Using Shape Grammars for Master Planning
Jan Halatsch, Antje Kunze, Gerhard Schmitt
Chair for Information Architecture, Switzerland
This paper describes the application of procedural modeling methods to automatically derive 3D models of high visual quality for a highly detailed and quick visualization of complex city models. We discuss the applicability of a procedural
modeling pipeline of shape grammars in urban planning to derive meaningful 3D
city models. Therefore we analyze CGA shape, a novel shape grammar for the
procedural modeling of CG architecture, for its usage in architectural planning
processes. Our system gives rise to three exciting applications in the field of city
modeling and city visualization enabling a quick decision making and iterative
design workflow: modeling of master plans, evaluation of built environments and
planning of urban open spaces. It is capable to produce building shells with high
visual quality and geometric detail. Context sensitive shape rules allow the user to
specify interactions between the entities of the hierarchical shape descriptions.
Selected examples demonstrate solutions to previously noted challenges.
Introduction
A shape grammar for master planning will be important for a) the quick
and detailed visualization of urban designs including buildings, roads, locations and associated vegetation, b) the visualization of planning laws, c)
the use of natural energy resources (solar and wind) and its effects on the
visual appearance of cities. Grammar-based 3D modelers applied to an
urban scale could be of great help for planning of future cities with an extremely reduced energy consumption due to its semantic system. One main
advantage of recent shape grammar systems is the increase of possible design variations, which would result in time-consuming tasks if done with
common planning methods. Manual tasks are difficult to handle in largescale landscape and urban planning projects especially when a large
portion of interdisciplinary expert knowledge is involved. To ensure the reliability and the quality of plans it takes a lot of human resources to manually
J.S. Gero and A.K. Goel (eds.), Design Computing and Cognition ’08,
© Springer Science + Business Media B.V. 2008
655
656
J. Halatsch, A. Kunze, G. Schmitt
model detailed 3D land- and cityscapes. In other industries beyond the
field of architecture next generation applications are already used in that
manner. If integrated in the architects‘ processes they could also dramatically raise design quality through (1) the added predictability of the aimed
urban design, (2) the possibility to simulate certain key measures for the
city‘s planning, like enhanced airflow through skillfully placed city layouts
along with flow optimized building facades, and (3) the opportunity to redesign large drafted urban models within seconds of what before could
have required months of work. Additionally, more sustainable designs can
be achieved if associated attributes would be included into the semantics of
such a system.
This paper explores how shape grammars can be integrated into the urban design process. We introduce a recently published system that can be
used to visualize master planning scenarios in high polygonal resolution.
We contributed to this system the ability to describe urban design patterns
in the form of hierarchical patterns. This system can be applied (1) to the
curriculum in education as a systematic learning tool to understand and
encode architecture, (2) to the field of architectural theory e.g. for the
analysis and synthetic construction of unbuilt city utopias, (3) to the field
of architectural history and archeology for the reconstruction of building
designs and cities [1, 2], (4) to the field of urban planning as well as to project development for the pre-visualization of large urban areas - especially
in cases where common building regulations exist (see Figure 1) and (5) to
evaluate simulation criteria on an urban scale e. g. in order to achieve
lower CO2 emissions within cities. This system had been introduced by
Parish and Müller [3] for the production of city-like structures and computer generated architecture by Müller et al. [2]. We introduce this system
as a novel framework for a grammar-based planning of urban environments. We extended CGA Shape to an urban visualization framework for
architects. Further we show how this shape grammar approach can be integrated into a common design workflow.
Related Work
A landmark in the formal theory of architecture was the introduction of
shape grammars by Stiny and Gips [4]. These shape grammars have been
used to analyze and describe designs as in the following examples, such as
Palladian villas [5], Mughul gardens [6], Frank Lloyd Wright's prairie
houses [7], or Alvaro Siza's houses at Malagueira [8]. However, the applicability of Stinys‘ original shape grammars as a practical modeling tool is
unclear, because it is hardly amenable for computer implementation. May-
Using Shape Grammars for Master Planning
657
all [9] presented an interesting implementation of a shape grammar specified for landscape design. The system produces urban scenes by generating
parcels with simple 3D buildings and by distributing vegetation objects.
Müller et al. [2] presented a more recent approach. They introduced a
novel attributed shape grammar called CGA Shape, which is suited for applications in computer graphics. Important contributions for the use of
shape grammars in urban design had been made by Brown and Johnson
[10] simulating mediaeval city blocks in London as well as by Beirão and
Duarte [11], they produced flexible urban plans with shape grammars. Ulmer et al. [12] introduced an approach to extend CGA Shape in order to
produce large-scale vegetation scenarios for urban environments. Shape
grammar rules of certain architectural configurations have been collected
within libraries. One such library exists in literary form and is described as
a collection of patterns. The use of patterns in architecture is founded on
Alexander et al. [13].
Fig. 1. Left: Stochastically with CGA Shape placed building masses on a given
urban field. Right: Automatic adjustments within the layout following regulations
concerning shape masses [3].
Crowe and Mitchell [14] practically applied Alexander’s patterns to describe specific landscape designs. The corresponding key components of
patterns in landscape design have been specified and illustrated in detail by
Bell [15]. Condensed as illustrated patterns, Turner [16] gives a comprehensive overview of landscape attributes in an urban context. Alexander
[17] describes the formal relationship between landscape design and urban
planning. Concerning the computational design curriculum there are only a
few examples for the use of shape grammars: Flemming [18], Knight [19,
20], Beirão and Duarte [11] and Colakoglu [21].
658
J. Halatsch, A. Kunze, G. Schmitt
Overview
This paper is organized as follows: Section 2 starts with an overview of the
CGA Shape grammar, and afterwards we introduces our novel extensions
needed for the encoding of urban open spaces. Section 3 deals with the
design of vegetation patterns and their implementation by using our grammar. In section 4, we describe three design studies as examples. Our examples show that by using this framework users can easily generate and
visualize large urban scenarios containing high-resolution building and
vegetation geometry. Finally we give conclusions and future work in Section 5.
A Shape Grammar for Master Planning
CGA Shape
In the following, we briefly introduce the main concepts of CGA Shape
but refer the reader to [2] for a more comprehensive description. CGA
shape stands for a novel shape grammar for the procedural modeling of
computer graphics (CG) architecture. CGA Shape is an extension of set
grammars, introduced by Wonka et al. [22]. The notation of the grammar
and general rules to add, scale, translate, and rotate shapes are inspired by
L-systems [23], but are extended for the modeling of architecture. While
parallel grammars like L-systems are suited to capture growth over time, a
sequential application of rules allows for the characterization of structure
i.e. the spatial distribution of features and components [24]. Therefore,
CGA Shape is a sequential grammar (similar to Chomsky grammars). The
CGA Shape framework consists of the shape definition, the production
process, the rule notation with shape operations, and an element repository.
Shape definition: A shape A consists of geometry, a symbol, and attributes. The most important attributes are the position P, three orthogonal
vectors X, Y, and Z, describing a local coordinate system, and a size vector S.
Production process: The production process can start with an arbitrary
configuration of shapes, called the initial shapes, and proceeds as follows:
(1) Select an active shape with symbol A in the set, (2) choose a rule with A
on the left hand side to compute a successor for A resulting in a new set of
shapes B, (3) mark the shape A as inactive and add the shapes B to the configuration and continue with step (1).
Rules: The CGA Shape production rules are defined in the following
form:
Using Shape Grammars for Master Planning
id:
659
predecessor: condition
 successor: prob
where id is a unique identifier for the rule, predecessor is a symbol
identifying a shape that is to be replaced with successor, and condition
is an optional guard (logical expression) that has to evaluate to true in order for the rule to be applied.
Conditional behavior: For example, it can be secured via condition
that the rule is only applied if the predecessor shape is not occluded by
another shape. To make designs stochastic, the rule is selected with probability prob (optional).
2D shape operations: The subdivision split can subdivide objects along
an axis. For example Subdiv(y,7,2,5){ A | B | A } splits the current
scope along the Y-axis in three parts with sizes 7, 2, and 5 and adds three
shapes A, B, and A accordingly. To enable size-independent rules, we also
include the letter r to denote relative values, e.g. Subdiv(y,1r,2,4r) creates three parts: the middle part has absolute size 2 and the remaining two
parts split the remaining space in a 1:4 proportion. We also employ a repeat split, e.g. Repeat(x,2){ A } places as many A shapes of approximate
size 2 as possible along the X-axis of the current shape. The component
split can divide a shape into its geometric components like 2D faces or 1D
edges, e.g. Comp(e){ B } splits the predecessor shape into edge shapes
with symbol B.
3D shape operations: Several different types of shape operations can be
applied to modify the successor shape, e.g. transformations like translating
or scaling.
Object placement: The insert command I(objId) adds, according to the
size of the current shape or according to the featured goal a) object distribution along lines or curves, b) stochastic distribution on surfaces, c) combination of a) and b), an instance of a geometry object with identifier objId (from the element repository).
Hierarchy: Conditional behavior, shape operations and object placement
are nested hierarchically due to the rules processing. This enables different
ordering schemes like the following: city, street-network, block, lot and
footprint.
Design patterns: A typical pattern described by Alexander [13] can be
interpreted as relational function of confirmed geometrical and spatial configurations. In the case of CGA shape patterns can represent a facade
design of an epoch or a given style, which can result in interchangeable terminal symbols. Otherwise patterns can describe relational proportions of
an object on a defined scope. The combination of hierarchical organization
and relative values for geometric configurations enables the introduction of
660
J. Halatsch, A. Kunze, G. Schmitt
design patterns. For example, street types like avenues consisting of street
and edging vegetation can be easily set up as well as parks or other geometric configurations consisting of nested rules or geometry objects [25].
Element repository: The library of 3D models consists mainly of basic
primitives, elementary architectural objects (e.g. ionic capitals) and plant
models. The elements are hierarchically organized in categories and types
and each element has a unique identifier, shader attributes and metadata.
While CGA Shape was presented by Müller et al. [2] the placement of
urban vegetation with shape rules was presented by Ulmer et al. [12]. The
hierarchical linkage of shape rules, the introduction of urban patterns and
the application of the grammar to master planning had been our contribution.
The CityEngine had been previously developed for the production of
city-like structures [3]. With Müller et al. [2] the CityEngine evolved into a
framework, which enables users to reconstruct architectural designs and
archaeological scenarios with shape grammar rules as given inputs. Within
this paper we explore the application of this evolved framework on urban
design.
The City Engine System
The CGA shape grammar had been implemented in C++ and integrated in
the CityEngine framework ([3] and continued by [2]). It can process urban
environments of any size i.e. ranging from a single lot up to a whole city.
The input data is represented in a GIS format and consists of different regions like streets, parcels, building footprints, patios etc and is supported
by importing bitmap data or geographic vector data like Google Earth’s
KML format, which can be used for building mass models. Furthermore,
these regions contain metadata information specified by the user in the
GUI of the CityEngine.
Since the CityEngine represents a 3D modeling application the user relies on many different views of the model: (1) A top view shows an overview on building lots, streets, blocks and building footprints. (2) Within a
three-dimensional view the partial derivation of the grammar can be analyzed on a mass model level. (3) A high detailed geometry visualization
shows the final geometry. (4) A text based editor serves as an input for the
shape grammar. Depending on the metadata attributes, the regions trigger
the selection and application of corresponding shape grammar rules.
Therefore, the region itself (usually a polygon) is fed as initial shape into
the grammar engine, which derives it into an elaborate design by applying
the selected rules. The resulting model can then be previewed in the
Using Shape Grammars for Master Planning
661
OpenGL viewer of the CityEngine or photo-realistically visualized in a 3D
application like Autodesk’s Maya.
Workflow for Master Planning using Shape Grammars
Fig. 2. Modeling urban designs with CGA Shape demands for prerequisites of
available data, which have to be refined for processing with the grammar interpreter.
Enhancing traditional design cycles with shape grammars
Since CGA Shape was originally designed for digital content creation purposes such as needed in archeology and entertainment, the system itself
provides many capabilities of a GIS-like application. The production processes can be constrained on different planning scales like districts, parks,
blocks, lots and the building itself. The placement of buildings, the use mix
and the type of buildings can be defined and controlled by bitmaps, which
can be layered and combined. In order to produce a grammar-based synthesized city model, given urban designs have to be analyzed first and preprocessed including all necessary design features and regulations (see
Figure 2). The preprocessing of design starts with the manual subdivision of a
given design into its functional geometric parts. This input can be for example a simple sketch, drawing or a photography. In the next step relational or fixed geometric parts can be identified and written down in CGA
shape syntax. Depending on the aimed level of detail to be described with
the syntax the user is able to construct repeating elements like windows,
plants, and street signals with a standard 3D modeling application. These
elements represent the terminal symbols of the grammar derivation. Finally
the model can be further redefined.
662
J. Halatsch, A. Kunze, G. Schmitt
Preparation steps - Analyzing given designs
Several manual preprocessing steps are needed before our system can start
with the generation of synthetic urban models. As stated above the system
interprets different planning scales. Therefore the inputs have to be prepared:
On city-scale geographic and socio-statistical measures like development boundaries and projected population densities, age mix of buildings
and/or population must be defined per area as well as use-mix and height
ratios. Further, open space systems need to be aligned with available vegetation objects, green area ratios and green area coverage maps. The definition of street parameters like highways and local street-networks are
needed for the creation and interconnection of districts and blocks.
For the work on a city block-level data relations between a block and its
buildings have to be defined e.g. the definition of the average and maximum size of lots. Further aspects relevant for the block-level are private
and open green spaces as well as mass-model dimensions, standard shape
type descriptions and use descriptions like dwellings, office buildings and
commercials.
Building designs must be analyzed for their spatial relationships of their
elements such as dimensions, facade rulings like floor heights and facade
elements like setbacks, and window geometry.
Encoding design into grammar
Master planning of urban environments operates on distinct major scales.
In order to give planners access to a manageable view in our grammar approach, we use the following region definitions: district zones, block
zones, lot zones, patio zones, and also vegetation regions (Figure 3).
District zones are created on both side of streets. They span over several
blocks if necessary and can be split by ordinating street crossings. They are
defined manually within the preprocessing steps, automatically generated
with given defaults or can be equally imported from other projects. Within
the generation process they do not take part in the creation of buildings nor
do they interact with their footprints.
Block zones are CE block polygons enclosed by streets. They can be
outfitted with vegetation objects, parks, or buildings only or as a combination of buildings and vegetation. Flanking streets can be defined by a
manually drawn street-graph or by an automatic creation process [3].
Using Shape Grammars for Master Planning
663
Fig. 3. Above: Layout of a typical urban environment produced with CGA Shape.
Below: Close-up on detailed urban scene showing block, lot and vegetation zones
[25].
Lot zones come into play when a combination of buildings and vegetation is desired. They are allocated to the regions list after the generation of
buildings and thus hold information about their footprints, providing
means to interact with its housing environment. Lot zones additionally get
building type attributes from the CityEngine framework. Attributes like
commercial, industrial and residential can be used later for a further selec-
664
J. Halatsch, A. Kunze, G. Schmitt
tion of suitable sub-patterns. A lot zone without a street edge is characterized as a patio zone. It serves to fill up the empty space in the center of
blocks, enclosed by buildings. Although patio zones share the possibility
of lot zones to include footprint information, their primary purpose is to
contain vegetation without housing. Areas that span several blocks or
cover big parts of the city and contain elements of vegetation are classified
as vegetation regions. Examples of vegetation regions include forests
within city areas, big parks or fields.
All zones can be controlled, arranged and designed with the 2D shape
operators. Further, the configurations can be saved as geometric design
patterns. Patterns can be divided into sets of attributes, which contain all
necessary textual information to describe urban areas. From a system’s
view the descriptions of each pattern – the grammar rules – can be stored
within a corresponding pattern text file. Each pattern text file can give reference to subordinated patterns if they exist. This permits the easy exchange of certain features like detail levels of the achieved models for
example to provide high resolution polygon models for offline visualization
needs or low resolution primitives for the use in real-time visualization.
Since the description of patterns starts at a high-level major pattern,
which contains a more human readable description, and ends at low-level
grammar descriptions. Our system offers the possibilities to keep things
simple for the non-grammar-expert and to keep things open for users that
want to get deep into the features of the grammar.
Building designs can be easily encoded for a quick visualization. Therefore a design provided by an architect has to be classified into its main
elements and their interrelationships like length and width among them.
Depending on the regularity of a proposed building design, its parts are
developed on a top-down approach. In theory there are no limitations to
produce non-regular geometry with CGA shape grammar. The computation of non-euclidean geometry itself can be processed for example with
the Catmull-Clark algorithm [26]. However it has to be implemented into
our framework.
In this shorthand example we would like to show how easy a design can
be refined into a more detailed model by using the above named tools. The
“Cruciform Skyscraper” (see Figure 4) comprises of regular vertical and
horizontal layouts. The horizontal section reveals four wings (see Figure
5), which can be interpreted as a trice rotation around the building center
done in multiple 90 degrees steps (see Figure 6).
Using Shape Grammars for Master Planning
665
Fig. 4. Drawing from Le Corbusier‘s “Contemporary City”.
Fig. 5. The analysis of the “Cruciform Skyscraper” by Le Corbusier.
The three-dimensional “body” of the building itself can be broken apart
into seven key measures (see Figures 7 and 8). The building‘s mass design
can be easily varied by changing values. It is possible to embed conditional
measures such as building regulations or further artist constrains as design
thresholds. Vertical elements for example the different facade types have
been divided into functional geometric groups, which are nested hierarchically.
Similar to the urban and landscape design - facades are described by 2D
and 3D shape operation tools. Geometric patterns are available as well.
Within the resulting design itself many of its features can be controlled and
distributed by using simple stochastic variations and conditions controlling
for example the look of the different openings on Le Cruiser’s facade of
the “Cruciform Skyscraper”, which can hardly be simulated effectively
with manual methods (see Figure 9).
666
J. Halatsch, A. Kunze, G. Schmitt
Fig. 6. Design can be easily modeled top-down.
Fig. 7. The programmed “Cruciform Skyscraper” visualized as a mass model.
BUILDING_H =
220
BUILDING_W = 100
GROUNDFLOOR_H = 6
WING_W = 16
SPINE_W = 50
TEETH_PROJ = 10
TEETH_DIST = 12
Fig. 8. Key measures for the appearance of the high-rise building.
Using Shape Grammars for Master Planning
667
Fig. 9. Detail of the modeled facade using the component split and stochastic
variation for the placement of the curtains.
Examples
Singapore, Pungol
The district of Pungol in Singapore consists of a large high density housing
area (see Figure 10). This area is aimed to answer the rising demands of
housing and the fast growing population. The district‘s layout and the
building shapes consist of regulated forms. The repetitive geometrical appearance of the facades is an ideal example that can be interpreted by our
grammar (see Figure 11).
Fig. 10. Typical street view of Pungol.
The rules set for this example starts with an occlusion rule, which assures that the open spaces of avenue zones do not get filled with items of
the following productions, e.g. buildings placed onto lots. Next, the district
is subsequently split into sub-regions, each containing a center avenue and
left and right building zones. Within the building zones courtyards can be
668
J. Halatsch, A. Kunze, G. Schmitt
filled with parks if the available width and length permit the placement of
vegetation (see Figures 12 and 13).
Fig. 11. Synthetic reconstruction of Pungol.
Fig. 12. The synthetic urban environment of Pungol produced by our system.
Fig. 13. Detailed view onto housing blocks with courtyards.
Avenue parameters like street width, tree density and tree types are adjusted automatically to the importance of the avenues. The buildings consist of length-dependent facade types. The associated facade rules react on
different sizes: For example a building length up to 20 meters features the
Using Shape Grammars for Master Planning
669
facade types A (Edges), B (standard element), C (balcony). Above 20 meters a type D (middle balcony) is introduced.
Student work
The following two examples feature student works produced within a student course at our Chair for Information Architecture fall 2007.
The first example shows the application and variation of one and the
same facade rule. The students started to analyze a facade design developed by themselves (see Figure 14). After encoding the design into the
shape grammar, the students had been able to apply the rule and its variation as a design pattern onto an urban environment (see Figure 15).
Fig. 14. An original facade designed by students.
Fig. 15. Rendering of a large-scale urban environment with different facade variations based on the original facade.
670
J. Halatsch, A. Kunze, G. Schmitt
The second example poses an artistic approach of creating a futuristic
looking city (see Figures 16 and 17). The students focused on the recursive
production calls of their skyscraper.
Fig. 16. The top view onto the “distorted city”.
Fig. 17. A detail view onto the model.
They used a Manhattan-like layout where they placed their “recursive
tower” with heights starting at 200 meters resulting in a “distorted city”
model. Colors had been applied to interpret and analyze the resulting distribution of stochastic rules done by the system.
Using Shape Grammars for Master Planning
671
Conclusion and Future Work
Automatic and semi-automatic approaches can drastically reduce and ease
complex and time consuming design and modeling tasks in the planning of
urban designs. Therefore we presented a stochastic shape grammar for
master planning as a novel approach for the automatic creation of large
urban scenarios in real or virtual 3D cities. We extended CGA Shape to an
urban visualization framework for architects. Our examples show that by
using this framework users can easily generate and visualize large urban
scenarios containing high-resolution building and vegetation geometry.
We showed that it is possible to implement this approach as a visualization method for design into existing design pipelines. The rule-based distribution and placement of objects like vegetation, buildings and the re-use
of design patterns in urban environments enables the planner to explore
further variations. The used procedural modeling methods are able to
automatically derive 3D models of high visual quality for a highly detailed
and quick visualization of complex city models. The system is also open
for iterative (re-) design purposes. Successful design findings of suitable
configurations can be stored within pattern collections. Also the complexity of procedural modeling became manageable with the help of reusable
patterns, which are not limited to landscapes and landscape objects. CGA
shape descriptions can be saved to a versioning database to encode and to
store design for easy application on virtual city models that had been generated by the system. Beside this, stored design data sets can contain geometry attributes, which feature sustainable design criteria. In upcoming
work we would like to explore how feedback (airflow, shading, movement
of people) from simulation can be integrated to a modeled design. We see
a strong potential in adding simulation based city layout optimization to
our framework. The reusability of design patterns for sustainable urban
design is another important research field. We will evaluate the inheritance
of design components and design patterns with the help of standard methods from computer science like UML - unified modeling language - and
test the applicability of object oriented design to shape grammars. Future
work also includes the generation of non-regular geometry.
We have created a promising shape grammar and applied it to the existing one of Müller et al. [2]. Our approach can be used for pre-visualization,
master planning, guided design variation and for digital content creation
purposes of the entertainment industries. Ongoing projects deal with the
pattern-based procedural design of the Science City ETH’s open spaces in
Zurich, Switzerland, and also with the challenging planning of High Density Housing in Singapore.
672
J. Halatsch, A. Kunze, G. Schmitt
Acknowledgments
We would like to thank Andreas Ulmer, Pascal Müller, Remo Burkhard,
and Martina Sehmi-Luck for their continuing support and for many helpful
discussions as well as the review team and our students Jonas Epper, Pascal Hendrickx, Lukas Treyer and Nicolas Zimmermann.
References
1. Guidi G, Frischer B, et al. (2005) Virtualizing ancient Rome: 3D acquisition
and modeling of a large plaster-of-paris model of imperial Rome. Videometrics VIII, JA Beraldin, SF El-Hakim, A Gruen, JS Walton (eds), SPIE 5665:
119-133
2. Müller P, Wonka P, Haegler S, Ulmer A, Van Gool L (2006) Procedural modeling of buildings. ACM Transactions on Graphics (TOG) 25(3): 614–623
3. Parish Y, Müller P (2001) Procedural modeling of cities. In E Fiume (ed),
Proceedings of ACM SIGGRAPH 2001, ACM Press: 301–308.
4. Stiny G, Gips J (1972) Shape grammars and the generative specification of
painting and sculpture. Information Processing 71: 1460-1465
5. Stiny G, Mitchell WJ (1978) Counting Palladian plans. Environmental and
Planning B 7: 189–198
6. Stiny G, Mitchell WJ (1980) The grammar of paradise: on the generation of
Mughal gardens. Environmental and Planning B 7: 209–226
7. Koning H, Eizenberg J (1981) The language of the prairie: Frank Lloyd
Wrights prairie houses. Environment and Planning B 8: 295–323
8. Duarte J (2001) Customizing mass housing: a discursive grammar for Siza’s
Malagueria houses. Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge
9. Mayall K, Endhall GB (2005) Landscape grammar 1: spatial grammar theory
and landscape planning. Environmental and Planning B 32: 895–920
10. Brown FE, Johnson JH (1984) An interactive computer model of urban development: the rules governing the morphology of mediaeval London. Environment and Planning B 12: 377–400
11. Beirão J, Duarte J (2005) Urban grammars: towards flexible urban design. In
Proceedings of eCAADe Conference 2005: 491–500
12. Ulmer A, Halatsch J, Kunze A, Müller P, Van Gool L (2007) Procedural design of urban open spaces. In Proceedings of eCAADe Conference 2007:
351–358
13. Alexander C, Ishi Kawa S, Silverstein M (1977) A pattern language: towns,
buildings, construction. Oxford University Press, New York
14. Crowe S, Mitchell M (1988) The pattern of landscape. Packard. Chichester
15. Bell S (1999) Landscape: pattern, perception, and process. E & FN Spon,
London
Using Shape Grammars for Master Planning
673
16. Turner T (1996) City as landscape. A post-postmodern view of design and
planning. E & FN Spon, London
17. Alexander C (2002) The nature of order: an essay on the art of building and
the nature of the universe. Center for Environmental Structure, Berkeley, CA
18. Flemming U (1986) The role of shape grammars in the analysis and creation
of designs. John Wiley & Sons: 213–244
19. Knight TW (1991) Designing with grammars. In GN Schmitt (ed), Proceedings of CAADFutures 1991, Vieweg: 33–48
20. Knight T (1999) Shape grammars in education and practice: history and prospects. International Journal of Design Computing 2
21. Colakoglu B (2006) Explorations in teaching design students to think and
produce computationally In Proceedings of eCAADe Conference 2006: 826–
831
22. Wonka P, Wimmer M, Sillion F, Ribarsky W (2003) Instant architecture.
ACM Transactions on Graphics 22(3): 669–677
23. Prusinkiewicz P, Lindenmayer A (1991) The algorithmic beauty of plants.
Springer Verlag
24. Prusinkiewicz P, Mündermann P, Karwowski R, Lane B (2001) The use of
positional information in the modeling of plants. In E Fiume (ed), Proceedings
of ACM SIGGRAPH 2001, ACM Press: 289–300
25. Ulmer A (2005) Modeling and rendering of plants in urban environments.
Master Thesis, Computer Vision Laboratory, ETH Zurich
26. Catmull E, Clark J (1978) Recursively generated B-spline surfaces on arbitrary topological surfaces. Computer-Aided Design 10(6): 350-355