1. Introduction
1.1. Background
1.1.1. Road infrastructure
Road infrastructure serves as a cornerstone for economic, social, and environmental activities within a society. It connects urban centers, towns, and rural areas, thereby facilitating the movement of people, goods, and services. As a critical element of public transportation, China’s road network surpassed 5.43 million kilometers in 2023, accommodating over 40 billion tonnes of freight and more than 11 billion passenger trips annually. Establishing transport networks with broader coverage and higher speeds is essential for enhancing global competitiveness in the transportation sector. Achieving this objective depends on the development and maintenance of extensive and efficient road networks, which are integral to national prosperity and strength.
Roads are often constructed in complex geological environments, traversing diverse terrains and landforms while being subjected to natural disasters and harsh environmental conditions. Heavy traffic loads and environmental stressors accelerate the degradation of road infrastructure, frequently shortening lifespans below design expectations. This degradation leads to reduced economic benefits and diminished traffic efficiency. Addressing these challenges has become increasingly critical in the context of urbanization and rising transportation demands.
In the current era of economic prosperity and technological advancement, road infrastructure has evolved beyond the goal of mere accessibility, with increasing emphasis on digitalization and intelligence as it moves toward future development
[1]. This transformation is supported by progressive policies. For example, Ministry of Transport of the People’s Republic of China has consistently prioritized the digital transformation of road infrastructure and the construction of smart highways. Similarly, the United States allocates 100 million dollars annually to support transportation digitalization and intelligent systems
[2]. European countries are also experimenting with smart highways that leverage advanced information technologies. The trajectory of road infrastructure development is unmistakably directed toward greater digitalization and intelligence, fostering advancements in road construction and operations.
From the perspective of life-cycle performance, implementing effective management strategies at each stage of road engineering is an essential approach. The life cycle of road infrastructure encompasses various stages, including design and construction, operation and use, maintenance and repair, rehabilitation and upgrades, and eventual decommissioning. The substantial costs associated with road engineering, particularly in network construction and maintenance, impose a significant burden on public finances. To address these challenges, various digital technologies, including digital twin (DT), have been proposed to enhance road performance, improve management practices, and optimize life-cycle costs.
1.1.2. Digital twin
(1)
The concept of DT. A DT creates a digital replication that serves as an identical counterpart to a specific physical object. The concept, originally proposed by the National Aeronautics and Space Administration (NASA) in the United States, has been widely adopted. NASA defines a DT as an integrated multi-physics, multi-scale, and probabilistic simulation of a vehicle or system that utilizes the best available physical models, sensor updates, fleet history, and other data to replicate the life of its physical counterpart
[3].
A precise definition of specific terms is critical during the initial stages of technological development. Therefore, it is necessary to differentiate DT from related concepts such as building information modeling (BIM), cyber–physical systems (CPSs), and digital shadows. BIM is a collaborative approach for creating, managing, and visualizing three-dimensional (3D) digital models of buildings and projects. Although BIM and DT are sometimes used interchangeably, notable differences exist. DT emphasizes the bidirectional data connection between a physical entity and its corresponding virtual model, whereas BIM lacks this real-time connection and is merely considered a foundation for DT construction
[4]. A CPS refers to an integrated system wherein physical components, such as sensors, actuators, and machinery, interact with digital components, including software, algorithms, and networks
[5],
[6]. In CPS, the mapping between physical and digital components is one-to-many, while in DT, it is strictly one-to-one. CPS prioritizes sensors and actuators, whereas DT requires a corresponding virtual model and a bidirectional data connection
[5],
[6]. A digital shadow represents a semi-automated digital model characterized by unidirectional data flow between a physical object and its digital counterpart
[7]. In contrast, a DT is a unique digital model that facilitates bidirectional data flow between the real world and its virtual twin. Based on the level of data flow automation, Kritzinger et al.
[7] classified DT into three subcategories: digital models, digital shadows, and DT. Yan et al.
[8] overviewed the differences between DT, BIM, and CPS. Furthermore,
Fig. 1 illustrates the comparison between DT and the analogous concepts.
This study references a five-dimensional (5D) model
[9] to discuss the related enabling technologies. Originally developed by Grieves and Vickers
[10], the DT model consists of three core components: physical entities, virtual entities, and connections. Tao et al.
[9] later expanded this model by incorporating twin data and services, resulting in a 5D DT framework, as represented in Eq. (1):
where M
DT indicates the DT models, PE indicates the physical entities, VM indicates the virtual models, Ss represents the DT services, DD represents the DT data, and CN is the connection between the components. This enhanced DT model has been successfully applied in various domains, including ship design, workshops, and smart cities
[11]. Its potential applications in the construction industry have also been demonstrated
[5].
In summary, a DT can be defined as an integrated framework with multiple layers, encompassing a digital model, twin data, and bidirectional connections. Static digital models, such as BIM, can be considered prototypes of DT. Although digital shadows share greater similarities with DT, they operate with less data-flow automation. While CPS is a conceptually parallel framework, its focus diverges from DT’s. Most DT implementations require physical entities, virtual models, and robust data connections.
(2)
The development of DT. Based on the number of publications indexed in the Web of Science that include DT,
Fig. 2 illustrates the three developmental stages of this transformative technology
[12]. In 2002, Grieves introduced the concept of DT during a speech on product life-cycle management (PLM). Subsequently, this concept found its way into his published works
[10]. However, no publicly available studies on this topic were initially documented. In 2005 and 2006, Grieves referred to the concept as the “mirrored spaces model” and the “information mirroring model,” terms that were later included in his published works
[10]. In 2010, NASA formally defined the DT of an aircraft, highlighting its significant potential within the aerospace sector
[3]. Following the publication of the first DT-related article by Shafto et al.
[3], researchers gradually began exploring the concept and its applications. In 2018, Tao et al.
[13] studied DT-driven products, manufacturing, and services using big data within the context of Industry 4.0. They also proposed a systematic 5D DT model
[9], reflecting the increasing maturity of DT theory. Since then, DT has been widely applied across various fields, resulting in a significant surge in recent publications.
Originating in the aerospace industry, DT has rapidly gained adoption across various fields, including smart production management
[14], DT-driven manufacturing system designs
[15], precision cardiology
[16], cloud-based battery monitoring in the energy sector
[17], and the integration of DT with wireless networks for advancing the sixth-generation mobile communication technology (6G) systems
[18]. This paradigm has been employed to forecast system health, estimate remaining useful time, and assess mission success probabilities
[19]. Recognizing the diverse application domains of DT, Dalibor et al.
[20] conducted a cross-domain systematic mapping study on software engineering for DT. Their research demonstrated that DT serves as the technological foundation for improving the understanding and management of existing systems, as well as for designing new systems across various domains. Moreover, Muctadir et al.
[21] conducted interviews with 19 professionals from different fields and identified substantial growth potential for adopting enhanced software engineering practices, processes, and expertise in leveraging DT. The potential applications of DT continue to evolve, often surpassing current expectations.
(3)
DT in infrastructures. The architecture, engineering, and construction (AEC) industries have increasingly employed DT principles to enhance infrastructure design and maintenance. University of Cambridge described DT in the built environment as “a realistic digital representation of assets, processes, and systems” that integrates data representing the physical environment in a digital format
[22]. AlBalkhy et al.
[23] defined DT in this context as a dynamic virtual replica of a physical twin, ranging from simple prototypes to sophisticated models, designed to support numerous functions throughout the lifecycle with appropriate enabling technologies. Boje et al.
[24] reviewed the status of BIM applications in construction and analyzed DT applications in adjacent fields, identifying gaps and proposing a construction-specific DT framework comprising three generations: monitoring platforms, intelligent semantic platforms, and agent-driven socio-technical platforms. Their work highlighted the ability of DT to address the semantic incompleteness inherent in BIM. West Cambridge demonstrated the successful development of a DT for its campus, showcasing its applicability at both building and city scales
[25]. Thus, the literature on DT applications within the AEC industry has expanded rapidly, with numerous review articles published between 2021 and 2023
[5],
[26],
[27] documenting the latest advancements.
Table 1 [5],
[8],
[24],
[26],
[27],
[28],
[44],
[45] summarizes recent review articles on DT applications in infrastructure, with a significant focus on construction, civil engineering, and transportation. Many review articles have specifically addressed bridge engineering, with emphases on maintenance
[32], digital transformation
[34], and structural health monitoring
[35]. Recent studies have expanded the scope to tunnel
[43] and subway construction
[44]. The fields of highways and urban roads embracing the digital era have garnered significant attention from researchers and institutions
[46],
[47]; however, comprehensive review articles addressing DT implementation within road engineering remain limited. Road engineering faces unique challenges, including large-scale operations, variable environmental conditions, complex structural designs, anisotropic materials, and fluctuating traffic loads. Existing studies have primarily examined progress in civil infrastructure
[30], transportation systems
[29], and specific road network applications
[31], as well as sensing technologies in asphalt pavements
[38]. The most recent article
[45] presents a detailed overview of state-of-the-art DT technologies for road pavements. However, current review articles lack a life cycle perspective, underscoring the necessity of this research.
1.2. Research scopes, contributions, and organization
The primary aim of this study was to analyze the applications of DT technology in road engineering. Specifically, the research investigates how DT can transform road engineering across its entire life cycle by systematically reviewing enabling technologies and practical implementations. The detailed objectives of the study are as follows:
(1) Summarize and analyze the current enabling technologies for DT applications in road engineering.
(2) Review state-of-the-art advancements in utilizing DT across the road engineering life cycle.
(3) Identify existing gaps and propose future directions for DT in road infrastructure.
The contributions of this study are multifaceted. First, it offers a comprehensive analysis of the current state of DT-enabled technologies within road engineering. Second, it examines application scenarios across the road engineering life cycle, highlighting the predominant use of DT in operations and maintenance (O&M) while identifying potential expansion into other stages. Finally, the study proposes a DT framework tailored for road engineering and provides recommendations for future research.
This research adopts an application-focused approach to exploring the development of DT technology in road infrastructure and aims to outline emerging trends in this field. Section 2 details the methodology employed in this review. Section 3 summarizes the related enabling technologies used in the industry. Section 4 reviews the primary DT applications across the road engineering life cycle. Section 5 discusses the opportunities and challenges associated with DT in road engineering. Finally, Section 6 presents the conclusions.
2. Literature review methodology
2.1. Document retrieval process
This study employed two widely recognized citation databases, Web of Science and Scopus, due to their comprehensive coverage of scholarly publications, including peer-reviewed journals, conference proceedings, and other pertinent sources. The search strategy was meticulously designed to ensure the retrieval of relevant literature. The primary query centered on DT, employing the keywords “DT,” “DTs,” and “digital twinning” to capture all related content. To address the diverse terminology used in road engineering, additional keywords such as “pavement,” “road,” “highway,” “expressway,” and “freeway” were incorporated. Searches were restricted to titles, abstracts, and keywords to maximize the relevance of the results. As a result, 1393 and 878 articles were retrieved from Web of Science and Scopus, respectively.
Fig. 3 illustrates the systematic approach employed for selecting relevant articles and excluding irrelevant or duplicate content.
The initial criteria for inclusion were clearly defined. Accepted publication types encompassed journal articles, conference papers, review articles, and book chapters. Excluded materials included patents, early-access publications, editorials, and retractions. Language restrictions were applied, focusing exclusively on English-language publications. The inclusion cut-off date was set as October 31, 2024. After the removal of duplicates from both databases, a total of 1165 unique publications remained.
The relevance of these publications was subsequently assessed through manual screening. Titles and abstracts were carefully reviewed to determine their alignment with the study’s objectives. Conference records that were inadvertently included in the databases were excluded. To ensure the literature review’s relevance, publications from unrelated disciplines were also removed. For instance, studies such as Ref.
[48], which included terms like “DT” and “road” in their titles but pertained to the medical field, were excluded. Similarly, articles with only superficial mentions of DT in their abstracts, without substantive engagement with the topic, were omitted. Following this step, 458 articles remained.
A comprehensive full-text review of the remaining publications was then conducted. Irrelevant documents were eliminated to ensure alignment with the study’s focus. Literature unrelated to pavement infrastructure, such as studies on traffic systems, autonomous driving, transportation vehicles, and roadside infrastructure, was excluded. This rigorous process ensured that the final selection of literature was directly related to road engineering. Ultimately, 80 articles were included in this review.
2.2. Bibliometric analysis
Publication trends and keyword co-occurrence patterns were analyzed to understand the growth and focus areas of DT-related research.
Fig. 4 illustrates the distribution of literature categories and the growth trends of DT-related publications. The number of publications has risen significantly, increasing from a single paper in 2019 to 30 papers in 2024, with the majority consisting of journal articles. The earliest collected article associated DT with lifecycle management of linear infrastructure
[49]. This was followed by a conference article that explored the application of DT to asphalt pavements
[50]. Additionally, some review articles investigated DT applications in transportation infrastructure
[29],
[31] or discussed technologies related to DT
[51],
[52]. The upward trend observed over the past five years underscores the growing importance of DT applications in infrastructure domains.
Bibliometric indicators are instrumental in summarizing and monitoring technological advancements within specialized fields
[53]. Using the analysis software VOSviewer (Centre for Science and Technology Studies, Leiden University, the Netherlands), a keyword knowledge graph relevant to DT in road engineering was constructed.
Fig. 5 illustrates this keyword atlas, where keywords are represented as nodes, and co-occurrence patterns form edges, visually depicting the interconnections among DT-related terms. The resulting map highlights the prominence of pivotal terms in road engineering, such as highway planning, asset management, pavement inspection, road condition prediction, decision-making, and maintenance. Certain keywords also signify the enabling methodologies for DT applications. For example, terms like BIM, point clouds, data integration, machine learning (ML), and artificial intelligence (AI) highlight specific pathways for implementing DT technologies. Emerging areas of research, including ML, AI, deep learning, and automation, have garnered increasing interest, signifying their role as current research hotspots. Notably, the strong association between the keywords “life cycle,” “DT,” and “road” emphasizes the criticality of adopting a lifecycle perspective for road infrastructure management.
Table 2 presents a detailed list of the top 15 keywords, ranked by their frequency of occurrence. This understanding of key terms can facilitate a more structured approach to organizing subsequent chapters.
Fig. 6 illustrates the distribution of research fields for the collected articles, classified according to the subject areas in the Web of Science. Each article spanned multiple topics, encompassing a total of 32 distinct subject areas, with the top 15 areas prominently highlighted. The chart reveals that engineering was the dominant subject area, accounting for 75% of the articles. This was followed by computer science, which appeared in more than 60% of the articles. Transportation and construction technologies were represented in over 50% and 40% of the articles, respectively. These findings underscore the pivotal role of integrating computer science and engineering in advancing DT technologies within road engineering. Transportation and construction technologies, in particular, embody the fundamental engineering characteristics of this field. The results further highlight the interdisciplinary and multi-technological nature of research on DT applications in road engineering.
3. Enabling technologies of DT for road engineering
Based on a 5D DT model
[9], the paradigm was adapted to the specific context of road pavement engineering. In this framework, physical entities correspond to existing road projects, twin data represent the information gathered from real-world road elements, and virtual models denote the creation of digital representations of these roads. Connections primarily involve the processing and interaction of data between real and virtual entities. DT services aim to enhance and optimize real-world road infrastructure across various stages through targeted applications. This section categorizes DT-enabling technologies into three domains: physical-world data acquisition, virtual model creation, and data processing and interaction. The enabling technologies across these domains are summarized, and their characteristics, limitations, and potential future directions are analyzed.
3.1. Physical-world data-acquiring technologies
Data that represent information from the physical world form the foundation for delivering DT services and enabling real-time insights. Physical entities include tangible components such as road structure layers (e.g., pavement layers) and surface texture, which constitute the objective existence of roads in the real world. A comprehensive approach to road engineering necessitates addressing aspects related to structure, materials, traffic, and environmental conditions.
Table 3 [46],
[46],
[54],
[55],
[81],
[82] presents the specific technologies and associated studies utilized for data acquisition in road engineering.
Data acquisition from the physical world forms the cornerstone for developing virtual representations of road infrastructure. Within this framework, the associated technologies are classified into two primary categories: virtual model creation and environmental perception.
Fig. 7 presents a comprehensive overview of these technologies as applied to road engineering. The process of obtaining detailed information about physical entities is described in this article as modeling data technology. Environmental perception technologies, on the other hand, focus on capturing parameters such as project location, real-time temperature, moisture levels, pavement conditions, and surrounding topography.
BIM models generate virtual representations by directly inheriting information from the design and construction phases
[83]. As a shared knowledge resource, BIM enhances the generation and management of information, thereby minimizing the need for repetitive data collection or reformatting
[84]. BIM facilitates data integration by employing a 3D model information database that spans the entire building life cycle, from design and construction to operation and eventual decommissioning. In the context of road engineering, BIM has been demonstrated to provide detailed information for each road element
[54].
Ground-penetrating radar (GPR) is a non-destructive technique that utilizes electromagnetic waves to acquire data without compromising the integrity of pavement structures. 3D GPR systems are capable of capturing internal pavement distress characteristics
[56] and enabling the 3D reconstruction of hidden defects
[85]. GPR-based detection technologies are instrumental in developing DT models for evaluating internal road distress, although research in this domain remains limited. Moreover, GPR facilitates environmental perception by assessing soil and rock properties, mapping geological formations, and monitoring water tables. Despite its significant potential, the full integration of GPR into DT frameworks has yet to be fully realized.
3D scanning involves analyzing physical objects or environments to capture 3D data related to their shape and, in some instances, additional attributes such as color. Point-cloud models derived from 3D scanning provide critical geometric information about existing structures
[56]. Furthermore, point-cloud data serve as valuable inputs for assessing pavement surface distress
[76]. Unlike GPR, which captures subsurface distress, 3D scanning focuses on surface-level distress. Mobile inspection vehicles are typically equipped with advanced scanning devices to facilitate this process
[65].
Unmanned aerial vehicles (UAVs) have become increasingly prominent due to their cost-effectiveness and superior performance compared with traditional land-based platforms requiring human operation. Equipped with sensors such as cameras and laser scanners, UAVs are primarily deployed to inspect infrastructure in environments that are hazardous or otherwise challenging for human surveyors. However, aerial perspectives often introduce data loss due to obstructions such as buildings and trees. Consequently, UAV-collected data frequently require preprocessing
[76] or integration with ground-based acquisition technologies to ensure comprehensive coverage
[79].
Maps serve as graphical representations, typically rendered to scale on a flat surface, that illustrate geographical features and spatial relationships within a specific area. Some studies have proposed leveraging map data to create expressway DT, supplementing physical altitude data with map-derived information
[46]. These studies involved downloading essential data, including road geographic locations, aerial imagery, DSM, and DTM, from online map databases to calibrate horizontal alignments, vertical alignments, and cross-sections. However, the applicability of this method is limited in regions where reliable map data services are unavailable.
Fig. 8 provides an overview of these modeling data technologies, which can be categorized into three purposes: geometric information, road internal structure, and additional information.
A geographic information system (GIS) is a computer system designed to create, manage, analyze, and map various types of data. GIS technology is capable of simultaneously collecting spatial data and performing geocoding and decoding. By integrating location-based information with descriptive data, GIS provides essential location-based analytics for obtaining geographic information relevant to projects
[86], particularly in macrotwin models
[54].
Sensors play a critical role in data perception within DT frameworks. A sensor is a device that responds to physical stimuli, such as heat, light, sound, pressure, magnetism, or motion, and generates an output signal based on the detected phenomenon. Real-time sensors are indispensable for capturing messages in DT. Examples include weigh-in-motion and tire pressure sensors for collecting traffic data, as well as temperature and moisture sensors embedded in pavement layers for capturing environmental data
[64]. Sensors can be utilized in diverse scenarios to acquire vital information, including road construction monitoring
[57], road geometric assessments
[69], and pavement crack detection
[65]. Moreover, self-sensing asphalt pavements have been explored, where embedded sensors facilitate automatic data perception, offering a promising avenue for enhanced infrastructure monitoring
[51].
Photography involves capturing images of physical objects using cameras or motion picture devices through the film’s sensitive response to light. This technique provides visual information regarding surface geometry, texture, and condition, contributing significantly to the perception of road surface health
[77]. On-board cameras are frequently employed in practical applications to support such analyses.
GNSS is an interplanetary radio navigation system that employs satellites as navigation stations. These satellites, equipped with specialized cameras and sensors, capture high-resolution images to enhance real-time positioning and ensure precision at specific road project locations
[58]. Liu and Ma
[82] proposed a method for positioning highway mileage markers by integrating a monocular camera with GNSS within a highway DT mapping system.
Fig. 9 overviews these environment perception technologies and divides their purposes into four categories: environment data, traffic data, geographic information, and road distress information.
3.2. Virtual models-creating technologies
Virtual entities serve as digital counterparts of physical entities within a DT environment. The adaptation of techniques from the AEC industry to road engineering requires specific technical considerations for creating virtual models.
Table 4 [38],
[46],
[50],
[54],
[55],
[57],
[81],
[85],
[87],
[90] lists the methodologies commonly employed in the creation of virtual models in road engineering.
BIM has emerged as the most widely utilized modeling approach in the reviewed studies, primarily relying on tools such as Civil 3D and Revit. The image-based method employs UAV photogrammetry and the SfM program, while the point-cloud model is derived from 3D scanning conducted using either aerial or terrestrial vehicles. Pan et al.
[70] introduced a scan-to-graph methodology to generate a graph-representation DT of highway segments from point-cloud data at a more granular level. GPR technology has proven effective in detecting and digitizing road structure defects
[85]. Fusion modeling integrates multiple methods to improve modeling accuracy. For instance, the combination of BIM and GIS facilitates the creation of multiscale DT geometries, effectively integrating macrotwin and microtwin representations
[91]. By incorporating dynamic data from sensors, cameras, and the IoT, Tang et al.
[61] enhanced the perception accuracy, especially in challenging scenarios such as mountain highways. The finite element method (FEM) is closely aligned with twinning mechanisms
[52]. Additionally, an unreal engine (UE) was employed to render pavement distress models
[76]. Maps serve as platforms for online modeling data, enabling the seamless integration of diverse information sources
[59].
Essentially, virtual models extend beyond simple geometric representations. These models encompass geometric, physical, behavioral, and rule-based components, collectively bridging the physical and digital domains
[9]. Together, these elements create a comprehensive digital representation of physical road systems. However, most studies have not fully incorporated all four model types. Instead, the majority have concentrated on geometric models, with some research focusing on physical models, such as FEM simulations
[92].
Fig. 10 presents a framework outlining enabling technologies for virtual modeling in road engineering.
The geometric model serves as the foundational component of a DT model, as it provides the essential spatial framework for positioning other relevant information. Technologies such as BIM, GIS, 3D scanning, and photogrammetry are commonly employed for geometric modeling. BIM is frequently used to create virtual models tailored for DT applications. For large-scale projects, including city infrastructure and road networks, GIS offers significant advantages in supporting model creation. Furthermore, point-cloud data and oblique photography play a critical role in developing detailed 3D pavement body models. Davletshina et al.
[88] introduced an automated method for constructing road DT geometries from point-cloud data. Their research also proposed a methodology for detecting and integrating geometric changes into road DT
[89]. The physical model was mainly simulated and displayed through FEM, GPR, numerical simulation, and additional attribute information. Shen and Wang
[64] developed an asphalt pavement modeling software based on semianalytical FEM to compute viscoelastic pavement responses under vehicle–tire–pavement interactions. GPR technology captures road internal structural signals and plays a crucial role in constructing a comprehensive road structure DT with detailed interior physical representations
[56]. Behavior models capture the dynamic behavior of physical entities, including their responses to environmental changes and the operational mechanisms of human activities
[36]. In road engineering, the behavioral model encompasses the progression of various degradation patterns such as crack development, rut formation, and pothole creation. The rule model governs behavior by setting relevant rules and procedures built using contextual normative systems and computer programs
[36]. AI models serve as common rule models in the AEC industry, as exemplified by the multiple time-series stacking (MTSS) model for pavement performance prediction
[67].
3.3. Data processing and interaction technologies
The interconnections among system components are essential for facilitating the seamless flow and interaction of data across various technologies and domains. Nonetheless, achieving the desired biconnective characteristics entails addressing the challenge of effectively linking and analyzing relevant information to establish meaningful connections with the physical world.
Table 5 [44],
[47],
[54],
[56],
[83],
[84],
[85],
[86] outlines the techniques employed for data processing and interaction in road engineering.
Siddiqa et al.
[97] emphasized that a comprehensive data lifecycle encompasses data collection, transmission, storage, pre-processing, analysis, and application. Accordingly,
Fig. 11 identifies four distinct levels of data processing and interaction—data transmission, storage, analysis, and presentation—in the context of DT in road engineering.
The initial stage is data transmission, which relies on wireless equipment such as laser scanner vehicles
[98] and portable video cameras
[73] to facilitate data transfer. Additionally, sensors play a vital role in establishing data communication in specific scenarios
[65]. IoT cloud platforms are widely used for data transmission and storage
[58], with the open-source platform FIWARE offering scalability and the ability to manage large volumes of data and sensor interactions, making it highly suitable for constructing road system DT
[93]. Another critical component is the software interface. For instance, lightweight physical engines can seamlessly integrate images
[76], and 3D BIM information can be transmitted across web servers, platforms, or applications
[68]. Furthermore, edge computing has demonstrated its value in reducing data transmission loads
[59].
The second stage involves data storage, which is indispensable for managing complex DT models. Depending on the intended application, processed data may be stored locally or in the cloud, ensuring both accessibility and analytical utility. Local storage typically employs BIM applications. For example, Appelt
[86] employed a 5D BIM system to consolidate construction information in Germany, while Cepa et al.
[54] successfully integrated BIM and GIS, subsequently uploading the resulting database to an intelligent management platform. Chang et al.
[99] used map-based geospatial software combined with high-precision GNSS data for intelligent road construction. Cloud-based data storage, on the other hand, offers enhanced convenience, as exemplified by its application in urban DT datasets for road inspection competitions
[100]. Consilvio et al.
[68] utilized an IoT database and a cloud platform to store 3D models.
The third stage, data analysis, is critical for extracting the full value of DT data. Intelligent algorithms and ML are among the most prevalent technologies employed. For example, Cao et al.
[65] implemented a segmentation algorithm to detect pavement cracks. Consilvio et al.
[68] used ML and computer-vision algorithms to segment point clouds. Pan et al.
[70] used deep learning techniques to segment pavements into carriageways and identify lane marking points. Sierra et al.
[77] proposed a pavement-cognitive DT that utilized advanced ML techniques for health monitoring. In terms of analysis platform, GIS provides an integrated environment for digital model processing and data analysis
[47]. Anantheswar et al.
[92] introduced an arbitrary Lagrange–Eulerian (ALE) framework for FEM applications. Liu et al.
[96] introduced an MC-DCGAN model for forecasting tire-pavement contact stresses. This model demonstrated greater computational efficiency while maintaining almost predictive accuracy comparable to FEM simulations.
The final stage is data presentation, which visually connects the DT to the physical world. Various methods, including graphs, charts, 3D models, and other visual representations, can be employed for this purpose. Numerous studies have relied on two-dimensional (2D) visual platforms, predominantly supported by BIM
[37],
[42]. Additionally, 3D engines such as UE4
[44] and UE5
[81] are frequently utilized to render road DT visualizations. For example, real-time simulation results of pavement distress can be directly displayed using a 3D engine
[65]. Charts provide an effective medium for conveying specialized information, such as carbon footprint assessments
[81].
Fig. 12 illustrates three common methods for displaying DT results.
3.4. Analysis of current enabling technologies
DT technology in road engineering is supported by a suite of enabling technologies, each characterized by distinct features and limitations. While these technologies are not uniquely specific to DT applications, they are indispensable for the development and ongoing maintenance of DT systems.
Table 6 provides a summary of the characteristics and limitations associated with these techniques.
In the process of acquiring data from the physical world, BIM and maps are utilized to integrate existing data effectively; however, these technologies are associated with high initial investment and maintenance costs and are less effective when the data are incomplete. D'Amico et al.
[101] applied BIM to obtain geometric information, while map data such as Digimaps provide insights into DSM, DTM, aerial photographs, and topography
[80]. GIS and GNSS supply essential geospatial and coordinate data, although their accuracy depends on satellite positioning. Methods including cameras, UAVs, and 3D scanning are valuable for acquiring detailed road surface texture and condition data but are constrained by specific environmental and equipment precision requirements. Sensors embedded in road structures offer real-time data on load and environmental conditions; however, their susceptibility to damage in harsh conditions and under heavy traffic limits their portability and durability.
In virtual model construction, BIM excels in integrating diverse datasets into informational models. GPR facilitates the creation of detailed internal structure models; however, it requires expert knowledge, complex data processing, and interpretation. The FEM is capable of simulating and forecasting outcomes based on material characteristics and structural parameters. UE is particularly effective in creating and rendering 3D virtual models. Fusion modeling is a more potent approach; however, it poses significant challenges in integrating multisource data. Researchers have explored different combinations of these methods, including BIM, GIS
[54], 3D laser scanning, and photography
[102]. Yang et al.
[90] proposed a new BIM + GIS framework for integrating pavement defects and modeling as well as transforming GPR data into distress models. D'Amico et al.
[101] proposed a novel approach for integrating GPR, mobile laser scanning (MLS), and synthetic aperture radar (SAR) data into a BIM to create a road DT prototype. However, despite these advancements, achieving seamless and comprehensive integration across geometric, physical, behavioral, and rule-based models within the DT framework remains challenging.
Data processing and interaction technologies are equally diverse. BIM serves as an information-management platform but faces challenges in terms of standardization and data security. GIS is a key platform for data visualization and analysis; however, its data integration is complex. The IoT enhances connectivity and automation but relies heavily on network connectivity and requires robust security measures. 3D engines provide high-quality visual environments but require high computer specifications. Although ML can learn from large datasets, it requires substantial-quality data and suffers from a lack of interpretability. Edge computing offers a scalable and cost-effective approach to data processing but must address data consistency issues.
Current studies predominantly focus on creating virtual–physical models and perceiving pavement surface conditions, lacking mechanism models and real-time connections in DT applications. Compared with cutting-edge research on other road infrastructures, the enabling technologies employed in road engineering require further innovation. In this regard, Gao et al.
[103] proposed a DT communication framework based on artificial intelligence of things (AIoT) to address time delays in bridge DT services. Lin et al.
[104] developed a bridge DT using three distinct FEM models for seismic collapse assessment of long-span cable-stayed bridges. In addition, Yu et al.
[105] leveraged an extended construction operation building information exchange (COBie) standard for tunnel twin data organization by integrating semantic web technologies to achieve data fusion. Despite these advancements, road DT technology remains insufficiently developed in terms of physical, behavioral, and rule models, primarily due to the intricate nature of pavement structures and material non-uniformity. These limitations result in inadequate data processing and connection-enabling technologies. As interest in DT technology grows, there is considerable potential for the development of related enabling technologies. Standardization efforts are crucial to facilitate the effective creation of DT from modular components. Such standardization will enable seamless integration of DT implementations in road engineering. Furthermore, these efforts will contribute to enhancing the overall functionality and scalability of DT implementations in road engineering and regulating the systematic application process of advanced enabling technologies.
4. Applications of DT at various life cycle phases of road engineering
The life cycle of road engineering is typically divided into four distinct phases: construction, use, maintenance, and end-of-life
[106], offering a transformative solution by integrating information across all stages of a product’s lifecycle. For instance, Tchana et al.
[49] proposed the development of a unique DT specifically tailored for the life-cycle management of linear infrastructure. Aligning with the concept of the “life cycle” illustrated in
Fig. 5, this study explores DT applications across four distinct phases: planning and design, construction and completion, operation and maintenance, and demolition and reconstruction.
4.1. Planning and design
Road planning and design encompass critical tasks such as route selection, pavement design, and budgeting. In this phase, DT effectively integrates engineering data with external environmental conditions, particularly at the macroscale level.
Fig. 13 illustrates the three main scenarios of DT during this phase: urban road planning, highway geometric design, and smart road framework development.
In the context of energy conservation and emission reduction, road planning requires a multidisciplinary approach to achieve optimal solutions. An innovative urban road planning method based on the DT-MCDM-GIS framework (MCDM refers to multi-criterial decision making), which incorporates DT, MCDM, and GIS, considers numerous factors such as land use and traffic congestion
[47]. This framework was successfully applied to urban road planning in Bromley, UK
[107]. DT provides data and processed information related to existing environments, surroundings, and relevant projects in this framework. Furthermore, Maserrat et al.
[108] proposed a Dempster–Shafer enhanced framework based on the DT-MCDM-GIS framework, assisting urban road planning decision-makers by identifying and evaluating alignments for road widening or new construction.
BIM further facilitates the adoption of DT during the design phase of road engineering projects
[28]. Wu et al.
[55] demonstrated this by applying BIM and DT to optimize the design and construction of a frame bridge. By leveraging physical models, sensor updates, and operational history, they optimized the construction scheme and mitigated potential collisions between the frame bridge and road infrastructure.
In road engineering, the integration of DT technology into the planning and design phases of an advanced road concept is indicative of its transformative potential. The concept of smart roads exemplifies this potential, as it promises to revolutionize transportation systems and road infrastructure by incorporating advanced technologies, including DT. Smart roads equipped with intelligent IoT systems enhance vehicular environmental awareness and improve decision-making capabilities within intelligent transportation frameworks
[109]. By employing networks of sensors and actuators, smart roads effectively manage traffic by delivering real-time information
[110]. Sun et al.
[111] reported a precise definition of smart roads, characterizing them as a road infrastructure seamlessly integrated with advanced network and communication technologies, where the infrastructure serves as the core component of smart roads. In the context of DT, Fu et al.
[112] introduced a three-layer conceptual framework for smart freeways based on DT, while Wang et al.
[40] outlined key research challenges in advancing DT-driven smart transportation infrastructure. Mao et al.
[113] proposed a road map for smart roads supported by IoT. Both DT and smart road concepts represent a visionary outlook for the future of transportation. However, their practical implementation remains challenging in the real world and is currently confined to the planning and design stages.
4.2. Construction and completion
The construction phase is a multifaceted process encompassing various stages, from initial arrangements to completion. Ensuring the quality, safety, and durability of road infrastructure is critical to delivering reliable pavement systems. The primary focus during this phase lies in monitoring construction quality and progress.
Fig. 14 illustrates three key applications of DT in this phase: construction management, completion model, and smart highway case.
During construction, DT technology supports resource allocation, material management, schedule coordination, and quality assurance, as demonstrated in the video link
[114]. Recent studies have integrated BIM with sensor data to leverage dynamic data querying, particularly when employing secondary raw materials for road construction
[57]. BIM enables the digital replication of construction projects, while sensors provide real-time updates, ensuring that DT models remain current. Additionally, geospatial tools such as Terra Veta (Transtec Group, Inc., USA) and GNSS have been used to integrate diverse construction data within the DT framework during pavement construction
[99]. In a pilot project conducted in Indonesia, maps and GPS-camera-based updates facilitated efficient road construction and management
[59]. Ellul et al.
[94] further highlighted the potential of DT for health, safety, and construction progress monitoring through case studies in highway construction projects.
The quality of road surfaces significantly influences driving comfort, necessitating precise and efficient pavement compaction. Intelligent compaction (IC) technology enhances quality standards while reducing costs; however, its primary application is real-time quality inspection within vehicle systems, limiting its use in advanced construction controls. Han et al.
[58] proposed an integrated framework combining BIM, IC, and IoT to establish a foundational DT platform for construction quality management. In this framework, BIM provides high-fidelity virtual representations of road infrastructure, facilitating data integration, visualization, and construction management. IC contributes compaction quality monitoring and control data, while IoT connects physical and digital assets by capturing real-time data from on-site operations.
In Germany, the A7 motorway widening project employed a 5D-BIM model that included a DT completion model
[86]. These as-built DT models significantly aid stakeholders during the O&M phase by incorporating 3D models, four-dimensional (4D) networked logistics and time management, and 5D quantities and cost/revenue analytics. The DT also forms the basis of an as-built model linked to the client’s maintenance system, ensuring traceability and manageability of asset information throughout the maintenance period.
Smart road construction provides another compelling application of DT. Within the context of smart roads, DT functions as an integrated representation of physical infrastructure and IoT-generated data, enabling the monitoring of vehicle driving states and surrounding road conditions. By combining IoT, digital maps, and road construction information models, Mao et al.
[113] developed two DT systems: one focused on road traffic and the other on road health monitoring and asset management. The latter system incorporates data from IoT devices, such as temperature, humidity, light, and vibration measurements, along with lane-level vehicle data. Thonhofer et al.
[72] created a collaborative, connected, and automated mobility decision-support DT platform for smart roads. By integrating IoT, semantic web technologies, cloud computing, and databases, this platform supports asset management, maintenance, and traffic management and control. Marai et al.
[71] produced a DT Box with a 360° camera, GPS device, and Internet dongle for connectivity. This device provides image, location, and environmental data on road infrastructure, enabling the creation of road asset DT models tailored to smart city environments.
4.3. Operation and maintenance
In the O&M phases, managing and accessing project data become increasingly challenging due to the complexities of road-serving environments. As the longest stage of a project’s lifecycle, ensuring the reliability and sustainability of pavement structures is of paramount importance.
Fig. 15 illustrates three primary applications of DT in this phase: monitoring pavement health conditions, asset operation management, and maintenance decision-making support.
4.3.1. Pavement conditions monitoring
Road operation and maintenance primarily focus on pavement distress detection, structural health evaluation, and informed maintenance decision-making. The identification of pavement distress characteristics forms the foundation of these activities. Researchers have sought to develop comprehensive DT models capable of capturing all distress data, including both surface and interior aspects.
A 3D model of the road surface, generated from UAV-captured images and SfM technology, has proven effective in identifying various pavement distresses and capturing the features of roadside facilities
[78]. Such DT models also integrate road distress data for enhanced monitoring
[66]. Additionally, Sierra et al.
[77] developed a cognitive DT specifically for pavements, which can accurately represent current conditions and detect pavement distress. Unlike standard DT models, cognitive DT are augmented with cognitive capabilities, enabling autonomous activities based on semantically interlinked digital models
[115]. In an urban DT-based intelligent road inspection competition, a DT dataset-based road inspection system was introduced to encourage the development of DT applications
[100]. To enhance recognition algorithms, a DT virtual data enhancement method based on rendering lightweight physics engines has been proposed to enrich crack datasets
[76]. Gooneratne et al.
[116] proposed a novel approach for monitoring road conditions by analyzing data from smartphones placed inside the backpacks of cyclists toward future long-term DT applications. However, these studies primarily focus on static condition monitoring and are not readily applicable to real-time DT implementations.
In contrast to the conventional methods used for monitoring pavement conditions, DT is designed for real-time condition monitoring and display based on advanced enabling technologies, particularly communication tools. Cao et al.
[65] proposed a multisensor data communication DT system for pavement crack detection. Utilizing an edge-based 3D crack segmentation algorithm, this system effectively identified pavement cracks after data collection and communication, thus advancing toward real-time monitoring, as shown in
Fig. 16. Mahmudah et al.
[117] investigated the application of edge AI and deep learning techniques for real-time road damage detection within the DT city framework, highlighting both the challenges and potential of deploying such systems. Heravi et al.
[118] proposed an edge AI-based remote road fixture monitoring system using small devices installed in personal vehicles. Furthermore, Barisic et al.
[60] established a thermal DT system based on temperature measurements from sensors installed in several highway sections, demonstrating its capability to continuously monitor asphalt pavement conditions.
To address internal road defects, DT models are often combined with non-destructive testing (NDT) technologies such as GPR. A 3D digital model constructed from GPR-measured data enables the identification of internal defects within road structures, thereby enhancing the fidelity of DT representations
[56]. Furthermore, NDT techniques like GPR are considered integral to future pavement management systems (PMSs), facilitating integration with digital pavement models such as BIM to produce comprehensive DT models
[101],
[119]. Despite these advancements, integrating a unified DT model that seamlessly combines surface and internal road conditions remains a significant challenge, primarily due to the complexity of interpreting GPR data and other related factors.
4.3.2. Road assets operation management
In the construction industry, the overarching goals of DT and asset management remain aligned, with the focus on utilizing digital asset information to enhance asset performance throughout the entire lifecycle. Specifically, in road infrastructure, roads function not only as critical operational components but also as fundamental assets. Road DT serves as the primary digital replica of the road infrastructure, effectively facilitating road asset operation management.
The adoption of DT in road infrastructure management is gaining traction among state transportation departments in the United States. For example, DT was initially utilized for data collection and governance in Utah, USA
[120]. The department developed a critical platform for DT, emphasizing its potential impact on asset management
[121]. While DT holds promise for road asset management, Vieira et al.
[122] highlighted its operational limitations. To explore the potential of DT further in physical asset management, they proposed a conceptual architecture
[37] and a value-based analysis method
[123]. Ammar et al.
[124] developed a roadmap for adopting DT in transportation asset data management and presented a case study focusing on guide hardware. Kodikara et al.
[125] regarded DT as an essential tool for the smart asset management of road pavements. Collectively, these studies underscore the growing potential of DT in this domain.
Road asset management primarily focuses on visual information during the operational phase. For instance, a campus road asset inventory was captured using a portable camera mounted on a vehicle, and the data were subsequently integrated into a DT model
[73]. Similarly, asset management of the metropolitan expressway in Japan relies on a smart infrastructure management system
[98], where 3D point-cloud data serve as the core DT technology. Indonesian researchers proposed a DT model for advanced asset management analysis and decision-making by meticulously analyzing the technical framework
[126]. Madrid, Spain, developed a semi-automated model based on common road GIS management
[54]. By integrating BIM and GIS, the DT model is transmitted to an intelligent management platform along with an external database to facilitate the visualization of available documents. In summary, the DT model effectively mirrors physical entities and stores actual operational status data recorded by road infrastructure, offering valuable support to facility managers in their decision-making processes.
4.3.3. Prediction and decision-making for maintenance
During the operational phase, minor damage is inevitable owing to environmental impacts and the aging of materials. Consequently, maintenance decision-making becomes essential to strike a balance between pavement conditions and repair costs. DT has applications through two distinct approaches: predicting pavement performance and supporting decision-making processes.
The DT model predicts the future state of physical entities. Ye et al.
[52] considered the DT as a physics-based FEM model for mechanistic and quantitative descriptions to predict bridge conditions. Yu et al.
[67] used the DT to predict the performance of a highway tunnel pavement. By digital mapping virtual BIM models to physical highways and using the MTSS algorithm, they accurately predicted the international roughness index (IRI) of the pavement by comparing the predicted values with actual measurements. Chen et al.
[95] demonstrated the effectiveness of ML approaches for predicting pavement performance states in DT road scenarios.
The aforementioned DT studies rely on data-driven performance predictions. However, physical-model-based prediction, such as finite element (FE) analysis, brings predictions closer to the essence of DT. Anantheswar et al.
[92] proposed a dynamic ALE framework for FE simulations specifically targeting pavement structures. This approach proved valuable within the context of DT, as it enabled simulations that could determine relevant parameters that align with future performance over time. Chen et al.
[127] integrated physics-based FE simulation data into an ML model, thereby enhancing prediction confidence. Ideally, DT should continuously collect real-time data from pavements and accurately predict their performance using appropriate physical models and mechanical paradigms. Although research and applications of DT are still in their early stages, combining DT with simulation methods holds significant promise for future predictions.
In road maintenance decision-making,
Fig. 17 illustrates the general architecture of the DT-based decision supporting tool
[68]. This system seamlessly integrates pavement data, including the IRI and sideways force coefficient, offering additional support functions. The system has demonstrated the potential to reduce intervention frequency by 10% and maintenance costs by 12% on the A24 Italian motorway. Lei et al.
[128] introduced a DT-based highway health maintenance system utilizing Kalman filtering for data denoising, thereby improving decision-making by accurately predicting damage types and trends. Considering the environmental impact of carbon emissions, Fang et al.
[81] developed a highway carbon footprint DT visual platform combined with a lightweight carbon footprint data collection method, and accurately quantified carbon emissions during highway operation and reconstruction. Zhu et al.
[129] explored the application of DT to address urban road defects caused by snow and water accumulation. They utilized real-time monitoring, early warning systems, and enhanced maintenance decision-making, ultimately enhancing road safety and operational efficiency. Yin et al.
[130] surveyed 183 highway professionals on the challenges and opportunities of DT in maintenance, revealing that complex decision-making, poor asset data quality, and incomplete road condition information are the primary inefficiencies in DT-based highway upkeep.
4.4. Demolition and reconstruction
At the end of the service cycle, demolition and reconstruction of roads are necessary to ensure sustainability. However, valuable knowledge about the behavior of a system or product is often lost during the demolition phase
[131]. DT technology provides a solution by retaining the entire lifecycle information of physical entities in a virtual space at minimal cost. In the road industry, the demolition phase tends to receive less attention in terms of DT applications than the renewal and reconstruction phases.
This review identified three limited studies concerning DT in the road reconstruction and expansion phases. Jiang et al.
[46] employed a DT method to assess the clearance of underpass roads during the widening of the main road. By constructing a BIM-based DT model of an underpass linked to online map data, they streamlined clearance inspections and underpass road redesigns without the need for field surveys. In a road-widening project in Germany, Appelt
[86] dynamically updated the actual construction model to allow direct visualization of construction logistics, progress, and quantities. Real-time DT facilitates fact-based, efficient management decisions. Additionally, the highway carbon footprint DT platform
[81] mentioned earlier was also applicable to decision support for highway reconstruction and expansion.
5. Discussion: Challenges and opportunities
5.1. The understanding of DT
DT technology in road engineering remains at a nascent stage. Despite its growing presence in the literature, its interpretations vary significantly. Some studies have underutilized the potential of DT technology or confused it with other technologies, such as BIM, in the built environment
[23]. Our study incorporates various examples, provides an explicit definition of DT, and refers to a 5D DT model
[9] to elucidate enabling technologies. However, a detailed understanding of DT in road engineering contexts remains scarce, diminishing the persuasiveness of its implementation. Although certain studies have proposed future visions for DT based on these methods
[56],
[58],
[62], the varying levels of DT knowledge among scholars have resulted in imbalances in its application, including a lack of motivation, different technological apprehensions, and implementation challenges. An urgent need exists to unify and refine the understanding of DT in road engineering to facilitate more effective applications. DT is defined herein as a framework encompassing digital models, twin data, and biconnections at various levels, while acknowledging current limitations in establishing dynamic, interactive two-way connections.
The literature reveals that DT can be built using diverse technologies. In road engineering, there are no rigid standards for building DT systems, employing tools, or applying different types of data. Similar to the findings in Refs.
[132],
[68], most DT frameworks adopt a four-layer structure: physical, database, server, and application layers. Despite differences in interpretation, a consensus on the general system structure has emerged in existing studies.
Fig. 18 illustrates the formulation of a DT framework for road engineering, which encompasses four aspects: data acquisition, model creation, interaction, and DT services. The initial step involves transforming the physical pavement entities into their virtual data counterparts. Relatively enabling technologies include 3D scanning, photography, sensors, BIM, GIS, GPR, maps, and satellite imagery. Virtual model creation encompasses various types of models, including geometric, physical, behavioral, and rule models, ensuring twinning characteristics. Interaction processes involve data transmission, storage, analysis, and visualization of results. DT services provide valuable applications throughout the road lifecycle, including design, construction, operation, maintenance, reconstruction, and demolition.
In the DT technology domain, various countries, including China, have released a suite of standard documents, as summarized in
Table 7 [133–140]. These documents, issued by different institutions, vary in their focus, addressing aspects such as the visualization of DT elements
[133], data representation and operability
[134], conceptual frameworks and terminology
[135], and overarching requirements
[136]. The International Organization for Standardization (ISO) series standards concentrate particularly on the implementation of DT in the manufacturing sector, detailing the general principles
[137], reference architecture
[138], digital representation
[139], and information exchange
[140] for an automation integrated system framework within the industry. However, for the construction sector and road engineering in particular, there remains a notable absence of more detailed DT-related specifications to guide the industry.
In summary, DT remains a cognitive disunity of the academic community in terms of understanding and lacks a unified standard to constrain the application path of all walks of life. Although numerous attempts and published papers have contributed to the practical applications and industrial contributions of DT, there is an urgent need for a common DT practice architecture. This architecture would not only enhance the collective understanding of DT but also facilitates the replication and learning of case studies. To this end, we advocate increased standardization of architectural patterns, communication protocols, and data storage solutions. Such transparency would significantly improve the industry’s understanding of DT architectures and their applications, paving the way for more informed and effective DT strategies.
5.2. Key enabling technologies
DT technology represents a transformative paradigm that cannot be achieved through a single technical tool. Across various domains, such as manufacturing and construction, significant differences exist in DT-enabling technologies. Although previous studies have thoroughly explored DT-enabling technologies within construction
[33], particularly focusing on the O&M phases
[36], the road sector has received comparatively less attention.
Fig. 19 presents the blueprint for current and future key enabling technologies for DT in road engineering.
The tools employed for digital pavement modeling lack the diversity observed in the construction sector. Common tools for virtual modeling include BIM, GIS, GPR, 3D scanning, and photography, with GPR being uniquely suited for internal pavement modeling. Advancements in integrating surface and internal models are urgently required to achieve a comprehensive road DT, which heavily depends on further development of GPR. Combining internal digital models with external surface data, future digital twinning enabled by GPR and laser scanning will yield more accurate monitoring and predictive capabilities. Automated modeling of road surface point clouds and rapid interpretation of GPR data are essential for establishing full-field DT models. To support these advancements, sophisticated ML algorithms must be developed for automated and precise data processing. Moreover, developing physical, rule-based, and behavioral models for road structures will enable advanced multiscale modeling. Future research should prioritize collaborative development of macroscopic behavioral models and microphysical models, accompanied by corresponding rule models.
Potential for development also lies in data acquisition and interaction technologies. Although sensing and measurement technologies have been extensively mentioned, the integration and fusion of advanced techniques remain underutilized. IoT technology is expected to integrate various sensing and measurement methods, facilitating DT development in road engineering
[45]. Data collected by autonomous vehicles demonstrates significant potential for structural health monitoring of roadway bridges
[141]. As a cost-effective NDT method, autonomous vehicles enable data collection across extensive road networks. Given the complexity of pavement structures, the fusion of twin data from multiple sources (e.g., devices, sensors, and systems) remains a challenge during data processing. Multisensor integration and advanced ML algorithms offer promising solutions. Additionally, existing research often lacks comprehensive processes for data interaction, spanning from data transmission to visual representation. Cloud-based solutions, combined with distributed approaches such as edge computing, could enhance data interaction capabilities. Real-time interoperability, a core characteristic of DT, presents significant research gaps in road engineering and requires focused efforts to address associated technical challenges. Emerging technologies such as the fifth-generation mobile communication technology (5G) and 6G for real-time interactions, generative AI for simplifying human–DT interaction, and XR for intuitive presentations promise to further advance DT services.
5.3. Future potential application
A review of existing studies reveals that DT can be applied throughout nearly the entire lifecycle of road engineering, albeit with varying frequencies across different stages.
Fig. 20 categorizes DT applications into six primary areas:
(1) Road design support: DT optimizes road planning and design by integrating diverse data sources and simulation scenarios. Relevant studies include urban road planning methods
[47], road geometric design
[55], and smart road frameworks
[112].
(2) Construction management: DT enhances construction processes by providing insights into progress, quality, and resource allocation. Nine studies were conducted on road construction site monitoring
[57], completion model provision
[86], and specific smart road cases
[71].
(3) Pavement condition monitoring: DT facilitates the real-time monitoring of pavement conditions, including distress detection and performance evaluation. Fourteen studies have focused on monitoring pavement surfaces
[66] or interior conditions
[56] using static images
[77] or real-time sensors
[60].
(4) Asset operation management: DT assists in managing road assets efficiently, ensuring optimal utilization and maintenance. Eleven relevant studies have explored the use of DT as a tool for infrastructure asset management at the theoretical level
[122] or in practical applications
[73].
(5) Maintenance support: DT guide maintenance decision-making by predicting deterioration and recommending timely interventions. Ten related studies have been identified in this article, highlighting the use of ML tailored for road DT
[95], FEM for pavement DT
[92], and a DT-based decision-supporting tool
[68].
(6) Reconstruction assistance: DT streamlines reconstruction efforts, minimizes disruptions, and improves overall efficiency. Just three articles were examined in this phase. Two of them provided the necessary information for road widening projects
[46],
[86], while one offered a highway carbon footprint for reconstruction and expansion
[81].
Current studies predominantly concentrate on the operational and maintenance stages, leaving applications in the design, construction, and demolition phases relatively underexplored. Furthermore, most existing applications are small-scale and fail to integrate into operational system environments. To enable the scalability of larger DT applications across their entire life cycle, standardized development and application processes are indispensable. The varying maturity levels of DT reflect a broad spectrum of challenges. In addition to the previously mentioned limitations concerning the understanding of DT and its technical tools, obstacles such as investment complexities, data privacy issues, security risks, and stakeholder trust also hinder progress. Focusing on DT development during the planning and design stages of road engineering can significantly enhance its utility and accessibility in subsequent stages. Moreover, it can help address data privacy concerns, strengthen security measures, and build stakeholder trust. The financial implications of DT projects are often overlooked in the current body of research. It is vital to assess the adequacy of DT across entire road networks and compare its economic benefits to project costs. Researchers should intensify their evaluation of DT applications by analyzing associated costs and benefits, with an emphasis on efficiency and minimal resource consumption. If lightweight DT demonstrates superior returns during application, its development is likely to accelerate. As DT prototypes undergo further validation and maturity levels increase, their potential applications, particularly in the underexplored demolition phase, are expected to expand. Although this phase is frequently neglected, future research is likely to uncover its potential in land planning and material utilization post-road demolition. A life-cycle perspective reveals an additional application in the production of raw materials, which can significantly reduce carbon emissions
[57]. Investigating the repurposing of materials from road demolition during the construction phase using DT presents an opportunity to connect demolition and construction stages, thereby promoting the sustainability of the entire life cycle.
Another promising avenue for DT is the development of smart roads, where DT serves as a central component. Most studies position DT as a core element of smart road architecture and explore its implementation through case studies. The DT road system and the smart road concept both exemplify the trajectory of road infrastructure development in the context of digital transformation, sharing key attributes such as informatization and data interconnectivity. However, many existing studies on smart roads, particularly those focused on autonomous driving, have neglected the pavement aspect
[113]. A future vision of smart roads extends beyond the IoT systems that enable vehicle–road coordination. It necessitates embedding intelligence within road infrastructure, a concept central to DT in road engineering. As a pioneering concept in road infrastructure, smart roads have the potential to integrate and amplify the advantages of advanced technologies. DT has the capacity to revolutionize road engineering within smart roads, but its successful adoption and implementation require a multifaceted approach. Achieving alignment between the overarching architecture and foundational applications of smart roads and DT is essential. This alignment hinges on policy support, technical advancements, pilot road projects, and user feedback, all of which are critical for the ongoing refinement and adaptation of DT within the smart road ecosystem.
6. Conclusions
By analyzing 80 collected articles, this study confirms the significant potential of DT-enabling technologies to advance road engineering. Despite its relatively brief development, DT-enabled technologies in road engineering encompass nearly all processes involved in constructing a DT, including virtual model building, real-time data perception, and data connection and interaction. However, current research predominantly focuses on the development of virtual geometry models and the acquisition of real-world data. The realization of a 5D DT concept in road engineering remains unachieved. Notably, the incorporation of related physical theories, mechanical rules, and behavioral codes, particularly for constructing multiscale and high-fidelity models, has been scarcely addressed in existing studies. Furthermore, the complexity of pavement structures and the inherent non-uniformity of materials pose significant challenges for data processing and connection-enabling technologies. Compared to data acquisition and model creation, data processing and interaction methods are frequently overlooked or insufficiently elaborated, potentially resulting in incomplete DT implementations.
The application of DT technology extends across the entire life cycle of road engineering. Related application scenarios are categorized into six situations spanning four dimensions, with the primary focus on the O&M stages. During the operational phase, DT monitors pavement conditions, supports asset operation management, and aids predictive decision-making for maintenance. Its primary application lies in pavement condition monitoring, while service performance prediction lacks robust simulation support grounded in real-time detection and monitoring data. In contrast, DT applications are less prevalent during the design, construction, and reconstruction stages, with a conspicuous gap in the demolition stage. Meanwhile, the integration of smart roads with DT technology has been explored extensively, encompassing both conceptual frameworks and engineering case studies. Nevertheless, specific application pathways and standardized approaches require further consensus.
In conclusion, this paper proposes a DT framework for road engineering that spans the entire life cycle, encompassing data collection, model creation, interaction, and application. The integration of DT with emerging technologies such as GPR and the IoT is anticipated to improve the monitoring, maintenance, and development of comprehensive DT. Additionally, advanced data fusion techniques and smart technologies, including 5G/6G, AI, and VR, are expected to expand the opportunities for DT applications. Standardization in DT development is emphasized as a critical step toward creating effective and scalable solutions from smaller DT components. Future research should prioritize the development of smart roads and predictive maintenance to address challenges such as real-time interaction, cost optimization, and model fidelity. To fully realize the transformative potential of DT, efforts must focus on establishing uniform industry standards, integrating advanced perception technologies, innovating data analysis and interaction methodologies, reducing development costs, expanding life-cycle applications, and ultimately achieving fully integrated and transformative DT systems in road engineering.
Acknowledgments
This work was supported by the National Key Research and Development Program of China (2022YFB2602103 and 2023YFA1008900).
Compliance with ethics guidelines
Yu Yan, Lei Ni, Lijun Sun, Ying Wang, and Jianing Zhou declare that they have no conflict of interest or financial conflicts to disclose.
PDF
(5630KB)
