AI-Driven Digital Twins in Industrialized Offsite Construction: A Systematic Review

Abstract

The increasing adoption of industrialized offsite construction (IOC) offers substantial benefits in efficiency, quality, and sustainability, yet presents persistent challenges related to data fragmentation, real-time monitoring, and coordination. This systematic review investigates the transformative role of artificial intelligence (AI)-enhanced digital twins (DTs) in addressing these challenges within IOC. Employing a hybrid re-view methodology—combining scientometric mapping and qualitative content analysis—52 relevant studies were analyzed to identify technological trends, implementation barriers, and emerging research themes. The findings reveal that AI-driven DTs enable dynamic scheduling, predictive maintenance, real-time quality control, and sustainable lifecycle management across all IOC phases. Seven thematic application clusters are identified, including logistics optimization, safety management, and data interoperability, supported by a layered architectural framework and key enabling technologies. This study contributes to the literature by providing an early synthesis that integrates technical, organizational, and strategic dimensions of AI-driven DT implementation in IOC context. It distinguishes DT applications in IOC from those in onsite construction and expands AI’s role beyond conventional data analytics toward agentive, autonomous decision-making. The proposed future research agenda offers strategic directions such as the development of DT maturity models, lifecycle-spanning integration strategies, scalable AI agent systems, and cost-effective DT solutions for small and medium enterprises.

1. Introduction

Industrialized offsite construction (IOC) refers to a systematic and integrated approach to the planning, design, fabrication, and assembly of building components or entire structures in a controlled offsite environment, using advanced manufacturing techniques and often digital tools, with the goal of enhancing construction speed, efficiency, safety, sustainability, and quality [1]. This approach has emerged as a response to limitations of traditional construction methods, which rely heavily on on-site labor and materials—often leading to inefficiencies, errors, and delays—especially with rising project complexity and delivery speed demands [2]. IOC offers several advantages across key performance areas: it enhances efficiency through faster delivery, improved decision-making, and flexible fabrication; supports higher quality and cost predictability while reducing rework; and improves safety, minimizes on-site labor demands, and promotes sustainability by lowering material waste and environmental impact [3]. Despite these benefits, IOC faces implementation challenges, including fragmented data systems, limited real-time monitoring capabilities, inadequate performance analytics, and complex supply chain coordination [4,5]. These challenges, which hinder the full potential of IOC, can be addressed through the integration of advanced digital technologies—such as automation, real-time data analytics, Internet of Things (IoT), and human–machine collaboration [6].

In the context of IOC, digital twins (DTs) powered by artificial intelligence (AI) play a pivotal role by creating accurate virtual replicas of physical assets. A DT is a dynamic, real-time virtual representation of a physical object, system, or process, maintained through a bidirectional linkage that enables continuous data exchange between the physical and digital counterparts [7]. This connection allows the virtual model not only to mirror and simulate the real-world entity using data from sensors, IoT devices, and other sources, but also to inform, influence, or control the physical system for enhanced analysis, prediction, and optimization [8]. AI significantly enhances the capabilities of DTs by enabling advanced data analytics, predictive modeling, and data-driven decision-making [9,10].

Recent market analyses indicate that the DT market in the construction sector was valued at approximately USD 41.98 billion in 2024 and is projected to reach USD 93.5 billion by 2029, growing at a compound annual rate of 17.3% [11]. Real-world applications further underscore this momentum. For instance, the Hong Kong–Zhuhai–Macau Bridge employs a comprehensive structural health monitoring system, featuring hundreds of embedded sensors across bridges and tunnels [12]. This system enables continuous real-time monitoring and predictive maintenance through data-driven anomaly detection, thereby enhancing operational safety and efficiency. Additionally, within the context of IOC, DTs facilitate efficient assembly and quality control by improving design precision [13], predicting potential issues [14], and optimizing supply chains [15]. As a key enabler of IOC, DTs integrate data-driven decision-making to reduce waste [16], accelerate timelines [17], promote sustainable construction practices [18], and improve project outcomes [6,19].

Despite these advantages, the implementation of DTs in IOC is accompanied by several critical challenges. One of the most pressing issues is the lack of standardization and compatibility across DT systems, which obstructs seamless data integration, model interoperability, and system scalability [20]. Additionally, limited efforts to assess the maturity levels of DT applications hinder the ability to benchmark progress and define structured development pathways [21]. There is also a lack of clarity regarding the core technological components, enabling frameworks, and integration mechanisms necessary for effective DT deployment in construction, often resulting in fragmented and inconsistent implementations [22]. In addition, given the critical role DTs play in addressing key IOC challenges, their increasing adoption demands focused scholarly investigation to explore their full potential, assess real-world applications, identify practical barriers, and bridge knowledge gaps. Motivated by these needs, this study investigates the application of AI-driven DTs in IOC. A systematic literature review is conducted to identify major trends, challenges, and opportunities in this underexplored domain, guided by the following research questions:

RQ1. 

(Descriptive) What is the current state of DT technology in IOC, including its applications, benefits, challenges, adoption barriers, and research gaps?

RQ2. 

(Technical) What are the key components and enabling technologies supporting the implementation of DT in IOC?

RQ3. 

(Forward-looking) How does AI enhance DTs, and what are the opportunities for advancing AI-driven DT in IOC?

To address these questions, the quantitative aspects of RQ1 are examined through a scientometric analysis, while the content-related elements are explored through systematic content synthesis and review. RQ2 is primarily addressed within the content synthesis and review by analyzing enabling technologies and implementation frameworks. Although RQ3 is briefly discussed throughout the review, it is explored more comprehensively in the discussion section, where the forward-looking implications of AI integration are critically analyzed.

2. Research Background

2.1. Industrialized Offsite Construction

IOC is reshaping the construction industry by relocating production to controlled factory environments for greater precision and efficiency [19]. IOC benefits span multiple key areas: Efficiency: Accelerated timelines and streamlined workflows [23]; Quality and Safety: Improved quality control and reduced on-site risks [24]; Sustainability: Lower material waste, emissions, and environmental impact [23]; Labor Optimization: Reduced dependence on on-site skilled labor [24]; Cost Predictability: More accurate budgeting and fewer delays due to rework [14]; and Design Flexibility: Modular components tailored to varied project needs [25]. Collectively, these advantages position IOC as a transformative solution. The evolution of IOC reflects a transition from basic prefabrication to digitally integrated manufacturing focused on precision, standardization, and sustainability [19]. IOC spans multiple sectors: in residential construction, it supports affordability and rapid deployment; in commercial projects, it delivers modular, adaptable solutions with shorter timelines; in healthcare and institutional settings, it ensures compliance and reduces disruptions; and in infrastructure, it expedites delivery of complex assets such as bridges, stations, and transit systems [26]. IOC component typologies span a spectrum of prefabrication levels, ranging from discrete elements, such as wall panels, beams, and columns, which are manufactured offsite and assembled onsite, to more integrated forms like non-volumetric assemblies, including floor cassettes and roof trusses that bundle multiple elements into single installable units. Further along the spectrum are volumetric modules, such as prefabricated bathroom pods or mechanical rooms, which are fully enclosed, pre-fitted units delivered to the site ready for installation. At the most comprehensive level are fully modular buildings, composed of multiple volumetric units combined to form entire structures, offering maximum efficiency and standardization, especially in repetitive-use facilities [27]. Material selection in IOC is performance-driven and tailored to project demands: steel is preferred for its structural strength, flexibility, and suitability for long spans; concrete is chosen for its durability, fire resistance, and acoustic properties; timber is valued for its renewability, low carbon footprint, and ease of handling; while hybrid systems combine two or more materials to optimize structural performance, environmental impact, and construction efficiency [26].

The lifecycle of IOC projects comprises a series of interconnected phases—design/planning, manufacturing, transportation, on-site assembly, operation/maintenance, and decommissioning—each contributing to improved efficiency, precision, and sustainability [21,23]. During design and planning, digital modeling and simulation tools are used to optimize specifications and workflows, resolving conflicts early; the manufacturing phase involves automated production and strict quality control for consistency; transportation relies on logistical coordination and route optimization for timely and secure delivery; on-site assembly benefits from prefabrication alignment strategies and integrated system hookups, expediting construction while minimizing disruptions; operation and maintenance leverage sensor-based monitoring and predictive analytics to extend asset lifespan and inform continuous improvement; finally, decommissioning uses standardized connections and documented disassembly to facilitate recovery, recycling, and adaptive reuse within the circular economy [26]. Unlike the fragmented and sequential nature of conventional construction, the integrated IOC model promotes collaboration and cohesive execution [28]. By shifting construction to controlled settings, IOC reduces waste, energy use, emissions, and environmental impact, and it helps address global challenges such as climate change, urbanization, and resource scarcity [26,29].

Given the aforementioned unique characteristics of IOC—such as its high degree of prefabrication, controlled production environments, modular assembly processes, and lifecycle-oriented workflows—DT emerges as a promising enabler to further enhance efficiency, precision, and sustainability [2]. For instance, DT-enabled manufacturing lines can synchronize real-time sensor data with production schedules to optimize throughput, integrated logistics DTs can predict transportation delays and adjust delivery sequencing, and asset-level DTs can provide predictive maintenance alerts for modular building systems [9,10]. However, successful implementation must align with the specific requirements and operational dynamics of IOC, as generic DT architectures may overlook the distinct typologies, material systems, and lifecycle phases that define this construction model. Adapting DT to discrete-element IOC may prioritize fine-grained component tracking, on-site integration with scanning technologies, and rapid assembly verification, while non-volumetric assemblies could benefit from DT-driven coordination between manufacturing tolerances and on-site fitting processes [27]. Volumetric modules may require DT models that integrate manufacturing, logistics, and installation sequencing into a single synchronized dataset, whereas fully modular buildings demand holistic, lifecycle-spanning DT systems capable of managing performance monitoring, energy optimization, and end-of-life circularity strategies. Material-specific adaptations could also be necessary, such as embedding IoT-enabled curing sensors in concrete elements, RFID-based traceability in timber modules, or structural health monitoring in steel assemblies [26]. Collectively, these considerations give rise to multiple hypotheses regarding the optimal adaptation of DT to IOC’s diverse characteristics, which warrant systematic investigation and refinement in the present study.

2.2. AI-Driven Digital Twin

The concept of DT has evolved considerably, generating ongoing debate regarding its definition and scope across various industries [22]. Originally developed within manufacturing and aerospace [30], DT has increasingly been adopted in architecture, engineering, and construction (AEC); however, its definition remains inconsistent and fragmented [5]. Some scholars frame DT primarily as a virtual modeling tool for lifecycle asset management [31], while others emphasize its capacity for real-time simulation and synchronization of physical processes [32,33]. The Center for Digital Built Britain, for instance, defines DT as a digital representation of physical assets, processes, or systems in the built environment, continuously updated with real-time data to maintain alignment with their physical counterparts. Complementing this, the United States Digital Twin Consortium describes a digital twin as a virtual representation of real-world entities and processes, synchronized at a specified frequency and fidelity [34]. This definition underscores DT’s potential as a dynamic, data-driven platform that enhances monitoring, predictive analysis, and decision-making throughout the construction lifecycle.

The lack of a consistent definition of DT has contributed to conceptual ambiguity, particularly in distinguishing it from related constructs such as building information modeling (BIM) and cyber–physical systems (CPS), which exhibit overlapping characteristics yet serve fundamentally different purposes [35]. BIM primarily facilitates design coordination, visualization, and documentation through a static or periodically updated digital representation of built assets, typically without real-time connectivity to their physical counterparts [36]. CPS focuses on the integration of computational algorithms with physical processes to enable real-time monitoring and control; however, it does not necessitate a comprehensive virtual model or a sustained digital–physical pairing [37]. In contrast, DT is defined by its persistent, bidirectional linkage between a physical asset and its digital replica, enabling continuous synchronization, simulation, and optimization. While BIM may serve as a foundational dataset and CPS provides the underlying control infrastructure, DT is uniquely characterized by its dynamic coupling of physical and digital systems, supporting data-driven decision-making across the asset lifecycle [35]. To elucidate this distinction, consider the scenario of a smart building. BIM offers a detailed digital representation used primarily during the design and construction phases, but this model typically remains static post-construction. CPS, such as a building automation system, enables real-time control of subsystems like HVAC and lighting through sensor feedback. A DT, however, synthesizes the static BIM model with real-time data streams from the CPS and additional sources—such as occupancy patterns and energy usage—to create a continuously updated, semantically rich virtual counterpart. This allows the simulation of operational scenarios (e.g., emergency response strategies or energy optimization), the prediction of equipment failures, and autonomous decision-making, thereby enhancing overall building performance and lifecycle management.

DT maturity models provide a structured framework to understand the evolving capabilities of DT systems as they progress through different levels of sophistication [38]. At Level 1 (Basic), DTs primarily function as static or periodically updated digital replicas of physical assets. This stage offers a foundational digital representation but lacks real-time interaction or dynamic responsiveness, serving mainly for visualization and record-keeping. Advancing to Level 2 (Intermediate), DTs integrate real-time data streams from sensors and IoT devices, allowing for continuous updates and enabling basic monitoring of the physical asset’s state. Unlike Level 1, this stage supports timely awareness of asset conditions but remains limited to descriptive insights without predictive or analytical functions. At Level 3 (Advanced), DTs incorporate simulation capabilities and predictive analytics, transitioning from purely descriptive models to proactive decision support tools. This level differentiates itself by enabling forecasting of potential future states, identifying trends or anomalies before they occur, which significantly enhances operational planning and risk management. Level 4 (Sophisticated) further expands DT functionality by supporting complex scenario modeling and in-depth diagnostics. This level allows users to test “what-if” scenarios and perform root-cause analysis to diagnose underlying issues. Unlike Level 3, which primarily focuses on prediction, Level 4 emphasizes optimization and advanced problem-solving by combining multiple data sources and analytical methods. Finally, at Level 5 (Autonomous), DTs achieve full self-governance through the integration of artificial intelligence and machine learning algorithms. This stage enables the DT system to make autonomous decisions, adapt to changing conditions in real-time, and self-optimize without human intervention. Unlike all previous levels, Level 5 DTs operate independently, continuously learning from new data to improve performance and respond dynamically to unforeseen events.

Achieving full DT maturity requires integration of multiple layers: (1) the physical layer, representing the actual asset; (2) the communication layer, facilitating real-time data exchange; (3) the sensing layer, responsible for data acquisition via sensors and IoT devices; (4) the storage layer, managing data archiving and retrieval; (5) the digital layer, maintaining the digital model and simulations; and (6) the service layer, delivering analytics, optimization, and decision support functions [35]. These layers collectively enable DTs to function as dynamic, real-time systems that bridge physical and digital environments. Figure 1 illustrates the structure and conceptualization of a high-maturity DT system tailored for construction applications.

AI is integral to the development and operation of DT systems, providing the computational intelligence necessary to support advanced functionalities across multiple layers [9]. AI enables predictive analytics, anomaly detection, and autonomous decision-making—capabilities essential for optimizing performance and synchronization [39]. While AI is most associated with the service layer—where it facilitates data analysis and informed decision-making—its influence extends to additional layers of the DT architecture [39]. In the sensing layer, AI improves data acquisition and processing through intelligent sensor networks and IoT integration [40]. In the virtual layer, it enhances dynamic modeling and simulation accuracy [40]. Within the communication layer, AI contributes to efficient data transmission, integration, and system interoperability [22]. By embedding AI throughout the DT architecture, these systems achieve greater accuracy, responsiveness, and adaptability for mature, self-optimizing DTs [9].

2.3. AI-Driven Digital Twin in Industrialized Offsite Construction

The integration of DT in IOC facilitates the convergence of physical and digital construction by offering a dynamic, real-time representation of assets and workflows. DTs support proactive decision-making, enhance design-to-delivery alignment, and drive process innovation—capabilities essential for greater adaptability and resilience [41]. However, despite their transformative potential, existing research on DT applications in IOC remains limited. Our preliminary review of AI-driven DT research in construction reveals three groups of studies. The first and most extensive group advances foundational insights into DT applications for conventional construction but engages only tangentially with IOC. For instance, Tuhaise et al. [22] identified key enablers of DT, such as data integration and interoperability; Boje et al. [42] introduced a semantic construction DT to overcome limitations in BIM-driven data exchange, referencing prefabrication as a component of value chain efficiency within a wider exploration of cyber–physical systems; AlBalkhy et al. [7] and Saif et al. [21] developed taxonomies of DT applications in the built environment, covering areas such as supply chain logistics and quality control; Yeom et al. [43] and Arsecularatne et al. [44] focused on DT-enabled operation phase energy optimization and carbon reduction; and Figueiredo et al. [45] and Yevu et al. [5] integrated DT with blockchain and IoT to support sustainability assessments and smart prefabrication supply chains. The second group examines the role of digital innovations in IOC, without an explicit focus on DTs. For instance, Xu et al. [16] reviewed automation in IOC, identifying technologies that enhance productivity, safety, and cost efficiency; Cheng et al. [3] reviewed digital tools in offsite and prefabricated construction, highlighting challenges in interoperability and standardization; and O’Connell et al. [46] analyzed the impact of emerging technologies, e.g., IoT, AI, and extended reality, emphasizing their potential to advance design quality, operational efficiency, and sustainability [47]. The third group, focusing on the application of DT in IOC, remains notably underdeveloped. Xie and Pan [48] investigated the role of DT in modular construction, emphasizing its potential to enhance logistics coordination, real-time monitoring, and automation, and identified key barriers to implementation, including sensor integration challenges, data security risks, and regulatory limitations.

While previous research has examined DTs in general construction contexts or addressed prefabrication tangentially, there appears to be no comprehensive DT framework that spans the full spectrum of IOC workflows—from design and planning to manufacturing and on-site assembly. Although technologies such as IoT and AI have been explored independently within offsite construction, their integration within DT systems remains largely unexamined. The transformative potential of AI-driven DTs in IOC is virtually unexplored, even though AI offers substantial promise in addressing core industry challenges through predictive analytics, computer vision-enabled quality control, and intelligent supply chain optimization [9]. This study presents a novel contribution by systematically exploring AI-centric DT frameworks tailored for IOC—as opposed to a supplementary tool—enabling real-time decision-making across modular workflows. It addresses key industry challenges, including the absence of integrated digital–physical quality control systems, poor synchronization between factory production and on-site assembly, and the lack of adaptive systems for managing design variability during manufacturing—barriers that conventional construction methods have yet to overcome [47].

3. Materials and Methods

We adopt a hybrid methodology that integrates systematic scientometric analysis with qualitative content review: Step (1) involves a structured search and screening protocol to ensure comprehensive and unbiased literature retrieval; in Step (2), the scientometric analysis quantitatively maps the existing body of literature, identifying influential publications, key thematic trends, patterns, gaps, and the temporal evolution of the field [6]; complementing this, in Step (3), the content review provides an in-depth qualitative examination of selected studies, offering critical insights into underlying theoretical frameworks, approaches, and contextual dimensions that quantitative methods alone cannot capture [49]. The overall process, illustrated in Figure 2, combines the breadth and objectivity of scientometric mapping with the depth and interpretive richness of content analysis.

3.1. Systematic Extraction and Evaluation of Articles

The review process involves (1) database selection, (2) scope and keyword definition, (3) development of a review protocol including inclusion and exclusion criteria, and (4) screening and evaluation of retrieved studies. To achieve broad and representative coverage, we selected three major academic databases—Scopus and Web of Science for their rigorous indexing standards and Google Scholar for a broader reach to gray literature and emerging studies [50]. To develop the keyword set, we employed an iterative refinement process, systematically testing and adjusting keywords based on preliminary search results to maximize coverage of relevant research [21]. The finalized set, presented in Box 1, spans three sets of terms: (1) construction industry terms (e.g., “construction”, “AEC industry”, “built environment”), (2) IOC terms (e.g., “offsite”, “prefabrication”, “modular”), and (3) DT terms (e.g., “digital twin”, “virtual twin”, “digital replica”). While terms like “cyber-physical system (CPS)” and “IoT-enabled BIM” are not technically equivalent to DTs, they were included to capture studies where such terminology may have been used imprecisely or interchangeably.

Box 1. Database Search String.

TITLE-ABS-KEY((“construction” OR “building” OR “AEC industry” OR “built environment”) AND (“offsite” OR “off-site” OR “prefabrication” OR “modular” OR “precast” OR “industrialized” OR “penalized”) AND ((“digital twin” OR “virtual twin” OR “digital replica” OR “cyber-physical system” OR “CPS” OR “cyber twin” OR “digital modeling” OR “digital mirroring”) OR ((“BIM” OR “Building Information Modeling” OR “Building Information Modelling”) AND (“Sensors” OR “IoT” OR “Internet of Things” OR “Real-time Data” OR “Dynamic Data” OR “Live Data” OR “Predictive Analytics” OR “Remote Sensing” OR “Sensing” OR “Monitoring Systems” OR “Real-time Monitoring”))))

Eligible sources for inclusion were published, peer-reviewed journal articles and conference papers written in English, specific to the AEC industry, without any date restrictions applied [51]. The review focused specifically on DT applications in IOC, including modular, prefabricated, and panelized construction methods. Excluded materials were non-peer-reviewed documents—such as book chapters, white papers, technical reports, theses, and dissertations—as well as studies outside AEC or unrelated to DT in IOC.

The screening and selection of studies followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol, which promotes transparency, reproducibility, and methodological rigor [52]. As presented in Figure 3, the process comprised identification, title and abstract screening, and full-text screening. During the identification stage, 505 publications were retrieved from the selected databases—Scopus (310), Web of Science (131), Google Scholar (64). The title and abstract screening was then conducted on 363 records, resulting in the removal of 368 studies that did not meet the inclusion criteria. Subsequently, 75 full-text articles were assessed for eligibility, of which 34 studies were excluded due to misalignment with the study scope or insufficient methodological quality. The final dataset included 41 studies, supplemented by 11 additional records identified through snowballing (i.e., backward and forward citation tracking).

3.2. Scientometric Analysis

The scientometric analysis addressed the quantitative dimension of RQ1 by mapping the evolution, trends, and intellectual structure of DT research in the context of IOC [53]. The selected dimensions of analysis included (1) publication years and sources; (2) geographic and author-level contributions; (3) methodological characteristics; and (4) keyword co-occurrence and frequency analysis. A suite of tools was used to support the analysis, including VOSviewer Online (v.1.2.4) for bibliometric network visualization, Bibliometrix (Stable Version) for thematic and citation mapping, Python’s NetworkX (v.3.5) for custom analysis, and ArcGIS Pro (v.3.5) for mapping [54,55]. We completed missing data where possible, excluded irreparable entries, formatted the dataset into tool-specific structures (e.g., adjacency matrices, citation networks), and ensured data quality through cross-verification, duplicate removal, and consistency checks [56].

3.3. Systematic Content Analysis

The systematic content analysis, the final stage of the methodology, provided a qualitative understanding of the literature [6] through: (1) content review, (2) protocol-based data extraction, (3) data synthesis and multi-step classification, and (4) interpretation of findings and future research directions [6]. Based on content review, the selected articles were examined to identify key themes, research methods, and knowledge gaps. Building on this review, a data extraction protocol was developed, consisting of structured questions designed to capture essential dimensions of each study [21], e.g., context, DT components, methodologies and validation techniques, and specific applications within IOC. Then, the extracted data underwent synthesis and multi-step classification through open coding, axial coding, and selective coding [21]. Open coding identified initial concepts directly from the data without preconceptions. Axial coding then organized these concepts into coherent thematic categories. Finally, selective coding consolidated these themes into overarching application domains. This iterative classification process enabled the identification of dominant patterns, emerging trends, and research gaps, while ensuring analytical transparency and depth.

4. Results

4.1. Scientometric Analysis Results

According to Figure 4, there was a modest emergence of scholarly interest in DTs in IOC in 2017, followed by a formative period through 2020 with only 1–2 publications per year. This limited output can be attributed to the early developmental stage of DTs, often explored under adjacent terms such as IoT-enabled BIM [57]. While DTs were already being examined in other areas of the construction industry during this time, their application to IOC was only beginning to emerge. From 2021 onward, research activity increased markedly, reaching a peak of 16 publications in 2023. This growth reflects the transition from foundational investigations toward more applied studies, supported by clearer conceptual frameworks established during the earlier phase. Publications from 2021 began to target well-defined use cases—such as quality assessment and supply chain coordination [15,58]—demonstrating growing maturity in the field. Journal articles represent the predominant type, likely due to the complex and interdisciplinary nature of DTs and IOC, which often require the depth and rigor afforded by long-form journal formats. Conference proceedings also gained visibility in 2023 (6 publications), suggesting growing interest among research communities. A slight decline was observed in 2024 (12 publications), which may reflect publication lag or saturation following the 2023 peak. The 2025 count (1) includes only outputs indexed through February and is thus expected to rise as the year progresses.

The bibliometric analysis identified 29 sources—18 journals and 11 conference proceedings—that have shaped research on DTs in IOC. As shown in Figure 5, and applying a minimum threshold of two publications, Automation in Construction emerges as the most influential journal, contributing 11 articles. Its prominence reflects the journal’s established focus on digital transformation and automation in the construction domain [6]. Other notable journals, such as Advanced Engineering Informatics, Computers in Industry, and Journal of Building Engineering, highlight the interdisciplinary scope of the field, spanning construction informatics, project management, and digital innovation. The inclusion of Sensors points to the growing integration of IoT and real-time monitoring technologies—core components in dynamic DT systems.

Conference proceedings serve as key outlets for disseminating emerging research. Noteworthy venues include the 2024 IEEE 20th International Conference on Automation Science and Engineering and the International Symposium on Automation and Robotics in Construction, both of which emphasize advancements in robotics, automation, and smart construction systems. Collectively, the diversity of these sources underscores the multidisciplinary nature of DT research in IOC.

The geographical distribution of publications, as illustrated in Figure 6, reveals Hong Kong and China as the most prolific contributors to DT research in IOC. Notably, both regions played pivotal roles during the early conceptualization of DTs in this context, contributing foundational frameworks and initial adoption models. While this early involvement has driven scholarly momentum, it may have also introduced a geographical favor into the research landscape. Many of the developed models and case applications appear tailored to the construction ecosystems of these two regions, potentially limiting their generalizability across diverse global contexts. As such, broader international engagement is essential to ensure that DT implementations in IOC are adaptable to varying regulatory, technological, and industrial environments. When analyzed in terms of citation impact—measured by average citations per publication—a different pattern emerges. New Zealand, despite producing only two papers, leads with a remarkable average of 168 citations per publication, followed by Hong Kong and China. This elevated impact can be attributed to New Zealand’s early and substantive contributions, particularly those offering pioneering implementation models and case studies. These foundational works have likely served as critical reference points, either as frameworks for refinement or as subjects of critique, thereby amplifying their academic influence.

As shown in Figure 7, the United States is the most interconnected node in the international collaboration network, engaging with seven countries—including the United Kingdom, Canada, and Australia—highlighting its role as a central research intermediary. The strongest bilateral link is between China and Hong Kong, reflecting close regional collaboration. While Australia, Canada, and the UK are also prominent, their international ties are less extensive. Additionally, the absence of high-performing countries—such as New Zealand—from major collaboration clusters suggests their contributions, while academically substantive, may be more specialized or independent.

Table 1. The most influential contributors and publications in digital twin research for industrialized offsite construction.

Author	NP	TC	H-Index	FC	Publication	Y	TC	NTC
Huang George Q.	9	757	6	2017	Zhong, R.Y., 2017 [32]	2017	336	37.33
Zhong Ray Y.	7	622	5	2017	Li, C.Z., 2018 [57]	2018	331	41.38
Xue Fan	5	947	5	2017	Li, X., 2022 [59]	2022	146	36.50
Jiang Yishuo	5	284	4	2021	Zhai, Y., 2019 [60]	2019	137	19.57
Shen Geoffrey Q.P.	4	874	4	2017	Lee, D., 2021 [15]	2021	131	26.20
Lu Weisheng	4	712	4	2017	Jiang, Y., 2022 [61]	2022	102	25.50
Li Ming	4	211	3	2022	Wu, L., 2022 [62]	2022	93	23.25
Li Xiao	3	570	3	2018	Jiang, Y., 2021 [63]	2021	74	14.80
Zhao Rui	3	280	3	2022	Zhou, J.X., 2021 [64]	2021	70	14.00
Liu Xinlai	3	181	3	2021	Tran, H., 2021 [65]	2021	59	11.80

summarizes the top contributing authors in the field, detailing each author’s number of publications (NP), total citations (TC), domain-specific h-index, year of first contribution (FC), most cited publication, its publication year (Y), total citations for that publication (TC), and normalized citations per year (NTC) to account for differences in publication timing. The author contribution analysis identified 171 researchers actively publishing on DTs in IOC. The analysis shows Huang George Q. as the most prolific author, followed by Zhong Ray Y. and Xue Fan. It is clear that these pioneer authors often collaborate closely, frequently within the same research labs. Their most impactful publications bridge foundational studies with recent highly cited works, highlighting the crucial role of early adoption models and frameworks. These pioneering contributions have fundamentally shaped the field’s direction, establishing conceptual foundations that continue to guide and inspire subsequent research and practical applications.

The research landscape is largely shaped by framework development and case studies. This dual focus directly reflects IOC’s critical need for a conceptual backbone to integrate complex DT functionalities (e.g., BIM, IoT, logistics) across its distinct manufacturing, transportation, and assembly phases [26]. Frameworks define these integration processes, while case studies provide the essential empirical testbed, applying and validating these models directly in IOC factory or project environments using real operational data. Bridging this gap, proof-of-concept implementations explore technical feasibility. Often conducted in simulations or lab settings that mirror IOC’s controlled production lines, these studies de-risk larger DT deployments by verifying specific components or strategies [17]. Beyond these applied methods, literature reviews and insight papers offer vital theoretical contributions, synthesizing existing knowledge and proposing new conceptual directions. Crucially for IOC, ontology development studies directly tackle the pervasive data interoperability challenge across its multi-stakeholder supply chains by creating structured semantic models [49]. Though less common, qualitative interview-based research provides essential human-centric insights from IOC practitioners on DT adoption and usability. This methodological mix highlights the field’s balance between foundational theory and the practical implementation challenges unique to IOC.

The sophistication of these methodologies is further reflected in their diverse, often complementary validation approaches; studies used an average of 1.8 methods. This multi-layered validation directly addresses the multifaceted nature of DT implementation in IOC, where technical performance, practical utility, and industry acceptance are all paramount [66]. Case study validation dominates, reflecting the critical need to test DT systems in active IOC projects for their real-world impact and scalability within prefabricated workflows. Simulation-based validation offers a crucial alternative for optimizing complex IOC processes (e.g., factory assembly lines, module logistics) where physical testing is prohibitive. Benchmarking compares DT system performance against IOC industry standards, gauging competitive advantage. Moreover, proof-of-concept validation and analytical methods collectively ensure the technical integrity of IOC-specific DT components, such as data models for modular elements. Crucially, industry feedback validation bridges the research–practice gap by involving IOC practitioners and ensuring that solutions are operationally viable. Finally, theoretical validation confirms the conceptual rigor of new DT frameworks for IOC before extensive empirical testing. Figure 8 summarizes the validation strategies and research methods.

Figure 9 displays the resulting keyword co-occurrence network: node size indicates keyword frequency, while link thickness denotes the strength of co-occurrence. Prominent keywords such as “digital twin” (central node), “building information modeling”, and “modular construction” exhibit the highest frequency and strongest interconnections, underscoring their foundational role in the field. Notably, “digital twin” shows strong linkages with “Internet of Things” and “cyber-physical systems” [67]. The network resolves into four thematic clusters: The red cluster (Digital Twin Core) represents the central research axis, emphasizing practical implementation frameworks that integrate DTs with modular integrated construction, on-site assembly, and smart construction. The green cluster (Data & Connectivity) underpins the infrastructure supporting DTs, encompassing BIM, IoT, and blockchain technologies, which enable secure data exchange and interoperability. The blue cluster (Industrial Automation) reflects the industry’s shift toward smart manufacturing, encompassing automation, digital industrialization, and offsite construction models. The yellow cluster (Advanced Simulation) focuses on predictive capabilities enabled by cyber–physical systems, advanced simulation techniques, and material-specific innovations such as timber prefabrication. These approaches support enhanced decision-making, system validation, and performance optimization across IOC.

Temporal keyword analysis reveals a clear progression, as depicted in Figure 10. The color gradient from blue to yellow visualizes this trajectory, demonstrating how research priorities have shifted over time. In the initial phase (2020–2021, blue to green tones), studies often focused on foundational enabling technologies such as BIM and IoT, which frequently co-occurred in the literature. These efforts reflect early attempts to digitally integrate construction workflows. Keywords like “prefabricated construction” and “on-site assembly services” appeared prominently, indicating interest in leveraging real-time data and prefabrication to enhance efficiency. Although the concept of DT had not yet fully emerged, its constituent technologies were actively explored. By 2021–2022 (green tones), DT gained visibility as a central research theme. Keywords such as “modular construction”, “blockchain”, and “cyber-physical systems” signaled increased attention to automation, data interoperability, and system integration. The rising presence of terms like “simulation” and “Industry 4.0” highlighted a move toward smart manufacturing. In the most recent phase (2022–2023, yellow tones), the literature reflects more advanced DT applications. Emerging terms such as “knowledge graph”, “software-defined manufacturing”, and “timber prefabrication” point to growing interest in AI-driven decision support and sustainable construction. The increased emphasis on “smart construction” and “automation” suggests a continued trajectory toward intelligent, adaptive, and integrated construction environments driven by DT frameworks. This temporal progression reflects the field’s shift from foundational digitization—characterized by the adoption of BIM and IoT—to the operational integration of DTs and modular construction methods. It culminates in advanced applications such as AI-driven automation and sustainable manufacturing.

4.2. Themes and Application Areas

This analysis identified 14 distinct themes related to the application of DT in IOC, which were further consolidated into seven overarching thematic clusters as depicted in Figure 11.

Theme 1.1: Assessing Digital Twin in IOC: As anticipated, studies assessing DT in IOC emerged as a prominent research focus, driven by the need to evaluate DT’s strengths, weaknesses, opportunities, and threats during the pre-adoption stage, and to assess whether adopted models aligned with intended objectives. Most studies in this theme employed qualitative methods, such as literature reviews and interviews. In some cases, DT was analyzed within the broader context of Industry 4.0 technologies in IOC. For instance, Xu et al. [16] conducted a systematic literature review on automation technologies in the manufacturing and assembly stages of IOC, analyzing 53 articles and identifying 22 tools, including AI, DT, and BIM. Their findings showed that DT supported seven key benefits in IC: interoperability, scheduling optimization, production traceability, production safety, manufacturability, quality assurance, and constructability—with its core contribution in enhancing assembly planning. These outcomes aligned with Cheng and Oconnell et al. [3,46], although Oconnell et al. [46] emphasized critical barriers such as lack of standardization, poor interoperability, cybersecurity risks, and limited real-world validation, which hindered DT’s full adoption.

Other papers in this cluster focused solely on DT in IOC. For example, Xie et al. [48] performed a three-step systematic review on DT in modular integrated construction (MiC), identifying opportunities such as real-time logistics coordination, progress monitoring, and AI-driven site management. However, they also reported challenges including sensor integration issues, cybersecurity and data privacy risks from cloud reliance, and regulatory barriers linked to unclear implementation guidelines. The study’s generalizability remained limited due to its small sample size (10 publications) and regional focus on Hong Kong. Consistently, Jaryani et al. [68] conducted a case study on a wood panelized construction firm, using interviews with professionals to define information needs for off-site construction management. The findings revealed challenges such as fragmented data integration due to multiple ERP systems, limited logistics and site tracking because of inadequate real-time monitoring, and inefficiencies arising from manual data collection through spreadsheets—challenges that DT integration could directly address [69].

Although the literature confirms growing attention to DT in IOC, Zhai et al. [60] examined modular housing perceptions in Australia using a focus group of 81 construction professionals. The findings showed that only 52% were aware of modular construction, with many perceiving modular housing as container-like boxes. Initially, 88% were neutral or unwilling to invest, but after a DT demonstration, 90% expressed a willingness to invest. These results highlighted persistent misconceptions about IOC, often shaped by regional context. Similar insights were reported by Lei et al. [70], who applied game theory to analyze stakeholder decision-making in DT adoption. Their findings demonstrated that addressing misunderstandings and lack of awareness could substantially increase the likelihood of financing IOC projects. Moreover, DT adoption depended heavily on economic factors. The analysis revealed equilibrium strategies where some stakeholders adopted DT while others delayed; however, government interventions—such as subsidies and incentives—could shift the equilibrium toward broader adoption.

Theme 1.2: Digital Twin Conceptualization: This theme focuses on conceptual frameworks and early-stage proofs of concept, primarily addressing theoretical rather than practical implementation aspects of DT in IOC. For instance, Čustović et al. [41] proposed a three-layered DT model integrating objectives, processes, and data/tools to improve coordination between prefabrication and on-site activities. Using hypothetical scenarios, the study demonstrated how DTs supported conflict resolution, resource optimization, and enhanced decision-making through bi-directional information flow. Similarly, Cole et al. [71] developed a DT architecture for precast concrete manufacturing, targeting temperature and moisture control during curing. The framework, applied in a small case from NVC Precast (Melbourne), combined IoT sensors, machine learning, and simulation across four layers—data acquisition, analytics, visualization, and feedback—to enable real-time monitoring and proactive control. The model enhanced sustainability, quality, and operational efficiency.

However, these models remained largely theoretical and lacked real-world validation. A more critical issue lay in their limited scalability—a gap addressed by Kosse et al. [72], who explored modularity in DTs across four dimensions: technical, process, spatial, and life cycle. Through a literature review, the study identified “information containers” as a practical approach to modular DT implementation, comparing Asset Administration Shell and Information Container for Linked Document Delivery as complementary solutions for semantic interoperability and machine-readability. To further bridge the implementation gap, Jaryani et al. [68] investigated the integration of knowledge graphs (KGs) into cognitive modular production. The study highlighted how KGs enhanced decision-making, interoperability, and automation in modular environments by supporting real-time analytics, predictive maintenance, and process optimization. An online-based survey of professionals from modular production, construction, and IT consultancy confirmed that KGs improved system interconnectivity and intelligent automation, reinforcing the potential of Cognitive DTs in advancing modular production. However, challenges in standardizing KGs across different modular production systems remained a critical issue the study failed to address.

Theme 2.1: On-Site Asset Tracking and Traceability: Since the initial implementation of DT applications in IOC in 2017, on-site component tracking and traceability has remained the most prominent research domain. Early work by Zhong et al. [32] introduced a multidimensional IoT-enabled BIM platform that established a three-phase process: RFID-based tracking during prefabrication, real-time monitoring during logistics, and automated verification during on-site assembly, with a special focus on the on-site assembly phase. This approach enhanced real-time visibility and inter-organizational coordination, producing quantifiable outcomes such as a 48.3% reduction in production paperwork, 40% shorter prefabrication lifecycles, 40% reduction in order picking time, and a 99.8% on-time delivery rate. These benefits underscored DT’s potential to streamline workflows and reduce errors in component handling and site installation. Comparable frameworks proposed by Li et al. [57], Zhou et al. [64], and Wang et al. [73] employed similar technological configurations—including RFID for component tracking, GPS for logistics data acquisition, and BIM for real-time visualization—and consistently reported gains in efficiency, decision-making, supply chain integration, and cost/time savings. However, these studies largely underdeveloped the service layer of DT systems, particularly in leveraging data analytics for predictive and adaptive decision-making—an area that gained traction in post-2020 research. For instance, Fischer et al. [74] advanced the analytical dimension by integrating extended value stream mapping, RFID-based tracking and tracing, and discrete-event simulation within a modular housing production context. By implementing this framework at Maxmodul (a subsidiary of Max Bögl Group, Sengenthal, Germany), the study identified critical production bottlenecks, reduced inventory by 64%, and accelerated lead times by 20%. This work illustrated the importance of data-driven simulation for system optimization, yet it also highlighted the complexity of translating RFID data into actionable intelligence, particularly in dynamic factory settings.

In the domain of robotics-integrated construction, Jiang et al. [61] proposed a DT-enabled Smart MiC System to enhance robotic on-site assembly via UWB positioning and real-time digital modeling. The system demonstrated improvements in control, navigation, and human–robot collaboration, with acceptable positioning accuracy (average deviation of 16.04 cm). However, the study highlighted key challenges: UWB’s inability to track module orientation, unstable outdoor positioning, limited scalability of robots due to environmental factors and workforce skills, and the need for AI autonomy to enable safer, adaptive automation through further machine learning research. Furthermore, Yang et al. [8] developed a 5D DT platform integrating IoT, BIM, complex event processing (CEP), and hierarchical finite-state machines (HFSM) for real-time material tracking in on-site fit-out construction. Using Automation Markup Language (AML) and OPC UA (Open Platform Communications Unified Architecture), it ensured smooth digital–physical data exchange, while CEP enabled automated, rule-based responses and HFSMs prioritized actions by urgency. Tested on a water pump house project, the platform enhanced coordination, reduced construction time, and improved traceability—demonstrating the effectiveness of combining real-time monitoring with intelligent control to address on-site communication challenges.

Collectively, while early DT models provided foundational benefits in traceability and monitoring, their limited analytic depth restricted predictive and autonomous decision-making capabilities. Later studies began to address these limitations through simulation, event logic, and robotic integration. Nevertheless, challenges persist around environmental variability, interoperability, and AI autonomy, indicating that while traceability is well-advanced in DT research, its full integration with intelligent decision support systems has remained a critical area for continued development.

Theme 2.2: Transportation and Logistics: This cluster examines DT applications aimed at enhancing tracking and traceability during the transportation and logistics phases of IOC. Lee et al. [15] proposed an integrated DT framework combining BIM and geographic information system (GIS) to improve supply chain coordination in modular construction. The approach leverages BIM for detailed module specifications (geometry, weight, delivery schedules) alongside GIS data on road networks, traffic conditions, and regulations. Real-time GPS data from in-transit modules dynamically updates the virtual model, enabling logistics scenario simulations via the Unity engine and route optimization using the Bing Maps API. A case study of a six-story modular apartment project in Seattle reported a significant reduction of 157.5 hours of idle time over 80 deliveries, primarily attributed to enhanced estimated time of arrival (ETA) accuracy and dynamic rerouting capabilities. This demonstrates the framework’s effectiveness in mitigating delivery delays and optimizing logistics operations. In comparison, Zhai et al. [60] developed an IoT-enabled BIM platform incorporating RFID, NFC, and GPS-based “smart trinity tags” for real-time asset tracking, complemented by a rule-based control system automating alerts, emergency responses, and progress updates. The cloud-based environment facilitated automated data collection, scheduling, and stakeholder coordination, reducing manual interventions and improving workflow integration. Practical application in a subsidized housing project confirmed notable gains in construction efficiency and collaboration. While Lee et al.’s [15] approach excels in leveraging spatial and traffic data for route optimization, Zhai et al.’s [60] platform emphasizes automation and real-time process control, illustrating complementary strengths. Together, these solutions highlight how diverse DT architectures address distinct logistics challenges, with their effectiveness contingent upon project scale, data availability, and stakeholder integration.

Theme 3.1: Data Security and Fast Indexing-Retrieval: Ensuring data security and integration is fundamental in DT applications, especially as DT evolved into a comprehensive framework for managing the entire lifecycle of IOC. Several studies explored blockchain as a key enabler of secure, transparent, and verifiable DT ecosystems. For example, Wu et al. [62] developed a Blockchain-Integrated IoT-BIM Platform (BIBP) to enhance off-site production management by addressing key IoT-BIM limitations, including data inconsistency, lack of trust, and version control. By integrating RFID sensors with Hyperledger Fabric and smart contracts coded in JavaScript, the system secured data provenance and automated verification tasks. Comparative analysis with conventional platforms showed improvements in data accuracy, traceability, and operational efficiency. Similarly, Li et al. [59] proposed a blockchain-enabled platform to tackle supply chain fragmentation in modular construction. The authors introduced a data–information–knowledge (DIK)-driven supply chain management (SCM) model within a three-layer system architecture (infrastructure, blockchain BIM, and software-as-a-service). Their prototype, validated on a modular student residence, demonstrated enhanced traceability, reduced latency, and lowered storage costs—confirming blockchain’s value in streamlining SCM processes.

Echoing these findings, Jiang et al. [75] and Dong et al. [76] advanced similar blockchain-integrated frameworks. Notably, Dong et al. [76] focused on real-time asset tracking and 4D change visualization. Their system automated data transfer between BIM and blockchain, used QR codes for physical tracking, and deployed smart contracts to detect schedule dependencies. A Navisworks plugin enabled 4D visualization, supporting transparent and collaborative project execution. The prototype validation confirmed the approach’s potential to increase trust and reduce manual inefficiencies. However, while these studies highlighted blockchain’s strengths in enhancing transparency and automation, they also shared limitations, namely network speed and the lack of offline communication capabilities. Addressing these gaps, Zhao et al. [4] introduced ChainPM, a Blockchain 3.0 framework optimized for DT-based construction project management. ChainPM improved on earlier blockchain models by enabling fast synchronization, offline functionality, advanced indexing, and complex querying. Validated through a modular construction case, ChainPM demonstrated superior performance in responsiveness and data integrity. Overall, while the reviewed solutions showcase significant advancements in securing and streamlining DT ecosystems in IOC, their comparative performance highlights a progression from foundational blockchain integration to sophisticated, high-performance frameworks; however, real-world scalability and network limitations remain areas for further research.

Theme 3.2: Data Integration and Interoperability: The challenge of information fragmentation and discontinuity remains critical in IOC, particularly when coordinating diverse technologies and multiple stakeholders. A key factor influencing interoperability in construction digital environments is the adoption of industry standards and platforms such as the Industry Foundation Classes (IFC), an open, international standard. IFC provides a widely accepted, open data schema designed to facilitate seamless information exchange across heterogeneous software tools in AEC projects. However, despite its foundational role, IFC’s complex data model and inconsistent implementation across platforms often hinder efficient integration of emerging technologies like AI-driven DTs especially in handling dynamic, real-time data streams that were not originally considered in the standard’s static design.

To address this, Jiang et al. [63] proposed a blockchain-enabled cyber–physical platform aimed at enhancing cross-enterprise information exchange. Using a user-centered design methodology, the framework integrated blockchain, IoT, and cyber–physical systems to support traceability, transparency, and real-time decision-making in modular integrated construction. The study demonstrated the full lifecycle of platform development and validated its impact on improving information continuity.

Focusing on production environments, Wen et al. [12] designed a four-layer DT framework to manage the prefabrication information of ultra-large, immersed tunnels. Applied in the Shenzhen–Zhongshan Link Project, the system integrated IoT, AI, BIM, and data management to optimize key operations such as concrete casting and section handling. The results showed improvements in process efficiency and intelligent monitoring, although the solution was highly specialized for factory-based tunnel fabrication. At the design-fabrication interface, Skoury et al. [13] developed a digital thread-based framework to improve data consistency and interoperability in robotic fabrication. By incorporating DTs, cyber–physical systems, and modular data schemas within a service-oriented architecture, the approach enhanced collaboration and reduced errors across design and fabrication domains. Its validation on the livMats Biomimetic Shell demonstrated effective support for co-design and automation in complex AEC projects.

Broadening the scope of interoperability, Ramonell et al. [77] proposed a knowledge graph-based DT system that integrated BIM, IoT, and other heterogeneous data sources. Built on cloud-based microservice architecture, the system enabled dynamic updates, efficient querying, and semantic interoperability. The framework was validated through several real-world demonstrators and was positioned as a versatile and scalable solution for DT development in built asset environments. While each study targeted distinct stages and contexts, a shared emphasis on real-time data integration and intelligent decision support was evident.

Theme 4.1: Dynamic Scheduling and Planning Optimization: Traditional scheduling in IOC often relied on fixed average production rates, which overlooked real-time variability and led to inaccurate estimates and inefficiencies. To address this, Alsakka et al. [18] developed DiTES, a DT system that integrated computer vision, ultrasonic sensors, machine learning, and 3D simulation to dynamically estimate cycle times and adjust production schedules in real-time. A case study at a wood-wall framing workstation showed an 81% reduction in scheduling deviations compared to conventional methods, highlighting DTs’ capacity to improve planning precision and operational responsiveness.

Expanding on real-time synchronization, Jiang et al. [17] proposed DT-SYNC, a DT-enabled model for planning, scheduling, and execution in precast on-site assembly. Using a hybrid validation approach—including numerical experiments and robotic testbeds—the study demonstrated improved coordination, reduced delays, and greater resilience to disruptions. Unlike DiTES, which focused on factory-level control, DT-SYNC emphasized on-site assembly synchronization.

Further advancing this concept, Jiang et al. [25] introduced an out-of-order (OoO) scheduling model inspired by CPU task sequencing. Leveraging DTs with IoT, computer vision, and real-time tracking, the model reordered tasks dynamically based on field conditions. Its rolling-horizon and forward heuristic mechanisms enabled adaptability to uncertainties such as material delays and labor fluctuations. The model reduced makespan by up to 15%, improved material handling efficiency, and outperformed both MILP and heuristic-based scheduling methods under real-time constraints. Collectively, these studies moved beyond static scheduling by leveraging DTs for real-time, adaptive planning.

Theme 4.2: Dynamic Resource and Layout Optimization: A core strength of DTs in IOC is their capacity to simulate, validate, and optimize production systems prior to physical deployment. Addressing high prefabrication costs, Wang et al. [78] used a DT-enabled 3D virtual environment to optimize facility layout through a modified genetic algorithm. By solving a mathematical model, the study improved workload balance, reduced transport distances, and minimized operational costs. This proactive optimization illustrated a pragmatic approach to improving prefabrication line efficiency.

Extending into robotics, Kaiser et al. [79] developed an automated method to generate DT models for simulating reconfigurable timber construction systems. Utilizing AML, the framework encoded geometry, kinematics, behavior, and interfaces, while simulations via ROS (www.ros.org, accessed on 11 August 2025) and Gazebo (www.gazebosim.org, accessed on 11 August 2025) supported rapid reconfiguration and integrity checks. Applied within the IntCDC system, this method enabled fast design iteration and adaptability—both vital for robotics-integrated construction workflows.

Building on these setup-focused strategies, Barkokebas et al. [80] investigated real-time optimization during fabrication by simulating multiskilled labor configurations. Using DTs as a surrogate for full-scale trials, the study employed discrete-event and continuous simulation to assess labor efficiency. The findings showed significant reductions in worker waiting times and enhanced labor flexibility. While multiskilling posed risks of diminished individual productivity, the DT-managed system offset these limitations, improving overall efficiency and reducing lean waste—underscoring DTs’ value in refining dynamic labor strategies.

Theme 5.1: Predictive Safety and Risk Management: While DTs in IOC are widely recognized for improving cost, time, and efficiency performance, their expanding role in predictive safety and risk management is becoming increasingly vital [81,82]. Zhao et al. [83] developed a DT framework for managing prefabricated component hoisting by integrating BIM, IoT sensors, and LoRa communication. The five-dimensional model—encompassing physical and virtual environments, data systems, service platforms, and interconnections—facilitated real-time tracking using IMU, GPS, and RFID sensors. Hoisting paths were optimized through Dijkstra’s algorithm, significantly reducing manual errors, collisions, and operation times. A case study validated its effectiveness in enhancing both safety and efficiency.

Extending this, Liu et al. [84] introduced a DT-based risk control framework that combined real-time monitoring, Bayesian networks, and a hierarchical safety model structured via DEMATEL-ISM methods. This framework enabled dynamic risk prediction, identified critical risk factors such as worker skill, equipment condition, and environmental variables, and supported proactive safety decisions. Case study outcomes indicated improved visualization, enhanced risk mitigation, and lower accident probabilities.

Reinforcing these developments, Li et al. [14] presented a DT platform for crane operations, showing that real-time localization significantly improved on-site safety. Addressing the limitations of vision-based methods—such as sensitivity to lighting and high deployment costs—Han et al. [85] integrated transformer-based pose estimation (PoseFormer) with UWB and IMU sensors. This hybrid system continuously updated the DT with precise position and orientation data, reducing average and final displacement errors by up to 87.29% and 56.95%, respectively. The result was improved stacking accuracy, real-time responsiveness, and enhanced installation safety.

Theme 5.2: Structural Health Monitoring and Maintenance: Structural health monitoring and maintenance remains a significant challenge in IOC due to installation inaccuracies, inefficient data transmission among stakeholders, and the limitations of traditional load rating methods, which tend to be costly, time-intensive, and disruptive. To address these challenges, Zhao et al. [67] proposed a DT framework that integrated RFID for component localization, a wireless sensor network for strain monitoring during installation, and LoRa communication for reliable data transmission to an on-site server. The collected data were visualized within a cloud-based BIM environment. Validated through a real-world application, the system enabled precise component tracking and satisfactory structural monitoring, thereby enhancing overall project performance by automating real-time data acquisition and analysis.

Complementing this, Ai et al. [86] and Rojas-Mercedes et al. [87] focused on improving load rating accuracy and structural risk assessment through DTs. Notably, Rojas-Mercedes et al. [87] developed seismic fragility curves for a precast reinforced concrete bridge in the Dominican Republic by integrating a structural health monitoring system with a DT. Real-time sensor data fed into a computational model to perform nonlinear incremental dynamic analyses, allowing probabilistic assessment of damage under varying seismic intensities. The study identified a 62% probability of extensive damage at peak ground acceleration (0.8 g) and a 30% probability of moderate damage at 0.6 g. The DT significantly enhanced fragility curve accuracy over traditional approaches, offering critical insights for disaster preparedness and post-earthquake recovery strategies.

Theme 6.1: Geometric Compliance Verification: DTs are transforming geometric quality control in IOC by embedding high-resolution 3D scanning into dimensional verification processes. In this regard, Tran et al. [65] proposed a DT framework for assessing the geometric fidelity of as-built prefabricated façades by comparing a 3D as-designed model against a 3D as-built semantic model derived from LiDAR point clouds. Automated correspondence identification enabled evaluation of completeness, correctness, and accuracy, facilitating rapid error detection and localization. Tests on both synthetic and real façade samples revealed that deviation visualization enhanced timely construction adjustments and supported rigorous quality assurance. Expanding this application into offsite manufacturing, Rausch et al. [88] presented a framework for geometric DTs during fabrication and assembly. The study evaluated three approaches—Scan-vs-BIM, Scan-to-BIM, and Parametric BIM Updating—using a commercial case study. Scan-vs-BIM demonstrated the highest accuracy, Parametric BIM Updating yielded semantically rich models, and Scan-to-BIM offered a balanced outcome. The authors advocated for hybrid implementations and advanced scanning technologies to optimize precision within offsite workflows. Advancing compliance and automation in prefabrication, Kirner et al. [89] introduced a semantic DT ontology tailored to robotic steel manufacturing. By incorporating process data, tolerances, and deviation metrics through semantic web standards (e.g., DSTV-NC), the model strengthened automation and resilience in production systems. A case study on robotic plasma cutting validated this approach, showing improved adaptability and consistent compliance with manufacturing tolerances.

Theme 6.2 Regulatory and Standard Compliance: Studies in this theme predominantly focus on automating rule-checking against codes and standards but often overlook the socio-technical challenges of operationalizing such automation across fragmented offsite workflows. Maturity frameworks for DT adoption—such as the Off-site Construction DT Assessment Model—provided a structured pathway for transitioning from manual documentation to automated compliance in panelized wood construction [90]. Case studies showed that about 78% of building code provisions could be translated into machine-readable formats, supporting real-time compliance during fabrication. However, the assumption that rule formalization alone ensured compliance omitted the variability in interpretation and exceptions embedded in many codes. The framework rightly stressed the need for phased digitalization, but broader institutional alignment, standard ontology development, and interoperability remained underexplored barriers to full automation.

Theme 7.1 Greenhouse Gas Mitigation: Integrating DTs into the prefabrication supply chain enhances smart construction by enabling real-time monitoring of carbon emissions and resource consumption across production, transportation, and on-site assembly according to Yevu et al. [5] Technologies such as RFID, AI, IoT sensors, and GPS supported accurate emissions tracking and carbon footprint estimation, improving logistics and operational efficiency. While real-time data acquisition enabled informed decision-making and sustainability optimization, assuming it automatically delivered environmental benefits overlooked challenges in data quality, granularity, and integration into planning. Greater focus is needed on how DTs support trade-offs between efficiency and embodied emissions, particularly in resource-constrained contexts.

Theme 7.2 Circularity in Offsite Construction: The potential of DTs to operationalize circular economy (CE) principles in offsite construction remains promising but conceptually underdeveloped. O’grady et al. [91] proposed a BIM-VR-DT framework via the Legacy Living Lab (L3), focusing on documentation, visualization, and education. While their findings indicated that 58% of materials were reusable and that the VR interface supported stakeholder learning, the study largely equated visibility with implementation. It underexamined how DTs could address deeper CE challenges, such as lifecycle traceability, material certification, or value chain coordination. The identified limitations—including the absence of robust material banks and the BIM model’s poor modularity representation—highlighted systemic gaps. The work pointed to the need for rethinking DTs not just as visual tools but as integrated platforms for circular SCM orchestration.

A detailed description and extended analysis of the reviewed papers are provided in the Supplementary Materials (Data S1).

4.3. Digital Twin in IOC Phases, Sectors, Components, and Materials

The literature has addressed various phases of the IOC project lifecycle, including planning and design, manufacturing, transportation, onsite assembly, operation and maintenance, and decommissioning.

The planning and design phase serves as a critical foundation in construction projects, defining the technical and organizational frameworks that guide all subsequent activities. A growing body of research (seven studies) emphasizes the role of advanced digital technologies, particularly BIM and DTs, in enhancing this phase [85]. These technologies are increasingly leveraged to improve design precision, simulate construction scenarios, and facilitate early coordination among stakeholders, ultimately reducing downstream errors and delays [13,92]. A relevant example can be seen in the work of Kaiser et al. [79], who developed a method for automatically generating DT models to simulate reconfigurable robotic fabrication systems in timber prefabrication. This approach enabled efficient layout validation and task simulation, addressing the limitations of manual DT adjustments. The study exemplifies how simulation-driven planning methods can significantly enhance the flexibility and responsiveness of design workflows, particularly in industrialized and off-site construction contexts.

The manufacturing phase has attracted considerable scholarly interest, with 23 studies underscoring its central role in ensuring the quality, efficiency, and reliability of prefabricated construction components. Research in this domain frequently explores automation, production process optimization, and real-time monitoring—highlighting the integration of IoT technologies and DTs as a core strategy to enhance traceability, safety, and quality assurance throughout manufacturing operations [16,93]. This emphasis is reflected in the work of Li et al. [59], who developed a Blockchain-Enabled IoT-BIM Platform to address supply chain fragmentation and trust issues in modular construction. Their platform integrates blockchain, IoT, and BIM within a data-information-knowledge framework, successfully improving information transparency and traceability in a modular student housing project. This example illustrates how advanced digital integration can streamline manufacturing workflows while enhancing the overall performance and reliability of modular construction supply chains.

Transportation represents a critical phase within the lifecycle of IOC, with 16 studies identifying the logistical complexities inherent in transferring prefabricated components from manufacturing sites to installation locations. These complexities often lead to schedule delays and increased costs, particularly when coordination and tracking mechanisms are insufficient [73]. In response, the literature highlights the adoption of IoT technologies and GIS-based routing systems as effective tools to optimize delivery schedules and minimize site-level disruptions [15]. A notable example is provided by Yang et al. [8], who developed a DT-enabled platform that integrates IoT, neural networks, and hierarchical finite state machines to enhance real-time monitoring and supply chain coordination. Although the study is applied to on-site fit-out construction, its focus on improving material visibility and traceability addresses core transportation challenges in prefabricated systems, reinforcing the importance of smart technologies in streamlining logistics across construction phases.

On-site assembly is the most extensively examined phase within the IOC process, with 29 studies dedicated to addressing its inherent complexities. The literature explores a range of strategies aimed at enhancing on-site operations, including real-time monitoring, automated tracking, and the optimization of hoisting and installation workflows [16,76]. Central to these advancements is the integration of DTs and IoT technologies, which are consistently highlighted for their ability to improve operational visibility, component traceability, and data-driven decision-making during assembly [61]. For instance, Zhou et al. [64] developed an IoT-enabled BIM platform to support on-site assembly services in prefabricated housing projects in Hong Kong. By incorporating technologies such as RFID, GPS, and smart sensors into the BIM environment, the platform demonstrated improved tracking, monitoring, and process integration. This example illustrates how digital platforms can streamline on-site operations and enhance the overall efficiency and reliability of prefabricated construction projects.

The operation and maintenance phase remains significantly underrepresented within the IOC literature, with only five studies addressing this critical stage. This gap reflects the sector’s traditional emphasis on earlier phases such as planning, manufacturing, and assembly, where the majority of interventions are typically concentrated. However, the limited existing research underscores the transformative potential of DTs and IoT technologies in post-construction contexts, particularly for real-time monitoring, condition assessment, and predictive maintenance [87]. A notable example is provided by Rojas-Mercedes et al. [87], who utilized a DT and structural health monitoring system to assess the seismic vulnerability of a precast reinforced concrete bridge in the Dominican Republic. By integrating sensor data with a computational model, the study developed seismic fragility curves to inform disaster mitigation and recovery planning. This case highlights how digital technologies can extend the value of prefabricated systems beyond construction, supporting long-term performance monitoring and resilience in infrastructure applications.

Decommissioning is the least explored phase in the IOC lifecycle, with only one study addressing its potential. This limited attention reflects a broader gap in lifecycle-oriented research, despite growing sustainability and circular economy imperatives. The existing study highlights the integration of DTs and BIM to facilitate the disassembly and reuse of building components [91]. By enabling detailed documentation and accurate tracking of elements throughout the building’s life, these technologies support systematic deconstruction, reduce construction waste, and minimize environmental impact [91]. This approach demonstrates the emerging role of digital tools in extending the value chain of prefabricated systems beyond their initial service life, aligning decommissioning practices with sustainable development goals.

Research on DTs in IOC spans a range of sectors, component typologies, and structural materials. Residential construction emerges as the most studied sector (15 studies), with a strong emphasis on improving design accuracy, real-time monitoring, and quality control through DT integration [57,65]. Infrastructure applications (five studies) primarily focus on logistics optimization and structural health monitoring [8,77], while commercial projects (three studies) concentrate on space utilization and facility management [88,91]. In contrast, healthcare and institutional settings remain underrepresented, with only one study addressing the use of DTs for enhancing operational efficiency and safety performance [25]. Regarding component typologies, non-volumetric pre-assembly—such as beams and slabs—dominates the literature (14 studies), followed by volumetric modules (8) and sub-assemblies or manufactured components (6). Modular buildings have received minimal scholarly attention, with only one study specifically addressing them. In terms of structural materials, concrete is the most frequently investigated (21 studies), followed by steel (7) and timber (6). Among multi-material strategies, hybrid systems combining steel and concrete are the most commonly explored, reflecting their prevalence in contemporary off-site construction practices. Figure 12 summarizes these findings.

4.4. Digital Twin Tools and Technologies

Sensing Layer: The sensing layer, also known as the data acquisition or perception layer, forms the primary interface between physical construction environments and the DT system. It collects real-time data from machines, components, and environments to ensure the virtual model remains synchronized with the physical system—supporting dynamic simulation, performance analysis, and informed decision-making [94]. As the foundation for communication, storage, and processing layers, it delivers high-fidelity data essential for analytics and integration, directly influencing the reliability of DT operations. According to the studies of DT in IOC, and as illustrated in Figure 13, this layer comprises five functional clusters. (1) Geometric and spatial sensing, using GPS, LiDAR, ultra-wideband, inertial measurement units, and accelerometers, captures dimensions and motion. (2) Identification and tracking, including RFID, near-field communication, and production sensors, focuses on real-time inventory and logistics visibility. (3) Environmental monitoring employs sensors like thermocouples, barometers, and humidity detectors to assess ambient conditions. (4) Visual and wearable sensing involves cameras and industrial wearables for image capture and worker monitoring. (5) Structural and mechanical monitoring uses strain sensors, fiber optics, and load cells to evaluate asset integrity and performance.

Different functional clusters within the sensing layer serve distinct application areas in IOC, with technology preferences shaped by their specific capabilities. Geometric and spatial sensing, primarily using GPS, is central to progress tracking and site coordination due to its effectiveness in asset localization and movement monitoring. Identification and tracking tools, particularly RFID, dominate supply chain and logistics optimization, offering cost-effective, real-time visibility of components and materials. Visual and wearable sensors, such as cameras, are mainly employed in safety and maintenance management for their versatility in monitoring worker activity and site conditions. Structural sensors, especially strain sensors, support quality control by detecting potential component failures and measuring load performance. Environmental sensors, including temperature and humidity monitors, are widely used in sustainability and circularity initiatives, tracking ambient conditions to ensure compliance and operational integrity.

A recurring trend in the reviewed studies is the integration of hybrid sensing systems, with 78% adopting two or more sensing technologies. The combination of GPS and RFID is particularly common, providing spatial tracking and identification simultaneously. However, challenges such as data synchronization in multi-sensor configurations and visual occlusions remain prevalent [18,48]. Timestamping and the addition of inertial measurement units are frequent strategies used to address these gaps [84,90]. In a study for managing on-site assembly services in prefabricated construction, Li et al. [57] suggest Edge computing to reduce latency in real-time processing [25]. Additionally, Ramonell et al. [77] reported RFID performance degradation due to material interference, which researchers mitigated by integrating vision-based methods and optimizing workspace layouts. The design and selection of sensing technologies directly influence DT accuracy. Multi-cluster systems demonstrate superior spatial fidelity compared to single-cluster setups. Even limited sensor integration shows substantial benefits. For instance, Rojas-Mercedes et al. [87], employing strain sensors for bridge structural health monitoring, reported a 40% improvement in failure prediction and an 87% reduction in displacement errors compared to human data records. However, several studies caution against excessive sensor deployment, as the complexity of data fusion may outweigh marginal accuracy gains.

The sensing layer’s data serve distinct but complementary roles: spatial and geometric data—such as point clouds and kinematic metrics—form the foundational digital scaffold, enabling precise tracking of assets, workers, and equipment across dynamic construction environments [14,65]. Production process data, particularly high-frequency metrics like task status and machine health, reveal bottlenecks and efficiency gaps in prefabrication workflows [90,95]. Inventory and material tracking data—the most extensively logged category—provide detailed visibility into supply chain states, which is critical for just-in-time delivery in modular construction [76]. Environmental and quality data act as compliance safeguards. Ambient condition tracking helps mitigate weather-related risks, while inspection records ensure regulatory adherence [62,83]. Structural performance data (e.g., strain and vibration) offer predictive insights into component durability [87]. The prevalence of worker-related metrics, including location and movements, reflects an industry shift toward safety-centric IoT integration [25]. Figure 14 presents the data types captured by the sensing layer in IOC.

Communication Layer: The communication layer plays a pivotal role in DT systems by enabling real-time data exchange between the physical and digital domains, thereby ensuring synchronization, responsiveness, and system scalability [96]. Rather than being a passive conduit, this layer actively shapes the reliability, latency, and interoperability of DT architectures across construction settings.

Across the reviewed studies, communication technologies are selected based on application-specific criteria such as transmission range, data volume, energy efficiency, and environmental conditions. Wi-Fi and cellular networks (4G/5G) are the most widely adopted technologies due to their high data throughput, scalability, and suitability for dynamic, mobile environments such as construction sites and modular factories [21]. Their support for wide-area coverage and ease of deployment make them ideal for real-time progress monitoring, remote control, and mobile worker tracking.

Short-range communication technologies, such as Bluetooth and Bluetooth Low Energy (BLE), are typically deployed in sensor-dense environments where proximity-based monitoring is critical—for example, in worker safety wearables or equipment status detection. BLE, in particular, is favored for its low power consumption, enabling long-term operation in battery-powered devices [22]. In contrast, long-range, low-power solutions like LoRa and LoRaWAN are leveraged in applications where centralized connectivity is impractical—such as monitoring prefabricated component movements or tracking environmental data over large construction sites [83]. These technologies strike a balance between energy efficiency and signal reach, although their limited bandwidth constrains them to transmitting small, periodic data packets.

In more controlled environments such as robotics labs or modular factories, wired networks like Ethernet and USB are preferred for their stability, low latency, and immunity to wireless interference. For instance, Ethernet is often integrated into robotic DTs, enabling high-frequency data transmission between physical and simulated models during control tasks or training cycles [13].

Communication protocols further determine how effectively data is structured, transmitted, and integrated across DT systems. MQTT is the most commonly adopted protocol due to its lightweight publish–subscribe architecture, which minimizes overhead and suits bandwidth-constrained applications with frequent sensor updates [21]. For secure, real-time data exchange in industrial settings, protocols like OPC UA, HTTP/HTTPS, and REST are employed to maintain interoperability across heterogeneous devices. AMQP and SFTP provide reliability and security in message queuing and file transfers, particularly for regulatory-compliant or multi-stakeholder environments [22].

At the data formatting level, XML and JSON facilitate interoperability by supporting structured and semi-structured data exchange. Specialized communication protocols are used to meet unique operational needs: Circle Redundancy supports fault-tolerant industrial automation; Ethernet KRL enables precise robotic control; and Python-based serial communication offers scripting flexibility in custom or modular DT applications [12,13,18]. Furthermore, the inclusion of Hyperledger Fabric Channels in one study signals growing interest in blockchain-enabled DTs, where secure, traceable, and decentralized communication is essential for collaborative construction project management [59]. Overall, communication in DTs is not a one-size-fits-all solution—it is a multi-layered, adaptive system where technology selection must align with spatial scale, energy constraints, latency tolerance, and data fidelity requirements

Storage Layer: The storage layer forms a critical backbone of DT architectures by managing the lifecycle of data—from initial capture to long-term archival and real-time analytics. Its structure and capabilities significantly influence the responsiveness, scalability, and interoperability of DT systems in IOC environments [21].

Cloud-based storage solutions from different technology providers such as Microsoft and Google dominate most implementations, reflecting the industry’s shift toward centralized, scalable infrastructures capable of supporting multi-user access, real-time processing, and integration with advanced analytics and visualization platforms [21]. These systems offer high availability and elasticity, which are particularly valuable in large-scale applications such as modular production coordination or cross-site monitoring. Cloud-native platforms also support seamless integration with machine learning pipelines, enabling predictive analytics for tasks like structural health monitoring or schedule optimization.

Despite the prominence of cloud systems, local storage solutions remain relevant in contexts where network connectivity is intermittent, data latency must be minimized, or systems are in prototype or experimental phases. Databases such as Microsoft Access, MySQL, and SQLite are often employed for such scenarios, providing lightweight and easily deployable alternatives for pilot-scale or localized DTs [77].

As DT implementations increasingly deal with heterogeneous and unstructured data, there is a growing reliance on NoSQL and non-relational database models. Document-based databases such as MongoDB [77] and CouchDB [59] support flexible data schemas, making them well-suited to dynamic construction environments where new data types (e.g., sensor metadata, component status, BIM annotations) are frequently introduced. Graph-based systems like Neo4j [77] allow for the modeling of complex relationships, such as interdependencies between prefabricated elements, supply chain actors, and construction phases—enhancing semantic query capabilities and data traceability.

Some studies have introduced semantic and ontology-driven storage frameworks to promote data interoperability and reasoning across distributed DT platforms. For example, Blazegraph enables ontology-based data management, which supports intelligent querying and integration with Linked Data sources [89]. Similarly, InterPlanetary File System (IPFS) has been explored as a decentralized, peer-to-peer file storage system, offering immutability and provenance—capabilities that are particularly beneficial in blockchain-integrated DTs or regulatory-compliant documentation workflows [63]. Collectively, these developments reflect a maturing DT ecosystem in IOC, where storage layer design is increasingly context-sensitive, tailored to factors such as project scale, collaboration intensity, data heterogeneity, and system resilience requirements. While MQTT-based transmission remains a common precursor to storage, the convergence of semantic modeling, cloud-native services, and distributed ledger technologies suggests a move toward more intelligent, decentralized, and interoperable architectures. Figure 15 summarizes the tools and technologies applied in DT-enabled IOC.

Digital Layer: The digital layer of DTs in IOC forms the foundation for establishing, visualizing, and maintaining the virtual counterpart of physical assets and processes [97]. This layer encompasses technologies that enable 3D modeling, geometric updates, immersive visualization, and simulation environments, ensuring accurate and dynamic synchronization between digital and physical domains throughout the project lifecycle [21]. Distinct from the service layer—which focuses on analytical interpretation and optimization—the digital layer is dedicated to the fidelity, continuity, and responsiveness of the virtual representation [35]. As depicted in Figure 16, we could identify four groups of the technologies supporting this layer: (1) 3D Modeling and BIM Tools, (2) Simulation and Game Engines, (3) Data Visualization and Monitoring, and (4) Extended Reality and Immersive Technologies.

Three-dimensional modeling and BIM tools constitute the foundation of digital model development, facilitating detailed virtual representations, parametric control, and the integration of geometric and semantic data [88]. The 3D BIM model has been the most frequently referenced concept, underscoring its central role in capturing both design intent and as-built conditions. Autodesk Revit (www.autodesk.com, accessed on 11 August 2025) is commonly employed for semantic modeling and parametric operations, often in combination with complementary tools. Rhinoceros 3D and Grasshopper 3D are primarily used in robotic prefabrication workflows, supporting complex geometry handling and automated modeling [89]. Additional platforms—including Autodesk Maya [63], SOLIDWORKS [61] (www.solidworks.com, accessed on 11 August 2025), Autodesk 3ds Max [91], BIM 360 [25] (www.autodesk.com, accessed on 11 August 2025), and Navisworks (www.autodesk.com, accessed on 11 August 2025) [67]—enable high-resolution modeling, visual simulation, and coordination within BIM-based project environments.

Simulation and game engines serve as powerful platforms for visualizing and modeling construction processes, particularly in robotic and offsite manufacturing contexts [12]. These tools enable the simulation of system configurations, motion sequences, and operational logic within immersive, interactive environments. Unity 3D (www.unity.com, accessed on 11 August 2025) is frequently utilized to model interactions among human operators, robotic systems, and prefabricated building components [15], while Unreal Engine (www.unrealengine.com, accessed on 11 August 2025) supports high-fidelity visualization of large-scale infrastructure scenarios [12]. Domain-specific platforms such as Gazebo, URDF, and XACRO facilitate the simulation of robotic kinematics and validation of motion planning strategies [95]. Additionally, Simio Software (www.simio.com, accessed on 11 August 2025) is employed to model factory layouts and workflow logic, supporting spatial planning and process visualization without engaging in detailed computational analysis [18].

Data visualization and monitoring comprises tools that translate complex datasets into intuitive visual formats to support human interpretation and operational decision-making [71]. Among these, data dashboards are most frequently cited, particularly for real-time monitoring of key performance indicators, construction progress, and sensor-derived feedback [71]. Visualization tools such as Gantt charts and dynamic location maps are commonly applied to schedule and logistics management [93], while deviation mapping platforms [89], automated alert systems [71], line-of-balance graphs [80], and S-curve charts [95] facilitate quality assurance and progress tracking. Collectively, these tools enhance both strategic oversight and real-time operational transparency.

Extended reality and immersive technologies refer to platforms that enable users to interact with DT environments through spatially immersive interfaces. Augmented reality (AR) and virtual reality (VR) have been employed to support applications such as assembly simulation, ergonomic analysis, and virtual construction walkthroughs [64]. Several studies [59,91] conceptualize XR as the integration of AR and VR to facilitate collaborative visualization and spatial decision-making. These technologies have proven particularly valuable in modular and prefabricated construction workflows, enhancing real-time alignment, stakeholder communication, and on-site validation [91].

Several studies highlight the integration of tools across software clusters to enable real-time visualization and simulation in construction robotics. For instance, Unity 3D is frequently coupled with Revit- or Rhino-based (www.rhino3d.com, accessed on 11 August 2025) BIM models to facilitate real-time visualization of robotic processes [17,89]. Autodesk BIM 360 serves both as a modeling environment and as a conduit for linking design models to real-time dashboards [67]. In robotic system simulations, formats such as URDF and XACRO are commonly integrated with platforms like Gazebo and Unity to model system behavior prior to physical deployment [95]. The literature also distinguishes between real-time and periodic visualization methods. Real-time visualization—enabled by platforms such as BIM 360, Unity, and interactive dashboards—is used for live monitoring of operations including module placement and deviation tracking [25,65]. In contrast, periodic visualization techniques, such as S-curves, line-of-balance charts, and post-scan deviation analyses, support retrospective evaluation and quality assurance [95]. This combination of real-time and periodic methods enables DTs to meet the varying feedback and responsiveness requirements of diverse construction activities.

Service Layer: This layer represents the analytical and cognitive engine of the DT framework in IOC. It operates beyond data visualization and static model handling, instead translating real-time data streams and historical records into actionable insights, predictive logic, and system optimizations [98]. This transformation hinges on a spectrum of algorithmic tools spanning machine learning, optimization, simulation, geometric validation, and decision support mechanisms. According to Figure 17, At the heart of this layer lies AI, particularly supervised learning techniques, which dominate in applications such as progress estimation, defect classification, and activity recognition from sensor or vision data. While neural networks—including deep learning architectures and fuzzy logic systems—offer the capacity to extract nonlinear patterns from complex datasets, their dependency on large volumes of annotated data and their limited interpretability raise concerns regarding their robustness and transparency in critical decision-making contexts [71]. Tools like PoseFormer (www.github.com/zczcwh/PoseFormer, accessed on 11 August 2025) have extended AI’s capabilities into spatial behavior tracking and pose estimation, supporting tasks such as ergonomic evaluation or equipment alignment [85]. However, such models often face practical limitations when confronted with noisy construction environments or occlusions in vision-based monitoring, pointing to a gap between algorithmic accuracy in controlled datasets and field applicability.

In optimization-intensive contexts such as layout configuration, resource sequencing, and robotic workflow planning, evolutionary algorithms—particularly genetic algorithms—have become prevalent due to their flexibility in handling multi-objective trade-offs and constraint-laden environments [16]. Yet, their convergence times and sensitivity to problem formulation can limit their suitability for real-time applications. Heuristic-based algorithms like Dijkstra’s and forward search strategies have proven more efficient for deterministic tasks such as path planning in robotic systems or connectivity analysis in modular assembly, though their capacity to handle uncertainty or dynamically changing environments remains restricted [83]. Meanwhile, quaternion-based estimation provides more precise modeling of spatial orientation for robotic joints and module alignment; however, it often requires integration with external simulation tools to validate mechanical feasibility [83].

Simulation and dynamic modeling form another critical dimension of the service layer, serving to forecast performance, test alternate scenarios, and evaluate the systemic impact of design or operational changes. Discrete-event simulation enables temporal modeling of prefabrication workflows, capturing dependencies and bottlenecks with granularity suitable for factory process optimization [74]. Continuous simulations and hybrid approaches—often supported by platforms such as Simio and Gazebo—offer more holistic insights into continuous system behavior, including robotic movements or structural responses under load [95]. However, such simulations remain heavily reliant on accurate input data and parameter tuning, and many studies neglect the incorporation of feedback loops from real-world sensors, undermining the representational fidelity and adaptability of these models over time. Despite the increasing availability of dynamic modeling platforms, integration challenges persist, particularly when linking process simulations with geometric or AI-driven components in real-time.

Geometric and spatial analytics underpin many quality assurance and assembly validation tasks within the DT ecosystem. Algorithms like Iterative Closest Point (ICP) and deviation analysis are routinely deployed for comparing as-designed and as-built geometries, providing dimensional feedback essential for ensuring fabrication precision and detecting installation errors [65]. While effective in capturing static discrepancies, these tools often function in isolation, lacking integration with predictive models that could foresee deviations based on preceding process trends. Finite Element Analysis (FEA), where integrated, contributes a performance-based layer to the DT, enabling structural behavior simulation under various load scenarios [86]. However, FEA integration is often limited to isolated studies in IOC, and its computational intensity and lack of real-time operability hinder its full deployment in dynamic assembly environments. This reflects a broader limitation: many advanced tools in the service layer operate as point solutions rather than being embedded in a coherent, interoperable ecosystem.

Perhaps most critically, the service layer serves as the gateway to real-time responsiveness and autonomous decision-making. Rule-based systems and complex event processing architectures enable DTs to respond to unplanned deviations, safety events, or resource shortages with conditional logic in IOC [8]. These are complemented by decision support systems that incorporate both historical data and predictive analytics to refine project strategies dynamically. Machine feedback systems, which interpret data streams from sensors and equipment during fabrication or robotic assembly, further enhance adaptability. Yet, the intelligence of these systems is still often constrained by pre-defined rules or black-box AI models, rather than by truly context-aware reasoning [35]. Moreover, while real-time systems are increasingly deployed during assembly and production, planning-phase applications remain dominated by static optimization and simulation tools, creating a temporal disconnect in DT intelligence across the project lifecycle.

Interoperability remains the linchpin of the service layer’s effectiveness. Emerging research indicates a move toward hybridization, where geometric analytics inform machine learning models, or where simulation outputs are looped into optimization routines to refine parameters iteratively. For instance, AI classifiers trained on geometric deviation data can inform tolerance-aware sequencing, and resource allocation algorithms can be dynamically adjusted based on simulation feedback [88]. These integrated workflows begin to shift the DT from a passive representation to an adaptive, learning system. However, the lack of standardized frameworks for cross-domain tool integration—especially across proprietary platforms—continues to limit scalability and robustness [18]. The service layer’s promise lies not simply in its computational strength, but in its capacity to orchestrate diverse forms of reasoning—geometric, temporal, statistical, and heuristic—into cohesive, context-aware intelligence that can evolve with the system it mirrors.

5. Discussion and Conclusion

5.1. Digital Twins in Industrialized Offsite vs. Onsite Construction

A comparative analysis of DT applications in IOC versus the broader construction and built environment reveals both overlapping priorities and distinct differences in focus and depth. Both domains emphasize core functional areas—including progress monitoring, data integration, quality control, supply chain optimization, and safety management—highlighting a shared foundation for DT adoption. However, the intensity and direction of focus within these categories diverge. For example, supply chain and logistics optimization is markedly more prominent in IOC studies, with increased attention to asset tracking, traceability, and transportation coordination, suggesting the coordination of prefabricated components between factory and site is critical to project performance [23]. Conversely, DT studies across the general construction sector exhibit a more balanced distribution of focus areas, with relatively greater emphasis on safety, risk management, and sustainability [99]. This broader orientation likely stems from the diversity of on-site processes and growing policy pressures related to environmental performance. DT research in IOC often delves deeper into production-oriented optimization strategies, such as dynamic scheduling, real-time progress tracking, and spatial layout planning—reinforcing IOC’s alignment with manufacturing principles. While both domains explore the integration of DT with Industry 4.0 technologies, IOC-related research typically treats this integration through the lens of feasibility, modular system compatibility, and implementation strategy. Broader construction studies, by comparison, frame such integration as part of overarching digital transformation initiatives.

The integration of DTs within IOC exhibits a distinct deployment pattern compared to other emerging technologies. While prior technological interventions in IOC have predominantly focused on early project stages—such as design coordination, constructability analysis, and multi-phase planning—DT applications increasingly emphasize execution-oriented phases, including manufacturing, transportation, and on-site assembly. This shift suggests that DTs function not only as strategic planning tools but also as operational systems that manage the real-time flow of physical components and data across the production-to-site continuum. Their capacity to synchronize virtual models with actual conditions enhances precision, facilitates logistics coordination, and supports timely decision-making during critical execution phases [38]. Unlike earlier technologies that addressed project phases in isolation, DTs enable continuity and interoperability across the entire lifecycle, positioning them as inherently integrative tools. Nevertheless, their currently limited application in the operation and end-of-life stages highlights a gap in leveraging DTs for feedback-driven sustainability and asset performance optimization.

Across structural systems and project types, DT implementation patterns align with broader IOC trends, yet reveal important nuances. Concrete-based systems remain predominant—likely due to their compatibility with standardized prefabrication methods and their widespread use in residential construction, which is the primary focus of most studies. Notably, most DT implementations are associated with non-volumetric components (e.g., slabs, beams) rather than full modular units. This may stem from the relative ease of modeling, tracking, and integrating discrete elements compared to complex volumetric assemblies. The technology remains primarily focused on components, with system-level integration still in its early stages.

While both IOC and conventional construction adopt a layered DT architecture—comprising sensing, communication, digital modeling, data integration, and service layers—IOC emphasizes tightly integrated, production-oriented implementations. In IOC environments, sensing technologies are routinely deployed in dynamic, factory-like conditions, often using hybrid configurations (e.g., GPS-RFID systems, inertial measurement units) to ensure high-fidelity spatial tracking and real-time monitoring of modular workflows. In contrast, general construction DT applications typically focus on environmental and occupant sensing (e.g., temperature, humidity, CO₂ levels), with greater attention to operational performance and building lifecycle management. Although both domains utilize similar tools—such as LiDAR, RFID, and computer vision—IOC applications more frequently address challenges related to high-frequency data synchronization, occlusion, and sensor fusion, reflecting the complexities of coordinating concurrent and prefabricated tasks [18]. At the digital and service layers, IOC implementations demonstrate a deeper level of tool integration. Whereas general construction commonly employs BIM, IoT platforms, and visualization tools (e.g., Revit, Unity, Three.js (www.threejs.org, accessed on 11 August 2025)) for asset tracking and facility optimization, IOC workflows extend these technologies to include robotic assembly systems, simulation engines, and programmable logic controllers.

The service layer in IOC often merges simulation, AI-driven prediction, and geometric analytics into integrated decision support systems for real-time coordination across manufacturing, logistics, and on-site assembly. Tools such as Unity and Simio are not only used for visualization but also for simulating prefabrication scenarios to optimize sequencing and minimize delays. Additionally, storage practices diverge: general DT applications tend to rely on cloud-based infrastructure for smart building management, while IOC implementations explore hybrid models (e.g., MongoDB, IPFS, LevelDB) to support decentralized data resilience and traceability. These differences suggest that although both domains increasingly converge on shared technologies, their implementation logic diverges based on project context. General construction DTs prioritize building performance, occupant safety, and operational efficiency, whereas IOC-focused DTs emphasize manufacturing precision, real-time synchronization, and logistical optimization, highlighting the adaptive and evolving role of DTs.

5.2. Trade-Offs in Digital Twin-Enabled Industrialized Offsite Construction

The integration of DTs in IOC offers transformative operational advantages, including enhanced real-time monitoring, predictive analytics, and system-wide optimization. These improvements support construction precision, minimize material waste, and strengthen quality control [100]. However, these benefits are accompanied by significant trade-offs that extend beyond technical performance, demanding a more nuanced socio-technical and organizational analysis.

Financial investment remains a core constraint. Achieving the data-intensive capabilities of DTs requires robust sensor networks, interconnected IoT systems, and high-performance computational infrastructure [22]. Such capital and operational expenditures can reinforce structural inequities, disproportionately favoring large firms with greater digital maturity while marginalizing small and medium-sized enterprises. The uneven capacity for digital investment thus creates a fragmented innovation landscape, where diffusion is constrained not by technological readiness alone but by institutional and financial asymmetries.

Environmental trade-offs, too, warrant critical scrutiny. While DTs are often promoted for their role in enabling energy-efficient processes and reducing construction rework, these sustainability benefits can be undermined by the ecological footprint of the digital systems themselves. Energy-intensive data processing, the embodied carbon in sensing hardware, and the accumulation of electronic waste contribute to an environmental paradox: DTs designed to support sustainability may also generate hidden environmental costs [101]. Addressing this contradiction requires a life-cycle perspective that evaluates net environmental impacts rather than relying on isolated efficiency metrics.

The transition to DT-enabled workflows also prompts fundamental shifts in organizational practices and labor dynamics. The demand for digitally literate professionals—ranging from data engineers to system integrators—alters the composition of construction teams and challenges existing skill structures [42]. While these developments open pathways for workforce upskilling and professional mobility, they introduce short- to medium-term disruptions. Many firms face transitional productivity losses during technology assimilation, especially where digital training is unstructured or misaligned with operational realities. More critically, the uneven distribution of digital literacy risks entrenching labor-market inequalities unless proactive strategies for inclusive training and career support are implemented [42].

Beyond organizational readiness, broader governance and regulatory structures remain underdeveloped in relation to the accelerating adoption of DTs. Ethical concerns related to data surveillance, cybersecurity vulnerabilities, and algorithmic opacity have yet to be systematically addressed within construction governance frameworks [6]. As DTs collect and process increasingly granular data on workers, materials, and operations, the risk of intrusive monitoring or biased decision automation grows. Regulatory frameworks must evolve not only to safeguard privacy and safety but to ensure that algorithmic decision-making aligns with principles of equity and accountability.

At the systemic level, the integration of DTs holds the potential to reshape IOC logistics and urban infrastructure delivery through predictive modeling, just-in-time coordination, and reduced site disruption [73]. However, persistent challenges in interoperability, data standardization, and legacy system integration continue to inhibit scalability. Without clear technical standards and cross-sectoral coordination, the promise of a connected, adaptive construction ecosystem remains aspirational. Moreover, the layering of new digital systems onto outdated institutional processes may generate unintended inefficiencies, reinforcing the importance of aligning technological innovation with organizational change and policy reform [21]. Collectively, this leads us to the fact that the adoption of DTs in IOC is not merely a technical upgrade but a socio-technical transformation. Its success depends on navigating the tensions between innovation and inclusion, sustainability and digital burden, and autonomy and governance.

5.3. AI in Digital Twin-Enabled Industrialized Offsite Construction

AI serves as a core enabler of DTs in IOC, advancing DTs from static models to adaptive systems capable of learning, prediction, and autonomous response [39,40]. AI enhances DTs by modeling system dynamics, identifying causal relationships, and generating insights for production sequencing, resource allocation, scheduling, and quality control [26]. AI integration transforms DTs from passive visualizations into adaptive, data-driven systems. Machine learning—particularly supervised and reinforcement learning—enables real-time anomaly detection, failure prediction, and workflow optimization. In IOC, predictive models trained on sensor data can anticipate equipment failures with high accuracy, while reinforcement learning dynamically allocates resources in response to changing conditions [18,89]. These capabilities position DTs as active agents in real-time decision-making and process optimization.

AI integration in DT-enabled IOC is most advanced at the service layer, where machine learning supports data aggregation, model deployment, and predictive analytics. Techniques such as supervised learning for defect detection, reinforcement learning for logistics, and clustering for pattern recognition have improved efficiency, with reported gains including 20% waste reduction and 25–30% higher throughput in modular facilities [70,102]. However, broader adoption is hindered by interoperability issues, inconsistent data quality, limited sensor integration, and scalability challenges. Additionally, applications in safety-critical contexts must address explainability, cybersecurity, and regulatory compliance to ensure trust and reliability [13]. At the sensing layer, edge AI and TinyML allow real-time, low-latency processing on embedded devices [103,104]. At the communication layer, AI manages network traffic, prioritizing critical data in sensor-dense environments. The storage layer benefits from AI-driven compression, retrieval, and privacy-preserving methods, particularly in federated systems [105]. At the modeling layer, physics-informed neural networks (PINNs) accelerate structural simulations by up to 70%, maintaining accuracy while enabling continuous model updates from real-world data [106].

Emerging AI paradigms, particularly large language models (LLMs), hold significant yet underexplored potential for enhancing DT-enabled IOC. While not yet widely adopted, LLMs can process unstructured data, synthesize insights, and enable natural language interaction—allowing construction personnel to query DT platforms for real-time forecasts, safety reports, and diagnostics without technical expertise [107]. Agentive AI systems, though still emerging in construction, offer promising support for decentralized decision-making in IOC [108]. Integrated with DT platforms, multi-agent systems could autonomously manage workflows and respond to disruptions in real time. While LLMs and agentive AI are not yet widely adopted, their convergence with DTs marks a shift toward fully autonomous, adaptive, and user-centric construction. Realizing this potential will require targeted research, cross-disciplinary collaboration, and pilot testing, aligning with the modular and data-driven trajectory of offsite construction [9]. AI in DT-enabled IOC is poised to advance across autonomy, interoperability, and resilience. Deep reinforcement learning and causal inference will support autonomous, context-aware decision-making. Semantic AI will enhance interoperability by aligning data across platforms. Self-healing models will improve resilience by detecting and correcting biases, ensuring robust performance in dynamic conditions [42,109].

5.4. Gaps, Limitations, and Recommendations for Future Research

Despite meaningful progress in leveraging AI–driven DTs to enhance IOC, several persistent gaps continue to hinder scalability, technological maturity, and broader industry uptake. These gaps delineate clear directions for targeted and actionable academic and industrial research.

Limited Scalability and Real-World Implementation: While DT technologies show promise in improving design coordination, performance monitoring, and predictive maintenance, their real-world adoption has been limited by a critical scalability gap. Most current applications remain confined to experimental or small-scale settings—such as modular housing projects—where they address isolated functions like logistics tracking or energy modeling [17]. These pilots demonstrate technical feasibility but fail to resolve the multidisciplinary integration and operational complexity required for large-scale construction.

This scalability challenge is evident even in advanced DT initiatives. The West Cambridge Digital Twin project (led by the Centre for Digital Built Britain) represents one of the few attempts to deploy DT systems beyond single structures, integrating buildings, infrastructure, and energy networks across an urban development zone [110]. Yet, despite robust funding and academic support, the project encountered persistent barriers to scalability, including fragmented data governance, mismatched stakeholder priorities, and incompatible digital infrastructure. These findings underscore that technical scalability alone is insufficient; institutional and operational adaptations are equally critical for expansion.

The most pressing scalability gap lies in high-stakes, complex projects. Unlike residential or modular pilots, large-scale infrastructure—such as high-rise commercial developments, transport hubs, or hospitals—demands rigorous safety protocols, real-time multi-system coordination, and compliance with dynamic regulatory frameworks. However, no empirical studies have validated whether current DT solutions can meet these demands at scale. Without evidence from such environments, the construction industry lacks a roadmap to bridge the divide between pilot-scale success and sector-wide transformation.

Finally, a narrow focus on residential projects and concrete-based systems reflects a bias that overlooks the complexity of other IOC typologies. Critical sectors such as commercial towers, healthcare infrastructure, and transport hubs—where the stakes for error are higher—are largely absent from empirical studies [26]. Additionally, the dominance of research in countries like China, Hong Kong, and the United States—where DT maturity is driven by strong academic and industrial ecosystems—may skew investigations toward region-specific needs, leaving other regions underrepresented. Similarly, non-volumetric approaches continue to dominate, sidelining important innovations in timber, steel, and hybrid assemblies.

This lack of context-sensitive, real-world evidence weakens industry trust and inhibits investment in DT platforms. Without field-validated demonstrations of scalability across diverse geographies and project types, stakeholders cannot justify transitioning from traditional workflows to AI-integrated systems.

Future studies must shift from laboratory trials to full-scale pilot implementations that integrate DTs across the entire construction lifecycle—from design through operations and decommissioning. Priority should be given to high-complexity environments (e.g., hospitals, airports), underrepresented construction methods (e.g., mass timber volumetric modules), and context-specific adaptations in regions beyond dominant research hubs. Research frameworks should emphasize interoperability, resilience under uncertainty, and performance benchmarking under real-time constraints.

Lack of Comprehensive Standardization for DT Levels of Development: The absence of standardized frameworks for defining and evaluating DT maturity levels poses a fundamental barrier to interoperability, industry benchmarking, and regulatory compliance [21]. At present, DT systems are deployed in silos, often customized to project-specific goals without reference to a common architecture or evaluation rubric. This leads to redundancies, inconsistent performance expectations, and a lack of transparency in procurement and implementation strategies.

Without standard maturity models akin to those developed for BIM, it is impossible to compare DT deployments across projects, verify system capability, or design scalable solutions that evolve over time. This also complicates the certification and regulation of DT-enabled construction practices.

Direction: Immediate attention should be directed to developing an internationally recognized DT Levels of Development (LoD) framework, detailing progressive benchmarks for functionality, integration, autonomy, and data fidelity. This taxonomy should be co-developed through academic–industry–regulatory partnerships and aligned with existing BIM and IoT standards to ensure consistency. Such a framework must include clearly defined metrics for data acquisition frequency, sensor resolution, AI autonomy levels, and decision support capabilities at each maturity stage. It would enable project stakeholders to calibrate expectations, align workflows, and evaluate return on investment more objectively.

Insufficient Integration of Digital Twins Across IOC Lifecycle Phases: Most existing research on DTs in IOC remains focused on early lifecycle phases—particularly design, planning, and production—while neglecting the equally critical stages of operation, maintenance, and end-of-life management. This myopic view restricts the long-term value that DTs could offer in optimizing sustainability, asset longevity, and resource circularity. For instance, digital systems rarely extend to predictive maintenance planning, adaptive operational control, or post-use disassembly strategies—key areas for future-ready construction.

The failure to consider the full lifecycle undermines the ability of DTs to support carbon tracking, facilities management, and material circularity—imperatives in the context of net-zero goals and sustainability mandates. This creates a disconnect between digital transformation ambitions and real sustainability outcomes.

Future studies must adopt a lifecycle systems-thinking approach, embedding DTs into all phases—from construction through to asset end-of-life. There is a critical need to develop AI models capable of integrating sensor data with lifecycle performance indicators such as embodied carbon, maintenance history, and operational energy use. Research should also explore the role of DTs in enabling dynamic decommissioning planning, material passports, and reuse logistics, thereby directly contributing to circular economy transitions in IOC.

Manual Configuration or Intervention: Despite the promise of automation, many existing DT systems still require substantial human intervention for tasks such as BIM parameter updates, model calibration, and reconciliation between real-time and design-intent data [48]. These manual inputs not only increase workload and reduce scalability but also introduce variability and potential for error, especially in large-scale, multi-stakeholder projects.

Manual intervention compromises one of the foundational goals of DTs: automated, real-time decision-making. It also contradicts the productivity and efficiency narratives used to justify DT investments, especially for resource-constrained firms and projects operating under tight timelines.

A critical research priority is the development of AI-agentive systems that can autonomously handle data interpretation, update synchronization, and real-time decision support. Large language models (LLMs), reinforcement learning, and digital multi-agent systems should be explored for their ability to orchestrate complex tasks across digital environments. For instance, AI-powered agents could automatically harmonize data streams between BIM and sensor platforms or generate maintenance alerts without human initiation. Such advancements would bring DTs closer to a truly autonomous state, unlocking their full transformative potential.

Data Integration, Interoperability, and Security Issues: One of the most persistent and technically challenging barriers to DT implementation in IOC is fragmented data ecosystems. Heterogeneous data formats, proprietary platforms, and limited interoperability between BIM, IoT, and ERP systems create information silos that undermine DT cohesion. Furthermore, the growing reliance on networked sensor systems introduces major vulnerabilities in cybersecurity, including threats such as data tampering, sensor spoofing, and model poisoning [13].

Without addressing these integration and security challenges, DT systems will remain brittle, opaque, and vulnerable—ultimately reducing their utility and trustworthiness for critical infrastructure. The lack of unified data protocols also impedes collaborative workflows across project stakeholders, resulting in duplication and data loss.

Future research should prioritize the development of open, semantic data models capable of harmonizing heterogeneous data streams into unified, queryable structures. This includes designing ontologies that bridge BIM metadata, IoT telemetry, and enterprise resource planning datasets. Simultaneously, blockchain and distributed ledger technologies should be investigated for secure, tamper-proof data exchange mechanisms. Establishing standardized encryption protocols, data access hierarchies, and digital ownership frameworks will be essential to securing DT ecosystems and ensuring regulatory compliance in sensitive construction sectors such as healthcare and defense.

Dependency on High-Precision Equipment: Many state-of-the-art DT implementations rely on advanced, often prohibitively expensive hardware such as LiDAR scanners, automated robotics, and high-resolution imaging systems [22]. While these technologies enhance data richness and model fidelity, they create significant accessibility barriers, particularly for small and medium-sized enterprises that dominate the construction sector.

This dependency fosters a two-tiered innovation system, where large firms can capitalize on digital advantages while smaller players are effectively excluded. It also slows down the broader democratization of DTs in the built environment, reinforcing digital inequity within the industry.

Researchers must explore cost-effective alternatives to expensive hardware, such as using low-cost depth cameras, photogrammetry, or edge-AI sensors with optimized performance-to-cost ratios. In parallel, design guidelines for scalable and modular DT systems—tailored specifically for SMEs—should be developed, including plug-and-play architectures and pre-trained AI modules. Economic modeling studies should investigate the return on investment of minimalist DT deployments, providing business cases that encourage adoption even among resource-limited stakeholders.

Unclear Cost–Benefit Analysis: Despite the widely discussed benefits of DT technologies in IOC, there is a notable lack of empirical evidence evaluating their financial performance. Most studies remain theoretical, with limited quantification of key metrics such as return on investment (ROI), total cost of ownership, payback period, or costs related to data management, training, and cybersecurity. This gap leaves stakeholders without the financial justification needed for capital-intensive adoption.

Real-world initiatives illustrate this issue. In Singapore’s HDB Smart Estate program, DTs improved asset monitoring and predictive maintenance but incurred substantial costs for data integration, secure cloud infrastructure, and inter-agency interoperability. These costs ultimately restricted broader deployment [111]. Similarly, the UK’s National Digital Twin program acknowledged that demonstrating tangible economic value is crucial to stakeholder engagement [11]. While early pilots showed potential for lifecycle gains, few offered frameworks to calculate financial returns or risk-reduction impacts.

Further complications arise from regional and typological variability. Cloud and cybersecurity expenses vary with local data regulations, while the cost of workforce upskilling depends on existing digital maturity. Such inconsistencies hinder the generalizability of findings and limit scalable adoption strategies.

The absence of rigorous, context-specific cost–benefit analysis makes DT adoption appear risky and uncertain. Without clear evidence of economic value, especially for diverse project types and geographies, organizations remain hesitant to transition from traditional methods. Future studies must develop practical economic models—such as lifecycle costing and ROI simulations—and encourage benchmarking across projects to support evidence-based investment in DT platforms.

There is an urgent need for comprehensive, sector-specific cost–benefit models that evaluate both tangible and intangible impacts of DTs. These models should consider not only capital and operational costs but also potential savings from reduced rework, enhanced safety, lifecycle asset performance, and carbon reduction. Longitudinal case studies across different project types (residential, commercial, infrastructure) should be conducted to draw comparative insights and provide industry benchmarks.

The future development and adoption of AI-driven DTs in IOC will be contingent upon resolving the above limitations through rigorous, cross-disciplinary research. These challenges—ranging from technical standardization to economic viability—are not merely operational details but foundational issues that determine whether DTs can move from innovation to industry norm. Addressing these gaps will require not only technological solutions but also new models of collaboration, regulation, and education within the AEC sector.