GA-Based Design Exploration of a House: An Interactive Computational Approach

Abstract

A design task usually begins with a phase where requirements and objectives are defined. Then, it continues through a creative and iterative process of generation, evaluation, and modification of design alternatives until the design team is satisfied. The main objective of this paper is to expand the concepts of interactive multi-objective design in the field of architecture design, and demonstrate the contribution of parametric modeling, genetic algorithms, computational simulation and an interactive programming environment in the process of decision making in the preliminary stage of architecture design. In this paper, an interactive GA-based computational method is employed to consider both technical and non-technical requirements in exploring the design solutions of a one-bedroom house unit. The design objectives, in this case, include thermal performance, material cost, and the resident’s lifestyle. Ultimately, it is shown how the interactive approach of the exploration method can help the decision maker with objective trade-offs.

Full text

The International Journal of Design Management and Professional Practice
Volume 11, Issue 3, 2017, www.designprinciplesandpractices.com
© Common Ground Research Networks, Anahita Khodadadi, All Rights Reserved
Permissions: support@cgnetworks.org
ISSN: 2325-162X (Print), ISSN: 2325-1638 (Online)
http://doi.org/10.18848/2325-162X/CGP/v11i03/13-25 (Article)
GA-Based Design Exploration of a House:
An Interactive Computational Approach
Anahita Khodadadi,
1
University of Michigan, USA
Abstract: A design task usually begins with a phase where requirements and objectives are defined. Then, it continues
through a creative and iterative process of generation, evaluation, and modification of design alternatives until the
design team is satisfied. The main objective of this paper is to expand the concepts of interactive multi-objective design in
the field of architecture design, and demonstrate the contribution of parametric modeling, genetic algorithms,
computational simulation and an interactive programming environment in the process of decision making in the
preliminary stage of architecture design. In this paper, an interactive GA-based computational method is employed to
consider both technical and non -technical requirements in exploring the design solutions of a one-bedroom house unit.
The design objectives, in this case, include thermal performance, material cost, and the resident ’s lifestyle. Ultimately, it
is shown how the interactive approach of the exploration method can help the decision maker with objective trade-offs.
Keywords: Genetic Algorithm, Interactive Approach, Performance Based Design, Building Simulation
Introduction
esign tasks usually begin with a conceptual design phase where requirements and
objectives are defined and synthesized into design alternatives (Pahl, et al. 2007). In a
traditional approach, the decision making is mainly based on the designer’s insight, basic
functions, and aesthetical aspects. Limited range of performances are evaluated, and this
increases the amount of post-engineering adjustments in the phase of design development
(Turrin, von Buelow and Stouffs 2011). Wang believes 75% of the product life-cycle cost is
determined during the preliminary phase of design (Wang 2002). Similar statements highlighted
the importance of numerical evaluation of the building performance in the conceptual design
phase. Thus, Computer Aided Conceptual Design approach came into play to prevent designers
from investment in poor design alternatives (Turrin, von Buelow and Stouffs 2011). Several
parametric tools and computational exploration methods such as Rhino plugins,
GenerativeComponents, StructuralComponents, and GeometryGym have been developed to
support the designer considering multiple objectives such as structural and environmental
performance, costs and architectural proportions in the early design stage (McNeel Europe.
2017a.; Geometry Gym Pty Ltd. 2017; Rolvink, Breider, and Coenders 2010; Bentley Systems
Inc. n.d.). Furthermore, some conceptual design tools provide a repeated cycle of divergence and
convergence to generate the widest possible range of alternatives and, at the same time, make it
possible to meaningfully manage the search space (Liu Y. C., Bligh and Chakrabarti 2003; Cross
1994; Pugh 1991). There, usually, the link of parametric modeling and evolutionary search
algorithms are employed. Parametric modeling, based on the consistent structure of dependencies
within a problem, allows generating a population of solutions conveniently and relatively fast.
An evolutionary algorithm, mainly a genetic algorithm (GA), provides the iterative process of
form generation, evaluation of solutions using a “fitness function,” modification of the
population until the designer is satisfied (Mendez Echenagucia 2013).
In this paper, the link of parametric modeling, computational simulations and a genetic
algorithm is represented through the early design stage of a one-bedroom housing unit. The focus
of the study is on the interactive procedure of form exploration that allows consideration of both
technical and non-technical requirements. In the following sections, first, the geometric
1
Corresponding Author: Anahita Khodadadi, 2000 Bonisteel Blvd., Taubman College of Architecture and Urban
Planning, University of Michigan, Ann Arbor, Michigan, 48109-2069, USA. email: anahitak@umich.edu
D

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
parametrization of the problem is described. Then, it is explained how the parametric model
supports the evaluation of performances including thermal demands and material cost.
Second, an interactive GA-based form exploration method called ParaGen, initially introduced
by Peter von Buelow (2007), is represented. In the section following that, it is discussed how
interactive approaches, such as ParaGen, assist the designer to make trade-off decisions among
various preferences.
Parametric Modeling
Figure 1: The Configuration of the House Unit
Source: Generated by the Khodadadi
14

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
A house is not merely a physical space. Its layout and configuration express “the meaning of
a home as a social unit” (Rapoport 1969). The house in this design task is to be occupied by a
maximum of three residents. The geometric configuration processing of the house unit is carried
out using Grasshopper, a plugin for Rhino (see Figure 1) (McNeel Europe. 2017b). Meanwhile,
DIVA plugin for Grasshopper, see (Solemma n.d.), and RS Means Cost Works, see
(RSMeansOnline 2015), are utilized to accomplish the thermal analyses and cost estimation
respectively. To run the thermal analysis using the DIVA plugin, the occupancy is set to be
residential. The weather file is based on Chicago, IL, OHare.Intl.AP data. The run period is
chosen to be annually, and the energy demands for heating, cooling, interior artificial lighting
and the total energy demand are the outputs.
Table 1: The Values of Some Parameters Used in Thermal Analysis of This Project
Parameters
Values
Acceptable Interval
The density of the people in the room:
Number of Residents ÷ Floor Area
Number of residents = [1,3]
Internal heat gain from light (W/m²)
13
10–18
Equipment load: Internal heat gain from
equipment (W/m²)
9
5–12
Set point heating
20 C
Set point cooling
26 C
Htg COP (COP or efficiency heating unit,
heat pump 3.5, boiler 0.95)
0.95
Clg COP (COP or efficiency cooling unit,
compression chiller 2.5)
2.5
Infiltration rate building envelope, Air
change per hour ACH
0.2
0.1–0.5
Fresh air supply mechanical ventilation
0.002
Source: Generated by the Khodadadi
Table 2: The Building Components Material and Their Assembly
Building Components
Assembly
Roof Assembly
Gravels = 2 cm
Plywood or wood panels = 0.75cm
Insulation board = 2 cm
Plaster: Gypsum board = 0.5 cm
Floor Assembly
Carpet = 0.25 cm
Lightweight concrete = 6 cm
Stone, sand and gravel 2240 kg/m³ = 4 cm
Façade Assembly
Wood siding = 0.5 cm
Plywood or wood panels = 0.5 cm
Batt Insulation = 5.5 cm
Plaster: Gypsum board = 0.5 cm
Windows
Glass material with 0.4 (Btu/hft²°F) U-value, 0.5 Solar Heat Gain Coefficient (min=0 and
max=1) and 0.5 Visual Transmittance (min=0 and max=1).
Source: Generated by the Khodadadi
The values of parameters required for thermal analysis were set in the Grasshopper model,
and the building components assembly (Ching 2001) is described in table 1 and table 2
respectively. Using the RSMeans Cost Work 2015 database, the Cost Data is set to “Residential
Square Foot Assembly,” the Labor Type and the Location are chosen to be “Residential” in
Chicago. There are many items such as site work, excavation, constructing the utility trench,
septic, interior fixtures and services, plumbing, HVAC, electrical, and protection that should be
considered to compute the actual building cost. In this study, only the costs of the items that are
15

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
dependent on the areas of the unit components are taken into account. Those items that cost the
same in all the units, such as the cost of the two interior doors, are also excluded. Similar to
many parametric models, the independent values of the variables in this problem are inputs, and
dependent variables are computed by receiving data from their related attributes, see (Turrin,
Kilian, and Sariyildiz 2009) for further discussion in this field.
ParaGen: A GA-Based Form Exploration Method
Within the preliminary phase of design, the designer is more engaged with reasoning,
argumentation, and decision-making (Rolvink, Mueller and Coenders 2014). Similar to sketching
on a tracing paper and examining different layouts of spaces, the designer needs to gather
information about the performance of each design alternative presenting to the client. Therefore,
some models are needed to be generated, evaluated and modified through an iterative procedure
to approach finally the ideal solutions that meet the design objectives. The described procedure
demands a considerable amount of time and effort and may cause exhaustion, mind blocking and
stop the designer exploring further possible solutions (Cross 2004). Hence, computational
research tools are developed to be utilized within the design process. They can facilitate
evaluation of a greater number of alternative solutions in diverse aspects, and allow the designer
to explore more possibilities.
Among an immense number computational search tools and search algorithms, evolutionary
algorithms can be traced back to the early 1950s when they were used by several biologists.
Within 1960 and 1970 genetic algorithms were developed at University of Michigan under the
direction of John Holland (Dasgupta and Michalewicz 1997). In 2006, at the World Congress on
Computational Intelligence, evolutionary multi-objective optimization (EMO) was recognized as
one of the three fastest growing research fields among all computational intelligence topics (Deb
2008). The reasons for this popularity were mentioned to be:
 No requirement for derivative information;
 Relative simplicity of implementation;
 Flexibility and a widespread applicability.
A genetic algorithm is population-based search method, where stochastic operators are utilized to
generate multiple good solutions. The variables of the parametric model are analogous to genes
and chromosomes. Mechanisms such as selection, recombination, and mutation are employed to
generate solutions and transmit the characteristics of a pair of solutions to the new offspring (Deb
2008). In this research, a Non Destructive Dynamic Population GA (NDDP GA) is used (von
Buelow, 2012). In an NDDP GA, the selection takes place at the population level. A population
of solutions is selected from a database of all current solutions using a SQL query. Next, two
parent solutions are randomly picked from the selected population to generate a new child
solution. The goal is to find an array of good solutions. In multi-objective problems, there is no
single “best” solutions. “Good” may be found on a Pareto front or in clusters. ParaGen is well
suited to explore these sets of good solutions using post processing graphic tools.
In this study, the ParaGen method is used to accomplish the form generation and
exploration. ParaGen steps of form exploration can be described as follows (von Buelow 2012)
(Khodadadi and von Buelow 2014):
1. The design problem is described in terms of parametric variables: a “chromosome.” The
house unit configuration is described in 27 geometrical variables, and several
parameters define its thermal performance and material cost. All these parameters build
a solution’s characteristics;
2. Initial, random values are assigned to the geometrical variables. An initial pool of
solutions is generated, and after evaluating their performance, all parameters of each
solution are stored in a database on the server;
16

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
3. A population of parents is dynamically pulled from the solution pool using the
determined criteria: the fitness function in the form of a SQL query;
4. Two parents are randomly chosen from the selected population;
5. Based on the HUX method (Syswerda 1989), a child is bred from the parents and may
inherit some characteristics of one parent or that of the other one. It is also possible that
a combination of some characteristics of both emerges for the child's chromosome. The
HUX principles are shown in figure 2. (Syswerda 1989).
Figure 2: A Schematic Demonstration of Half Uniform Crossover (HUX).
Source: Khodadadi, von Buelow 2014
6. The chromosome (the values of the 27 geometrical values) of the child is processed into
a geometric solution (in Rhino + Grasshopper);
7. The performance of the solution is evaluated using various simulation software. In this
study, DIVA plugin in Grasshopper is used to evaluate the thermal performance. Since
the parametric model is built in the Grasshopper environment, using the DIVA plugin
for Grasshopper facilitates the process and saves time. The material cost is also
estimated using an equation in an Excel spreadsheet;
8. The resulting performance values, which are the thermal and lighting outcomes, along
with the geometric values, are uploaded to the database. Images are also included and
linked to the solution.
Through an iterative process, steps 3 through 8 will take place until satisficing solutions are
found. The solution database and design interface are located on an internet server whose
interface is a web site. The designer takes the first step and defines the configuration of the house
unit parametrically. In step 2, a C code along with a PHP script on the server are used to generate
random values within the acceptable range of each variable. When these values are assigned to a
solution, its thermal performance and material cost are evaluated, and along with the geometrical
characteristics are stored in the database. In fact, step 2 repeats a sufficient number of times to
create the initial pool of solutions. Steps 3, 4 and 5 are taken using the C and PHP codes. Figure
3 illustrates the process of form exploration of the housing unit.
ParaGen is based on a Non-Destructive Dynamic Population GA approach. None of the
defective or poor performing solution is removed (killed off) from the database. Therefore, if the
design criteria are changed, the solutions previously known as defective and poor ones may also
come into play. Also, ill solutions may inform the designer as to what makes a good solution
(von Buelow 2013).
17

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
Figure 3: A Scheme of the Paragen Method Used for Computational Search in Architecture Design of a Housing Unit
Source: Generated by the Khodadadi
Interactive Exploration Approach
Within the design task, there are some subjective goals, such as beauty and peace, that cannot be
parameterized. Also, the suitability of the solutions regarding some subjective preferences cannot
be measured or numerically evaluated. One person may find a certain amount of area small, but
another one may not. Hence, it is a privilege for an exploration method if the decision maker
(DM), who might be the designer, the user or the client, can progressively be engaged with the
exploration process, and specify and express a preference at each iteration. Therefore, some
techniques such as visualization are developed to help (Miettinen, Ruiz and Wierzbicki 2008).
The ParaGen method provides this desirable interaction in three ways. First, it allows the
designer or decision maker to specify or adjust a preference, and (re)define the fitness function at
the beginning of each set of iterations. Second, the ParaGen web interface allows the user to filter
and sort the solutions based on any combination of geometry or performance data (von Buelow
2012). Similar to utilizing a telescope to observe one or some specific stars, the ParaGen web
filters can be applied to focus on a well-performing set of solutions. Thus, the designer can
inspect this manageable quantity of filtered solutions and make selections for further breeding if
18

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
it is required. Third, ParaGen improves the readability of the information by two visualization
techniques:
1. The web interface provides both images and key performance values associated with
each solution (von Buelow 2012). The images may display the geometrical
configuration of the solutions or diagrams indicating some certain performance
assessments (see Figure 4).
2. The ParaGen website provides post processing though scatter point graphs of solution
clusters or Pareto fronts and parallel point graphs. These tools combined with graphic
image arrays as shone in Figure 4, help the user explore the most desirable solutions
(von Buelow 2012). In a graph, the points shown represent solutions that have been
filtered by a certain set of criteria.
Figure 4: The Set of Solutions That Have a Maximum 30 M2 Living Room Area
and a Maximum 6 M2 Living Room Window
Source: Generated by the Khodadadi
Discussion about Trade-off and Decision Making
Through a multi-objective design task a set of alternatives will be provided, then the designer
makes the final decision and chooses the most preferred one. Within the decision-making
process, there is usually a trade-off among the design objectives, which is an exchange of loss on
one aspect and gained benefits in another aspect (Miettinen, Ruiz and Wierzbicki 2008). The
Pareto optimal front allows for studying the suitability of solutions regarding some quantitative
values. As an example, the Average Heating Energy demand of the units designed for two
residents and the corresponding values of the Material Cost may be studied. Figure 5 displays a
plot where all the solutions with any amount of total floor area are included. Then, one may
search for a solution that is less expensive regarding Material Costs and, at the same time, has
less Average Heating Energy demand. It is possible to study other properties of each design
alternative such as total floor area while clicking on the corresponding point on the graph and
choose the most suitable solution.
In this paper, the emphasis is on the benefits of the interactive approach of ParaGen that
enable the designer to make a better decision regarding the qualitative design objectives. User’s
lifestyle and personal preferences may vary in each design case. Thus, a design solution that is
19

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
the most suitable for one may be inappropriate for the other. Among the provided solutions in the
given example, one scenario might be a couple who enjoy serving their guests at home and desire
a house with a large living room of at least 40 m2 area. Additionally, they may prefer to have as
much daylight as possible, but not be exposed to the outside. They may also think of having a
large balcony where they can stand or sit and look at the garden. They may look for a house with
less average total energy demand and set 26000 kWh as a filter to find a good solution. Also, the
material cost, as a portion of total cost of the house may be critical for them, and 65000 USD
might be set as a limit. Table 3 shows the suitable filtering criteria and Figures 6 and 7 display a
suitable solution regarding the described lifestyle and preferences.
Figure 5: The Average Heating Energy Demand vs. the Material Cost of the Units Designed for Two Residents
Source: Generated by the Khodadadi
Table 3: The Set of Filters which Include Performance Values and Lifestyle Considerations
Properties
Acceptable value
LivingRoomArea
>
40 m2 /
323 ft2
Area of living room window
=>
2.5 m2 /
6.5 ft2
Z coordinate of the living room window
>
60 cm /
2 ft
Balcony depth
>
50 cm /
21 ft2
Bathroom area
>
4 m2 /
2 ft
Average Total Energy Demand
<
26000 kWh
Material cost
<
65000 USD
Source: Generated by the Khodadadi
20

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
Figure 6: The Furnished Plan of the Suitable Unit
Source: Generated by the Khodadadi
Figure 7: Solution 374, the Suitable Solution
Source: Generated by the Khodadadi
Figure 8: The Average Total Energy vs. Material Cost—Solution 374 is Highlighted in the Graph.
Source: Generated by the Khodadadi
Figure 8 illustrates the preferred solution, with the ID 374. The graph presents the Average
Total Energy vs. Material Cost among the house units with a living room area greater than 40 m2
and minimum 50 cm balcony depth. The graph indicates that the solution is significantly energy
efficient and, also, affordable for the residents. On the other hand, other residents may have
different preferences. They may not have large gatherings in their house and prefer a unit with a
smaller living room that needs less effort to be cleaned up and arranged. Also, they may look for
a unit with less exposure to outside view. Then, filters may be determined differently, and a set
of alternatives displayed in Figure 4 may be more suitable for them.
If a specific filter doesn’t provide any solution, or a broader variety is demanded, it is
possible to set the fitness function to minimize or maximize a particular property. For example,
one may set the fitness function to minimize the cost and restrict the Average Heating Energy of
the newly generated solutions between 100 and 3000 kWh. In Figure 9, the solutions which
emerged after 6 iterations are highlighted.
It is also possible to search for solutions using Pareto sets. A Pareto set itself can be simply
used as the selection population for breeding new solutions. This allows to focus on certain areas
of the set and generate more new solutions within limiting criteria (von Buelow 2017; Emami,
21

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
Giles and von Buelow 2017). Figure 10 illustrates a Pareto front (level 1 in a red) where the total
area of windows and the total energy demand are compared. If one may want to search more
solutions with less total energy demand and greater window area, this Pareto set can be used in
the breeding population. The width of the solutions set can also be increased to include more
solutions (see Figure 11).
Figure 9: The Generated Solutions with Minimized Cost and the
Average Heating Energy is between 100 and 3000kWh
Source: Generated by the Khodadadi
Conclusion
In short, in multi-objective design projects where building performance, aesthetic issues,
stability, costs and construction requirements and designer’s or user’s preferences should be
considered, the evolutionary form exploration method seems to be a helpful means of decision
making. The application of a genetic algorithm through the form generation and exploration
process allows setting the fitness function regarding the design objectives and achieving better
solutions in few runs.
This research work contributes to the development of the existing method, ParaGen. This
design search method is based on providing several suitable solutions instead of one single best
solution. Then, it is possible to consider subjective issues and make appropriate decisions. In this
paper, the benefits of ParaGen interactive approach have been discussed. Accordingly, a one-
bedroom housing prototype was considered, and both quantitative and qualitative design
objectives have been taken into account. It is represented how the Pareto front assists the
decision maker to study the design alternatives and choose the suitable one regarding the
quantitative objectives. Then, the benefits of ParaGen interactive approach, where qualitative
objectives may vary from one resident to the other, have been discussed. The decision maker,
who may be the designer or user, is able to provide information about one’s preferences to set the
fitness functions at the beginning of each iteration. Later, when several possibilities have been
explored, and the design alternatives are generated, the ParaGen visual interface with images
attached to each solution, as well as the provided graphs, allow the designer to consider the
qualitative objectives and to facilitate the objective trade-offs. The strength of ParaGen is that it
is user-friendly enough to facilitate a suitable interaction between the decision maker and the
22

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
automated form exploration process. Albeit, the decision maker should either have programming
knowledge or be assisted in setting up the parametric model of the problem. In addition, the
algorithm is flexible for further modifications due to the designer’s aspiration. Furthermore,
similar to other interactive search and optimization methods, the designer also can learn about the
interdependencies of the design parameters and obtain a realistic expectation of final solution.
As the last notation, further work is required in the development of multi-objective design
tasks where both quantitative and qualitative objectives are considered. The focus of this paper
has been on the application of an interactive approach in decision making. Additional studies can
be carried out about the possibility of parametrization of some socio-cultural aspects of design
such as lifestyle.
Figure 10: Windows Area vs. Maximum of Total Energy Demand: A Plot Showing Pareto Frontier Level 1 Set in Red
Source: Generated by the Khodadadi
Figure 11: Windows Area vs. Maximum of Total Energy Demand: A plot Showing Pareto Set Including Level New
Generated Solution Using Pareto Levels 1–3 Set in Red
Source: Generated by the Khodadadi
23

THE INTERNATIONAL JOURNAL OF DESIGN MANAGEMENT AND PROFE SSIONAL PRACTICE
Acknowledgement
My thanks to Professor Peter von Buelow at the University of Michigan for all his support and,
also, for providing the opportunity to utilize ParaGen tool to accomplish this study.
REFERENCES
Bentley Systems Inc. n.d. GenerativeComponents: Design and Explore the Unimaginable.
Accessed May 24, 2017. https://www.bentley.com.
Ching, Francis. D. K. 2001. Building Construction Illustrated. 5th. New York: John Wiley.
Cross, Nigel. 1994. Engineering Design, Methods-Strategies for Product, Design. Chichester:
John Wiley & Sons.
Cross, Nigel. 2004. “Expertise in Design: An Overview.” Design Studies 25 (5): 427–441.
Dasgupta, Dipankar, and Zbigniew Michalewicz. 1997. Evolutionary Algorithms in Engineering
Applications. Milton Keynes, UK: The Open University.
de Chiara, Joseph, and Mike Crosbie. 2012. Time-Saver Standards for Building Types. 4th.
New York: McGraw-Hill Professional Publishing.
Deb, Kalyanmoy. 2008. “Introduction to Evolutionary Multiobjective Optimization.” In
Multiobjective Optimization: Interactive and Evolutionary Approaches, edited by
Jürgen Branke, Kalyanmoy Deb, Kaisa Miettinen, and Roman Słowinski, 59–96.
London: Springer.
Emami, Niloufar, Harry Giles, and Peter von Buelow. 2017. “Continuous to Discrete:
Computational Performative Design and Search of Shell Structures.” Proceedings of the
IASS Annual Symposium 2017 “Interfaces: Architecture, Engineering, Science.”
Hamburg, Germany.
Geometry Gym Pty Ltd. 2017. GeometryGym: OpenBIM tools for Architects, Engineers and the
Construction Industry. Accessed May 24, 2017. https://geometrygym.wordpress.com.
Holland, John Henry. 1975. Adaptation in Natural and Artificial Systems. Ann Arbor, Michigan:
The University of Michigan Press.
Khodadadi, Anahita, and Peter von Buelow. 2014. “Performance Based Exploration of
Generative Design Solutions Using Formex Algebra.” International Journal of
Architectural Computing 12 (3): 321–338.
Liu Y. C., T. Bligh, and Amaresh Chakrabarti. 2003. “Towards an ‘Ideal’ Approach for Concept
Generation.” Design Studies 24 (4): 341–355.
McNeel Europe. 2017a. Food4Rhino: Apps for Rhino and Grasshopper. Accessed May 24, 2017.
http://www.food4rhino.com.
McNeel Europe. 2017b. Grasshopper Extension for Rhinoceros. Seattle, Washington.
Mendez Echenagucia, Tomas. 2013. Computational Search in Architectural Design. A thesis
submitted for the degree of “Doctor of Philosophy,” Torino: Politecnico di Torino.
Miettinen, Kaisa, Francisco Ruiz, and Andrzej P. Wierzbicki. 2008. “Introduction to
Multiobjective Optimization: Interactive Approaches.” In Multiobjective Optimization:
Interactive and Evolutionary Approaches, edited by Jürgen Branke, Kalyanmoy Deb,
Kaisa Miettinen and Roman Słowinski, 27–57. London: Springer.
Miettinen, Kaisa, Kalyanmoy Deb, Johannes Jahn, Wlodzimierz Ogryczak, Koji Shimoyama,
and Rudolf Vetschera. 2008. “Future Challenges.” In Multiobjective Optimization:
Interactive and Evolutionary Approaches, edited by Jürgen Branke, Kalyanmoy Deb,
Kaisa Miettinen and Roman Słowinski, 435–61. London: Springer.
Neufert, Ernst, and Peter Neufert. 2012. Neufert Architect’s Data. 4th. New Jersey:
Wiley-Blackwell.
24

KHODADADI: GA-BASED DESIGN EXPLOR ATION OF A HOUSE
Pahl, Gerhard, W. Beitz, Jörg Feldhusen, and Karl-Heinrich Grote. 2007. Engineering Design: A
Systematic Approach. London: Springer.
Pugh, S. 1991. Total Design: Integrated Methods for Successful Product Engineering.
Wokingham: Addison Wesley.
Rapoport, Amos. 1969. House Form and Culture. Englewood Cliffs: Prentice-hall.
Rolvink, Anke, Caitlin Mueller, and Jeroen Coenders. 2014. “State on the Art of Computational
Tools for Conceptual Structural Design.” IASS-SLTE 2014 Symposium, “Shells,
Membranes and Spatial Structures: Footprints.” Brasilia: International Association for
Shell and Spatial Structures (IASS).
Rolvink, Anke, Janwillem Breider, and Jeroen Coenders. 2010. “Structural Components: A
Toolbox for Conceptual Structural Design.” Proceedings of the International
Association of Bridge and Structural Engineering, 768–769.
RSMeansOnline. 11/20/ 2015. http://rsmeansonline.com/SearchData.
Solemma, LLC. n.d. Diva for Rhino—Environmental Analysis for Buildings.
Syswerda, Gilbert. 1989. “Uniform Crossover in Genetic Algorithms.” Edited by Morgan
Kaufmann. Third International Conference on Genetic Algorithms, 2–9.
Turrin, Michela, Axel Kilian, Rudi Stouffs, and Sevil Sariyildiz. 2009. “Digital Design
Exploration of Structural Morphologies Integrating Adaptable Modules: A Design
Process Based on Parametric Modeling.” Cultures and Visions: CAADFutures.
Montreal: Les Presses de l'Universite de Montreal
Turrin, Michela, Peter von Buelow, and Rudi Stouffs. 2011. “Design Explorations of
Performance Driven Geometry in Architectural Design Using Parametric Modeling and
Genetic Algorithms.” Advanced Engineering Informatics 25 (4): 656–675.
von Buelow, Peter. 2007. An Intelligent Genetic Design Tool (IGDT) : Applied to the
Exploration of Architectural Trussed Structural Systems. Dissertation, University of
Stuttgart.
———. “Choosing Parents to Produce Better Preforming Children: A Comparison of Selection
Methods Used for Evolutionary Search.” 2017. Proceedings of the IASS Annual
Symposium 2017 Proceedings of the IASS Annual Symposium 2017 “Interfaces:
Architecture, Engineering, Science.” Hamburg, Germany.
von Buelow, Peter. 2012. “ParaGen: Performative Exploration of generative systems.” Journal of
the International Association for Shell and Spatial Structures 53 (4): 271–284.
———. “Techniques for more Productive Genetic Design: Exploration with GAs using Non-
Destructive Dynamic Populations.” 2013. ADAPTIVE ARCHITECTURE: Proceedings
of the 33rd Annual Conference of the Association for Computer Aided Design in
Architecture. Cambridge Ontario.
Wang, J. 2002. “Improved Engineering Design Concept Selection Using Fuzzy Sets.”
International Journal of Computer Integrated Manufacturing 15 (1): 18–27.
ABOUT THE AUTHOR
Anahita Khodadadi: Doctoral Candidate, Taubman College of Architecture and Urban Planning,
University of Michigan, Ann Arbor, Michigan, USA
25