SEEP: A Cognitive-Driven Framework for Encoding Spatial Experience in Generative Design
Description
In the rapidly evolving field of generative design, artificial intelligence has demonstrated remarkable capabilities in optimizing geometric forms and environmental performance. However, there remains a systematic gap in encoding the structured dimensions of human spatial cognition.
To address this deficit, we propose SEEP (Spatial Experience Encoding Parameters), a framework that quantifies abstract spatial experiences into computable six-dimensional vector sequences: Openness (O), Sequential Rhythm (R), Landmark Visibility (LV), Visual Field Continuity (VFC), Visual Angle Change (VAC), and Visual Depth Penetration (VDP).
Through a cross-cultural design experiment—translating the spatial rhythm of a traditional Chinese garden (Lion Grove Garden in Suzhou) into a modern campus layout in Milan—we validated the framework's effectiveness. The experimental results demonstrate that SEEP-driven generative solutions achieved high-fidelity reconstruction of the target spatial sequence (Openness MAD=0.026, Rhythm MAD=0.037) while maintaining full compliance with complex functional requirements (3,264 m² mixed-use program, 100% satisfied).
This proves that experience optimization and functional efficiency are not a zero-sum game; AI-driven design can precisely reconstruct the cognitive rhythm of cultural spaces while meeting physical constraints. This research provides a methodological foundation for human-computer interaction in architectural design, advancing the paradigm shift from "geometry-oriented" to "cognition-driven" generative design.