In its parallel pursuit of an increased competitivity for design offices and more pleasurable and easier workflows for designers, artificial design intelligence is a technical, intellectual, and political challenge. While human-machine cooperation has become commonplace through Computer Aided Design (CAD) tools, a more improved collaboration and better support appear possible only through an endeavor into a kind of artificial design intelligence, which is more sensitive to the human perception of affairs. Considered as part of the broader Computational Design studies, the research program of this quest can be called Artificial / Autonomous / Automated Design (AD). The current available level of Artificial Intelligence (AI) for design is limited and a viable aim for current AD would be to develop design assistants that are capable of producing drafts for various design tasks. Thus, the overall aim of this thesis is the development of approaches, techniques, and tools towards artificial design assistants that offer a capability for generating drafts for sub-tasks within design processes. The main technology explored for this aim is Evolutionary Computation (EC), and the target design domain is architecture. The two connected research questions of the study concern, first, the investigation of the ways to develop an architectural design assistant, and secondly, the utilization of EC for the development of such assistants. While developing approaches, techniques, and computational tools for such an assistant, the study also carries out a broad theoretical investigation into the main problems, challenges, and requirements towards such assistants on a rather overall level. Therefore, the research is shaped as a parallel investigation of three main threads interwoven along several levels, moving from a more general level to specific applications. The three research threads comprise, first, theoretical discussions and speculations with regard to both existing literature and the proposals and applications of the thesis; secondly, proposals for descriptive and prescriptive models, mappings, summary illustrations, task structures, decomposition schemes, and integratory frameworks; and finally, experimental applications of these proposals. This tripartite progression allows an evaluation of each proposal both conceptually and practically; thereby, enabling a progressive improvement of the understanding regarding the research question, while producing concrete outputs on the way. Besides theoretical and interpretative examinations, the thesis investigates its subject through a set of practical and speculative proposals, which function as both research instruments and the outputs of the study. The first main output of the study is the “design_proxy” approach (d_p), which is an integrated approach for draft making design assistants. It is an outcome of both theoretical examinations and experimental applications, and proposes an integration of, (1) flexible and relaxed task definitions and representations (instead of strict formalisms), (2) intuitive interfaces that make use of usual design media, (3) evaluation of solution proposals through their similarity to given examples, and (4) a dynamic evolutionary approach for solution generation. The design_proxy approach may be useful for AD researchers that aim at developing practical design assistants, as has been examined and demonstrated with the two applications, i.e., design_proxy.graphics and design_proxy.layout. The second main output, the “Interleaved Evolutionary Algorithm” (IEA, or Interleaved EA) is a novel evolutionary algorithm proposed and used as the underlying generative mechanism of design_proxybased design assistants. The Interleaved EA is a dynamic, adaptive, and multi-objective EA, in which one of the objectives leads the evolution until its fitness progression stagnates; in the sense that the settings and fitness values of this objective is used for most evolutionary decisions. In this way, the Interleaved EA enables the use of different settings and operators for each of the objectives within an overall task, which would be the same for all objectives in a regular multi-objective EA. This property gives the algorithm a modular structure, which offers an improvable method for the utilization of domain-specific knowledge for each sub-task, i.e., objective. The Interleaved EA can be used by Evolutionary Computation (EC) researchers and by practitioners who employ EC for their tasks. As a third main output, the “Architectural Stem Cells Framework” is a conceptual framework for architectural design assistants. It proposes a dynamic and multi-layered method for combining a set of design assistants for larger tasks in architectural design. The first component of the framework is a layer-based, parallel task decomposition approach, which aims at obtaining a dynamic parallelization of sub-tasks within a more complicated problem. The second component of the framework is a conception for the development mechanisms for building drafts, i.e., Architectural Stem Cells (ASC). An ASC can be conceived as a semantically marked geometric structure, which contains the information that specifies the possibilities and constraints for how an abstract building may develop from an undetailed stage to a fully developed building draft. ASCs are required for re-integrating the separated task layers of an architectural problem through solution-based development. The ASC Framework brings together many of the ideas of this thesis for a practical research agenda and it is presented to the AD researchers in architecture. Finally, the “design_proxy.layout” (d_p.layout) is an architectural layout design assistant based on the design_proxy approach and the IEA. The system uses a relaxed problem definition (producing draft layouts) and a flexible layout representation that permits the overlapping of design units and boundaries. User interaction with the system is carried out through intuitive 2D graphics and the functional evaluations are performed by measuring the similarity of a proposal to existing layouts. Functioning in an integrated manner, these properties make the system a practicable and enjoying design assistant, which was demonstrated through two workshop cases. The d_p.layout is a versatile and robust layout design assistant that can be used by architects in their design processes.

Recent Artificial Intelligence studies have achieved substantial improvements in practical tasks by using extensive amounts of data. We assume that a substantial part of the data to guide artificial design technologies resides in existing design examples. Developing ways to use this data may enable improvements in intelligent design tools, with the hope that these may provide more effective design workflows and more productive design practices. Such improvements may result in more in-depth evaluations of potentials and alternatives for design situations; hence better planning for the spatial environment. Various approaches have been developed to use representations of architectural examples for artificially tackling architectural design tasks. This study presents a review of the historical development of these approaches, with an overall aim to investigate where and how design examples have been used for practical computational design applications. The review encompasses traditional and recent Shape Grammar and Procedural Modeling studies, Case-Based Design, Similarity-Based Evaluation and Design, and recent studies on the architectural uses of Machine Vision, Semantic Modeling, Machine Learning, and Classification. The emphasis of the review is on the studies that aim at designing or generating new design examples, particularly for building layouts, facades, envelopes, and massing. For a comparative evaluation of the current capabilities of the examined lineages of studies, we propose a minimum set of design capabilities, and assess each study through this framework. This reveals the overall patterns of already covered requirements. The review shows that initial hand-operated SGs gave way to automatic generation, which in turn developed into automated SG extraction, through increasing levels of computational capabilities. Case Based Design has been neglected; however, it can be reinvigorated through novel Al techniques. On the other hand, Similarity-Based Evaluation may complement and balance the orientation towards technical performance. Machine Learning and Computer Vision appear as potential intermediaries for connecting these threads. There are example-based studies towards almost all aspects of artificial design; yet, these have not been tackled adequately or definitively. In particular, dynamic process control is still only a future potential. On the other hand, the examined research lineages have the potential to assume complementary roles for more capable and multifaceted design systems. As an overall result, example-based research perspectives raise important possibilities for intelligent design systems. (C) 2017 Elsevier Ltd. All rights reserved.

This study aims to provide an overview on the intersection and interaction between architecture, urban modeling, planning fields and computer vision field. The reflection of the methods and approaches of fields such as visual recognition, natural language processing, data mining and data visualization onto architecture and urban studies are investigated and potentials of inter/transdisciplinary encounters are unfolded through research questions. This study can be considered as an initial attempt to constitute a common and sustainable platform for advancing efficient inter/transdisciplinary and cross-domain collaborations.

We propose a novel application of similarity-based evaluation for layout design. Basing evaluations on a pool of existing designs or quickly delivered graphical specifications enables an Evolutionary Algorithm to become an open-ended system, where problem definitions can be kept open for fast updating. Through its flexible and open-ended aspects, the approach is able to develop draft layouts for an unlimited variety of tasks, which are not implied within the initial task definition. This enables a versatile computational layout design assistant, which can be easily adopted within regular design processes through usual graphic media. In this paper, the functioning of the approach is described and verified; the versatility of the resulting assistant, its intuitive graphical interface, and its usage scenario are demonstrated.

In this paper we explore the land-use patterns of a historical mixed-use quarter of Istanbul by focusing on the reciprocal relationships of housing and commercial uses. We are concerned with the distribution of land-use patterns based on spatial adjacency and land-use patterns within the buildings. We measure reciprocal relationships of housing and commercial uses by using GIS and Data Mining in a complementary way and we test if we can re-generate these relationships by evolutionary computation to further support land-use allocation in inner city regenerations or new urban developments.