Kotsopoulos S, Ph.D. Dissertation, Massachusetts Institute of Technology, 2005
3. Design Concept and Method
The design approach is influenced by three factors: a) the absence of a predetermined site, b) the specifications
of the building program, and c) the provision for low construction-cost. The proposed design approach follows the
general principles of the polyomino house: Starting from an initial number of rooms,, the objective is to develop
compositional rules able to generate one or two storey houses of variable size and morphology. The
systematization of the ground plan is the method employed for the attainment of this objective. In order to achieve
certain preferable room adjacencies spatial relationships and rules are set out. The definition of the “parts”
(spatial vocabulary of rooms) is followed by the construction of “wholes” (designs). The overall process can be
characterized prescriptive and introspective.Prescriptive because each designer determines prescriptive
sequences of rules that guide the production of house arrangements. Introspective because each potential
designer could possibly form rules on the basis of personal preference.
The computational framework defined in Stiny & Gips 1972, and extended in Stiny 1976; 1980; 1991 is employed.
The members of a spatial vocabulary belonging to some algebra Uij are composed with the aid of production
rules. Similar efforts referring to the construction of rule systems for 0-dimentional languages exist in Carnap 1937
and in Chomsky 1957. In design, Mitchell (1974) reduced the plans of minimum cost dwellings into minimal
representations (partis). The parti becomes the basis on which several dwelling variations are generated. Stiny
and Mitchell (1978) capture the production of Palladian villa plans in a computational system of eight productive
stages. Other published examples include rule-based systems for the generation of Frank Lloyd Wright’s prairie
houses (Koning and Eizenberg 1981), Japanese tea-room designs (Knight 1981), Queen Ann houses (Flemming
1987), Taiwanese houses (Chiou and Krishnamurti 1995), Yingzao fashi houses (Li 2000), and Alvaro Siza’s
houses (Duarte 2001).
The novelty of the proposed approach is that it captures the exploratory effort of a design process from scratch
without dealing with analysis of existing designs. It involves testing and selection of rules, as the designer
explores their possible outcomes. First, the spatial entities and rules are described and then, are tested. The
heuristics of the process are organized in three levels. Each level contains rules that achieve some objective.
At the top level of formation, the rules produce parti diagrams. At the middle level of transformation, a chosen parti
is transformed to a boundary-layout. At the third level of refinement, the rules apply on the boundary-layout to
determine its tectonic details (doors, windows etc). The overall process is not linear, as the output of any level may
influence both the preceding and the subsequent levels. The proposed framework can be sketched out as follows:
R: { | Formation | | Transformation | | Refinement |
| | | | | |
| A1 ---> F1 | <==> | G1 ---> M1 | <==> | N1 ---> W1 |
| : | | : | | : |
| An ---> Fn | | Gn ---> Mn | | Nn ---> Wn |
}
where A1,…, An,are elements in S.
articulation of candidate rules, while the digital the rule-testing. The digital interpreter is used only in the first two
levels of formation and transformation. Through an iterative process of formation, transformation and refinement,
the rules are evaluated and redefined according to their compliance to programmatic, intuitive, or other criteria.
Retrospectively the rules can be available in constructing production systems, or grammars.