A Hybrid Artificial Intelligence Approach for Automated Architectural Space Layout Planning Using Fuzzy Maps

Architectural spatial layout planning requires balancing multiple qualitative objectives—proximity, privacy, accessibility, and view quality—that are difficult to formalize using crisp rules. Although recent AI methods can automate floor-plan generation, many remain difficult to interpret and offer limited direct control over design rationale. This study presents a fuzzy map-based generative framework in which five spatial maps (proximity, adjacency, privacy, view, and accessibility) are computed for each candidate placement, merged through a fuzzy decision tree, and evaluated by Mamdani-type inference. A circulation-first strategy is used to establish a spatial backbone before room allocation, and multi-scenario generation is introduced to explore diverse corridor topologies and layout outcomes. Candidate placements are ranked through integrated fuzzy reasoning and constraint-aware scoring, with explicit handling of must-adjacency, corridor access, compactness, wet-zone clustering, and view zoning. Validation on a residential testbed demonstrates that the method can produce high-quality, interpretable layouts with strong connectivity and functional coherence, while failure-case analysis identifies topology-driven causes of low-performance scenarios. Quantitative results show stable performance across successful scenarios and reveal how corridor geometry directly affects feasible placement regions and final plan quality. The framework contributes a transparent, modifiable generative mechanism in which fuzzy maps operate not only as evaluation tools but also as the primary engine for sequential architectural layout synthesis.