Retrieval-Augmented Generation for Precedent-Based Support in Early-Stage Architectural Design

 

Early-stage architectural design is highly dependent on precedent cases and accumulated domain knowledge, which help designers explore concepts, establish design logic, and reduce uncertainty. However, existing digital assistance tools struggle to effectively support this phase due to the dominance of visual information and the linguistic diversity found in architectural descriptions. This study addresses these challenges by proposing a retrieval-augmented generation (RAG) framework specifically tailored to architectural design contexts.

Challenges in Precedent-Based Architectural Assistance

Architectural precedents are complex, multimodal, and context-sensitive, combining drawings, images, diagrams, and textual narratives. Traditional retrieval systems often fail to capture the underlying design logic or to align visual data with semantic descriptions. These limitations reduce retrieval accuracy and restrict the usefulness of precedent recommendations during conceptual design.

Architecture-Oriented Knowledge Graph Design

To overcome these issues, the proposed framework introduces a lightweight, architecture-specific knowledge graph that represents core design logic. Instead of relying on overly complex ontologies, the graph captures relationships among spatial concepts, functional intentions, and design strategies. This structure enables efficient organization and reasoning over precedent knowledge relevant to early design exploration.

Multimodal Knowledge Extraction Pipeline

A dedicated knowledge extraction pipeline is developed to process both visual and textual data from architectural precedents. Visual features are interpreted to identify spatial and formal characteristics, while textual descriptions are analyzed to extract semantic and conceptual information. The integration of these modalities ensures a more comprehensive and structured representation of architectural knowledge.

Retrieval, Aggregation, and Question Answering Mechanisms

The framework employs retrieval-augmented generation techniques to aggregate relevant precedent information and support question answering. By combining graph-based retrieval with generative reasoning, the system consolidates multiple precedents into coherent design insights. This approach allows designers to query design intent, spatial strategies, and functional solutions in a natural and intuitive manner.

Experimental Evaluation and Design Support Implications

Experimental results demonstrate that the proposed framework achieves higher retrieval accuracy, broader precedent coverage, and improved user experience compared to conventional methods. By enhancing access to structured precedent knowledge, the system advances intelligent assistance for early-stage architectural design. The study highlights the potential of RAG and knowledge graphs to support more informed, creative, and efficient design decision-making.

Architecture Engineers Awards

🔗 Nominate now! 👉 https://architectureengineers.com/award-nomination/?ecategory=Awards&rcategory=Awardee 🌐 Visit: architectureengineers.com 📩 Contact: contact@architectureengineers.com Get Connected Here: ***************** Instagram :  https://www.instagram.com/architecture_engineers_awards/ Facebook :  https://www.facebook.com/profile.php?id=61576995475934 Tumblr :   https://www.tumblr.com/architectureengineersawards Pinterest :   https://in.pinterest.com/architectureengineersawards/ Blogger :   https://architectureengineers.blogspot.com/ Twitter :   https://x.com/Architectu54920 YouTube :  https://www.youtube.com/@Architechtureengineer LinkedIn :  https://www.linkedin.com/in/architecture-engineer-01a044361/

#AIinArchitecture
#ComputationalDesign
#ArchitecturalKnowledge
#DesignReasoning
#GenerativeAI
#HumanCenteredDesign
#InformationRetrieval
#DesignInnovation
#ConceptualDesign
#BuiltEnvironment
#DigitalArchitecture