Extracting Actionable Information from the Site Context Using A Phenotype-Based Strategy
BanLiang Ling, Bige Tunçer · Proceedings of the International Conference on Computer-Aided Architectural Design Research in Asia · 2024
As more computational design tools are developed, solution generation has been accelerated to provide real time feedback. However, a human designer is still required to translate generated data into actionable information. This is especially so for diverse design scenarios, where the data structure differs, and the computer is unable to draw conclusions across both scenarios. The site context is one key parameter that contributes towards the difference in scenarios. In short, how can an algorithm extract design-related information from diverse scenarios? To address this issue, a phenotype-based strategy is proposed as a representation method, and it re-parameterises diverse site conditions by focusing on their geometrical properties. Instead of parameterising the site context, street-view images are captured, and Gabor filters extract relevant geometrical properties, such that site conditions with different compositions, forms, and density can be organised. This method quantifies compositional and density-based properties of the surrounding building blocks, thereby enabling the computer to digest generated information and provide design suggestions. A new sample site is then used to demonstrate a query of the phenotype space, where suggestions about solar radiation performance is feedback to a human designer.