Scholarlink

<p>To address the challenges of global urbanization and housing shortages, implementing practical densification approaches often necessitates tailoring solutions based on the local context and prevailing housing typologies. However, such expansion strategies have been limited to unidirectional stacking or single-direction extensions while also heavily relying on designers' previous experience and subjective judgment. Therefore, this study proposes a novel machine learning (ML)-based framework for generating multi-directional house extension options, enabling efficient and contextually appropriate residential densification. Unlike existing approaches limited to unidirectional expansion and subjective designer input, our framework automatically identifies suitable land surfaces, conducts voxel-based generation of extensions, and incorporates customizable, structurally valid prefabricated components. A case study of Walthamstow, a neighborhood in North London, UK, demonstrates the framework's potential for significant residential densification. Key findings reveal that our proposed data-driven approach can generate scalable densification solutions tailored to diverse residential building types and neighborhoods, offering a promising strategy to reduce urban sprawl, alleviate the housing crisis, and minimize environmental impact through efficient, automated, and contextually sensitive design. This ML-based framework significantly advances automated densification strategies, providing a practical tool for sustainable urban development.</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

Machine learning-based urban densification: extending roof ridge lines for sustainable housing extension using generative adversarial networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月10日

查看全文
  • Yangzhi Li,
  • Jingwei Li,
  • Qiwei Song

摘要

To address the challenges of global urbanization and housing shortages, implementing practical densification approaches often necessitates tailoring solutions based on the local context and prevailing housing typologies. However, such expansion strategies have been limited to unidirectional stacking or single-direction extensions while also heavily relying on designers' previous experience and subjective judgment. Therefore, this study proposes a novel machine learning (ML)-based framework for generating multi-directional house extension options, enabling efficient and contextually appropriate residential densification. Unlike existing approaches limited to unidirectional expansion and subjective designer input, our framework automatically identifies suitable land surfaces, conducts voxel-based generation of extensions, and incorporates customizable, structurally valid prefabricated components. A case study of Walthamstow, a neighborhood in North London, UK, demonstrates the framework's potential for significant residential densification. Key findings reveal that our proposed data-driven approach can generate scalable densification solutions tailored to diverse residential building types and neighborhoods, offering a promising strategy to reduce urban sprawl, alleviate the housing crisis, and minimize environmental impact through efficient, automated, and contextually sensitive design. This ML-based framework significantly advances automated densification strategies, providing a practical tool for sustainable urban development.