Sie sind auf Seite 1von 8

Using Shape Grammars and Computational Design Synthesis Systems to Define and Create in Styles.

Abstract
Inspired by the Design Synthesis and Shape Generation project at the Universities of Leeds and Strathclyde as well as the Open University, this paper looks at the process of defining shapes and styles and more formally the practice of using shape grammars as a means of communicating and also developing designs. Communicating a design or style to another designer or client is a difficult process with a lot of room for miscomprehension and misinterpretation. This paper looks at using shape grammars as a process of reducing these errors in understanding and also as a process of creating new designs, to a style, which may not have been seen using traditional methods. Using these methods the paper presents a set of designs as evidence for the successful implementation of shape grammars in describing and creating products within a style. The conclusion weighs up the benefits of using shape grammars against the costs of having to learn the methodology as well as the time and costs of developing computational systems as a means of using shape grammars effectively in the modern day, suggesting that an early educational approach may be more effective at this time, while programs and computational abilities are so far behind actual needs.

1. Introduction
The generation of new styles and patterns of shapes is something that is perhaps overlooked as a normal, everyday creative process. It is seen by many as natural creativity and then not given much thought thereafter. A firm will have a style and the designers will develop patterns and shapes in this style, maybe then create a new style, but how are these styles defined and how does a designer then create products consistent to this style? The aim of this paper is to look into three key areas of style development each developing from the last. To start with it will look at how a designer goes about creating products consistent with a style as mentioned before. If a firm is producing crockery for a large department store, John Lewis for example, it will need to be constantly updating its range with new products and therefore it will need to be making a variety of products that fit within the style set by the brand or that brands particular range at the time. This concept of creating a variety of designs within a style is where the problem is. How can a style be defined and communicated to clients and/or other designers? If the firm only employs one designer and this designer has developed and set the style and does all of the work then perhaps there is no issue here, but that is a very rare circumstance. Even still, how does this designer go about recreating something of the same style and same design as a previous work? It is not easy to look at the final design and then write down how exactly they got to it in a form another designer will be able to read and understand. The use of shape grammars is a method of developing and defining a style which someone else can then use or read, and this will be explained and tested in this paper. This application of shape grammars leads on to how they can be and are already implemented within design today, with the major question being, how can computational systems help in the generation and development of products in a given style? This paper is split into two major sections. The first, a literature review, follows this introduction and this looks at all of the key areas of style design, examining shapes and styles, then looking in detail at shape grammars and computational design synthesis systems with two relevant case studies detailed to contextualise the content to. This literature review gives a rich source of research theory to then relate to the rest of the paper. The second section is concerned with the implementation and application of shape grammars and computational design synthesis systems to a real style development problem. This includes a method of what was done, the results of this and then a discussion of the outcomes. Having included some firsthand and contextualised this to the research theory the paper then sets out a range of conclusions to the problems and research questions set out in this introductory piece. This firsthand research is used to test the theory and give rise to any issues found in it. 1

2. Literature Review
2.1 Background: Shapes are very difficult to define in one easy sentence, and that ambiguity is what makes this topic so interesting. Put simply, a shape is the outer form of something, composed of lines, points and planes (Stiny 2006). It can then essentially be anything from something drawn on a piece of paper to a shape found in a 3D object or even say a building. Shapes are visual and therefore are cognitively computed meaning the designer cannot do your seeing for you and that you may not see exactly what they saw or how they went about designing the shape (Stiny 2006). This is where the notion of rules was realised as a way to show others, as well as yourself, how a shape was developed. This can then allow another designer to reproduce a shape just as was first produced or at least a shape to a style as set before, and it is these styles that this paper is looking at. Seeing-moving-seeing (Schn and Wiggins 1992) is a novel but evident approach to looking at the act of designing. The process is carried out by trialling through reflective conversation with the materials at hand. An original shape is made and then seen by the designer. The designer will then discover a new site to draw in relation to the shape they have seen and again sees the outcome, and so on. The moving to a new site allows a new shape to be drawn. The work of Schn and Wiggins concludes that designing is more than conceptual as a process, and that there is an almost continual progression in a designers mind with discoveries in features and relations which inform, direct and kindle further designing, be it within the same design problem or an entirely different one. In this way a designer may develop their own style or maybe a better understanding of say a brands style that they are working for. 2.2 Styles So how do you define a style? Firstly the type of style must be defined. A style could be a trend or movement, anything from Neo-Classical to Retro-Futurism for example, or it could be something founded by an individual over time and influenced by their surroundings. It could of course also be that of a brand as previously mentioned. In general terms a style can be recognised as designs having similar features, created by a single person, a collective, across geographical areas, or over time (Chan 2000). Away from the study of forms and syntax others suggest a style is just a way of doing things and that a pattern of the same ways of doing things defines a style (Sparshott 1965, cited in Chan 2001 p.278), but this seems non quantitative and therefore is difficult to apply, especially in the development of computational systems for design. Chan (2001) suggests that understanding the process of creating a style improves the chances of changing the style of an individual, diversifying their design aptitude. This is something that may happen intuitively and almost subconsciously or may be taught. This is something for consideration when developing computational design aids. 2.3 Shape Grammars Shape grammars are essentially rules or definitions for shapes which visually describe the form and structure a design. Before these can be developed and implemented a style definition must be settled upon. This is why understanding the meaning of style is so essential. A style definition is a simple description of a style, an example of which is found in the results section (Figure: 4.1 and 4.4), and it must fully restrict the design to the intended style when interpreted by somebody else. A grammar will then take on this definition as rules. A grammar is intended to have a limited number of 4 2 rules that can generate an unlimited number of designs. The resulting 3 language created is then the set of designs produced (Stiny 2006 p.18). Figure: 1.1 - adapted from Stiny (2006 p.10). 2

The concept of using rules to create a design is nothing new. Imagine the join the dots game from childhood. The dots are numbered and therefore the shape is predefined and if followed correctly would not create any surprise pictures (Stiny 2006 p10). This is shown in Figure: 1.1 beside this text. By removing the numbers that are defining the design there is a lot more room for creativity and, to an extent, generative shape exploration (Figure: 1.1). It allows you to use more lines and join them anywhere, but again still restricts the design to a set number of shapes, but more than previously as shown in Figure: 1.1. The concept of shape grammars allows a design to take on many more facets, and a simple shape rule could lead to a rumpus of shapes depending on how you apply a rule, creating different spatial relations between shapes to then build upon; the generative potential of a rule (March 1996). Explaining shape grammars is best achieved visually. Below is a worked example of a shape grammar taken from Formal Engineering Design Synthesis (Antonsson and Cagan 2001) but a chapter actually written by George Stiny. This example shows the simplest form of shape rule where there is a shape A and B and the rule is simply AB. The shapes next to each other determine the rule and are separated by an arrow showing the direction of the rule. The cross symbols below the shapes are the registration marks and these are there to fix the relationship between the two shapes. When these two marks register the relationship is clear.

+
Figure: 1.2 Reproduced from Antonsson and Cagan (2001 p.37-40)

In this example then, Figure: 1.2, the rule says if you find shape A, replace it with shape B. This creates the shape shown underneath the rule. This rule is continuous as the new shape now contains a square. Figure: 1.3 shows this continuation for one more application of the rule. The square shape that was found has been filled with shape B and as can be seen this rule will continue until it is too small to complete any longer or until it meets a termination rule, if predetermined (Antonsson and Cagan Chapter 2, 2001). This can be repeated with a 3D shape just as effectively. In the results section of this paper, Figure 4.5 shows a real design problem with shape grammars applied. Rules can do more than simply add shapes. A rule could delete shapes, rotate Figure: 1.3 Reproduced from them, translate them, scale Antonsson and Cagan (2001 p.37-40) them, reflect them or just completely replace them with something else, like a pattern for example. This process is called generative shape exploration (Prats et. al. 2006) and is the route to discovering shapes and patterns which could have previously evaded a designer.

As well as a variation of rules there are also three types of grammar to choose from. These are sub-division, addition and grid based grammars and their use varies between designs dependant on the circumstances. These types of grammar can lead to more complex looking designs. An example of this is the ice-ray lattices (see Figure: 1.4) created by sub-division of grammars where a larger original shape is divided sequentially to create a new more complex design

Figure: 1.4 - Reproduced from March and Stiny (1985 p.50)

within a style. The design is confined by the boundary of the outer original shape so the technique is useful for the design of products (Stiny 1977, cited in March and Stiny 1985 p.50). 2.4 Computational Design Synthesis Systems Prats et al (2006 p.212) describe the process of generative shape exploration in three stages. Interpretation and decomposition, analysis and generation in the structure shape to broaden exploration, and reinterpretation and refinement. It is said that this could provide a structure of the design process it its early stages. This then creates a platform from which to build computational aids for exploring product design problems, but mapping a human process into a computational is an extremely difficult task. The difference between cognitive computing and actual computing are obviously vast. Tasks like sketching, painting and poetry require dense, ambiguous and amorphous symbol systems. These inarticulate and non-discursive thought processes do not fit the precise, rigid, discrete, and unambiguous processes that a computer needs to function correctly. This theory is adapted from Vinod Goels (1995) theory of computational cognition but it carries the same, if not more value, when applied to electronic computation. To create a system that can identify and understand a shape a designer has sketched is not therefore an easy task. A designer will have had a thought process drawing the shape and this is something the computer cannot feasibly re-enact. A computer needs a detailed input of what the shape is that is being defined. Figure: 1.5 shows the basic relationship within the system established by the Design Synthesis and Shape Generation project (2008). The project looks to produce a system that uses computer vision to enable a shape synthesis system to see the subshapes within a design. From this the system is then able to replace these shapes with other shapes according to a shape grammar rule. This creates a system where the designer feeds the system designer shapes which the system can then identify. The system will then apply the defined rules and thereby generate new shapes in the form of shape networks. These generated shapes can then be studied by a designer and worked on, inputting them into the system again afterwards. This method allows a designer to see sub-shapes that would perhaps not been identified before (University of Leeds 2008). Shape grammars here provide the computer with the rules it needs to work.
Figure: 1.5 - Reproduced from Design Synthesis and Shape Generation: A Project Overview. (University of Leeds 2008)

Current programs that exist have begun working towards creating this system, but none fully solve this problem. Teaching a computer every possible shape is not feasible, for the moment anyway, but you can have a computer recognise specific mathematical shapes with restrictive qualities such as dimensionalities. The other way around this is to have the computer see and read the shape on the same premise as optical character recognition works, but this has its disadvantages too (McKay 2013). The methodology and preceding results sections use one such program, SD2, and this will be explained and shown visually within these sections of the report. 2.5 Applications within Product Design 2.5.1 The Language of Coffeemakers (Agarwal and Cagan 1998) Coffeemakers are a fantastic example of how shape grammars can be applied to product design. There is a vast selection of designs on the market, all formed around the same basic functionality. The water is heated, rises as steam and condensates dropping in through the coffee filter. Given this basic functionality it is then the job of a designer to make the product stand out. This may be through other features, visually appealing form, or creating value for money. Generating shapes around these functions is achieved with function labels which maintain the proper function-to-form sequence on a product where the function drives its form. Shape grammars can then be manipulated around these to create a 3D shape by creating a grammar view from the front side and top of the product. Agarwal and Cagan (1998) detail this process resulting in a vast selection of designs made by grammars, some of which turned out very visually similar to designs on the market at the time of publication. The grammars 4

they created had an infinite number of outcomes meaning an infinite number of coffeemaker designs. Patently not all of the designs are feasible and things like materials, colour and finish constrict which designs can be used, but it is interesting to see how many designs can be created using shape rules over natural shape creation. 2.5.2 Speaking the Buick Language (McCormack, Cagan and Vogel 2004) The second example is a study of the brand image and style associated with vehicle manufacturer Buick. This company have built and maintained a distinctive image and style, just as many others have (e.g Coca Cola), and all designs keep this style whilst evolving with time. McCormack, Cagan and Vogel propose that these styles can be encoded into shape grammars which can then be used to develop new styles consistent to the brand. These styles had been developed over time and by different lead designers but had maintained elements throughout. The front end of the car is detailed over time and shape grammars were developed from these old designs to establish the style. These grammars are then trialled to establish new designs in the Buick style. Again, as with the coffeemakers, the grammars produced designs that looked like existing products as well as an infinite number of new designs. This process is one which could be used to explore new designs and new concepts like crossover type vehicles, as well as testing the limits of what is maybe achievable within the brand image and within design itself.

3. Method
The aim of this report was to explore the different ways of communicating and developing a style. This section describes the first hand action research carried out, implementing the theory into a real product design problem to assess the methods and allow a better, contextualised discussion. The study method follows the structure set out in the literature review and was carried out by students at the University of Leeds. The first stage was to create an informal style definition. This is the description by which everything else would be developed and discovered. This was done by simply creating a pattern and then describing it in one or two sentences. These descriptions were then tested and evaluated by another student to see how effective they were at describing the style. Any refinements necessary where then made by the designer, forming a more formal description before the next stage. This step allowed one to see how a style can be communicated to a client or another designer, as well has the basic process of creating a style. The second stage was to translate the style and pattern that had been created into a set of shape grammar rules. The style definition was used to aid this and the shape grammars were then tested and evaluated again to see how effectively they communicate the rules intended. Again any refinements were made at this stage to ensure the grammar would work when applied. The shape grammar and associated rules developed had to be able to develop a range of designs when applied. These designs were then applied to crockery forms to better visualise the outcomes of the development. This allowed one to see how a designer may create a product that is consistent within a style. The final stage was to implement the shape grammar within a computational design synthesis program. This program was SD2, chosen as it had the best capability to produce varying designs for the crockery products being designed for. The grammar was created outside of the program and then moved into the program to be turned about using two rules. This stage finished off the process and assessed firsthand how a computational system could help in the generation and development of products in a given style.

4. Results
Figure: 4.1 - Informal Style Definition A pattern made of circles. The circles are arranged around each corner of a square on in a random fashion. All of the circles are of a different radius but aligned to a reference point (corner) on the grid. Interpretations of the Informal Style Definition.

Figure: 4.2 Authors own interpretation.

Figure: 4.3 - Secondary Designers Interpretation (Padgett, B. 2013. University of Leeds PDES5185 Workshop. Unpublished.)

Figure: 4.4 - Formal Style Definition A pattern made of circles. The circles are arranged around each corner of a square on in a random fashion. All of the circles are of a different radius but aligned to a reference point (corner) on the grid. The excess circle found outside of a bounding square is removed and a new square consists of 4 new corner points. Figure: 4.5 - Shape grammar created by author from definition

Interpretations of the Shape Grammar (Figure: 4.5).

Figure 4.6: Authors own interpretation.

Figure: 4.7 Secondary Designers Interpretation (Padgett, B. 2013. University of Leeds PDES5185 Workshop. Unpublished.)

Implementations of the shape grammar developed.

Figure: 4.8 - (Right) Designs for two square ceramic plates made using computational techniques. Figure: 4.9 - (Below) Shape grammar used to develop tile designs.

5. Discussion of Results
This study gives a realistic account of the challenges in describing visual style and also the difficulties in recreating things in a style. There are many strengths and weaknesses evident in all of the methods evaluated here, and this section explains these, offering some routes around issues encountered. The first stage was to create a description of a visual style using text, the style definition. This is by far the hardest way of communicating a style, primarily because a style is visual. It is extremely difficult to visualise a text description and it is therefore very unlikely that another designer will read a text description and produce something in exactly the style the author of the style had intended. The study here showed that although the author of the design has relayed elements of the design to the next designer, some key details were lost in translation from text to a visual design. The circles were not cut from outside of the squares as the author had not thought about the area outside of the bounding square. The author had assumed that the excess circle was removed whereas the second designer did not. This difficulty in describing a visual style textually is why shape grammars were developed and that was the next stage in this study. The use of a shape grammar did show a better outcome. The second designer did show a very similar design to the authors intentions, with the only real visual difference being a slight simplification of the design where the circles did not cross each other at any point. This may have been that the designer did not see this possibility, which is shown in the grammar, or it is more likely just that the designer was trying the grammar out and that given more opportunity, the shapes and style would have begun to mature and develop. This study has shown the effectiveness of shape grammars in communication visual style, but only to another person that understands shape grammars. A client is unlikely to devote time to learning a new technique of describing styles; they may even believe that it was not a problem in the first place. Shape grammars are fairly simple when you understand them, but getting around the concept and the possibilities can be difficult. Given these two issues, implementation of shape grammars is something of a challenge in itself. A design company could implement the method as the norm, having every employee learn and understand the method and use it, showing a client only the finished designs. Having an advanced computational system which can simplify the process of shape grammars and allow you to undo shapes and changes could be a very useful tool. The results above in Figure: 4.8 were built using the SD2 program, although the shape rules were already set and not altered in the program. This is because the program is difficult to get to grips with and this is the issue with current computational systems; there is a lot more development to be done before they can be on a level with 3D CAD systems and the like. In terms of developing a program it may be beneficial developing an educational program to develop students aptitude for seeing shapes and using a simple shape grammar approach. Schn and Wiggins (1992) do suggest that there should be an emphasis on traditional design techniques which allow the designer to see things in new ways. Developing a program that practically does the designing for you would remove the input of the designer and also limit the designers learning, and therefore overall ability. It is suggested that any system developed should enhance the designer's capacity to capture, store, manipulate, manage and reflect on what he sees and not just remove the need of a designer altogether, something that could be more beneficial to people on an early educational level as a means and way of thinking. 7

6. Conclusions
How do designers create products that are consistent with a style? There is no one way in which designers create products consistent with a style. Every designer has their own process and it differs from person to person. Most designers would probably fit into a loose general category; the seeingmoving-seeing type (Schn and Wiggins 1992) but it is difficult to show this from the studies carried out for this paper and results found. The process of creating a style without the use of a practice like shape grammars is purely cognitive, although through the means of a canvas to hold visualisations and thoughts. This means that it is to an extent immeasurable, but it is possible to communicate a style onto another designer, be it vocally, textually or visually with a method like shape grammars. How can a style be defined and communicated to clients and/or other designers? This paper has looked at the use of shape grammars over traditional methods of communication. It was clear in the results that the communication was much more successful where shape grammars were the language, giving very similar designs at both ends of the conversation. Traditional methods of using drawings and text and having discussions about designs are clearly a common practice and without any study or comparison in results between the two methods it is difficult to state which works more efficiently. Shape grammars are efficient and give clear definition of styles but require an understanding of the shape grammars methodology, which could be time consuming, and it could be that having that undefined element gives rise to more flair and originality within the style. How can computational systems help in the generation and development of products in a given style? The current programs, as smart and brilliant as they are, are rather rudimentary in terms of what is needed and even what was intended as a long term goal for the DSSG project. At present the computational systems do offer a view into a different way of thinking and designing, but do not offer all that much to the generation and development of products in a given style due to their many limitations and difficult user interfaces. In time, and as the software progresses, it will be easier to see how much help these systems could potentially offer. The feelings of the author are that an advanced build of this type of system, possibly combined with a 3D CAA system could wield a super design package with limitless capabilities. For now though it use would be better suited as an educational tool, providing people with a different way of thinking which can be taken away and used elsewhere in design, if only on a latent cognitive level.

7. References:
Agarwal, M., J. Cagan. 1998. A blend of different tastes: the language of coffeemakers. Environment and Planning B. 25, 205226. Antonsson, E.K., J. Cagan. 2001. Formal Engineering Design Synthesis. UK: Cambridge University Press Chan, C.S. 2000. Can style be measured? Design Studies. 21(2000), 277291. Chan, C.S. 2001. An examination of the forces that generate a style. Design Studies. 22(4), 319-346. Goel, V. 1995. Sketches of Thought. USA: MIT Press. March, L., G. Stiny. 1985. Spatial systems in architecture and design: some history and logic. Environment and Planning B: Planning and Design. 12(1), 31-53. March, L. 1996. Rulebound unruliness. Environment and Planning B: Planning and Design. 23(4), 391-399. McCormack J.P., J. Cagan, C.M. Vogel. 2004. Speaking the Buick language: capturing, understanding, and exploring brand identity with shape grammars. Design Studies. 25(1), 1-29. McKay, A. 2013. Computational Design Synthesis workshop, PDES5185M Design Research. University of Leeds, 22 February. Prats, M., C. Earl, S. Garner, I. Jowers. 2006. Shape exploration of designs in a style: towards generation of product designs. Artificial Intelligence for Engineering Design, Analysis and Manufacturing. 20(3), 201-215. Schn, D.A., G. Wiggins. 1992. Kinds of seeing and their functions in designing. Design Studies. 13(2), 135156. Stiny, G. 2006. Shape: talking about seeing and doing. USA: MIT Press. University of Leeds. 2008. Design Synthesis and Shape Generation: A Project Overview [online].[Accessed 10 March 2013]. Available from: http://www.engineering.leeds.ac.uk/dssg/objectives.htm

Das könnte Ihnen auch gefallen