The application of XR in contract management is not reflected in the previous research work. The application in cost control is only reflected in one study, which is not analyzed as a separate section. Therefore, the research hotspot analysis was synthesized from progress control, quality control, and safety management. According to these research hotspots, the authors proposed an overall analytical framework from two perspectives: (1) a management perspective and (2) a technical perspective. Specifically, the proposed analytical framework from the management perspective is based on the textbook of the Constructor Examination in China and the related literature on construction management [
24,
39]. Besides, the proposed analytical framework from the technical perspective is based on the XR technical architecture proposed by the China Academy of Information and Communications Technology (CAICT). The specific analysis framework is shown in
Figure 9. In addition, the analysis of each area is carried out from two aspects: (1) state-of-the-art studies of XR applications and (2) challenges related to XR applications.
4.2.1. Progress Control
Among the seventy-nine studies included in the analysis, nine studies were related to research topics in progress control. Overall, these relevant studies approach the progress control issues through the visualization of schedule, the automatic inspection of progress deviations, and the on-site visualization of inspection data. Moreover, among the nine XR systems developed in these studies, seven are AR systems, and the other two are VR and MR systems.
Regarding schedule visualization, on-site workers cannot interact with the 4D model (e.g., measuring distance, inspecting component types), which is not conducive to workers’ perception of the workspace [
40]. In terms of on-site progress inspection, the traditional progress inspection methods are affected by subjective experience and are not conducive to the real-time grasp of the site progress. To address the issues mentioned above, Tallgren et al. [
40] developed a VR-based project planning system that enables users to view and interact with 4D models in an immersive manner. Kopsida and Brilakis [
8] proposed an indoor progress inspection method based on MR to inspect the progress deviations automatically. AR has the potential for on-site monitoring, project information retrieval, and real-time information sharing [
4,
5,
6,
41]. Therefore, some researchers have developed AR applications to realize the in situ presentation of project information and progress inspection data on-site building entities [
7,
42,
43].
The above studies take XR as an in situ visualization tool and interactive tool for construction information. They have examined XR in the following engineering scenarios: (1) an immersive view of the construction process of complex building components and (2) in situ visualization of progress inspection data.
Using XR tools to assist progress control still has some application challenges. For example, Bae et al. [
6] tracked and located the position and direction of mobile AR devices by matching the on-site image with the pre-established 3D point cloud model and then enhanced the progress information associated with the building elements in the on-site image. However, this tracking and positioning method has two challenges: (1) high-resolution on-site images are conducive to accurate positioning, but there is a problem with the long matching time between 2D images and 3D point cloud models; (2) if AR devices only include building elements with few feature points (e.g., windows) in the camera’s view, the matching accuracy will be low. The above two challenges will lead to the low real-time level and accuracy of the site progress inspection information access. Kopsida and Brilakis [
8] inspected whether the 3D objects included in the as-planned model exist in the actual scene by comparing the as-planned data (referring to the pre-established BIM model) with the as-built data (the spatial surface meshes covering the real object captured by MR devices). In this way, the site construction progress can be inspected. However, this inspection method has the following challenges: (1) the accuracy of mesh data captured by MR devices is limited; (2) the scanning results of MR devices are affected by weather conditions, observation angle of the object, and object material type (e.g., tiling, or surfaces with strong light absorption). This inspection method is unsuitable for inspecting small irregular objects’ progress, scanning outdoor environments in sunny weather, scanning objects at a shallow angle, and examining the progress of construction tasks related to the material texture. The specific application challenges mentioned in the previous research work in progress control are shown in
Table 2.
4.2.2. Quality Control
Among the seventy-nine studies included in the analysis, a total of sixteen studies were related to research topics in quality control. Overall, these relevant studies approach the quality control issues through construction operation instruction and on-site quality inspection. Moreover, among the sixteen XR systems developed in the above studies, fifteen are AR systems, and the other is a VR system. Therefore, this section focuses on the application of AR in quality control.
Traditional manual-based operating instruction hinders the high integration of workers’ information retrieval and understanding and workers’ task operation. Traditional quality inspection methods are subjective, time-consuming, and cannot guarantee the accuracy of inspection results. Some researchers proved through experiments that AR can improve the information retrieval process and alleviate the problem of information overload [
44]. Hou et al. [
9,
10,
11,
45] developed AR systems to instruct workers in assembly or installation. These systems can improve the effectiveness of operation instruction from the perspective of reducing the number of visual transitions during construction [
10] and the obstacles to workers’ understanding of complex information [
11]. Park et al. [
12,
46,
47,
48,
49,
50] proposed the construction defect inspection method based on BIM and AR, which realized the automatic on-site inspection of the size deviation or omission of building elements. Chi et al. [
13,
51] developed an AR-based method for inspecting and repairing reinforcement defects, which realized the visualization of reinforcement inspection results and rework instruction. Liu et al. [
52] developed a prototype of an AR system to realize the integration of the BIM model and unmanned aerial vehicles (UAVs) aerial video, which can carry out comprehensive quality inspections of buildings.
The above studies take AR as an on-site retrieval tool for quality information. They have examined AR in the following engineering scenarios: (1) in-process instruction of construction quality (e.g., virtual annotations-based on-site assembly or installation instruction), and (2) post-inspection of construction quality (e.g., automatic on-site inspection for dimensional deviations of building entities).
Using XR tools to assist quality control still has some application challenges. For example, Fazel et al. [
45,
51] found that the field-of-view provided by Helmet Mounted Display (HMD) was limited when they tested the effect of AR tools on workers’ assembly instruction. It indicates that using AR instruction tools could limit the ability of workers to perceive their surroundings at complex construction sites. Park et al. [
47,
49] used the method of combining AR, BIM, and image-matching technology to automatically inspect the defects of building elements caused by size errors at construction sites. However, the effectiveness of the inspection results of this method heavily depends on whether the angle of the cameras of mobile AR devices at construction sites is consistent with the angle of the pre-established standard images. This method has the limitation of low matching accuracy and long matching time between images from different angles. Moreover, this method lacks an effective real-time transfer mechanism of inspection results based on AR annotation between inspectors and on-site workers. Yao et al. [
53] combined sensing devices to collect the position information of the impact compactor and carried out real-time visualization on the VR management platform. This method realizes a real-time three-dimensional interactive display for the virtual scene’s compaction process to control the operation’s quality. However, this method has the limitation of low transfer efficiency of position information between the mobile station of the Beidou Navigation Satellite System, cloud server, and VR management platform. The specific application challenges mentioned in the previous research work in quality control are shown in
Table 3.
4.2.3. Safety Management
The contents of construction safety management include hazard identification and prevention, safety education and training, safety inspection and instruction, and safety accident handling. The typical applications of XR-based construction safety management include three aspects of hazard identification, safety education and training, and safety inspection and instruction [
24,
39]. Therefore, the studies related to safety incident handling retrieved in this review are limited. Among the seventy-nine studies included in the analysis, fifty-three studies were related to research topics in safety management. Moreover, among the fifty-three XR systems developed in these studies, forty-three are VR systems. These VR systems are used for hazard identification and safety training, and the other AR or MR systems are used for safety inspections.
The traditional hazard identification methods are challenging to identify unsafe construction environments or unsafe worker behaviors from the perspective of construction participants. The combination of BIM and VR can assess the rationality of a workspace or observe the behavior of construction participants by simulating site conditions and designing immersive interactions. The VR-based safety evaluation method can improve the authenticity of on-site evaluation. For example, Sydora et al. [
54,
55,
56] developed VR-based crane simulation systems to evaluate the availability of construction space during the lifting process. Getuli et al. [
57] used VR to simulate construction activities (e.g., plate installation), which can realize the identification and replanning of unreasonable installation space. In addition, VR plays a stimulus role for the participants. VR can be used to observe the reaction and behavior of construction participants in a virtual hazardous environment and participants are not exposed to the real risk of physical injury [
58]. For example, Kurien et al. [
14,
15,
59] create VR construction environments and simulate construction tasks to identify unsafe operating positions by capturing workers’ movements during operation. Habibnezhad et al. [
60,
61,
62,
63,
64] combined VR with physiological sensing to extract the physiological reaction data of workers in virtual hazard scenarios and then form representative priors of unsafe actions. Through experiments, Tixier et al. [
62,
65] proved that even if workers can correctly identify construction site hazards, they will still engage in dangerous behaviors.
The above studies take VR as a research tool to help the safety management team extract and analyze the potential risks of the construction site to explore how to develop and implement effective safety training and safety inspection plans. They have examined VR in the following engineering scenarios: (1) identifying unsafe operating postures for workers combined with ergonomics, and (2) identifying unreasonable spatial layouts in crane operation.
- 2.
Safety education and training
Traditional online safety education and training methods lack interaction and feedback, whereas on-site safety education and training cannot ensure the safety of workers. Nykanen et al. [
66,
67,
68] have proved through experiments that VR-based safety education and training can improve workers’ safety performance. Therefore, VR-based safety education and training methods are currently advocated, which will not bring any actual risk while improving training effectiveness. Many studies on machinery operation training and construction technology training are based on immersive VR systems [
25,
69,
70]. For example, Song et al. [
16,
17,
71] developed VR-based training systems for workers and machinery to work together. Then, the systems are evaluated from the aspects of system availability and workers’ psychological indicators, and the conclusion is that the VR-based safety training system is effective. Joshi et al. [
72] developed an immersive VR training system for prestressed concrete construction to train workers in the three aspects of wearing personal protective equipment, avoiding dangerous work areas, and performing tension operations. The above studies are related to the development and evaluation of safety training systems, and the existing studies also involve the formulation and evaluation of safety training content. For example, Shi et al. [
73,
74] proposed that theories (e.g., cognitive retrieval and memory method and positive reinforcement learning theory) could be embedded in the safety training content to enhance the training effectiveness.
The above studies use the assumed construction scenarios in VR systems to convey abstract safety knowledge to workers to enhance workers’ perceptions of hazard sources and understanding of safe behaviors. They have examined VR in the following engineering scenarios: (1) VR safety experience pavilion and (2) construction machinery (e.g., crane, forklift) operation training and construction technology training.
- 3.
Safety inspection and instruction
Safety inspection mainly refers to the inspectors inspecting the conditions of workers, equipment, and environment during construction and conveying risk information to workers on time [
24]. Traditional safety inspection methods are affected by the subjective experience of safety inspectors, and the effectiveness of risk information transfer between inspectors and on-site workers is low. Therefore, some studies have studied automated methods for on-site safety inspections. For example, Atherinis et al. [
26,
75] proposed a VR/AR-based method of automatic inspection of collective protective equipment (CPE), which can improve the accuracy of inspecting the number or location of equipment components. Chen et al. [
76] integrated AR, BIM, and path planning into the crane remote operation system, which can automatically identify potential risks (e.g., overload) and issue warnings. Fenais et al. [
77,
78] developed VR/AR-based mobile platforms to capture information about underground facilities, which can inspect whether the earthwork excavation route will pose a threat to underground facilities in real time. Dong et al. [
79,
80] combined VR systems, sensing devices, and real-time positioning systems to realize real-time inspection of workers’ unsafe behaviors (e.g., misuse of personal protective equipment (PPE)). Promoting the real-time transfer of risk information and visual interaction for safety inspection is also important. Wu et al. [
18,
19] developed AR/MR-based real-time alarm systems to provide workers with real-time safety warnings in environmental vision to help them determine their safety conditions.
The above studies use XR tools to automatically inspect safety facilities at construction sites and enhance hazard information in the real environment. They have examined VR/AR/MR in the following engineering scenarios: (1) automatic inspection of collective protective equipment, (2) monitoring and warning of unsafe personal conditions, and (3) in situ real-time visualization of underground public facility information.
Using XR tools to assist safety management still has some application challenges. For example, Shi et al. [
15,
74] used the vision-based motion capture method to capture the movement information of workers working in VR environments to identify unsafe operating postures. However, this method has limitations of occlusion, motion range limited by sensor depth range, and motion speed limited. Pooladvand et al. [
55,
70] allowed workers to perform operations (e.g., lifting tasks and handling construction material) in the VR system to identify unreasonable workspaces at construction sites (e.g., the potential collision between lifting objects and surrounding objects) and unsafe behaviors of workers (e.g., cutting into dangerous areas). However, the VR system can only provide visual stimulation for users and only support single-user and single-device interaction. This system has limitations of lack of auditory and tactile feedback and not supporting multiuser interaction with multiple devices simultaneously. Kim et al. [
19] transformed and matched the workers’ perspective image captured by wearable AR devices with the global perspective image captured by the construction site camcorders to track and locate the position and direction of AR devices. This study has calculated, extracted, and enhanced the hazard information of the distance between the workers and the dangerous equipment and the direction of the dangerous equipment relative to the workers. However, challenges still exist, such as large errors of image transformation from different perspectives, occlusion, and high dependence on network and computing systems. These challenging problems could reduce the effectiveness of hazard information and its real-time transfer. Ramos-Hurtado et al. [
26] used mobile AR devices to compare the BIM model of Collective Protective Equipment (CPE) with the real elements at construction sites to inspect whether the safety equipment exists. However, in this method, the positioning of the BIM model is realized by manual placement after planeDetection with mobile AR devices, which has the limitation of low positioning accuracy of the model. This method is not suitable for inspecting the size of safety equipment. Wu et al. [
18] used MR to anchor the virtual hazard warning sign and the real hazard source to transfer hazard information to on-site workers. However, this method has not realized the automatic alignment between virtual signs and real hazard sources and the sharing of spatial coordinates among different MR devices. The specific application challenges mentioned in the previous research work in safety management are shown in
Table 4.
The above analysis has obtained the specific application challenges related to XR in the construction management domain. Finally, it is necessary to summarize and rank these application challenges mentioned in the previous research work, aiming to find the key application challenges and pave the way for the analysis of the future research direction of XR in
Section 5. According to the proportion results, the highest-ranking application challenges mentioned in the previous research work are poor perception and interaction of XR. In addition, other challenges of XR include (1) adverse physiological reactions to users (e.g., dizziness), (2) The differences between the virtual scenarios and the real world are still large, and so on.
Table 5 shows the summary and ranking of application challenges related to XR in construction management.