Literature Review

Posted: January 5th, 2023

2.0 Literature Review

This chapter reviews the existing literature related to artificial intelligence technologies and their application in the construction industry, and which is being and can be specifically used to enhance construction safety. After exploring what the literature says about artificial intelligence and its applications in the construction industry, specific information related to the application of artificial intelligence technologies to promote construction safety is discussed.

2.1 Construction Safety

Construction safety is a highly researched topic in response to the high prevalence of construction accidents, injuries, and fatalities. Many researchers have documented diverse issues related to construction safety, such as hazard prevalence, diversity of incidences, risk assessment, risk detection, and risk reduction. For instance, Sanni-Anibire, et al. (2020) noted that different countries had different prevalence of construction-related accidents and injuries. The United States and Ireland led the world in construction incidences, where construction sites accounted for 50% of all the injuries and accidents in the two countries. However, although countries like the United Kingdom and Japan had 25% and 40% of their safety issues attributed to the construction industry, respectively, the construction side remain the most dangerous workplace across all industries. Therefore, Sanni-Anibire, et al. (2020) supported an increased focus on construction-related safety to reduce injuries and fatalities, and hopefully achieve risk-free construction sites.

A risk assessment performed by Sanni-Anibire, et al. (2020) revealed that falling objects presented the highest safety risk, while high winds were the most significant cause of this hazard in a construction site. In contrast, falls, trips and slips registered the best safety performance in construction sides in Saudi Arabia. Li (2018) noted that traditional approaches to construction safety were gradually giving way to innovative automated systems. Common automated technologies being employed in the construction industry to enhance safety include building information modelling, robotics, additive manufacturing, software engineering, virtual reality, and internet-of-things. However, Sherratt and Dainty (2017) revealed what they termed as the ‘zero paradox’ in which major accidents were more likely in construction sites of large companies that have adopted the zero accident or injury policy than in sites that did not have a policy. They used evidence gathered from construction sites in the United Kingdom, which demonstrated that the zero policy was not only ineffective in enhancing site safety but was also pointless in some case. These findings direct attention towards realistic approaches to construction site safety using quantitative data rather than managerial branding strategies that did not influence safety practices at the construction sites. From a different perspective, Alruqi and Hallowell (2019) argued that safety performance in construction sites could be reliably assessed using safety critical success factors that targeted standards such as personal protection equipment, owner involvement, safety resources, safety staffing, safety inspections, safety record-keeping, safety observations, and safety incentive programs. Therefore, the effectiveness of any construction safety interventions should be critically assessed using these critical success factors and any safety enhancing technology should pass this test to prove its worthiness for adoption in construction sites.      

2.2 Construction Safety Technology

Many studies have documented various technologies being used to enhance construction safety. Some of these technologies have been adopted from other industries, while others have been developed for the construction industry exclusively. For instance, Shohet et al. (2019) demonstrated that information and communication technology could be used to make systems that simultaneously promote safety and quality at construction sites. They noted that the construction industry regularly struggled with promoting construction site safety without compromising the quality of work. In a study that used a mobile application for commanding, controlling, and communicating in construction sites. Shohet et al. (2019) showed the system could reduce unsafe construction site practices by 90 % while delivering 30% improvement in quality. Similarly, Mneymneh, Abbas and Khoury (2017) revealed that the computer-based tools commonly used to enhance safety in construction sites included computer vision systems using processes to detect hardhats worn by construction workers on site. However, these systems over-predicted and falsely unwanted identified objects as hardhats, thus presenting inaccurate detections. Besides, these systems required to be used along with closed circuit television (CCTV) and were yet to be tested in real-life conditions in a construction site where lighting conditions, colour, image clarity, background contrast, and orientation varied widely in construction sites. However, the findings from the experiment conducted by Mneymneh, Abbas and Khoury (2017) revealed that well-trained cascade classifier was robust in detecting hardhats in indoor construction sites rapidly and efficiently. However, they noted that further research was needed to improve the helmet detection process and reduce false detections that combined image and colour differentiation techniques with classifiers. Similarly, Li, et al. (2018) noted that enlarged reality and virtual reality technologies were being widely researched upon for their potential in enhancing construction safety, paving way for their application in the existing construction projects, which has become increasingly complex and hazardous. These technologies were being used to hazard identification, safety inspection, and safety education and training within the construction industry. However, they noted that the adoption of these technologies could benefit from multidisciplinary research. In the same manner. Marefat, Toosi, and Hasankhanlo (2019), evaluated the influence of building information modelling (BIM) on construction safety and the barriers that delayed its adoption in construction projects. Their study revealed that although BIM enhanced construction safety, its adoption was undermined by the lack of social infrastructure, well-trained personnel, proper guidance, and government support. However, Riaz et al. (2017) revealed that BIM could be made more effective by integrating it with wireless sensor technology. They noted that these two technologies were compatible provided the hardware and software were selected precisely and optimised accordingly. In this regard, Park, Kim, and Cho (2017) demonstrated the workability of an integrated automated safety system employing BIM and Bluetooth low energy (BLE) mobile tracking sensors, which can operate from the cloud.  

With the construction workers being responsible for their personal and workplace safety, some technologies have targeted the monitoring of safety behaviour and observance to safety protocols by construction workers in construction sites. For instance, Awolusi, Marks and Hallowell (2018) reviewed wearable technologies already in use for personal safety in construction sites, although the construction industry remained behind other industries of the adoption of wearable technologies. They noted that the integration of two or more sensors could help overcome the limitation of each one and complement each other, conferring more benefits in safety enhancement. They also advised that research should venture into the conversion of commercially available wearable devices to construction-safety specific ones. Besides, Awolusi, Marks and Hallowell (2018) argued that wearable technologies were well-suited for personalised safety monitoring in construction sites. Similarly, Rubaiyat et al. (2016) investigated the use of computer vision and machine learning methods for automatically detecting helmet use in a construction site. Their study revealed that a system using colour detection techniques could accurately detect white, red, blue, and yellow construction helmets and differentiate them from caps. However, this system could be scaled up using upper body systems to detect wearers of construction helmets of all colours with better accuracy to promote observance to helmet safety protocols in a construction site.  

Many construction safety technologies rely on data obtained from various hazard situations at construction sites. Tixier, Hallowell, and Rajagopalan (2017) acknowledged the benefits of using quantitative data for safety risk analysis rather than subjective data obtained from managerial perspectives, which was risky with biases that undermined safety decision-making in construction sites. They also argued that quantitative data was critical in automating construction safety interventions because it accounted for various situations of hazards in construction sites.  Similarly, Ayhan and Tokdemir (2020) revealed that artificial neural networks (ANNs) and latent class clustering analysis (LCCA) could be used to predict construction incidents and their outcomes, facilitating the formulation of practical construction incident prediction systems and interventions. 

Gheisari and Esmaeili (2016) noted that several studies had suggested the application of technologies like the global positioning system (GPS) and radio-frequency identification (RFID) for reducing potential human errors and enhancing safety performance in construction sites. However, they hailed the unmanned aerial system (UAS), otherwise known as drone technology, as an emerging technology that held great promise for enhancing construction safety due to its inherent advantages. Gheisari and Esmaeili (2016) argued that the ability of drones to move faster than human beings, the capability of reaching inaccessible areas in construction sites, and the possibility of equipping them with wireless sensors, video cameras and communication hardware for transferring data in real-time, gave drones a clear advantage over other construction site technologies. Specific safety practices that could benefit from the unmanned aerial system include operating in the blind spots of heavy construction machinery, in unprotected openings and edges of construction structures and in areas populated by cranes and boomed vehicles. Besides, drones could be used for safety inspection in construction sites that were hazardous to human inspectors. This is because drones had features, such as sense-and-avoid capabilities, high-precision outdoor navigation, and real-time video communication using video sensors. They recommended that unmanned vehicles be adapted for the construction industry to enhance further the safety of complex construction projects. 

2.3 Artificial Intelligence in Construction Safety

To understand its application in the construction industry, it is critical to understand artificial intelligence and its related technologies. Rao (2021) described artificial intelligence as an aggregative terminology describing the mimicry of human cognitive functions by machines, while machine learning was the application of statistical techniques to facilitate machines to learn from data without being programmed. Consequently, machine learning as part of artificial intelligence technology and algorithms were the central mechanisms of machine operations. Prieto (2019) revealed that artificial intelligence leveraged big data and was already being used in self-driving cars and intelligent assistants.

The construction industry is renowned for being a late adopter of new technologies. Despite being valued at more than $10 trillion annually, estimated to maintain a growth rate of 4.2% through 2023, and having sophisticated clientele, the engineering and construction industry was significantly under digitised (Blanco, et al., 2018; Rao, 2021) observed that over Blanco, et al., (2018) goes on to note that the construction industry had lagged behind the automotive, transport and logistics, and telecommunications industries, particularly in the adoption of artificial intelligence technologies. However, Blanco, et al. (2018) revealed that the construction industry could adopt artificial intelligence-powered applications from other sectors. The gave the example of reinforced learning algorithms from the transport industry, which could be used in project planning and scheduling and predictive artificial intelligence solutions from the pharmaceutical industry, which could be adopted for forecasting outcomes and lowering research and development costs through project risk forecasting, constructability determination, and structural stability assessment. Likewise, Niu et al. (2019) revealed that occupational health and safety (OHS) systems in the construction industry has evolved through two waves already; a first wave incorporated hard technologies, such as personal protective equipment for construction workers, while the second wave was characterised by soft technologies, which included the development of a safety culture in construction sites using innovative managerial approaches for addressing construction site hazards. However, a third wave was underway, in which artificial intelligence technologies were being used to improve construction safety. In this regard, Niu et al. (2019) developed a smart construction object enabled occupational health and safety management system (SCO-enabled OHS management system), which leveraged smart elements, such as autonomy, communicativeness, and awareness to identify hazardous situations in a construction site and respond to them quickly and freely. Their smart system was tested in a smart tower crane, which provided numerous data to facilitate developing of an artificial intelligence-based system that could allow a group of tower cranes near identify and respond to hazardous situations at construction sites promptly and autonomously, preventing the occurrence of fatal accidents (Niu et al., 2019). Niu et al. (2019) noted that their study paved the way towards developing a smarter and safer construction site empty of fatal accidents.     

Pan and Zhang (2021) noted that artificial intelligence was increasingly being applied in construction engineering and management (CEM) and had become the focus of demanding research contemporary research activities. Their critical review of existing literature revealed that on-going research was concentrated process mining, natural language processing, information fusion, intelligence optimization, computer vision, and knowledge reasoning and representation to promote data driven applications. Similarly, it revealed that the construction technologies employing artificial intelligence that were overrepresented in the literature published in the last decade, were digital twins and building information modelling. However, Pan and Zhang (2021) advised that future research should focus on 4D printing, artificial intelligence of things, cloud augmented and virtual reality, block chains, and smart robotics, which would help bridge the divide between construction engineering and management and artificial intelligence technologies, and promote the automation of construction process. When explaining 4D printing, Pan and Zhang (2021) noted that this technology was an enhancement of the 3D printing often used to make construction models and help visualise safety risks. It goes ahead by creating 3D models that can interact with the environment and change when certain environmental parameters, like wind, temperature, and loads  are changed, thus helping identify structural weakness and safety risks in the structure (Pan & Zhang, 2021). Similarly, in predicting the trends of innovative solutions in the construction industry between 2016 and 2018, Kapliński (2018) predicted that increase and virtual reality technologies would gain prominence because of their potential for monitoring construction sites, detecting safety risks and error in advance, and preventing costly mistakes. Similarly, Mathern, Ek, and Rempling (2019) argued that in an industry where engineer shortages were growing, artificial intelligence agents that could mimic engineers by learning to perform tasks, such as structural design, would become increasingly relevant in the future. Although, Mathern, Ek, and Rempling (2019) noted that trained artificial intelligence processes could be used to deliver sustainable designs to match the environmental requirements of the construction industry. Similarly, Mohammadpour, Karan and Asadi (2019) predicted that smart robotics, blockchains, and artificial intelligence of things (AIoT) were the next frontier of artificial intelligence application in construction engineering and management.     

Mohammadpour, A., Karan, E. and Asadi, S., 2019) addressed the challenges of implementing artificial intelligence technology in the architecture, engineering, and construction industry, which was highly under digitised, and noted that inappropriate information gathering, and poor selection of appropriate artificial intelligence techniques were the two most significant in barriers obstructing technology adoption.

Vision-based methods also include using LED rings, camera modules, real-time tracking, and detection, using stereo systems, and including videos. The goal is to have technology that tracks a persons’ movement and sensors their activities according to their environment. For instance, the real-time visual technologies would warn individuals about impending danger and facilitate precise risk assessments.  Computer vision technology is critical in identifying errors and providing an active approach towards construction safety. Guo & Skitmore (2017), classifies various technologies, such as infrared centres and computing videos and simulations that can be classified in this category. For instance, the infrared lights can reflect on a person movement and help them identify they almost stepped on an object. Sensors may trigger warnings whenever there are objects that can cause accidents on the ground. Additionally, the cameras will provide extra vision that ensures safety protocols are followed by employees. They also showcase workplace routines that help to establish how injuries will occur. Guo & Skitmore (2017), also highlighted that visualization technologies are used to enhance training opportunities for construction workers. 

Fire is among the most frequent problems affecting construction sites and buildings. According to Khan et al. (2020), traditional approaches to fire prevention are insufficient to handle risks, especially at the building stage when construction workers work on systems. Hence, the authors suggest using computer visions that provide visual programming tools to detect any problems with fire systems and recommend changes. Real-time visual technology monitors the building or site and provides plans based on the visualized programming tools (Li et al., 2018). It also examines potential errors and facilitates suggestions that promote safety. The technologies can also be used to identify hazards, enhance inspection, and continually monitor the construction site to remove objects that can cause accidents and falls.

Natural language processing technologies also impact construction safety by providing data that guides recommendations. Tixier et al. (2016), suggest that the attribute-based approach to describe injuries uses as a minimal description of the circumstances under which employee safety was compromised. The method does not provide a context-based analysis, facilitating the need for natural language processing technologies. NLPs provide automated content analysis that allows construction experts to examine large-scale injury reports and have specific data detailing their occurrence (Kim & Chi, 2019, p. 5). The technology uses procedures that classify language characters and images to provide a perfect narrative that enables appropriate recommendations. NLPs encompass technologies that can process natural language and represent it in data to be used for quality improvements.

After an injury occurs, a detailed report indicates what happened and provides as much information as possible to enable appropriate action. Zhang et al. (2019) argues that text mining and NLP technologies are ideal in interpreting texts and ensuring they are correctly represented in policies and decisions about workplace safety. Construction safety begins with an understanding of injuries because they ultimately guide decision-making. NLPs often provide algorithms that read injuries and identify objects that cause them (Zhang et al., 2019). For example, supervisors may use these technologies to explore how a fall occurred at the workplace. They could discover that the actions leading to this accident started with an earlier action from a construction worker who left an object unattended on the ground (Li et al., 2018; Zhang et al., 2019). Recommendations would suggest protocols that direct workers on how to safeguard the tools they use and proper storage. All this is only possible with NLPs that enhance a detailed understanding of injury.

Even though this research is still at the developmental stage, some studies have been conducted to determine the effectiveness of this method in construction safety. For instance, Guo et al (2018) tested object tracking using a computer vision technology in construction safety. The research entailed object recognition or tacking and sensory warnings whenever a risky situation was discovered. Applications indicated a 70% effectiveness rate for these visualization technologies. Accident rates reduced significantly when objects on the surface are tracked using visual technologies and the framework for this entailed recognition and tracking, assessments, and prediction at the developmental level, the functional level entailed tracking and recognition with key questions such as: is this object harmful? Where is this object? What is the next action for this object? Would this action be considered unsafe? Such questions led to the development of a framework in which object identification, tracking, and assessment reduced construction injuries by nearly 70%. The study also shows that this process enhances a safety culture and proactive management regarding object placement. Further studies reflect a continuous use of visual technologies such as sensors and infrared lighting to track objects and reduce accidents.

Another experiment required using NLPs to enhance construction safety by exposing accident reports. Previous data indicates that 30% of construction safety issues are related to poor reports preventing appropriate management practices. Therefore, Zhang et al (2019), provided data regarding the effectiveness of NLPs, which can be used for a framework addressing construction safety. Their research explored text mining in 1000 accident reports to discover common problems that prevent a proactive approach. Results indicated that data mining improves how reports are understood by more than 50%.

The language translated by the NLP technology would enable the construction site to identify what caused the accident and determine the most effective way to reduce or prevent a similar injury in the future. Moreover, the NLP would determine whether a chemical, motion, electricity, or another element was directly responsible for the accident.

2.4 Chapter Summary

To conclude, the reviewed literature demonstrated that the construction industry was considering new technologies, such as artificial intelligence, to improve safety performance in construction sites. However, the adoption of artificial technology in construction safety lagged that in other industries and many applications were still at the experimental and development stages. However, the literature indicated glaring gaps between conceptualisation of applications of artificial intelligence in construction safety and the actual application in construction sites. Despite these gaps, the literature revealed that the research questions could be answered using information from existing secondary data.   

2.0 Methodology

This chapter outlines the methodology employed in conducting this secondary research. It is guided by the aim of this study, which is to develop a framework for using artificial intelligence to enhance safety in the construction industry. Since this study intends to i) identify artificial intelligence technologies that can enhance construction safety, ii) explore the barriers in using artificial intelligence technologies in construction safety, and iii) develop a conceptual framework that will direct the use of artificial intelligence in construction safety, qualitative research that leverages the secondary data already available, is reasonable and justifiable. The rest of the chapter is structured to present the various components of the methodology used. It commences with the research design, followed by the and sampling techniques, and then, the data collection methods used in the study. Finally, the procedure used to analyse the data to generate meaning is outlined.

3.1 Research Design

Meaningful research should have a framework of methods and techniques suitable for answering the research question. The descriptive research design was selected for this study for its suitability in addressing the research question, which is whether artificial intelligence can be used to enhance construction safety. While there have been significant advancements in vision-based technologies and natural language processing, the related ideas and theories need to be expanded, explored, and explained to suit improving safety in the construction industry. In this regard, the descriptive research design should help elaborating on the unexplored aspects of construction safety using artificial intelligence and fill in the existing gaps in the construction industry (Silverman, 2021).

3.2 Sample and Sampling Techniques

Secondary data is used in this study. Secondary data is data that has already been collected in previous studies or accrued by organisations and stored in their publications or sources, individually. Since a huge amount of data on artificial intelligence technology and its application in various industries already exists, data that is relevant for providing understandings into the application of artificial intelligence in construction safety needs to be filtered systematically.

It was intended for the data sources to cover as many themes related to artificial intelligence and construction safety as possible. However, it was anticipated that some secondary sources would have similar data and therefore, not add value to the study. Therefore, stratified sampling was the preferred sampling method, which was selected because it helped categorise the data sources based on the themes related to artificial intelligence technologies applicable in the construction safety (Silverman, 2021). Therefore, before searching for the secondary sources, specific divisions representing thematic areas based on the types of artificial intelligence-enabled technologies applicable in the construction industry were identified first. The thematic categories include vision-based technologies and natural language processing technologies. After that, secondary sources in each thematic category were sampled randomly until saturation was attained. In other words, sources were retrieved until they provide new information related to the thematic area. This sampling approach was selected because it guaranteed greater representativeness of the different artificial intelligence technology applicable in the construction industry and that can be adapted for safety purposes (Saunders, et al., 2018).

Secondary sources were retrieved based on predetermined sampling criteria. Secondary sources were accepted if they were a) from reputable sources, published within the last five years, published in English, and addressed specific artificial intelligence technologies (Silverman, 2021). Reputable sources of secondary data included governmental agencies, construction and technology companies, and reputable online databases. These included peer-reviewed journal articles, industry reports, and company reports provided reliable data. In addition, social artefacts published by reputable people and organisations, which included letters, commentaries, interviews, and news delivered current information. Similarly, 5 years since publication ensured that only current sources were retrieved, while the English language of publication guaranteed that the researcher, supervisors, and most of the target audience could understand the sources. Further, relevance was guaranteed by ensuring that the secondary sources addressed specific artificial intelligence technologies applicable in the construction industry.

3.3 Data Collection

Data was collected from the secondary sources. The researcher read through the publications to identify the relevant data and information, which was then highlighted using the computer highlight feature. Before mining the sources for data, the secondary sources were retrieved from online sources and downloaded into a computer (Corti, 2018). Google Scholar and Microsoft Academic were the search engines used to query the various online data repositories. Besides, valuable information was retrieved from the websites of government agencies, private construction companies, and technology developers and venders. Keywords like artificial intelligence in the construction safety, vision-based technologies for the construction safety, natural language processing in the construction safety, and technology for construction safety were used to narrow down the search to relevant sources. The secondary data collection using online tools was preferred because it saved time and resources, while accessing huge amounts of publications from diverse national and international locations under severe resource constraints (Johnston, 2017; Silverman, 2021). Secondary sources also delivered data obtained from diverse methodologies used in different primary studies. This presents depth into the quality of data available, while saving time that would have been used to perform the numerous primary studies to obtain the same data first-hand. Besides, the methodologies of acquiring and analysing the data in the secondary sources saves time and effort that would have been used experimenting with different methodologies and leading the studies prior to primary data collection. Besides, the approach was appropriate to the prevailing restricted research environment occasioned by the on-going coronavirus pandemic, in which minimal interaction, unnecessary travel and avoidable contact among people is encouraged to stop the spread of infections.

3.4 Data analysis

Thematic analysis was employed in this study as part of the secondary analysis. This approach identifies and interprets the partners of meanings emerging from the secondary, which are condensed into themes and subthemes (Ruggiano and Perry, 2019). This analytical technique is systematic, with technical steps that begin with familiarising with the data by perusing the secondary sources, coding, generating themes from observed patterns, reviewing these themes, and defining and naming the emerging overarching themes and subthemes before writing up the analysis report (Ruggiano and Perry, 2019). For the coding process, the reflexive approach was used because it is flexible and therefore accommodates new themes and modification of the existing ones as the analysis proceeds.

3.4 ChapterSummary

The qualitative study used secondary sources for data related to the artificial intelligence technologies that can be used to enhance construction safety. The sources were retrieved online and sampled into categories reflecting the different types of artificial intelligence technologies using the stratified sampling technique. After that, thematic sampling was used to analyse the data from the secondary sources that can be used to meet the study objectives. 

5 Discussion

5.1 Introduction

In this chapter, the findings presented in chapter 4 are contextualised alongside existing literature. The discussion is topically arranged according to the objectives of the study, which are:

1. Identifying AI technologies that can enhance construction safety.
2. Exploring the barriers presented when using AI in construction safety.

5.2 AI technologies for construction safety

The results revealed that the construction industry was employing several artificial intelligence technologies to enhance construction safety. These technologies included robotics, drones, computer vision, and machine learning. The media has highlighted the increased application of artificial intelligence technology in the construction industry to enhance construction safety to support what has been documented in peer-reviewed secondary sources and construction industry reports. For instance, Walch (2020) of Forbes indicated that the construction industry had evolved considerably by adopting several artificial intelligence technologies. Specific technologies include autonomous robots to perform dangerous tasks at the construction site, drones to continuously monitor construction sites and identify hazardous situations, and machine learning to analyse huge amounts of data and identify potential risks to safety faster than human beings (Walch, 2020).

However, these technologies were mostly applied by large construction firms with the resources to invest substantially in this area. For instance, Skanska, a large Swedish-based construction company, has enhanced its construction safety and promoted an injury-free construction environment using an image and video analytics platform developed by SmartVid (Pemberton, 2019). In addition, several technology firms are building artificial intelligence applications that can be incorporated into the construction industry to enhance construction safety. For instance, Newmetrix, a Boston-based firm, develops algorithms that analyse construction site photographs for construction workers not wearing protective equipment and raising timely alarms. The software can also compute construction site risks to help project managers institute safety measures (Rao, 2021).

Built Robotics, is a California-based firm that develops and builds robotic upgrade kits fitted onto construction equipment to improve their safe deployment. This company has fitted bulldozers, skid steers and excavators with sensors and software that enables them to operate autonomously to prevent construction workers from working in hazardous construction sites (Pro Crew Schedule, 2021).  Likewise, Versatile Natures is a California-based technology company that develops artificial intelligence and internet-of-things platforms that collect and analyse data at the construction site and help construction project managers to make timely and informed decisions on enhancing construction site safety (Pro Crew Schedule, 2021). Similarly, Nucon, a tech firm based in Singapore, has developed an artificial intelligence engine that helped construction firms learn from their mistakes and avoid unsafe practices at the construction site in future projects (Pro Crew Schedule, 2021). Also, Astralink, an Israeli firm, has integrated augmented reality (AR) and building information modelling (BIM) using deep learning and computer vision to prevent the occurrence of construction errors in real-time, thus enhancing construction safety (Pro Crew Schedule, 2021).

However, many other artificial intelligence technologies were in the experimentation stage because they exhibited potential for use to enhance construction safety. These technologies are developed by large construction companies working with software companies that are skilled in developing artificial intelligence solutions. Knight (2019) reported about an artificial intelligence system that was being developed by Boston-based Suffolk in collaboration with a computer vision company, SmartVid and EmTech Next. This system shall use deep-learning algorithm that has been trained using construction site accident records and images. It is expected to monitor the safety of a construction site by identifying and flagging situations that endanger the lives of construction workers, such as being to close to dangerous equipment and machinery, and workers not wearing hardhats and gloves (Knight, 2019). Altogether, this technology is intended to track and predict dangerous construction site situations and activities, and providing a prior warning to forestall injuries and fatalities.

There was potential for the construction industry to adopt several artificial intelligence technologies to promote construction safety. These technologies were already in use in other industries but were yet to be adopted by the construction industry. 

5.2 Barriers hindering deployment of AI in construction safety

The construction industry presented several barriers that hindered the deployment of artificial intelligence technologies to enhance construction safety. These barriers included mistrust of artificial intelligence technologies among construction project managers and workers, lack of sufficient resources to develop artificial intelligence infrastructure, traditional organisational cultures that resisted change by avoiding investing in new technologies, the high cost of artificial technology equipment, the lack of standardized tools and protocols, the lack of government policy and regulations, lack of training in the use of artificial intelligence technologies, and the anxiety about the possible replacement of construction workers by machines. These barriers can be categorised into three thematic areas; a) resources and investment in artificial intelligence, capacity development, and regulatory environment.

Resources and Investment in Artificial Intelligence

Underinvestment in developing artificial intelligence technologies that are usable in the construction industry to promote construction safety is prevalent and similar to that in other industries with longstanding histories. Many studies have decried the slowness with which the construction industries adopted new and emerging technologies because of organisational culture hindrances and barriers. For instance, Akinosho et al. (2020) noted that the construction industry lagged behind education, finance, healthcare, and entertainment industries in the pace of adopting artificial intelligence technologies because of the steep learning curve involved because of the preference for practical and hands-on approaches for working at construction sites. In addition, the World Economic Forum (2016) noted that construction companies persisted with conservative corporate cultures because they operated in traditional settings that did not justify radical changes to existing workplace practices. In the same vein, Delgado et al. (2019) explained that cost-benefit studies on the applications of robotics and other artificial intelligence-enabled technologies in the construction industry were scarce and provided limited evidence of the safety enhancing capabilities in a real construction site environment.

Capacity Development

The construction industry has not developed sufficient capacity among project managers and construction workers to enable them embrace artificial intelligence technologies to enhance construction safety. The lack of capacity development leads to the mistrust of new technologies in industries that have relied for a long time in traditional methods. In this respect, Akinosho et al. (2020) noted that that construction companies lacked in-house expertise in artificial technology making it difficult for construction engineers to adopt these technologies because they did not have the requisite technological background and expertise. The same challenge was extended to third party artificial intelligence solution developers who had no construction background and therefore did not understand the unique problems that artificial technology would solve in construction sites (Akinosho et al., 2020). Similarly, a report by the World Economic Forum (2016) indicated that the construction industry did not collaborate sufficiently with suppliers and that decisions were made in a project-to-project basis. The lack of longstanding supplier relationships stifled the collaborative development of artificial technology solutions for safety enhancement in construction sites.

Regulatory Environment

The regulatory environment is not conducive for the development, adoption, and deployment of artificial intelligence technologies to enhance construction safety. Governments always lag behind technologies because they create policies long after new technologies have been developed. Delgado et al. (2019) faulted government for not facilitating the adoption of new technologies in the construction industry, considering they were the largest construction clients. They noted that governments were not employing the numerous tools in their disposal to promote technology-enabled construction safety, including mandates, regulatory frameworks, and policies (Delgado et al., 2019). Delgado et al. (2019) give the example of the United Kingdom government mandating public construction projects to employ building information modelling (BIM) level 2, which went a long way in fostering the adoption of the technology. In the same vein, some policies related to technology use and data placed challenges on construction firms intending to use artificial intelligence to analyse data collected from construction workers. In this regard, Akinosho et al. (2020) observed that the General Data Protection Regulation (GDPR) in force in the European Union challenged construction companies that wished to apply deep learning solutions because under this regulation, collecting and dealing with personal data without the owner’s consent is a privacy violation. Consequently, companies that had collected huge amounts of data before the enactment of the GDPR, cannot use it to develop artificial intelligence solutions before seeking consent from the individual owners. Moreover, Akinosho et al. (2020) noted that analysing such personal data could introduce unintended biases based on race and gender, which would be discriminatory. In other words, when construction companies analyse the data related to construction safety, the findings are likely to reveal safety trends that are prevalent among construction workers of a certain race or gender. In turn, it would be discriminatory to assign and generalise safety levels and adherence to specific races or gender of construction workers.

5.3 Chapter Summary

Although the construction industry is ripe for adopting artificial intelligence technologies to promote safety in construction sites, it encountered several barriers related to underinvestment, capacity, and regulatory environment. These barriers had persisted due to the conservatory culture employed by the industries and perpetuated by large construction firms. Similarly, governments contributed significantly to these barriers because they did not inspire innovation and adoption of new technologies as the largest customers of the construction industry, considering that they often undertook many of the largest infrastructure projects in their countries. Overcoming these barriers can promote the adoption of artificial intelligence technologies in construction sites, which have been demonstrated to have the ability to promote safety and secure workers in large and complex construction sites.

References

Akinosho, T. D., Oyedele, L. O., Bilal, M., Ajayi, A. O., Delgado, M. D., Akinade, O. O., & Ahmed, A. A. (2020). Deep learning in the construction industry: A review of present status and future innovations. Journal of Building Engineering, 101827.

Alruqi, W.M. and Hallowell, M.R., 2019. Critical success factors for construction safety: review and meta-analysis of safety leading indicators. Journal of Construction Engineering and Management, 145(3), p.04019005.

Awolusi, I., Marks, E. and Hallowell, M., 2018. Wearable technology for personalized construction safety monitoring and trending: Review of applicable devices. Automation in Construction, 85, pp.96-106.

Ayhan, B.U. and Tokdemir, O.B., 2020. Accident analysis for construction safety using latent class clustering and artificial neural networks. Journal of Construction Engineering and Management, 146(3), p.04019114.

Benny, D. and Jaishree, D., 2017. Construction safety management and accident control measures. International Journal of Civil Engineering and Technology, 8(4), pp.611-617.

Chen, Y., McCabe, B. and Hyatt, D., 2018. A resilience safety climate model predicting construction safety performance. Safety Science, 109, pp.434-445.

Delgado, J. M. D., Oyedele, L., Ajayi, A., Akanbi, L., Akinade, O., Bilal, M., & Owolabi, H. (2019). Robotics and automated systems in construction: Understanding industry-specific challenges for adoption. Journal of Building Engineering, 26, 100868.

Gheisari, M. and Esmaeili, B., 2016. Unmanned aerial systems (UAS) for construction safety applications. In Construction Research Congress 2016 (pp. 2642-2650).

Knight, W. (2019). Artificial intelligence sees construction site accidents before they happen. MIT Technology Review. Retrieved from https://www.technologyreview.com/2019/06/14/134944/artificial-intelligence-sees-construction-site-accidents-before-they-happen/

Li, R.Y.M., 2018. An economic analysis on automated construction safety: Internet of things, artificial intelligence and 3D printing. Springer.

Li, X., Yi, W., Chi, H.L., Wang, X. and Chan, A.P., 2018. A critical review of virtual and augmented reality (VR/AR) applications in construction safety. Automation in Construction, 86, pp.150-162.

Marefat, A., Toosi, H. and Hasankhanlo, R.M., 2019. A BIM approach for construction safety: Applications, barriers and solutions. Engineering, Construction and Architectural Management, 26(9), pp. 1855-1877.

Mneymneh, B.E., Abbas, M. and Khoury, H., 2017. Automated hardhat detection for construction safety applications. Procedia Engineering, 196, pp.895-902.

Niu, Y., Lu, W., Xue, F., Liu, D., Chen, K., Fang, D. and Anumba, C., 2019. Towards the “third wave”: An SCO-enabled occupational health and safety management system for construction. Safety Science, 111, pp.213-223.

Oswald, D., Sherratt, F., Smith, S. and Dainty, A., 2018. An exploration into the implications of the ‘compensation culture’ on construction safety. Safety Science, 109, pp.294-302.

Pan, Y. and Zhang, L., 2021. Roles of artificial intelligence in construction engineering and management: A critical review and future trends. Automation in Construction, 122, p.103517.

Park, J., Kim, K. and Cho, Y.K., 2017. Framework of automated construction-safety monitoring using cloud-enabled BIM and BLE mobile tracking sensors. Journal of Construction Engineering and Management, 143(2), p.05016019.

Pemberton, L. (2019). Creating the ultimate onsite safety assistant using machine learning on construction projects. Retrieved from https://www.prescouter.com/2019/07/using-machine-learning-construction-projects-safety/

Pro Crew Schedule (2021). 12 tech startups powering construction site with artificial intelligence (AI). Retrieved from https://www.procrewschedule.com/12-tech-startups-powering-construction-site-with-artificial-intelligence-ai/

Rao, S. (2021). The benefits of AI in construction. Retrieved from https://constructible.trimble.com/construction-industry/the-benefits-of-ai-in-construction

Riaz, Z., Parn, E.A., Edwards, D.J., Arslan, M., Shen, C. and Pena-Mora, F., 2017. BIM and sensor-based data management system for construction safety monitoring. Journal of Engineering, Design and Technology, 15(6), pp. 738-753.

Rubaiyat, A.H., Toma, T.T., Kalantari-Khandani, M., Rahman, S.A., Chen, L., Ye, Y. and Pan, C.S., 2016, October. Automatic detection of helmet uses for construction safety. In 2016 IEEE/WIC/ACM International Conference on Web Intelligence Workshops (WIW) (pp. 135-142). IEEE.

Sanni-Anibire, M.O., Mahmoud, A.S., Hassanain, M.A. and Salami, B.A., 2020. A risk assessment approach for enhancing construction safety performance. Safety Science, 121, pp.15-29.

Sherratt, F. and Dainty, A.R., 2017. UK construction safety: a zero paradox? Policy and Practice in Health and Safety, 15(2), pp.108-116.

Shohet, I.M., Wei, H.H., Skibniewski, M.J., Tak, B. and Revivi, M., 2019. Integrated communication, control, and command of construction safety and quality. Journal of Construction Engineering and Management, 145(9), p.04019051.

Tixier, A.J.P., Hallowell, M.R. and Rajagopalan, B., 2017. Construction safety risk modeling and simulation. Risk Analysis, 37(10), pp.1917-1935.

Tixier, A.J.P., Hallowell, M.R., Rajagopalan, B. and Bowman, D., 2017. Construction safety clash detection: identifying safety incompatibilities among fundamental attributes using data mining. Automation in Construction, 74, pp.39-54.

Walch, K. (2020). AI transforming the construction industry. Forbes. Retrieved from https://www.forbes.com/sites/cognitiveworld/2020/06/06/ai-transforming-the-construction-industry/#:~:text=AI%20is%20also%20helping%20to,on%2024%2F7%20without%20distraction.

World Economic Forum (2016). Shaping the future of construction: A breakthrough in mindset and technology. Retrieved from http://www3.weforum.org/docs/WEF_Shaping_the_Future_of_Construction_full_report__.pdf