杨天人 金鹰 方舟

Yang Tianren, Jin Ying, Fang Zhou

关键词：城市系统；人工智能；多源数据；规划实施评估；空间规划编制；空间形态生成

Keywords：Urban Systems; Artificial Intelligence; Multi-source Data; Planning Evaluation; Spatial Planning; Urban Fabric Generation

摘要：

随着城市多源数据兴起，城市系统模型与人工智能技术成为建立数据间内在关联（如非线性与因果性关系）的核心基础。本文以两者的典型代表——“空间均衡模型”与“对抗生成网络”为例，总结梳理两者在城市研究与实践中的理论基础、优势与局限以及应用场景。城市系统模型更适用于支持大尺度的规划实施评估（通过反事实模拟）与空间规划编制（通过情景预测），而人工智能技术则更适用于基于现状案例与规划指引的小尺度城市空间形态生成。基于两者的优势互补性，跨尺度的模型耦合可以为探索因地制宜、多维度共赢的城市决策提供可量化、可解释的科学依据。本文解释了多源数据在规划与设计中囿于现状描绘的局限与原因，挖掘了其在模型支撑下辨析城市问题与优化空间决策的应用潜力。

Abstract：

With the emergence of multi-source urban data, urban systems models and artificial intelligence play a fundamental role in unraveling the relationships (e.g., nonlinear and causal) between different urban sectors. Using spatial equilibrium frameworks and generative adversarial networks as examples, this paper summarizes the theoretical foundations, strengths and weaknesses, and application scenarios of the two types of models. Urban systems models are suitable for large-scale applications, including planning evaluation (through counterfactual simulations) and spatial planning (through scenario forecast). Comparatively, artificial intelligence is more useful to support plan-making (based on examples and planning guidance) at a finer spatial level. A complementary use of the two models can derive quantifiable and explainable evidence for urban decision-making that considers the trade-offs across urban sectors and geographic units. This paper identifies the constrained use of multi-source data in analyzing past trends and advocates their potential usages to explain urban issues and optimize spatial strategies based on model development.

杨天人，香港大学建筑学院城市规划与设计系，助理教授；剑桥大学马丁建筑与城市研究中心，研究员

金鹰（通信作者），剑桥大学马丁建筑与城市研究中心，教授。yj242@cam.ac.uk

方舟，剑桥大学马丁建筑与城市研究中心，博士研究生

• [1] 杨天人, 吴志强. 美国城市规划院校2000—2014 年研究动态[J]. 城市规划学刊, 2017(4): 10-19.

  [2] 吴志强. 人工智能辅助城市规划[J]. 时代建筑, 2018(1): 6-11.

  [3] WEBER R, CRANE R. The Oxford handbook of urban planning[M]. Oxford: Oxford University Press, 2015.

  [4] PEARL J, MACKENZIE D. The book of why: the new science of cause and effect[M]. Oxford: Basic Books, 2018.

  [5] SCH?LKOPF B. Causality for machine learning [J/OL]. arXiv e-prints, 2019[2021-02-01]. https://arxiv.org/abs/1911.10500.

  [6] 张钹, 朱军, 苏航. 迈向第三代人工智能[J]. 中国科学: 信息科学, 2020, 50(9): 1281-1302.

  [7] LEE D B. Requiem for large-scale models[J]. Journal of the American Institute of Planners, 1973, 39(3): 163-178.

  [8] BATTY M. Urban modeling[M] // KITCHIN R, THRIFT N. International encyclopedia of human geography. Oxford: Elsevier, 2009: 51-58.

  [9] 万励, 金鹰. 国外应用城市模型发展回顾与新型空间政策模型综述[J]. 城市规划学刊, 2014(1): 81-91.

  [10] ARROW K J, DEBREU G. Existence of an equilibrium for a competitive economy[J]. Econometrica, 1954, 22: 265-290.

  [11] BURFISHER M E. Introduction to computable general equilibrium models[M]. Cambridge: Cambridge University Press, 2011.

  [12] YANG T. Long-term prospects of land value uplift in planned new urban centres: measurement, modelling and predictions[D]. Cambridge: University of Cambridge, 2020.

  [13] JIN Y, ECHENIQUE M, HARGREAVES A. A recursive spatial equilibrium model for planning large-scale urban change[J]. Environment and planning b: planning and design, 2013, 40(6): 1027-1050.

  [14] HARTMANN S, WEINMANN M, WESSEL R, et al. StreetGAN: towards road network synthesis with generative adversarial networks[C] // 25th International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision, 2017, 133-142.

  [15] DEAL B, PAN H, TIMM S, et al. The role of multidirectional temporal analysis in scenario planning exercises and planning support systems[J]. Computers, environment and urban systems, 2017, 64: 91-102.

  [16] YANG T. Understanding commuting patterns and changes: counterfactual analysis in a planning support framework[J]. Environment and planning b: urban analytics and city science, 2020, 47(8): 1440-1455.

  [17] YANG T, JIN Y, YAN L, et al. Aspirations and realities of polycentric development: insights from multi-source data into the emerging urban form of Shanghai[J]. Environment and planning b: urban analytics and city science, 2019, 46(7): 1264-1280.

  [18] FANG Z, YANG T, JIN Y. DeepStreet: a deep learning powered urban street network generation module[J/OL]. arXiv e-prints, 2020[2021-02-01]. https://arxiv.org/abs/2010.04365.

  [19] FANG Z, QI J, YANG T, et al. “Reading” cities with computer vision: a new multi-spatial scale urban fabric dataset and a novel convolutional neural network solution for urban fabric classification tasks[C]. 28th International Conference on Advances in Geographic Information Systems (SIGSPATIAL’20), Seattle, 2020.

  [20] FANG Z, JIN Y, YANG T. Incorporating planning intelligence into deep learning: a planning support tool for street network design[J/OL]. arXiv e-prints, 2020[2021-02-01]. https://arxiv.org/abs/2010.04536.

  
  

• 收起