Harnessing Design Space: A Similarity-Based Exploration Method for Generative Design

Abstract

Working with multiple alternatives is a central activity in design; therefore, we expect computational systems to support such work. There is a need to find out the tool features supporting this central activity so that we can build new systems. To explore such features, we propose a method that aims to enable interaction with a large number of design alternatives by similarity-based exploration. Using existing data analysis and visualization techniques adopting similarity-based search, we formalized the method and its elements by focusing on systematic filtering, clustering, and choosing alternatives. We present a scenario on developing conceptual designs for a residential apartment to illustrate how the method can be applied, as well as to reveal the limitation of current tools and the potential interactive clustering and filtering features for the new systems coupled with parametric design.

Navigating Design Spaces: Finding Designs, Design Collections, and Design Subspaces

“…These designs can be obtained by different processes, namely, by randomly generating values within a narrow range close to a design's parameter values, as in Autodesk Refinery, 7 or by estimating the similarity degree (given by an appropriate metric) between pairs of design alternativesand selecting designs that have a higher degree, as described in the similarity-based exploration method. 4 While this is a valuable process for generating alternatives, consideration should be given to not overwhelming users with very similar variations. 1 Hybrid designs share features of two or more designs.…”

“…3 Nevertheless, there are several exploration strategies, such as: parametric exploration (manual manipulation of parameter values), similarity-based exploration (semi-automated exploration of parametrically close alternatives) and optimization-based exploration (automatic selection of alternatives according to certain goals). 4 These approaches possess different degrees of automation, ranging from more manual to more automatic, which affect users' intervention in the search process. They also have different degrees of computational cost, which influence the speed in the search for solutions: for example, optimization-based methods are more expensive (as they involve cycles of generation-analysis-regeneration), while parametric methods are less expensive (as they only involve cycles of generation).…”

“…Unfortunately, AD tools present limitations regarding solution exploration and visualization, as they mostly address black-box exploration methods and often present a single model at a time, instead of multiple models. 4 Other tools are more oriented towards navigating GDSs, such as design space explorers, although these are more focused on the visualization and interaction aspects of the exploration process. 2,5 White-box methods can be dispersedly found in specific GDSs.…”

Quantifying diversity in parametric design: a comparison of possible metrics

“…Furthermore, when considering the human-centric context of architectural or structural design, there has been no rigorous investigation as to whether or not the given metrics accurately translate to a human perception of design diversity, and how the use of different diversity metrics might influence the conceptual design process. When design computation applications require diversity measurement, many researchers borrow an existing methodology, propose their own that is problem specific (Yousif et al, 2017), or suggest that designers themselves should choose one interactively (Erhan et al, 2015), since the metric itself was not the focus of their research.…”

Integrating expertise and parametric analysis for a data-driven decision-making practice

“…New computational frameworks and methods integrate generative design, parametric modeling, performance-based analysis, and data analytics into integrated workflows supporting optimization and decisionmaking [11][12][13] and allow us to revisit these questions. In a short period, parametric analysis evolved from a plausible framework 14,15 to actual implementations.…”