Scholarlink

<p>Functional layout planning in apartment design often demands time-intensive iteration to balance optimization and creativity. This study presents a hybrid AI framework that integrates Genetic Algorithms (GA), Variational Autoencoders (VAE), and Pix2Pix models to generate optimized yet adaptable floor plans. A GA navigates the VAE’s latent space to minimize internal movement by optimizing room centroids, followed by Pix2Pix-based rendering of these layouts into visual plans. Compared to direct sampling, the GA-VAE approach yields higher fitness scores and more coherent spatial configurations. While Pix2Pix outputs serve as draft visualizations (FID: 39), the framework allows for human-in-the-loop refinement. Limitations include dataset scale, single-objective focus, and the need for manual adjustments. Future work will target multi-objective optimization, enhanced visual fidelity, and integration of real-world architectural constraints.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a hybrid model between generative adversarial networks (GANs) and Genetic Algorithms (GA) in apartment layout optimization

  • 期刊论文

  • 发表日期：

    2025年11月11日

查看全文
  • Vu Hong Son Pham,
  • Duy Hoa Nguyen,
  • Thi Bich Huyen Vo

摘要

Functional layout planning in apartment design often demands time-intensive iteration to balance optimization and creativity. This study presents a hybrid AI framework that integrates Genetic Algorithms (GA), Variational Autoencoders (VAE), and Pix2Pix models to generate optimized yet adaptable floor plans. A GA navigates the VAE’s latent space to minimize internal movement by optimizing room centroids, followed by Pix2Pix-based rendering of these layouts into visual plans. Compared to direct sampling, the GA-VAE approach yields higher fitness scores and more coherent spatial configurations. While Pix2Pix outputs serve as draft visualizations (FID: 39), the framework allows for human-in-the-loop refinement. Limitations include dataset scale, single-objective focus, and the need for manual adjustments. Future work will target multi-objective optimization, enhanced visual fidelity, and integration of real-world architectural constraints.