Optimizing for Orientation in Complex Spaces

Abstract

AbstractIn response to the increasing demand for collaboration and knowledge exchange within Postfordist network society, both virtual and physical spaces are becoming more and more complex. Therefore the orientation within these increasingly complex and information-rich scenes becomes a problem that architectural design must address. The goal of this research is to upgrade architectural design competency in this respect by setting up a workflow for evaluating and optimizing the legibility of complex scenes. This paper introduces a novel research approach focused on the recognizability of salient interaction offerings within complex spatial settings, by using machine learning. A systematic workflow is being developed for simulations that appraise and rank design proposals with respect to the trade-off between scene complexity and legibility. The authors explore the research through a series of simulation experiments concerned with semantic segmentation, i.e. with distinguishing and classifying relevant features in a large complex visual field. The paper first describes the method of setting up the measurement of complexity and ease of recognition, and then illustrates how a trained neural network can be used to evaluate and rank a series of design proposals (with systematically varied degree of complexity) on the basis of their recognizability. While the paper found that the hypothesis of a statistical inverse correlation or trade-off between complexity and recognizability holds, for each degree of complexity there are several design options with different degrees of recognizability. Therefore this approach allows to optimize the design of complex scenes in terms of the crucial criterion of legibility.