感知·推演·生成·仿真:智能化城市设计方法研究  

PERCEPTION,DEDUCTION,GENERATION,AND SIMULATION:A STUDY ON INTELLIGENT URBAN DESIGN METHODS

作　　者：杨俊宴[1,2,3] 盛华星 史北祥[2] 崔澳 孙昊成 马巍[4] YANG Junyan;SHENG Huaxing;SHI Beixiang;CUI Ao;SUN Haocheng;MA Wei

机构地区：[1]东南大学 [2]东南大学建筑学院 [3]东南大学智慧城市研究院 [4]东南大学无锡校区建筑学院

出　　处：《城市规划》2025年第4期65-73,94,共10页City Planning Review

基　　金：国家自然科学基金重大项目课题(52394224);江苏省重点研发计划项目(BE2023799)。

摘　　要：在新一代多模态、可解释人工智能技术的推动下,人机交互的数字化城市设计正孕育着大量新的发展可能。本文在第四代城市设计范式内,探讨了基于深度学习、大语言模型、混合现实为核心的智能化城市设计方法,提出了“感知—推演—生成—仿真”的探索,感知包括街景、人群行为和城市意象等的采集感知;推演涉及发展需求、格局结构和形态等的预判决策;生成包括设施布局、三维场景等的生成与优化;仿真包括空间模型、人体舒适度和行为活动的交互验证。在此基础上,结合南京中央路城市设计案例,展现智能技术在城市设计中的革新尝试,为城市设计的科学性和人本性提升提供参考。Driven by the new generation of multimodal and explainable artificial intelligence technologies,human-computer interactive digital urban design is fostering numerous new development possibilities.Within the framework of the fourth-generation urban design paradigm,this paper explores intelligent urban design methods centered on deep learning,large language models,and mixed reality.We propose an exploratory framework of“Perception-Deduction-Generation-Simulation”.Perception involves the collection and sensing of street views,crowd behavior patterns,and urban imagery.Deduction entails predictive decision-making regarding development needs,spatial structures,and morphological patterns.Generation includes the creation and optimization of facility layouts and three-dimensional scenes.Simulation involves interactive validation of spatial models,physiological comfort,and behavioral dynamics.Based on this framework,this study applies an urban design case of Zhongyang Road in Nanjing to showcase innovative attempts to integrate intelligent technologies into urban design,providing insights for enhancing its scientific rigor and human-centered approach.

关 键 词：智能化城市设计 人工智能 智能感知 智能推演 智能生成 城市设计 

分 类 号：TU984[建筑科学—城市规划与设计] D63[政治法律—政治学]