隨著數位化與智慧化浪潮席捲建築與室內設計領域，傳統設計流程卻仍高度仰賴設計師經驗與手動建模，在廚房設計的前期構思與平面配置階段，設計師必須不斷重繪與反覆修訂圖面，耗費大量時間與人力。
為解決上述問題，本研究提出一套結合生成式人工智慧(Generative AI)與建築資訊模型(BIM)的AI廚房設計輔助平台，以廚房設計為實證案例。系統架構由「Prompt與指令設定模組」、「智慧庫資料模組」與「輸出控制模組」三大元件組成，並採雙階段AI工作流程：第一階段由GPTs-1模組根據使用者輸入之空間尺寸、門窗配置及家電需求，自動生成結構化設計建議；第二階段由GPTs-2模組將該建議轉譯為Dynamo可執行之Python腳本，最終在Revit中自動構建三維BIM模型。模擬驗證結果顯示，平台產出具有一致性及合理性，與資料庫內容可以達到高度吻合，且經專業廚房設計師驗證具可行性與價值。本研究成果示範了以生成式AI為輔助的自動建模流程，具備高度可擴展性，未來可應用於各類室內空間與多模態互動場景。

Despite the rapid digitalization and intelligent transformation of the architecture and interior-design industries, conventional workflows still depend heavily on designers’ experiential judgment and manual modeling. In kitchen design, iterative redrawing and repeated revisions during the early concept‐development and layout stages consume substantial time and human resources.
To address these limitations, this study develops an AI-assisted kitchen-design platform that synergistically integrates Generative Artificial Intelligence (Generative AI) with Building Information Modeling (BIM). The platform serves as an empirical case for evaluating AI-enabled workflows in kitchen design. The system architecture comprises three core modules—(1) Prompt and Command Configuration, (2) Knowledge-Base Management, and (3) Output Control—and employs a two-stage AI workflow. In Stage 1, the GPTs-1 module automatically generates structured design recommendations from user-provided room dimensions, openings, and appliance requirements. In Stage 2, the GPTs-2 module translates these recommendations into Python scripts executable in Dynamo, thereby enabling the automatic construction of three-dimensional BIM models in Revit. Simulation and case-study evaluations demonstrate that the platform produces designs of high consistency and rationality, achieving a near-perfect match with the underlying knowledge base. Validation by professional kitchen designers further confirms its practical feasibility and value. The proposed generative-AI-driven modeling pipeline exhibits high scalability and is readily applicable to diverse interior spaces and multimodal interactive scenarios.

[1] Buhler, K. (2025). https://www.sequoiacap.com/article/ai-50-2025/
[2] Houzz Inc. (2024). https://www.houzz.com/press-release/us-houzz-home-study-2024-renovation-trends
[3] The Brainy Insights. (2023). https://www.globenewswire.com/news-release/2023/12/04/home-improvement-services-market-to-reach-usd-717-6-billion-by-2032
[4] Bhonde, S., Hartmann, T., & Fischer, M. (2022). Evaluating the use of virtual reality for architectural design review. ITcon, 27, 222–237.
[5] Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., … Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
[6] Chaillou, S. (2020). ArchiGAN: Artificial intelligence in architectural floor‐plan design. In H. H. Amor & P. J. Sharrott (Eds.), CAAD Futures 2019: Computer‐Aided Architectural Design Futures (pp. 35–50). Springer.
[7] Chen, J., Shao, Z., & Hu, B. (2023). Generating interior design from text: A new diffusion model–based method for efficient creative design. Buildings, 13, 1861.
[8] Du, C., Esser, S., Nousias, S., & Borrmann, A. (2024). Text2BIM: Generating building models using a large language model–based multi‐agent framework [Preprint]. arXiv.
[9] Feng, Y., Wang, X., Wong, K. K., Wang, S., Lu, Y., Zhu, M., Wang, B., & Chen, W. (2023). PromptMagician: Interactive prompt engineering for text-to-image creation [Preprint]. arXiv.
[10] Gao, Y., Cheng, P., & Liang, J. (2023). DiffuScene: High‐fidelity text‐to‐interior rendering via conditional diffusion models. Computer Graphics Forum, 42(7), 155–168.
[11] He, Z., Wang, Y.-H., & Zhang, J. (2025). Generative AIBIM: An automatic and intelligent structural design pipeline integrating BIM and generative AI. Information Fusion, 114, 102654.
[12] Henderson, P., Subr, K., & Ferrari, V. (2019). Automatic generation of constrained furniture layouts. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 1711–1720).
[13] Hsu, Y.-C., Fontaine, M. C., Earle, S., Edwards, M., Togelius, J., & Nikolaidis, S. (2022). Generating diverse indoor furniture arrangements [Poster]. In SIGGRAPH ’22 Posters (pp. 1–2).
[14] Le, M.-H., Chu, C.-B., Le, K.-D., Nguyen, T. V., Tran, M.-T., & Le, T.-N. (2023). VIDES: Virtual interior design via natural language and visual guidance [Preprint]. arXiv.
[15] Li, C., Zhang, T., Du, X., Zhang, Y., & Xie, H. (2024). Generative AI for architectural design: A literature review [Preprint]. arXiv.
[16] Liu, R., & Ismail, A. I. (2024). Application and challenges of generative AI in architectural design: A case study of GPT-4. Evolutionary Studies in Imaginative Culture, 8(S2), 990–1000.
[17] Ma, J., Erdogmus, E., & Cha, S. H. (2023). Integration of a choice modeling approach with immersive virtual environments for accurate space utilization prediction. Journal of Building Engineering, 76, 107126.
[18] Mertens, A., Hamarat, Y., & Elsen, C. (2022). Interactions between architects and end-users during housing design processes: A systematic literature review. Archnet-IJAR: International Journal of Architectural Research, 17(6), 1201–1225.
[19] Mohamed, M. F. M., Kamel, I., & Abdelaal, A. H. (2021). CAD using preference compared to hand drafting in architectural working drawings coursework. Ain Shams Engineering Journal, 12(3), 2975–2984.
[20] Pejic, J., & Pejic, P. (2022). Linear kitchen layout design via machine learning. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 36(9), 1–12.
[21] Rane, N., Choudhary, S., & Rane, J. (2023). Integrating Building Information Modelling (BIM) with ChatGPT, Bard and similar generative artificial intelligence in the architecture, engineering and construction industry: Applications, framework and challenges [Preprint]. SSRN.
[22] Sanguinetti, P., Abdelmohsen, S., Lee, J., Lee, J., Sheward, H., & Eastman, C. M. (2012). General system architecture for BIM: An integrated approach for design and analysis. Advanced Engineering Informatics, 26(2), 317–333.
[23] Steinfeld, E., & Maisel, J. (2012). Universal design: Creating inclusive environments. John Wiley & Sons.
[24] Sun, M., Xu, Z., Wang, L., & Nießner, M. (2022). Text2Room: Extracting textured 3D meshes from text-guided image generation. In Proceedings of the European Conference on Computer Vision (pp. 7909–7927).
[25] Tanasra, H., Rott Shaham, T., Michaeli, T., Austern, G., & Barath, S. (2023). Automation in interior space planning: Utilizing conditional generative adversarial network models to create furniture layouts. Buildings, 13(7), 1793.
[26] Vartanian, O., Navarrete, G., Palumbo, L., & Chatterjee, A. (2021). Individual differences in preference for architectural interiors. Journal of Environmental Psychology, 77, 101668.
[27] Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E. H., Le, Q., & Zhou, D. (2022). Chain-of-thought prompting elicits reasoning in large language models [Preprint]. arXiv.
[28] White, J., Fu, Q., Hays, S., Sandborn, M., Olea, C., Gilbert, H., … & Schmidt, D. C. (2023). A prompt pattern catalog to enhance prompt engineering with ChatGPT [Preprint]. arXiv.
[29] Wu, H., Li, Q., Zhang, R., & Chen, X. (2024). LayoutGPT: Conversational large-language-model framework for real-time interior layout generation. Automation in Construction, 159, 105080.
[30] Zheng, J., & Fischer, M. (2023). BIM‐GPT: A prompt‐based virtual assistant framework for BIM information retrieval [Preprint]. arXiv.
[1]普克拉室內平台. (2022)，https://www.puloapp.com/2022-interior-design-trend-resport/
[2]接案平台PRO360 達人網. (2021)，https://www.pro360.com.tw/
[3]王柔心(2022)，以原宇宙概念建構營建協作平台-以室內裝修為例，國立成功大學碩士論文
[4]張鎧丞(2024)，應用生成式AI打造沉浸式室內裝修體驗-以廚房系統櫃為例，國立成功大學碩士論文