Twenty-First International Conference on Interdisciplinary Social Science

Abstract

The configuration of architectural space plays a central role in shaping spatial experience. Graph-based representations offer a structured method for modelling and analysing floor plans by abstracting spatial relationships into nodes and edges. These representations enable the evaluation of spatial qualities such as accessibility, visibility, connectivity, and circulation, which influence how users perceive and navigate the built environment. This study presents a systematic review of peer-reviewed literature published between 2010 and 2025 that employs graph-based methods to represent, analyse, evaluate, or generate architectural floor plans. The reviewed studies are categorized by graph types (e.g., axial, visibility, topological, semantic), generation techniques (e.g., BIM-derived, image-based), computational methods (e.g., graph neural networks, evolutionary algorithms), and application domains (e.g., healthcare, education, housing). Eight analytical dimensions are used to extract and compare methodological approaches and implementation strategies. The review identifies current trends, recurring methods, and notable gaps in the literature, emphasizing the growing role of graph-based models in computational design workflows. These models are demonstrated to support both analytical evaluation and generative design, providing a foundation for the further integration of graph-based thinking into architectural research and practice. Ultimately, the findings highlight the potential of graph-based approaches to enhance the design and understanding of spatial experiences.