| 333 | 0 | 24 |
| 下载次数 | 被引频次 | 阅读次数 |
人群行为与建筑空间形态之间的关联性构成了建筑设计的基本起点。然而，空间行为作为一种高度复杂的规律，往往难以被建筑师以直观方式认知。本研究通过整合图像识别和图像生成两种人工智能模型，提出了针对空间行为的人机协作设计方法，并以商场中庭空间平面设计为例进行验证。首先，通过智能行为感知技术，实地采集商场中庭空间内的微观行为数据，并转化为人群热力图；进一步，将人群热力图和中庭平面图进行匹配训练，基于生成对抗网络（GANs）构建“图对图”的双向输出模型；最后，通过将模型嵌入建筑设计流程，构建一种基于“空间—行为”预测和“行为—空间”生成的建筑设计方法。本文验证了生成对抗网络处理建筑微观行为数据的有效性，可以在建筑设计早期阶段为空间行为推演与优化提供智能决策支撑。
Abstract：The correlation between human behavior and spatial layout constitutes the fundamental starting point of architectural design. However,spatial behavior,as a highly complex pattern,is often difficult for architects to intuitively perceive. This study proposes a human–machine collaborative design method for spatial behavior by integrating image recognition and image generation AI models,and verifies it through the floor plan design of shopping mall atriums. First,micro-scale behavior data within the atrium space are collected through AI behavior sensing technologies and converted into crowd heatmaps. Then,the crowd heatmaps and atrium floor plans are matched and trained to construct a bidirectional “image-to-image” output model based on Generative Adversarial Networks(GANs). Finally,by embedding the model into the architectural design process,a design method is established that enables “space-to-behavior” prediction and “behavior-to-space” generation. This study verifies the effectiveness of GANs in processing micro-scale behavioral data in architecture and demonstrates that it can provide intelligent decision-making support for spatial behavior speculation and optimization at the early stage of architectural design.
[1]Stanislas C.ArchiGAN:Artificial Intelligence x Architecture[C].Archite ctural Intelligence:Selected Papers from the 1st International Conference on Computational Design and Robotic Fabrication(CDRF 2019),2019:117-127.
[2]Huang W,Zheng H.Architectural drawings recognition and generation through machine learning[C].Proceedings of the38th annual conference of the association for computer aided design in architecture,2018:18-20.
[3]李飚,张佳石,卢德格尔·霍夫施塔特,等.算法模型解析设计黑箱[J].建筑师,2019(1):94-99.
[4]He Q,Li Z,Gao W.Predictive models for daylig ht performance of general floorplans based on CNN and GAN:A proof-of-concept study[J].Building and Environment,2021,206:108346.
[5]Zeng Y,Li Y,Du P.Smart fire detection analysis in complex building floorplans powered by GAN[J].Journal of Building Engineering,2023 79 107858.DOI:10.1016/j.jobe.2023.107858.
[6]Wang L.Zhang H,Liu X.Exploring the synergy of building massing and façade design through evolutionary optimization[J].Frontiers of Architectural Research,2022,1 1(4):761-780.
[7]Millon H.The Architectural Theory of Francesco di Gio rgio[J].The Art Bulletin,1958(40):257-261.
[8]闫超,李特,袁烽.具身形态学:基于身体行为的计算性几何[J].建筑师,2023(5):4-11.
[9]Lin X,Liang Y,Zhang Y.IE-GAN:a data-driven crowd simulation method via generative adversa rial networks[J].Multimedia Tools and Applications,2023,83(15):45207-45240.
[10]Helbing D,Molnar P.Social Force Model for Pedestrian Dynamics[J].Physical Review E,1995,51(5):4282-4286.
[11]Lee K H,Choi M G,Hong Q.Group behavior from video:a data-driven approach to crowd simulation[J].Proceedings of the2007 ACM SIGGRAPH/Eurographics symposium on compute r animation,2007:109-118.
[12]Ambrose,M.A.Body|Form|Space:Geometric translations of the body in motion[C].The Third International Conference of the Arab Society for Computer Aided Architectural Design(ASCAAD 2007).Alexandria(Egypt),2007:431-438.
[13]Zhao Y,Zhe G,Hao Y.Behavioral Data Analysis and Visualization System Base on UWB Interior Positioning Technology[C].Proceedings of the 23rd CAADRIA ConferenceVolume 2.Beijing(China),2018:217-226.
[14]Schumacher,P.Advancing Social Functionality via Agent Based Parametric Semiology[J].Architectural Design:Parametricism 2.0-Rethinking Architecture's Agenda for the21st Century,2016:108-113.
[15]C reswell A,White T,Dumoulin V.Generative Adversarial Networks:An Overview[J].IEEE Signal Processing Magazine,2018,35(1):53-65.
[16]Sun Y,Jiang L,Zheng H.A Machine Learning Method of Predicting Behavior Vitality Using Open Source Data[C]//AC AD IA2020:Distributed Proximities.O nline and Global,2020:160-168.
[17]陈雷,张伶伶,陈子墨,等.基于图像生成式人工智能的“人—机—人”建筑设计教学模式研究[J].建筑师,2025(5):134-143.
[18]Isola P,Zhu J Y,Zhou T,et al.Image-to-image translation with conditional adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Honolulu,HI,USA:IEEE,2017:5967-5976.https://doi.org/10.48550/a rXiv.1611.07004.
基本信息:
中图分类号:TU201
引用信息:
[1]金衍孜,方亦文,谢雪颖,等.人因智能设计——行为性能驱动的建筑空间生成设计方法[J].建筑师,2026,No.239(01):6-15.
2025-12-08
2025-12-08
2025-12-08