Immersive Technologies for Occupational Safety in Horizontal Transportation Construction: A Systematic Review

Abstract

The construction industry remains among the most hazardous, with workers in horizontal transportation infrastructure facing additional risks from dynamic work zones, live traffic exposure, and variable environmental conditions. Immersive technologies such as Virtual Reality (VR) and Augmented Reality (AR) offer new approaches to accident analysis and prevention, yet their applications toward improving occupational safety in transportation construction have not been comprehensively reviewed. This paper presents a systematic review of 54 studies published between 2016 and 2025 collected from two online databases (Transportation Research International Documentation and Web of Science). This review synthesizes how immersive technologies contribute to occupational risk assessment, safety training, and real-time hazard monitoring in the construction of roads, bridges, tunnels, and work zones. Each study is classified across two dimensions: the immersive medium (VR, AR, etc.) and the operational context within the construction lifecycle (onsite tools, offsite monitoring and planning, simulation-based analysis, and workforce education). This dual classification is the first to systematically map immersive technology applications for occupational safety, specifically within horizontal transportation infrastructure. The findings of this review demonstrate the unique use cases of each immersive medium, revealing that VR is primarily used for controlled experimentation and full-immersion remote analysis, whereas AR and handheld devices are preferred for field-deployed applications. Despite these promising capabilities, widespread adoption remains limited by hardware constraints, challenging field conditions, and organizational resistance. This suggests that future work should focus on safety systems tested in real-world settings and rigorously evaluated by domain experts to enable their integration into standard workplace risk management practices.

1. Introduction

Construction is a cornerstone of the United States’ economy, accounting for 2.9% of its workforce and 4.3% of its GDP as of 2023 [1]. Given the nation’s aging infrastructure and the urgent need for renovation, significant investment is anticipated in the coming years [2]. This investment is particularly critical for horizontal transportation infrastructure, such as roads, bridges, and tunnels, which underpin national mobility and commerce [3]. However, work zones represent particularly high-risk segments of the roadway network, as fatalities have increased by 21% between 2015 and 2023 [4], with struck-by vehicle incidents representing the leading cause of death in highway construction zones [5]. These environments pose elevated risks due to the dynamic nature of work zones, exposure to live traffic, and variable environmental conditions [6]. As these projects grow in scale and complexity, methods for construction safety, monitoring, and planning are undergoing massive transformation driven by advancing technologies [7]. Industry 5.0 technologies such as reality capture [8], Digital Twins [9], and Generative Artificial Intelligence [10] are accelerating construction data integration, creating an increasing need for enhanced visualization interfaces [11]. Immersive technologies, including Virtual Reality (VR) and Augmented Reality (AR), offer a promising solution by providing construction professionals with intuitive interfaces to interact with spatial data and complex infrastructure systems remotely, thereby reducing exposure to on-site risks [12].

Recent studies have documented the potential of immersive technology to improve the construction sector in areas such as safety education [13], model viewing [14], and construction planning support [15] for both horizontal and vertical infrastructure. Reported safety benefits of leveraging immersive technologies in construction include measurable improvements in hazard recognition [16] and reduced errors during construction processes [17]. However, the majority of this research has focused on vertical building construction [18,19], leaving a critical gap in understanding how these technologies apply to the distinct occupational safety challenges of horizontal transportation infrastructure where environmental exposure is constant and coordination with live traffic introduces unique hazards [6,20]. Horizontal transportation infrastructure construction refers to projects such as roads, bridges, and tunnels that extend linearly across the landscape, in contrast to vertical building construction that is typically confined to discrete building sites.

The rapid adoption of reality capture and Digital Twins in civil engineering has created a modern environment rich in data [21], yet new methods for leveraging this information to improve worker safety in transportation infrastructure are still under development [22]. As these technologies spread across various areas of transportation infrastructure and construction, a thorough review of the state-of-the-art developments is needed to synthesize safety-related findings and outline clear directions for future research.

A critical dimension of construction safety is the occupational risks inspectors face during structural assessments and quality monitoring. The National Bridge Inspection Standards require biennial inspection of all highway bridges over 20 feet (6.1 m) in length [23]. Inspections such as these expose inspectors to hazards such as live traffic, work at heights, and confined spaces [24]. The required inspection of infrastructure means that exposure to hazards is constant. More efficient inspection methods could improve occupational safety not only for the inspector but also enable early detection of hazardous conditions, such as deteriorating structural elements, before workers are exposed to these risks. Remote sensing technologies paired with immersive visualization platforms offer the potential to conduct thorough structural assessments while substantially reducing or eliminating inspector exposure to traffic, fall, and confined space hazards, while simultaneously improving the speed and consistency of defect detection that protects the broader construction workforce [17,25,26]. Current reviews on immersive technologies in related sectors have consistently highlighted their potential to improve safety, productivity, and decision-making; however, none have focused specifically on occupational safety in horizontal transportation construction, underscoring the need for the present study. Li et al. [27] reviewed VR/AR in construction safety, establishing a taxonomy which categorizes applications by technology characteristics, application domains, safety scenarios, and evaluation methods. Abbasnejad et al. [22] reviewed Industry 4.0 technologies for sustainable transportation infrastructure and found that AR and VR receive less research attention than other Industry 4.0 technologies in transportation construction. Pentury et al. [28] reviewed digitalization and computerization technologies in road construction and identified applications in pavement quality control, construction management, remote inspection, and earthwork management; however, the review excludes immersive technologies from its scope, despite covering adjacent digital technologies such as BIM, 3D modeling, computer simulation, and sensor systems. Assaf et al. [29] reviewed VR applications in offsite construction clustering the applications based on application area. Sudhakaran et al. [30] reviewed extended reality applications of road user behavior and traffic interactions. Park et al. [18] conducted a text-mining-based systematic review of digital twins in construction and identified VR as one of six core technologies for digital twin visualization, and transportation as a major sector with applications. Wang et al. [31] reviewed applications of VR in construction engineering education and found generally positive outcomes, including improved student motivation, engagement, and learning performance across applications, including construction safety training. However, their review focuses predominantly on building construction and does not address transportation infrastructure contexts. Babalola et al. [19] reviewed construction health and safety management, revealing that despite the applications in hazard identification and safety training, there is a lack of field deployment and limited research on the comparative effectiveness of different immersive mediums.

Collectively, these reviews demonstrate that while immersive technology research in general construction has matured substantially, with proven effectiveness in safety training and established taxonomies for implementation challenges, horizontal transportation infrastructure remains underrepresented, with no comprehensive systematic review addressing immersive applications for occupational safety across the construction of bridges, roads, tunnels, and work zones holistically. This systematic review addresses the knowledge gap identified in Section 1 by examining how immersive technologies are currently used in occupational safety for the construction of horizontal transportation infrastructure. It is guided by the following primary research question: How are immersive technologies currently being applied in the construction of horizontal transportation infrastructure for occupational safety? Specifically, the review addresses three supporting questions:

1.

What types of immersive technologies are being deployed for occupational safety and for which phase or task in the construction process?

2.

What capabilities and benefits of immersive technologies for occupational safety have been demonstrated in research-based implementations

3.

What technical or methodological limitations constrain the widespread adoption of immersive technologies for occupational safety?

Answering these questions will enable practitioners and researchers to identify both the most promising application contexts for current deployment, their benefits, limitations, and which areas to prioritize for further research before immersive technologies can achieve widespread adoption in this domain.

2. Methodology

This review examines the integration of immersive technologies for occupational safety in horizontal transportation construction. This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [32] (See Supplementary Materials). This review was not prospectively registered and no protocol was developed. The defined scope of this literature review is “immersive technology applications for occupational safety in transportation infrastructure construction” with a focus on horizontal structures including roadways, bridges, and tunnels. The technologies included in this review should be directly applicable to the project’s safety management, construction monitoring, or workforce training; therefore, technologies such as driving simulators are omitted unless they are directly applicable to the construction process. To initiate the literature review, articles for potential inclusion in this study were collected from two distinct data sources. First, the Transportation Research International Documentation (TRID) database was included, as it is the world’s largest and most comprehensive bibliographic resource for transportation research [33]. Second, the Web of Science Core Collection, a multidisciplinary citation index that covers literature across many disciplines [34]. Results were restricted to records containing available abstracts and a legitimate DOI. To maintain currency, the date range for this search was 2016–2025, as many of these technologies have only recently seen consumer-ready products. All corresponding titles and summaries were subsequently downloaded for review as of November 2025. Upon several iterations, the following queries were used in the final literature search, resulting in the most relevant collection:

• (“augmented reality” OR “virtual reality” OR “extended reality” OR “mixed reality” OR “immersive technolog*”) AND (“horizontal construction” OR “horizontal infrastructure” OR “transportation infrastructure” OR “transportation project*” OR “highway” OR “roadway” OR “heavy civil” OR “civil infrastructure” OR “work zone” OR “bridge construction” OR “pavement design” OR “pavement installation” OR “pavement construction” OR “pavement maintenance” OR “pavement management” OR “paving” OR “asphalt paving” OR “concrete paving” OR “road surface”)
• (“augmented reality” OR “virtual reality” OR “extended reality” OR “mixed reality” OR “immersive technolog*”) AND (“bridge inspection” OR “Road inspection” OR “flagger” OR “road construction” OR “pavement inspection”)

The WOS search identified 302 records, and the TRID search identified 200. Duplicate removal using title matching was employed, removing 60 records from the identified list. Thus, 442 articles were collected for the subsequent screening phase.

Importantly, the search strategy was not limited to keywords related to occupational safety. Since safety in transportation infrastructure construction is influenced by more than just tools that directly warn of hazards or prevent accidents, the search also includes terms such as “pavement management” and “bridge inspection”. Safety in this context also depends on a broader set of activities across the project’s lifecycle, such as structural inspection, defect detection, construction monitoring, and quality assurance. These activities help prevent the conditions that can lead to worker injuries and premature structural degradation. By using a more refined search and classifying each study’s safety relevance during analysis, this review captures a broader range of articles potentially within this review’s scope, rather than limiting itself to studies that explicitly label themselves as safety research.

A three-step exclusion process was employed to determine if the article’s abstract was within the review’s scope: (1) determine if the paper is within the field of horizontal transportation infrastructure projects, (2) determine if the paper is within the scope of virtual/augmented realities for occupational safety, and (3) determine if the article is a review paper. Review articles were excluded unless they contained a substantially original research component that directly contributes to this review. The screening was performed in that order. As such, articles that were excluded during the first step may have also been outside of the scope in other regards, but are marked as excluded under the first criterion. During the abstract screening process, 342 articles were excluded for the first criterion, 4 for the second, and 18 for the third. In total, 78 full-text articles were assessed for eligibility within this study.

Finally, a deeper, more extensive analysis of the full texts was conducted to determine eligibility for this review. Six articles did not have a full text available and were removed. Upon detailed examination of the remaining 72 texts, 14 articles were determined to be outside the defined scope based on the same three-step exclusion process used for abstracts. For example, although Azofeifa et al. [35] passed initial abstract screening, on full-text review, it was revealed that the VR application in engineering education was not specifically related to horizontal construction. Similarly, although Jung et al. [36] developed 3D environments for sight distance analysis, there was no immersive technology and was thus removed at this stage. Additionally, four articles were removed for poor publication quality (i.e., publication in non peer-reviewed journals or excessive grammatical errors). After these final exclusions, 54 studies met all criteria and were included in the final review. Figure 1 visualizes the process performed via the PRISMA diagram. WOS accounted for the majority of unique hits with 34, while TRID contributed 7 unique records; 13 studies appeared in both databases. The abstract screening and full-text review process were conducted by a single reviewer. To verify the reproducibility of the screening process, a secondary reviewer independently screened a random 10% sample of abstracts, yielding an observed agreement of 93% and a Cohen’s $\kappa$ of 0.80, indicating substantial agreement [37]. Any potential screening inconsistencies would be further resolved during the full-text review stage. As such, the overall methodology remains highly reproducible.

3. Literature Review

3.1. Bibliometric Overview

Following the identification and screening phases, a bibliometric overview was conducted to establish a baseline understanding of the current research landscape. Figure 2 shows the publication timeline of the included articles, spanning from 2016 to 2025 as previously mentioned for research currency, which is the chosen publication time frame. Additionally, the Figure 2 reproduces the publication trends from Babalola et al.’s review on immersive technology applications for safety and health management in the construction sector [19]. This comparison provides a visualization of immersive technology-based safety research trends across both general construction and horizontal infrastructure. The timeline shows a distinct maturation in the field, with a sharp increase in annual publications in 2021. Between 2016–2020, there were an average of 0.8 publications per year within the review’s scope. Between 2021–2025, the second half of the publication timeline, there was an average of 10. This is consistent with observed trends, as virtual reality saw a significant increase in usage and private equity investment around this time [38,39]. Based on comparative trends from Babalola et al.’s study, it is indicated that increased research activity in immersive technologies for construction began a few years earlier than in the specific domain considered in this review. However, the overall growth patterns and acceleration in publication volume around 2021 are broadly consistent. Figure 3 visualizes the distribution of publishers for included articles, among those with more than one appearance within the selected articles. Advanced Engineering Informatics is the most common, followed by Automation in Construction, Transportation Research Record, Sensors, and Construction Research Congress. Buildings is the least frequent journal to have multiple appearances. Additionally, 31 publishers have one paper.

To characterize the research landscape, a keyword co-occurrence analysis on the article abstracts was performed using VOSviewer (version 1.6.20). Figure 4 presents this network visualization, where node size indicates keyword frequency. The analysis reveals four distinct research clusters. The green cluster centers on structural inspection and digital infrastructure-based safety applications, emphasizing terms such as ‘inspection’, ‘bridge’, ‘model’, and ‘digital twin’. The red cluster focuses on field operations and human factors, dominated by terms including ‘safety’, ‘work zone’, ‘worker’, and ‘roadway work zone’, highlighting research that examines occupational hazards, behavioral responses, and real-time risk management in active construction environments. The yellow cluster occupies the upper portion of the network, connecting terms such as ‘BIM,’ ‘building information modeling,’ ‘efficiency,’ and ‘effectiveness,’ representing studies that evaluate digital modeling capabilities and safety outcomes. The blue cluster bridges multiple domains with terms like ‘construction industry,’ ‘work,’ ‘training,’ and ‘documentation,’ reflecting cross-cutting themes in workforce development and organizational implementation. This thematic structure indicates that immersive technology research on safety in horizontal infrastructure encompasses both digital infrastructure systems for inspection and monitoring (green and yellow clusters) and field-based applications that address worker exposure to hazards in dynamic construction environments (red cluster).

To further understand the literature collection, a dual classification was employed to systematically label each article’s (1) immersive medium and (2) tool class. The first set of definitions (See

Table 1. Immersive medium definitions.

Classification	Definition
Virtual Reality	A fully immersive environment in which the user is utilizing a head-mounted display (HMD) for visualization.
Augmented Reality	A technology that superimposes digital content over the user’s view of the physical world using a see-through HMD or smart glasses.
Handheld Device	Superimposed digital content over the user’s view of the physical world using a handheld device such as a tablet or smartphone.
3D Keyboard Mouse	A non-immersive virtual environment displayed on a standard 2D monitor, where the user navigates and interacts using traditional input devices like a keyboard and mouse.
CAVE	A room-sized, immersive virtual reality space where 3D imagery is projected onto the walls (and occasionally the floor and ceiling) to surround the user in real-life space.
Laser Projection	Superimposed digital content onto the real-world environment by means of lasers.

) clarifies the immersive medium, detailing the hardware and display technologies used to deliver the experience. Applications range from fully enclosed virtual environments such as Virtual Reality and CAVE systems, to mixed-reality approaches using Augmented Reality headgear or Handheld Devices, to semi-immersive methods such as 3D Keyboard/Mouse setups. The second set of definitions (See

Table 2. Tool class definitions.

Classification	Definition
Onsite Tool	Field-based applications deployed directly at the physical job site to enhance labor productivity and safety.
Offsite Viewer	Platforms used to visualize or monitor project data remotely without being physically present.
Simulation	Virtual environments that extend a base “truth” model by adding specific calculations or features to investigate performance and derive design insights.
Education	Applications designed to facilitate learning and training for students or professionals.

) categorizes the tool class and specifies how each immersive medium supports occupational safety functions across the construction lifecycle. This classification identifies whether the tools are used directly at the construction location (Onsite Tool), for remote oversight and planning (Offsite Viewer), for safety analysis and behavioral research (Simulation), or for workforce safety training purposes (Education).

Based on these definitions, each article was assigned to the immersive medium and tool class. In some cases, multiple immersive mediums or multiple tool classes appeared in the article. As such, some received multiple labels, and the sum of the frequency counts in Figure 5 and Figure 6 exceeds the number of articles in this study. For example, Sabeti et al. [40] is classified into two tool classes (‘Simulation’ and ‘Onsite Tool’) and three immersive mediums (‘Virtual Reality’, ‘Augmented Reality’, and ‘3D Keyboard Mouse’) as the methodology involves multiple distinct experiments to test augmented reality work zone alerts (an Onsite tool) in a simulated immersive environment. Findings were then synthesized narratively, organized thematically by immersive medium and tool class to enable cross-study comparisons across application domains. Figure 5, which illustrates the frequency of each technology, reveals that Virtual Reality is the most widely studied (29). Augmented Reality follows as the second most frequent, with 16 occurrences, indicating significant but less frequent use than VR. Handheld devices (6), 3D Keyboard Mouse (6), CAVE systems (3), and Laser Projection (1) comprise the rest of the articles with more specialized use cases.

Unlike the immersive medium frequency, the tool class categorizations are significantly more evenly distributed. A cross-tabulation of immersive medium by tool class is reported in

Table 3. Immersive medium by tool class cross-tabulation from Figure 6.

Tool Class	Articles	VR	AR	HD	3DKM	CAVE	LP	Classifications
Onsite Tool	19	3	13	5	1	0	1	23
Offsite Viewer	16	11	3	1	2	0	0	17
Simulation	18	16	2	0	3	1	0	22
Education	10	6	0	0	2	3	0	11
Total	54	29	16	6	6	3	1	73

VR = Virtual Reality, AR = Augmented Reality, HD = Handheld Device, 3DKM = 3D Keyboard Mouse, LP = Laser Projection. Articles = unique articles assigned to this tool class. Classifications = total immersive medium assignments within this tool class, reflecting that some articles received multiple classifications.

and visualized in Figure 6. Onsite tools (19) are the most common VR/AR tools among the articles. Simulation and offsite viewer are the second- and third-most common, with 18 and 16 appearances, respectively. Lastly, education is the minority class with 10 occurrences. The relatively uniform distribution across these four safety-relevant categories suggests that immersive technologies are being explored more broadly across the occupational safety landscape, rather than concentrating in a single application area. To illustrate the relationship between immersive mediums and their application contexts, Figure 6 additionally presents a hierarchical visualization of the distribution across tool classes. The inner ring segments the articles by tool class, while the outer ring subdivides each category by the specific immersive medium employed. This visualization reveals distinct medium preferences within each safety-related application domain.

Table 4. Categorization of reviewed articles by tool class and immersive medium.

Author (Year)	Immersive Medium(s)
Onsite Tool
Arvikar et al., 2025 [41]	Handheld Device
Awadallah and Sadhu, 2023 [42]	Augmented Reality
Binni et al., 2025 [43]	Augmented Reality
John Samuel et al., 2022 [44]	Handheld Device
El Kassis et al., 2023 [45]	Augmented Reality
Martins et al., 2024 [46]	Handheld Device
Mohamed and Tran, 2022 [47]	Augmented Reality, Virtual Reality
Mohammadkhorasani et al., 2023 [48]	Augmented Reality
Mojidra et al., 2024 [49]	Augmented Reality
Nguyen et al., 2022 [50]	Augmented Reality
Nilnoree and Mizutani 2025 [51]	Handheld Device
Pantoja-Rosero and Salamone, 2025 [52]	Augmented Reality
Sabeti et al., 2021 [53]	Augmented Reality
Sabeti et al., 2022 [54]	Augmented Reality
Sabeti et al. 2024 [55]	Augmented Reality
Sabeti et al., 2024 [40] [Also: Simulation]	Augmented Reality, Virtual Reality
Tschulik et al., 2025 [56]	Laser Projection
Xu et al., 2022 [57]	Augmented Reality, Virtual Reality
Yu et al., 2022 [58]	Handheld Device
Offsite Viewer
Alhady et al., 2024 [59]	Augmented Reality, Virtual Reality
Carter et al., 2024 [60]	Augmented Reality
Choi et al., 2023 [61]	Virtual Reality
d’Avigneau et al., 2025 [62]	Virtual Reality
Offsite Viewer
Fawad et al., 2024 [63] [Also: Simulation]	Augmented Reality
Kong et al., 2025 [17]	Virtual Reality
Moradi and Assaf, 2022 [64] [Also: Simulation]	3D Keyboard Mouse
Omer et al., 2019 [25]	Virtual Reality
Omer et al., 2021 [26]	Virtual Reality
Wang et al., 2023 [65]	Virtual Reality
Wilson Simao et al., 2023 [66]	Handheld Device
Yiǧit and Uysal, 2025 [67]	Virtual Reality
Simulation
Aati et al., 2020 [68] [Also: Education]	Virtual Reality
Ardecani and Shoghli, 2025 [69]	Virtual Reality
Ardecani et al., 2025 [70]	Virtual Reality
Ergan et al., 2022 [71]	Virtual Reality
Hao et al., 2021 [72] [Also: Offsite]	Virtual Reality
Kim et al., 2021 [73]	Virtual Reality
Kim et al., 2021 [74]	Virtual Reality
Lu and Ergan, 2025 [75]	Virtual Reality
Lu et al., 2025 [76]	Virtual Reality
Marzouk and Elsayed, 2024 [77] [Also: Offsite]	Virtual Reality
Saeidi et al., 2019 [78]	3D Keyboard Mouse, Virtual Reality
Shen et al., 2022 [79] [Also: Offsite]	Virtual Reality
Zhang et al., 2025 [80]	3D Keyboard Mouse
Zou et al., 2020 [81]	Virtual Reality
Education
Arif, 2021 [82]	CAVE
Eiris et al., 2022 [83] [Also: Offsite]	3D Keyboard Mouse
Li et al., 2022 [84]	3D Keyboard Mouse
Li et al., 2023 [85]	Virtual Reality
Luleci et al., 2024 [86] [Also: Simulation]	CAVE, Virtual Reality
Qing and Edara, 2024 [87]	Virtual Reality
Sakib et al., 2021 [88]	Virtual Reality
Tanbour et al., 2024 [89]	CAVE
Yu et al., 2022 [90] [Also: Simulation]	Virtual Reality

builds upon the bibliometric trends visualized in Figure 5 and Figure 6 by listing each reviewed article and its classifications. This dual classification schema highlights an intuitive pattern in the research landscape: Virtual Reality is predominantly used in simulation and offsite viewing, whereas Augmented Reality and handheld visualization dominate onsite tools in applications such as defect detection and progress tracking. Education applications do not leverage Augmented Reality or handheld devices, but they are fairly evenly distributed across the CAVE, Virtual Reality, and 3D Keyboard Mouse. This mapping also establishes the technical foundation for the subsequent in-depth review, enabling a more targeted discussion of how each study applies VR/AR technologies to address specific challenges in transportation infrastructure.

To enhance discussion of the technical landscape of the reviewed literature,

Table 5. Computational methods and algorithms appearing across multiple articles.

Key Technology/Algorithm	Representative Articles
Computer Vision & Defect Detection
Deep Learning Detection	Sabeti et al. [53]; Awadallah and Sadhu [42]; Pantoja-Rosero and Salamone [52]
Feature-Point Tracking	Mohammadkhorasani et al. [48]; Mojidra et al. [49]
UAV Photogrammetry	Kong et al. [17]; Yiǧit and Uysal [67]
LiDAR/Point Cloud Processing	Nilnoree and Mizutani [51]; Nguyen et al. [50]; Luleci et al. [86]; Omer et al. [25]; Omer et al. [26]
Spatial Registration
BIM Registration	Binni et al. [43]; Nguyen et al. [50]; Martins et al. [46]; John Samuel et al. [44]; Arvikar et al. [41]
SLAM-based Tracking	Martins et al. [46]; John Samuel et al. [44]
Physiological Signal Processing
EEG Analysis	Ardecani and Shoghli [69]; Ardecani et al. [70]; Kim et al. [74]
Biosignal Monitoring	Kim et al. [74]; Sakib et al. [88]; Zou et al. [81]; Kim et al. [73]
Simulation & Modeling
Hardware-in-the-Loop	Ergan et al. [71]; Lu et al. [76]; Zhang et al. [80]
Traffic Simulation	Marzouk and Elsayed [77]; Wilson Simao et al. [66]; Saeidi et al. [78]
Digital Twin Integration	Alhady et al. [59]; Fawad et al. [63]; d’Avigneau et al. [62]; Shen et al. [79]; Carter et al. [60]
Sensor Integration & Alerting
Real-Time Sensor Streaming	Carter et al. [60]; Fawad et al. [63]; Sabeti et al. [54]
Multimodal Alert Systems	Sabeti et al. [55]; Sabeti et al. [40]; Lu and Ergan [75]; Zou et al. [81]

summarizes the key computational methods and algorithms identified across the relevant articles, organized by method type.

A co-authorship network analysis, also visualized with VOSviewer in Figure 7, reveals the collaborative structure among the most prolific authors in the reviewed literature. Only authors with at least two studies in this review are analyzed. Seven distinct clusters emerged. The largest cluster, colored red and based at the University of New Mexico, comprises a group that collectively advances AR-based bridge inspection and fatigue crack detection methodologies [48,49,57]. This cluster centers on F. Moreu, who is an author on every publication in this cluster. The second-largest, highlighted in green and based at New York University, collaborates on VR-based work zone safety simulations and hardware-in-the-loop platforms to study worker-vehicle interactions [71,75,76,80,81]. The most prominent author in this cluster is S. Ergan, who appears on all four publications. The third-largest cluster, colored blue and based at the University of North Carolina Charlotte, studies AR safety warning systems and neurophysiological responses in highway work zones [40,53,54,55,69,70]. O. Shoghli is the most frequently appearing author in this group, with authorship on all six documents. The smaller clusters focus on collaborations in topics such as VR-based behavioral conditioning [73,74], remote bridge inspection [25,26], VR-based flagger training [68,87], and AR-based structural damage detection [42].

3.2. Detailed Review of Articles

To contextualize the bibliometric trends and classifications presented in the previous figures and tables, the following section provides a detailed review of each article included in this study. The review is organized according to the tool classes defined earlier: onsite tools, offsite viewers, simulations, and education platforms. This structure enables synthesis by emphasizing how each immersive medium addresses specific occupational safety challenges in transportation infrastructure construction. By grouping studies by the deployment of immersive technologies, the review highlights common design strategies, emerging research challenges, and distinctive contributions across different operational contexts. Onsite tools exhibit the greatest diversity in immersive mediums, with Augmented Reality (13) dominating due to its suitability for overlaying safety-critical digital information or inspection-based content onto physical infrastructure, followed by handheld devices (5), Virtual Reality (3), and a single instance of laser projection. Simulation applications demonstrate a pronounced preference for Virtual Reality (15), reflecting the need for fully immersive environments to replicate complex work zone scenarios and investigate worker-vehicle interactions. Offsite viewers similarly favor Virtual Reality (11) for remote inspection and digital twin visualization, though the category also includes Augmented Reality (3), handheld devices (1), and 3D keyboard/mouse interfaces (2). Educational applications show the most balanced distribution among immersive mediums, utilizing Virtual Reality (6), CAVE systems (3), and 3D keyboard/mouse setups (2), suggesting that educational contexts accommodate a wider range of immersive mediums depending on classroom accessibility and group-based learning requirements.

3.2.1. Onsite Tools

Onsite tools, defined as “Field-based applications deployed directly at the physical job site to enhance labor productivity and safety” (See Table 2), represent the most heavily studied application area of immersive technologies in horizontal infrastructure construction. These tools are particularly safety-oriented because they operate within active construction environments where workers face immediate hazards from live traffic, heavy equipment, and structural deficiencies. In doing so, these onsite tools have the potential to protect workers, improve the efficacy and consistency of inspections, and reduce inspectors’ exposure to live traffic and other on-site risks. However, the efficacy of these applications is intrinsically linked to the underlying hardware capabilities. To this end, Xu et al. [57] evaluated the state of the art across 16 AR HMDs and developed a framework for civil infrastructure operations. The study established a strong linear correlation between price and weight (r = 0.91), indicating that heavier, more expensive units are currently required to support the advanced processing needed for complex visualization tasks such as real-time defect highlighting. Referencing ANSI Z89.1, which establishes 425 g as the maximum recommended weight for Class A/C safety hard hats used in civil infrastructure environments, the authors note that several high-capability devices commonly reported in related studies exceed this threshold. This finding carries direct safety implications, as prolonged use of heavy headsets in active work zones introduces ergonomic risks and may compromise worker attention to surrounding hazards.

A significant body of research focuses on bridge inspection and defect documentation, in which AR systems streamline visual assessments and improve the accuracy of digital records. Early and accurate defect detection is essential for both structural integrity and occupational safety. Mohammadkhorasani et al. [48] integrated a fatigue crack detection model with a HoloLens-based augmented reality overlay system to highlight feature points directly on concrete bridge structures. This process was validated through testing with industry stakeholders, and achieved an error rate of only ~3%. A similar approach was explored by Mojidra et al. [49], who used an AR headset in conjunction with cloud computing to detect cracks under AASHTO fatigue loading, achieving a crack detection intersection rate of 0.73 for in-plane and 0.66 for out-of-plane cracks. As in the previously mentioned article, detected feature points are highlighted over real-life features. To further enhance on-site inspection, Pantoja-Rosero and Salamone [52] developed real-time crack detection via AI-based damage segmentation, trained on images from traditional visual inspections, using a YOLO11 model, with near real-time inference of 0.66 s per frame on the HoloLens 2. Collectively, these studies demonstrate that AR enhances structural inspection by integrating automated defect recognition with human expertise. Awadallah et al. [42] extend this line of work by combining YOLOv5 crack detection with an AR-assisted measurement interface in which inspectors place 3D segmentation points to compute crack lengths, scoring measurement errors below 2% and demonstrating the feasibility of AR-supported defect quantification. Overall, AR-enabled automated detection reduces the time inspectors must spend exposed to traffic hazards, working at heights, and in close proximity to deteriorated structural elements. By streamlining the inspection process and improving defect recognition accuracy, these systems minimize inspector exposure to onsite risks while preventing structural failures that endanger workers and the public.

In addition to crack detection, researchers have leveraged AR for registration and model alignment, enabling practical applications in the inspection of transportation infrastructure. Binni et al. [43] proposed an AR-based automatic registration framework that leverages knowledge graph technology to register images to BIM models for inspection. This approach dynamically updates the virtual model’s geographical coordinates to reflect real-world environmental changes, such as tectonic displacements. Similarly, Nguyen et al. [50] developed an automatic registration framework but utilized drone-based laser scanning. In contrast to these approaches, Martins et al. [46] developed a mixed-reality application specifically for handheld devices to track BIM model alignment with onsite conditions. Samuel et al. [44] designed an AR handheld device to update BIM or digital-twin models of structural defects based on visual inspections, demonstrating a 50% reduction in inspection time, and thus, risk exposure. These approaches move towards AR as an interactive tool for updating and maintaining digital models, rather than a more limited visualization tool. Digital documentation methodologies, such as that developed by Nilnoree and Mizutani [51], support similar efforts by integrating iPhone Lidar-based AR scanning for roadway management and the recording of large-scale construction changes, achieving post-alignment distance accuracy within 6.3 cm and reducing measurement time from 20–30 min to 2–3 min per round compared to traditional survey tools.

Another primary stream within onsite tools focuses on visualizing BIM models for site management during early construction stages. Yu et al. [58] developed a handheld AR-based measurement tool to enhance placement accuracy of model projections on infrastructure projects, leveraging model simplification techniques to reduce computational demands on mobile devices, with post-improvement mean displacement errors of 0.066 m longitudinally and 0.013 m transversely, satisfying the expert-defined tolerance threshold of 0.1 m. Arvikar et al. [41] integrated project scheduling with AR visualization, enabling experts to overlay relevant BIM sections onto site photos in real time across five bridge construction sites, with evaluation by 30 on-site professionals yielding an overall system satisfaction score of 85.65%. These systems underscore the importance of aligning AR capabilities with on-site usability and hardware limitations.

Safety-focused research constitutes another cluster, addressing the high hazard levels in active highway work zones. Sabeti et al. [54] developed a Wi-Fi–enabled AR hazard warning system that triggers real-time alerts for vehicle intrusions in under 10 milliseconds using Vuzix smart glasses. Sabeti et al. [53] similarly developed AR glasses integrated with a deep learning model trained for vehicle detection and classification, enabling real-time wireless warnings. As a follow-up project, Sabeti et al. [40] conducted an experiment evaluating the effect of alert modality on worker reaction time to AR safety warnings. Such findings indicated that multimodal warnings, combining visual and haptic alerts, elicited the fastest reaction time. Building on these results, Sabeti et al. [55] designed a multimodal methodology for work zone safety alerts using the aforementioned multimodal alerts, and validated it via user trials. There was a significant correlation between perceived trust and usability under this study. El Kassis et al. [45] studied video feed of construction workers wearing AR tools for safety communication on uncontrolled construction sites, and defined 14 factors that influence the use of AR for these activities. Their study concluded that AR can significantly improve communication efficiency, but its success relies on addressing challenges related to hardware usability, environmental conditions, connectivity, and user familiarity. Additionally, it was noted that many of the negative factors may be improved through training and task repetition.

Beyond traditional viewing mediums for spatial computing, such as immersive headsets or handheld devices, Tschulik et al. [56] introduced a laser projection-based AR technology for assisting heavy machine operators through overlayed guidance for low lighting conditions, with accuracy evaluation demonstrating projection deviations within 6 cm horizontally across distances up to 25 m under real-world snow groomer conditions. The methodology included real-time calculations of ground surface properties to ensure that projected overlays remained stable and spatially accurate during machine operation. The overall industry readiness for immersive tools for on-site management was assessed by Mohamed and Tran [47], in a survey of U.S. state Departments of Transportation (DOTs) which revealed that only one state had deployed VR or AR inspection systems as of 2022. The DOT in question, which reported deploying VR/AR inspection systems, responded that this technology was used for earthwork inspection and quantity determination. Agencies viewed training limitations, poor connectivity, cost concerns, and resistance to technological change as primary barriers to implementation.

3.2.2. Offsite Viewers

Offsite viewers, defined as “Platforms used to visualize or monitor project data remotely without being physically present” (See Table 2), represent a rapidly expanding category of emerging technologies that enable offsite inspection, public engagement, and data-driven decision-making. These remote visualization capabilities fundamentally improve occupational safety by eliminating workers’ exposure to hazardous conditions prevalent during inspections, such as live traffic and work near moving construction vehicles. One segment in this collection of work is the use of VR environments to explore 3D scene captures (generated from methods such as photogrammetry or laser scanning) of construction sites. Yiǧit and Uysal [67] demonstrated offsite viewing of unmanned aerial vehicle (UAV) photogrammetry data for offsite structural inspection. In an expert-based review, the technology yielded an overall satisfaction score of 8.8/10 and visual clarity score of 9/10. As a method to assess the visual fidelity of these photogrammetry scans, Kong et al. [17] developed a framework for pixel-level visual fidelity assessment. This framework provides a clear and interpretable approach for verifying model quality before employing these scans in remote structural safety assessments of transportation infrastructure. Similarly, Choi et al. [61] leveraged 360° image viewing via VR for infrastructure inspection and assessed the method using data from two bridges, with the resulting smart inspection system reducing inspection time from 2–4 weeks to a single preliminary and two main inspections.

Further work towards remote site inspection includes Omer et al. [26], who developed a case study comparing conventional inspection methods and VR-based inspection methods, and found that VR methods produced highly repeatable results, with inspectors consistently identifying the same defects and ratings across settings, unlike the greater variability seen in field inspections. This repeatability represents a key safety advantage, as it reduces the likelihood that critical defects are overlooked due to inspector fatigue, environmental distractions, or inconsistent field conditions. Although limitations in texture quality due to point-cloud resolution were reported. In earlier work by Omer et al. [25], a smartphone, configured as a VR display via a headset-style adapter, was used as a viewer for bridge inspection, demonstrating that even low-cost VR platforms can support repeatable point-cloud viewing for construction inspection. Wang et al. [65] further advanced workflows for remote inspection using VR-based point cloud viewers by integrating user-centric interaction techniques, including graphical menus, voice commands, gestural interaction, and object selection, positioning, and rotation. This workflow achieved a System Usability Scale score of 77.83 across 22 participants, which is categorized as “Excellent”. Such interaction provides users with multiple views of a scan and enables real-time marking of damage.

Several studies demonstrate the integration between BIM and immersive technologies for digital twin applications. Alhady et al. [59] used AR and VR technologies to visualize bridge and roadway designs for preconstruction adjustments, noting that VR supports deep technical investigation while AR provides intuitive context or nontechnical stakeholders. Data driven 3D environments were further advanced by d’Avigneau et al. [62], with the introduction of CAMHighways dataset, a large-scale point-cloud-derived environment designed specifically to support AR and VR research for roadway infrastructure management. The resulting model spans more than 40 km providing a mobile-mapping-derived digital twin containing georeferenced meshes and labeled infrastructure defects. Carter et al. [60] developed a AR system to enable real-time visualization of structural vibration data by streaming sensor measurements, demonstrating accurate visualization of time-domain, frequency-domain, and modal-shape parameters across three sensors at 100 Hz with no performance degradation over 60-second collection windows. Lastly, Fawad et al. [63] proposed structural health monitoring framework that combines finite element modeling, sensor networks, and BIM-based AR visualization to present live structural performance for infrastructure construction. This application is also classified as a simulation technique (See Table 1).

Public engagement in the literature is identified as a promising application area, including the development of an AR system for public meetings by Wilson Simao et al. [66]. The proposed system used handheld devices as an AR viewer to see proposed improvements to local transportation infrastructure by hovering over target images of the location, allowing stakeholders to visualize roadway design interactively. In parallel, Moradi and Assaf [64] developed an augmented-reality decision-support framework that overlays predicted pavement distresses and recommended maintenance actions onto a desktop-based 3D model of an urban road network. Their approach demonstrates how AR-based visualization can help city councils better anticipate the consequences of pavement management decisions by presenting analysis results in an interactive, spatially aligned format.

3.2.3. Simulation

Simulation technologies, defined as “Virtual environments that extend a base “truth” model by adding specific calculations or features to investigate performance and derive design insights” (See Table 2), enable a base model representing real site conditions to undergo experiments while leveraging immersive technologies for visualization in the construction industry. One type of work in this domain focuses on replicating the complex roadway environments to understand worker-vehicle interaction in a safe and controlled experiment. Marzouk and Elsayed [77] developed a VR-based bridge information modeling workflow to visualize the impacts of construction on work zone traffic, demonstrating a methodology for enabling advanced planning strategies for construction sequencing. Zou et al. [81] used a VR work zone paired with wearable sensors to study how alarm factors influence workers’ responsiveness to safety warnings around live traffic. The findings indicated that modality and frequency affected behavior, but duration did not.

Considering the dangers associated with work zone construction for workers, simulation research leveraging VR has investigated cognitive behavior to further understand safety. Kim et al. [73] demonstrated that repeated exposure to simulated struck-by risks in VR leads to risk habituation, where participants become desensitized to hazards. However, the study also found that behavioral interventions, such as vividly illustrating the consequences of these hazards, can meaningfully reduce this habituation, highlighting VR’s potential for controlled behavioral conditioning before field deployment. Additionally, Kim et al. [74] used biosignals collected during a VR construction task to identify physiological markers correlated with inattentiveness, achieving a classification accuracy of 72.2% using multimodal biosignals (electrodermal activity, eye tracking, and contextual features), demonstrating the potential for automated worker attention monitoring. The results showed significant differences in neural and bodily responses between attentive and inattentive states. Ardecani and Shoghli [69] examined stress responses to AR safety warnings using a fully immersive VR environment containing roadway tasks of varying intensities. The study found that moderate-intensity tasks caused a much stronger electrodermal response than light-intensity tasks. It also showed that AR warning signals were processed within 125 ms under low physical load but were delayed to 125–250 ms under moderate exertion, indicating that higher task intensity reduces the cognitive bandwidth available to process safety warnings. Ardecani et al. [70] similarly used Electroencephalography (EEG) signals to quantify situational awareness, attention, and cognitive load associated with AR warning delivery, finding that beta and gamma power increased post-warning within 125 ms under light-intensity tasks but peaked between 125–250 ms under moderate-intensity tasks, with the temporal delay indicating that physical workload reduces the speed of cognitive engagement with safety warnings.

Building on these physiological insights, recent research has shifted toward the development of integrated hardware platforms and the optimization of specific safety interventions. Ergan et al. [71] developed a two-way hardware-in-the-loop platform that synchronizes VR with traffic simulation to test sensor systems throughout interactions between workers and vehicles. Similarly, Lu and Ergan [75] evaluated the effectiveness of different wearable alarm attributes using VR environments, identifying sensory modality as the most influential alarm attribute, more than twice as influential as duration and over six times more influential than repetitions, with haptic-auditory alarms at 350 ms duration most frequently improving worker safety behavior. Building on these findings, Lu et al. [76] applied reinforcement learning to determine optimal alarm control policies for consistent worker reactions, using VR environments to collect behavior data such as safe/unsafe positioning, gaze direction, and post-alarm movement. Notably, this study found no evidence of alarm fatigue.

Simulation research also extends into training and operation assessment, particularly to simulate different weather or lighting conditions. Aati et al. [68] implemented one of the first VR-based work zone inspection training systems, enabling DOT engineers to drive through simulated rural and urban work zones under varying environmental conditions, with evaluation by 34 MoDOT staff revealing an average immersive module score of 79%, with 88% of participants scoring above 70%, and 97% agreeing that VR was useful for inspector training. Meanwhile, Shen et al. [79] used a digital-twin based VR and driving-simulator setup to evaluate tunnel lighting performance, validating VR-based results against field measurements. The results demonstrated that lighting uniformity within the simulated environment could be matched to that observed in the real-world tunnel. Hao et al. [72] similarly developed a BIM-VR design simulation for a tunnel project to determine real time site distance calculations based on varying driving conditions. Another shared VR environment was developed by Saeidi et al. [78], which enabled drivers and flaggers to interact within a coordinated VR highway work zone, demonstrating that co-presence can reveal distraction patterns and mismatches in worker-driver perception. Zhang et al. [80] integrated VR-CARLA co-simulation and eye tracking to analyze driver behavior near work zones across 20 participants, finding that workers exhibiting risky behaviors consistently received higher gaze fixation ratios than safely behaving workers regardless of warning sign presence, and reporting high participant satisfaction (7.78/10) and minimal motion sickness (1.75/4). These works are critical for flagger safety and struck-by prevention in active work zone environments.

3.2.4. Education

Educational applications, defined as “Applications designed to facilitate learning and training for students or professionals” (See Table 2), focus on improving knowledge transfer or skill development in transportation infrastructure construction. Often, this is done through experiential learning environments that integrate complex data sources into an immersive interface. Luleci et al. [86] fused LiDAR point clouds with dynamic structural health monitoring data in a VR environment using both HMDs and a VR table, allowing students to visualize vibration mode shapes of a pedestrian bridge. Similarly, Li et al. [85] developed an immersive VR teaching platform for bridge engineering which organizes educational concepts based on an ontology-based semantic hierarchy. The evaluation identified a 40% increase in students’ interest in learning, demonstrating how immersive interaction can enhance engagement.

Another segment of educational studies apply to safety and emergency response training. Yu et al. [90] integrated BIM and VR to create a tunnel fire evacuation and emergency response training system, with evaluation across 32 staff over four months showing that operation scores for the central control room doubled (84 to 173) and operation time was reduced from 40 to 17 min, with comparable improvements observed for the emergency response group. Sakib et al. [88] similarly investigated VR-based drone operation training, using physiological data and machine learning to predict trainee performance. The results showed 83% accuracy in performance prediction and no significant increase in stress levels compared with outdoor training. In another safety focused application, Qing and Edara [87] developed a VR-based work zone flagger training program integrated into an active MoDOT training course, achieving a System Usability Scale score of 78.4/100 across 28 participants, with over 82% reporting improved focus during training and 93% reporting a more positive overall training experience.

Immersive environments are also leveraged for inspection and operational skill training. Li et al. [84] developed a methodology for VR-based training and assessment of bridge inspections. Operational and biometric data, as well as post-study assessment, indicated that the system supports skill development for inspection personnel without requiring access to live bridge structures. Complementing this, Eiris et al. [83] introduced VR-OnSite, a web-based virtual environment that enables onsite site visits for construction education. During a pilot study in a railway transportation course, significant increases in students’ perceived knowledge transfer were reported.

Alternative immersive setups are similarly explored for classroom deployment. Tanbour et al. [89] developed a VR wall (CAVE) system to simulate highway bridge construction processes, enabling group-based visualization and discussion in an academic setting without requiring individual headsets, thereby expanding accessibility for immersive instruction. Arif [82] developed a CAVE system for infrastructure management education, in which senior undergraduate students conducted virtual walk-through inspections and applied NBIS-style condition ratings to modeled bridge components. Evaluation with 69 students showed improved concentration and strong perceived proximity to full-scale structural elements.

Collectively, these studies demonstrate that immersive educational platforms support not only improved engagement and conceptual understanding but also measurable gains in procedural efficiency, situational awareness, and skill acquisition. Across bridge engineering, tunnel safety, inspection training, and infrastructure management education, these technologies enable repeatable and engaging learning experiences that are difficult to replicate through conventional instructional methods alone. By allowing high-risk scenarios to be rehearsed virtually rather than in active work zones or hazardous environments, and by using engaging instruction techniques to teach advanced construction topics, immersive education ultimately strengthens safety knowledge acquisition.

4. Discussion

The systematic review of immersive technologies in horizontal transportation infrastructure reveals a field undergoing rapid maturation, transitioning from conceptual exploration to data-driven, automated applications. The publication timeline illustrates a clear shift in research intensity. Between 2016 and 2020, the industry averaged only 0.8 publications per year within the review’s scope. This increased to an average of 10 publications per year between 2021 and 2025, consistent with the maturation of hardware and increased investment in immersive technologies [38,39]. Compared with literature on similar technologies in offsite construction [29] and general safety [19], research on horizontal infrastructure commenced later but experienced a comparable surge in activity in 2021. As such, the large increase in research interest reflects not only the growing technical feasibility of these systems but also a broader recognition of their potential to enhance safety, efficiency, and decision-making across the lifecycle of horizontal transportation infrastructure projects.

Virtual Reality was the most frequent medium (29 occurrences), primarily because it dominates simulation and offsite viewing environments, where full immersion is critical for replicating complex roadway scenarios. Augmented Reality (16 occurrences) and handheld devices (6 occurrences) are increasingly favored for onsite tools, where digital overlays must coexist with the physical job site. This medium-specific specialization is further reflected in the computational methods employed across the literature (Table 5), where biosignal processing and simulation-driven algorithms are concentrated in VR-based studies, while computer vision and spatial registration methods predominate in AR onsite applications.

The application of immersive technologies is concentrated on critical components of the horizontal transportation network, most notably bridges, roadways, and tunnels. However, each infrastructure component demands a distinct technological focus tailored to its unique operational challenges. Research on bridges primarily focuses on structural health monitoring and inspection, where AR is used for on-site defect detection [42,48,49,52], such as highlighting and measuring fatigue cracks, whereas VR is most often used for remote assessments of 3D data [17,25,26,61,67]. In contrast, applications for roadways and work zones more often shift the focus toward safety management and behavioral conditioning. Employing VR simulations to study worker-vehicle interactions [71,75,78,80,81] and the effectiveness of multimodal safety warnings in high-risk traffic environments [40,53,54,55]. Onsite roadway tools also prioritize real-time site management, such as using AR handheld devices for pavement management [64] and LiDAR-based scanning to document large-scale construction changes [51]. Tunnels constitute a specialized niche in which immersive environments are used to investigate site-specific environmental factors, including lighting uniformity [79], sight distance calculations under varying driving conditions [72], and emergency fire evacuation procedures [90]. This application-specific diversity indicates that the selection of an immersive medium is fundamentally governed by the technical or safety objectives of the infrastructure component under consideration.

A significant trend identified in this review is the increasing interest in moving VR/AR from a strictly visual tool to an active component of an automated data ecosystem. Researchers have integrated features, such as YOLO-based damage segmentation models, to automatically detect and measure structural defects, such as fatigue cracks [42,49,52]. This allows inspectors to see data-based insights while maintaining their human expertise.

While the reviewed studies collectively demonstrate that AR can meaningfully enhance automated defect detection in bridge inspection, the underlying computer vision architectures vary considerably, and their relative performance under outdoor, field-deployed conditions has not yet been systematically compared. By improving the speed and reliability of defect detection, these AR-enabled systems can reduce the time inspectors must spend in hazardous locations, thereby limiting their exposure to site-specific risks. Mohammadkhorasani et al. [48] and Mojidra et al. [49] both employed feature-point tracking algorithms, specifically the Shi-Tomasi detector paired with the Kanade-Lucas-Tomasi (KLT) tracker, to identify fatigue cracking. In contrast, Awadallah and Sadhu [42] and Pantoja-Rosero and Salamone [52] adopted YOLO-based detection architectures. These two approaches reflect different assumptions about the inspection, as motion-based methods are most well-suited for fatigue monitoring of steel structures under traffic loading, and detection-based methods are more broadly applicable across damage types and material structures. More work must be conducted to test these algorithms under the variable lighting conditions, complex backgrounds, and movement relevant to active horizontal construction environments as to most improve occupational safety.

Furthermore, immersive technologies are among the primary interfaces for maintaining dynamic digital twins of infrastructure [59,60,62,63]. Rather than relying on static BIM models, new frameworks use AR for automatic model registration [43,50], allowing virtual models to be updated in real time to reflect environmental changes or construction progress. The registration methods underlying these frameworks vary substantially in their technical approach. That is, localization and alignment approaches in the literature range from centimeter-level 6-DoF pose estimation via GNSS-VIO fusion [43], SLAM-based markerless AR alignment [44,46] and geometry-driven methods using RANSAC plane segmentation and ICP for LiDAR registration [51]. These contrasting strategies highlight a consistent challenge in horizontal infrastructure: open linear worksites lack stable reference features and variable lighting degrades visual tracking. Systematic evaluation of these registration approaches under horizontal construction field conditions remains an important direction for future research, as reliable real-time model alignment is necessary for advanced hazard detection and structural systems that most directly reduce inspectors’ exposure to traffic, fall, and confined-space risks.

Handheld AR devices are being used to scan and record large-scale changes via phone-based LiDAR [51]. Through interactive documentation, immersive technologies are positioned as a critical tool for future work in construction lifecycle management. Additionally, the literature increasingly recognizes immersive visualization as a natural medium for digital twin applications [18], given its inherent compatibility with three-dimensional spatial data.

From an accident prevention perspective, the reviewed technologies demonstrate measurable risk reduction across multiple hazard categories. AR-based defect detection systems reduce inspectors’ exposure to traffic hazards by enabling faster, more accurate structural assessments [42,48,52]. VR-based behavioral training mitigates risk habituation and improves worker response times to struck-by hazards [73,74]. Real-time AR warning systems enhance reaction times through multimodal alerts that directly target the leading causes of work zone fatalities [40,53,55]. Collectively, these findings indicate that immersive technologies serve as both predictive tools for accident analysis and proactive interventions for accident prevention.

The use of immersive technologies provides a unique platform for understanding workers’ cognitive and physiological responses in high-risk environments. The use of EEG and other biosignals in VR simulations has allowed researchers to identify how physiological markers are affected during high-risk tasks [69,74,88], such as construction near live traffic, in controlled, safe environments. As smart technologies are embedded into daily construction site operations, research highlights the importance of leveraging multimodal warnings to enhance worker safety. The literature demonstrated that safety technology achieves significantly improved outcomes when multiple sensory channels are combined [40,55,75,81]. Oftentimes, this means a combination of visual, auditory, and tactile alerts to ensure critical safety information is received even in sensory-taxing environments. However, these technological gains must be balanced against the risk habituation phenomenon, in which repeated exposure to simulated struck-by risks can desensitize workers to hazards [74]. Fortunately, targeted behavioral interventions that vividly illustrate the consequences of these risks can meaningfully reduce this effect [73], reinforcing the utility of VR for controlled behavioral conditioning before field deployment.

The transition of immersive technologies from controlled research environments to standard industry practice is fundamentally restricted by several factors. Xu et al. [57] reported a strong linear correlation (r = 0.91) between HMD price and weight across commercially available devices, indicating that the advanced processing power required for complex civil engineering visualization necessitates heavier, more expensive hardware. According to ANSI Z89.1, which sets 425 g as the maximum recommended weight for a Class A/C hard hat, the report identified that several high-capability devices exceed the referenced threshold. As such, there are significant ergonomic and cost barriers. Onsite use of immersive technologies will require advances in lightweight hardware to support prolonged use in active highway environments. Additionally, these ergonomic constraints have direct safety implications beyond worker comfort as heavy HMDs increase physical fatigue, which impairs attention and reaction time in active work zones. Furthermore, environmental challenges unique to horizontal construction, such as poor connectivity and unstable outdoor lighting, continue to hinder the stability of digital overlays [45,47]. While innovations such as laser-projection-based AR are emerging to address guidance in low-light conditions [56], the hardware’s inability to consistently perform in variable field environments remains a critical technical bottleneck.

This technical maturity gap directly feeds into a broader institutional resistance within the industry. Despite a sharp increase in academic publications, to an average of 10 per year, a 2022 survey of state Departments of Transportation revealed that only a single state had officially deployed immersive systems for inspection tasks, specifically for monitoring earthwork quantities [47]. This disparity suggests that bridging the implementation gap requires more than just hardware refinement; standardized protocols and significant investments in organizational readiness are also needed to move these technologies into standard practice.

The current literature demonstrates a robust foundation for immersive technology in horizontal construction, yet several critical gaps prevent its evolution from a niche research interest to an industry standard. While many studies focus on specific phases, such as design visualization [59,66] or defect detection [48,52], there is a notable lack of comprehensive frameworks that span the entire lifecycle of transportation infrastructure. Current research is often segmented, with limited interoperability between the digital twins used in construction and those required for long-term operation and maintenance. Future research should prioritize developing unified methodologies to ensure that data captured during construction remain functionally useful for management and maintenance years after project completion. Until immersive technologies are integrated into a unified data ecosystem, they will remain isolated methodologies rather than industry-wide standards.

Additional research is needed to develop intuitive, low-barrier interfaces for construction workers without specialized technical expertise, particularly in horizontal construction applications such as roadway, pavement, and utility projects. While current research shows that these technologies function in controlled settings [55,58,65], the high cognitive load of navigating complex menus or managing misaligned overlays can detract from primary safety and construction tasks [45], especially in large, traffic-exposed, and spatially distributed work zones typical of horizontal infrastructure. Future studies should prioritize user-centered design (UX/UI) tailored to these domain applications.

For wearable safety alert systems specifically, the reviewed literature suggests that haptic-auditory combinations at durations of at least 350 ms consistently produce the strongest worker safety responses [75], and where physiological monitoring is integrated, electrodermal response has been identified as the most practical real-time signal for distinguishing stress levels and informing adaptive warning delivery [69]. On the software side in particular, more research should be conducted to reduce the technical threshold for field deployment. Current immersive applications frequently rely on manual calibration workflows and software that requires specialized expertise to configure and maintain [45,47].

Future research should similarly explore the development of flexible, parameter-based design methodologies to ensure that immersive simulations are more generalizable across diverse infrastructure contexts, including varying roadway geometries, staging conditions, and traffic control configurations. By shifting from static, site-specific virtual environments to procedural models that allow variables to be programmatically adjusted, the industry can validate safety and operational protocols across a much broader spectrum of scenarios. Additionally, to ensure effective stakeholder adoption, these technologies should be systematically evaluated by industry experts [48,67], such as construction inspectors and health and safety managers working in horizontal construction environments, and iteratively refined based on their feedback.

It is noted that this review is based on only two data sources. Despite TRID and WOS being extensive databases, a limitation of this work is that relevant studies indexed elsewhere may not have been captured. In particular, WOS’s Conference Proceedings Citation Index indexes proceedings volumes through a publisher-submission process rather than comprehensively, so short-form conference papers common in computer science and HCI venues, such as IEEE VR conferences, may not have been fully captured. Based on a supplementary search of IEEE Xplore using the search strings described in Section 2, the vast majority of papers omitted from WOS are short-form conference papers that were consistently excluded in this review due to the publication quality threshold. However, future reviews in this domain may benefit from including databases such as IEEE Xplore to increase the visibility of preliminary and emerging findings.

Since articles were manually labeled based on immersive technology and tool class, some subjectivity is unavoidable and may affect the resulting classifications. Another limitation is that abstract screening and full-text review were performed by a single reviewer. However, the 10% spot-check mitigated this limitation by validating the reproducibility of the inclusion and exclusion process. Finally, no formal quality appraisal of the included studies was conducted. Instead, study quality was addressed implicitly by reporting evaluation methodologies and outcomes throughout the narrative review, and articles lacking a clear methodological contribution were excluded during full-text screening. This is a limitation of the current work, and future reviews would benefit from a domain-specific quality assessment framework.

5. Conclusions

This systematic review synthesized recent research on the application of immersive technologies, primarily VR, AR, and related extended reality tools, in horizontal transportation infrastructure construction. By reviewing 54 studies published between 2016 and 2025, this work addressed a critical gap in the literature by focusing specifically on roads, bridges, tunnels, and work zones, which present operational, environmental, and safety challenges distinct from those found in vertical construction. The findings demonstrate that immersive technologies are no longer conceptual tools in this domain but are increasingly deployed for inspection, safety management, simulation, and education, with clear evidence of technical maturation since 2021.

Across application contexts, distinct patterns emerged in both technology selection and functional use. VR dominates simulation and offsite viewing applications, where full immersion enables controlled experimentation, behavioral analysis, and remote inspection of complex spatial data. In contrast, AR and handheld devices are most prevalent in onsite tools, supporting real-time defect detection, model registration, and progress tracking within active construction environments. These differences highlight that the adoption of immersive technology in horizontal infrastructure is fundamentally task-driven, with infrastructure type and operational objectives governing both hardware and software choices.

The review also reveals a broader shift in immersive technologies from passive visualization interfaces toward active components of integrated, data-driven ecosystems. Many studies demonstrate integration between immersive visualization, computer vision, sensor networks, and digital twins, enabling automated defect recognition, real-time model updating, and interactive decision support. At the same time, immersive environments have proven particularly valuable for investigating human factors such as attention, cognitive load, stress, and risk habituation under hazardous work zone conditions that would be unsafe or impractical to study in the field. These capabilities position immersive technologies as uniquely powerful tools for both systems monitoring and human-factors research aimed at accident prevention.

Despite these advances, significant barriers continue to limit widespread industry adoption. Hardware cost, weight, and ergonomics constrain extended onsite use, while environmental factors such as connectivity limitations and lighting challenge the reliability of field deployments. Additionally, the review highlights a persistent gap between research and institutional implementation, with limited adoption by transportation agencies despite demonstrated benefits. Addressing this gap will require not only continued hardware innovation but also standardized workflows, interoperable data frameworks, and user-centered interface design that reduces cognitive burden for field personnel. Future research should prioritize lifecycle-spanning digital twin integration, scalable simulation frameworks, and validation through longitudinal field studies and expert evaluation to support the transition of immersive technologies from experimental tools to standard practice in horizontal transportation infrastructure construction.