摘要
基于多智能体深度强化学习算法与生成对抗网络,提出风环境性能驱动的计算性生成设计与交互式优化方法,并展开实践应用。实践结果表明:该方法可有效提高方案设计阶段风环境性能评估效率,增强性能导向的街区空间形态自适应调控能力,提升复杂设计场景下室外风环境优化潜能。
Based on multi-intelligent deep reinforcement learning algorithms and generative adversarial networks,a computational generative design and interactive optimization method driven by wind environment performance is proposed and applied in practice.The results of a case study shows that the proposed method could effectively improve the evaluation efficiency of the outdoor wind environment in the early stage of design,strengthen the ability of self-adaptive morphological control of performance-oriented urban blocks design,and improve the potential of outdoor wind environment optimization efficiency under the complex design scenario.
出处
《建筑学报》
CSSCI
北大核心
2022年第S01期31-38,共8页
Architectural Journal
基金
国家自然科学基金项目(51908410)
上海市青年科技英才扬帆计划(19YF1451000)
上海市级科技重大专项项目(2021SHZDZX0100)
中央高校基本科研业务费专项资金
关键词
室外风环境
街区形态
生成式设计
深度强化学习
生成对抗网络
MADRL-GAN
outdoor wind environment
urban block morphology
generative design
deep reinforcement learning
generative adversarial network
MADRL-GAN
