"Transit Oriented Development Assistive Interface (TODAI) - A Machine Learning Powered Computational Urban Design Tool for TOD"

Abstract

Transit-oriented Development(TOD) is widely regarded as a sustainable development paradigm for its sensible space planning and promotion of public transit access. Research in providing decision support tools of TOD may contribute to the Sustainable Development Goals, especially towards sustainable cities and communities (SDG goal 11).While the existing Geographic Information System(GIS) approach may well inform TOD planning, computational design, simulation, and visualisation techniques can further enhance this process. The research aims to provide a data-driven, computational-aided planning support system (PSS) to enhance the TOD decision-making process. The research adopts an action research methodology, which iteratively designs experiments and inquires through situating the research question in real-world practice. A work-in-progress prototype is provided - Transit-Oriented Development Assistive Interface (TODAI), along with an experiment in a newly proposed metro station in Sydney, Australia. TODAI provides real-time visualisation of urban forms and analytical data indicators reflecting key considerations relevant to TOD performance. A regressive machine learning model (XGBoost) is used to make predictions of analytical indicators, promptly producing outcomes that may otherwise require a costly computational operation.