Title: Conceptual stage predictive and collaborative design using machine learning
Type: Conference paper
Authors: Margaret Wang, Emil Poulsen, Dan Reynolds, Robert Otani
Publisher: International Association for Shell and Spatial Structures
Massive amounts of information gathered from multiple sources in the process of designing a building are a perfect fit for machine learning application. Traditional approaches to building design involve individual disciplines working separately and eventually piecing the parts together. Current major roadblocks are the slow process, entrenched patterns of information processing, and lack of resources to generate better options. The opportunity to improve the modus operandi lies in faster design prototyping using machine learning models to bypass resource intensive design, streamlining collaboration and making use of information earlier in a project. This paper presents an application, Asterisk, which directly addresses the conceptual level phase of design, showcasing features that target each of these roadblocks to aid an overhaul of existing paradigms.
Keywords: machine learning, embodied carbon, design exploration, conceptual design, design space, automation, prediction