Performance-Based Generative Architectural Design Integrating Environmental Constraints Using GANs
With the rapid advancement of deep learning technologies, generative artificial intelligence has become an influential tool in architectural design research and practice. While existing generative design studies largely emphasize spatial organization and functional relationships, environmental performance considerations are often treated as post-design evaluations rather than integral design drivers. This research addresses this limitation by proposing a performance-based generative framework that embeds environmental constraints directly into the architectural generation process. Limitations of Conventional Generative Design Approaches Most GAN-based architectural design models focus on visual similarity, spatial plausibility, or functional compliance, neglecting critical environmental performance metrics. As a result, generated design outcomes frequently require extensive post-processing or simulation-based refinement. This section discusses how the absence of performance f...