Multi-Agent Decision-Making in Construction Engineering and Management: A Systematic Review

Abstract

Construction engineering and management (CEM) involves multiple stakeholders, complex interest relationships, and conflicts. All stakeholders must comprehensively consider the interests relating to a project to make decisions. Appropriate multi-agent decision-making can effectively coordinate and integrate the needs or expectations of all stakeholders, which can reduce conflicts, improve the success probability of the project, maximize the overall returns on interest, and contribute to the project’s sustainability. Existing studies have begun to explore the associated theoretical framework and practical methods for multi-agent decision-making. However, early studies mainly focused on the conceptual theories of decision-making models and processes, such as stakeholder analysis, decision evaluation, and risk management. In recent years, increasing research has concerned the application of multi-agent decision-making in CEM. This paper is the first to use a comprehensive review approach to analyze multi-agent decision-making in CEM, providing an overall perspective. In this paper, 105 journal papers are identified and classified into four categories: (1) main concerns regarding multi-agent decision-making in CEM; (2) tools and methods of multi-agent decision-making in practice; (3) research tools and methods of multi-agent decision-making in CEM; (4) critical points on how to solve multi-agent decision-making problems. The findings of this study offer references for future trends in four regards, namely (1) knowledge management, (2) decision resilience, (3) the fusion of many techniques, and (4) technologies for future research.

1. Introduction

The construction industry faces new challenges brought about by the proliferation of construction projects, new cooperation modes between stakeholders, and digital transformation [1]. CEM has run into numerous roadblocks in various complex contexts, including massive information flows, tasks involving multiple stakeholders, diversified subjects, and many risk factors, necessitating a re-evaluation of the industry’s future trajectory [2]. The construction industry requires thoughtful decision-making. Due to its intricate and time-consuming nature, multi-agent decision-making has emerged as one of the most significant issues facing the construction industry [3]. Multi-agent decision-making offers significant advantages over other theoretical approaches. It improves the inclusivity and comprehensiveness of decision-making by combining stakeholders’ experiences from many areas of expertise. Mutual correction among different stakeholders effectively spreads risks and minimizes the incidence of accidents, increasing the legitimacy and transparency of choices and lowering subsequent disagreements. By utilizing all available information and knowledge from many stakeholders, the decision’s reliability can be reviewed more broadly, enhancing the decision’s long-term efficacy. Multi-agent decision-making also encourages project innovation in terms of sustainable development and management from a fresh perspective.

Studying the creative and cooperative behaviors of the participants in the collaborative conduct of projects is essential from a practical perspective [4]. For instance, the general contractor in the Hong Kong–Zhuhai–Macao Bridge project successfully executed the submerged tube docking window forecast and guarantee system, as well as the real-time motion attitude monitoring system, thanks to extensive collaboration. This enabled the seamless installation of immersed tubes in a deep trench [5]. However, there are many instances of projects failing because multi-agent collaboration was not given sufficient thought. Taking the Dabo Power Plant as an example, disagreement over the stakeholders’ payment of electricity tariffs has crippled India’s independent power plants and severely harmed the country’s attempts to draw in foreign investment.

The rise of diverse delivery, investment, and finance methods and business models creates a complicated network of relationships and new avenues for collaboration. Stakeholders must adjust to these models and make judgments accordingly. Divergent perspectives and attitudes among decision-makers on a project frequently result in conflicts and controversies due to their lack of adaptive and dynamic behaviors [6]. Often, the goal of the parties engaged in a construction project is to maximize their interests, while they ignore the goal of achieving the relatively equal optimization of the needs of all parties and the best project performance overall [7]. This fuels construction and management process divergences [8].

Many prior attempts have been made to solve those problems. Multi-criteria decision analysis (MCDA) is suggested to enhance the project’s chances of success and improve decision realization [9]. The two most useful tools are visualization technology and BIM. Marzouk, M. et al. (2023) attempted to examine collaboration in the context of green construction management processes and employed BIM dynamic simulation tools to increase construction efficiency and address the issues of achieving multi-stakeholder decision-making collaboration [10]. In the process of researching the relationship between BIM technology and digital twins, it was determined that a close relationship between data management and BIM technology is essential for project decision-making as it accurately simulates the project’s current state, forecasts its future development, and offers an approach to reducing risks [11,12].

However, most academics from various organizations approach multi-agent decision-making from a single model and algorithmic perspective, and the present research in this area needs more holistic cognition. To support researchers who are involved in multi-agent decision-making with their analysis of the development trend and exploration of future research directions, this paper gives an overview of the topic’s current research status and clusters, analyzes and summarizes the characteristics of each cluster, and investigates the current research status and gaps. This study lays a solid foundation for improving administrative management tools and policy promulgation. Furthermore, it provides a theoretical basis for updating the standards in the industry of CEM and introducing incentives related to multi-agent decision-making.

2. Materials and Methods

2.1. Search Rules

In this study, we retrieved journal articles from 2012 to 2023 using the keywords “multi-agent decision AND construction”, “collaborative decision AND construction”, “multi-agent decision AND construction”, or “multi-stakeholder decision AND construction” in the Web of Science core collection. A total of 137 papers were chosen for additional analysis and review [13]. This ensured a high-quality coverage of the existing literature on multi-agent decision-making. When they are employed in different nations and locations, the terms “multi-agent decision”, “collaboration decision”, “multi-agent construction decision”, and “multi-stakeholder decision” have similar meanings. After their extraction, all 137 records were downloaded as complete text files (TXT). The files include essential information for literature analysis, such as author, title, source, abstract, cited references, document type, times cited, keywords, publisher information, etc. The procedure for retrieving papers is as follows in Figure 1. The guidelines on Preferred Reporting Items for Systematic Reviews and Metanalyses (PRISMA) were used (see Supplementary Materials).

• Topics were scanned with the search rule mentioned above in the Web of Science database for 1 January 2012 to 6 January 2023 and the language of English. This rule retrieved 645 papers (including articles, proceedings papers, editorials, and reviews).
• The abstract of each paper was read to exclude irrelevant ones. Finally, 137 papers were selected for the literature review.
• Papers in conference proceedings and editorials were rejected, leaving 105 articles.

2.2. Bibliometric Approach

Bibliometric analysis is a quantitative and visualization approach to document analysis that applies mathematical and statistical methods to reveal the bibliographic features and regular patterns of a body of literature, such as attribute distributions, network relationships, research hotspots, and trends. As a software tool for constructing and visualizing bibliometric networks, VOSviewer 1.6.19 was employed to generate network diagrams to visualize the relationships between authors, articles, and keyword terms of the literature. The strength of the bibliometric approach is that it shows the research development quantitatively and visually, which is conducive to mining deep information from the literature as a whole. The results of bibliometric analysis can also assist in the design of a logical framework for the literature review [13,14].

3. Bibliometric Analysis

3.1. Journals

Initially, 645 papers were identified by literature retrieval following the search rules, including journal articles, reviews, and books. Then, the authors skimmed the full abstracts of the 645 papers to remove the irrelevant ones. After the manual screening, 137 journal articles were finally selected for further analysis and their distribution in core journals from which were extracted is shown in

Table 1. Identified academic journals and articles.

ID	SOURCE	NUMBER
1	Architectural Engineering and Design Management	16
2	Journal of Construction Engineering and Management	8
3	Automation In Construction	8
4	Building and Environment	7
5	Buildings	6
6	Construction Innovation-England	5
7	Construction Management and Economics	4
8	Energies	4
9	Engineering Construction and Architectural Management	3
10	Frontiers of Engineering Management	3
11	IEEE Access	3
12	International Journal of Construction Management	2
13	International Journal of Disaster Risk Reduction	2
14	Journal of Building Engineering	2
15	Journal of Civil Engineering and Management	2
16	Journal of Computing in Civil Engineering	2
17	Journal of Information Technology in Construction	2
18	Journal of Legal Affairs and Dispute Resolution in Engineering and Construction	2
19	Journal of Management in Engineering	2
20	Safety Science	2
21	Smart and Sustainable Built Environment	2
22	Sustainability	4
23	Sustainable Cities and Society	2

.

After further reading the abstracts and papers’ content, a total of 105 papers were taken.

3.2. Years

By tracing and studying the number of papers published in recent years, the annual changes in the number of papers published and the cumulative number of papers published from 2012 to 2023 are plotted, as shown in Figure 2. (the deadline for the 2024 paper count is 20 July). The number of relevant papers has fluctuated, especially since 2018, which indicates that researchers worldwide have become increasingly interested in multi-agent decision-making over the past decade.

3.3. Countries and Territories

Overall, the selected papers were published by authors from 50 countries/regions worldwide.

Table 2. List of national distribution.

ID	Country	Documents
1	China	44
2	USA	21
3	Australia	17
4	United Kingdom	11
5	Canada	10
6	Malaysia	7
7	Netherlands	7
8	Egypt	5
9	France	4
10	New Zealand	4
11	Singapore	4
12	Switzerland	4

lists the countries with four or more relevant publications. China is the top contributor, accounting for over one third of all publications; the United States and Australia come in second and third, with contributions of 15.33% and 12.41%, respectively. The table also lists other European nations, such as the UK, the Netherlands, France, and so on, making European nations the majority.

Research collaborations between countries with four or more publications are demonstrated in Figure 3. China occupies the central position in the figure with 44 documents, connecting closely to Australia and New Zealand in the same cluster. China is also linked directly to other countries not in the same clusters, such as the UK, Canada, Netherlands, Singapore and France. The network diagram shows that international collaboration on the research topic has become more popular, and more researchers from different countries tend to join the collaboration even though they are different in terms of culture, language, and geographical location. With the development of large-scale projects, international collaborations are increasingly popular. Despite the differences in culture, language, and geography of different countries, more and more researchers are inclined to join the collaboration and promote smart decision-making.

3.4. Co-Occurrence Analysis

A co-occurrence analysis of text items is employed to uncover the logical structure of current research in Figure 4. The frequency with which two items appear in the same record refers to the co-occurrence relationship of text items. A cluster represents a significant area of study or current line of inquiry. Key text pieces were found in the titles and abstracts of the chosen publications to conduct the co-occurrence analysis. The frequency of these text items occurring together in the same document was then calculated, which indicates the correlation between the text items.

The threshold of the text item occurrences was set as two, which defines the minimum frequency of occurrences of a text item in a single document. Of the 855 identified terms, 157 met the threshold. It was then selected for creating the co-occurrence map shown in Figure 3, which was divided into 11 clusters: red (24 items), green (24 items), blue (16 items), yellow (14 items), purple (14 items), wathet blue (14 items), orange (12 items), brown (12 items), pinkish purple (10 items), pink (9 items), light green (8 items). Nodes in the network map stand for the selected text items, and the size of the nodes represents the frequency of occurrences, as shown in Figure 4.

Four additional categories were identified after secondary analysis and summing the 11 clusters. The first new category is the largest of the four categories, including the red, brown, and pinkish purple clusters, in which key items include “design management, communication, energy, environment, health, technology, quality”. The second category has two clusters, in which key items include “BIM, information, building information modeling, integrated design process”. The third category mainly includes the green, purple, and orange clusters, the key items of which are agent-based modeling, life-cycle construction, software, knowledge management, optimization, supply chain management, and network. The fourth category has clusters of three colors, including light blue, blue, and pink clusters, whose keywords are “collaboration, trust, innovation, perceptions, success factors, governance”.

The literature on each of these elements was compiled to identify and classify the subjects related to the various classifications. The following four combing types were established by recognizing several characteristics of the four orientations connected to multi-subject decision-making, as illustrated in Figure 5.

4. Critical Review

Based on the four categories in Section 3.4, a critical review was conducted from four perspectives in Figure 6:

(1)

Primary concerns for multi-agent decision-making in CEM.

(2)

Tools and methods to do multi-agent decision-making in practice.

(3)

Research tools and methods of multi-agent decision-making in CEM.

(4)

Key points to solving multi-agent decision-making problems.

4.1. Main Concerns for Multi-Agent Decision-Making in CEM

Multi-stakeholder decision-making in construction and management is interesting for a variety of reasons. Initially, decisions must reflect the various stakeholders’ interests in safety [1]. Cooperative decision-making can guarantee that a building’s design, material selection, construction techniques, and maintenance procedures all adhere to pertinent safety standards; cooperative multi-stakeholder decision-making can lessen a building’s environmental impact and achieve a small carbon footprint [14]; and at the design stage, there may be competing needs and expectations of various stakeholders. When decision-making involves multiple stakeholders, it is simpler to balance their interests and guarantee that the design solution will satisfy everyone’s needs and be completed on schedule and under budget [15].

4.1.1. Safety Management

In construction projects, using mechanical equipment and personnel specifications can result in an array of safety concerns that may result in significant financial losses or even fatalities [16]. Various stakeholders are involved in the safety management of construction projects, and their cooperation affects the decision-making process for safe building and the effectiveness of safety governance [17]. The safety management network of CEM is a complex social–technical system, and poor contractual and regulatory relationships between stakeholders indirectly lead to accidents. Procedure, creativity, meritocracy, and risk are the characteristics of multi-agent decision-making in safety management [18]. Using scientific safety decision-making approaches and procedures to select possibilities, decision-making on safety management activities can increase the quality of decision-making. For example, a case study in Denmark proved the importance of multi-agent decision-making for on-site construction safety. Multiple project participants, such as project owners, project managers, supervisors, and site managers, examined the on-site safety management of three construction projects and made reasonable decisions on personnel adjustments to avoid the project’s economic losses due to safety problems [19].

4.1.2. Green and Low Carbon

The construction industry has a high energy consumption and carbon emission ratio. Energy saving and emission reduction are urgent. The construction phase accounts for 20–50% of total carbon emissions [20], and the operation and maintenance phase accounts for about 55–60% [21]. Multiple stakeholders involved in the whole project process need to make decisions about green and low carbon during the project implementation. There are also secondary stakeholders such as the government, investors, and neighbors. The stakeholder relationships differ among different retrofit options, so the choice of options significantly impacts the project’s quality.

Green construction technology is one of the concerns of multi-stakeholder decision-making. It also helps to optimize the emission reduction during the construction phase of the building. The adoption rate of green technology depends on the different interactive characteristics of the decision-makers [18,21]. The use of OSC technology during the project process maintains a different attitude among stakeholders’ views. Tezel, A. et al. (2023) asked about the current status and future development proposals of OSC on UK highways by interviewing senior project managers of five major highway projects in the UK [22]. The suggestions include reducing early decision-making errors, freezing the status of each stage promptly, coordinating off-site construction (OSC) decision-making among project stakeholders, and proposing a framework for multi-agency collaborative OSC decision-making, which supports multi-agent synergistic decision-making [23,24]. To promote green sustainable development, the introduction of relevant policies by the state can influence multi-agent decision-making at the macro level, improve policy performance, and stimulate the continuous progress of green professional practitioners [3,25]. Li, Y. et al. (2021) used content mining and social network analysis to provide a comprehensive approach for evaluating the policy performance of green building initiatives to assess green building policies in China [26,27]. The behavioral consistency of the governmental agencies indicates that they are capable of coordinating the resources, accelerating the transfer of information in the implementation of the decisions, and forming an adaptive mechanism of the collaborative network to direct the enhancement of green building initiatives.

4.1.3. Design Management

The design phase takes place in the pre-project phase and continuously impacts decision-making throughout the project life cycle [28]. Proper design provides clear direction to the stakeholders involved in the project for decision-making, allowing the project to proceed smoothly [29]. Currently, design management is supported by more mature technologies. For example, the construction process of the Sydney Opera House used design management techniques to integrate management and design through the cross-analysis of different documents and the use of a digital project collaboration platform [30]. Further influences of decisions in the design phase are taken into account. Liu, N. et al. (2022) consider the interactions among construction decision-makers that influence the quality of design decisions and their safety and health outcomes in construction projects. Social–technical network theory helps to obtain an effective interaction network that provides decision-makers with appropriate sources of knowledge to improve the safety of design outcomes and reduce the uncertainty associated with decision-making [31].

4.1.4. Dynamic Planning

The quality of decision-making in current construction projects depends on the dynamic adjustment of the various stakeholders in the project’s work. As a result, wise decision-makers often make dynamic plans to coordinate the interests of all parties in a complex and changing environment for the best decision-making results. The interactive behavior between stakeholders is important to the overall project performance, which in turn affects the project’s outcome [6]. Therefore, different incentives, rewards and penalties, and various distributional strategies are ways to provide dynamic adjustments to the final decision. Research is conducted to consider the impact of rewards and punishments, innovation benefit distribution, and other factors on the behavior of decision-makers. Pramono, RWD (2022) selected eight residential development projects as case studies and constructed a framework involving different stakeholders to evaluate and simulate urban land use decision-making [32]. Results show that the realization of urban development results from the interaction of stakeholders, and the optimal decision-making is reached through the games and adjustments of the multi-actors.

4.2. Tools and Methods to Do Multi-Agent Decision-Making in Practice

Stakeholders involved in CEM have different needs, so it is necessary to consider the interests of all parties systematically to make decisions. Validating the scientific nature of decision-making requires applicable practical tools [33]. BIM technology can integrate information from various aspects, such as architecture, structure, and equipment, providing a collaborative work platform that facilitates information sharing and collaboration among different stakeholders. Information sharing deals with different and heterogeneous data, and data information technology can handle massive amounts of data to deliver predictive analytics, real-time decision assistance, and intelligent suggestions to suit the needs of multiple stakeholders. To further help all parties understand the project, data visualization and design through extended-reality technology, virtual reality (VR) construction, and OSC technology are combined to reduce construction disruption and meet different stakeholders’ needs. Life-Cycle Analysis (LCA) is now commonly utilized in sustainable building and project evaluation to help all parties better understand projects’ long-term implications and make informed decisions.

4.2.1. BIM Technologies

The application of BIM technology during practice in the field of engineering and construction provides technological support for multi-agent decision-making. Previous research on BIM is shown in

Table 3. The Usage of BIM for Multi-Agent Decision-Making.

Project Number	Design	Construction	Decision-Making	Cost	Production	Information Data	Simulation	Collaboration	Reference
1						√		√	[6]
2			√	√				√	[9]
3							√		[10]
4		√		√					[11]
5			√			√			[17]
6	√					√			[18]
7	√								[20]
8	√	√				√		√	[29]
9			√			√	√		[31]
10	√						√		[32]
11		√	√				√		[35]
12		√						√	[36]
13	√					√			[37]
14			√	√					[38]
15		√	√			√	√		[39]
16			√			√	√		[40]
17									[41]
18				√	√		√		[42]
19			√					√	[43]
20			√	√					[44]
21	√		√				√	√	[45]
22			√					√	[46]
23	√				√		√	√	[47]

“√” indicates that the author has conducted relevant research in their paper.

. As a way of integrating information, BIM is used throughout the entire life cycle of a construction project, enabling all participants to access the necessary information at the right time, helping to understand the project, and reducing common errors. A rapid and virtual environment commonly used to simulate and explore various construction project options, whose goal is to improve project interfaces, facilitate communication between all stakeholders, and provide a framework for building data-rich product models [34]. A collaborative framework for green construction management, utilizing automated and semiautomated simulations to facilitate performance-based multi-agent decision-making, used BIM dynamic simulation tools to improve performance and meet the Green Pyramid Rating System (GPRS) [10]. BIM technology can simulate realistic scenarios in construction management, but other phases dwarf the implementation of BIM in the operations phase. In a Facilities Management (FM) environment, BIM can actively store documents and other data used in the FM phase, with great potential to support multi-agency collaboration [35].

Construction presents uncertainty and risk due to unforeseen circumstances, intricate designs, and ineffective information management. BIM in the construction industry provides solutions to these problems through effective data information modeling. BIM technology and digital twins provide opportunities for decision support in complex situations and comprehensive data management of construction projects in the construction industry [36,37]. In the practice of engineering and construction, Sharafat, A. et al. (2021) novel multi-model tunnel information modeling (TIM) framework based on BIM. The framework facilitates data sharing, information integration, data accessibility, design optimization, project communication, efficient project management, and scientific decision-making during construction [35]. BIM technology has also shown its advantages for railroad project cost control [48], decision support, avoiding extra work due to design errors, helping in prefabrication and facility management, etc. Integration of BIM in construction is becoming a global trend [38].

In the growing field of new technologies in engineering and construction, BIM technology has emerged with several new approaches. The integration between 4D BIM and digital project documents provides a new collective decision-making device based on 4D stepping forward. These elements help to propose 4D BIM as a decision-making (DM) support for Architecture Engineering Construction projects [39]. The 4D BIM technology can also be applied to reduce problems and conflicts through 3D coordination involving new practices, contractual standards, and related equipment and software, understanding of new tasks for multiple subjects, increased subject participation through coordination, and enhanced communication [40].

4.2.2. Data Information Technology

Data information technology is becoming increasingly feasible in CEM as data accumulates and computational power grows. It can help different parties understand the project’s position and make more informed decisions. The construction industry generates a large amount of heterogeneous and dynamic data, which is characterized by being dispersed throughout projects LCA. Instant and accurate access to these data is fundamental to management, decision-making, and analysis for participants involved in different project life-cycle phases [41]. Different scholars in recent years have investigated models for data collection, analysis, integration, etc., and applied them in different scenarios. Wu et al. (2023) proposed an information technology integration tool for the construction industry that helps insert the building performance requirements for different stakeholders throughout the construction project phases to support integrating this information in the BIM process [40]. Rodrigues, F et al. (2022) recognize that the combination of 3D parametric models with business intelligence tools is in line with the current trends in the construction industry, leading to the visualization of the data and its interconnection with the same database (BIM model) to provide technical support for sustainable construction management [39]. Dong, Jichang, et al. (2021) explored the application of information technology such as Big Data and Artificial Intelligence (AI) as an example. They build a pavement management system through information integration to optimize the workflow, improve decision-making efficiency, and provide automated and intelligent solutions for relevant management departments and companies [41].

The dynamic and increasing complexity of construction projects brings many challenges to project planning and control, incorporating resource specifications and limitations as well as complex multi-factorial constraints into the planning process and integrating intelligent information gathering and decision support systems to achieve progressive refinement [11]. Information intelligence also provides the convenience of remote sharing, examining the prospects offered by telematics and analytics for more informed and integrated collaborative management decisions on building sites.

4.2.3. Off-Site Construction

Off-site construction (OSC) techniques arose due to the construction industry’s need for efficiency, sustainability, and quality assurance, as well as to fulfill the needs of different stakeholders. In the past, decision-making systems in prefabricated construction lacked efficiency and collaboration because decision-related information was stored and managed in heterogeneous systems by different stakeholders, who existed spatially compartmentalized. Thus, subjective decision-making in the field of OSC technology is also worthy of attention [44]. Empirical evidence on the performance of Prefabricated Prefinished Volumetric Construction (PPVC) projects continues to show that the ultimate success of a PPVC project is directly related to the key decisions made in the early stages of the PPVC project life cycle. Wuni et al. (2020) conducted a questionnaire survey by creating a list of nine CSFs related to the early stages of the entire PPVC project life cycle to identify and assess the key factors at different stages. They concluded that the most influential CSFs include good collaboration between project participants, effective communication and information sharing, effective stakeholder management, etc. [49,50].

Prefabricated construction cannot be separated from the logistics and supply chain, and each stakeholder’s decision-making affects the realization of the project’s agentives, the development of the design plan, the selection of materials and equipment, and the conduct of construction. Therefore, the cooperation and coordination of the various stakeholders in both design and cost is crucial to ensure the project’s smooth implementation and successful completion [51]. In exploring Modular Integrated Construction (MIC), Tezel, A et al. (2023) considers the manufacturability, logistics, and assembly constraints of OSC components to design OSC components [22]. Design for Manufacturing and Assembly (DFMA) and an intelligent asset monitoring system were used to coordinate the OSC decisions of project stakeholders [47]. Decisions made during the design phase directly impact project functionality, efficiency, and the subsequent cost and construction process. Currently, widely used DFMA does not correlate cost estimation with OSC. Vakaj, E. et al. (2023) proposed a new domain ontology [28]. Off-site Housing Ontology (OHO) uses the NEON methodology framework to support cost estimation, considering products, resources, and production processes, and to provide a solid foundation for improving the efficiency of collaborative design management. Hussein M et al. (2023) constructed hybrid multi-agent simulation models and revealed the interaction effects of logistics and construction decisions through the design of experiments. The results show that some logistics decisions can significantly affect construction KPIs, and decisions that cannot affect KPIs need to enhance cooperation among stakeholders [52].

However, current construction is still a combination of on-site and OSC to be more relevant. Karamoozian (2020) utilized the Analytic Hierarchy Process (AHP), which combines off-site and on-site construction methods, to plan a multi-criteria DMAS, which identifies the key decision-making agents and housing choices for fast, appropriate, and cost-effective large-scale housing, and helps the decision-makers to make better decisions [53]. The emergence of various technologies requires us to integrate them and to improve the efficiency of decision-making. He proposed a conceptual framework for BIM and PHP integration, including Smart BIM Platform, Smart Work Package, and Smart PHP agents, which helps in the development of a system architecture for BIM and PHP to benefit various stakeholders, facilitate integration.

4.2.4. Extended Reality

Extended Reality uses in construction mostly involve improving construction program design, increasing construction safety, and decorating and renovating buildings. Extended Reality (XR) refers to three upcoming technologies: virtual reality (VR), augmented reality (AR), and mixed reality.

The application of extended reality in the whole life cycle of construction helps the stakeholders to make decisions more clearly and intuitively. MR is a visualization technology that improves the visual perception of a facility by overlaying 3D virtual agents and textual information on views of real-world building agents, discussing how a collaborative BIM-based MR approach can be developed to support facility site tasks. Interactive Virtual Collaboration (IVC) mode communicates with the office while supporting effective decision-making by field staff on existing and potential problems [51]. AR provides useful auxiliary information to field staff who operate and maintain facilities. MR can provide a way to remotely operate facilities in an immersive environment. The combination of the two technologies can simultaneously support communication and improve collaboration between subjects that do not intersect in space. The key to implementing VR in the design assessment phase is the creation of collaborative virtual environments that allow multiple stakeholders to interact, communicate, and collaborate virtually at the same time during the design assessment phase, providing them with informed decision-making or consensus before the start of construction activities [9]. VR is more intuitive than traditional tools and improves planning accuracy and collaboration. It brings significant economic savings and benefits to the construction industry. Future research should investigate the real-world cost-benefit ratio of VR and streamline its technical implementation in construction projects, contributing to the current body of knowledge, providing real-life economic benefits, and addressing research gaps in academia and industry to promote the widespread use of VR [22].

4.3. Research Tools and Methods of Multi-Agent Decision-Making in CEM

Many stakeholders are involved in the process of engineering construction and management, and the cooperation and conflict between stakeholders and the relationship between them constitute a largely invisible network [52]. The network perspective is used to reveal the relationship among stakeholders, communication channels, and the way information is transferred, thus helping to understand the information flow and influence in multi-agent decision-making. The “invisible web” influences decision-making at the macro level and is used to explore the behavior and interactions between subjects to help future researchers better understand the dynamic evolution of decision-making behaviors. The behavior of decision-making parties is often accompanied by competition and conflict. To improve the quality of multi-agent decision-making, through the system dynamics, ABM simulation [53], and other methods, the actual data of the construction of engineering projects into the simulation and simulation, the use of evolutionary game and game theory is widely used to predict the results of different decision-making strategies, to provide a more rational and robust decision-making basis for the cooperation of multiple parties [54]. A lack of decision-making often accompanies game-based prediction, MCDM methods to find the optimal decision-making scheme, the overall interests of the maximization of the balance of the interests of all parties, to improve the efficiency of decision-making and the quality of the results [55].

4.3.1. Network Perspective

Due to the numerous person movements, the CEM participants base has gradually increased, constituting a great network of relationships. Studying the relationship of information flow among multiple stakeholders through networks can help build more robust networks of organizations and achieve better decision-making procedures [56,57]. Therefore, a network perspective was adopted to study stakeholders’ decision-making, to understand the interaction patterns among stakeholders through a network perspective, to identify key players and decision-influencing factors, and thus to improve the transparency and efficiency of the decision-making process. The social network perspective has a unique advantage in clarifying the relationship between multiple stakeholders in the engineering field, and the social network analysis method is used as the basis for constructing a model of collaborative safety governance structure to provide theoretical support for safety management [1].

Social network analysis (SNA) has received increasing attention as a tool for mapping network relationships. It can be used to assess the flow of information between project teams, synergize multi-agent decision-making efforts, and improve performance [55]. The whole blueprint for collaborative action can be seen clearly. The classical network governance theory (NG) is used as an entry point to introduce it into the background. The construction of the model is realizing the network governance with benign interaction of multi-participants [12]. To understand the complex collaborative process, an integrated framework consisting of three parts, namely data collection system updating, data preprocessing, and social network analysis, is constructed for pairing and mining collaborative networks in construction projects to increase the frequency of collaboration and communication further. In recent years, many scholars have recognized the potential of combining network perspectives with engineering and construction. Heikinheimo et al. (2018) innovatively investigated the fit between Actor-Network Theory and the field of engineering by investigating cooperation, collaboration, and decision-making in architects’ firms during the construction period, demonstrating the importance of collaborative decision-making in construction projects [58].

4.3.2. Simulation

The construction process is a complicated system comprising a range of procedures, which are frequently broken down into smaller procedures for simplicity of comprehension and simulation. With the concept of lean construction, the problem of planning and managing material supply in the construction industry needs to take into account the behavior of its decision-makers. Robles et al. (2022) used system dynamics, supported by the AHP supplier selection methodology [59], ABC classification evaluation and monitoring, to develop a model that can optimize the use of resources utilizing the Vensim data analysis software (Vensim PLE 9.2.3) and to identify the key variables to evaluate their contribution and to make the right decisions. The wide range of applications of system dynamics and the use of system dynamics for synergistic management between different regions can be effectively improved to enhance the implementation ability of cross-regional policies in practice [60].

Multi-agent decision-making in construction is related to ABM, and social emergence is a key idea in ABM. In this society, it is not the simple addition of all individual qualities but the consequence of individual characteristics-guided interaction. The interaction between multiple subjects in the process of engineering construction is also complex and variable; it explores the application of ABM in the field of engineering construction. ABM is a method of modeling systems to simulate agent system behavior with the organizational behavior of a Multi-Agent System (MAS) to ensure system reliability. MAS is a powerful tool for studying complex resource allocation and competition in the process of concurrent execution of construction programs that focuses on designing agents to solve specific practical or engineering problems. Together, these studies focus on applying MAS and simulation models to improve the efficiency of project management and the accuracy of decision-making for successful project advancement and maximization of overall benefits. These methods are expected to address several issues in architectural and engineering project management, such as risk management, design revisions, and layout planning.

In CEM, the common goal of various studies is to improve decision-making efficiency and project performance, with particular attention paid to the application of MAS and simulation models. Some experts and scholars focused on solving the problems of congestion and cumbersome processes and analyzes the behaviors of each agent on the project schedule control by defining the mechanisms of interaction and competition between the agents. The impact of each agent’s behavior on project schedule control is analyzed by defining the interaction and competition mechanism between agents. This method is expected to play an active role in improving construction efficiency and optimizing decision-making.

Meanwhile, risk management is of great significance in engineering and construction projects. The MAS for risk allocation in PPP projects aims to determine appropriate risk allocation decisions for an efficient and accurate risk allocation scheme.

This approach helps to reduce project risks, improve decision-making efficiency, and ensure the smooth progress of the project. On the other hand, design changes are often part of the risk in construction projects and, therefore, need to be corrected on time to avoid negative impacts on project performance. The study by Du et al. (2019) addresses the risk of design changes in prefabricated assembled building projects by proposing a multi-agent-based simulation model for evaluating the effectiveness of design change management and testing the validity of different management strategies to provide support for managerial decision-making [30].

There are some methods for dealing with uncertainty in multi-agent decision-making in CEM and reducing variability in non-linear analysis. Jiang S et al. (2022) have innovatively introduced trade associations and idle penalties to establish a three-party dynamic evolutionary game model of government, developers, and buyers [3]. Sensitivity analysis of decision-making behavior is carried out through numerical simulation. Durdyev (2022) identified 20 barriers affecting the prioritization of engineering decisions using a hybrid parsimonious fuzzy AHP approach [34]. Mooselu has established a risk-based optimization framework to form a multilayer perceptron artificial network (MLP-ANN) agent model [8].

In addition, Construction Site Layout Planning is an important area of concern that is directly related to cost, time, and quality. Researchers constructed an agent-based decentralized two-layer mathematical model covering multiple stakeholder decisions and iteratively updated individual decisions to find the optimal integrated solution to reduce potential stakeholder conflicts during the construction phase to improve project efficiency [7]. The scholars used agent modeling and neural network techniques to simulate the realization and dynamic adjustment of project benefits as a complex integrated system to maximize and sustain the overall benefits of PPP projects [6].

4.3.3. Game Theory

Game theory can help discover conflicts of interest and opportunities for collaboration among stakeholders. Multiple people are involved in engineering building projects, and their interests are frequently linked in complex ways. The advantage of evolutionary games is that they can better simulate and predict the strategic adjustments of the participants in a dynamic environment, thus providing a more accurate reference for decision-making. Decision-making in the development process needs to consider the decision-makers’ choices, and a hierarchical analysis method based on satisfactory options is used, and an agent system based on game theory is employed in the coalition formation process to combine value-based decision-making, group decision-making, and collaborative support [61]. The coalition formation agreement option is applied to select the most appropriate program. With the decrease in new construction projects and the gradual rise of retrofit projects, green retrofit projects have great research potential worldwide. Some applied evolutionary game theory to determine the behavioral multiple interaction mechanisms of agents and then simulated it to obtain a method to stimulate the interaction of multiple agents (government, developer, and occupants) participating in the green retrofit of commercial buildings in China [3,25], others used evolutionary game systems to reveal the evolutionary trends under the tripartite game conditions of the government, developers, and home buyers, as well as the evolutionary stabilization strategies of the dynamic tripartite to provide targeted measures and suggestions for the development of Green Residential Buildings (GRBs), emphasizing the importance of multi-agent collaboration for decision-making. Green renovation can realize the sustainable development of the project. However, the use of green supply chain management in the construction process can also realize the purpose of green and low carbon. Former scholars consider the characteristics of the multi-agents with solid dynamics and opportunistic behaviors. They set up an evolutionary game model to analyze and simulate the equilibrium determinants to formulate the subsidy coefficients according to the different simulation results. The green concept needs to run through the whole process of engineering construction, and reducing environmental pollution is also one of the links. Then they explored the evolutionary decision-making process and stabilization strategies among three stakeholders in the construction waste recycling industry such as governmental agencies), waste recycling enterprises, and waste producers, which provides a powerful tool and a theoretical framework to the stakeholders and provides an opportunity for them to make decisions on the green supply chain. Powerful tools and theoretical frameworks offer smarter decision-making and more sustainable development [62].

4.3.4. Multi-Criteria Decision-Making

Because of the changing nature of the business environment and a lack of stakeholder understanding, multi-stakeholder construction projects may pose risks. When performing project management duties, essential risk indicators for the project must be quantified. Managing these risks necessitates effective risk mitigation measures for assessing and analyzing their severity. The presence of information asymmetry makes it challenging to achieve Pareto efficiency. Risk assessment for these projects can be a critical component of the MCDM process to maintain a balanced degree of satisfaction for all parties. In real-world problems, the assessment of project risk is often uncertain or even incomplete; they evaluated risk criterion weights by extending the D-domain analytical network process (ANP) approach to handle fuzzy information and used known fuzzy preference relationships to determine the preference registration of each decision-maker’s decision matrix. Based on this, an extended multi-attributive border approximation area comparison (MABAC) method is proposed to rank and select risk response decisions. The D-ANP-NABAC method. It is based on MCDM and extends different decision-making frameworks, and effective decision-making requires the integration of theoretical perspectives with linked data and decision-maker behavior. The DMAS [63] assistant system can organize the necessary information and procedures to enable decision-makers to make informed and effective decisions [50]. To ensure stakeholder accessibility and mobility during construction, the construction supply chain uses the multi-criteria decision analysis methodology, which is progressively improved based on different cases to adapt it to the construction environment and help to recognize the necessity of applying building flow in projects [63]. As one of the MCDAs, Choosing by Advantages (CBA) creates a participatory, transparent, and auditable decision-making process based on comparing the advantages of alternatives to determine the significance of these advantages. Kpamma et al. (2018) selected a hospital in the Brong Ahafo region of Ghana to conduct a case study to enable stakeholders to decide on the design options using the CBA methodology [64]. The results showed that encouraging dialogue and respect among stakeholders to understand each other’s values, as well as determining the transparency of the design options, helped to create and maintain the quality of the collaboration and improve the efficiency of the decision-making. Marzouk et al. (2018) used agent-based simulation to model worker behavior in an evacuation situation [65]. The MassMotion simulation platform was utilized to implement the agent-based simulation. The Ranking and Selection statistical program determines the best simulation model configuration among the four alternatives considered. The technique for ranking similarity to ideal solutions can determine the best construction method alternatives, enhance construction safety awareness, and reach critical decisions.

4.4. Key Points to Solve Multi-Agent Decision-Making Problems

To solve the problem of multi-subject decision-making, the following keys to solving the problem are proposed through literature combing: collaboration is the core of multi-agent decision-making, and through collaborative work, all parties can integrate resources and experiences to solve problems together. Collaborating parties need to build trust because it establishes the basis for cooperation and interaction among stakeholders. Lack of trust may lead to problems such as information confidentiality, risk-sharing, and resource-sharing, while building trust can promote more effective decision-making cooperation [66]. Rewards and incentives can guide the active participation of all parties and ensure that their interests are aligned with the project agentives. Performance assessments are used to measure and monitor the contributions and performance of all parties to provide feedback and opportunities for improvement in the decision-making process. In practice, these elements are often used in combination to facilitate collaboration and knowledge-sharing, build trust, and motivate and evaluate performance to ensure the successful implementation of multi-actor decision-making [55].

4.4.1. Collaboration

Collaboration is the process by which stakeholders analyzing a problem from different perspectives can constructively discuss their differences and propose solutions beyond their limited knowledge [67]. Mutual trust, full commitment of all parties, openness between all parties, and win-win attitudes are adopted in cooperative contracting in the Singapore construction industry [68]. The dynamics of cooperation is a complex process that is usually influenced by the other party’s intention to cooperate. A quantitative framework is proposed using game theory to analyze the dynamic interactions of collaboration in construction projects, to promote collaboration by reducing tacit knowledge, to provide better benefit returns to stakeholders, and to deepen the understanding of collaborative economics [69]. Sun, CS et al. (2021) explored the internal logic of BIM collaborative applications by analyzing the dynamic behavior of owners and general contractors and considered the impact of different strategies on multi-stakeholder decision-making through numerical simulation, which helps researchers to think about the collaboration dynamics among stakeholders in a project, and also helps participants to choose the right strategies for improved collaboration [68].

4.4.2. Collaborative Innovation

Collaborative innovation among engineering and construction participants includes two aspects [17]. Participants need to collaborate with the innovation activities of other partners, and participants must also contribute to the innovation activities for which they are responsible.

For mega-engineering projects, co-innovation among different stakeholders is critical for successful implementation. Liu, N. et al. (2022) explored the roles of partners’ non-mediated power (expert power and referent power) [31], shared mental models, and team innovation efficacy in improving co-innovation performance based on organizational identity theory and social cognitive theory. To explore the influencing factors affecting participants’ willingness to participate in collaborative innovation, analyzing the collaborative innovation behaviors of participants in megaprojects by constructing a reward and punishment incentive mechanism, combining with the evolutionary game theory and prospect theory, and established a game model between different participants to facilitate the participants’ choice of an evolutionarily stable strategy for participating in synergy. Therefore, factors such as participation cost, synergy coefficient, synergy benefit distribution ratio, and risk preference will be in different degrees [69]. To combine technological innovation with engineering construction and make innovation serve megaprojects, unit participation in construction synergistic innovation is considered to be an effective solution [70].

On the one hand, due to the uncertainty of the construction site environment, the demand for innovation has a dynamic character. Participants have a better understanding of the actual situation at the construction site, so their participation can improve the innovation efficiency. On the other hand, as a general user of innovation results, the collaborative innovation cooperation of the participants is not only conducive to the application and promotion of innovation results and improves the implementation efficiency of mega projects [32]. In the development of the green building materials industry, integrated building materials supply chain enterprise collaborative innovation has become an important way of green collaborative innovation. Promoting the technological innovation of the green building materials supply chain is a key strategy to realize the development of the green building materials industry. The green innovation partner selection method is a reasonable and effective dynamic selection decision-making method, which can be used to improve the co-innovation capability of the integrated building materials supply chain [71].

4.4.3. Trust

With the introduction of relevant policies, an environment for sharing credit information in the construction industry has been created. As a result, stakeholders can use this credit information to select suitable partners. Stakeholders’ credit is disseminated through the Internet.

Zhang, N. et al. (2022) introduced and validated a credit network recommendation model based on a collaborative filtering algorithm to improve the efficiency of selecting partners [69]. The contractor credit is used for owners to select efficient contractors, the owner neighbor set is used to calculate the comprehensive trust of the decision-making owner in the previous owner, and the time decay function is used to correct the differences in the trust relationship between the owner and the contractor over time [72]. Based on the differentiated needs of projects with different values during the implementation, each participant should focus on factors related to collaborative attitudes [35].

Trust can be an effective mechanism for more informed decision-making to accommodate the ever-changing nature of this multi-organizational dynamic collaboration in the construction industry. The trust-based Collaborative Value Estimation (CoVE) methodology was proposed to support and sustain the various disciplines of construction project cooperation and improve collaborative performance. Trust was identifiedas an information asymmetry risk minimization strategy based on principal-agent theory through a survey of project managers and an analysis of multi-attribute utility theory to enhance communication between project participants and provide an ethical basis for multi-agent decision-making.

4.4.4. Incentives

Incentives include the payment of bonuses or incentives in exchange for employees’ performance. Incentives are an effective technique to improve collaboration and develop confidence among project participants in the near term during a multi-actor collaborative process. The degree of teamwork in the construction industry can improve team performance and thus increase the likelihood of successfully delivering a project. Analyzing the data through a questionnaire survey identified these dimensions of team effectiveness as providing a rigorous basis for the development of useful team-building strategies. These strategies aim to integrate a collaborative environment between project stakeholders to improve project performance. Commonalities can be found through questionnaires. Alleman et al. (2020) found that substantial allowances are a best practice tool to improve design efficiency, incentivize bidders for the effort they put into preparing technical proposals [71], and encourage more bidders, therefore increasing competition, as well as reducing risk. At the same time, allowances promote fair procurement, design, and other processes that build trust with the various parties involved, therefore increasing collaborative and innovative project decision-making [73]. According to Pryke, performance incentives can promote productivity and project performance, which is another area to explore in terms of long-term collaborative relationships among construction stakeholders. The success of long-term relationships can be assessed using SNA to explore whether strong relationships impact the quality of decision-making [55]. In the actual management process, He et al. (2022) suggest the cross-district collaborative management of construction wastes [45]. Clear management decisions and the establishment of a sound and dynamic reward and punishment mechanism can provide implications for the management practice of multi-stakeholder decision-making [60]. At the same time, as the number of new construction projects decreases and construction and demolition waste threatens the environment, regulators face many challenges in collaborative management and decision-making. An information-based solution that integrates multiple technologies to monitor violations of law in real time during the waste disposal process collects accurate data to assess the performance of stakeholders and strengthens collaboration between regulators [2].

4.4.5. Performance Evaluation

When unanticipated problems arise in complicated and unpredictable scenarios during the construction process, it is critical to address problems and make dynamic team decisions. Therefore, the analysis of standardized team performance assessment indicators is important for distributed teamwork and decision-making by stakeholders. To verify the role of non-mediated rights in collaborative innovation of different stakeholders, based on organizational identity theory and social cognitive theory. Deriving the collaborative mechanism of multi-subjects through the shared psychological model, exploring the availability of which has a positive impact on the performance of collaborative innovation, and providing suggestions for the wise decision-making of project collaborators to promote the participants’ collaborative innovation, and provides suggestions for project collaborators to make wise decisions to promote participants’ collaborative innovation. The project delivery system (PDS) provides references for project performance measures and improves the accuracy and fairness of performance assessment. In PDS, an indicator system for decision determinants of the project delivery system is established. A PDS decision simulation model was constructed to analyze multi-agent choice preference, which provides a theoretical reference for the scientific decision-making of construction enterprises [3]. Emerging project delivery methods increasingly encourage collaboration between multiple subjects to develop long-term stable relationships between project parties. Rahmani (2021) found that integrated project delivery models that emphasize early engagement are important for working relationships, improving price and scope certainty, and innovation [18]. Participants engage early and share risks and rewards, therefore reconciling the goals of project stakeholders collaboratively, improving AEC project outcomes through Activity-Based Costing methodology [51].

5. Future Research Directions

The systematic development of multi-agent decision-making was suggested to improve the quality, efficiency, and adaptability of multi-agent decision-making and to cope with the challenges of a complex and changing reality [73]. The evolution of a multi-agent decision-making system is promoted through the complementarity of three aspects: knowledge management, decision resilience, and the integration of multi-tools and multi-methods [64].

5.1. Knowledge Management

In the process of multi-agent decision-making, the information asymmetry of each stakeholder requires strengthening knowledge management [69]. Specifically, knowledge management includes constructing a new knowledge-sharing platform, integrating various types of information, and improving the comprehensiveness and accuracy of decision-making. Multi-agent involves knowledge in multiple fields, often scattered in different departments and systems, forming knowledge fragments. Integrating knowledge fragments through the knowledge management platform creates a more complete knowledge map to provide more comprehensive information support for decision-makers [26].

For example, they are utilizing the knowledge map to integrate green technologies, environmental policies, and market trends to help enterprises make sustainable development decisions. Also, a green and low-carbon knowledge map for cities should be built to support urban planning decisions, optimize energy use, and mitigate the impact of climate change [74].

5.2. Resilience

The current engineering and construction field faces many challenges, which puts higher demands on decision-making systems. Emphasizing decision-making resilience means that the system needs to adapt to environmental changes quickly and adjust decision-making strategies in time to maintain effectiveness [75]. Temporally, decision-making systems need to focus more on utilizing real-time data to make timely adjustments in a rapidly changing environment. Spatially, the global perspective of the decision-making system is emphasized to improve the system’s robustness to shocks in different geographies through cross-geographic cooperation and information sharing [2].

5.3. Methodological Development

In the future, multi-agent decision-making will require smarter and more comprehensive problem-solving methods. Multi-tool and multi-method integration will be a key trend [33]. They also integrate approaches faced in practice, such as technical complexity, high costs, and resource requirements, difficulties in updating and iterating the technology, and difficulties in ensuring security and privacy. This includes the introduction of smarter decision support systems that leverage technologies such as AI, machine learning, and deep learning. At the same time, decision-making methods will integrate qualitative and quantitative analyses more comprehensively, reducing the one-sidedness of decisions through comprehensive assessments from multiple perspectives. Simulation models and complex systems are combined to consider the nonlinearity, dynamics, and interconnectivity of the system to ensure that the decision-making model is more in line with the real situation and to provide more realistic decision-making support through the simulation of large-scale data [76]. The information resources necessary for the decision-making process or certain important decision-making factors are dispersed over a more extensive range of activities, a class of organizational decision-making, or distributed decision-making. Make full use of the information system, combined with distributed technology, to make the decision-making system more flexible and responsive and adapt to multi-agent decision-making’s complexity [77].

6. Conclusions

Multi-agent decision-making is essential since it offers a broad perspective on decision-making. It is a technique used in engineering construction, but it also represents an innovation in making decisions. It gives stakeholders greater voice and engagement and encourages the growth of cooperative spirit, integrated thinking, and information exchange [78].

This research aims to present the state of current knowledge of multi-agent decision-making in CEM and to provide future directions for research. To find pertinent literature for debate, the Web of Science was searched using a visual analysis tool. After reviewing 120 papers on the application of multi-agent decision-making in CEM, four groups were formed: (1) Focus areas for multi-agent decision-making (such as safety management, green and low carbon, design management, and dynamic planning). (2) Practical application tools and procedures (e.g., BIM, AI, off-site construction, and virtual reality). (3) Methodologies and theoretical application tools include network viewpoint, simulation, and game theory. (4) Critical points for problem solutions (e.g., teamwork, collaboration, creativity, trust, and incentives). For example, land use planning in cities must consider the interests of a wide range of stakeholders, including developers, citizens, and the government planning department. Their decisions must strike a balance between sustainable development and environmental protection. This research for future directions focuses on (1) Constructing a knowledge map among stakeholders by integrating knowledge and information fragments to provide comprehensive information support for decision-makers. (2) Multi-agent decision-making resilience, improving decision-making resilience in both time and space to cope with shocks caused by various extreme events. (3) Management methods and tools, seeking more intelligent and systematic management tools to integrate management with reality.

This study provides a systematic overview of multi-agent decision-making and valuable references for scholars and industry practitioners. This study helps scholars to have a comprehensive understanding of multi-agent decision-making. This study helps industry practitioners pay attention to multi-agent synergy in CEM, which has resulted in high-quality construction products and promotes the industry’s sustainable development.

Some limitations exist in this research. The data are retrieved from the database of ISI Web of Science; more comprehensive data can be acquired from other databases in the future. Multi-agent decision-making for CEM may have occurred and may not have been captured because the methodology relied upon published literature only. Moreover, only articles published in English are studied in this research.