Paul Laseau introduces the concept of graphic language – the idea of designing
with graphic vocabulary and grammar based on symbols, such as shapes, lines ...
14
Chapter Two Related Work
This chapter reviews existing studies of design and design representations, including freehand sketching, 3-D modeling and several computational sketching or modeling programs. The purpose of this chapter is to establish background and underlying principles for this thesis by examining the significance of freehand drawing and 3-D modeling for architectural design and the role of drawing conventions in freehand sketches. The chapter is divided into six sections. Section 2.1 looks at studies of freehand drawings in design. Section 2.2 discusses the use of drawing conventions in freehand sketches. Section 2.3 discusses how freehand drawings support abstraction and ambiguity. Section 2.4 discusses the role of 3-D modeling in architectural design; Section 2.5 reviews relevant researches on computational sketching programs. Finally, Section 2.6 discusses the findings of the related work review.
2.1. The Importance of Freehand Drawings
Drawing plays an important role in architectural design. Architects use drawings to explore ideas and solutions, to examine alternatives and to test design concepts. People use the terms of “diagram”, “sketch” and “schematic drawing” somewhat interchangeably. Here the term “freehand drawing” and “sketches” mainly refer to the drawings designers make during the early phases of design process.
Several studies concerned with the relation between design and freehand drawings have argued that freehand drawings are the crucial means by which designers visualize design ideas.
15 Design researcher B. Lawson, in his book “Design in Mind”, documents interviews with ten well-known architects and discusses their working drawings (Lawson, 1994). All ten architects emphasized the significance of drawing. For example, Santiago Calatrava regarded freehand drawings as a dialogue between drawn elements on paper and internal thoughts in the designer’s mind. Herman Hertzberger argued that drawing is crucial in his design process, and described drawing as a “communication between my brain and my paper”. Lawson concluded unsurprisingly that the act of drawing plays an important role for the designers.
E. Robbins’s book “Why Architects Draw”, also examines freehand drawings made by several renowned professional architects (Robbins, 1994). Robbins argued that drawing serves as a means to carry out the designer’s internal conversation and evaluation about a design.
Louis Kahn saw freehand sketching as a way that architects visualize the design problem, and also as a way that architects explore design solutions for a problem (Kahn, 1931). Kahn regarded freehand drawings as a means of representing the design problem. Kahn mentioned that architects must interact with their drawings in order to interact with the design problem.
Studies of design drawings have also become of interest to cognitive science researchers. These researchers have argued that drawing is important to design because it is an external representation that helps designers solve problems and generate ideas. Larkin and Simon argued that a diagram is a representation created to externalize and visualize problems (Larkin & Simon, 1987). They argued that freehand drawings help designers recognize thoughts, identify problems and seek solutions. In other words, a designer's observations and thoughts about a design problem become more easily accessible through the act of making drawings.
16 2.2. The Use of Drawing Conventions
Designers use drawing to think about architectural concepts and design concerns. Architectural design involves many different design tasks, such as spatial layout, form study, and lighting arrangement. When thinking about different design problems, designers employ different drawing conventions into freehand drawings. Several studies have focused on the use of drawing conventions in freehand sketches.
Architect Daniel Herbert in “Architectural Study Drawings” argued that architects use graphic conventions to communicate with themselves and others (Herbert, 1993). Among the drawings in his book, several drawing conventions can be identified, such as symbols for axial lines, entrances, views, and dimensions. Herbert argued that designers use graphic conventions to simplify the complexity of design and symbolize intangible design concerns.
Fraser and Henmi’s book “Envisioning Architecture” examined the relationships of different architectural drawing types (Fraser & Henmi, 1994). They argued that freehand drawings that different architects use to develop design concepts vary in type and style. However, several drawing conventions can be identified among different architects’ drawings, such as lines indicating architectural objects, or human figures in section drawings. Fraser and Henmi argued that designers visualize intangible design concepts such as circulation, function, and employ symbolic representations in their drawings in order to abstract the complexity of information.
Paul Laseau introduces the concept of graphic language – the idea of designing with graphic vocabulary and grammar based on symbols, such as shapes, lines, and text (Laseau, 1989). He argues that a clearly defined graphic language is important to design thinking as well as communication between designers. A language system helps designers “structure” their thinking with the visualization of design thoughts. Diagramming, based on this graphic language and composed of simple elements, enables
17 designers to clarify order and hierarchy among different design concerns. Among the drawings in his book, text labeled inside closed shapes (ovals or rectangles) can be identified and are mainly used to indicate the functions of different spaces. (Figure 2.1)
Figure 2.1, Text labels inside containing shapes represent different functions of spaces. Source: Laseau, P., 1989
Ellen Yi-Luen Do’s doctoral dissertation reports on several empirical studies that examined the relation between drawing conventions and the design tasks with which they are associated (Do, 1998). The goal of the studies was to determine what is in design drawings that a computer program should recognize and hence support. Do found that designers use certain symbols and configurations of drawn elements when thinking about certain design concerns, and that it is possible to associate designer’s drawing symbols with designer’s intentions. The key findings are briefly described below: •
Designers have preferences for using plans or sections to illustrate certain architectural concepts. For example, floor plans are commonly used for exploring relationships between different functional spaces. Figure 2.2 shows that the plan view is mostly used for illustrating the relationships of functional spaces.
Figure 2.2, Plan view is mostly used for illustrating the relationships of functional spaces. Source: Ellen Do, 1998
18 •
Designers share drawing conventions and use primitives from a limited universe of geometric shapes such as lines, ovals, bubbles, or rectangles and arrange them in highly conventional ways in order to create their drawings. Figure 2.3 shows that primitives (drawing elements) used in architectural drawings include lines, simple geometric shapes and hatching.
Figure 2.3 shows primitives used in drawings include lines, simple geometric shapes and hatching. Source: Ellen Do, 1998
•
Plan drawings concerned with spatial layout usually contain more text labels than section drawings. Functional spaces in a floor-plan drawing are mostly labeled by placing text inside a shape (oval, blob, or rectangle). Figure 2.4 shows that designers use text to label different functional spaces.
Figure 2.4, Designers use text to label different functional spaces. Source: Ellen Do, 1998
2.3. Abstraction & Ambiguity
Gross and Do examined the fundamental characteristics of freehand drawing as background for implementing a computer-based sketching environment, Electronic Cocktail Napkin (Gross, Do, 1996).
19 They argued that a freehand drawing in the early phase of design process indicates not only a designer’s decisions but also the associated degrees of ambiguity, precision, and commitment. They mentioned that graphical abstraction is a common technique in early conceptual design. They argued that symbols enable designers to work with components without specifying their detailed configuration. Likewise, an ambiguous representation is a means to postpone structured development, while still enabling a designer to work with preliminary architectural ideas during conceptual design. Designers may have several alternative interpretations in mind. They draw a formless shape (bubble) that could be interpreted in one of several ways. Further, Gross and Do also mentioned that designers usually need only rough dimensions to decide on a basic layout. A hand-made sketch ensures that a drawing preserves the look and sense of quick and rough thinking.
Freehand drawing supports abstract and ambiguous representations. Abstraction permits postponing detailed measurement and allows structured configurations to be replaced by sophisticated components. Ambiguity permits engaging several alternatives for the selection or identity of an element. These graphic techniques all serve to provide designers explicit visual hints for problems that remain to be resolved, and also reassure the designer that certain decisions remain open and that the design is still flexible.
2.4. The Importance of 3-D Architectural Models
Three-dimensional modeling is another common medium that architects use to explore design solution. 3-D modeling provides designers even stronger evidences for visualizing design concerns than 2-D sketching.
Professor Marguerite Koepke in her book “Model Graphics” introduced some quick and simple methods of building study models for architectural design (Koepke, 1988). Koepke stated that models are the best way to represent 3-D space and are the most easily understood presentation tool. “Anyone can understand something that s/he can see.” (p.2)
20 Koepke argued that models, as study tools, support the designer with “crystallizing” ideas, providing a precise way to assess spatial form and sequential relationships. Friedrich Kurrent in his book “Scale Models: Houses of the 20th Century” shows some 200 scale models that were primarily designed by distinguished 20th century architects (Kurrent, 1999). All the scale models were actually built by his students for design studios directed by Kurrent over about twenty years. Those exercises were intended principally to train students to visualize space in 3-D, which Kurrent thought is the architect’s prime qualification. He argued that 3-D models are invaluable to designers when checking designs and establishing forms, and they also enhance communication between designers and those who are involved in the planning and construction processes. Kurrent also argued that a physical model “speaks a language that can be instantly understood and enables spatial relations to be perceived at a single glance”.
The book “Product Design Models”, by Roberto Lucci and Paolo Orlandini, is an introduction to simple techniques of model-making for industrial and architectural design (Lucci, Orlandini, 1990). Lucci and Orlandini see the model as an important tool for the designers. Model-making is a means of designing by “sketching” in three dimensions, and is an activity deeply integrated into the creative process. Model-making, as a representational tool for communicating, also has educational implications. Modelmaking helps students verify design thoughts, and also helps instructors understand students’ design ideas. However, Lucci and Orlandini argued that the model is not the goal of the design. Therefore, the effort required for building models should be minimal in order to encourage its use as a design tool.
Ervin and Hasbrouck in the book “Landscape Modeling: Digital Techniques for Landscape Visualization” describe their perspective on methods and procedures for visualizing the landscape with digital modeling techniques (Ervin, Hasbrouck, 2001). They regarded model-making as part of a larger decision-making process, in which the models are used for further analysis and discussion of design. They argued that there are
21 “no perfect or neutral representations”; rather there are a number of important decisions to be made in choosing and making a representation for a specific design purpose. All models are imperfect, containing simplifications and abstractions. Designers usually build different types of models in order to explore various design concerns. Choosing and using modeling techniques appropriate to the task is an essential part of designing with models. Digital modeling techniques allow designers to build and use models more efficiently and effectively. Designers can build multi-purpose models that integrate with several options for flexibly switching model appearances or functions to fit any specific design purpose.
2.4.1. Discussion – A Bridge between 2-D and 3-D
Three-dimensional modeling plays an important role in architectural design, as does 2-D freehand drawing. However, 2-D freehand drawing, due to its convenience, is commonly chosen by most architects to start designing. Model-making is a time-consuming task, and most architects do not move to making models until they complete a relatively clear framework for design solutions. However, as discussed in the previous section, models are the best way to represent 3-D space and are the most easily understood presentation tool (Koepke, 1988). The earlier an architect can incorporate 3-D modeling into design process, the more information he can gain for further modifying design solutions. Twodimensional freehand drawing and three-dimensional modeling each has advantages and contributions to an architectural design process. A convenient tool that can bridge the gap between drawing and modeling would be valuable and helpful for designers who are developing a concept and, at the same time, trying to a relatively accurate sense about the space they are working with. Moreover, as Lucci and Orlandini suggest, the effort required for building models should be minimal in order to encourage the use of this design tool. Therefore, a computationally based design tool that combines the advantages of both 2-D and 3-D media could be of significant use to designers.
22 Section 2.5 describes several computer programs that aim to either support freehand sketching or 3-D modeling. Section 2.6 further discusses the findings of related-work reviews.
2.5. Computational Support of Sketching
With the observation that drawing and modeling both play important roles in architectural design, the issue becomes how computational design tools can support freehand sketching, and enable architects to generate 3-D forms easily. This section briefly introduces several studies on computational tools that aim to support freehand sketching. The section is divides into two parts: 1) systems that support sketching; 2) systems that generate 3-D forms by sketching.
2.5.1. Systems that Support Sketching
Takeo Igarashi’s program “Pegasus” is a prototype 2-D drawing application for penbased systems (Igarashi, 2000). Using Pegasus, designers can construct precise geometric diagrams easily with freehand stroke input. Two interaction techniques are introduced: Interactive Beautification and Predictive Drawing. Interactive Beautification beautifies the user’s freehand stroke input by considering possible geometric constraints among drawn segments. Multiple candidates are generated for users to choose from. Predictive Drawing attempts to predict the user’s next drawing operation based on the spatial relationships among existing drawn segments on the screen (Figure 2.5).
Figure 2.5, Interactive Beautification beautifies the user’s freehand stroke (left four). Predictive Drawing predicts the next drawings and shows them as multiple candidates (right four). Source: Igarashi, 2000
23 James Arvo and Kevin Novins introduce a new sketching interface, “Fluid Sketches” that tightly couples recognition and morphing of freehand sketches to create a continuous and immediate beautified form of feedback (Arvo, Novins, 2000). By means of smooth and continual shape transformations, a user’s raw input strokes are continuously morphed into ideal geometric shapes based on geometric constraints among drawn segments. For example, when a user sketches a rough box, even though some segments look like a circular arc, Fluid Sketches will beautify the input strokes and morph toward a rectangular box. In addition, Fluid Sketches also can immediately show the ideal shape corresponding to the hand-drawn strokes. For example, a user can immediately see the entire extent of a circle when only drawing a small portion of it (Figure 2.6).
Figure 2.6, Fluid Sketches beautifies the user’s rough sketch box stroke and morphs toward a rectangular box (left two rows). Fluid Sketches shows the ideal shape corresponding to a small portion of a hand-drawn stroke (right). Source: Arvo, Novins, 2000
J. Landay and B. Myers’s program “SILK” is an interactive tool that allows interface designers to design an interface by directly sketching a mockup scheme using an electronic pad and stylus (Landay & Myers, 1995). The designers are allowed to draw several pre-defined shapes to represent interface components, such as a circle for a radio button, a long rectangular box for a slider box. The designers can also illustrate behaviors by sketching in a storyboard, which specifies how the screen should change in response to end-user actions. The designers then can play the proposed interface to explore the look and behavior of the interface (Figure 2.7). Based on the same sketch-based interface builder concept, the “DENIM” program, by J. Lin, M. Newman, J. Hong, and J. Landay, allows web designers to quickly sketch out proposed pages, create links among them, and interact with the pages using the program’s run mode (Lin, Newman, Hong, Landay,
24 1996). The designers draw to layout each individual page, arrange links among different pages in a storyboard, and can view the site structure from a site map (Figure 2.7).
Figure 2.7, A User sketches an interface scheme in the SILK storyboard and makes a dialog box appear when a button is pressed (left). Source: Landay, Myers, 1995. The DENIM shows the web site structure that is created by sketching (right). Source: Lin, Newman, Hong, Landay, 1996
The “Electronic Cocktail Napkin” (Gross, Do, 1996), by Mark D. Gross and Ellen YiLuen Do, is a computer environment for making hand drawn sketches and diagrams. It aims to support freehand drawing and thinking of the sort that designers often do on the back of an envelope or cocktail napkin for creative design. Beyond the act of drawing, the Cocktail Napkin also aims to support intelligent design critiquing, retrieval of relevant information from related database, and simulation. The Cocktail Napkin can recognize and interpret the elements and configurations of drawings made by designers, and then the system retrieves relevant information. The Cocktail Napkin recognizes multi-stroke glyphs and allows designers to customize the program to recognize personally defined symbols (Figure 2.8).
Figure 2.8, The Cocktail Napkin Drawing Board and Sketchbook. Source: Gross, Do, 1996.
Ellen Yi-Luen Do in her doctoral dissertation proposed a prototype computational drawing environment called “Right-Tool-Right-Time” (Do, 1998). Designers use
25 different symbols in freehand drawings to explore design concepts. She argued that an intelligent computer-aided-design environment should be able to infer designer's intentions from all drawn elements in the drawing and suggest designers with appropriate computational tools based on this inference for the task at hand. For example, a layout bubble diagram may activate other design schemes with similar configurations (Figure 2.9). This research developed an integrated digital sketching environment to support freehand sketching for all early design activities and provide designers the right tools at the right time.
Figure 2.9, A bubble diagram activates a similar design scheme for reference (left). A diagram retrieves relevant information from a database (right). Source: Do, 1998
2.5.2. Systems that Generate 3-D Forms by Sketching
Takeo Igarashi’s program “Teddy” is a sketching interface for quickly and easily generating freeform 3-D models (Igarashi, 2000). Users draw several 2-D freeform strokes on the screen and then the system constructs 3-D polygonal surfaces to generate a 3-D model. The system supports the modeling operation to construct a 3-D polygonal surface from a 2-D silhouette drawn by users (Figure 2.10). The system inflates the region surrounded by the silhouette to make wide areas fat and narrow areas thin in order to generate 3-D shapes.
Figure 2.10, Users construct 3-D models simply by drawing silhouettes. Source: Igarashi, 2000
26 J. Cohen, J. Hughes, and R. Zeleznik introduced “Harold”, an interactive drawing system that allows a user to create 3-D scenes from 2-D input (Cohen, Hughes, Zeleznik, 2000). The interface paradigm in Harold is freehand drawing. All objects in a 3-D scene are created simply by drawing with 2-D input devices, and are collections of planar strokes that are reoriented in a view-dependent way as the camera moves through the world. For example, users draw several curves to indicate a hill on the far background in the scene and draw lines to crate a tree in the front side of the scene. Harold then determines the relative positions of two objects in a perspective view. The users can then change viewpoints and navigate through the scene in 3-D (Figure 2.11). The virtual world created in Harold maintains a hand-drawn appearance as a user navigates through it.
Figure 2.11, The User draws a hill on the background and a tree in the front. Harold then determines the relative positions of two objects in a perspective view (a). The user changes the viewpoint in the scene (b). Source: Cohen, Hughes, Zeleznik, 2000
“Digital Clay” explored the concept of the direct generation of 3-D form from freehand drawing input (Schweikardt, Gross, 1998). Digital Clay derives 3-D geometry from a 2-D projected sketch. The program walks the lines of the sketch to identify convex and concave edges, then assigns 3-D coordinates to vertices, and finally generates a 3-D model that the designer can view and edit. The program can make assumptions about hidden edges and surfaces, or it can leave that for the designer to specify later (Figure 2.12). Several similar projects are based on the same concept of deriving 3-D objects from 2-D projected drawings. I. Grimstead and R. Martin describe a method of reconstruct a Boundary-Representation (B-rep) model from a single hidden-line removed sketch of a 3-D object (Grimstead, Martin, 1995). H. Lipson and M. Shpitalni introduce an Optimization-Based algorithm for reconstructing a 3-D model from a single 2-D edge-
27 vertex graph (Lipson, Shpitalni, 1996). “Quick-Sketch”, introduced by L. Eggli, B. Bruderlin, and G. Elber, also allows users to create 3-D models by making 2-D projected drawings (1995).
Figure 2.12 shows the stages of interpreting a sketch to 3-D model in VRML. Source: Schweikardt, 1998
“Sketch VR”, by Ellen Yi-Luen Do, is a pen-based computing environment that generates 3-D objects in a VRML format (Virtual Reality Modeling Language) by recognizing simple geometric shapes in a 2-D view (Do, 2000). To create an architectural space designers only simply draw lines and circles in a simple “Cocktail Napkin” 2-D sketch to indicate the appearances and locations of walls and columns. Designers draw in 2-D and a 3-D world is created accordingly. Similarly, designers can generate 3-D models of interior by drawing diagrams to indicate furniture placements (Figure 2.13). The 3-D scene generation in Sketch VR includes three levels of complexity: simple extrusion of lines and curves (walls, columns), solid modeling from a given geometric object representation (spheres, cones and boxes), and complex configuration with objects from a graphics library (furniture layout).
Figure 2.13, A sketch floor plan generates a corresponding VRML 3-D scene. Source: Do, 2000
28 2.6. Discussion
As discussed above, designers have preferences for using plans or sections to illustrate certain architectural concepts (Do, 1998). Especially, among various drawing types, floor plans are commonly used for exploring relationships between different functional spaces in 2-D. In addition, architects also apply drawing conventions into their sketches in order to identify different design concerns. Architects label different spaces with different text in the floor-plan sketches in order to identify their functions (Do, 1998). A label can be seen as a representation of the space with a certain spatial configuration. By recognizing designer’s drawing symbols in a sketch floor plan, it is possible to infer designer’s concept about the spatial configuration of the space. Furthermore, it is also possible to build a computer modeling tool that creates 3-D space model based on the recognition of designer’s drawing symbols in sketch floor plans.
The Electronic Cocktail Napkin can recognize and interpret drawn elements and configurations of freehand drawings, and then retrieves relevant information for designers. The Right-Tool-Right-Time computer-aided-design environment can infer designer's intentions by recognizing all drawn elements and identifying the task context in a drawing. Based on this inference, the system then suggests appropriate computational tools for the task at hand. Both systems focus on information retrieval based on the recognition of drawn elements and configurations. However, 3-D modeling based on the recognition of drawing symbols has not been specifically embedded into either system.
On the other hand, Do’s program Sketch VR does generate 3-D objects in a VRML format by recognizing simple geometric shapes in a 2-D view (Ellen Do, 2000). The generation of 3-D objects is accomplished by an object-oriented translation from a 2-D drawn symbol to a 3-D object. Sketch VR conceives every single drawn element that contains either single or multiple strokes, as an object that will be generated in 3-D format. For example, a line will be translated to a wall; a box surrounded by four smaller rectangles will be translated to a dining table with four chairs. The same concept of the
29 object-oriented translation is also carried by both Digital Clay and Teddy. These two systems both translate 2-D freeform envelopes into 3-D polygonal surfaces that enclose and generate a 3-D model. However, when designers draw symbols into a sketch floor plan in order to identify either the functionality or the basic spatial characteristic, designers may think about the space instead of only focusing on any single element. For the early schematic development, designers need a computer modeling tool that allows them to think of the basic characteristics of a space, instead of a single element, and use a symbol to represent the spatial configuration of this space. Then the tool creates a model corresponding to the spatial configuration that the symbol represents.
This chapter provides the background for this thesis. The following two chapters will discuss the computational implementation of the SpaceMaker. The SpaceMaker is a symbol-based 3-D modeling tool that allows users to apply symbols into a sketch floorplan in order to arrange the spatial layout, and the program converts the floor plan into 3D models according to the use of the symbols. Chapter Three will follow this chapter to take a closer look at freehand drawing and examine what is in a freehand drawing that a 3-D computer modeling tool should recognize. Chapter Four focuses on the computational implementation.
