Using Cellular Automata to Challenge

Cookie-Cutter Architecture

 

 

 

Abstract

While economically efficient and easy to plan, over-simplistic residential “cookie-cutter” architecture – ubiquitous in high-density urban environments - results in obvious deficiencies for inhabitants including lack of individual adaptability in the immediate living environment as well as of larger-scale urban identity. This paper investigates into the possibility of applying cellular automata systems to the planning of high-density architecture. Following a review of previous approaches to cellular automata systems in architectural design, I aim at examining more concrete strategies of mapping dynamic processes in cellular automata to residential building morphology in high-density architecture. I discuss the possible use of various forms of cellular automata to generate emergent patterns based on local activity and identity within larger networks and the mapping of automata to aspects of designing and building architectural form. Because of her high living density, her extensive use of cookie-cutter architecture and her richness in emergent illegal building patterns, I use the city of Hong Kong as a case study for this investigation. The paper concludes by outlining possible strategies for developing cellular-automata-based alternatives to the pervasive monotony of commonplace high-density cookie-cutter architecture.

1. Introduction

High-density residential architecture accompanies the development of high-density living environments where land for development is scarce and where large numbers of inhabitants have to be accommodated. With urban population densities growing rapidly, the provision of housing is constrained by efficiency of production and cost, with extensive regulatory prescription to guide designers to desired results[TK1] . Symptomatic for architectural solutions to this kind of need is large-scale planning that typically uses repetition with little variation – resulting in “cookie-cutter” architecture. The results are monotony and lack of identity in the built environment on different levels of scale. As a possible architectural design approach to this challenge, this paper outlines the potential of applying Cellular Automata (CA) as a generative design tool that aims at creating variance and identity while attaining performance goals within a high-rise residential building framework.

In the context of increasing density of urban development, generative design approaches are becoming potentially more relevant. Generative design proposes ways to create and structure design output with the help of design tools or systems instead of the continuous intervention of a human designer. In generative design strategies, problems are typically approached through the development of a generative design tool, which is used to generate solutions for a given design task. Generative design tools are often variations of techniques developed in sciences other than architecture – e.g. mathematics, physics and biology. CA have played a more prominent role in fields of study other than architecture: in Mathematics, CA have been studied as an alternative to differential equations, in Physics CA are used to model gas behaviour and in Urban Planning, CA are employed in the simulation of urban development (see for instance [13]). In the field of architectural design, however, CA have not yet been studied exhaustively – I outline key examples of CA related architectural research below. High-density residential architecture and CA can both be described as expressions of logic in space (see [5], p.45) determined by the massively parallel dynamics and interactions of individual units or cells, and grid structures that facilitate their interaction and provide general framework functions. These parallels can be taken as a starting point to apply CA to various aspects of planning in mass housing architecture, for which I propose Hong Kong as a high-density urban test bed. This paper positions further research in the topic and is intended to lay out a framework for more detailed studies.

2. Approaches to Cellular Automata in Architectural Design

Stanislaw Ulam, who is said to have initiated scientific interest in “cellular automata games” before the availability of suitable computational facilities, has used physical models for his early investigations into the field [1]. Otherwise, mathematical CA research was more interested in computational outcomes and the possibility to produce “chaotic” structures from simple rules and less interested in form finding and physical manifestations. Later developments, such as Conway’s Game of Life and various subsequent CA applications in the Artificial Intelligence field, did not change this until Frazer [5] began to pioneer CA in the field of architecture and developed various physical digital cellular automata models to visualize dynamic architectural processes.

Figure 1: Reconstruction of a three-dimensional CA system designed by Schrandt and Ulam

 

The Generator Project, designed by Price (see [10], pp. 84-89) in 1976, was an early attempt to bring architecture closer to the spatial and logical versatility of a CA system. The design intention behind the project was to enable buildings to perform as catalysts and facilitators to dynamically encourage social and spatial interaction. The Generator Project consisted of a set of volumetric units, screens and barriers that could be freely arranged on a grid with the help of a mobile crane. Each unit was to be fitted with a logic circuit that would communicate with a central computer, which would keep track of configuration changes and even suggest possible alternatives.  Never having been built, the project nonetheless gives an excellent example of a design that relies on dynamic reconfiguration to suit changes in user needs as well as on digital processing of spatial logic. However, though each unit contained a logic circuit, they would still depend on central computer control.

More recent discussions of CA applications to pattern-formation in architecture can be found in Coates [3] and Krawcyk [8]. Whereas Coates emphasises the role of CA as a tool for exploring function-driven generation of architectural morphology, Krawcyk focuses on the spatialisation of CA-generated volumetric clusters. Fischer et al. [4] describe applications of cellular automata to model morphogenetic and developmental form finding processes in architecture and construction. These examples indicate the potential versatility of CA in the architectural field. The possibility of structural design with cellular automata however has not yet been extensively explored despite existing related approaches in the engineering field, as in Finite Element Analysis (for example see [12]).

3. High Density Residential Buildings in Hong Kong

Accommodating some of the highest population densities in the world, Hong Kong houses a variety of exemplary residential building types, which demonstrate the problems of high-density architectural design. While elements of some building types are partly based on local conditions such as land ownership patterns, plot sizes and building regulations, they also demonstrate problems that other cities with rapidly increasing population densities might also encounter in the future. Rapid urban development in the future will not only have to deal with problems of density and identity. It also requires the development of highly time-efficient and competitive construction and design methods. China’s plan to urbanise the populations of twenty counties per year for the coming two decades [2] for example illustrates the scale of future requirements. Increasing project sizes often result in increased architectural monotony, as large-scale, post-war modernist housing developments in Europe have shown. The Hong Kong case study demonstrates how similar building types also exhibit a lack of variation on the individual residential unit level. In this section I outline the characteristics of what I believe to be the most relevant types of high-density housing in Hong Kong.

 

The monolithic block is an early example of mass housing, with an overall building form mostly restricted to a rectangular box shape for space efficiency, structural and construction reasons. Buildings of this type, shown on the left of figure 2, were mainly built during the 1970s and early 1980s, both by private investors and the Hong Kong Government and are typically inhibited by roughly 1500 inhabitants each. In private developments, it occurs generally as a singular building, with repetition limited to the level of the individual unit. In larger government housing projects, estates can contain as many as a dozen blocks. While the simple form provides an efficient use of the land available for construction, it also reduces building surface areas that provide inhabitants with sunlight and air. A monolithic block’s building facades usually do not show any architectural articulation apart from occasional crude ornaments. Inhabitants, however, very often change the building façade adding plants, individual clothes drying racks, commercial signage or entire balconies and room extensions. This phenomenon has lessened since building regulations have been enforced more strictly during the past few years.

 

 

Figure 2: Monolithic block and cookie-cutter high-rise buildings in Hong Kong

 

More recent examples of high-rise residential buildings in Hong Kong are based on plans that are geometrically less simplistic. This is mostly due to recently introduced building regulations that require every bathroom and kitchen to have direct external ventilation. This more recent type of building, shown on the right of figure 2, is just as rigid as the monolithic block, but significantly taller: the plan is extruded to up to sixty floors without much variation. Economically efficient for the developers and/or government, this type of building is rarely built as a single instance – in most cases, it is copied several times to form an urban unit of a larger scale. Several self same high-rise buildings often share one common podium that occupies an entire urban block and houses commercial uses such as shopping malls and restaurants as well as car parking spaces. With the need for residential quarters ever increasing, the Hong Kong Government has focused on the development of new towns – large-scale developments extending smaller towns outside of the inner city. New towns typically consist of a small old village centre surrounded by massive residential identical high-rise buildings. As a result of this development, newly added city blocks do neither connect well to the existing urban fabric nor exhibit any local identity. New developments all over Hong Kong are disconcertingly similar. The recent public housing design competition (in fact the only one in the history of Hong Kong to date) in 2001 for Shui Chuen O has acknowledged this deficit in declaring diversity and identity of character as one of the central aims of the competition, next to improved building sustainability [9].

 

 

In older urban areas, where land ownership is constricted to small lots, a further building type has developed out of the need to make use of very expensive land: the pencil tower. Buildings of this type are developed in older urban districts, on areas as small as 10m x 20m, with a podium of two to four floors for commercial use and a high-rise tower of twenty to thirty floors on top. The residential pencil tower floor plan typically consists of only one or two apartments, and its outer shape does not vary on different floors. Even though pencil towers are high-rise buildings, their small footprint has little effect on the urban pattern at the ground level. Given the highly monotonous and extremely small unit footprints – typically inhabited by families with 4-6 members, inhabitants of large-scale residential architecture frequently customise their individual living environments. Since changes or additions to buildings contravene building regulations, they are generally referred to as illegal structures [11]. The individualised changes of apartment interiors and building facades are motivated by a need for additional living space as well as the desire for a garden or simply the wish to convert the anonymous unit into an individualised home. Inhabitant-initiated changes occur only slowly, as the inhabitants’ needs and preferences change over time and give rise to building activities to adjust the living environment accordingly. Thus, illegal structures as those shown in figure 3 most often appear on the facades of older buildings. While illegal structures often alter the appearance of monolithic buildings and inside living conditions to the better, they are being targeted by recently enforced government policies. Governmental mass-media campaigns encourage owners of illegal structures to remove them. The elimination of illegal structures ensures safety and proper building maintenance, but it also removes the occupants’ ability to live in a personally adapted living environment.

 

 

Figure 3: Balcony conversions and illegal structure in Hong Kong

 

The factors determining the shape of residential architecture in Hong Kong are mostly related to aspects of regulation, economy and efficiency. Since much of Hong Kong’s housing development has been in reaction to the rapidly increasing population during the previous decades, it is argued there has been little room for alternative developments in residential architecture. Unplanned building activities have been a significant feature of urban architecture in Hong Kong, for example, but high-density residential architecture has neither sought to address the problem of urban monotony nor utilised the potential dynamics resulting from the individual inhabitants’ activities. With improving standards and increasing expectations of individual inhabitants, though, one key design challenge of the future will be to tackle the problems of efficient, high-density housing while at the same time providing high-quality, diverse living environments.

4. CA and Architecture

Numerous architectural proposals for the design of large building projects have diverged from the conventional approach to design homogeneous, monolithic monuments. During the 1960s, the fascination of architects with the aesthetic appeal of fine-grained conglomerates resulted in design proposals that featured cell-like units in large numbers – as in the extensively published projects of Archigram or the Metabolists. Many of these projects were concerned with residential architecture, reacting against the large-scale development of mass-produced identical residential buildings built all over Europe and North America after the Second World War. Cell-type residential structures were assumed to provide the inhabitant with greater flexibility in choosing a living environment and relocating it according to changing needs. Residential cells were conceived as elements within a larger structural framework that provided structural support as well as infrastructure, often extending far beyond the scope of individual buildings. Most of these projects are well known and have significantly influenced architectural theory, but they have nevertheless remained at proposal stages. A rare built example is habitat 67, a residential complex designed by Safdie in Montreal for the 1967 Expo. Habitat 67 consists of cell-like, prefabricated concrete residential units that are stacked on top of each other in an irregular way, leaning against a supporting concrete structure. Since residential units can stretch across several concrete cells, unit plans are varied throughout the building and mixed with gardens and verandas on the roofs of lower units. While the cells of habitat 67 are immobile, the building nonetheless manages to provide a diversity of residential units, individual living conditions and neighbourhoods.

The influential architectural design paradigm proposed by Habraken [6] in the early 1960s relates to both efficiency and inhabitant-oriented diversity in residential architecture. His open building approach divides buildings into support (structure) and infill (interior fittings and removable parts). Since infill elements are more related to the immediate way of inhabiting a building, Habraken suggested that these parts of the building were to be changed and improved by the inhabitants over time, while the support elements are designed to last for a long time without interfering with the changing infill. This type of structure allows more dynamic buildings that adapt to their inhabitants and give them opportunities to bring in their own personal preferences. The open building approach is only slowly gaining momentum, and it is remains widely ignored in the context of high-density architecture.

Figure 4: The vertical village project, elevation and model

 

As an alternative to current residential architecture in Hong Kong, I have proposed a ‘vertical village’: a high-rise structure that provides a rigid frame structure with completely flexible internal spaces and external additions (see figure 4). The building is seen not only as a stack of residential units, but also as a community similar to that of a village. Flexible unit sizes, removable floors and walls as well as a reconfigurable service infrastructure provide the basis for spatial diversity within the building. On the exterior, inhabitants can change the façade of their flat or add semi pre-fabricated extension modules to their living space using a crane that forms an integral element of the constantly changing building. Within the overall structure, a network of small community spaces provides the platform for local neighbourhoods, enabling communication between inhabitants. The ‘vertical village’ relates to many aspects of architecture that CA can be applied to in a generative design approach such as small-scale variance replaces top-down planned repetition, building form is seen as a process over time, and community spaces are an outcome of neighbourhood negotiation.

While the need for individual character and local expression of identity seems to become an issue in the context of high-density residential developments in Hong Kong, current architectural design practice is primarily concerned with efficiency. For example, the organizers of the Shui Chuen O public housing competition, who initially asked for individual character in residential developments, awarded the first prize to the entry with the most efficient and economic proposal, not the greatest variation and individualisation. One reason for the monotony of current residential high-rise design is that designing for variety usually requires increased design effort of the human designer as well as increased complexity of the design that is harder to manage and might easily cause complications during construction. Alternatives to simple repetitiveness can be developed by addressing individual unit layouts, three-dimensional building form as well as structure and services. While each aspect can be considered in separate generative design scenarios, their interactions can also be addressed. Such an approach can lead to greater variation while conforming to multiple performance criteria. The relationship of structure and inhabitable space in residential buildings, for example, typically results in one element dominating over the other due to sequential human planning. In an integrated generative design tool, however, structural and spatial aspects can be linked together with mutual feedback relationships, adjusting one to the other during the development of the building. Such simultaneous developments can be implemented in CA systems and interconnections made between a variety of scales, enabling integrated growth of structures similar to that observed in natural growth processes. Researchers in the generative computer-aided design field have recently begun to aim at integrating building performance apects with generative computation (see Kolarevic [7]). The research described below addresses in particular this issue of integrating and complying with multiple performance criteria within one generative process. This more organic view on residential building design results in architectural design that is less prone to the monotonous homogeneity resulting from rigid structural frameworks. It also allows to see large-scale, high-density residential buildings not as monuments, but as dynamic systems that harbour sub-systems on smaller scales while still maintaining an overall coherence. Diversity in high-density residential buildings should not result in an unrelated conglomerate of individual units, but rather in architecture that takes individual differences into account while supplying efficient infrastructure and a sense of overall identity.

Figure 5: Three-dimensionally mapped game of life variations

 

5. Applying Cellular Automata to High Density Architecture

In high-density residential buildings, variety can be generated either by the dynamics of the inhabitation process over time based on a homogeneous initial condition, or it can be a characteristic of the initial architectural design. Both approaches can certainly also be combined. In both cases, decentralised development of a large number of parallel units is essential – for simulation as well as for building design purposes. As with CA systems used in other sciences, architectural applications can and should be adjusted to the problem they focus on, which means they can be based on non-homogeneous, non-linear or even mobile cell systems. Furthermore, individual cells can be enabled not only to determine their states within their neighbourhoods, but also to initiate state changes in other cells. This relates closely to the dynamics commonly observed in the growth of illegal building structures in Hong Kong: while global rule sets (regulations or the feasibility of a particular type of addition) define the overall development conditions, inhabitants initiate building activities, influencing surrounding residential neighbourhoods.

To address the challenges of designing high-density residential architecture, two areas appear to be of particular interest. Firstly, the issue of monotony in plan and three-dimensional configuration, and secondly, the problem of designing inhabitable spaces, structure and services in an integrated way. Apart from potentially supporting the automation the design process to some extent, generative design techniques are well suited to develop variety at scales of high detail resolution that might otherwise be neglected by human designers for pragmatic reasons. The simplicity of CA systems might provide a handle to complex patterns arising from the interaction of large numbers of elements needed to describe the diversity potentially contained within large-scale residential building. The simplest means to achieve variety in high-rise architecture is to consider only the building exterior: building facades can be designed to allow for changes on the individual unit level. If this process is governed by a set of simple rules, it closely resembles classic two-dimensional CA systems like the Game of Life. While ornamental patterns have been used on facades before to alleviate the austere expression of highly monotonous building exteriors (see left on figure 6), I also suggest the integration of functional and structural aspects as well as to address the human desire for variance. In many Hong Kong buildings, similar processes – balcony conversions and added cantilevered structures - have already resulted in generating individualised building facades (see left on figure 6).

Since high density residential buildings usually come in clusters, massing studies of building forms within larger urban frameworks need to be considered as well. Seen as volumetric representations, generative CA systems can be useful in spatial composition tasks on the building massing level as well as in generating unique floor plans on every floor. As typical high-density residential projects often house as many as thousands of inhabitants, decentralised zoning within a building is required. Where buildings become mega-structures, taking on the role of city quarters rather than traditional homes, urban-scale zoning for housing people of similar needs becomes important: clusters of special-needs residences, for example, need to be located within a specific neighbourhood, allowing shared use of facilities. The Ulam model shown in figure 1 gives a simple demonstration of how CA can be used to develop consistent density while simultaneously differentiating building mass into connected zones of use.

Figure 6: Façade ornament (Bunka Gakuen Campus in Tokyo, Japan) and illegal balcony conversions (Wan Chai district, Hong Kong)

 

I intend to investigate the potential of CA in developing structural and spatial aspects together in decentralised, massively parallel systems of cells using cellular rule sets that encompass various aspects of overall architectural design considerations. In a similar way, infrastructure and services within large buildings can be developed simultaneously with building massing and internal organisation. This process has analogies to growth in Nature where cells develop together with support networks that provide the necessary nutrients for growth. Natural processes have a fundamental bottom-up logic, which inspire the use of CA in generative design of residential structures. Structures like the support skeletons in open housing designs could be modelled to perform in ways that are similar to the supporting and connecting functions of extracellular matrices in biology. CA also have the potential to complement traditional top-down design strategies with localised, decentralised bottom-up design generation, which I intend to explore through the development of interactive experimental generative tools.

 

4. Conclusions

Up to now, generative, cellular-automata based design strategies have primarily been applied to ornamental design aspects, producing visual pattern variations. They have rarely addressed basic building design issues, common problems of high-density architecture or applied functional or integral structural aspects. I propose the use of CA as a design platform for generative design processes using large numbers of massively parallel cells, facilitating the design of structures that are characterised by local variation as well as coherence within a larger context. Apart from two-dimensional variations on building facades, CA units can represent volumetric or relational aspects of individual units within buildings. Generative architectural design using CA can also enable the emergence of decentralised bottom-up processes within top-down design frameworks, facilitating processes that focus on local neighbourhood relationships within predefined frameworks. They are applicable to a broad range of architectural design scales from interior design issues to the layout of individual residential units up to the planning of urban structures. In my further research I plan to focus on the integration of structural, spatial and functional aspects as well as the design of problem-specific cellular automata, hoping to develop alternatives for overly pragmatic mass housing designs.

Acknowledgements

I gratefully acknowledge the kind support of Prof. Kvan of the Hong Kong University. I also acknowledge Sheila Cooke’s friendly permission to reproduce her photograph on the right-hand side of figure 6, and Catty Chan Fung Yin’s friendly permission for reproducing her photograph of an illegal building structure in Yau Ma Tei, Hong Kong in figure 3.

References

[1]          Burks, A. W. (ed.): Essays on Cellular Automata. University of Illinois Press, Chicago 1970.

[2]          Cering, D.: Wèi lái èr shí nián wǒ guó huì zēng duō shăo xīn shì zhèn? Outlook Weekly No. 3, August 14th 2000.

[3]          Coates, P., Healy, N., Lamb, C. and Voon, W.L.: The Use of Cellular Automata to Explore Bottom-Up Architectonic Rules. Paper presented at Eurographics UK Chapter 14th Annual Conference held at Imperial College London/UK, 1996.

[4]          Fischer, T., Burry, M. and Frazer, J.: How to Plant a Subway System. In: Chiu, Mao-Lin et al. (eds.): Digital Design - Research and Practice. The Proceedings of The Tenth International Conference on Computer Aided Architectural Design Futures. Kluwer Academic Publishers, Dordrecht, Boston and London 2003, pp. 403-412.

[5]          Frazer, J.: An Evolutionary Architecture. Architectural Association, London 1995.

[6]          Habraken, N. J. et al.: Variations: the systematic design of supports. Cambridge, Mass.: Laboratory of Architecture and Planning at MIT 1976.

[7]          Kolarevic, B.: Computing the Performative in Architecture, In: Dokonal, Wolfgang and Hirschberg, Urs (eds.): Digital Design. Proceedings of the 21st International eCAADe Eonference, Graz University of Technology, Austria, 2003, pp. 457-464

[8]          Krawcyk, R.: Architectural Interpretation of Cellular Automata. In: Soddu, C. (ed.): The 5th International Conference on Generative Art 2002.Generative Design Lab, DiAP, Politectnico di Milano University, Italy 2002, pp. 7.1-7.8.

[9]          Shui Chuen O Architectural Design Competition. Published by the Hong Kong Housing Authority, Hong Kong 2001.

[10]      Spiller, N. (ed.): Cyber Reader – Critical writings for the digital era. Phaidon Press Limited, London and New York 2002, pp. 84-89.

[11]      Wojtowicz, J.: Illegal Facades. Privately published, Hong Kong 1984.

[12]      Xie, M. M. and Steven, G. P.: Evolutionary Structural Optimization. Springer, London 1997.

[13]      Yeh, A.G.O. and Li, X.: Urban Simulation Using Neural Networks and Cellular Automata for Land Use Planning. Symposium on Geospatial Theory, Processing and Applications, Ottawa, Canada 2002.


 [TK1]The problem in HK is that we have prescriptive controls instead of performance controls. This latter is where CA has great potential.