Generative Design

GD is frequently used in architectural design to explore a variety of design options. There are various generative engines which have been reviewed in other places [Muehlbauer et al. 2017b], [Singh and Gu 2012] and examined in their historical context in Kitchley [2006, p. 24-51]. The use of grammar evolution for architectural design is one example of a GD technique that allows for the integration of performance criteria in the computational process [Muehlbauer et al. 2017a]. For the sake of this critical literature review, I want to focus on various conceptual considerations about GD before discussing the application of GD in decision support for architectural design.

The concept of GD can be tracked back to Aristotle, who put forward a metaphor in which nature is said to often combine sets of parts that together form a coherent whole (Mitchell[1977, p. 29]). 5 Building on ideas found in the architectural discourse of

5

In his book titled ‘The parts of Animals’, Aristotle describes the combination of “homogeneous” and “heterogeneous parts” in material systems to achieve different functions and characteristics necessary to the movement of the body (Aristotle[-350, s. 10]).

Leonardo da Vinci6 and the Bauhaus movement7, the term ‘generative systems’8 started to be used in artistic discourse in 1970, as argued by Sheridan [Sheridan 1983]. An earlier connection between the term ‘generative’ and the meaning discussed in this section is present in the work of Eduardo Mac Entyre as part of a group called ‘Arte Generativo’. The functional aspects of “generative design system” links back to a description of Dell K. Allen and Ronald P. Millett and [Beeby and Thompson 1979]. Jon McCormack lists “self-organisation” 9, “evolutionary systems” and “generative grammar” as GD methods [McCormack et al. 2004], all of which combine design elements to artefacts in emerging processes. Each of the generative algorithms can be used as a generative engine in the sense [Frazer 2016] mentioned - as a geometric system for building morphogenetic capabilities. These capabilities can be integrated in CAD systems to facilitate design space exploration, if combined with designer interaction.

The architectural design process is still frequently discussed throughout the field of architecture. A manifold of interpretations of the design process can be expressed because of the complexity of design in architecture [Bayazit 2004]. While Alexander, Ishikawa and Silverstein [Alexander et al. 1977] argue that design solutions can be composed of hierarchically organised patterns, Patrick Schumacher [Schumacher 2011-2012] develops a set of rules to simultaneously describe “Parametricism” as an architectural style and as a set of design methods. I argue that design intent in the context of computational archi- tecture needs to be represented in a way that suits a particular computational process, as described, for example, by John Frazer in his book ‘An Evolutionary Architecture’ [Frazer

6_{Leonardo da Vinci and Durand used a rule-based approach to generate architectural plans [Ediz and} Cagdas 2004].

7

The central idea that contributed to this context from my perspective is the use of technology to develop new stylistic expressions and a different view on creativity that was based on a holistic under- standing of the practice of art and crafts. Understanding compositional rules as an important tool to generate harmonic expression in abstract geometric compositions was highly influenced by the De Stijl movement around Theo van Doesburg.

8_{The use of the term “generative system” started in most likely 1817, pointing toward the reproduction} system of animals [Robertson and Baillie 1817].

9_{A wide range of algorithms can be unfolded from this term that assembles parts based on local} interactions to create larger entities [McCormack et al. 2004]. In contrast to [Fischer and Herr 2001], I consider the term “emergent system” as a term encompassing generative systems and other processes with emergent properties. These emergent properties are characteristics of a system that appear in unforeseen ways during the interaction of its parts, or as Pia Ednie-Brown defines: “the process and outcome of combining things to form a whole”, therefore rendering the term an “issue of composition”Ednie-Brown [2007, p. 15]. However, we return to [Fischer and Herr 2001] to understand the range of algorithmic models that are inscribed in the term of self-organisation: e.g. growth models, cellular automata, and swarm modelling. Again, I want to challenge the listing found in [Fischer and Herr 2001] and exclude parametric design from GD, as a means of differentiation of the terms in context of this PhD study. Parametric models per se have a hierarchical, static definition that follows a left-to-right logic. Now, based on recent implementation of loops and optimisation methods, GD can also be integrated in parametric design software, therefore blurring the boundaries of traditional terminology. In the further unfolding of this thesis, I refer to parametric design in the terms just described, while I use the term ‘visual programming editor’, if I refer to parametric design environments. Again, I need to raise awareness that this term is to be seen as different from “visual calculation” or “visual computation”, which describe the use of shape grammar for design composition [Stiny and Gun 2012]

1995] to allow the designer to exploit the capabilities of GD processes in a performative way, usually referred to as digital morphogenesis, as described by Stanislav Roudavski [Roudavski 2009]. As a basis for discussion of Typogenetic Design as a kind of aesthetic decision support, I shall briefly review some of the theories and methods frequently used in the context of computational design.