AI-ENHANCED PERFORMATIVE BUILDING DESIGN OPTIMIZATION AND EXPLORATION: A design framework combining computational design optimization and generative A

When using computational optimization for early-stage architectural design, most optimization applications often produce abstract design geometries with minimal details and information in relation to architectural design, such as design languages and styles. Meanwhile, Generative AI (GAI), including Natural Language Processing (NLP) and Computer Vision (CV), hold great potential to assist designers in efficiently exploring architectural design references, but the generated images are often blamed for having limited relevance to the context and building performance. To address the limitation in computational optimization and leverage the capability of GAI in design exploration, this study proposes a design framework that incorporates Performative/Performance-based Design Optimization (PDO) and GAI programs for early-stage architectural design. A case study is demonstrated by designing a high-rise mixed-use residential tower in Hong Kong. The result shows that the PDO-GAI approach can help designers efficiently proceed with both diverging exploration and converging development.