Computational Generation of Hutong Neighbourhood Maps a Shape Grammar for Procedural Modelling to Represent the Historical Beijing Urban Structure
Total Page:16
File Type:pdf, Size:1020Kb
Computational Generation of Hutong Neighbourhood Maps A shape grammar for procedural modelling to represent the historical Beijing urban structure Yuyang Wang1, Asterios Agkathidis2, Andrew Crompton3 1,2,3University of Liverpool 1,2,3{Y.Wang179|a3lab|Andrew.Crompton}@liv.ac.uk The Hutong neighbourhoods, composed of Chinese courtyard dwellings (Siheyuan), are historically and socially significant urban spaces that embody traditional Chinese philosophy. There is a global interest in these spaces many of which face oblivion, this study presents a formal grammar for parametrically generating Hutong neighbourhoods, which replicates the morphological characteristics of historic cases. This research investigates traditional principles of urban planning of ancient Beijing, based on examples on the historical map Qianlong Jingcheng Quantu, to derive the lost design rules. These rules are then used to build up a procedural modelling process, which reveals the development of Beijing urban structure from Yuan (1271-1368) to Qing (1368-1911) dynasty. The grammar presented in the procedural modelling process contributes to the understanding of the generation of Hutong neighbourhoods. In support of traditional Chinese urban design, the grammar will be translated into an algorithm to be implemented in a computational environment by visual scripting that will generate new instances of Hutong neighbourhoods. Keywords: urban morphology, Siheyuan neighbourhood, procedural modelling, shape grammar BACKGROUND formed due to population increase, after which they Hutong neighbourhoods are historically and socially were composed of variants of a vernacular dwelling significant part of Chinese urban heritage that em- type- Chinese courtyard housing (Siheyuan). Since bodying traditional Chinese philosophy. The ur- these neighbourhoods usually existed between two ban area of ancient Beijing was divided into four- Hutong (alley), they were called Hutong neighbour- sided sectors as residential neighbourhoods under hoods, whose basic units are Siheyuan. A typical Si- the constraint of an orthogonal grid system. Dur- heyuan consists of axially aligned courtyards sym- ing the Qing dynasty, these neighbourhoods were re- metrically surrounded by individual buildings, con- Rule based systems and shape grammar - Volume 2 - eCAADe 39 | 303 nected by orthogonally located corridors, walls, and yards of Siheyuan into five categories to understand gates, as illustrated in Figure 1. how the different Siheyuan layouts were deformed to adapt themselves in the various site contexts within the Hutong neighbourhoods (Figure 2). Figure 1 One common conclusion of these studies is that Beijing Siheyuan, the rules underlying the evolution of Hutong neigh- Ma, 1999. bourhoods are complex. However, computational approaches are powerful to make generative rules explicit and to develop the design iterations. The Shape Grammar, invented by George Stiny (1980) to interpret the generative rules of objects graph- ically, was implemented by Rudi Stouffs (2018) us- ing algorithms developed in Rhino/Grasshopper en- vironment. An analytical shape grammar has its ad- vantage in straightforwardly demonstrating the mor- phological iteration of shapes, which was employed It is noted that current architects lack physical ex- to explain historical styles or languages of designs in amples to study their design knowledge. Duan’s procedural modelling approaches, such as ‘Rome Re- (2007, 2016) measuring surveys indicate most his- born 2.0’ project (Dylla et al., 2008) and urban blocks torical Siheyuan and Hutong neighbourhoods have of nineteenth-century Camden London Flora et al. been damaged and even destroyed. Meanwhile, it is (2018). noted there is an increasing interest in Hutong neigh- Figure 2 bourhood and Siheyuan, but, due to the linguistic A section of difficulties in understanding historical literature, it is Qianlong hard for architects and urban designers to learn the Jingcheng Quantu original design knowledge of traditional Chinese ar- (Qianlong Capital chitecture (Di, 2016). To explicate the lost historical Map, 1750) design principles for contemporary architects and ur- indicating the ban designers and provide a tool for rapidly gener- urban planning of ating Hutong neighbourhoods replicating the tradi- historical Beijing is tional morphological characteristics may help them based on a grid in their design work. system. Since the Hutong neighbourhoods were com- posed of Siheyuan dwellings containing a large num- ber of variants, morphological studies of the neigh- bourhoods were linked to the typological analysis of Siheyuan. Li (2009) revealed the principles explain- ing how the archetype of Siheyuan evolved variants that compose Hutong neighbourhoods on Beijing’s AIMS, METHODS, AND MATERIALS orthogonal grid. Similarly, Liu (2019) investigated Our overall aim is to investigate the design rules many Siheyuan precedents within five Hutong neigh- underlying historical Hutong neighbourhood design bourhoods drawn on Qianlong Jingcheng Quantu in order to translate them into an algorithmic tool (Qianlong Capital Map, 1750) by classifying the court- to generate variations of traditional Beijing Hutong neighbourhoods. To achieve this aim, the following 304 | eCAADe 39 - Rule based systems and shape grammar - Volume 2 questions are addressed: entation, as demonstrated in Rule 2a in Figure 3. We defined these neighbourhoods as ‘Type A Neigh- • How could the historical rules of Hutong neigh- bourhood’ (TAN). It is noted that the form of some bourhood generation and development be neighbourhoods was re-planned in the Qing dy- translated into a shape grammar to simulate nasty. In the south-north orientation, two neigh- its procedural modelling? bourhoods were combined as a whole and were di- • How could such a shape grammar be imple- vided into three new neighbourhoods, with an al- mented as an interactive algorithmic tool in ley width of 10.01 metres between each of the two a computational environment to generate Hu- adjacent new neighbourhoods as shown in Rule 1 tong neighbourhood variants replicating their in Figure 3. We defined these new neighbourhoods original morphological characteristics? as ‘Type B Neighbourhoods’ (TBN). The neighbour- The current paper is focusing on the first question. hoods planned in the Yuan dynasty (TAN) were di- The derivation of rules come from historical litera- vided into 67.67x67.67m Siheyuan plots. Although ture (E, 1739, and Song, 1369) and secondary sources the development of neighbourhoods in the Qing dy- from other studies, such as the morphological studies nasty changed the depth of both Siheyuan plots and on Hutong neighbourhoods by Li (2009), Li and Wang neighbourhoods, the width remained. Rule 2b repre- (2006), Liu (2019), and Ni (2009). The Qianlong Cap- sents the division of a TBN into Siheyuan plots. ital Map recorded the urban plan of Beijing around 1750 at a scale of 1: 650, including all the Hutong Defining Siheyuan plots neighbourhoods and Siheyuans they contained. This Division of plots into sub-plots. A Siheyuan plot in map is also used for rules derivation. the Yuan dynasty could be divided into two to four sub-plots in the east-west orientation. Each sub-plot THE HUTONG NEIGHBOURHOOD GRAM- could include several courtyards aligned in south- north orientation (Type A sub-plot, TASP) or an empty MAR space (Type B sub-plot, TBSP). Based on this finding, By observing the Qianlong Capital Map, it is noted Li and Wang’s (2006) investigation of examples on that there were a large number of Hutong neigh- the Qianlong Capital Map concluded on several types bourhood variants, which resulted in the diverse ur- of sub-plots. As shown in Rules 3 in Figure 4, there ban structure of ancient Beijing within its orthog- are six dividing patterns, in which the plot is consti- onal grid system. We chose the most representa- tuted of one or two sub-plot types. We found that tive the east-west oriented Hutong neighbourhoods, the width of most TBSP is within the range or much from the Yuan dynasty to be our case study. smaller than TASP. Therefore, we define the width of TASP ranges from 12 to 35 metres and TBSP from 5 to Generation of Hutong neighbourhoods 12 metres. The six division patterns of the plots are plan and Siheyuan plots defined as Rules 3a to 3f respectively. In the Yuan dynasty, each neighbourhood’s depth (length in south-north orientation) was planned to Sub-plot combination. Some TASPs and TBSPs were be 67.76 metres and its width (length in east-west combined to regenerate new sub-plots. According orientation) was 677.6 metres. The Hutong alley be- to Liu’s investigation (2019) of selected Siheyuan ex- tween two adjacent neighbourhoods was 9.24 me- amples shown on the map, we inferred the principles tres wide. These key urban elements and dimen- that one TASP could be combined with one adjacent sions are determining the initial shapes of the Hu- TBSP to become a combined sub-plot, called Type AB tong grammar. Each neighbourhood was averagely sub-plot (TABSP). Then the TABSP could be used to divided into ten plots aligned in an east-west ori- construct a Siheyuan. This combination is defined as Rule based systems and shape grammar - Volume 2 - eCAADe 39 | 305 Figure 3 Rule 1 trasnforms TAN into TBN and Rule 2 divides TAN/TBN into plots. Rule 4a and Rule 4b, shown in Figure 4. Since the TAB- south orientation, constrained by the depth of the SPS are wider than other types, it enables a Siheyuan neighbourhood, is 3 or 4 in most cases. For simpli- to be built on it and contains two courtyards in the fication, we assume it is 4 in TAN and 3 in TBN. The east-west orientation. Normally, on a TABSP, a set of number in the east-west orientation, constrained by courtyards are aligned in a south-north orientation, the width of the sub-plot, is 2 or 3 mostly. We define it with a side courtyard next to these courtyards in par- as 3 if the width is bigger than 20 metres, otherwise allel and connecting the south and north boundaries. 2. Rule 5 is defined as the disaggregation of TABSP Sub-plot disaggregation and recombination.