ALIGNING CURATORIAL INTENT WITH PERCEPTION-BASED NAVIGATION: SOM AND ISOVIST APPROACHES IN EXHIBITION DESIGN

Abstract

Designing intuitive indoor exhibition spaces requires aligning the intended narrative sequence with visitors' spontaneous movement. Evaluating this alignment before construction remains a challenge in architectural planning. This research presents a computational framework to assess spatial layouts by comparing planned exhibition routes with simulated visitor behaviour. The study employs two complementary computational approaches within a temporary indoor exhibition context: the Self-Organising Map (SOM) and Isovist-based agent simulation. The SOM algorithm generates an optimised sequence of exhibit nodes, representing the intended visitor route, while the Isovist simulation predicts movement based on spatial geometry, visibility, and obstacle avoidance. An image-based, pixel-level comparison using OpenCV quantifies the overlap between SOM-generated paths and Isovist trajectories, revealing areas of alignment and divergence. Results show partial overlap, indicating that while SOM captures structured circulation, natural visitor movement is more adaptive and variable. The methodology highlights spatial inefficiencies and navigational bottlenecks, providing architects and planners with a diagnostic tool to optimise exhibition layouts, enhance circulation, and ensure key points of interest are naturally discoverable.