Generative artificial intelligence (AI) in built environment design and planning – A state-of-the-art review
Haolan Zhang, Ruichuan Zhang
Abstract
Despite numerous studies on adopting, implementing, and developing generative design approaches within the architectural, engineering, and construction (AEC) sectors, there remains a limited understanding of the capabilities and constraints of generative artificial intelligence (AI) in specific applications for built environment design and planning. This review paper aims to bridge this gap by providing a systematic review guided by a framework encompassing three main related application areas in building development – site layout, interior, and exterior design, and three main categories of generative AI algorithms – rule-based AI and expert systems, optimization and metaheuristics, and machine learning algorithms, with a focus on state-of-the-art deep learning algorithms. We collected, reviewed, and analyzed 179 state-of-the-art studies in the past decade, consolidating siloed knowledge of user-centric design constraints and objectives, hybrid generative AI methods, data sources for development and testing, as well as benchmarking methods and metrics for assessing design performance, thereby providing a comprehensive understanding of the efficacy of generative AI technologies across diverse design contexts. • Review state-of-the-art generative AI approaches, focusing on deep learning. • Address three design and planning areas – site layout, interior, and exterior design. • Selected and analyzed 179 state-of-the-art studies. • Detail user-centric, data-driven, hybrid-method, and benchmarking strategies.