Optimal Transport-based Affective User Requirements Realignment for Automated Spatial Design Revision

초록

Modern residential occupants require their affective intents to be reflected in spatial design to achieve satisfaction. However, current generative architectural design studies focus on morphological and functional requirements, overlooking affective intent. This study proposes an affective user requirements realignment framework for automated, minimally disruptive spatial design revision based on Optimal Transport. The proposed BIM-integrated framework (i) represents spatial experience as a multidimensional affective distribution derived from 1,638 annotated responses across 54 parametric configurations, (ii) predicts the distribution from discrete design parameters using a lightweight Transformer regressor, and (iii) selects a minimally disruptive revision via entropic-regularized Wasserstein distance to a user-specified target affect across 6,480 candidate configurations. The selected parameter set is automatically applied through a Revit C# add-in to support closed-loop design review. Results demonstrated consistent convergence toward target affects (reduction: 0.036-0.290) with positive directional alignment across all ten affective dimensions (mean cosine similarity = 0.502), demonstrating the feasibility of affect-preserving revision in BIM workflows as a proof-of-concept.

키워드