A learning automated 3D architecture synthesis model: demonstrating a computer governed design of minimal apartment units based on human perceptual and physical needs. (4th July 2019)

• Record Type: Journal Article
• Authors: Fisher-Gewirtzman, Dafna; Polak, Nir
• Is Part Of: Architectural science review. Volume 62:Number 4(2019)
• Page Start: 301
• Page End: 312
• Publication Date: 2019-07-04
• Subjects: Automated architecture design -- parametric design -- design optimization -- machine learning -- crowdsourcing; Architecture -- Periodicals; Building -- Periodicals; Engineering -- Periodicals; 720.5
• DOI: 10.1080/00038628.2019.1611537
• Languages: English
• ISSNs: 0003-8628

Abstract

ABSTRACT: This paper presents a learning model for the automated generation of built environments, demonstrated by the creation of minimal apartments situated in dense urban settings. The research utilizes the techniques of parametric modelling, multi-criteria optimization and supervised machine learning to provide 3D configurations of minimal apartments with improved visibility from significant viewpoints, and with a functional layout defined by 'the wisdom of the crowd'. The model seeks to maximize the measured 3D visibility in generated units – an attribute associated with low perceived density, recognized as having a positive effect on the well-being of dwellers. Simultaneously, the model engages a learning process, through which the classification of the suitability of each apartment is refined through 'the wisdom of the crowd', collected through an open web-simulation. This automated design model, regarding both perceptual and physical needs, demonstrates its potential for future use in the development of larger-scale densified urban environments.