Blurring the boundaries between real and artificial in architecture and urban design through the use of artificial intelligence

Abstract

This doctoral thesis explores the use of three-dimensional (3D) technologies for architectural representation and modelling of both ‘real world’ and ‘artificially generated’ 3D objects. Current 3D architectural representation has reached high levels of reality, to the extent that it’s often hard to distinguish between ‘pictures from the real world’ and ‘artificially generated’ 3D renderings. The thesis will make use of the latest available 3D technologies in combination with artificial intelligence (AI) processes to increase the visual realism and geometrical precision of 3D models. The first line of research relates to architectural representation and visualization, by exploring the use of ‘Light Imaging Detection And Ranging’ (LIDAR) technology and proposing a point-based rendering (PBR) methodology, to seamless merge models obtained directly from the ‘real world’ with ‘artificially generated’ ones. The second line of research is related to geometrical architectural modelling, and proposes the use of evolutionary computation and self-organization logic to achieve more geometrical realism and accuracy in the 3D modelling process, by exploring the idea of auto-form generation. The research consists of three case studies, and algorithms are proposed for each one. The first one is related to 3D visualization through LIDAR scans and PBR rendering, the second to geometrical generation through evolutionary morphogenesis and the third one to human made self-organized systems (cities). From the results obtained from each case study final conclusions will be draw. The final objective is to determine the efficiency of using point-based technologies and artificial intelligence as a methodology to further blur the boundaries between ‘real world’ 3D models and ‘artificially generated’ ones.