URBAN ANALYTICS AND GENERATIVE DEEP LEARNING FOR CONTEXT RESPONSIVE DESIGN IN DIGITAL TWINS A Singapore Study

Abstract

City Digital Twins (CDT) can play a pivotal role in consolidating and visualising complex urban big data, promoting rapid and informed decision making in contemporary cities. Beyond rich contextual data, these tools offer features like interactivity, 3D models and data visualisation, making them ideal for urban planning and design explorations. However current CDT implementations primarily focus on data visualisation and lack any robust design support. Simultaneously, generative urban prototyping occurs in specialised tools, detached from this rich contextual data. This study, on the Virtual Singapore (VSg) CDT platform, explores how the platform’s existing data, interactivity and 3D visualisation capabilities can facilitate rapid proto typing through generative machine learning techniques trained on the city’s unique planning texture; and discusses the challenges and limitations of the platform in supporting the development of such tools, along with potential improvements. The study aims to advance CDTs beyond static data consumption and visualisation towards generative tools for urban planning and decision-making processes.