中国的城市经过三十多年的高速发展, 到2017年城镇化率已升至58.5%。随着城市的不断发展扩张, 很多原本处在城市边缘的区域逐渐被纳入到中心城区的范围, 其在城市中的功能定位也可能发生改变。这些区域的空间已经不再适应城市发展的需要, 成为城市中有待更新的区域。在城市更新中, 如何处理建成环境与小尺度城市区域的发展需求之间的关系, 是我们需要在设计中思考的问题。

　　 笔者将以研究生一年级的数字城市设计课题为例, 试图通过Java编程的方式, 来探索在城市建成区环境下的一种可能的城市形态生成方法。

　　 本次设计课题的场地位于江苏省南京市明城墙金川门至神策门及其附近区域 (图1, 2) , 属于具有历史文脉的城市地段。在场地南部有明城墙遗迹, 在场地北部还有民国时期修建的铁路线, 现在已经废弃。我们将这两个区域作为主要体现场地历史文脉的景观区, 在设计中进行充分的展现 (图3) 。

　　 在设计前期, 我们对场地现状进行了调研并确定了整个场地的规划策略。根据现状和规划方案, 我们以街区为单位确定其功能定位, 设定了四种街区类型:商业+商务型、商业+居住型、创业园区、居住区 (图4) , 场地中的每一个街区都被定义为一种类型。考虑到场地内现存居住区大部分状况良好, 且居住建筑形态较为单一, 因此我们保留了场地内的居住区, 而对其他类型的街区进行重新生成。

　　 我们将带有场地条件信息和街区类型的街区平面轮廓作为输入条件, 其中场地条件信息包括边界、各等级道路以及景观。我们以街区为单位进行城市形态的生成, 主要包括场地准备和建筑生成两个阶段 (图5) 。

　　 场地准备阶段主要包括以下几个步骤, 在每个街区中进行地块划分, 生成支路、道路退线, 而后对地块进行细分生成网格 (图6) 。采用网格的方式对地块进行细分, 是因为格网系统能够将建筑空间约束在特定的模数之下, 使我们能够较为简单地控制建筑形态^[1]。场地准备阶段不是本文重点, 在此不进行详细介绍。

　　 在场地准备的基础上, 建筑生成阶段主要需要考虑内部因素和外部因素两方面, 内部因素包括功能、结构、空间品质对建筑形态的要求, 外部因素包括道路、公共交通站点、景观、气候条件等。

　　 外部因素会对街区中地块的土地价值分布造成影响, 继而影响每个地块的土地开发强度, 即各个地块的容积率。在对场地的整体规划中, 我们根据功能定位, 为每个街区设定了一个整体容积率。本次设计中, 我们将主干道、地铁站点、景观带 (图7) 作为主要外部影响因素, 以各个地块到主干道、地铁站点、景观带的距离作为参数, 将整体容积率分配到各个地块上 (图8, 9) 。通过调整每个街区中容积率的分布, 来控制整体建筑群的高度变化, 以形成适当的城市形态。基本计算方法如下:

　　 distA:各地块中心点到景观带边界的最小距离

　　 distB:各地块中心点到主干道的最小距离

　　 distC:各地块中心点到各地铁站点的最小距离

　　 DistN:对地块N的三种距离的综合评价

　　 FARN:地块N的容积率

　　 本次设计中主要根据不同建筑功能的要求, 设置三种体量组合类型:裙房+塔楼、单栋楼房、多层组团。对于不同的街区类型, 采用不同的规则选取网格中应该被建筑体量所占据的格子, 用以生成建筑体量。我们通过编程的方式来制定这些规则。

　　 在商业+商务与商业+居住的街区中, 为提升经济效益, 靠近主干道、地铁站点的地块采用裙房+塔楼形态 (图10) ;为减少对景观的遮挡, 靠近景观区的地块采用单栋楼房的形态。创业园区中对办公空间的需求较少, 但需要更多的交流和更好的户外环境, 因此采用多层组团的形态形成内院, 提供更多的公共空间 (图11) 。

　　 根据不同建筑类型, 我们为不同建筑体量设置四种立面形式, 分别是居住塔楼、办公塔楼、商业裙房、创业园办公楼。生成立面的方式与体量生成类似, 首先将立面按照适当的模数进行网格细分, 再通过人工编程制定不同的选取格子的规则, 生成不同的立面形式。

　　 在本次设计课程中, 我们对城市建成区中基于规则的建筑生成方法做了初步探索。作为一种辅助城市设计的方式, 基于规则的城市形态生成方法, 能够较好地将设计者对设计问题的前期分析结论纳入到方案生成过程中, 同时能够即时反馈设计结果, 帮助设计者高效快速地深化设计方案。但这一方法仍有以下不足和有待深化完善之处:

　　 (1) 在设计前期阶段, 仍然采用基于经验的传统调研分析与规划设计方式, 没有充分结合多种城市数据进行分析^[2]。

　　 (2) 立面生成涉及到很多设计意象和建筑性能上的问题, 在本次设计中没有做深入探讨。

　　 (3) 基于规则的生成方式缺少自我评价和优化的过程, 仍然是一个试错的过程, 而不是一个多目标优化的过程。

　　 (4) 目前尚未建立完整的人为因素评价体系。传统环境行为学方法的效率低下、样本数量有限导致结果不够客观可信、难以深入挖掘, 产生超出主观经验之外的新认知等弊端。近年来以机器学习为代表的新技术、新数据的介入, 使我们有可能去测度以往“不可测度”的物质空间对人的感受和行为的影响, 并辅助我们进行城市设计方案的评价和优化^[3]。

　　 After 30 years effort of reform and opening-up, China’s urbanization rate has risen to 58.5%in 2017 from 17.9%.As a city continues to expand, many areas originally located on the edge of the city are gradually being included in the central area of the city, and their functional orientation may change.And urban space in these areas is no longer adapted to the needs of urban development.This situation often leads to the development of these areas backwardness.And these areas become where need to be renovated in the city.In urban renewal, how to deal with the relationship between built environment and requirements of urban development of small-scale urban areas is a problem that requires us to think in the process of designing.

　　 Taking the digital urban design work of the first-grade of postgraduates in School of Architecture, Southeast University as an example, we attempt to explore a possible method for generating urban form in urban built environment through the way of Java programming.

　　 The site of this project is around the part of the Ming City Wall between the Jinchuan gate and the Shence gate in Nanjing, Jiangsu Province, which is a historic area.In the southern part of the site, there are the remains of the Ming City Wall.In the northern part of the site, there is a railway track constructed in the period of the Republic of China which has been abandoned.In our design, we regard these two parts as landscape source that mainly reflect the historical context of the site.

　　 We conducted a survey on the current status of our site and determined the development planning strategy of this area.According to the current status and our planning, we set up four block types:commercial+office, commercial+residential, innovation park and residential area, each block in our site was defined as one of these types.In view of the fact that most of the existing residential areas in the site are in good condition and the architectural form of residential buildings are relatively simple, we remained the residential areas in our site and regenerated buildings in blocks of the other types.

　　 We use the outline and type information of each block and site condition information, such as boundary, roads in different level and landscape as input information.We take a block as the unit to generate the urban form.The generation process mainly includes two stages, the block preparation stage and the building generation stage.

　　 Block preparation stage includes the following steps, divide each block into lots, generate branch roads, make the legal distance between lots and roads, subdivide each lot to make grid.We use grids to subdivide because the grid system is able to constrain the building space by a certain modulus and allows us to control the architectural form more easily (GUO Zifeng, 2017, pp.35-41) .Because this is not the main content of this article, we will not describe this part in detail.

　　 The building generation stage is based on the block preparation stage, mainly considering internal factors and external factors.Internal factors include functional requirements for architectural forms, structural requirements for architectural forms, the requirements of space quality to architectural forms.External factor includes traffic, public transport station, landscape, climatic condition, etc.

　　 External factor may affect the distribution of land values in each block, which in turn affects the intensity of land development in different part of each block, which means the FAR of each lot should be different in each block.In the master planning of this site, we set a total FAR for each block based on the functional orientation.In this project, we take the main road, subway station and landscape as the external factors.We take the distance from each lot to main road, subway station and landscape as parameters to distribute the total FAR to each lot.By controlling the FAR of each lot, we can control the architectural form to generate a proper urban form.Following is the calculate process of each lot’s FAR.

　　 distA:the minimum distance between the centre of a lot and Landscape boundary

　　 distB:the minimum distance between the centre of a lot and main street

　　 distC:the minimum distance between the centre of a lot and metro stations

　　 DistN:the evaluation result of lot N on the three kinds of distance

　　 FARmax, FARmin:the maximum and minimum FAR in a block, they are set by us according to the planning restriction and the reality of this block.

　　 FARN:the FAR of lot N

　　 a, b, c is the weight of each kind of distance

　　 In this project, we set three combinations of architectural volume, according to the requirements of different building functions:podium+tower, single building, clustering multi-layer building.We need different rules about how to select those small squares to generate different architectural volume.By manual programming, we define the rules.

　　 In commercial+office and commercial+residential blocks, the lot which close to the main road, the subway station, should use podium+tower form, in order to maximize the land value.For lots which near the landscape area, we make it in single building form, in order to reduce the block on the landscape view.In the innovation park, users’requirement for office space is less, but they need more communication and better outdoor environment, so we use clustering multi-layer buildings to create more public space.

　　 According to different building type, we set up four façade forms for different building volume, namely residential towers, office towers, commercial podiums and office buildings in Innovation Park.The way to generate facade is similar to that of volume generation.First, facades are subdivided into grids according to a appropriate modulus.Then, through manual programming, different rules for selecting grids are defined.We use these rules to generate different facade forms.

　　 In this curriculum, we made a preliminary exploration of the rule-based building generation method in the urban built environment.As a way to assist urban design, this method can incorporate the pre-analysis conclusion of designers into the generation process and can feedback the design results instantly to help designers modify the design quickly and efficiently.

　　 However, this method still has many deficiencies that need to be perfected.

　　 In the early stage of this design, we still use the traditional empirical research methods based on experience and general planning design methods.Multiple sources city data is not fully used (SHENG Qiang, YANG Tao and LIU Ning, 2017, pp.38-43) .

　　 Façade generation involves many detail design problems, which are not our focus point, so we have not considered it carefully.

　　 The rule-based generation method lacks self-evaluation and optimization process, which means it is still a trial and error process, not a multi-objective optimization process.

　　 The complete human factor evaluation system for urban design has not been established yet.The traditional methods of environmental behavior science have many problems.In recent years, new data and new technologies, represented by machine learning, have made it possible for us to measure the impact of physical space on human feelings and behaviors which used to be regarded as unmeasured, and to assist us in the evaluation and optimization of urban design proposals. (YE Yu and DAI Xiaoling, 2017, pp.6-13)