An exploration of computational approaches to support systematic innovation in architectural design

© 2018 Dr. Camilo CruzDesign can be broadly defined as an activity aimed at transforming an existing condition into a desired one (Simon, 1969), via the generation of things—artefacts—that display new physical order (Alexander, 1964). Design is intentional and purposeful, as it starts from the identification of a need or from a desire to change what exists. The task of designers, then, is to formalise these needs and desires into new artefacts, which when deployed on the existing condition, improve its characteristics. Every design endeavour addresses practical issues, where the goal is to ‘make things work’. However, in some cases — architectural design, for instance — designers have to deal with issues that exceed the satisfaction of practical requirements, as the results of their actions will also have impact in their cultural, social, and physical domains. This tension between the artistic ‘spontaneity’ of the design practice, and the practical aspects addressed by it, is arguably what separates design from plain problem-solving. which has been the centre of an ongoing discussion in regards to design methodology, and has defined a spectrum of approaches within the design community, that ranges between rationalism and artistry. One of the consequences that the attempts to develop systematic methods for designing, was the adoption of computers, by researchers and, later on, practitioners. And with computers, came expectations of a radical transformation. A new way of designing was supposed to emerge. But these expectations have been only partially met. Positive impacts from the incorporation of digital tools into the designer’s workflows are undeniable, as computers have allowed for more efficient productivity chains, and improved performance of the artefacts being designed. Another clear example — more related to the creative aspect of design—is the way in which newly acquired capabilities in digital representation, have enabled designers to work with complex geometries that were unimaginable 50 years ago. Nevertheless, the development of computation methods for design seems to have gone down a pathway driven by efficiency and convenience, rather than one focused on the definition of methodological connections between design and computing. Design exploration still relies almost exclusively on the capabilities of designers to frame the problem, evolve that framing and establishing connections between the requirements defined by it and the attributes of the solution candidates they can generate. And all these activities are still done by ‘sketching’ (Parthenios, 2005) — understood as the direct translation of ideas to form, based on implicit decision-making — to which digital tools are mere instrumental support. In contrast to what has happened in computational design, where the real impact of technological development has not met the expectations set by its founders, groundbreaking steps towards capitalising on the advantages that computation methods offer for the exploration of novel solutions to particular problems, have been taken in other areas of research. One example of these is artificial intelligence. This thesis explores ways in which the use of computation methods can contribute to the development of innovative design solutions in architecture. The research adopts an exploratory strategy, which requires two main tasks to be addressed. The first one is to systematically formulate what designing is, in order to define a road map for the exploration of the possible roles that computation methods can play in design. The second one is the actual exploration of the capabilities of computation methods to contribute to the generation of novelty in the design process. These tasks are addressed via the development of a conceptual model of design, that builds on well established theoretical approaches, coupled with the implementation of minimalistic generative computation methods inspired on evolutionary computing. The aim of the study is to advance the understanding of the roles that computation methods can play in the architectural design process, in order to enable designers to reach areas of the space of design alternatives that have not been explored before. Through the synthesis of a model of design, which supports the development of methods that enable systematic design exploration, this thesis contributes to advancing the field of digital design. The articulation of a conceptual framework that incorporates principles of optimisation, yet encourages spontaneity and authorship, can be extended and used as a roadmap for the development of new methods. Additionally, the proof-of-concept computation methods developed for this study, serve as an example of the avenues that this roadmap opens for exploration in digital design