Improving the Scalability of Generative Representations

Applied Intelligence, 2002

Design tools that aim not only to analyse and evaluate, but also to generate and explore alternative design proposals are now under development. An evolutionary paradigm is presented as a basis for creating such tools. First, the evolutionary paradigm is shown to be the only successful design system on which this new phase of design tool could be based. Secondly, any characterisation of design as a search problem is argued to be a serious misconception. Instead it is proposed that evolutionary design systems should be seen as generative processes that are able to evaluate their own output. Thirdly, a generic framework for generative evolutionary design systems is presented. Fourth, the generative process is introduced as a key element within this generic framework. The role of the environment within this process is fundamental. Finally, the direction of future research within the evolutionary design paradigm is discussed with possible short and long term goals being presented.

2020

The word "design" has been used to describe the theory of evolution by Creationists. Scientists object, however, and point to the unguided nature of evolution as evidence that no design has taken place. In this paper, I discuss the meaning of the word design, especially in the context of more recent developments in computer software, specifically procedural generation, and how this may change our understanding of not only the word design, but the very nature of evolutionary change.

Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 2015

2001

This dissertation dwells in the interstitial spaces between the fields of architecture, environmental design and computation. It introduces a Generative Design System that draws on evolutionary concepts to incorporate adaptation paradigms into the architectural design process. The initial aim of the project focused on helping architects improving the environmental performance of buildings, but the final conclusions of the thesis transcend this realm to question the process of incorporating computational generative systems in the broader context of architectural design. The Generative System [GS] uses a Genetic Algorithm as the search and optimization engine. The evaluation of solutions in terms of environmental performance is done using DOE2.1E. The GS is first tested within a restricted domain, where the optimal solution is previously known, to allow for the evaluation of the system's performance in locating high quality solutions. Results are very satisfactory and provide confidence to extend the GS to complex building layouts. Comparative studies using other heuristic search procedures like Simulated Annealing are also performed. The GS is then applied to an existing building by Alvaro Siza, to study the system's behavior in a complex architectural domain, and to assess its capability for encoding language constraints, so that solutions generated may be within certain design intentions. An extension to multicriteria problems is presented, using a Pareto-based method. The GS successfully finds well-defined Pareto fronts providing information on best trade-offs between conflicting objectives. The method is open-ended, as it leaves the final decision-making to the architect. Examples include finding best trade-offs between costs of construction materials, annual energy consumption in buildings, and greenhouse gas emissions embedded in materials. I would like to thank the several institutions that contributed to funding this research: The Fundagao para a Ciencia e a Tecnologia, Praxis XXI program, Portugal; the Fundagao Cultural Luso-Americana, Fulbright program; the V. Kahn-Rassmussen Foundation; and MIT, through a Rosenblith Fellowship. This work is first dedicated to my parents. To my father, who has always guided me through life with his clarity of view, his sharp mind and his uncompromising honesty. To my mother, who has that kind of wisdom that one only gains if one passes through life with sensitivity and awareness. Then to my sisters and brother, Ana, Gabe and Miguel, for always being there for me. And to those late night talks with Miguel, drinking whiskey and talking about all kinds of things, some of them so important for this thesis, until four or five in the morning. I would like to thank Les Norford, my advisor, for always believing in my ideas and my work, making me come to MIT, and for all the valuable advice he gave me throughout this academic endeavor; Bill Mitchell, for bringing a precious contribution to this research drawing from his extensive knowledge of the field and theoretical background; Julie Dorsey, for looking at my work with her inquisitive mind and forcing me to face the higher-level questions; Peter Testa, for his creative advice and conversations about the use of generative systems in architecture; Alvaro Siza, for accepting to participate in this research and providing all the necessary information; Jorge Bastos and the late Frederico George, for pushing me towards doing this Ph.D.

2008

Abstract. We follow up on the work of Ebner[1] in studying representations for evolutionary design of objects. We adopt both the method and the simulation framework, and perform more thorough experiments. We design new representations, both direct and indirect, and compare their performance to the original work. We design and make use of a specialised system for distributed computing that integrates smoothly with the EO library[5]. First, we confirm the results of Ebner with VRML scene graphs representation. Next, we demonstrate how both of the new representations based on triangular mesh perform significantly better. Finally, we study and improve the performance of the distributed system that we utilised to run our experiments on tens of computing nodes. 1

Since the time of Alan Turing's death, in the early nineteen-fifties, largely due to beliefs spawned by Turing's formative work into machine intelligence, dreams that the design process can be automated by an artificial system, have been theoretical catnip in the field of Architecture. The first wave of interest in this subject can be identified to have peaked around 1972 with Nicholas Negroponte's experiments for an 'Architecture Machine' at MIT. In developmental biology, the incentives that drive natural growth systems are referred to as morphogens. It is acknowledged in this field that without morphogens, systems would not be able to grow. It is also possible to observe morphogens in the human creative design process.Since the initial excitement into the potential of artificial design systems in the late sixties, it is fair to say that relatively small progress has been made: the architect's position as designer, has not received any serious artificial challenger. It is therefore interesting to note the general absence of the concept of Morphogens in historical and contemporary proposals for artificial design systems. However, in 2002, Manuel DeLanda published an article on the requirements of an artificial process which will be used in this essay to identify the morphogens of an artificial design process.The main body of this essay is formed into three sections documenting the proposed morphogens. The first of these is defined as a requirement to grow, which can be seen in nature in phototropism. The second, is to have large populations of problems and solvers to encourage,competitiveness; this will also be used to explore the idea of community design with reference to the open source community. The third incentive will be used to organise both the proposed Design Engine, and the resultant design, calibrating the cybernetic relationship between the human designer(s) and the artificial tools.This essay seeks to extrapolate morphogens from a wide variety of sources develop a proposal for a Universal Design Engine, to be tested in the subsequent thesis design project. The Engine will be formed from a small community of unskilled human designers, or an individual,with morphogenetically motivated artificial tools

2020

Due to recent developments in the field of additive manufacturing enormous advantages have become in product design and manufacturing process. Before the appearance of additive manufacturing, developing very complex or light weight structures was difficult to manufacture. The development of artificial intelligent technology helps to develop new collaborative tools and algorithms. Generative design approach is one of them. The outcome model from a generative design study is not depending only from designer/engineer experience or his knowledge. Designers can react with sophisticated algorithms through CAD programs to specify the shape and the topology of the model. A significant tool on a generative design system is topology optimization which is able to generate different solutions. The changes in design process are significant. A rough conceptual design (sketch) or a 3d model is first prepared. Then, boundary conditions, safety factor, manufacturing limitations and materials propert...

With computational design strategies, particularly evolutionary generative systems, the understanding of the design activity has been changing. Accordingly, the designerly behaviour and the designer's interaction with and the involvement in the design process have been evolving. Evolutionary design systems generate a large number of design solutions that expand design spaces immensely. Especially in multi-modal fitness landscapes, the existence of many optimal design alternatives complicates the designer's cognitive involvement in design. Moreover, particularly in automated design generation, due to the immense expansion in design search space, the designer's visual interaction with the design artifacts is diminished. Accordingly, the proposed research problematizes two major issues of evolutionary design systems; (1) the broad, dense and non-structured design search spaces and (2) the decrease in the designer's involvement in decision-making and evaluation due to the automated design generation. As a solution, perception-based design space structures are proposed as an evaluative structuring strategy. These network structures are intuitive, case-based, observer-dependent and subjective maps of the designer's cognitive world. By encouraging the integration of designer's cognitive abilities in the design synthesis, these structures aim to bridge the gap between the design artefact and the designer by acting as a mediator during the generation process. Alongside, these structures aim to manage the complexity in design search space by providing an environment for designerly evaluation and decision-making. With those aspects, perception-based design space structures are a designer-centric (human-centric) approach for automated design processes and are based on the designer's perception and identification of common visual features. The identification of the common features of instances, visual resemblance is chosen as a criterion for forming perception-based design structures within the scope of this research, on the basis of Rudolph Arnheim's Visual Thinking that highlights forming categories as one of the major ability of a human cognition. Within this scope, a case study is conducted within a group of designers with an existing design space of chairs that are generated by IDEA, an evolutionary system, experimented by Celestino Soddu. In this framework, this paper discusses and presents the experimental study about evolutionary generative design search spaces and perception-based design space structures. Abstract With computational design strategies, particularly evolutionary generative systems, the understanding of the design activity has been changing. Accordingly, the designerly behaviour and the designer's interaction with and the involvement in the design process have been evolving. Evolutionary design systems generate a large number of design solutions that immensely expand design spaces. Especially in multi-modal fitness landscapes, the generation of many optimal design alternatives complicates the designer's cognitive involvement in design. Moreover, particularly in automated design generation, due to the immense expansion in the design search space, the designer's visual interaction with the design artefacts is diminished. The proposed research problematizes two major issues of evolutionary design systems; (1) the broad, dense and non-structured design search spaces and (2) the decreased level of designer involvement in decision-making and evaluation due to the automated design generation. As a solution, perception-based design space structures are proposed as an evaluative structuring strategy. Within this scope, a case study is conducted within a group of designers with an existing design space of chairs that are generated by IDEA, an evolutionary system, experimented by Celestino Soddu. In this framework, this paper discusses and presents the experimental study about evolutionary generative design search spaces and perception-based design space structures.

Design Studies, 2012

Design creativity techniques encourage divergent thinking. But how well do the existing generative design techniques support this requirement? How can these general techniques be augmented for supporting design exploration and creativity? This paper investigates these questions through a review of five different generative design techniques used in architectural design that includes cellular automata, genetic algorithms, L-systems, shape grammars, and swarm intelligence. Based on the literature on design cognition and the recent theoretical works on digital design thinking, this paper proposes the need for an integrated generative design framework to enhance design exploration support for human designers. Potential challenges and strategies towards developing such an integrated framework are discussed.