Automation in Construction 166 (2024) 105638

Contents lists available at ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Review

Generative Design in the Built Environment

Zhi Xian Chew a , Jing Ying Wong a, * , Yu Hoe Tang a , Chun Chieh Yip b , Tomas Maul c
a
Department of Civil Engineering, University of Nottingham Malaysia, Semenyih 43500, Selangor, Malaysia
b
Department of Civil Engineering, Universiti Tunku Abdul Rahman, Bandar Sungai Long, 43000, Cheras, Kajang, Selangor, Malaysia
c
School of Computer Science, University of Nottingham Malaysia, Semenyih 43500, Selangor, Malaysia

A R T I C L E I N F O A B S T R A C T

Keywords: Generative design (GD) has gained widespread attention in the built environment domain, revolutionising
Generative design traditional methodologies in the field. It operates on a set of rules, utilising algorithmic and parametric modelling
Building information modelling (BIM) techniques to explore and enhance design options. By determining the existing challenges and ongoing research
Sustainability
in 85 representative papers, this paper seeks to explore the implications of GD within built environment, sus-
Artificial intelligence (AI)
tainability, BIM, and AI design practices, clarifying its potential to shape a more sustainable, resilient, and in-
Built environment
clusive built environment for future generations. It contributes to the dialogue surrounding the utilisation of GD
as a tool for building and infrastructure design practices via a content analysis-based review method. Incorpo-
rating advanced technologies could facilitate a more effective and immediate response to design methodologies,
thus promoting enhanced innovation and sustainability in design practices. The selection of, and proficiency in,
programming languages constitute important aspects in advancing GD aligned with built environment objectives.

1. Introduction provide the setting for human activity, including buildings, infrastruc-
ture, and other physical spaces. It integrates various fields such as ar-
Generative design (GD), a paradigm-shifting approach, has been chitecture, urban planning, civil engineering, landscape architecture,
rapidly gaining attention in the built environment domain, revolutio- and transportation engineering, all of which contribute to the design,
nising traditional methodologies within the field. GD, a method within construction, and management of these spaces. For instance, within civil
design exploration, typically involves parametrically varying design engineering, GD holds promise during the design phase, where it in-
geometry and assessing the performance of the resulting output designs volves considerations spanning structural integrity, efficiency, con-
[1]. As the demands on infrastructure continue to evolve, propelled by struction process, and cost-effectiveness. For instance, structural designs
population growth, urbanisation, and environmental concerns, the ne- for buildings primarily depend on drawings that are translated into
cessity for innovative design solutions becomes increasingly pro- structured or image data. Engineers then finalise the design process,
nounced. This paper explores the developing landscape of GD, considering architectural design alongside multiple constraints such as
specifically within building design practices, illuminating its trans- compliance and economy [3]. By utilising the power of generative al-
formative potential across various aspects of the built environment. GD gorithms, engineers can navigate complex design spaces with unprece-
stands at the forefront of innovative approaches to architectural and dented efficiency, optimising designs for performance while adhering to
engineering creativity, utilising algorithms to explore the variants of a regulatory requirements and project constraints.
design beyond what is currently possible using the traditional design As building energy consumption accounts for 30%–40% of the total
process. In a core approach, it emulates nature’s evolutionary process, of society’s energy consumption, there is a great potential for energy-
by using parameters and goals to quickly explore thousands of design saving at the building design stage [4]. Hence, GD also plays a vital
variants to find the best solution [2]. This proactive approach allows for role in promoting sustainability within building design practices. By
real-time decision-making that incorporates considerations of multiple embedding sustainability criteria into the GD process, engineers can
factors such as structural performance, material efficiency, and envi- systematically explore design alternatives that minimise environmental
ronmental impact during the design phase itself. impact, reduce resource consumption, and enhance resilience to climate
The built environment consists of the man-made surroundings that change. Through iterative optimisation, generative design enables the

* Corresponding author at: Department of Civil Engineering, University of Nottingham Malaysia, Semenyih 43500, Selangor, Malaysia
E-mail address: [email protected] (J.Y. Wong).

https://doi.org/10.1016/j.autcon.2024.105638
Received 30 March 2024; Received in revised form 19 July 2024; Accepted 19 July 2024
Available online 27 July 2024
0926-5805/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

creation of environmentally responsible solutions that balance perfor-

mance, cost, and environmental considerations, thereby advancing the
goals of sustainable development.
Moreover, the integration of GD principles with Building Informa-
tion Modelling (BIM) systems further enhances collaboration and facil-
itates data-driven decision-making throughout the project lifecycle. This
integration empowers stakeholders to explore design alternatives
comprehensively, providing a clear understanding of the implications of
each decision on construction, operation, and maintenance. With the
advent and rising popularity of the BIM concept, the infrastructure for
deploying a comprehensive GD approach is increasingly becoming
accessible. BIM models, offering extensive geometric and non-geometric
data on buildings throughout their lifecycle, have proven effective in
addressing complex optimisation challenges in building design [5].
In recent years, the convergence of advanced computational tech-
niques, artificial intelligence (AI), and parametric modelling has cata-
lysed a paradigm shift in design processes. GD transcends conventional
methodologies by employing algorithms to iteratively generate and
evaluate countless design alternatives based on specified criteria and
constraints. This iterative exploration not only accelerates the design
process but also promotes creativity, enabling engineers to conceive
solutions that might otherwise remain undiscovered. As generative AI is
capable of generating informative and suggestive human-like responses
to user inputs, it is necessary to study to what extent generative AI will
be relevant in triggering innovative outcomes [6].
AI augments the capabilities of GD by enabling algorithms to learn
from data and adaptively optimise designs based on evolving project
requirements and constraints. Machine learning (ML) algorithms, inte-
grated into GD workflows, can analyse a wide range of datasets to
identify patterns, predict performance outcomes, and recommend
design strategies that optimise for desired objectives [7]. This symbiotic
relationship between AI and generative design empowers engineers to
Fig. 1. Process of generative design.
tackle increasingly complex design challenges with greater precision,
efficiency, and adaptability.
In summary, GD represents a transformative perspective within the 2.1. Generative design
built environment, offering unprecedented opportunities to reimagine
design practices. By integrating principles of sustainability, BIM, and The groundwork for GD was laid in the 1960s and 1970s when re-
artificial intelligence, GD not only accelerates the design process but also searchers and engineers began exploring the use of computers in design
promotes innovation, collaboration, and sustainability across the entire processes. This era witnessed the initial attempts to apply computational
project lifecycle. This paper seeks to explore the multifaceted implica- methods to generate and optimise designs. Subsequently, following the
tions of GD within built environment, sustainability, BIM, and AI design introduction of parametric Computer-Aided Design (CAD) tools, the
practices, clarifying its potential to shape a more sustainable, resilient, exploration of GD took on a more practical focus.
and inclusive built environment for future generations. By defining the Notably, it was only within the initial decade of the 21st century that
existing challenges and ongoing research initiatives, this paper con- initial studies emerged; wherein generative design was subject to aca-
tributes to the ongoing dialogue surrounding the utilisation of GD as a demic research for its potential applications within the architecture field
tool for building and infrastructure design practices within the broader [8]. GD found early adoption in industries such as the aerospace,
field of built environment, with a primary focus on the design phase. automotive, and architecture industries. In aerospace, companies like
GD intersects with various disciplines, including architecture, engi- Airbus utilise generative design to create lightweight and structurally
neering, environmental science, and computer science. This review optimised components for aircraft, improving fuel efficiency. Similarly,
paper contributes to bridging interdisciplinary perspectives by exploring in the automotive sector, manufacturers have employed generative
the integration of GD principles with sustainability, BIM, and artificial design to innovate vehicle components, enhancing both performance
intelligence as well as within the context of the built environment. By and fuel economy. In architecture, the technology has been applied to
synthesizing insights from various fields, the paper promotes cross- generate complex and efficient building structures, pushing the
pollination of ideas and encourages interdisciplinary collaboration. By boundaries of design possibilities and sustainability.
identifying knowledge gaps, unanswered questions, and areas ripe for GD is a rules-driven iterative design process, which is based on
exploration, the paper serves as a roadmap for researchers seeking to algorithmic and parametric modelling to automatically explore, iterate,
contribute to the ongoing evolution of the field. and optimise design possibilities by defining high-level constraints and
goals [9]. The most powerful capability of GD is it can automatically
2. Overview of generative design explore and iterate design possibilities, and permute the best solutions
for human designers and decision makers [10]. GD systems in the
This section aims to clarify the requirements of GD by investigating building sector mainly focus on architectural, structural engineering,
background knowledge of GD, the process of GD, as well as AI and and urban design applications [11]. Given the increased demand for
machine learning in GD. construction materials and their corresponding adverse effects on the
environment, the application of GD methods to structures is worthy of
further research with a particular focus on material reduction and al-
ternatives. For instance, GD systems can be utilised for sustainability-

2
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Fig. 2. Review framework.

3
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

related objectives including daylight maximisation and energy-efficient objectives and constraints. The collaboration of AI and generative design
designs, based on parameters such as materials, construction methods not only expedites the design iteration process but also opens up new
and structural components [12]. frontiers for creativity, producing designs that are not only aesthetically
pleasing but also functionally optimised and context aware.
2.2. Process of GD
3. Review methodology
The GD process can be broken down into three components from the
perspective of design: (1) “a design framework”, (2) “a means of 3.1. Review design
generating variations”, and (3) “a means of determining preferable
outcomes” [13]. As shown in Fig. 1, GD is an iterative process which can GD in the built environment, such as civil engineering, has evolved
be summarised into four steps [2]: significantly over the past decade, driven by advancements in compu-
Step 1: In the initial phase, designers define design goals and ob- tational capabilities and design methodologies. Initially emerging as a
jectives, incorporating constraints and parameters such as material concept in the early 2000s, the application of generative design princi-
properties, performance criteria and load cases. ples gained attention around 2015. This innovative approach has since
Step 2: Subsequently, by employing algorithms, the generative transformed the industry by enabling engineers to explore a multitude of
design solution produces a wide range of designs, potentially executing optimised design solutions, considering factors such as structural effi-
performance analyses to inform decision-making. ciency, material utilisation, and environmental impact, ultimately
Step 3: Upon receipt of preliminary outcomes, designers or engineers leading to more innovative and sustainable infrastructure designs.
evaluate various options, iteratively adjusting parameters and objectives This review was conducted using a content analysis-based approach
to refine the problem formulation. This iterative process, guided by a which allows for both qualitative and quantitative analysis as well as
combination of human intuition and artificial intelligence, continues provides valuable historical and practical insights over time [16]. An
until the most relevant solution is obtained. extensive literature was discussed in this paper which was reviewed
Step 4: Finally, the ultimate design undergoes further evaluation and from Scopus and Google Scholar databases. Both databases provide
manufacturing by the designer, ensuring alignment with the established extensive coverage across various disciplines and offer insights into the
design goals. scholarly impact of publications through citation metrics. Specific
search queries were used to filter papers relevant to this study to ensure
the quality of the sorted papers. The research framework is shown in
2.3. Artificial intelligence in generative design Fig. 2. The major topics to be analysed were:

To automatically create designs that satisfy performance re- • The correlation between the built environment design process and
quirements, GD, which defines the design parameters and produces establishing research goals with GD.
potential solutions, integrates artificial intelligence (AI) and machine • The correlation between the sustainable design process and estab-
learning algorithms which perform generative and predictive processes lishing research goals with GD.
that contribute towards choosing the best solutions according to the • The correlation between the BIM design process and establishing
design criteria [14]. AI contributes to GD through the development of research goals with GD.
intelligent algorithms, which typically analyse extensive datasets. AI • The correlation between the AI design process and establishing
focuses on data-driven mapping and navigation strategies when it comes research goals with GD.
to decision-making in GD. AI systems can learn how to map the asso- • The relationship between programming languages and the develop-
ciations between thousands of constituent choices of a compound design ment of different GD elements.
decision, and can thus help users navigate gigantic decision spaces,
where different decisions can have a variety of consequences with
different impacts [15]. 3.2. Review process and criteria
Machine learning techniques enable algorithms to learn from pat-
terns, past designs, and user preferences, allowing for the creation of The “OR” and “AND” Boolean operators were used to refine queries
more context-aware and adaptive generative models and solutions. This in order to capture relevant documents regarding the generative design
integration allows for more intelligent decision-making, as AI algo- implementation in the built environment. The query string used was
rithms continuously learn from iterative design cycles, adapting and “Generative design” OR “Parametric design” OR “Algorithm design”
refining their approaches over time. Generative Adversarial Networks AND “Built Environment” OR “Building” OR “Engineering” OR “Mate-
(GANs) and Residual Networks are examples of AI solutions that can be rials” OR “Construction” OR “Project planning” OR “Sustainable build-
employed to automatically generate new designs or form predictions ing” OR “Environment” AND “Artificial Intelligence”. The search results
(discrete classes or continuous values) respectively. Furthermore, AI were expanded to include publications on GD in urban planning, civil
enhances the efficiency of simulations and analyses, providing designers engineering, sustainable development, and AI on for improving building
with comprehensive insights into structural integrity, material consid- designs efficiency and reducing environmental footprints. Parametric
erations, and performance metrics. Genetic algorithms and reinforce- design and algorithm design were included as relevant keywords since
ment learning, constitute other relevant AI approaches which can be generative design uses algorithms to explore and generate design solu-
employed to predict and optimise designs considering multiple tions, often incorporating parametric design principles for flexibility.

Table 1
Top 5 journal publication sources from 2014 to 2024.
Journal 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 Total

Automation in Construction 1 1 3 1 2 6 2 16
Energy and Buildings 1 2 1 1 1 6
Buildings 1 2 2 5
Advanced Engineering Informatics 1 1 2 4
Applied Energy 2 1 3
Total 1 4 4 3 5 9 7 1 34

4
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

explore how generative design, particularly in cooperation with BIM and

AI, contributes to and enhances the overall design process within the
built environment. These strategies go beyond specific tasks, aiming to
improve and support the broader design workflow, emphasising effi-
ciency and effectiveness in the design process. Fig. 3 shows the publi-
cations classified in terms of the two categories. 38 publications
(accounting for 45%) are sorted under the design process enhancement
category, while the other 47 publications (accounting for 55%) are
sorted under the design for specific task solution category. Further
analysis of GD and different design processes are presented in the sec-
tions below.

4.2. Correlation between the built environment design process and

establishing research goals with GD

Fig. 3. Publications classified into two categories. The built environment incorporates various disciplines and activities
that focus on the planning, design, construction, and management of
During the review, the documents were focused on articles and confer- human-made spaces, including buildings, infrastructure, and urban
ence papers, with the publication language set to English and the pub- areas, to support and enhance modern society. Table 2 offers an analysis
lication dates limited to the 2014–2024 range. The literature obtained of the reviewed publications, presenting a detailed breakdown of their
from databases was sorted manually by reading the abstracts, followed key aspects. The objectives section summarises the primary contribu-
by discussions and conclusions. Next, all papers were browsed tions of the collected studies, providing their main focal points. Mean-
completely to ensure that they covered a subject relevant to the built while, the characteristics section outlines the developmental features
environment field. A Microsoft Excel workbook was customised to identified within the reviewed literature. In addition, an evaluation of
organise the key information of the collected papers, based on devel- the publications was conducted, investigating various factors, including
oping GD in the built environment, sustainable building design, Building an assessment of whether the targeted issues were effectively addressed
Information Modelling (BIM), and AI. For instance, the publication year and if there was an improvement in the design workflow. Additionally,
for each paper, the aim of the research, as well as programming lan- considerations were made regarding the universality of the proposed
guages and software used were recorded for further analysis and solutions, the intensity of programming utilisation, the advancement
summarisation. achieved in design outcomes, and the duration of development.
A total of 85 papers on advancements and implications of GD in the The majority of the research focuses on establishing new method-
built environment were collected and reviewed from Scopus and Google ologies, integrated with GD to design particular elements, to plan certain
Scholar databases. Among all reviewed papers, 76 publications (89%) activities in construction, as well as to improve the design workflow
are journal articles, 6 publications (7%) are conference articles, and 3 [17–39]. The application of GD primarily aims to optimise the design of
publications (4%) are institutional publications. Institutional publica- elements or structures, offering enhanced solutions for decision-making
tions refer to scholarly papers, including research articles, theses, and processes. Furthermore, GD has been employed to automate various
other academic materials, produced, and disseminated by the uni- aspects of the design process, including structural design, urban plan-
versity’s faculty, researchers, and students. Table 1 summarises the top 5 ning, and parking lot layout. Certain studies have extended the scope of
journal publication sources from 2014 to 2024, which are Automation in GD to consider other aspects such as safety risk and cost, covering
Construction, Energy and Buildings, Buildings, Advanced Engineering broader evaluations in design and increasing the maturity of the
Informatics and Applied Energy. framework proposed [23,31,33].
Several studies focused on targeted solutions or to solve a specific
4. Analysis of generative Design in the Built Environment problem, resulting in lower universality which is not applicable to other
related types of designs. For example, the development of optimised
4.1. Publications of establishing research goals with GD in the built wind turbine foundation design [18] may not be directly transferable to
environment the foundation design of high-rise buildings due to differing structural
characteristics. On the other hand, some studies presented solutions
The 85 collected publications allow for the identification of the with high universality, capable of application across various design
correlations between GD research goals and the built environment, scenarios, such as space exploration and transit-oriented development in
sustainability, BIM as well as the AI design process. The approach of urban planning [37,38].
classifying the articles into the 4 suitable categories is based on the The validation of research outcomes often involves case studies and
objectives of the research, following the methodologies outlined. For comparative analyses. Models and data derived from past projects and
instance, a research aim to solve the complex optimisation problem of designs served as benchmarks to evaluate the effectiveness of developed
photovoltaic (PV) module design in high-rise buildings to achieve sus- GD systems. Solutions generated across different scenarios were
tainable urban development, which is presented using a BIM-GD solu- compared based on different scenarios to assess the comprehensive
tion [5]. Therefore, it is categorised into the sustainable section instead coverage of the developed systems. Most of these studies demonstrated
of the BIM section as it aims to promote the utilisation of renewable efficiency in terms of workability, practicality, compatibility, and time
energy by enhancing the PV design method with GD. efficiency, thereby reinforcing the potential of novel design methodol-
Moreover, the papers are classified into two categories: design for ogies. Nevertheless, it is noteworthy that many of these investigations
specific task solutions and design process enhancement. Design for are still in their initial stages, with the methodologies and frameworks
specific task solutions refers to articles focused on generative design exhibiting identifiable limitations. For instance, more relevant param-
techniques tailored to address specific design challenges or tasks within eters are required to generate optimal solutions and new algorithms and
the built environment. These solutions are characterised by their tar- technologies are needed for higher accuracy of the solutions. While
geted and specialised nature, aiming for precision in solving particular further research is needed to enhance the maturity of design outcomes, it
design problems. Design process enhancement gathers articles that is evident that the development of GD in the field of built environment
has shown significant potential. The evolving landscape of GD in this

5
 Z.X. Chew et al.
Table 2
Correlation between built environment and establishing research goals with GD.
Category Objectives Characteristics Evaluations Publications Programming Method Programming
Language

Visual
Automated optimization process of
Dynamo in Revit Programming
steel through arch bridges [17].
Language (VPL)
Random Forest, Textual
Optimization of wind turbine
Feedforward Neural Programming
foundations [18].
Network (FFNN) Language (TPL)
• Propose automated optimisation of • Managed to solve the targeted Minimum-weight design of diagrid tall Grasshopper in Rhino,
VPL
structural elements. design problems. buildings [19]. Galapagos
• Develop integrated methods based • Targeted solutions lack Optimal joint for treelike column
not mentioned N/A
on generative design and additive universality, limiting broad structure [20].
manufacturing. applicability. Automated schematic design of framed
StructGAN-KNWL TPL
• Create a parametric model to be • Development is tube structures [21].
applied in the early stages of the straightforward with minimal Parametric model for the steel pipe
Grasshopper in Rhino VPL
project. programming bridge [22].
Design for • Develop integrated energy implementation. Design of pile slabs and foundations
Grasshopper in Rhino VPL
specific task performance-driven generative • Somewhat insufficiently [23].
solution design methodology. advanced to address complex Evolutionary Program
Modular lightweight steel framed
• Automate three-dimensional visual task at the current phase. for the Space Allocation TPL
dwellings in hot climates [24].
modelling and conduct structural • Short development process. Problem (EPSAP)
calculations. BIM software for
Asphalt pavement design [25]. N/A
• Develop a multi-objective optimi- parametric modelling
zation (FRMOO) ensembled gener- • Focused on built environment design Highway Construction Contract Time
not mentioned N/A
ative design system. practices in aspects of structural Estimation [26].
• Present a framework contributing engineering, infrastructure, construction, Optimisation of large-diameter tunnel
Grasshopper in Rhino VPL
to planning optimisation. urban planning and geotechnics. construction project [27].
6

• Research-based studies. Off-Deployment Traffic Estimation

not mentioned TPL
• Mainly applied to early design stages. [28].
• Mostly evaluated by case studies and Truss layout design and optimization
eifForm TPL
comparative analysis. [29].
• Aimed to increase design efficiency, Optimisation of public transit network
not mentioned N/A
reduce cost, and improve the quality of planning [30].
design outputs. Parametric design for automatic
Python TPL
structural design [31].
• Managed to improve design
Structural optimization at the Grasshopper in Rhino,
workflow in terms of VPL and TPL
• Automate the design process and conceptual design phase [32]. Galapagos
efficiency, time, and cost.
evaluation. Planning and design of a mountainous
• Solutions with high
• Propose a workflow for structural highway in terms of safety risks and NSGA-II algorithm TPL
universality, enabling broad
optimization at the conceptual cost burden [33].
applicability.
design phase. Parametric multi-level of detail
• Development is complex with
• Conduct appropriate planning and modelling method for extra-long Dynamo in Revit, Python VPL and TPL
relatively more programming

Automation in Construction 166 (2024) 105638

design. highway tunnel [34].
implementations.
Design Process • Propose a parametric multi-level Parking lot design [35]. not mentioned N/A
• Relatively fundamental to
Enhancement modelling method. Generative planning for construction
accommodate complex design
• Develop multi-objective optimiza- safety surveillance camera installation
at the current phase. Dynamo in Revit VPL
tion (FRMOO) for the design considering different site layouts with
• Lengthy development process.
process. time [36].
• Propose a framework that Generative urban design and rapid
Grasshopper in Rhino VPL
integrates the generative design solution space exploration [37].
method and data-driven decision- Generative urban design proposals for
Grasshopper in Rhino VPL
making approach. transit-oriented development [38].
Approximation of soil properties
across sites based on sparsely sampled not mentioned N/A
geotechnical data [39].
 Z.X. Chew et al.
Table 3
Correlation between sustainable design and establishing research goals with GD.
Category Objectives Characteristics Evaluations Publications Programming Method Programming
Language

Optimised the amount of materials

• Managed to solve the targeted MATLAB TPL
used in concrete beams [12].
design problems related to
Enhanced mass-customization
materials, costs, and GHG Parametric algorithm TPL
capabilities in building industry [40].
emissions.
Automated parametric design into
• Targeted solutions lack
modular construction to reduce Grasshopper in Rhino VPL
universality, limiting broad
• Develop economical and construction waste [41].
applicability.
environmentally sustainable Integrating Financial and Energy Goals
• Development is straightforward
reinforced concrete structures. for Automated Neighbourhood Layout Genetic Algorithm TPL
with minimal programming
• Enhance prefabrication design [42].
implementation.
capabilities in the building Embodied emissions (EE) reduction at
• Lacking maturity to address
industry. the conceptual stage of high-rise Dynamo in Revit VPL
complex task at the current
• Automate urban layout residential buildings [44].
phase.
considering financial and energy Automated parametric model for
• Short development process. Grasshopper in Rhino VPL
goals. sustainable corbel dwellings [49].
Design for
• Develop a tool to reduce (EE) at Performance-based mixture design
specific task
the conceptual stage. model to generate engineered
solution Python TPL
• Automated parametric model for • Managed to solve the targeted cementitious composite (ECC)
sustainable structural elements design problems such as mixtures [50].
and building design. materials, costs, and GHG Optimization for cost-optimal and low-
• Develop performance-based • Focused on sustainable design practices in emissions. carbon design of high-rise reinforced Genetic Algorithm TPL
design models. aspects of prefabrication, structural, • Solutions with high universality, concrete structures [45].
• Automate the generation of construction, urban planning, and the enabling broad applicability. Automated the generation of optimal
optimal building designs in environment. • Development is complex with building designs in respect of their Life Python 3 with
TPL
respect of environmental and • Research-based studies. relatively more programming Cycle Carbon Footprint (LCCF) and Life EnergyPlus
7

financial concerns. • Mainly applied to early design stages. implementations. Cycle Costs (LCC) [46].
• Mainly validated by case studies, comparative • Relatively fundamental to Optimization of photovoltaic (PV) Dynamo in Revit,
analysis, and illustrative examples. accommodate complex design at modules layout on high-rise building Grasshopper in Rhino, VPL
• Aimed to enhance the efficiency of the design the current phase. skins [5]. Project Refinery
process, minimise waste and costs, optimise • Lengthy development process. Parametric design to minimise the
energy performance within buildings, and embodied GHG emissions of a zero- Grasshopper in Rhino VPL
achieve lower GHG emissions to mitigate emission building [47].
environmental pollution. • Managed to improve design Methodology for assessing walkability
Grasshopper in Rhino VPL
workflow in respect of in urban environments [43].
environmental performance of
• Develop a framework for a
buildings.
national LCA tool.
• Targeted solutions lack
• Develop an algorithm-aided
universality, limiting broad
design workflow that optimizes
applicability.
the building shape. Framework for a national life cycle
• Development is straightforward
• Propose a framework based on assessment (LCA) tool to support early C# TPL

Automation in Construction 166 (2024) 105638

with minimal programming
various performance parameters decision [48].
implementation.
Design Process in building.
• Lacking maturity to address
Enhancement • Develop a parametric generative
complex tasks at the current
algorithm to automate
phase.
performance-oriented urban
• Short development process.
design.
Maximising energy efficiency and
• Develop an automated • Managed to improve design Dynamo in Revit,
daylight performance in office VPL
optimisation framework for workflow with respect to the Grasshopper in Rhino
buildings [51].
sustainable building design environmental performance of
Empowered the process of sustainable
process. buildings.
energy use and management in Minitab, Revit, Green
• Solutions with high universality, N/A
buildings in different climatic Building Studio
enabling broad applicability.
conditions [52].
(continued on next page)
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

domain indicates its ability to contribute substantially to the design

process, facilitating the generation of high-quality solutions.

Programming

VPL and TPL

VPL and TPL

VPL and TPL

Language
4.3. Correlation between the sustainable design process and establishing
research goals with GD

VPL
Grasshopper in Rhino, Sustainable design practices in the built environment are becoming

Grasshopper in Rhino,
Programming Method

Grasshopper in Rhino
increasingly vital, driven by the necessity to create structures that

Dynamo in Revit,
harmonize with ecological concerns while meeting the functional de-
mands of modern society. The integration of sustainability principles in
construction and engineering practices has propelled the exploration of
Python

Python

Python
innovative methodologies such as GD. This intersection between sus-
tainability and GD is a dynamic area of research that holds significant
promise for shaping the future of the built environment. Table 3 pro-
minimising energy consumption in the
layout focusing on health and comfort

vides an analysis of the evaluated publications, presenting a detailed

Chinese urban residences based on

Automated GD-based performance

Automated GD schemes of typical

performance-oriented design flow

Automated GD of residence space

sustainable building design [54].

breakdown of their key components.

Comprehensive exploration of

The primary focus of the studies revolves around proposing inte-

optimization framework for

grated approaches or frameworks for Generative Design (GD) associated

with environmental considerations, including materials, waste, green-
building scale [56].
performance [55].

house gas (GHG) emissions, and cost. An illustrative example is the

application of an automated GD approach to attain precise material
Publications

quantities for reinforced concrete beams, revealing reductions of 63% in

material costs and 57% in CO2 emissions per beam [12]. Within pre-
[53].

fabrication activities, GD emerges as a potential tool for mass custom-

ization within the building industry and facilitates an automated
accommodate complex design at

workflow in modular coordination, effectively minimising construction

• Lengthy development process.
• Development is complex with
relatively more programming

waste during the design stage [40,41]. In urban design and planning, GD
• Relatively fundamental to

proves instrumental in automating neighbourhood layouts, harmonising

financial and energy goals, and yielding findings at a higher stage [42].
the current phase.
implementations.

Additionally, a methodological framework for planning walkability in

urban environments that emphasises factors such as the integration of
distance to amenities, street-level greenery, and human perspective was
Evaluations

proposed, aiming to reduce the use of fuel transportation [43].

GHG emissions stand as a crucial concern within the built environ-
ment which bears a responsibility much like other industries. Therefore,
the studies proposed the automated generation of optimal building de-
signs, specifically for high-rise, reinforced concrete, and zero-emission
buildings, thereby minimising GHG emissions [44–48]. These studies
successfully generated optimal building designs that not only reduce
GHG emissions but also lower material costs. Apart from that, an auto-
mated parametric model for sustainable corbel dwellings was intro-
duced, designed for construction using standard equipment, local
materials, and lower-skilled workers [78]. The advancement extends to
the development of a GD framework for the layout design of photovol-
taic modules on tall building exteriors. This innovative approach results
in a more uniform arrangement, maximising surface capacity,
Characteristics

enhancing energy generation due to less self-shadow effect, and

reducing installation and technical complexities [79]. Moreover,
performance-based design workflows, developed through GD, prove
instrumental in promoting sustainable energy use and designing green
buildings. These workflows consider aspects such as daylight perfor-
mance, health and comfort, and energy efficiency [80–86]. To validate
the efficiency of the methodologies developed, all studies were evalu-
ated through case studies, comparative analyses, and illustrative ex-
amples. The findings demonstrated the efficiency and effectiveness of
these approaches in improving the quality of building designs and
enhancing the sustainability of the design process.
Objectives

4.4. Correlation between the BIM design process and establishing research
goals with GD
Table 3 (continued )

The utilisation of Building Information Modelling (BIM) has emerged

as a transformative approach, revolutionising traditional design and
Category

construction processes. BIM integrates diverse aspects of a project,

advancing collaboration among multidisciplinary teams, improving
communication, and providing a comprehensive digital representation

8
 Z.X. Chew et al.
Table 4
Correlation between BIM design and establishing research goals with GD.
Category Objectives Characteristics Evaluations Publications Programming Method Programming
Language

Minimum weight plane truss design Grasshopper in Rhino,

VPL
[57]. Speckle, Dynamo
• Managed to solve the targeted
Method for designing prequalified Dynamo in Revit, Project
design problems in structural VPL
connections [58]. Refinery
performance.
Autodesk Revit Application
• Targeted solutions lack Generative framework for building
Programming Interface TPL
universality, limiting broad panelization design [60].
(API) with Visual Studio C#
applicability.
Automated structural design of RC
• Development is straightforward Genetic Algorithm TPL
• Develop optimised structural wall-slab buildings [61].
with minimal programming
element design approaches. Intelligent automation and
implementation.
• Develop BIM-based designs to optimization of reinforced concrete not mentioned N/A
• Lacking maturity to address
automate and optimise modular dual systems [62].
complex tasks at the current
designs and constructions. Multi-objective optimization in ArchiCAD, Grasshopper in
Design for phase. VPL
• Automate optimal structure floor tile planning [64]. Rhino
specific task • Short development process.
design of buildings. Drywall installation planning in
solution Dynamo in Revit VPL
• Optimise layout design based on prefabricated construction [65].
multi-objective optimisation • Managed to solve the targeted Automated design and planning for Revit using application
approach. design problems in automation boarding of light-frame residential programming interface TPL
• Automated infrastructure design. buildings [66]. (API) in the C# language.
design and construction. • Solutions with high universality, Parametric BIM approach to foster
enabling broad applicability. bridge project design and analysis Grasshopper in Rhino VPL
• Focused on integrating GD in BIM design
• Development is complex with [67].
practices in aspects of prefabrication,
relatively more programming Automated tunnel segment
structural, construction, urban planning,
implementations. assembly and deviation control Dynamo in Revit VPL
hydraulics, and infrastructure.
• Relatively fundamental to [68].
9

• Research-based studies.
accommodate complex design at Graph data model for automated
• Mainly applied to early design stages.
the current phase. generative design of modular Dynamo in Revit VPL
• Mainly validated by case studies,
• Lengthy development process. buildings [69].
comparative analysis, and illustrative
• Managed to improve design Early-stage parametric
examples.
workflow in structural design and optimisation of the structural Dynamo in Revit VPL
• Aimed to develop BIM-based approaches and
construction. design [59]
workflows to generate design solutions for
• Targeted solutions lack
more precise decision making.
• Automate the optimisation of universality, limiting broad
the structural design of applicability.
buildings. • Development is straightforward
• Develop GD processes in with minimal programming Automated modular construction in
Dynamo in Revit VPL
deconstructing and reusing implementation. congested urban areas [70].
waste materials in construction. • Lacking maturity to address
• Automate infrastructure design complex tasks at the current
and construction processes. phase.
Design Process

Automation in Construction 166 (2024) 105638

• Develop BIM-based framework • Short development process.
Enhancement
to automate and optimise • Managed to improve design Automated binding architectural
modular designs and workflow in respect of structural models to structural models for tall Dynamo in Revit, Python VPL and TPL
constructions. and environmental performance. buildings [63].
• Automate carbon, cost, and life • Solutions with high universality, Integrating three-dimensional road
cycle energy calculations in enabling broad applicability. design and pavement structure Dynamo in Revit VPL
buildings. • Development is complex with analysis [71].
• Automate water and sewerage relatively more programming Fully automated design-through-
Dynamo in Revit, Python,
system design. implementations. analysis workflow for segmental VPL and TPL
C++
• Relatively fundamental to tunnel linings [72].
accommodate complex design at Reality data characterization of
the current phase. recovered construction materials Grasshopper in Rhino, C# VPL and TPL
• Lengthy development process. [73].
(continued on next page)
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

of the project. This approach not only streamlines project delivery but
also facilitates data-driven decision-making, promoting efficiency, ac-

Programming

VPL and TPL

VPL and TPL

curacy, and sustainability throughout the entire lifecycle of projects.
Language
The integration of GD within Building Information Modelling (BIM)
represents a cutting-edge advancement in the built environment, spe-

N/A

VPL

VPL
cifically offering a dynamic approach to conceptualising and optimising
building and infrastructure projects. By utilising algorithms and
computational tools, generative design within the BIM framework en-

Dynamo in Revit, Python

Dynamo in Revit, Python

ables engineers to explore a wide range of design possibilities, consid-
Programming Method

NSGA-II algorithm，
ering multiple parameters and constraints simultaneously. This
Dynamo in Revit
transformative integration empowers engineers to achieve innovative,
not mentioned

efficient, and contextually responsive solutions, promoting a new era of

Refinery

intelligent and data-driven design solutions. Table 4 provides an anal-

ysis of the evaluated publications, presenting a detailed breakdown of
their key components.
The primary focus of the studies revolves around developing GD and
Allocate categories of buildings into

generation for modular multi-story

BIM-integrated design practices. This involves the development of

Parametric BIM-based lifecycle

Automated GD for sustainable

Automated design process for

domestic drinking-water and
Building information models

methodologies aimed at designing structural components for plane

performance prediction and
optimisation for residential

residential buildings [76].

offsite construction [77].

trusses with minimal weight and cost, as well as moment-resisting steel

sewerage systems [78].

connections [57,58]. An automated prototype to perform the structural

design, analysis and optimisation for tall buildings at the early stage was
buildings [74].

proposed, which could also be used by users with or without extensive

Publications

lands [75].

knowledge and background in programming [59]. Moreover, the

effectiveness of a BIM-based generative framework is demonstrated in
enhancing building panelization design productivity [60]. Traditionally,
engineers initiate structural design after receiving architectural layouts
from architects, leading to time-consuming and intensive processes. To
address this, innovative approaches and intelligent frameworks have
been proposed to streamline and automate the interface between ar-
chitecture and structural engineering in building projects [61–63].
Sustainability considerations play a crucial role in the intersection of
BIM and GD processes. Studies have explored the deconstruction and
reuse of waste building materials, automated calculation and analysis of
Evaluations

embodied carbon, and lifecycle energy simulation for residential

buildings and offsite construction. These initiatives showcase the po-
tential for energy conservation and carbon reduction across different
building applications [74,77]. BIM-based generative design extends its
benefits beyond time efficiency, contributing to waste reduction and
optimised material usage in construction. Examples include optimised
floor tile planning [64], drywall installation layout [65], and light-frame
residential buildings boarding [66] which achieved significant waste
rate and material usage reductions from the evaluations. Furthermore,
the applicability of BIM goes beyond building structures to include
infrastructure, such as bridge and road engineering, enhancing design
flexibility and decision-making efficiency [67,71]. In tunnel design, BIM
aids in automating tunnel segment assembly, deviation control, and
Characteristics

analysing lining scenarios [68,72]. The collaboration of BIM and GD is

particularly beneficial in modular construction, automating and opti-
mising modular layout planning and calculation for various building
designs in urban and residential areas [69,70,76].
The studies also propose a workflow for allocating houses on a plot of
land, demonstrating the potential for automating building allocation in
real estate and urban planning [75]. Additionally, a workflow for do-
mestic drinking water and sewerage systems aims to reduce subjectivity
and errors in the design process [78]. The comprehensive evaluation
through case studies, comparative analysis, and illustrative examples
showed the capabilities of BIM in various built environment fields,
Objectives

consisting of structural engineering, prefabrication, infrastructure,

urban planning, environmental considerations, and hydraulics. Further
enhancing BIM with GD will provide more significant benefits to de-
Table 4 (continued )

signers and engineers in terms of work efficiency, duration, and

improved design solutions during the design phase.
Category

10
 Table 5
Correlation between AI design and establishing research goals with GD.

Z.X. Chew et al.

Category Objectives Characteristics Evaluations Publications Programming Method Programming
Language

• Managed to solve the targeted Intelligent generative design for

GAN TPL
design problems. shear wall building [80].
• Targeted solutions lack Automated component layout
universality, limiting broad design of steel frame-brace Python TPL
applicability. structures [81].
• Develop automated
• Development is straightforward Railway Fastener Fault Diagnosis GAN and Residual
innovative structural design TPL
with minimal programming [82]. Network (ResNet)
approaches.
implementation. Multi-condition GD of aerial
• Propose structural crack
• Lacking maturity to address building machine (ABM) Grasshopper in Rhino VPL
detection methods.
complex tasks at the current considering uncertainties [83].
• Propose infrastructure fault
phase. Automated tower crane layout
diagnosis solutions. CraneGAN TPL
Design for specific • Short development process. planning (TCLP) system [84].
• Improve construction
task solution • Managed to solve the targeted Crack Detection Based on GAN and GAN, Convolutional
equipment operability and TPL
design problems with algorithmic Deep Learning [85]. Neural Network (CNN)
layout planning.
design. Prediction of long-term highway
• Prediction and optimisation not mentioned N/A
• Solutions with high universality, traffic speed [86].
of traffic speed and road
enabling broad applicability. Road network synthesis on GAN
network generation. GAN TPL
• Development is complex with [87].
• Flood frequency estimation
relatively more programming Road traffic optimisation [88]. Data mining techniques TPL
and forecasting.
implementations. Multi-site stochastic weather
• Relatively fundamental to generator for flood frequency Python TPL
accommodate complex design at estimation [89].
the current phase. Time-series GAN for flood
• Focused on integrating GD in AI design practices in TimeGAN, RTSGAN TPL
• Lengthy development process. forecasting [90].
aspects of prefabrication, structural, construction,
• Managed to improve the design Multivariate relationships between
urban planning, hydraulics, environment, and Grasshopper in Rhino,
workflow in structural design and campus design parameters and VPL and TPL
infrastructure. Python
urban planning. energy performance [91].
11

• Research-based studies.
• Targeted solutions lack
• Mainly applied to early design stages.
universality, limiting broad
• Mainly validated by case studies, comparative
applicability.
analysis, and numerical analysis.
• Development is straightforward
• Aimed to explore AI-driven approaches and work- Automated clash resolution for Gan, Pix2pix, DRL
with minimal programming
flows to generate and simulate solutions for more reinforcement steel design in (Deep Reinforcement TPL
implementation.
accurate and efficient decision making. precast concrete wall panels [92]. Learning)
• Lacking maturity to address
[79]
complex tasks at the current
• Support decision-making for
phase.
urban design.
• Short development process.
• Automate generation of
GD preference of a ML assisted
clash-free structural designs. Python TPL
residential site plan layout [93].
• Generate residential site plan
Component-based ML for
Design Process layouts.
performance prediction in building Python, Matlab TPL
Enhancement • Rapid evaluation of wind
design [94].
environment. • Managed to improve design

Automation in Construction 166 (2024) 105638

Rapid prediction of Low-Velocity
• Performance prediction for workflow using ML. Grasshopper in Rhino,
Areas (LVAs) around rectangular- VPL and TPL
building design. • Solutions with high universality, Python
form buildings [95].
• Automate modular building enabling broad applicability.
Approximate model predictive
design. • Development is complex with
building control via machine Matlab, ANSI C TPL
relatively more programming
learning [79].
implementations.
A predictive model for daylight Grasshopper in Rhino,
• Relatively fundamental to VPL and TPL
performance [96]. Python
accommodate complex design at
Assessing environmental
the current phase. Dynamo in Revit,
performance in the early building VPL and TPL
• Lengthy development process. Matlab
design stage [97].
Smart design for sustainable city
not mentioned N/A
development [98].
Automated modular housing
Python TPL
design generation [99].
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

4.5. Correlation between the AI design process and establishing research commendable stability and acceptable quality in handling large-scale
goals with GD construction cases [84]. A comprehensive framework was developed
to investigate generative design preferences within a machine learning-
Incorporating AI into built environment design processes offers assisted residential site plan layout. This study revealed that machine
innovative opportunities for optimising project outcomes. By utilising learning can effectively optimise two particular elements, namely plaza
machine learning (ML) algorithms and predictive modelling, engineers paving and landscaped green spaces, aligning them with residential site
can enhance decision-making, facilitate design iterations, and improve layout plans [93]. The integration of AI and GD methodologies extended
overall project efficiency. The integration of AI in the built environment to road engineering, such as the prediction of long-term highway traffic
domain not only accelerates design workflows but also supports adap- speed [86]. Moreover, an efficient road traffic optimisation system uti-
tive, data-driven solutions, marking a transformative shift towards more lising image processing and clustering algorithms demonstrated efficacy
intelligent methodology development. The integration of generative in both day and night conditions [88].
design within AI frameworks revolutionises built environment practices StreetGAN, an innovative example-based approach relying on GANs,
by automating the creation of optimal design solutions. Utilising algo- was proposed for synthesizing road networks, producing structurally
rithms to explore various design possibilities, generative design pro- sound and visually similar novel road network patches [87]. Addressing
poses innovation and efficiency by generating diverse, high-performing the critical issue of flooding, the study explored the implementation of
alternatives tailored to project-specific constraints. This symbiotic GANs as a data-driven multi-site stochastic weather generator for flood
relationship between generative design and AI enables civil engineers to frequency estimation [89]. Besides, the application of Time-series GAN
discover novel solutions, optimise resource utilisation, and push the (TimeGAN) and Real-world Time Series GAN (RTSGAN) proved effec-
boundaries of traditional design approaches, ultimately contributing to tive in generating synthetic flood time series, providing valuable flood
the evolution of more sustainable and resilient design solutions. Table 5 scenarios with complex spatial-temporal correlations. The introduction
shows an analysis of the reviewed publications, presenting a detailed of synthetic datasets holds promise for training classical machine
breakdown of their key components. learning forecasting models, particularly at longer forecast horizons
The primary focus of the studies centers around advancing integrated [90]. In the domain of urban planning, artificial intelligence (AI) plays a
design practices by incorporating GD and AI. One key aspect involves crucial role in supporting the creation of energy-efficient and visually
the application of structural design methodologies, appealing urban environments [91]. Notably, previous research has
demonstrated by the utilisation of Generative Adversarial Networks explored a Smart Design framework, emphasising urban design
(GANs) to generate shear wall building designs [80]. GANs, as machine decision-making supported by artificial intelligence-aided design
learning models with a generator-discriminator architecture. (AIAD) [98].
trained in an adversarial manner, exhibit the ability to iteratively ML become a valuable tool in predicting various performance aspects
produce high-quality, realistic outputs, a capability well-established in in building design, capable of providing detailed and simplified simu-
various domains, including image synthesis, style transfer, and creative lations for parameterised components during the early design phase
content generation. Several innovative GAN-based methods have been [94]. A comprehensive study introduced a versatile framework for
proposed, such as FrameGan, which automates the layout design of steel predictive building control, incorporating an arbitrary optimization-
frame-brace structures [110]. Furthermore, GAN integration with deep based control algorithm [79]. Additionally, predictive models for
reinforcement learning has demonstrated efficacy in novel crack daylight performance based on multimodal data [96] and a quantitative
detection methods, employing a three-stage detection model [111], and method utilising parametric design and ML for assessing environmental
in the automated generation of clash-free rebar designs for prefabricated performance during early decision stages in building projects were also
concrete wall panels [112]. The collaboration between AI and GD ex- proposed [97]. In prefabrication, an innovative automated modular
tends its impact to address challenges in rail fastener fault diagnosis and housing design system leveraging a module configuration algorithm and
enhance the detection accuracy of labour-intensive manual patrol in- a coupled Generative Adversarial Network (CoGAN) framework was
spection tasks [82]. developed in [99]. This approach successfully generated an extensive
In construction, an innovative multi-condition generative design and array of design alternatives swiftly, allowing for efficient evaluation
deep learning ensembled framework was introduced to enhance the within specified constraints. All studies went through evaluation
visualisation, operability, and intelligence of aerial building machines, through case studies, comparative analyses, and numerical analysis to
resulting in a significant 16.59% improvement under normal construc- validate the efficiency of the methodologies developed. The findings
tion conditions and 16.91% enhancement under extreme wind scenarios demonstrated the efficiency and effectiveness of AI with GD in built
[83]. Additionally, a novel image-based automatic tower crane layout environment designs. The utilisation of GAN and other ML algorithms in
planning (TCLP) system tailored for high-rise buildings demonstrated the design models proved to be especially promising and impactful.

4.6. Programming languages and software used to develop GD in built

environment design practices

In the domain of built environment design practices, the application

of generative design is facilitated by a variety of programming languages
and software tools. Various programming languages such as Python,
JavaScript, and Ruby are commonly employed to script algorithms that
drive generative design processes, enabling architects and designers to
iteratively explore diverse solutions. Additionally, specialised software
platforms, including but not limited to Rhino with Grasshopper, Revit
with Dynamo, Autodesk Generative Design, and Fusion360 have
emerged as integral tools for implementing generative design method-
ologies. These platforms provide intuitive interfaces for creating para-
metric models, optimising designs based on specified criteria, and
generating innovative solutions that respond dynamically to changing
parameters. The seamless integration of programming languages and
Fig. 4. Summary of programming languages application from publications. dedicated software contributes to the evolution of generative design

12
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Table 6 Table 7
Software and available programming language for GD. AI and ML algorithms that can be scripted in TPLs.
Software Developer First BIM Available Programming Algorithm Python MATLAB C# C++
released Language for GD
Random Forest √ √ √ √
VPLs TPLs FFNN (Feedforward Neural Network) √ √ √ √
GAN (Generative Adversarial Network) √ √ √ √
Dynamo, Python,
Revit Autodesk 2000 Yes Evolutionary algorithms √ √ √ √
Project Refinery C++, C#
NSGA-II (Non-dominated Sorting
Rhinoceros Robert Python, √ √ √ √
1998 No Grasshopper Genetic Algorithm II)
3D McNeel C##
Parametric algorithms √ √ √ √
ArchiCAD Graphisoft 1987 Yes Grasshopper N/A
Genetic algorithms √ √ √ √
ResNet √ √ √
CNN (Convolutional Neural Network) √ √ √
methodologies, offering creativity and efficiency in shaping the built Pix2pix √
Limited
√ √
Support
environment. Pix2pixHD √ √ √
In the context of generative design and computational methodolo- DRL (Deep Reinforcement Learning) √ √ √
gies, visual programming languages (VPL) and textual programming
languages (TPL) serve as distinct platforms for expressing and manipu-
environment designs not only expedites the exploration of design pos-
lating design algorithms. VPLs, demonstrated by tools like Grasshopper
sibilities but also aids in addressing complex, multi-dimensional opti-
and Dynamo, enable designers to create and manipulate algorithms
misation challenges, thereby providing a significant transition towards
through a visual interface, using graphical elements and nodes to
more efficient, creative, and data-informed design practices.
represent code logic. This visual approach enhances accessibility and
Table 7 presents the common TPLs used for programming a variety of
provides a more intuitive understanding of the design process.
AI and ML algorithms. While all these algorithms can be implemented in
Conversely, TPLs such as Python and C# rely on text-based syntax for
most programming languages, including Python, MATLAB, C#, and
algorithmic expression, providing a more traditional programming
C++, the availability of specific libraries, ease of implementation, and
experience. Each method offers unique advantages, with VPLs empha-
performance might vary between different programming languages.
sising visual clarity and TPLs offering greater control over elaborated
Python is a popular choice for AI development due to its extensive
algorithmic details. The choice between VPL and TPL in generative
provision of frameworks, libraries and community support. Deep
design practices often depends on the designer’s preferences, expertise,
learning frameworks like TensorFlow and PyTorch are commonly used
and the specific demands of the design task at hand. Fig. 4 depicts a
for neural network implementations in Python. MATLAB is often
summary of the programming languages adopted in the reviewed
preferred for its ease of use in numerical and scientific computing,
studies. 29 publications (accounting for 34%) utilised VPLs, 31 publi-
making it well-suited for certain types of applications. However, it may
cations (accounting for 36%) utilised TPL, 14 publications (accounting
not be the first choice for some researchers who prefer cost-free and
for 16%) utilised both VPL and TPL, while 11 publications (accounting
more versatile languages like Python which has an extensive collection
for 13%) did not mention the specific programming language(s) used.
of machine learning and AI libraries for developing complex AI models.
The integration of VPL and TPL in generative design processes en-
hances the versatility and efficiency of algorithmic exploration. By
employing a dual approach, designers can utilise the intuitive visual 4.7. GD programming language designer suitability
representation of VPLs for conceptual clarity while employing the pre-
cision and expressiveness of TPLs for fine-grained control, thereby In advancing GD aligned with built environment objectives, the se-
proposing a comprehensive and adaptive computational design frame- lection of, and proficiency in, programming languages play vital roles,
work in the built environment. For an extra-long highway tunnel design, serving as a critical pathway for translating innovative concepts into
a parametric tunnel model was established by positioning, stretching, or stable computational frameworks and solutions. Designers enjoy the
placing the unit model in the Revit and Dynamo workspace, while a flexibility of selecting programming languages based on their distinct
Python script in Dynamo was applied for generating tunnel lining types objectives and design aspirations. Research investigations may employ
with structures in it [34]. In a sustainable offsite construction study, a VPL, TPL exclusively, or collaboratively utilise both VPL and TPL within
prototype was developed and interfaced within Revit, using Dynamo the chosen design domain. Table 8 compares the adoption of VPL and
programming, Python scripting, and a GD tool where its main func- TPL approaches regarding different criteria, such as their advantages,
tionality and key features were created by Dynamo and its key algo- disadvantages, flexibility, learning curve, debugging, and AI/ML inte-
rithms were programmed with Python [77]. The solution serves to gration. The comparison is summarised by reviewing the research out-
automatically generate parametric BIM for offsite construction, deter- comes and performance levels, in addition to the authors’ remarks
mine the optimal building topological arrangement and element type, sourced from the gathered publications.
and evaluate the embodied carbon and construction cost associated with Given the varied characteristics of tasks and applications highlighted
geometric models. Table 6 presents a summary of the software and in the studies, the suitability of VPL and TPL in built environment design
programming languages used for GD from the reviewed literature. Dy- becomes a critical consideration. VPL, with its graphical representation
namo in Revit and Grasshopper in Rhinoceros 3D are the most common and intuitive interface, may offer an accessible entry point for designers
VPLs used for GD workflows. less experienced in traditional coding, facilitating collaborative efforts
The integration of AI algorithms has become instrumental in opti- and accelerating design exploration. This visual approach aligns well
mising and innovating design processes. Algorithms such as genetic al- with the creative and exploratory aspects of generative design, as
gorithms (GAs), Non-dominated Sorting Genetic Algorithm II (NSGA-II), demonstrated in the use of a steel pipe bridge design [22], modular
and GANs are examples that have found application in diverse aspects of construction waste coordination [41], and drywall installation planning
design optimization. Genetic algorithms emulate the process of natural for prefabricated construction [65], and road design and pavement
selection, evolving design solutions through successive generations, structural analysis [71].
while NSGA-II excels in multi-objective optimisation, enabling the On the other hand, the studies are also divided into complex struc-
exploration of optimal solutions. GANs, with their capacity for genera- tural and environmental considerations, where the precision and gran-
tive modelling, facilitate the creation of novel and realistic design al- ular control offered by TPL become crucial. TPL may be preferred for
ternatives. The incorporation of these AI algorithms in built tasks requiring complex algorithmic control, such as crack detection
[85], tower crane layout planning [84], road traffic optimisation [88]

13
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Table 8 Table 8 (continued )

Comparison of VPL and TPL regarding different criteria. Criteria VPLs TPLs
Criteria VPLs TPLs
machine learning • Provide a high degree of
• Utilise visual elements to algorithms seamlessly into customisation and control
• Rely on text-based syntax for
create algorithms, the visual workflow. over AI implementations via
coding, emphasising written
emphasising graphical specialised AI frameworks
statements and commands to
representations. and libraries in languages
define program logic.
• Visually construct and • Enhance accessibility for like Python or C#.
Definition • Execution is typically (but
manipulate program logic, designers exploring AI- • Allow for the detailed coding
not necessarily) sequential,
via nodes, blocks, or driven design solutions. of intricate AI algorithms
following a traditional
diagrams, providing an thanks to the text-based na-
approach to programming
intuitive approach to ture of TPLs.
languages.
programming. • VPLs in-built with ML • Enhance designers with the
• Offer accessibility, • Provide precise control, functionalities allow users flexibility to implement ML
especially for non- making them suitable for to drag and drop ML com- algorithms with precision,
programmers, with an complex tasks. ponents, streamlining the utilising comprehensive
intuitive interface for rapid • Allow a versatile array of integration of predictive libraries such as scikit-learn
prototyping and design applications across a variety models into design (Python) or TensorFlow
Advantages
exploration. of programming scenarios, ML Integration workflows. (JavaScript).
• Provide visual clarity that thanks to their high • Simplify the application of • This textual control enables
helps in effective expressiveness. ML in design tasks, fine-tuning of models and
communication of design • Provide extensive libraries to enabling designers to seamless integration of
concepts. enhance functionality. utilise predictive analytics advanced ML techniques into
• Have a steeper learning without extensive coding various applications within
• May struggle with complex curve, making them less knowledge. the built environment.
algorithms due to a lack of approachable for beginners.
Disadvantages fine-grained control. • May challenge the
• May limit the handling of conceptualization of code and flood frequency estimation [89]. The combination of GD and AI in
highly specialised tasks. structure, due to the absence these scenarios demands a higher degree of customisation and fine-
of a visual representation.
tuning, making TPL a suitable choice for engineers and designers
• Python
• Dynamo • MATLAB seeking explicit control over the underlying algorithms.
Example Tools
• Grasshopper • C# As for beginners, an approach that incorporates both VPL and TPL
• C++ elements might be beneficial. Initiating the learning process with a VPL
• Are preferred for complex allows beginners to learn foundational concepts and gain confidence in
• Excel in rapid prototyping, algorithmic computations,
algorithmic exploration, performance optimization,
manipulating visual representations of algorithms. As proficiency grows,
Use case and conceptual design, and large-scale software transitioning to TPLs can deepen their understanding, providing a more
particularly in creative and development. comprehensive and versatile skill set. Platforms that offer a combination
design-oriented fields. • Allow crucial fine-grained of both programming languages or seamlessly integrate VPLs and TPLs
control.
can provide beginners with a balanced learning environment. Moreover,
• Enable precise control over
every aspect of an algorithm, resources such as tutorials, interactive coding exercises, and community
• Allow quick iterations,
thanks to their extensive support play crucial roles in easing the learning journey for beginners,
emphasising creativity and
Flexibility
iterative refinement in
flexibility. regardless of the programming language chosen. All in all, the suitability
• Are advantageous for of using VPL or TPL for beginners in generative design within the built
design exploration.
complicated algorithmic
details and optimisation.
environment depends on an appropriate methodology that considers the
• Exhibit lower learning learner’s background, preferences, and the learning curve associated
curves. with each programming platform.
• Exhibit steeper learning
• Facilitate accessibility to
curves.
individuals without a
• Require a deep 5. Analysis and research trends
Learning Curve strong programming
understanding of syntax and
background.
programming principles for
• Assist in understanding The majority of the research reviewed concentrates on developing
mastery.
basic concepts quickly new methodologies integrated with GD to optimise the design of specific
through visual elements. elements, plan construction activities, and enhance overall design
• Involve debugging through
• Benefit from a visual
interpreting error messages
workflows. GD is found to be successfully employed for automating
representation that helps in various aspects, including structural designs [19,20], urban planning
and using debugging tools.
quickly identifying issues.
Debugging • Demand a more conventional [37,38], and parking lot layout [35]. While some studies extend the
• Simplify the debugging
process, especially for
approach to identifying and scope of GD to address broader concerns such as safety risk and cost
rectifying issues within the
beginners. [33], others focus on targeted solutions, resulting in varying levels of
code.
• Have extensive and diverse universality. The integration of sustainability principles in construction
communities across and engineering practices has given rise to innovative methodologies,
• Avail abundant resources
and tutorials due to their
industries. such as decreasing significant percentages of carbon emissions and cost
• Benefit from a wide range reduction, which save material usage [44,45].
Community popularity in creative and
ecosystem of libraries,
Support friendly applications,
frameworks, and resources,
The integration of GD within the BIM framework is identified as a
particularly those used in cutting-edge advancement, offering a dynamic approach to con-
making them applicable to a
design fields.
wide array of programming ceptualising, and optimising design projects. Through algorithms and
needs. computational tools, this integration enables engineers to explore a
• Facilitate AI integration • Offer extensive support for AI
multitude of design possibilities simultaneously, considering multiple
through visual AI blocks or integration through
AI Integration
libraries, enabling dedicated libraries and parameters and constraints [49,52]. The collaboration between BIM and
designers to incorporate frameworks. GD extends beyond building structures to include infrastructure,
modular construction, and tunnel design, thereby enhancing design

14
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Fig. 5. Research trends of GD in the built environment.

Fig. 6. Yearly counts of GD adoption in different built environment fields.

flexibility and decision-making efficiency. Furthermore, the integration used in various built environment fields, with structural design
GD with AI accelerates design workflows, promotes data-driven deci- employing GD the most for structures, connections, and truss designs. In
sion-making, and introduces more intelligent methodology develop- addition, considering environmental aspects in the design process has
ment. GAN-based models are the most utilised ML approach to develop a gained attention and GD has shown its ability to address environmental
unique design approach that is suitable for the research. challenges and perform optimisation for GHG emissions, material usage
Based on an analysis of the studies focusing on the built environment, as well as cost. Fig. 7 shows the count of GD adoptions in various built
the research can be categorised into a few aspects, which includes environment fields for the 4 categorised design processes. It shows that
structural, construction, infrastructure, urban, environment, prefabri- the environmental aspect occupies a significant proportion in the sus-
cation, hydraulics, and geotechnics. Of all the papers reviewed, struc- tainability design process and that BIM approaches have proved their
tural was the most studied for methods integrating GD, subjected to ability and versatility in integrating with GD in different design fields.
28%, followed by 22% for environment, 16% for infrastructure, 12% for Moreover, the capabilities of AI have proven their usefulness across
construction, 9% for prefabrication, 7% for urban, 4% for hydraulics, various built environment design domains. The adaptability of AI
and 1% for geotechnics. technologies in complementing and enhancing GD processes confirms
Fig. 5 illustrates the increase in the application of GD within the built the synergy between AI and innovative design methodologies, reflecting
environment design process, which rises particularly in the current a promising path for the future evolution of integrated AI-GD systems
decade, namely the 2020s. It is noteworthy that the limited research within design practices.
output in the year 2020 might be due to the global pandemic, which led
to the postponement of study works worldwide. In recent years, there
has been a significant rise in the integration of AI into GD design
methodologies, mainly utilising TPL. Fig. 6 provides the count of GD

15
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Fig. 7. GD adoption counts in built environment fields in relation to the 4 design processes.

6. Challenges and possible approaches for GD in the built conventional engineering software tools that, while integrated with
environment architectural designs, fall short in their ability to compute or generate
optimal designs [63]. Furthermore, a crucial restriction in advancing GD
6.1. Challenges applications is the insufficient of high-quality data. The lack of relevant
parameters obstructs the development of an effective design framework,
Despite the promising results of adopting GD in the built environ- emphasising the need for comprehensive enhancements in various as-
ment domain, it is still a great challenge to incorporate GD in the cor- pects [38,81]. The depth of information and the degree of para-
responding design methodologies. One significant limitation lies in the metrisation in some GD models demand improvements to elevate their
inadequacy of existing optimisation algorithms, which have yet to align efficiency within the built environment context [34].
seamlessly with GD methodologies [29]. The current technological The successful implementation of GD in built environment design
landscape, although evolving, faces immaturity and inflexibility in ac- requires effective collaboration among various disciplines, such as ar-
commodating a wide range of design practices [32]. Certain software chitecture, civil engineering, and urban planning. Achieving seamless
might need improvements in terms of the functionality and operability integration and communication between professionals with varied
of GD, as it is neither completely compatible with GD nor user-friendly expertise can be a significant challenge [61]. Adhering to existing
[58]. This inefficiency is caused by a significant reliance on building codes, standards, and regulations is crucial in the built

Fig. 8. Challenges of utilising GD approaches in the design phase.

16
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

designs are not only optimal in the short term but also sustainable over
the life cycle of a building requires further research and development
[74].
Fig. 8 summarises the challenges of utilising GD approaches in the
design phase after reviewing the ongoing research initiatives.

6.2. Possible approaches

For future directions in the BIM design process, the incorporation of

augmented reality (AR) and virtual reality (VR) technologies into BIM-
GD workflows can provide immersive experiences for project stake-
holders, facilitating better visualisation, design validation, and stake-
holder engagement as shown in Fig. 9. Efforts towards standardisation
and interoperability between various BIM and GD platforms can pro-
mote a more collaborative and integrated approach within the built
environment community. Ensuring user-friendly interfaces in BIM-GD
tools can enable convenient access to these technologies, allowing en-
gineers and designers with varying levels of expertise to utilise the
benefits seamlessly.
Fig. 9. Incorporation of AR and VR into BIM-GD.
As AI-driven generative models become increasingly developed,
there is a growing need for transparent and interpretable decision-
environment. GD methodologies may face challenges in ensuring making processes. Future studies should emphasise the development
compliance with these regulatory frameworks, potentially limiting their of AI models that not only generate designs but also provide explana-
applicability [48]. Generating complex designs through GD often de- tions for the rationale behind each decision, in line with the evolving
mands advanced computational resources. Access to high-performance fields of interpretable and explainable AI. This aspect is crucial for
computing infrastructure and addressing the associated costs could be gaining the trust of engineers and stakeholders. In addition, exploring
a constraint for some designers and organisations. Moreover, the suc- collaborative frameworks where AI systems work collaboratively with
cessful adoption of matured GD in the built environment requires a human designers can lead to more innovative and contextually sensitive
skilled workforce expert in both the technical aspects of GD algorithms solutions. Understanding how AI tools can enrich and amplify human
and the integration of these tools into traditional design processes. The creativity in the design process is a promising path for future investi-
need for comprehensive training sessions and programs is also a chal- gation. Considering the unpredictable nature of real-world construction
lenge. Apart from that, evaluating the long-term sustainability of scenarios, future research should prioritise the development of genera-
structures designed using GD poses challenges. Ensuring that generated tive models that exhibit resilience, incorporating adaptive algorithms

Fig. 10. Proposed possible approaches particularly in design phase.

17
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Fig. 11. Summary of review objectives and answers.

capable of adjusting to dynamic and unforeseen challenges during the 6.3. Perspective of developing built environment design objectives in GD
project timeline.
The evolution of AI-GD models into real-time decision support sys- The new proposed perspective addresses a Hybrid Programming
tems can revolutionise the built environment industry. Future research Approach for Generative Design (HyPAGeD), acknowledging that built
efforts should focus on developing adaptive systems that provide environment design tasks often span a spectrum of complexities, from
instantaneous feedback and suggestions during the design and con- creative exploration and conceptualisation to the integration of struc-
struction phases, facilitating more flexible decision-making. Aside from tural, environmental, and other considerations. In the creative and
that, the integration of advanced analytical tools, such as reinforcement exploratory phases, where the emphasis is on ideation and collabora-
learning and deep learning algorithms, can further enhance the capa- tion, VPL serves as an instrumental tool, providing a visually accessible
bilities of AI-driven generative design. These tools can provide more platform. The studies showcased the effectiveness of VPL in tasks such as
specialised learning mechanisms, adaptability to dynamic project con- arch bridge design [46], diagrid-like structure design [48], and the
straints, and improved optimisation of design parameters. design of sustainable corbel dwellings [78], validating its ability to
One approach for exploration involves refining GD methodologies to streamline the creative process.
incorporate the life cycle of buildings, ensuring a more comprehensive On the other hand, when encountered with tasks demanding preci-
understanding of the environmental impact of structures from concep- sion, fine-tuning, and algorithmic control, TPL acts as an essential
tualization to demolition. Additionally, the integration of real-time counterpart within the HyPAGeD framework. In scenarios involving
environmental data and predictive modelling within GD frameworks shear wall design [109], automated steel frame-brace design [110], and
could enhance the adaptability of designs to changing environmental clash-free rebar designs in prefabricated concrete wall panels [112],
conditions. Collaborative efforts between researchers, practitioners, and TPL’s textual representation affords engineers and designers the neces-
policymakers are essential to establish standardised sustainability met- sary granularity and explicit control over the underlying algorithms.
rics and benchmarks, providing a cohesive and universally applicable The HyPAGeD perspective encourages a seamless transition between
framework for evaluating the environmental performance of GD- VPL and TPL, allowing designers and engineers to navigate the complex
generated designs. As the field continues to evolve, future research design landscape with flexibility, accuracy, and efficiency. This inte-
should navigate towards more comprehensive, adaptable, and ethically grated approach not only accommodates the varied requirements within
grounded approaches that contribute not only to the efficiency of design built environment design but also bridges the gap between the creative
processes but also to the long-term sustainability of the built and analytical aspects of the discipline.
environment. Moreover, the interdisciplinary applications of AI and GD method-
All in all, Fig. 10 summarises the proposed possible approaches ologies emphasise the need for a comprehensive perspective in built
particularly in the design phase after reviewing the ongoing research environment design. By embracing the HyPAGeD framework, designers
initiatives. A significant future direction involves refining AI-driven GD can employ the collaborative and exploratory strengths of VPL alongside
models to explicitly incorporate sustainability metrics. The integration the precision and control offered by TPL, resulting in a comprehensive
of ecological considerations, life-cycle assessments, and environmental and adaptive design process. This perspective positions GD not merely as
impact analyses within the generative design process can contribute to a tool but as a dynamic and responsive framework that evolves with the
the creation of more sustainable and environmentally friendly design demands of each project, ultimately redefining the way designers and
solutions. engineers conceptualise, plan, and execute within the domain of built
environment design.

18
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

7. Conclusion Acknowledgements

This comprehensive review paper presents the transformative impact This research was funded by the Ministry of Higher Education
of integrating GD in built environment sustainability, BIM, and AI design (MOHE) of Malaysia under the Fundamental Research Grant Scheme
processes. Through a thorough examination of the literature, various (Grant No: FRGS/1/2024/SSI12/UNIM/02/1). We gratefully acknowl-
applications spanning structural design, construction, infrastructure, edge the grant support.
environment, prefabrication, hydraulics, geotechnics, and urban plan-
ning showcase the effectiveness of this integration. 85 publications References
demonstrated a relationship between establishing research goals with
GD and its application in built environment design processes, with the [1] S. Oh, Y. Jung, S. Kim, I. Lee, N. Kang, Deep Generative Design:Integration of
TopologyOptimization and Generative Models, J. Mech. Des. (2019), https://doi.
research papers classified into two main categories: design for specific org/10.48550/arXiv.1903.01548.
task solutions, focusing on targeted GD techniques addressing particular [2] M. McKnight, Generative design: what it is? How is it being used? Why it’s a game
design challenges, and design process enhancement, which contributes changer, KnE Engineering. (2017), https://doi.org/10.18502/keg.v2i2.612.
[3] W. Liao, X. Lu, Y. Fei, Y. Gu, Y. Huang, Generative AI design for building structures,
to and improves the broader design workflow within the built Autom. Constr. (2023), https://doi.org/10.1016/j.autcon.2023.105187.
environment. [4] C. Li, Y. Chen, Modeling and optimization method for building energy performance
The selection of, and proficiency in, programming languages in the design stage, J. Build. Engin. (2024), https://doi.org/10.1016/j.
jobe.2024.109019.
constitute an important aspect in advancing GD aligned with built [5] F. Vahdatikhaki, N. Salimzadeh, A. Hammad, Optimization of PV modules layout
environment objectives. The choice between VPL and TPL is dependent on high-rise building skins using a BIM-based generative design approach, Energy
upon the specific characteristics of different design tasks within the & Buildings. (2021), https://doi.org/10.1016/j.enbuild.2021.111787.
[6] J. Jauhiainen, The Metaverse: innovations and generative AI, International Journal
field. While VPL offers accessibility and collaborative potential, partic-
of Innovation Studies. (2024), https://doi.org/10.1016/j.ijis.2024.04.004.
ularly for the visual and exploratory aspects of GD. TPL becomes [7] A. Barcaui, A. Monat, Who is better in project planning? Generative artificial
essential for tasks demanding precision and algorithmic control, such as intelligence or project managers? Project Leadership and Society. (2023) https://
complex structural considerations and environmental analyses. A doi.org/10.1016/j.plas.2023.100101.
[8] L. Barbieri, M. Muzzupappa, Performance-driven engineering design approaches
balanced approach that incorporates both VPL and TPL elements, based on generative design and topology optimization tools: a comparative study,
catering to the varied expertise and learning trajectories of designers, Appl. Sci. (2022), https://doi.org/10.3390/app12042106.
becomes a beneficial strategy for advancing research and development [9] W. Ma, X. Wang, J. Wang, X. Xiang, J. Sun, Generative Design in Building
Information Modelling (BIM): Approaches and Requirements., Sensors, 2021,
within the design domain. https://doi.org/10.3390/s21165439.
Future directions in BIM-GD processes emphasise the integration of [10] A. Di, Generative Design for project optimization (S), in: International Conferences
AR and VR technologies for enhanced stakeholder engagement and on Distributed Multimedia Systems,Proceedings of the 27th International
Conference on Distributed Multimedia Systems, 2021, https://doi.org/10.18293/
design visualisation. Standardisation and user-friendly interfaces are dmsviva21-014 [accessed7 December 2023].
important for encouraging collaboration and accessibility within the [11] S. BuHamdan, A. Alwisy, A. Bouferguene, Generative systems in the architecture,
built environment community. The evolution of AI-driven generative engineering and construction industry: a systematic review and analysis, Int. J.
Archit. Comput. (2021) 226–249, https://doi.org/10.1177/1478077120934126.
models into transparent, collaborative, and real-time decision support [12] F. Alsakka, A. Haddad, F. Ezzedine, G. Salami, M. Dabaghi, F. Hamzeh, Generative
systems is vital for addressing dynamic challenges and enriching the design for more economical and environmentally sustainable reinforced concrete
design process, with a specific emphasis on incorporating sustainability structures, J. Clean. Prod. (2023), https://doi.org/10.1016/j.jclepro.2022.135829.
[13] R. Jaisawal, V. Agrawal, Generative design method (GDM) – a state of art, IOP
metrics. These sustainable approaches, integrating ecological consider-
Conference Series: Materials Science and Engineering. (2021), https://doi.org/
ations, life-cycle assessments, and environmental impact analyses 10.1088/1757-899x/1104/1/012036.
within the GD process, are essential for creating environmentally [14] J. Srivastava, H. Kawakami, Systematic Review of Difference between Topology
friendly and resilient design solutions, contributing to the long-term Optimization and Generative Design, International Federation Automatic Control
(IFAC)-PapersOnLine, 2023, https://doi.org/10.1016/j.ifacol.2023.10.307.
sustainability of the built environment. [15] P. Nourian, S. Azadi, R. Uijtendaal, N. Bai, Augmented Computational Design:
In conclusion, this paper aims to contribute to the ongoing dialogue Methodical Application of Artificial Intelligence in Generative Design, 2023
surrounding the utilisation of GD as a tool for design practices within the [accessed 4 December 2023].
[16] C. Zhang, G. Gholipour, A.A. Mousavi, Blast loads induced responses of RC
broader built environment field, with a primary focus on the design structural members: state-of-the-art review, Compos. Part B Eng. (2020), https://
phase. Fig. 11 depicts the summarised review objectives and answers doi.org/10.1016/j.compositesb.2020.108066.
obtained from the analysis. The literature reviewed indicates the [17] K. Korus, M. Salamak, M. Jasinski, Optimization of geometric parameters of arch
bridges using visual programming FEM components and genetic algorithm, Eng.
importance of using advanced GD technologies in enhancing design Struct. (2021), https://doi.org/10.1016/j.engstruct.2021.112465.
quality and workflows, to provide outstanding solutions for better de- [18] Q. Shen, F. Vahdatikhaki, H. Voordijk, J. Gucht, L. Meer, Metamodel-based
cision making. generative design of wind turbine foundations, Autom. Constr. (2022), https://doi.
org/10.1016/j.autcon.2022.104233.
[19] F. Cascone, D. Faiella, V. Tomei, E. Mele, A structural grammar approach for the
CRediT authorship contribution statement generative design of diagrid-like structures, Buildings (2021), https://doi.org/
10.3390/buildings11030090.
[20] H. Wang, W. Du, Y. Zhao, Y. Wang, R. Hao, M. Yang, Joints for treelike column
Zhi Xian Chew: Writing – original draft, Validation, Methodology,
structures based on generative design andadditive manufacturing, J. Constr. Steel
Investigation, Formal analysis, Data curation. Jing Ying Wong: Writing Res. (2021), https://doi.org/10.1016/j.jcsr.2021.106794.
– review & editing, Supervision, Project administration, Funding [21] Y. Fei, W. Liao, Y. Huang, X. Lu, Knowledge-enhanced generative adversarial
acquisition, Conceptualization. Yu Hoe Tang: Writing – review & networks for schematic design of framed tube structures, Autom. Constr. (2022),
https://doi.org/10.1016/j.autcon.2022.104619.
editing, Supervision. Chun Chieh Yip: Writing – review & editing. [22] D. Samoilova, The Parametric Design of a Steel Pipe Bridge, Hame University, 2021
Tomas Maul: Supervision, Writing – review & editing. [accessed 20 December 2023].
[23] L. Ohls, Parametric Design Process for Pile Slabs and Foundations, Aalto
University, 2020 [accessed 22 December 2023].
Declaration of competing interest [24] E. Rodrigues, N. Soares, M.S. Fernandes, A.R. Gaspar, Á. Gomes, J.J. Costa, An
integrated energy performance-driven generative design methodology to foster
We hereby declare that we do not have any conflict of interest in modular lightweight steel framed dwellings in hot climates, Energy Sustain. Dev.
(2018) 21–36, https://doi.org/10.1016/j.esd.2018.02.006.
connection with the work submitted. [25] F. Tang, T. Ma, Y. Guan, Z. Zhang, Parametric modeling and structure verification
of asphalt pavement based on BIM-ABAQUS, Autom. Constr. (2019), https://doi.
Data availability org/10.1016/j.autcon.2019.103066.
[26] D. Zhai, Y. Shan, R.E. Sturgill, T.R.B. Taylor, P.M. Goodrum, Using parametric
modeling to estimate highway construction contract time, Transportation Research
Data will be made available on request.

19
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

Record: Journal of the Transportation Research Board. (2016) 1–9, https://doi. [52] M.K. Najjar, V.W.Y. Tam, L.T. Di Gregorio, A.C.J. Evangelista, A.W.A. Hammad,
org/10.3141/2573-01. A. Haddad, Integrating parametric analysis with building information modeling to
[27] P. Lin, L. Zhang, R. Tiong, Multi-objective optimization ensembled generative improve energy performance of construction projects, Energies (2019), https://doi.
design for large-diameter tunnel passing underneath existing tunnels, Appl. Soft org/10.3390/en12081515.
Comput. (2023), https://doi.org/10.1016/j.asoc.2023.110448. [53] J. Zhang, N. Liu, S. Wang, Generative design and performance optimization of
[28] Y. Zhang, Y. Li, X. Zhou, X. Kong, J. Luo, Off-deployment traffic estimation — a residential buildings based on parametric algorithm, Energ. Buildings (2021),
traffic generative adversarial networks approach, IEEE Trans Big Data (2022), https://doi.org/10.1016/j.enbuild.2021.111033.
https://doi.org/10.1109/tbdata.2020.3014511. [54] Y. Ji, W. Wang, Y. He, L. Li, H. Zhang, T. Zhang, Performance in generation: an
[29] A. Hooshmand, M.I. Campbell, Truss layout design and optimization using a automatic generalizable generative-design-based performance optimization
generative synthesis approach, Computers and Structures (2016) 1–28, https://doi. framework for sustainable building design, Energ. Buildings (2023), https://doi.
org/10.1016/j.compstruc.2015.09.010. org/10.1016/j.enbuild.2023.113512.
[30] K. Wong, Developing a Generative Design Framework for Optimising Public Transit [55] Y. Zhou, Y. Wang, C. Li, L. Ding, Z. Yang, Health and comfort oriented automatic
Network Planning, University of Toronto, 2022 (accessed 6 December 2023). generative design and optimization of residence space layout: an integrated data-
[31] R. Rempling, A. Mathern, D. Ramos, S. Fernández, Automatic structural design by a driven and knowledge-based approach, Developments in the Built Environ. (2024),
set-based parametric design method, Autom. Constr. (2019), https://doi.org/ https://doi.org/10.1016/j.dibe.2023.100318.
10.1016/j.autcon.2019.102936. [56] E. Touloupaki, T. Theodosiou, Optimization of building form to minimize energy
[32] A. Birkemo, K. Samindim, Samarakoon, Application of Generative Design for consumption through parametric modelling, Procedia Environ. Sci. (2017),
Structural Optimization at the Conceptual Design Phase, University of Stavanger, https://doi.org/10.1016/j.proenv.2017.03.114.
2021 (accessed 22 December 2023). [57] B. Malgıt, Ü. Işıkdağ, G. Bekdaş, M. Yücel, A generative design-to-BIM workflow for
[33] C. Li, L. Ding, B. Zhong, Highway Planning and design in the qinghai–tibet plateau minimum weight plane truss design, J. Const. (2022), https://doi.org/10.7764/
of china: a cost–safety balance perspective, Engineering (2019), https://doi.org/ RDLC.21.2.473.
10.1016/j.eng.2018.12.008. [58] D. Henríquez, R.F. Herrera, J.C. Vielma, Method for designing prequalified
[34] A. Rusuli, T. Li, W. Zhang, H. Bai, S. Wang, A parametric multi-level of detail connections using generative design, Buildings (2022), https://doi.org/10.3390/
modeling method for extra-long highway tunnel, IOP Conference Series: Earth and buildings12101579.
Environmental Science. 072051 (2021), https://doi.org/10.1088/1755-1315/ [59] T. Hamidavi, S. Abrishami, P. Ponterosso, D. Begg, N. Nanos, OSD: a framework for
861/7/072051. the early stage parametric optimisation of the structural design in BIM-based
[35] A. Babushkina, M. Petrochenko, A. Kukina, N. Astafieva, Optimizing Parking Lot platform, Constr. Innov. (2020), https://doi.org/10.1108/CI-11-2019-0126.
Design by Generative Design Approach, E3S Web of Conferences, 2021, https:// [60] H. Liu, Y. Zhang, Z. Lei, H.X. Li, S. Han, Design for Manufacturing and Assembly: A
doi.org/10.1051/e3sconf/202126304042. BIM-Enabled Generative Framework for Building Panelization Design, Advances in
[36] S.V.-T. Tran, T.L. Nguyen, H.-L. Chi, D. Lee, C. Park, Generative planning for Civil Engineering, 2021, https://doi.org/10.1155/2021/5554551 pp1–14.
construction safety surveillance camera installation in 4D BIM environment, [61] S. Tafraout, N. Bourahla, Y. Bourahla, A. Mebarki, Automatic structural design of
Autom. Constr. 104103 (2022), https://doi.org/10.1016/j.autcon.2021.104103. RC wall-slab buildings using a genetic algorithm with application in BIM
[37] Y. Sun, T. Dogan, Generative Methods for Urban Design and Rapid Solution Space environment, Autom. Constr. (2019), https://doi.org/10.1016/j.
Exploration, Urban Analytics and City Science, Environment and Planning B, 2023, autcon.2019.102901.
https://doi.org/10.1177/23998083221142191. [62] N. Bourahla, S. Larfi, K. Souaci, Y. Bourahla, S. Tafraout, Intelligent automation
[38] L. Shen, J. Ochoa, H. Bao, X. Huang, W. Yuan, M. White, N. Langenheim, and optimization of reinforced concrete dual systems for earthquake resisting
A parametric framework to assess generative Urban Design proposals for transit- buildings in a BIM environment, J. Build. Engin. (2023), https://doi.org/10.1016/
oriented development, Buildings (2022), https://doi.org/10.3390/ j.jobe.2023.107111.
buildings12111971. [63] T. Hamidavi, S. Abrishami, M.R. Hosseini, Towards intelligent structural design of
[39] W. Chen, J. Ding, T. Wang, D. Connolly, X. Wan, Soil property recovery from buildings: a BIM-based solution, J. Build. Engin. (2020), https://doi.org/10.1016/
incomplete in-situ geotechnical test data using a hybrid deep generative j.jobe.2020.101685.
framework, Eng. Geol. (2023), https://doi.org/10.1016/j.enggeo.2023.107332. [64] S. Wu, N. Zhang, X. Luo, W.-Z. Lu, Multi-objective optimization in floor tile
[40] G.P. Monizza, E. Rauch, D.T. Matt, Parametric and Generative Design Techniques planning: coupling BIM and parametric design, Autom. Constr. (2022), https://doi.
for Mass-Customization in Building Industry: A Case Study for Glued-Laminated org/10.1016/j.autcon.2022.104384.
Timber, Procedia CIRP, 2017, https://doi.org/10.1016/j.procir.2017.01.051 [65] J.D. Cuellar Lobo, Z. Lei, H. Liu, H.X. Li, S. Han, Building Information Modelling-
pp392–397. (BIM-) Based Generative Design for Drywall Installation Planning in Prefabricated
[41] S. Banihashemi, A. Tabadkani, M.R. Hosseini, Integration of parametric design into Construction, Advances in Civil Engineering (2021), https://doi.org/10.1155/
modular coordination: a constructionwaste reduction workflow, Autom. Constr. 2021/6638236 pp1–16.
(2017), https://doi.org/10.1016/j.autcon.2017.12.026. [66] H. Liu, G. Singh, M. Lu, A. Bouferguene, M. Al-Hussein, BIM-based automated
[42] D. Nagy, L. Villaggi, D. Benjamin, Generative Urban Design: Integrating Financial design and planning for boarding of light-frame residential buildings, Automation
and Energy Goals for Automated Neighborhood Layout, in: Proceedings of the in Construction (2018), https://doi.org/10.1016/j.autcon.2018.02.001
2018 Symposium on Simulation for Architecture and Urban Design (SimAUD pp235–249.
2018), 2018, https://doi.org/10.22360/simaud.2018.simaud.025 [accessed 6 [67] A. Girardet, C. Boton, A parametric BIM approach to foster bridge project design
December 2023]. and analysis, Autom. Constr. (2021), https://doi.org/10.1016/j.
[43] D. Kumalasari, M. Koeva, F. Vahdatikhaki, D. Antonova, M. Kuffer, Planning autcon.2021.103679.
walkable cities: generative design approach towards digital twin implementation, [68] J. Gong, T. Bao, Z. Zhu, H. Yu, Y. Li, BIM-based framework of automatic tunnel
Remote Sens. (2023), https://doi.org/10.3390/rs15041088. segment assembly and deviation control, Underground Space. (2023), https://doi.
[44] J. Zaraza, B. McCabe, M. Duhamel, D. Posen, Generative design to reduce org/10.1016/j.undsp.2023.09.005.
embodied GHG emissions of high-rise buildings, Autom. Constr. (2022), https:// [69] Vincent J.L. Gan, BIM-based graph data model for automatic generative design of
doi.org/10.1016/j.autcon.2022.104274. modular buildings, Automation in Construction, 2021, https://doi.org/10.1016/j.
[45] J.L. Vincent, C.L. Gan, K.T. Tse Wong, C.P. Cheng Jack, M.C. Irene, C.M. Chan Lo, autcon.2021.104062.
Parametric modelling and evolutionary optimization for cost-optimal andlow- [70] Y. Wei, H. Choi, Z. Lei, A generative design approach for modular construction in
carbon design of high-rise reinforced concrete buildings, Advanced Engineering congested urban areas, Smart and Sustainable Built Environ. (2021), https://doi.
Informatics, 2019, https://doi.org/10.1016/j.aei.2019.100962. org/10.1108/SASBE-04-2021-0068.
[46] Y. Schwartz, R. Raslan, I. Korolija, Ejan Mumovic, A decision support tool for [71] F. Tang, T. Ma, J. Zhang, Y. Guan, L. Chen, Integrating three-dimensional road
building design: an integrated generative design, optimisation and life cycle design and pavement structure analysis based on BIM, Autom. Constr. (2020),
performance approach, Int. J. Archit. Comput. (2021), https://doi.org/10.1177/ https://doi.org/10.1016/j.autcon.2020.103152.
1478077121999802. [72] J. Ninić, H.G. Bui, G. Meschke, BIM-to-IGA: a fully automatic design-through-
[47] G. Lobaccaro, A.H. Wiberg, G. Ceci, M. Manni, N. Lolli, U. Berardi, Parametric analysis workflow for segmented tunnel linings, Adv. Eng. Inform. (2020), https://
design to minimize the embodied GHG emissions in a ZEB, Energ. Buildings (2018) doi.org/10.1016/j.aei.2020.101137.
106–123, https://doi.org/10.1016/j.enbuild.2018.02.025. [73] J. Zhang, A. Olumo, Reality data characterization of recovered construction
[48] S.R. Hassan, N.A. Megahed, O.M. Abo Eleinen, A.M. Hassan, Toward a national life materials for generative design, 2022: The Digital Reality of Tomorrow, 2022,
cycle assessment tool: generative design for early decision support, Energ. https://doi.org/10.57922/tcrc.623.
Buildings (2022) 112144, https://doi.org/10.1016/j.enbuild.2022.112144. [74] J. Gan, K. Li, X. Li, E. Mok, P. Ho, J. Law, J. Lau, R. Kwok, R. Yau, Parametric BIM-
[49] Y. Zhang, L. Tatarintseva, T. Clewlow, E. Clark, G. Botsford, K. Shea, A template based lifecycle performance prediction and optimisation for residential buildings
design and automated parametric model for sustainable corbel dwellings with using alternative materials and designs, Buildings (2023), https://doi.org/
interlocking blocks, Developments in the Built Environ. (2023), https://doi.org/ 10.3390/buildings13040904.
10.1016/j.dibe.2023.100148. [75] I. Perez-Martinez, M. Martinez-Rojas, J.M. Soto-Hidalgo, A preliminary approach
[50] J. Yu, Y. Weng, J. Yu, W. Chen, S. Lu, K. Yu, H. Wang, Generative AI for to allocate categories of buildings into lands based on generative design, in: IEEE
performance-based design of engineered cementitious composite, Composites Part International Conference on Fuzzy Systems (FUZZ-IEEE), 2020, https://doi.org/
B. (2023), https://doi.org/10.1016/j.compositesb.2023.110993. 10.1109/fuzz48607.2020.9177853.
[51] J. Ratajczak, D. Siegele, E. Niederwieser, Maximizing energy efficiency and [76] S.S. Pibal, K. Khoss, I. Kovacic, Framework of an algorithm-aided BIM approach for
daylight performance inOffice buildings in BIM through RBFOpt model-based modular residential building information models, Int. J. Archit. Comput. (2022),
optimization: the GENIUS project, Buildings (2023), https://doi.org/10.3390/ https://doi.org/10.1177/14780771221138320.
buildings13071790.

20
Z.X. Chew et al. Automation in Construction 166 (2024) 105638

[77] V.J.L. Gan, BIM-based building geometric modeling and automatic generative [89] H. Kang, M. Mirzaei, S. Lai, A. Dehghani, A. Dehghani, D. Ames, Implementing
Design for Sustainable Offsite Construction, J. Constr. Eng. Manag. (2022), https:// generative adversarial network (GAN) as a data-driven multi-site stochastic
doi.org/10.1061/(asce)co.1943-7862.0002369. weather generator for flood frequency estimation, Environ. Model. Softw. (2023),
[78] E. Atencio, P. Araya, F. Oyarce, R.F. Herrera, F. Muñoz-La Rivera, F. Lozano- https://doi.org/10.1016/j.envsoft.2023.105896.
Galant, Towards the Integration and Automation of the Design Process for [90] P. Weng, Y. Tian, Y. Liu, Y. Zheng, Time-series generative adversarial networks for
Domestic Drinking-Water and Sewerage Systems with BIM, Applied Sciences, flood forecasting, J. Hydrol. (2023) 129702, https://doi.org/10.1016/j.
2022, https://doi.org/10.3390/app12189063. jhydrol.2023.129702.
[79] J. Drgoňa, D. Picard, M. Kvasnica, L. Helsen, Approximate model predictive [91] S. Chang, N. Saha, D. Castro-Lacouture, P.P.-J. Yang, Multivariate relationships
building control via machine learning, Appl. Energy (2018), https://doi.org/ between campus design parameters and energy performance using reinforcement
10.1016/j.apenergy.2018.02.156. learning and parametric modeling, Appl. Energy (2019) 253–264, https://doi.org/
[80] W. Liao, Y. Huang, Z. Zheng, X. Lu, Intelligent generative structural design method 10.1016/j.apenergy.2019.04.109.
for shear wall building based on “fused-text-image-to-image” generative [92] P. Liu, H. Qi, J. Liu, L. Feng, D. Li, J. Guo, Automated clash resolution for
adversarial networks, Expert Syst. Appl. (2022), https://doi.org/10.1016/j. reinforcement steel design in precast concrete wall panels via generative
eswa.2022.118530. adversarial network and reinforcement learning, Adv. Eng. Inform. (2023), https://
[81] B. Fu, Y. Gao, W. Wang, Dual generative adversarial networks for automated doi.org/10.1016/j.aei.2023.102131.
component layout design of steel frame-brace structures, Autom. Constr. (2023), [93] P. Sun, F. Yan, Q. He, H. Liu, The development of an experimental framework to
https://doi.org/10.1016/j.autcon.2022.104661. explore the generative design preference of a machine learning-assisted residential
[82] D. Yao, Q. Sun, J. Yang, H. Liu, J. Zhang, Railway Fastener Fault Diagnosis Based site plan layout, Land (2023), https://doi.org/10.3390/land12091776.
on Generative Adversarial Network and Residual Network Model, Shock and [94] P. Geyer, S. Singaravel, Component-based machine learning for performance
Vibration, 2020, https://doi.org/10.1155/2020/8823050 pp1–15. prediction in building design, Appl. Energy (2018), https://doi.org/10.1016/j.
[83] J. Wang, K. Chen, H. Yang, L. Zhang, Ensemble deep learning enabled multi- apenergy.2018.07.011.
condition generative design of aerial building machine considering uncertainties, [95] Y. He, X.-H. Liu, H.-L. Zhang, W. Zheng, F.-Y. Zhao, M. Aurel Schnabel, Y. Mei,
Autom. Constr. (2023), https://doi.org/10.1016/j.autcon.2023.105134. Hybrid Framework for Rapid Evaluation of Wind Environment around Buildings
[84] R. Li, H.-L. Chi, Z. Peng, X. Li, A. Chan, Automatic tower crane layout planning through Parametric Design, CFD Simulation, Image Processing and Machine
system for high-rise building construction using generative adversarial network, Learning, Sustainable Cities and Society, 2021, https://doi.org/10.1016/j.
Adv. Eng. Inform. (2023), https://doi.org/10.1016/j.aei.2023.102202. scs.2021.103092.
[85] G. Chen, S. Teng, M. Lin, X. Yang, X. Sun, Crack detection based on generative [96] X. Li, Y. Yuan, G. Liu, Z. Han, R. Stouffs, A predictive model for daylight
adversarial networks and deep learning, KSCE J. Civ. Eng. (2022), https://doi.org/ performance based on multimodal generative adversarial networks at the early
10.1007/s12205-022-0518-2. design stage, Energy & Buildings. (2024), https://doi.org/10.1016/j.
[86] G. Zou, Z. Lai, C. Ma, Y. Li, T. Wang, A novel spatio-temporal generative inference enbuild.2023.113876.
network for predicting the long-term highway traffic speed, Transp. Res. C (2023), [97] K. Feng, W. Lu, Y. Wang, Assessing environmental performance in early building
https://doi.org/10.1016/j.trc.2023.104263. design stage: an integrated parametric design and machine learning method,
[87] S. Hartmann, M. Weinmann, R. Wessel, R. Klein, R. Network, StreetGAN: Towards Sustain. Cities Soc. (2019), https://doi.org/10.1016/j.scs.2019.101596.
Road Network Synthesis with Generative Adversarial Networks, WSCG 2017– 25th [98] S. Quan, J. Park, A. Economou, S. Lee, Artificial intelligence-aided design: smart
Computer Graphics, Visualization and Computer Vision, 2017 [accessed 22 design for sustainable city development, Environment and Planning B: Urban
December 2023]. Analytics and City Science. (2019), https://doi.org/10.1177/2399808319867946.
[88] A. Chaurasia, A. Gautam, R. Rajkumar, A.S. Chander, Road traffic optimization [99] P. Ghannad, Y.-C. Lee, Automated modular housing design using a module
using image processing and clustering algorithms, Adv. Eng. Softw. (2023), configuration algorithm and a coupled generative adversarial network (CoGAN),
https://doi.org/10.1016/j.advengsoft.2023.103460. Autom. Constr. (2022), https://doi.org/10.1016/j.autcon.2022.104234.

21