary focus was on understanding
the role of computer-controlled prototyping and fabrication in the creative act
of design and making form.

The pavilion is a Visitor pavilion celebrating sports in America. It will be

located in Washington, on Maryland Avenue close to the Capitol Building.
[Figure 3.2.2.1] It will provide up-to-date information about sporting events as
well as enable visitors to learn about American Sports. The program requires
a public space, information center, exhibition space, space for audio visual dis-
play, and lockable storage totaling 4000 to 6000 square feet. The Visitor
Pavilion is a permanent structure , but is only intended to be used seasonally,
from May through September, so that it will not be heated or cooled. How-
ever, it will rely on passive based ventilation and cooling together with a self
sufficient energy source-photovoltaic systems.

The design concept was derived from considering two major program issues-
fabrication of components and control of natural lights. Initially 'a kit of
parts' idea was developed to investigate how building components can be
manipulated for prototyping and fabrication. Their architectural forms are
generated from responding to the program and site. Then to organize and
articulate the space for controlling natural light, the 'Five Faces' concept was
introduced; east face was created for admitting the light in half and rejecting in
Figure 3.2.2.1 Site in half; west face for denying the light; south face for shutting off the light but
allowing indirect light; north face for celebrating the light; and finally, sky
face for welcoming the sunlight to utilize the photovoltaic panels. Aestheti-
cally these elements meant to create five distinctive facades, but the challenge
was to assemble them harmoniously as one building entity.

Inside space was divided into two zones by a wide bridge which functions as
both gallery and circulation path. One is a space for celebrating light, serving
as an entry and lobby; the other is a space for blessing darkness where major
artifacts are displayed. The entire circulation system was designed to experi-
ence four different walls so that the circulation starts with the northern wall
 (leading visitors to the second floor), moves to the eastern wall, crosses the
major pavilion space to arrive at the western wall, and then ends with southern
wall (taking visitors to the ground level). Spaces under the bridge accommo-
date a video display area, office, restrooms, and storage. Steel was used for
North South structural components- columns and beams; aluminum was used for non-
structural elements, such as window, exhibition panel, and fins for shading.
The ground floor was made of black marble so that it could reflect surrounding
components, thereby casting interesting shadows with a combination of direct
East ast and indirect lights.

At the initial design stage, 4 distinctive wall elements, a roof form, and two
interior volumes were created with some of my existing vocabulary intro-
Public Private duced. A compositional rule was set up to manipulate and organize the
vocabulary elements. [Figure 3.2.2.2] The first scheme of the project was
composed by the intial shapes based on this composition rule. [Figure 3.2.2.3]
After this scheme, there were three computatinal transformations as can be
Sky seen from Figure 3.2.2.4, Figure 3.2.2.5, Figure 3.2.2.6). In this process of
making a building, the shape rule of 'addition and deletion' was applied for the
Initial Vocabulary
transformation of the vocabulary elements. Figure 3.2.2.7 shows how initial
shapes are manipulated and transformed by adding new shapes and deleting
old shapes in response to changing needs of the design states.

The bottom-up design approach which I attempted here is not a perfect dem-
onstration, but an investigative experiment. There was some inconsistency in
the way vocabulary elements are manipulated and replaced. Some initial
shapes maintained their original characteristics of geometry with some
changes in scale, proportion, and texture throughout the design development,
Cubic Vocabulary but other shapes were discarded to meet the changing design conditions with
the generation of new shapes. Figure 3.2.2.8 show numerous sketches of
building components in the design process; some of them were gone and oth-
E
ers survived in the final stage of design.
-:1 S
Through this experimentation, the bottom-up design approach was found to be
N an efficient method of exploring form and space to take advantage of the com-
W puter capability (that is, inserting, deleting, moving, rotating, and copying).
Compositional rule From the beginning, a few objects were manipulated to compose a space by
moving and rotating them, and to build up complex spaces by the simple func-
tions of inserting and copying with deleting unnecessary objects. The bottom-
Figure 3.2.2.2
up method did not seem to pretermine the final form of design unlike the
Figure 3.2.2.3 Scheme 1

40
 Sketch of forms

Forms for
Scheme 2 Figure 3.2.2.4 Scheme 2
 00

Forms for
Scheme 3 Figure 3.2.2.5 Scheme 3
-40F

Final form Sketch of forms

Figure 3.2.2.6 Scheme 4

 T 0 T
to T 00

N MO, N T

*% T N N

E E TN

T T T

T: transformation Forms for Forms for Forms for Forms fo

N: new shape Scheme 1 Scheme 2 Scheme 3 Scheme 4
TN: T+N
E: existing shape
Figure 3.2.2.7 Manipulation of vocabulary elements
Figure 3.2.2.8 Sketch 1 of building components
 00

il
Figure 3.2.2.8 Sketch 2 of building components
El~zzu
I

0
7z
Figure 3.2.2.8 Sketch 3 of building components
top-down method. Even though a designer has a abstract picture of a final
building form in the inital design stage, my belief is that the bottom-up method
is open to the changes of design states for shaping a building form and pro-
vides quick alternatives for design decision-making in order to determine a
final form. While the top-down method tends to have a fixed direction to get
to the final building form, the bottom-up method provides flexible ways of
manipulating shapes to meet the intention of a designer. In addition, a
designer can understand the inside space much better in the process of making
a building in the bottom-up fashion than in the top-down fashion. Because the
space is constructed with smaller objects from the beginning design stage in
the bottom-up method, the designer can have the clear perception of larger
space which is made out of the smaller vocabulary elements even though the
geometry of the larger space becoms dense and complex.

From my experiments with two design projects, I believe that the bottom-up
method is a better design technique than the top-down method to utilize the
computer capabilities of manipulating objects, such as inserting, copying, and
stretching. Furthermore, the bottom-up method seems to explore inside spaces
better inherently in the process of making a building away from the issue of
'computer use' in design. It provides a designer with a better way of exploring
complex spaces by providing a wide array of design alternatives for solving
architectural problems. But there is a general consensus among CAD
researchers and designers that both the top-down and bottom-up methods are
complimentary and it may be impossibe to determine one absolute way of des-
ing a building because of the subjective and artistic nature of architectural
design. Also my experiment was limited to only two projects. More design
experiments need to be excuted to find out the advantages and disadvantages of
both bottom-up and top-down methods in the use of computer in the design
process.
 Axonometric View from North-West

Bird's Eye View

from South-West

Axonometric View from South-East

Second Floor

First Floor
Perspective view from North

51
3.3 Knowledge Application

Manipulating and composing vocabulary elements is an integral part of the

design process. Complex design issues require the flexibility and adaptability
of the vocabulary elements to meet different spatial conditions at various con-
texts. By encoding a knowledge of vocabulary elements into a program, one
can transform an object type into numerous subtypes, thus testing numerous
spatial ideas with a wide range of vocabulary elements. This enables a
designer to come up with better design alternatives for solving an architectural
problem.

3.3.1 Computation of Design Vocabulary

Designing a building is like making a sppeech. Before making a speech, we

must know words before sentences are made, and we must know rules of com-
position before paragraphs are constructed. A good speech can be made only
with a good combination of words, sentences, and paragraphs. When we
design a building, we must recognize the basic elements of architectural
vocabulary which can be manipulated for solving a design problem. The basic
architectural forms serve to lay out the organization of space and their trans-
formation into a building. They become the critical means of construction
which provides a primary tool for designer.

A built environment can be segmented into parts by name as shown in Table

3.3.1.1. Designers describe a building by combining these basic elements. In
this process of combining the vocabulary elements, designers must know the
mutual relationships beteen the basic elements in order to compose them har-
moniously for the making of an architectural space. Table 3.3.1.2 shows the
numerous shapes that are composed of smaller elements. The composition
rule is based on Georgy Stiny's shape rule of 'addition' and 'subtraction' for
the computational application. The shapes can be further combined into archi-
tectural spaces [Table 3.3.1.3] which can be eventually assembled to create a
complete building identity. [Table 3.3.1.4]

Once the vocabulary elements are defined and constructed with parametric
variation capability in a computer, they can be reused and transformed easily
by creating various derivatives from them to meet the changing needs of
design states. This CAD-generated form can be a powerful tool for design if
one can make it as a prototype, and create various derivatives from it with
Taxonomy
Table 3.3.1.1

arch elevator panelboard

baluster equipment room pantry
balcony escalator parapet
bathroom fence parlor
beam fire detector pattern
bed fire extinguisher pavilion
bedroom fireplace pergola
bench flag photovoltaics
bollard floor planter
bookstacks fountain pool
bridge freezer porch
bumper gallery portico
cabinet gateway railing
carport garage ramp
ceiling garage door roof
chair grid shade
chimney grille shelf
cloakroom gutters sign
clock handrail sill
closet hallway sink
column jamb screen
connector kitchen skylight
corridor ladder sofa
courtyard lighting soffit
deck lamp stack
desk laundry stair
den lavatory step
diningroom livingroom step-ladder
dishwasher linen stool
door mailbox storage
downspouts mirror switch
dressingroom monument table
drinking fountain nosing tower
duct nook window
eaves overhead door walkway
Shape Algebra
Table 3.3.1.2

Balcony = (+) railing Roof =

(+ -) overhang (+)ceiling
(+ ) door Tower = (+)beam
(+) plane (+)joist
(+ -) flag
(+ -) support (+)deck
(+ -) ornament
(+ -) capital (+)roofing
Chair = (+) shaft
(+) plane (+ -) base Skylight =
(+) back (+)ver-beam
(+) side Window = (+-)slope-beam
(+ -) leg (+)hor-beam
(+-)lintel
(+) mullion
Chimney = (+) frame Door =
(+ -) cap (+ -) sill (+)plane
(+) flue (+-)shading device (+-)frame
(+) fireplace (+ -) overhang
(+-)curtain Frame =
Column = (+)ver-beam
(+ -) capital Ex-wall= (+)hor-beam
(+) shaft (+)coping
(+ -) base (+-)gutter Ceiling =
(+-)sprout (+)lighting
Desk = (+-)window (+)grille
(+) plane (+-)gateway (+-)duct
(+ -) ornament (+-)door
(+) leg (+-)opening Mailbox =
(+-)box
Gateway = Ramp = (+-)shaft
(+)facade (+)rail (+-)base
(+) door (+)walk
(+-) portal In-wall =
(+-) canopy Truss = (+-)opening
(+)top-cord (+-)grille
Pergola = (+)bottom-cord (+-)lighting
(+) beam (+)diagonal (+-)drinking-fountain
(+) column (+-)bench
Floor = (+-)shelf
Stairs = (+)ceiling (+-)panelboard
(+) tread (+)beam
(+) steps (+)joist Rail =
(+ -) stringer (+)deck (+)horizontal bar
(+ -) body (+-)flooring (+)vertical bar
Space Algebra Built Worlds
Table 3.3.1.3 Table 3.3.1.4

Entry(public) = Bedroom = Museum

(+)gateway (+-)bathroom
(+-)vestibule (+)bed
(+)desk School
(+)lobby
(+-)dresser
(+)closet House
Lobby = (+)chair
(+)furniture (+-)balcony High-rise
(+)colonnade
(+-)elevator
(+-)stair
Dining = Shopping Mall
(+-)skylight (+)kitchen
(+)table
(+)wall (+)chair Supermarket
(+-)glass
(+-)monument Library
Walkway =
(+-)pattern
Entry(private)= (+-)street-lighting Theater
(+)gateway (+-)bench
(+-)vestibule (+-)kiosk Bank
(+)livingroom (+-)bollard
(+-)trash-container Prison
Livingroom =
(+)furniture Courtyard = Furniture Store
(+)colonnade (+-)wall
(+-)chimney (+-)glass Airport
(+-)stair (+-)fountain
(+-)skylight (+-)monument Restaurant
Closet = Hallway = Parking Garage
(+)shelf (+-)in-wall
(+-)drawer (+-)glass
(+-)door Car Dealer
Kitchen = (+-)ceiling
(+-)skylight Aquarium
(+)refrigerator
(+)dishwasher
(+)sink Plaza = Church
(+)cabinet (+-)monument
(+)oven (+-)tree Nursing Home
(+-)wall-oven (+)walkway
(+)houseware (+)pattern Gymnasium
 parametric variation capability, so that one can superimpose and transform
them for spatial explorations in alternative compositions. This enables a
designer to explore form and space and to produce a great number of plausible
schemes efficiently for solving design problems.

3.3.2 Top-down and Bottom-up Methods

[1] NITROS is short for Nagakura's Network- With the design knowledge acquired from previous design experiments, two
interfaced Inter-Type Relation Operating Shell. computer programs were written using Autolisp with two knowledge-base
Developed by Takehiko Nagakura, NITROS design approaches: one is bottom-up design method and the other top-down
interface enables a user to assemble a complex method. NITROS [1], a software shell, was used to implement the programs
geometry by instantiating a formal type, trans- in Autocad environment. Through the interface with HTML, the NITROS
forming it into another, parametrically revising can be dispatched to an Autocad environment; defined formal types (that is,
the resulting assembly, and editing its constraint
pergola and gateway) can be loaded.
progagation paths.
In top-down method, one starts with a simple volume, and then articulating it
into smaller volumes. Smaller subshapes are constrained parametrically to fit
within a large shape so that they can affect each other when any of the shapes
changes. Figure 3.3.2.1 shows how a simple rectangular box is articulated into
a detailed pergola. The pergola is composed of a row of columns and capitals.
If a parameter of columns is changed, not only the shape of columns is
changed, but also the form of capitals is affected. By parametric variation, the
pergola can be transformed into multi-shpaes in response to the progrmam
and site. Figure 3.3.2.2 shows the interface between NITROS and HTML for
dispatching, loading, and transformation of the pergola.

In bottom-up method, one starts with a detailed element of form, and then adds
new elements to generate a desired shape. Individual vocabulay elements are
parameterized independently in the beginning, but they are combined differ-
ently to create different types of shapes. In Figure 3.3.2.3, a door is the begin-
ning element of the composition. There are three optional elements (that is, a
portal, a wall-face, and a canopy). One can choose different combinations of
elements to obtain different results. The door and a wall-face can be combined
to create an ordinary facade and then a portal or a canopy or a combinatin of a
portal and a canopy can be added to the door for the creation of different types
of facades. It seems that the bottom-up design technique provides a effiecint
method of manipulating shapes for spatial exploration. There are limitless
possibilities of combining one vocabulary element with others. For example,
the door can be combined with other elements to create not only a gateway, but
Top-down
Process Pergola = Capital + Shaft Parametric
Variations
Figure 3.3.2.1 Top-down Tree
Figure 3.2.2.2 Pergola loaded and dispatched from Netscape to Autocad through NITROS

58
 Bottom-up process of
making a gateway

Gateway = Door + (Wallface/Portal/Canopy) Parametric

Variations
Door Variations Figure 3.3.2.3 Bottom-Up Tree
Figure 3.2.2.4 Gateway loaded and dispatched from Netscape to Autocad through NITROS

60
also a hallway, a lobby, and any type of occupied space. Figure 3.3.2.4 shows
the interface between NITROS and HTML for dispatching, loading, and trans-
formation of the gateway.

It seems that top-down and bottom-up design methods are complimentary.

When some shapes are inter-dependent, they can be parameterized to fit within
a geometry resulting from their combination in top-down fashion. Otherwise
sub-shapes are designed using the bottom-up method to be compatible with
any other shapes parametrically so that their combination can produce a
desired shape for a particular situation. However, from my own experience,
the bottom-up method seems to be a more efficient method of manipulating
shapes for creative design. The method is more flexible for adaptation to dif-
ferent conditions at various contexts. It is not limited to a pre-determined
form, but provides limitless possibilities for combining vocabulary elements
for various design alternatives.
Cubic Vocabulary
Reflection

Design is the art of assembling forms harmoniously to create space for human
needs. From design concept to final production, designers manipulate forms
until the forms are transformed into a building entity. The broad knowledge
of architectural vocabulary navigates a designer to understand design activity,
and enables him or her to lay out forms to create space.

This thesis developed a set of architectural vocabulary for experimenting

with design projects. Based on the Aldo Rossi's methodology, my own
vocabulary was developed to investigate how forms could be manipulated,
transformed, and composed with top-down and bottom-up methods. By
actual design experimentation, the advantages and disadvantages of the two
methods were discovered: the current computational theory was evaluated:
and a new direction for developing software tools for creative design was
explored.

The project to design a school was approached in the top-down fashion.

From the beginning, simple volumes were created in response to program
requirements, and then gradually detailed into smaller, concrete volumes.
Because it was my first attempt to use a computer in the design process,
several obstacles were met right from the design development.
Communicating ideas was difficult due to my inexperience in the use of a
computer in the design process. But the most difficult was the top-down-
design itself. Each time the design state was changed, most of the volumes
had to be redesigned from the beginning to meet a new design state. The top-
down design was inefficient for keeping up with the speed of changing ideas,
and difficult to explore spatial relationships between objects - especially for
understanding the inside space as complex geometries are overlapped.

In contrast to the school project, the pavilion project was developed with a
bottom-up design method. In response to the program requirement, initial
detailed forms were created; some of these forms were recycled from my
existing vocabulary while others were newly introduced, and then built up to
organize a space. A set of rules were created to compose forms in the early
stage. These rules enabled me to simplify ideas from complex design issues
and articulate form elements with mutual relationships.
With experience gained from two design experiments, two computer
programs were written using Autolisp to investigate the role of design
knowledge encoded by programs. It was not an easy task to learn the
program, clarify the relationship between building components, and
implement the parametric variation with top-down and bottom-up methods.
But it seems that the encoded knowledge is extremely powerful once all the
building components are defined parametrically with mutual relationships.
The parametric variation seems to provide a designer with a wide range of
design alternatives in response to the changing needs of design states.

In architecture, there is no absolute way of designing a building. Due to the

artistic nature of design, none can suggest or teach a design doctrine which
seems to solve all architectural problems. However, my belief is that there is
a more efficient design method in the use of a computer and the bottom-up
design method can be a powerful technique to explore form and space with
the idea of typing vocabulary. This conclusion about the design methodology
is not meant to be universally accepted, but to set out my journey and search
for an principle to which I can look up for simplicity in architectural life.
Illustration

Figure 1.1, p. 12 (Ungers, Oswald Mathias: 1994)

Figure 1.1.1 and Figure 1.1.2, p. 13 (Madrazo, Leandro: 1994)

Figure 1.1.3, Figure 1.1.4, and Figure 1.1.5, p. 15

(Thiis-Evensen, Thomas: 1988)

Figure 1.2.4, p. 18 (Adjmi, Morris: 1991/ Arnell, Peter and Bickford: 1985)

Figure 2.1.2, p.2 3 (Stiny, George: 1985)

Figure 2.1.3, p. 23 (Mitchell, William J.: 1990)

Figure 2.3.1 - Figure 2.3.6, pp. 27-28(McCullough, Malcolm: 1990)

Figure 3.1, p. 32 (Adjmi, Morris: 1991)

Bibliography

Adjmi, Morris. Aldo Rossi: Architecture 1981-1991. New York:

Princeton Architectural Press, 1991.

Anderson, Stanford. Types and Conventions in Time: Toward a

Historyfor the Durationand Change ofArtifacts. Perspecta 18, 1982.

Antoniades, Constantine G. The Function of Repetition in Typology. Thesis

Arch. 1988. M.Arch.

Argan, Giulio Carlo. "On the Typology of Architecture", English translation

by Joseph Rykwert, Architectural Design, Dec. 1963.

Arnell, Peter and Bickford, Ted, ed. Aldo Rossi: Buildings and Projects. New
York: Rizzoli, 1985.

Baker, Geoffrey H. Design Strategies in Architecture: an Approach to the

Analysis of Form. London: Van Nostrand Reinhold, 1989.

Baker, Geoffrey H. Le Corbusier: An Analysis of Form. New York: Van

Nostrand Reinhold, 1984.

Baudini, Micha. Typology as Form of Convention. AA Files, no 6, May 1984:

73-82.
Ching, Francis D. K. Architecture, Form, Space, and Order. New York: Van
Nostrand Reinhold, 1979.

Colquhoun, Alan. "Typology and Design Method," in Essays in

Architectural Criticism. Cambridge: MIT Press.

Frank, Karen A. et al. Ordering Space: Types in Architecture and Design.

New York: Van Nostrand Reinhold, 1994.

Gero, John J. et al. "StructuralDesign Knowledge as Prototypes." Sydney:

The Architectural Computing Unit. Department of Architecutal Science,
University of Sydney, 1988.
Frampton, Kenneth. et al. Oswald Mathias Ungers: Works in Progress.
New York: Rizzoli, 1981.

Jacobs, Stephen Paul. The CAD Design Studio: 3D Modeling as a

FundamentalDesign Skill. New York: McGraw-Hill, INC., 1991.

Jules, Frederick. Form/Space and the Language of Architecture. Milwaukee:

Center for Architecture and Urban Planning Research, University of
Wisconsin-Milwaukee, 1979.

Jung, C.G. "Abstracts of The Collected Works of C.G. Jung," A Guide To

Collected Works, Volumes I - XVII: Bollinger Series XX . Princeton
University Press, 1968.

Hermander, Antonio. J.H.L. Durand's Architectural Theory. A study in the

History of Rational Building Design.

Lansdown, J and Roast, C. "The possibilities and problems of knowledge-

based systems for design" Environment and Planning B: Planning and
Design, Volumne 14, 1987. Pp. 255-266

Lavin, Sylvia. Quatremerede Quincy. Pp. 86-100.

Madrazo, Leandro. "Durandand the Science of Architecture" Journal of

Architectural Education, September, 1994, p. 12-24.

March, Lionel, ed. The Architecture of Form. Cambridge: Cambridge

University Press, 1976.

McCullough, Malcolm. et al. The Electronic Design Studio: Architectural

Knowledge and Media in the Computer Era.Cambridge: MIT Press, 1990.

Mitchell, William J. The Logic of Architecture: Design, Computation, and

Cognition. Cambridge: MIT Press, 1990.

Mitchell, William J. et al. Digital Design Media: A Handbookfor Architects

and Design Professionals.2nd ed. New York: Van Nostrand Reinhold, 1995.
Mitchell, William J. et al. The Art of Computer Graphics Programming: A
Structured Introduction for Architects and Designers. New York: Van
Nostrand Rinhold, 1987.

Moneo, Rafael. "On Typology", Oppositions 13, 1978.

Moos, Stanislaus von. Le Corbusier: Elements of a Synthesis. Cambridge:

MIT Press, 1979.

Nagakura, Takehiko. Form-processing: A System for Architectural Design.

PHD dissertation at Harvard University, 1995.

Oechslin, Werner. Premises for the Resumption of the Discussion of

Typology. Assemblage 1, 1986:37-52.

O'Regan, John, ed. Aldo Rossi: Selected Writing and Projects. London:
Architectural Design, 1983.

Quatremere de Quincy, Type. Introduction by Anthony Vidler, Oppositions 8.

Cambridge:MIT Press, 1977.

Rossi, Aldo. A Scientific Autobiography. Trans. Lawrence Venuti.

Cambridge: MIT Press, 1981

Smith, Max J. Space-Form: Computer-Aided-Designfor Architecture. Salt

Lake City: Architectural Research, Information Processing Systems,
University of Utah, 1968.

Stiny, George. "Computing with Form and Meaning in Architecture," JAE,

Fall, 1985, p. 7-19.

Stiny, G. and Mitchell, W.J. "The Palladiangrammar," Environment and

Planning B 5:5-18, 1978.

Schmidt, Gerhard. Microcomputer Aided Design for Architects and

Designers. New York: John Wiley & Sons, 1988.

Ungers, Oswald Mathias. Architectureas theme. New York: Rizzoli

Schirmbeck, Egon. Idea, Form, and Architecture: Design Principles in
Contemporary Architecture. Trans. Henry J. Cowan. New York: Van
Nostrand Reinhold, 1987.

Ungers, Oswald Mathias. Oswald Mathias Ungers Trans. Pamela Johnston.

London: Studiopaperback, 1994.

Ungers, Oswald Mathias. City Metaphors. Kaln: Verlag der Buchhandlung

Walther Konig, 1982.

Thiis-Evensen, Thomas. Archetypes in Architecture. Oslo: Norwegian

University Press, 1988.

Vidler, Anthony. The Idea of Type: The Transformation of the Academic

Ideal, 1750-1830. Oppositions 8. 1977.

Vidler, Anthony. The Production of Types. Oppositions 8. 1977:95-115.

Villari, Sergio. Sergio Villari:J.N.L. Durand(1760-1834):Art and Science of

Architecture. Trans. Eli Gottlieb New York:Rizzoli, 1990.

Yessios, Chris I. ed. New Ideas and Directionsfor the 1990's: ACADIA 89
Proceedings. Gainesville, University of Florida: ACADIA, 1989.