| 74 | 0 | 97 |
| 下载次数 | 被引频次 | 阅读次数 |
针对当前城市建筑布局设计在满足现有建筑设计规范的基础上，亟需进一步提升环境性能的问题与挑战，综述并提出了一系列结合机器学习的研究方法。首先，系统解析了城市建筑布局设计问题，提取了设计变量、约束条件和设计目标3个核心要素。在此基础上，依据3要素间的内在关系，将设计问题拆解成生成、预测和优化3个方面，并分别构建了相应的机器学习模型。在生成方面，提出了基于集成学习的辅助生成模型，用以从高维的潜在设计空间中识别设计可行域。在预测方面，构建了基于空频特性增强的建筑室外风环境预测模型，实现不同设计方案环境性能的快速评估；在优化方面，提出了基于课程化多智能体的建筑布局性能优化模型，从而在多目标性能空间中搜索并获得最优布局方案。研究为提升城市建筑布局设计的效率和质量提供了新路径，有助于提高城市环境宜居性。
Abstract：In response to the challenges of enhancing environmental performance while complying with existing design codes in urban residential planning, this paper presents a comprehensive methodology that incorporates artificial intelligence techniques. The study begins with a systematic analysis of the urban building layout design problem. Three core components are identified, which include design variables, constraint conditions, and performance objectives. Based on the relationships among these components, the design process is divided into three main phases, including generation, prediction, and optimization. Each phase is addressed using corresponding machine learning models. In the generation phase, an ensemble learning-based model is proposed to identify feasible design regions within a highdimensional latent space. In the prediction phase, a spatial-frequency-enhanced model is developed to evaluate outdoor wind performance, allowing for efficient assessment of environmental metrics across design alternatives. In the optimization phase, a curriculum-based multi-agent model is employed to search for optimal layout solutions within a multi-objective performance space. This integrated framework enhances the efficiency, adaptability, and environmental quality of urban building layout design, contributing to the creation of more livable residential environments.
[1] FABOLUDE G, KNOBLE C, VU A, et al. Smart cities,smart systems:A comprehensive review of system dynamics model applications in urban studies in the big data era[J]. Geography and Sustainability, 2025, 6(1),100246.
[2] LU Y. Urban form and urban heat island:towards a cool built environment[M]. Beijing:Springer Nature,2025.
[3] ZHU Z, WU M, LIU N, et al. Multi-scale exploration of the influence mechanism of 2D/3D morphological characteristics of island cities on LST[J]. Ecological Indicators, 2025, 178:114127.
[4] JOHN P B, FOREST L F, MICHEAL L. A multiobjective feedback methodfor evaluating sequential conceptual building design decisions[J]. Automation in Construction,2014(45):136-150.
[5] CHAILLOU S. AI+architecture:Towards a new approach[J]. Harvard University, 2019:1-8.
[6]刘羿伯，吴梓溶，孙澄.面向新一代人工智能的城市设计：范式转型与数智赋能[J].城市规划学刊，2025(1):64-70.LIU Y B , WU X R, SUN C. Urban design for the new generation artificial intelligence:paradigm shift and digital-intelligent empowerment[J]. Urban Planning Forum, 2025(1):64-70(in Chinese)
[7] STINY G, GIPS J. Shape grammars and the generative specification of painting and sculpture[A]//HANS J K.IFIP Congress(2)[C]. Amsterdam:Elsevier Science Pub. Co,1971, 2(3):125-135.
[8] MARCH L. Forty Years of Shape and Shape Grammars,1971-2011[J]. Nexus Network Journal, 2011, 13(1):5-13..
[9] STOUFFS R, JANSSEN P. A rule-based generative analysis methodfor urban planning[J]. Morphological Analysis of Cultural DNA. Singapore:Springer Press,2017:25-36.
[10] SCHANK R C. Dynamic memory:A theory of reminding and learning[J]. Computers and People,1982:1-8.
[11]杨俊宴，朱骁.人工智能城市设计在街区尺度的逐级交互式设计模式探索[J].国际城市规划，2021,36(02):7-15.YANG J, ZHU X. Exploration of AI-based interactive urban design at the block scale[J]. Urban Planning International, 2021, 36(2):7-15.(in Chinese)
[12]曹宇琦，虞志淳.绿色建筑性能化数字化设计方法综述[J].建筑节能（中英文），2023,51(01):47-53.CAO Y, YU Z. A review of performance-oriented and digital design methods for green buildings[J]. Building Energy Efficiency(Chinese&English), 2023, 51(01):47-53.(in Chinese)
[13]韩昀松，赵昕，孙澄.风环境性能驱动的寒地街区计算性设计方法研究与实践[J].当代建筑，2022(06):19-23.HAN Y, ZHAO X, SUN C. Performance-driven computational design of cold-climate urban blocks based on wind environment simulation[J]. Contemporary Architecture, 2022(6):19-23.(in Chinese)
[14] ALOTAIBI A, AHMED M. Neural architecture search for generative adversarial networks:a comprehensive review and critical analysis[J]. Applied Sciences,2025, 15(7):3623.
[15] SALIU N, ELEZI K. The transformative integration of artificial intelligence in architectural practice:From generative design to sustainable building performance[J]. European Chronicle, 2025, 10(1):66-73.
[16] HAN Z, YAN W, LIU G. A Performance-Based Urban Block Generative Design Using Deep Reinforcement Learning and Computer Vision[A]//PHILIP F Y. The International Conference on Computational Design and Robotic Fabrication[C]. Singapore:Springer, 2020:134-143.
[17]姚佳伟，黄辰宇，付斌等.深度强化学习支持下风环境性能驱动的设计研究与实践[J].建筑学报，2022,25(S1):31-38.YAO J, HUANG C, FU B, et al. A study and practice of performance-driven wind environment design supported by deep reinforcement learning[J]. Architectural Journal,2022,25(S1):31-38.(in Chinese)
[18] HUANG C, ZHANG G, YAO J, et al. Accelerated environmental performance-driven urban design with generative adversarial network[J]. Building and Environment, 2022, 224:109575.
[19] DUERING S, CHRONIC A, KOENIG R. Optimizing Urban Systems:Integrated optimization of spatial configurations[A]//AZAM K. Proceedings of the 11th annual symposium on simulation for architecture and urban design[C]. Newyork:Curran Associates, Inc,2020:1-7.
[20] WANG P, GUO M, HAN Y, et al. Ensemble learningbased hierarchical retrieval of similar cases for site planning[J]. Journal of Computational Design and Engineering, 2021, 8(6):1548-1561.
[21] WANG P, GUO M, CAO Y, et al. Pedestrian wind flow prediction using spatial-frequency generative adversarial network[J]. Building Simulation, 2024, 17(2):319-334.
[22]王鹏跃.基于机器学习的城市住区建筑布局设计方法研究[D].北京：北京建筑大学，2023.WANG P Y. Research on urban residential layout design method based on machine learning[D]. Beijing:Beijing University of Civil Engineering and Architecture,2023.(in Chinese)
基本信息:
DOI：10.19740/j.2096-9872.2025.05.11
中图分类号:TU984
引用信息:
[1]赵玲玲,王鹏跃.基于机器学习的城市建筑布局设计方法研究[J].北京建筑大学学报,2025,41(05):120-127.DOI:10.19740/j.2096-9872.2025.05.11.
基金信息:
国家自然科学基金项目（62271036）; 北京市自然科学基金面上项目（4232021）