Scholarlink

As the demand for high-rise residential buildings (HRBs) increases, conventional manual design process demanding multiple revisions and expertise encounters significant challenges. Although several methods have attempted to automate HRB layout design, applicable solutions across diverse scenarios are still lacking. We present a framework that streamlines the HRB layout design process with reinforcement learning by treating every suite as an autonomous agent. The design requirements of functionality, safety, living comfort and efficiency are considered. Based on the design drawing input, we establish the environment through abstraction and rescale, and search for its initial state satisfying part of the design requirements. With a deep Q-network (DQN) algorithm, the suite agents are then autonomously aggregated towards the optimal solution determined by the user-specified preference. This proposed framework is demonstrated capable of producing practical and innovative layouts for various goals, highlighting its potential to automate HRB layout design.

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

Automatic Framework for High-Rise Residential Building Layout Design Using Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yunzhu Liao,
  • Xuhong Zhou,
  • Jiepeng Liu,
  • Liang Feng,
  • Hongtuo Qi,
  • Gan Luo

摘要

As the demand for high-rise residential buildings (HRBs) increases, conventional manual design process demanding multiple revisions and expertise encounters significant challenges. Although several methods have attempted to automate HRB layout design, applicable solutions across diverse scenarios are still lacking. We present a framework that streamlines the HRB layout design process with reinforcement learning by treating every suite as an autonomous agent. The design requirements of functionality, safety, living comfort and efficiency are considered. Based on the design drawing input, we establish the environment through abstraction and rescale, and search for its initial state satisfying part of the design requirements. With a deep Q-network (DQN) algorithm, the suite agents are then autonomously aggregated towards the optimal solution determined by the user-specified preference. This proposed framework is demonstrated capable of producing practical and innovative layouts for various goals, highlighting its potential to automate HRB layout design.