A 3D Shape Generative Method For Aesthetic Product Design

A 3D shape generative method for aesthetic
product design
Jorge Alcaide-Marzal, Jose Antonio Diego-Mas and Gonzalo Acosta-Zazueta,
Institute for Research and Innovation in Bioengineering (i3B), Universitat
Polit
ecnica de Val
encia, Camino de Vera s/n, 46022, Valencia, Spain
This work describes a generative method for the exploration of product shapes in
the conceptual design stage. The method is based on three concepts: the notion
of grammars to capture product appearance, the implementation of sketching
transformation rules to produce design variations and the use of a parametric
modeller to build shapes. We represent product solutions as 3D sketches using
combinations of basic shapes arranged in simple and schematic product
structures. This procedure allows creating many varied configurations with a
minimal number of shapes, and facilitates the adaptation of the generative model
to different products. The performance of the method is demonstrated through
several examples from the literature.
Ó 2019 Elsevier Ltd. All rights reserved.
Keywords: aesthetics, computer aided design, conceptual design, creativity,
generative design
V
isual appearance is frequently a fundamental attribute of a product.
During the conceptual stage, designers explore the solutions space
searching for suitable appealing concepts to develop in the following
stages. Many authors have studied the processes that lead to successful prod-
uct concepts, as well as the obstacles to creativity that hamper the designer
tasks at this stage (Boden, 2004; Cross, 1997, 2008; Dorst Cross, 2001).
In many design disciplines, sketching is key to generate concepts (Do, Gross,
Neiman, Zimring, 2000; Goel, 1995; Goldschmidt, 1991; Obrenovic, 2009;
Schon Wiggins, 1992; Suwa Tversky, 1997; van Sommers, 1984). The
task of sketching produces situations of reinterpretation and emergence which
favour creativity (Prats, Lim, Jowers, Garner, Chase, 2009; Schon
Wiggins, 1992; Tovey, Porter, Newman, 2003). On the other hand, one of
the most analysed creativity blockers is fixation (Jansson Smith, 1991). By
this mechanism, the designer is trapped in a limited space of familiar solutions
instead of opening to unexplored paths (Crilly, 2015; Crilly Cardoso, 2017;
Vasconcelos Crilly, 2016). This is such an important blocker that it might
deserve its own research field (Crilly Cardoso, 2017).
Corresponding author:
Jorge Alcaide-Marzal.
jalcaide@dpi.upv.es
www.elsevier.com/locate/destud
0142-694X Design Studies xxx (xxxx) xxx
https://doi.org/10.1016/j.destud.2019.11.003 1
Ó 2019 Elsevier Ltd. All rights reserved.
Please cite this article as: Alcaide-Marzal, J et al., A 3D shape generative method for aesthetic product design, Design
Studies, https://doi.org/10.1016/j.destud.2019.11.003

The scientific community has conducted intensive research on methods and
techniques aiming to foster creativity (Wang Nickerson, 2017). Some works
focus specifically on the field of computer aided creativity (Bernal, Haymaker,
Eastman, 2015; Gabriel, Monticolo, Camargo, Bourgault, 2016; Gero,
2000; Liu, Li, Pan, Li, 2011; Lubart, 2005), and many researchers have an-
alysed the influence of the use of computers on the creative performance of de-
signers (Chang, Chien, Lin, Yiching, Hsieh, 2016; Company, Contero,
Varley, Aleixos, Naya, 2009; Robertson Radcliffe, 2009).
One of the proposed approaches to computer aided conceptual design is gener-
ative design (Bernal et al., 2015). Although wide spread in architecture, it is a
relatively new discipline in industrial design (Krish, 2011; Singh Gu, 2012).
Generative design may constitute a way of overcoming fixation: by using the
computer to produce alternatives, the designer is able to explore a wider range
of solutions, facilitating emergence. Actually Bernal et al. (2015: p 170), hold
that the generative exploration of solutions nearly reproduces the “co-evolving
dialog between problem and solution that characterizes expert designers”
described in Dorst and Cross (2001). Hyun and Lee, (2018) point out the
need for flexible computational methods focused on form exploration and
capable of producing a high number of concepts.
This paper analyses the use of generative shape modelling to explore product
aesthetic alternatives. We propose a computer model for stylistic product
design using a combination of three concepts: the notion of product gram-
mars, the transformation rules identified in traditional paper-and-pencil
sketches studies and the use of an interactive parametric modeller.
1 Using computers for creative product styling
1.1 From a representation tool to an idea generator
The role of computers in the conceptual stage of the design process has been a
matter of discussion for long (Bernal et al., 2015; Dekker, 1992; Lubart, 2005;
Tovey, 1997; van Dijk, 1995; Van Elsas Vergeest, 1998; Vuletic et al., 2018).
The special nature of this stage poses important obstacles to the use of CAD.
Conceptual design problems are poorly defined and the available information
is still ambiguous and unstructured, while CAD systems require structured
and accurate data. In addition, it is difficult to replicate the creative workflow
of hand sketching within CAD software. Another drawback is that producing
conceptual solutions in CAD is a slow process for the amount of solutions
needed at this stage (Krish, 2011).
From the point of view of conceptual shape design, the scientific community
has driven its efforts towards two main areas: the use of CAD tools compared
to paper and pencil sketching (computer as a digital representation tool) and
2 Design Studies Vol xxx No. xxx Month xxxx

the use of computational power to increase human productivity (computer as a
complementary exploration tool).
The first research corpus studies the suitability of CAD systems as representa-
tion tools for conceptual design. Due to the thoroughly studied relationship
between paper and pencil sketching and creativity, researchers try to conceive
and evaluate computer tools able to emulate the traditional sketching experi-
ence. This research focuses on the interaction between designer and computer
as a “drawing tool”. The computer is a means to express the designer ideas
more efficiently.
Therefore, most of these works consider the implementation of sketch gestures
in the modelling process (Company et al., 2009; Cook Agah, 2009; Olsen,
Samavati, Sousa, Jorge, 2009; Samavati, Sousa, Jorge, 2008; van Dijk
Mayer, 1997). Other works focus on the analysis of software interfaces
(Hewett, 2005; Machado, Gomes, Walter, 2009; Selker, 2005) or the use
of new paradigms in modelling and representation, most of them also sketch
based (Alcaide-Marzal, Diego-M
as, Asensio-Cuesta, Piqueras-Fiszman,
2013; Bae, Balakrishnan, Singh, 2008; Igarashi, Matsuoka, Tanaka,
1999; Rahimian Ibrahim, 2011). In the 90s, some works used the term Com-
puter Aided Conceptual Design to refer to this kind of software (van Dijk,
1995; Van Elsas Vergeest, 1998).
The efforts to transfer the creative advantages of sketching to conceptual com-
puter tools is in our opinion a very valuable contribution of this approach.
This is also recommended by several authors (Lim et al., 2008; Tovey et al.,
2003). We have considered this when designing the computer model.
The second research corpus investigates how to use the power of computers in
order to help designers perform creative tasks more efficiently, or even to make
the computers produce solutions. This is the computational creativity approach
(Gabriel et al., 2016), which includes any computer application capable of
generating product shapes not directly conceived by the designer. The com-
puter works in this case as a system producing variations from the information
given by the designer, thus increasing the exploration of solutions. Algorithmic
design procedures constitute the main research interest in this area. Shape
grammars and evolutionary models are the most frequently used.
According to Bernal et al. (2015), the research focus has evolved from the first
approach (computer as a digital representation tool) to the second one (com-
puter as a tool capable of transforming information and generating ideas). The
authors, however, raise several limitations to this second approach: As afore-
mentioned, conceptual problems present unstructured, non-hierarchical data,
difficult to handle by CAD systems; moreover, the designers’ expertise and
3D shape generative method 3

instinct is not easy to implement in computer systems. These obstacles prevent
the computer from performing as a designer.
We agree, but we consider that despite these limitations, the second approach
exploits more efficiently the computer’s capabilities. According to Stones and
Cassidy (2010), designing digitally is more related to computers allowing de-
signers to explore solutions beyond what they can draw or imagine than to de-
signers using computers to represent shapes more efficiently. This is what
Mitchell called “digitally mediated design” (Mitchell, 2005). The computer
does not replace the designer, but neither is it a mere representation tool. Prop-
erly used, computers may complement the designer’s abilities, enabling the
exploration of much larger solution fields and overcoming fixation.
As we will describe later, our work aims to merge both approaches: we propose
a generative modeller (computational approach), but provided with a high de-
gree of interactivity. Through this interaction, the designer is able to perform
exploratory actions inspired by sketching creative strategies.
1.2 Generative product design
Generative design involves the algorithmic modification of an initial product
state according to defined rules. Many applications can be found in architec-
ture, as the nature of architectural objects allows an easy generative shape
exploration and the use of optimisation methods (Chase, 2005; Gu
Behbahani, 2018; Rodrigues, Amaral, Gaspar, Gomes, 2015; Shea, Aish,
Gourtovaia, 2005; Singh Gu, 2012). On the contrary, applications in
product design are less numerous. Due to this, generative product design still
lacks formal methodologies for its implementation (Krish, 2011). Nevertheless
new contributions are appearing in fields such as graphic layouts (Cleveland,
2010), electronic consumer products (Lin Lee, 2013), jewellery (Kielarova,
Pradujphongphet, Bohez, 2013) and interface design (Troiano Birtolo,
2014).
Shape grammars are among the most utilised generative methods. They have
been widely used since Stiny and Gips introduced them in their seminal work
(Stiny Gips, 1972). The work of Agarwal and Cagan (1996) is one of the first
using shape grammars for product design. Hsiao and Chen (1997) combine
shape grammars with a semantic approach to define significant shape struc-
tures for an office chair. Later, a work by Hsiao and Wang (1998) used a
similar approach to produce car designs.
An interesting contribution of shape grammars is the formal research of brand
identity features. The work of McCormack, Cagan, and Vogel (2004) presents
a study in which shape grammars are used to capture Buick brand identity and
generate alternative coherent designs. Similar studies can be found related to

Harley motorcycles (Pugliese Cagan, 2002), shampoo bottles (Chen
McKay, 2004), Coca Cola bottles (McKay, Ang, Chau, de Pennington,
2006) or Porsche frontends (Aqeel, 2015).
The idea of shape grammars is very appealing as a tool to assist conceptual
design by automatic generation of solutions. However, there is a need for
methodologies to adapt the generic concept to specific design goals (Ruiz-
Montiel et al., 2013). The complexity of object representations is also an
obstacle to further development of shape grammars based systems (Chase,
2002).
Parametric modelling offers an alternative way of implementing shape gram-
mars, although this approach may miss some of the grammars strengths
(Eloy, Pauwels, Economou, 2018). According to Krish (2011: p 91), nowa-
days it is possible to build most of the grammar definitions through CAD func-
tions. Parametric modelling allows the hierarchical representation of
relationships within the product geometry (replacing the grammar rules),
and permits the exploration of alternatives by modifying the parameters
(Bernal et al., 2015). In Granadeiro, Duarte, Correia, and Leal (2013), the au-
thors describe some works performing adaptations from shape grammars to
parametric design systems. The conversion consists of translating the transfor-
mation rules of shape grammars into parametric equations and variables.
Turrin, von Buelow, and Stouffs (2011) propose a combination of parametric
modelling and interactive genetic algorithms to control the solutions
generation.
Genetic algorithms are widely used for computational design exploration.
Renner and Ek
art (2003) offer a review of their application to computer-
aided design. Since some early works such as Bentley (1998); Gero (1996);
Hybs and Gero (1992), many studies have tested the use of evolutionary algo-
rithms in product design: Lamp holders (Liu, Tang, Frazer, 2004); shape
optimisation for mobile phones (Sun, Frazer, Mingxi, 2007); wine glasses
profiles (Su Zhang, 2010). Shieh, Li, and Yang (2018) use a combination
of Kansei and evolutionary algorithms for vase designs. In O’Neill et al.
(2010) and Lee, Herawan, and Noraziah (2012), the authors combine genetic
algorithms with shape grammars.
However, genetic algorithms focus on design optimisation, something that
may hamper their use in conceptual, ill-defined design problems. Based on
this genetic concept Krish (2011), proposes a different approach called Gener-
ative Design Method. Instead of a fitness function, it uses randomness
controlled by means of the variability of each parameter (gene). This para-
metric method is simple and effective, and adequately set up is able to generate
many different solutions, depending on the product. Moreover, by avoiding
the need for a fitness function, it constitutes a suitable system for subjective

design problems, such as styling design. We will discuss this issue in the next
section, as our proposal lies in a similar approach: we also use parametric
modelling instead of shape grammars and random generation as an explora-
tion procedure.
Although randomness allows introducing unexpectedness into the process, it
may also generate excessive divergence. Some works use different algorithms
to ensure a homogeneous space exploration (Gunpinar Gunpinar, 2018;
Khan Awan, 2018; Khan Gunpinar, 2018). Dang et al. (2015) propose
a probability density function combined with machine learning to avoid
pure random concepts, focusing generation on desired areas of the design
space.
1.3 The aesthetic factor
The consideration of subjective requirements, especially aesthetic ones, is an
ongoing research problem when using generative procedures. Shape grammars
use transformation rules that lead from one configuration to another, but they
lack selection mechanisms to determine the aesthetic validity of the produced
solutions. On the other hand, genetic algorithms use an evaluable performance
of the solutions in order to check the fitness of each individual in each gener-
ation. However, this is complicated when dealing with subjective factors. In
these cases, the use of human interaction is a widely utilised option (Liu
et al., 2004).
There are many examples of research work using interactive genetic algorithms
(IGA) in different product disciplines: fashion design (Gong, Hao, Zhou,
Sun, 2007; Kim Cho, 2000; Mok et al., 2013), car styling (Cluzel,
Yannou, Dihlmann, 2012; Dou, Zong, Nan, 2016; Kelly, Papalambros,
Seifert, 2008), perfume bottles (Kielarova Sansri, 2017), garment design
(Hu, Ding, Zhang, Yan, 2008). An interactive evolutionary system using
shape grammars is proposed in Lee and Tang (2006) to design cameras.
Hsiao, Chiu, and Lu (2010) use genetic algorithms and Kansei for styling of
coffee machines. Nevertheless, including human interaction into the evolu-
tionary system slows down its performance. Humans cannot assess the alter-
natives as fast as the system is capable of generating them (Renner Ek
art,
2003), and they easily feel tired in this kind of task (Gong et al., 2007).
Therefore, some research focuses on the automation of the evaluation stage.
Lo, Ko, and Hsiao (2015) quantify product aesthetics by means of an evalua-
tion equation built from aesthetic theories. They utilise it as a fitness function
to make product solutions evolve towards more appealing ones. Gu, Xi Tang,
and Frazer (2006) propose an IGA combined with neural networks. Similarly
Sun, Gong, and Zhang (2011), propose an IGA with semi-supervised learning
to design sunglass lenses. Hsiao and Tsai (2005) use trained fuzzy neural

networks as fitness function to design electronic door locks. Khan and Awan
(2018) propose an approach using Jaya algorithms to find optimal solutions.
They use it to design several products such as wine glasses, ceiling lamps, mo-
torbikes or loudspeakers.
The complete automation of the exploration process is a remarkable goal.
Despite this, in our opinion, the use of interactive systems results in a
designer-computer cooperation that closely replicates the evolving exploration
via sketches. As a means to implement these cooperative dynamics in our pro-
posal, we aim to provide the generative system with a high degree of
interactivity.
2 Proposed methodology
The analysis of the state of the art describes very promising research areas.
Most of the analysed works seek to combine the capabilities of the existing
CAD systems with the generative production of solutions (for instance, linking
the generation of solutions with their corresponding mechanical analysis).
Actually, an effective and immediate connection between computer-
generated concepts and CAD utilities is very desirable. However, we think
that currently this goal restricts the potential of a generative system to produce
aesthetic solutions. Moreover, this approach moves away from the typical pa-
per and pencil sketching process, resulting in several of the disadvantages that
designers find when using traditional CAD systems in the conceptual design
stage.
According to Bernal et al. (2015) reinterpretation may also occur when using
computers. Unexpected shapes produced by a generative system are poten-
tially capable of suggesting new solution fields to explore. These authors
also point out the lack of reformulation capabilities (the possibility of modi-
fying the solution to explore a reinterpretation) as a drawback of generative
design. Nevertheless, if the generative system facilitates interactivity to the
designer, enabling the modification of any parameter at any time, then refor-
mulation is possible.
Hence, our proposal is an interactive generative system producing conceptual
shapes, comparable to pencil sketches. We refer to it as Conceptual Generative
Model (CGM). Following some authors (Liu et al., 2004; Stones Cassidy,
2010), it is not conceived as a geometric modeller, but as a style explorer. In
fact, some authors recommend separating the conceptual computer system
from detailed CAD modellers, as they work with different objects (van Dijk,
1995). At this stage, solutions do not need to be completely correct or fulfil
exact functional requirements. On the contrary, as the purpose of these models
is producing emergence, they may even be intuitively incorrect (Hsiao Chen,
1997).

Therefore, we have prioritised the variability of the generated shapes over their
topological precision and consequently over the immediate CAD connection.
The CGM may produce geometrically wrong solutions as long as they commu-
nicate a suitable visual idea of their aesthetic appearance. For instance, in gen-
eral the CGM does not perform Boolean operations, to increase operational
speed. Consequently, solutions may present overlapping geometry. Actually,
most of them do. The reason is that we do not expect the CGM to produce
detailed CAD models, but conceptual “3D sketch” models useful to inspire
varied and hopefully unexpected and original shapes. This strong simplifica-
tion serves another purpose: a high generality and variability of the CGM.
We aim to define an open model, valid for almost any product and easy to
adjust and modify to include new geometric features (Hoisl Shea, 2011).
We have based the CGM on three main concepts, which we will describe in the
following sections:
a) The idea of product grammar as a set of geometric features and relation-
ships describing the appearance of a particular product.
b) Sketching shape rules: Some studies have identified general transforma-
tion rules successfully used by designers to explore solutions when sketch-
ing. The CGM will make use of these rules through interactivity.
c) Parametric modelling: The product grammar and the shape rules are
embodied as parametric operations into a 3D generative model. This
model is in turn comprised by three main areas, as we will explain in sec-
tion 2.3.
2.1 Defining an initial product grammar
In order to define an initial product grammar, we need to analyse a wide range
of existing products (Figure 1). As mentioned above, we use the concept of
grammar in a broad sense, referring to the geometric elements and restrictions
that confer a product its visual identity.
Chen and Owen (1998) present a generative system that splits products into
spatial configurable volumes. A set of rules determines their relationships,
incorporating geometric as well as stylistic information to each volume. The
CGM operates in a similar way, using a grammar based on a product structure
and a product vocabulary (Figure 2).
a) PRODUCT STRUCTURE: The product structure is a schematic repre-
sentation, by means of boxes, of the main volumes that constitute the
whole product shape. The key function of the product structure is con-
trolling the position and proportion of the vocabulary elements.
b) PRODUCT VOCABULARY: The product vocabulary includes a wide
range of simple shapes able to populate coherently each of the boxes of

a particular product structure. That is, the placement of these shapes into
the structure boxes produces feasible products. The product vocabulary
provides visual identity and variety to the product.
Generally, detecting the structure and vocabulary of a product will not require
a thorough market analysis. However, performing it results in a more flexible
product structure and a richer vocabulary, thus increasing the CGM’s ability
to produce varied concepts.
Our goal is to keep the vocabulary as general as possible, so it will preferably
be restricted to combinations of simple geometries (boxes, spheres, cylinders,
revolutions), as long as they are capable of representing most of the visual fea-
tures detected in the product.
2.2 Setting shape transformation rules
Shape grammars use transformation rules to produce new shapes from exist-
ing ones. These transformations may be translations, rotations, substitutions
. Rules are applied by identifying an existing geometry, analysing the appli-
cable rules and deciding which one to apply. They are usually specific to the
product under study.
Using this concept Prats et al. (2009), identified several “general transforma-
tion rules” that designers utilise to transform sketches when exploring solu-
tions (Table 1). These generic rules are outline transformations, structure
transformations, substitution, addition, deletion, cutting and view change. The
authors recommend considering them when conceiving a computer aided
design system. Some other works have already used them (Alcaide-Marzal
et al., 2013; Kielarova, Pradujphongphet, Bohez, 2015).
Figure 1 Product grammar analysis for perfume bottles and implementation into the CGM

In this work, we have implemented these rules as different parametric opera-
tions within the CGM. We aim to provide it with the same general mechanisms
used by designers to produce creative solutions when they are sketching. Some
of these rules are embedded in the operations utilised to build the product vo-
cabulary and structure. Others are added as secondary transformations to in-
crease shape variability.
2.3 Building the Conceptual Generative Model (CGM)
This stage involves defining an initial parametric model capable of producing
the required geometry. We have built the CGM using Grasshopper 1.0.0006
for Rhinoceros V6. This software constitutes a very convenient way to imple-
ment generative design procedures, obtaining an immediate geometric result.
Lately, its use in research involving algorithmic and generative design has
increased notably.
As mentioned before, the CGM presents three main areas, working sequen-
tially (Figure 3):
a) The first area is devoted to the construction of the product vocabulary. It
allows building 2D profiles (basic polygons and circles, polylines or
random curves) and 3D operations (parametric operations producing
3D geometry such as extrusions, revolutions or lofts). This area imple-
ments shape rules (outline transformations, substitutions and additions)
aimed to build and modify the vocabulary elements. These elements are
sent to the second area.
b) The second area permits controlling the product structure. It allows
defining the general disposition of the product’s main volumes, repre-
sented by boxes that will receive the vocabulary shapes. This area includes
shape rules (structure transformations, substitutions, additions or
Figure 2 Morphing vocabulary elements into product structure boxes

Table 1 Shape rules (Prats et al., 2009) and some examples of implementation into the CGM as parametric operations.
Outline transformations are mostly present in the product vocabulary operations. Structure transformations may be found
in the product structure operations (a change of position, for instance) or in secondary transformations (an array).
Figure 3 The CGM implemented in Grasshopper

deletions) to build and modify the boxes that constitute the structure
elements.
c) The third area allows applying secondary transformations to the vocabu-
lary shapes once morphed into the structure, multiplying the variety of
shapes. These transformations are produced by shape rules such as struc-
ture transformations (translation, rotation, reflection, array, twist,
scaling), as well as addition and cutting.
From a genetic perspective, the CGM represents the genotype of the product
and the parameters represent the genes. Each possible solution is a phenotype,
characterised by a unique set of parameter values. Each parameter in the
CGM presents both a manual and a random input (Figure 4).
The CGM includes two types of parameters:
a) Numerical parameters: determine the value of a dimension, such as the
diameter of a circle or the height of a box. Grasshopper allows assigning
an interval for these parameters, thus defining what Krish calls the explo-
ration envelope, the limits of the space of solutions (Krish, 2011). Each
numerical parameter includes a control to set its level of randomness,
that is, how much the random value and the manual one may differ.
Hence, the designer may control the variability of each dimension
(parameter variability, PV).
b) Categorical parameters: determine if a given geometry is present in a so-
lution. These parameters control for instance which 2D profile is used in a
loft operation, which shape is placed into a structure box or whether a
secondary transformation is applied or not. From a genetic point of
view, we may consider the geometries controlled by categorical parame-
ters as traits of the product, which may or may not be expressed in the
phenotype. Each categorical parameter includes a control to determine
the probability of occurrence (PO) of the corresponding trait.
The CGM begins by producing an initial solution, which results from applying
any set of parameter values. Krish (2011) recommends using simple pheno-
types to start the exploration. In order to generate solutions randomly, the
designer defines the probability of occurrence of each trait (PO) and the vari-
ability of each parameter (PV). Once defined these values, each run of the
Grasshopper solver will produce a new solution. The designer may decide in
each run which parameters to set manually and which ones to randomise
and to which degree. Every time the CGM generates a random concept, the
designer may manually change any parameter to modify the solution.
Figure 5 shows some examples of shape exploration.

3 Exploratory case studies
It is difficult to compare the performance of the proposed model against
models described previously in the literature, as most of them consider not
only aesthetic shape generation but also functional requirements. Therefore,
to validate the model we have developed versions of the CGM for several of
the existing examples, analysing their performance and ability to represent
Figure 4 A subsystem of the CGM product vocabulary area, aimed to build a 6 point symmetric profile with three possible types of output:
polyline, interpolated curve and a combination of arc and curve. This subsystem includes controls to set manually the (X,Y) coordinates of
the points defining the profile, a control to set the PV of these values (if set to 0, there is no random effect) and a control to define the PO
of each type of output. The PO Solver produces outputs according to these probabilities. The (X,Y) values are numerical parameters; the
PO Solver output is a categorical parameter
Figure 5 Application of rules starting from several 2D profiles

plausible and highly varied product geometries effectively. We present these
tests in the following sections. The first cases are described a little more in
depth, to show the CGM possibilities.
3.1 Perfume bottles
Our first CGM product test is a perfume bottle, the same product utilised in
the optimisation study described in Kielarova and Sansri, (2017). They
consider both objective and aesthetic requirements. A similar product
(shampoo bottles) is used in Chen and McKay, (2004), using a shape grammar
approach to capture brand image features.
Figure 6 represents the bottle CGM grammar, based on the analysis previously
shown in Figure 1:
The product vocabulary consists of 16 configurable 2D shapes and 8 3D op-
erations. The 3D operations are mostly general, while many of the 2D
shapes derive from the product analysis.
Kielarova and Sansri (2017) divide the bottle into three volumes (base, body
and neck). We consider three volumes as well, but differently distributed
(bottle body, cap base and cap body), comprising the product structure.
We have included two more optional volumes to add particular aesthetic
details.
The bottle grammar is extended with 10 secondary transformations.
Considering just the shape combinations that we can achieve with this initial
grammar, we get 56 main shape families (16 vertical extrusions, 16 horizontal
Figure 6 Perfume bottle grammar

extrusions, 16 lofts, revolutions, cylinders, spheres, cones, boxes and 3 platonic
shapes), which we may use almost indistinctly to create bottles bodies, cap ba-
ses or cap bodies.
As mentioned before, we use simple phenotypes as the starting point to explore
the solutions space. In order to demonstrate the potential of the CGM to pro-
duce different alternatives, Figure 7 shows some of the results, grouped in rows
by bottle body families. The first element in each row is the initial phenotype of
the corresponding family.
All these variations come from the same genotype. Unexpected shapes arise
easily when the CGM utilises secondary transformations. Random curves
also generate interesting and organic geometries. Figure 8 displays the shape
rules used to build all the solutions of the first row in Figure 7 from the initial
one. It is a symbolic representation of the transformations applied. These rules
are parametric operations and they need the corresponding parameter values
to define entirely the resulting solution.
Although we have displayed all the transformations from the starting pheno-
type, each solution is reachable from any other. In fact, most of these solutions
derive from intermediate ones through random exploration or manual refor-
mulation. The designer may apply shape rules in reverse as well, cancelling
some features to attain a more appealing solution. Figure 9 shows an example
of manual versus random control of parameter values.
An important sketching strategy is the design families exploration (Lim et al.,
2008). It consists of generating versions of a particular design solution, what
Goel called “vertical transformations” (Goel, 1995). The CGM provides this
sketching technique by controlling the probability of occurrence (PO) of
each trait. Let us suppose that the designer comes up with a solution that is
appealing due to a particular trait (a bottle including an array, for instance).
Then, it is possible to explore many versions of this solution by producing
random variations with a high PO for the corresponding trait (the array in
this case), thus ensuring its presence.
In Figure 10, we present different concept generations. Those in column A are
produced with PO ¼ 15% for all traits, so they display a reasonable variability.
Groups in column B represent generations with a particular PO raised to 95%
(for “array” in the first row, for “non-uniform scale” in the second one).
Finally, groups in column C keep these values, increasing the PO for a partic-
ular shape (“curve revolutions” and “vertical extrusions from symmetric 6-point
curves”) and decreasing the PO for the rest of traits to 5%. Allowing at least a
minimum probability for the rest of traits enhances variability and favours un-
expected solutions. In this case, the two solutions highlighted in column B

present the same combined traits (array and non-uniform scale) despite com-
ing from diﬀerent initial phenotypes.
This mechanism emulates the “probability density function” proposed in
Dang et al. (2015). Assigning higher probabilities to desired traits, the designer
may control the style of the product while keeping other parameters variable.
Supplementary video related to this article can be found at https://doi.org/
10.1016/j.destud.2019.11.003.
Figure 7 Several body families of perfume bottles
Figure 8 Shape rules applied from initial phenotype to achieve each variation. As the rules are implemented through parametric operations, they
need numerical information to complete the transformation. For instance, a SCALE rule needs a value for the scale factor, or an ARRAY rule
requires a value for the number of instances

The following is the supplementary data related to this article: The whole
exploration of the perfume bottle space turned out too large to be fully
described here. We have included a video showing parts of these sessions.
Figure 9 Bottle body and cap body variations. An example of manual vs random control of parameters
Figure 10 Exploring shape families using probability of occurrence (PO)

3.2. Table lamps
As generality is an important goal of the proposed model, a new case study is
prepared to check the transformation of a product structure into another one.
We use table lamps, following the work of Liu et al. (2004). The approach of
this research is very similar to ours, as it focuses on product aesthetic perfor-
mance, and the designer interactively guides the evolution of the system.
Liu et al. (2004) use a four part structure (base, body, light and shade). We
divide the table lamp into three main volumes (base, body and lampshade).
Therefore, we only need to adjust the proportions of boxes in the bottle struc-
ture to obtain the lamp structure (Figure 11). The bottle product vocabulary is
only modiﬁed to add two operations (sweep shapes and hollow extrusions), in
order to include some geometries relevant to the lamp morphology. The rest of
the grammar remains the same.
Figure 12 shows some results grouped by base or by body shape families.
A very interesting feature of the proposed CGM is the possibility of transfer-
ring traits between products. Once the sweep and hollow extrusion operations
were added to the lamp CGM, we thought of inserting them into the bottle vo-
cabulary, increasing the initial bottle grammar. Two entirely new families of
concepts emerged in this test, extending the design space (Figure 13). This
CGM feature represents a good example of the “analogy building” creative
strategy, based on transferring knowledge design from one situation to
another (Goel, 1997). It also allows generating stylistically coherent products
(bottles and lamps in this case) by using common traits (Chen Owen, 1998).
Figure 11 Adapting the perfume bottle structure to the lamp CGM. It is worth noting that in this example the vocabulary elements are the same
in both cases. The product structure makes them take the suitable proportions to look either like a bottle or like a lamp

10.1016/j.destud.2019.11.003.
The following is the supplementary data related to this article: As in the pre-
vious example, we have included a video showing part of the exploration
session.
3.3. More complex examples
Both previous case studies present a very close product structure and few geo-
metric restrictions. A ﬁnal test was performed using other examples from the
literature review: Jewellery rings (Kielarova et al., 2013), coﬀee makers
(Agarwal Cagan, 1996; Hsiao et al., 2010), office chairs (Hsiao Chen,
1997), wheel rims (Gunpinar Gunpinar, 2018; Hoisl Shea, 2011; Khan
Figure 12 Several lamp families by base and body
Figure 13 New perfume bottles families, generated by adding elements from the table lamp vocabulary to the bottle vocabulary (analogy
building)

Gunpinar, 2018; Lim et al., 2008) and motorbikes (Khan Awan, 2018;
Pugliese Cagan, 2002). The goal was to check the CGM’s ability to adapt
to very diﬀerent products with more complex geometric requirements.
After analysing the grammar of each product, we adapted the CGM to each
structure and vocabulary. Several sets of 20 alternatives were produced, using
random generation combined with manual adjustments (Figure 14 and
Figure 15).
10.1016/j.destud.2019.11.003.
The following is/are the supplementary data related to this article: Videos
attached in this section show some raw results and other explorations per-
formed during the exploration sessions.
Figure 14 Groups of solutions for office chairs, jewellery rings and coﬀee makers

Table 2 summarises the adaptation of the CGM to these products. Using the
bottle CGM as a base, we describe the changes needed to obtain the models for
the rest of the products.
Obtaining the ring CGM was almost immediate, as the vocabulary elements
used in the bottle grammar were able to produce most of the shapes identified
in the ring study, except for the circular ones. As for the product structure, we
only added a circle under the three-box structure of the bottle to allow the
placement of circular vocabulary elements. The coffee maker CGM was also
relatively easy to achieve, although we had to increase the vocabulary and
include some operations.
The office chair and the motorbike needed larger modifications. While previ-
ous examples shared a product structure of vertical stacked boxes, we could
not use this configuration to represent the geometry of the chair and the
motorbike. In these cases, we built a product structure based on lines defining
a “skeletal” representation of the product. We placed the structure boxes along
Figure 15 Groups of solutions for rims and motorbikes

these lines (Figure 16). This allowed an easy reorientation of the volumes,
which were mostly not vertical. This kind of lineal product structure has
proved more flexible and adequate to complex product configurations and
will be generalised in future versions of the CGM.
As mentioned above, we wanted the CGM to be easily adaptable to a wide va-
riety of products with minimal modifications. Therefore, many of the grammar
elements should be general and reusable in different products. Table 3 summa-
rises our analysis about grammar usage in each CGM test:
a) Black dots represent grammar elements (columns) initially included in the
product grammar and utilised with coherent results in the corresponding
structure part (rows). We consider a result coherent if its geometry intu-
itively respects the basic functionality of the product. A black dot indi-
cates a strong and immediate relationship between the grammar
element and the structure part (cylinders for bottle body and cones for
lamp shade are clear examples).
b) White dots represent grammar elements initially not included in the cor-
responding structure part, but eventually used with coherent results. A
white dot indicates a feasible, but not immediate, relationship between
the grammar element and the structure part.
The last column represents in a colour scale the degree of grammar usage for
each part. It is a qualitative representation based on the number of elements
used. In the cases studied, perfume bottles, table lamps and rings have made
a more intensive use of the grammar elements. The wheel rims and the motor-
bike, due to their functional configuration, are the most restricted products.
We have used this information to build a genetic correlation matrix (Table 4),
comparing the genetic load (the amount of grammar usage) between the seven
product CGMs. This is a qualitative approach aimed to illustrate the CGM
potential to adapt to different products.
Table 2 Grammar adaptation and performance (average and maximum generation times, AGT and MGT) for the seven prod-
ucts. Generation times correspond to the software running on an AMD Ryzen 5 1400 QuadCore 3.2 GHz. Total average time
needed to produce a new random solution was 0.812 s

We assign a value of 1 to black dots and 0.5 to white ones. These values are
indicative of the relationship strength. They intend to express numerically
the suitability of a given grammar element to adapt to a particular structure
part. Each cell “g” of the genetic correlation matrix (Table 4) is built using
the values of cells “a” from the grammar usage table (Table 3) following
this expression:
gij ¼
Pn
k¼0aik ajk
Pn
k¼0aik aik
Each cell gij in Table 4 represents the ratio between grammar usage of part i
(row) and part j (column). Let us refer to part i as A and to part j as B.
Figure 16 “Stacked boxes” structure vs “skeletal lines” structure
Table 3 Grammar usage by product part

Conceptually, the score in cell AB expresses how much of the grammar of part
B is present in part A:
Values very close to 1 indicate that grammar of A includes most of the
grammar of B. Values will exceed 1 if, in addition, A presents many black
dots where B presents white dots, in Table 3. This indicates that part A
shows a stronger relationship with grammar elements.
If parts A and B share a very similar grammar (many common dots in Table
3) they will show values very close to 1 both in cells AB and BA.
Values below 1 indicate that A uses only a fraction of the grammar elements
present in B.
Parts whose rows present many high scores (we have highlighted those above
0.7) include many of the grammar elements used in other parts. Generally,
they utilise a great percentage of the whole grammar and show high shape va-
riety. On the contrary, those with many low-score cells make a lower grammar
usage and present less shape variety, mainly due to geometric restrictions.
From Tables 3 and 4, we observe that:
Products whose geometry is less functionally restricted (bottle, lamp or
ring) make more use of the grammar possibilities and include many ele-
ments utilised in other products. They are also more likely to add new
shapes to their initial grammars during exploration. These products may
act as inspiration when exploring others, as they present a high variety of
shapes and a high probability of transferring them.
Table 4 Genetic correlation matrix.

Even those products very limited by their geometric functionality present
parts with a high potential variability. It is possible to adapt grammar ele-
ments from less restricted products to these parts, by analogy building.
Parts with similar functional restrictions make use of similar grammar ele-
ments. This is the case for the tubular parts or the seats in chair and motor-
bike. Shape exploration strategies used in one of these parts could be useful
for the rest.
4 Discussion
The CGM performed satisfactorily in all products tested. The generation of
solutions proved very fruitful even in the most complex cases, and time needed
to produce a solution was below 1 s on average.
The generality of the CGM has been demonstrated by the application to the
exploration of design spaces from very different products. Despite this, we
consider that this feature of the model may be improved. More configurable
product structures are desirable to simplify the adaptation to different prod-
ucts. On the contrary, adapting the vocabulary has been reasonably easy,
although models of products with high geometric functionality required
more specific vocabulary elements.
The addition of new items to an initial grammar increases exponentially the
design space, which benefits exploration. In our opinion, the modular flexi-
bility of the CGM constitutes one of its main advantages and makes it outper-
form previous methods in this aspect. It enhances generality, makes the CGM
scalable, allows analogy building and even makes possible the use of prede-
fined CAD or previous CGM geometry inside a product model.
The CGM enables the communication cycle between designer and computer
described by Prats et al. (2009). Apart from stimulating emergence through
randomness, we have shown that it provides designers with other different
sketching creative strategies described in the literature such as reformulation,
design families exploration, analogy building, combination or transformation.
We consider that the performance of these features in design practice using the
CGM deserves a particular study and will be subject of future works.
Despite these good results, we are aware that this is an exploratory study and
the approach needs more development. The construction and fine-tuning of
the first model (bottles) was hard and very demanding, and many aspects of
the CGM could be improved and optimised. The use of a specific software
is also a limitation. The designer must master it to achieve a high degree of
exploration, as many interesting shapes emerge by playing with the CGM at
its lowest level, modifying the Grasshopper file definition. It is possible to

prepare user-friendly interfaces for Grasshopper definitions; however, accord-
ing to our experience this greatly restricts the model interactivity.
Finally, although the tests presented permit an internal validation of the pro-
posal, an external validation would confirm to which extent the method is use-
ful. These complementary studies are part of future works.
5 Conclusions
In this work, we have presented and tested a generative design technique
focused on obtaining a high number of valid aesthetic proposals for product
design. The goal was to build a general parametric model able to produce
many different and coherent conceptual shapes for very different products,
with minimum changes.
Despite the aforementioned limitations, the results are satisfactory and open
an interesting space of research on the adaptation of design creativity theories
to generative computer techniques. In this sense, we are working on the
enhancement and encoding of CGM elements, the generation of adaptable
product structures and the exploration of workflow strategies. It is possible
to connect several instances of the CGM to explore product parts separately.
Designers may create and reutilise repositories of CGM generated geometry to
increase performance. The consideration of particular cases is also under
study. For instance, products characterised by just one main volume with sig-
nificant surface variations (shoes, helmets, electronic devices) or very organic
shapes need a specific approach. A connection with some machine learning
mechanism to help designers to evaluate concepts is also under development.
As a final reflection, we consider that an actual implementation of computer
aided conceptual design requires changing the way we design today. We
have to look at generative systems as objects of design rather than as tools
for designing. In Lin and Lee (2013), the authors differentiate between gener-
ative designers (people who design the generative system) and industrial de-
signers (people who use the generative system to design products). It is true
that designers may not be used to this kind of tool yet, and that paper and pen-
cil are paramount when producing conceptual solutions. However, the process
of defining the generative system is actually very similar to the act of sketching.
Creating geometric relationships and playing with them efficiently to explore
shapes requires knowledge about product features, aesthetics intuition and
imagination. Generative systems may help designers to be more creative, but
designing and modifying them also needs creativity. There is not a sequential
process based on creating the system and then using it to create solutions. We
are creating all solutions at once, because building the generative system simul-
taneously builds the space of solutions to explore. The system is open and
evolves with the problem throughout the creative process. Just as sketches do.

This research did not receive any specific grant from funding agencies in the
public, commercial, or not-for-profit sectors.
Declaration of Competing Interests
The authors declare that they have no known competing financial interests or
personal relationships that could have appeared to influence the work reported
in this paper.
References
Agarwal, M., Cagan, J. (1996). A blend of different tastes: The language of cof-
feemakers. Environment and Planning B: Planning and Design, 25(2), 205e226.
https://doi.org/10.1068/b250205.
Alcaide-Marzal, J., Diego-M
as, J. A., Asensio-Cuesta, S., Piqueras-Fiszman, B.
(2013). An exploratory study on the use of digital sculpting in conceptual prod-
uct design. Design Studies, 34(2), 264e284. https://doi.org/10.1016/
J.DESTUD.2012.09.001.
Aqeel, A. B. (2015). Development of visual aspect of Porsche brand using CAD
technology. Procedia Technology, 20, 170e177. https://doi.org/10.1016/
J.PROTCY.2015.07.028.
Bae, S.-H., Balakrishnan, R., Singh, K. (2008). ILoveSketch: as-natural-as-
possible sketching system for creating 3D curve models. In S. B. Cousins,
M. Beaudouin-Lafon (Eds.), Proceedings of the 21st annual ACM symposium
on User interface software and technology (pp. 151e160). ACM. https://
doi.org/10.1145/1449715.1449740.
Bentley, P. J. (1998). Aspects of evolutionary design by computers. In Roy, R.,
Furuhashi, T., Chawdhry, P. K. (Eds.). Advances in soft computing engineer-
ing design and manufacturing, Vol. 44. London: Springer. Retrieved from.
http://eprints.ucl.ac.uk/171653/.
Bernal, M., Haymaker, J. R., Eastman, C. (2015). On the role of computational
support for designers in action. Design Studies, 41, 163e182. https://doi.org/
10.1016/J.DESTUD.2015.08.001.
Boden, M. A. (2004). The creative mind : Myths and mechanisms (2nd ed.).
Routledge.
Chang, Y.-S., Chien, Y.-H., Lin, H.-C., Yiching, M. C., Hsieh, H.-H. (2016).
Effects of 3D CAD applications on the design creativity of students with
different representational abilities. Computers in Human Behavior, 65,
107e113. https://doi.org/10.1016/J.CHB.2016.08.024.
Chase, S. C. (2002). A model for user interaction in grammar-based design sys-
tems. Automation in Construction, 11(2), 161e172. https://doi.org/10.1016/
S0926-5805(00)00101-1.
Chase, S. C. (2005). Generative design tools for novice designers: Issues for selec-
tion. Automation in Construction, 14(6), 689e698. https://doi.org/10.1016/
j.autcon.2004.12.004.
Chen, X., McKay, A. (2004). Package shape design principles to support brand
identity. In 14th IAPRI world conference on packaging (pp. 1e14).
Chen, K., Owen, C. L. (1998). A study of computer-supported formal design.
Design Studies, 19(3), 331e359. https://doi.org/10.1016/S0142-694X(98)00002-
7.
Cleveland, P. (2010). Style based automated graphic layouts. Design Studies,
31(1), 3e25. https://doi.org/10.1016/J.DESTUD.2009.06.003.

Cluzel, F., Yannou, B., Dihlmann, M. (2012). Using evolutionary design to
interactively sketch car silhouettes and stimulate designer’s creativity. Engi-
neering Applications of Artificial Intelligence, 25(7), 1413e1424. https://
doi.org/10.1016/j.engappai.2012.02.011.
Company, P., Contero, M., Varley, P., Aleixos, N., Naya, F. (2009). Computer-
aided sketching as a tool to promote innovation in the new product develop-
ment process. Computers in Industry, 60(8), 592e603. https://doi.org/
10.1016/j.compind.2009.05.018.
Cook, M. T., Agah, A. (2009). A survey of sketch-based 3-D modeling tech-
niques. Interacting with Computers, 21(3), 201e211. https://doi.org/10.1016/
j.intcom.2009.05.004.
Crilly, N. (2015). Fixation and creativity in concept development: The attitudes
and practices of expert designers. Design Studies, 38, 54e91. https://doi.org/
10.1016/J.DESTUD.2015.01.002.
Crilly, N., Cardoso, C. (2017). Where next for research on fixation, inspiration
and creativity in design? Design Studies, 50, 1e38. https://doi.org/10.1016/
J.DESTUD.2017.02.001.
Cross, N. (1997). Descriptive models of creative design: Application to an
example. Design Studies, 18(4), 427e440. https://doi.org/10.1016/S0142-
694X(97)00010-0.
Cross, N. (2008)Engineering design methods: Strategies for product design, Vol. 58.
Wiley.
Dang, M., Lienhard, S., Ceylan, D., Neubert, B., Wonka, P., Pauly, M. (2015).
Interactive design of probability density functions for shape grammars. ACM
Transactions on Graphics, 34(6), 1e13. https://doi.org/10.1145/
2816795.2818069.
Dekker, K. (1992). A future interface for computer-aided styling. Design Studies,
13(1), 42e53. https://doi.org/10.1016/0142-694X(92)80004-I.
van Dijk, C. G. C. (1995). New insights in computer-aided conceptual design.
Design Studies, 16(1), 62e80. https://doi.org/10.1016/0142-694X(95)90647-X.
van Dijk, C. G. C., Mayer, A. A. C. (1997). Sketch input for conceptual surface
design. Computers in Industry, 34(1), 125e137. https://doi.org/10.1016/S0166-
3615(96)00073-5.
Do, E. Y.-L., Gross, M. D., Neiman, B., Zimring, C. (2000). Intentions in and
relations among design drawings. Design Studies, 21(5), 483e503. https://
doi.org/10.1016/S0142-694X(00)00020-X.
Dorst, K., Cross, N. (2001). Creativity in the design process: Co-evolution of
problemesolution. Design Studies, 22(5), 425e437. https://doi.org/10.1016/
S0142-694X(01)00009-6.
Dou, R., Zong, C., Nan, G. (2016). Multi-stage interactive genetic algorithm
for collaborative product customization. Knowledge-Based Systems, 92,
43e54. https://doi.org/10.1016/J.KNOSYS.2015.10.013.
Eloy, S., Pauwels, P., Economou, A. (2018). AI EDAM special issue: Advances
in implemented shape grammars: Solutions and applications. Artificial Intelli-
gence for Engineering Design, Analysis and Manufacturing, 32(2), 131e137.
https://doi.org/10.1017/S0890060417000634.
Gabriel, A., Monticolo, D., Camargo, M., Bourgault, M. (2016). Creativity
support systems: A systematic mapping study. Thinking Skills and Creativity,
21, 109e122. https://doi.org/10.1016/J.TSC.2016.05.009.
Gero, J. S. (1996). Creativity, emergence and evolution in design. Knowledge-
Based Systems, 9(7), 435e448. https://doi.org/10.1016/S0950-7051(96)01054-4.

Gero, J. S. (2000). Computational models of innovative and creative design pro-
cesses. Technological Forecasting and Social Change, 64(2), 183e196. https://
doi.org/10.1016/S0040-1625(99)00105-5.
Goel, V. (1995). Sketches of thought. Cambridge: MIT Press.
Goel, V. (1997). Design, analogy, and creativity. IEEE Expert, 12(3), 62e70.
https://doi.org/10.1109/64.590078.
Goldschmidt, G. (1991). The dialectics of sketching. Creativity Research Journal,
4(2), 123e143.
Gong, D.-W., Hao, G.-S., Zhou, Y., Sun, X.-Y. (2007). Interactive genetic al-
gorithms with multi-population adaptive hierarchy and their application in
fashion design. Applied Mathematics and Computation, 185(2), 1098e1108.
https://doi.org/10.1016/j.amc.2006.07.043.
Granadeiro, V., Duarte, J. P., Correia, J. R., Leal, V. M. S. (2013). Building
envelope shape design in early stages of the design process: Integrating archi-
tectural design systems and energy simulation. Automation in Construction, 32,
196e209. https://doi.org/10.1016/j.autcon.2012.12.003.
Gu, N., Behbahani, P. A. (2018). Shape grammars: A key generative design al-
gorithm. In B. Sriraman (Ed.), Handbook of the mathematics of the arts and sci-
ences (pp. 1e21). Cham: Springer International Publishing. https://doi.org/
10.1007/978-3-319-70658-0_7-1.
Gunpinar, E., Gunpinar, S. (2018). A shape sampling technique via particle
tracing for CAD models. Graphical Models, 96, 11e29. https://doi.org/
10.1016/J.GMOD.2018.01.003.
Gu, Z., Xi Tang, M., Frazer, J. H. (2006). Capturing aesthetic intention during
interactive evolution. Computer-Aided Design, 38(3), 224e237. https://doi.org/
10.1016/J.CAD.2005.10.008.
Hewett, T. T. (2005). Informing the design of computer-based environments to
support creativity. International Journal of Human-Computer Studies, 63(4),
383e409. https://doi.org/10.1016/j.ijhcs.2005.04.004.
Hoisl, F., Shea, K. (2011). An interactive, visual approach to developing and
applying parametric 3D spatial grammars. Artiﬁcial Intelligence for Engineer-
ing Design, Analysis and Manufacturing, 25, 333e356. https://doi.org/10.1017/
S0890060411000205.
Hsiao, S.-W., Chen, C.-H. (1997). A semantic and shape grammar based
approach for product design. Design Studies, 18(3), 275e296. https://doi.org/
10.1016/S0142-694X(97)00037-9.
Hsiao, S.-W., Chiu, F.-Y., Lu, S.-H. (2010). Product-form design model based
on genetic algorithms. International Journal of Industrial Ergonomics, 40(3),
237e246. https://doi.org/10.1016/J.ERGON.2010.01.009.
Hsiao, S.-W., Tsai, H.-C. (2005). Applying a hybrid approach based on fuzzy
neural network and genetic algorithm to product form design. International
Journal of Industrial Ergonomics, 35(5), 411e428. https://doi.org/10.1016/
j.ergon.2004.10.007.
Hsiao, S.-W., Wang, H.-P. (1998). Applying the semantic transformation
method to product form design. Design Studies, 19(3), 309e330. https://
doi.org/10.1016/S0142-694X(98)00009-X.
Hu, Z.-H., Ding, Y.-S., Zhang, W.-B., Yan, Q. (2008). An interactive co-
evolutionary CAD system for garment pattern design. Computer-Aided Design,
40(12), 1094e1104. https://doi.org/10.1016/J.CAD.2008.10.010.
Hybs, I., Gero, J. S. (1992). An evolutionary process model of design. Design
Studies, 13(3), 273e290. https://doi.org/10.1016/0142-694X(92)90216-W.
Hyun, K. H., Lee, J.-H. (2018). Balancing homogeneity and heterogeneity in
design exploration by synthesizing novel design alternatives based on genetic

algorithm and strategic styling decision. Advanced Engineering Informatics, 38,
113e128. https://doi.org/10.1016/J.AEI.2018.06.005.
Igarashi, T., Matsuoka, S., Tanaka, H. (1999). Teddy: A sketching interface for
3D freeform design. In Proceedings of the 26th Annual Conference on Computer
Graphics and Interactive Techniques - SIGGRAPH ’99 (pp. 409e416). https://
doi.org/10.1145/311535.311602.
Jansson, D. G., Smith, S. M. (1991). Design fixation. Design Studies, 12(1),
3e11. https://doi.org/10.1016/0142-694X(91)90003-F.
Kelly, J., Papalambros, P. Y., Seifert, C. M. (2008). Interactive genetic algo-
rithms for use as creativity enhancement tools. In AAAI spring symposium:
Creative intelligent systems.
Khan, S., Awan, M. J. (2018). A generative design technique for exploring
shape variations. Advanced Engineering Informatics, 38, 712e724. https://
doi.org/10.1016/J.AEI.2018.10.005.
Khan, S., Gunpinar, E. (2018). Sampling CAD models via an extended teach-
ingelearning-based optimization technique. Computer-Aided Design, 100,
52e67. https://doi.org/10.1016/j.cad.2018.03.003.
Kielarova, S. W., Pradujphongphet, P., Bohez, I. L. J. (2013). An approach of
generative design system: Jewelry design application. In IEEE International
Conference on Industrial Engineering and Engineering Management (pp.
1329e1333). https://doi.org/10.1109/IEEM.2013.6962626.
Kielarova, S. W., Pradujphongphet, P., Bohez, E. L. J. (2015). New interactive-
generative design system: Hybrid of shape grammar and evolutionary design -
an application of jewelry design. In Advances in swarm and computational intel-
ligence (pp. 302e313). Springer. https://doi.org/10.1007/978-3-319-20466-
6_33.
Kielarova, S. W., Sansri, S. (2017). Shape optimization in product design using
interactive genetic algorithm integrated with multi-objective optimization. In
Multi-disciplinary trends in artificial intelligence, Vol. 10607 (pp. 76e86).
https://doi.org/10.1007/978-3-319-69456-6.
Kim, H.-S., Cho, S.-B. (2000). Application of interactive genetic algorithm to
fashion design. Engineering Applications of Artificial Intelligence, 13(6),
635e644. https://doi.org/10.1016/S0952-1976(00)00045-2.
Krish, S. (2011). A practical generative design method. Computer-Aided Design,
43(1), 88e100. https://doi.org/10.1016/j.cad.2010.09.009.
Lee, H. C., Herawan, T., Noraziah, A. (2012). Evolutionary grammars based
design framework for product innovation. Procedia Technology, 1, 132e136.
https://doi.org/10.1016/j.protcy.2012.02.026.
Lee, H., Tang, M. (2006). Generating stylistically consistent product form de-
signs using interactive evolutionary parametric shape grammars. In 2006 7th
international conference on computer-aided industrial design and conceptual
design (pp. 1e6). Hangzhou, China: IEEE. https://doi.org/10.1109/
CAIDCD.2006.329391.
Lim, S., Prats, M., Jowers, I., Chase, S., Garner, S., Mckay, A. (2008). Shape
exploration in design: Formalising and supporting a transformational process.
International Journal of Architectural Computing, 6(4), 415e433. https://
doi.org/10.1260/147807708787523303.
Lin, M.-H., Lee, L.-C. (2013). An experimental study for applying generative
design to electronic consumer products. Part IV LNCS, 8015, 392e401.
https://doi.org/10.1007/978-3-642-39253-5_43.
Liu, X., Li, Y., Pan, P., Li, W. (2011). Research on computer-aided creative
design platform based on creativity model. Expert Systems with Applications,
38(8), 9973e9990. https://doi.org/10.1016/J.ESWA.2011.02.032.

Liu, H., Tang, M., Frazer, J. H. (2004). Supporting creative design in a visual
evolutionary computing environment. Advances in Engineering Software, 35(5),
261e271. https://doi.org/10.1016/j.advengsoft.2004.03.006.
Lo, C.-H., Ko, Y.-C., Hsiao, S.-W. (2015). A study that applies aesthetic theory
and genetic algorithms to product form optimization. Advanced Engineering
Informatics, 29(3), 662e679. https://doi.org/10.1016/J.AEI.2015.06.004.
Lubart, T. (2005). How can computers be partners in the creative process: Clas-
sification and commentary on the Special Issue. International Journal of
Human-Computer Studies, 63(4e5), 365e369. https://doi.org/10.1016/
j.ijhcs.2005.04.002.
Machado, T. L. D. A., Gomes, A. S., Walter, M. (2009). A comparison study:
Sketch-based interfaces versus WIMP interfaces in three dimensional modeling
tasks. Latin American Web Congress, 2009, 29e35. https://doi.org/10.1109/
LA-WEB.2009.22.
McCormack, J. P., Cagan, J., Vogel, C. M. (2004). Speaking the Buick lan-
guage: Capturing, understanding, and exploring brand identity with shape
grammars. Design Studies, 25(1), 1e29. https://doi.org/10.1016/S0142-
694X(03)00023-1.
McKay, A., Ang, M. C., Chau, H. H., de Pennington, A. (2006). Combining
evolutionary algorithms and shape grammars to generate branded product
design. In Design computing and cognition ’06 (pp. 521e539). Dordrecht:
Springer Netherlands. https://doi.org/10.1007/978-1-4020-5131-9_27.
Mitchell, W. J. (2005). Constructing complexity. In Computer aided architectural
design futures 2005 - proceedings of the 11th international CAAD futures confer-
ence (pp. 41e50).
Mok, P. Y., Xu, J., Wang, X. X., Fan, J. T., Kwok, Y. L., Xin, J. H. (2013). An
IGA-based design support system for realistic and practical fashion designs.
CAD Computer Aided Design, 45(11). https://doi.org/10.1016/
j.cad.2013.06.014.
Obrenovic, Z. (2009). Sketching and creativity. Creativity.
Olsen, L., Samavati, F. F., Sousa, M. C., Jorge, J. A. (2009). Sketch-based
modeling: A survey. Computers Graphics, 33(1), 85e103. https://doi.org/
10.1016/j.cag.2008.09.013.
O’Neill, M., Swafford, J. M., Hemberg, E., McDermott, J., Byrne, J.,
Brabazon, A. (2010). Evolutionary design using grammatical evolution and
shape grammars: Designing a shelter. International Journal of Design Engineer-
ing, 3(1).
Prats, M., Lim, S., Jowers, I., Garner, S. W., Chase, S. (2009). Transforming
shape in design: Observations from studies of sketching. Design Studies,
30(5), 503e520. https://doi.org/10.1016/j.destud.2009.04.002.
Pugliese, M. J., Cagan, J. (2002). Capturing a rebel: Modeling the Harley-
Davidson brand through a motorcycle shape grammar. Research in Engineer-
ing Design. https://doi.org/10.1007/s00163-002-0013-1.
Rahimian, F. P., Ibrahim, R. (2011). Impacts of VR 3D sketching on novice
designers’ spatial cognition in collaborative conceptual architectural design.
Design Studies, 32(3), 255e291. https://doi.org/10.1016/j.destud.2010.10.003.
Renner, G., Ek
art, A. (2003). Genetic algorithms in computer aided design.
Computer-Aided Design, 35(8), 709e726. https://doi.org/10.1016/S0010-
4485(03)00003-4.
Robertson, B. F., Radcliffe, D. F. (2009). Impact of CAD tools on creative
problem solving in engineering design. Computer-Aided Design, 41(3),
136e146. https://doi.org/10.1016/j.cad.2008.06.007.

Rodrigues, E., Amaral, A. R., Gaspar, A. R., Gomes,
A. (2015). An approach
to urban quarter design using building generative design and thermal perfor-
mance optimization. Energy Procedia, 78, 2899e2904. https://doi.org/
10.1016/j.egypro.2015.11.662.
Ruiz-Montiel, M., Boned, J., Gavilanes, J., Jim
enez, E., Mandow, L., P
erez-de-
la-Cruz, J.-L. (2013). Design with shape grammars and reinforcement learning.
Advanced Engineering Informatics, 27(2), 230e245. https://doi.org/10.1016/
j.aei.2012.12.004.
Samavati, L. O. F. F., Sousa, M. C., Jorge, J. (2008). A taxonomy of modeling
techniques using sketch-based interfaces. In Eurographics 2008 (pp. 39e57).
https://doi.org/10.2312/egst.20081044.
Schon, D. A., Wiggins, G. (1992). Kinds of seeing and their functions in
designing. Design Studies, 13(2), 135e156. https://doi.org/10.1016/0142-
694X(92)90268-F.
Selker, T. (2005). Fostering motivation and creativity for computer users. Interna-
tional Journal of Human-Computer Studies, 63(4e5), 410e421. https://doi.org/
10.1016/J.IJHCS.2005.04.005.
Shea, K., Aish, R., Gourtovaia, M. (2005). Towards integrated performance-
driven generative design tools. Automation in Construction, 14(2), 253e264.
https://doi.org/10.1016/j.autcon.2004.07.002.
Shieh, M.-D., Li, Y., Yang, C.-C. (2018). Comparison of multi-objective evolu-
tionary algorithms in hybrid Kansei engineering system for product form
design. Advanced Engineering Informatics, 36, 31e42. https://doi.org/10.1016/
J.AEI.2018.02.002.
Singh, V., Gu, N. (2012). Towards an integrated generative design framework.
Design Studies, 33(2), 185e207. https://doi.org/10.1016/
J.DESTUD.2011.06.001.
van Sommers, P. (1984). Drawing and cognition: Descriptive and experimental
studies of graphic production processes. Cambridge University Press.
Stiny, G., Gips, J. (1972). Shape grammars and the generative speciﬁcation of
painting and sculpture. In Information processing 71 proceedings of the IFIP
congress, Vol. 2 (pp. 1460e1465).
Stones, C., Cassidy, T. (2010). Seeing and discovering: How do student de-
signers reinterpret sketches and digital marks during graphic design ideation?
Design Studies, 31(5), 439e460. https://doi.org/10.1016/j.destud.2010.05.003.
Sun, J., Frazer, J. H., Mingxi, T. (2007). Shape optimisation using evolutionary
techniques in product design. Computers Industrial Engineering, 53(2),
200e205. https://doi.org/10.1016/J.CIE.2007.06.010.
Sun, X., Gong, D., Zhang, W. (2012). Interactive genetic algorithms with large
population and semi-supervised learning. Applied Soft Computing, 12(9),
3004e3013. https://doi.org/10.1016/J.ASOC.2012.04.021.
Suwa, M., Tversky, B. (1997). What do architects and students perceive in their
design sketches? A protocol analysis. Design Studies, 18(4), 385e403. https://
doi.org/10.1016/S0142-694X(97)00008-2.
Su, J., Zhang, S. (2010). Research on product shape innovation design method
with human-computer interaction through genetic algorithm. In 2010 IEEE
11th international conference on computer-aided industrial design conceptual
design (pp. 301e305).
Tovey, M. (1997). Styling and design: Intuition and analysis in industrial design.
Design Studies, 18(1), 5e31. https://doi.org/10.1016/S0142-694X(96)00006-3.
Tovey, M., Porter, S., Newman, R. (2003). Sketching, concept development and
automotive design. Design Studies, 24(2), 135e153. https://doi.org/10.1016/
S0142-694X(02)00035-2.

Troiano, L., Birtolo, C. (2014). Genetic algorithms supporting generative
design of user interfaces: Examples. Information Sciences, 259, 433e451.
https://doi.org/10.1016/j.ins.2012.01.006.
Turrin, M., von Buelow, P., Stouffs, R. (2011). Design explorations of perfor-
mance driven geometry in architectural design using parametric modeling and
genetic algorithms. Advanced Engineering Informatics, 25(4), 656e675. https://
doi.org/10.1016/j.aei.2011.07.009.
Van Elsas, P., Vergeest, J. (1998). New functionality for computer-aided con-
ceptual design: The displacement feature. Design Studies, 19(1), 81e102.
https://doi.org/10.1016/S0142-694X(97)00016-1.
Vasconcelos, L. A., Crilly, N. (2016). Inspiration and fixation: Questions,
methods, findings, and challenges. Design Studies, 42, 1e32. https://doi.org/
10.1016/J.DESTUD.2015.11.001.
Vuletic, T., Duffy, A., Hay, L., McTeague, C., Pidgeon, L., Grealy, M. (2018).
The challenges in computer supported conceptual engineering design. Com-
puters in Industry, 95, 22e37. https://doi.org/10.1016/
J.COMPIND.2017.11.003.
Wang, K., Nickerson, J. V. (2017). A literature review on individual creativity
support systems. Computers in Human Behavior, 74, 139e151. https://doi.org/
10.1016/J.CHB.2017.04.035.

A 3D Shape Generative Method For Aesthetic Product Design

More Related Content

Similar to A 3D Shape Generative Method For Aesthetic Product Design (20)

More from Martha Brown (20)

Recently uploaded (20)

A 3D Shape Generative Method For Aesthetic Product Design