Deep Reinforcement Learning: A New Approach to Automatic Planning of Highrise Buildings
SUN Chengyu;SONG Xiaodong;
In the context of highrise and high-density development,automatic planning of building clusters is a promosing area.However,to satisfy the sunlight requirments is still full of difficulties.The latest deep reinforcement learning theories in artificial intelligence provides a new approach to the problem.This study proposes a two-stage approach.Firstly,an initial layout will be generated according to the land development indices including total floor area and building heights.The plan will then be optimized in terms of sunlight and requirements according to a framework of"one axis,dualistic model,and four essential elements",which is summarized from the theories.After five automation experiments of highrise building planning in Beijing,Shenyang,Zhengzhou,Shanghai,and Fuzhou,the new approach is proved efficient and reliable.
CAJViewer7.0 supports all the CNKI file formats; AdobeReader only supports the PDF format.