Overview of Emerging Technologies for Improving the Performance of Heavy-Duty Construction Machines

Construction equipment is one of the most significant resources in large construction projects, accounting for a considerable portion of the project budget. Improving the performance of heavy machinery can increase efficiency and reduce costs. However, research on boosting the machine efficiency is limited. This study adopts a mixed review methodology (systematic review and bibliometric analysis) and evaluates emerging technologies such as digital twin, cyber physical systems, geographic information systems, global navigation satellite systems, onboard instrumentation systems, radio frequency identification, internet of things, telematics, machine learning, deep learning, and computer vision for machine productivity, and provides insights into how they can be used to improve the performance of construction equipment. This study defined three major equipment operating areas—monitoring and control, tracking and navigation, and data-driven performance optimization—classified technologies and explored how they can increase machine productivity. Other circumstantial issues affecting machine operation and loopholes in the existing innovative tools used in machine processes have also been highlighted. This study contributes to the goal of deploying digital tools and outlines future directions for the development of automated machines to optimize project efficiency.

INDEX TERMS Digital models, earthmoving, equipment productivity, mobile equipment, emerging, technologies, tracking, sensing.

17
In recent years, practitioners and researchers have focused 18 on improving the performance of the construction sector, 19 which has a reputation for poor productivity, with only a 1% 20 increase over the previous two decades [1]. Even though con-21 struction is one of the largest sectors in the world, account- 22 ing for 13% of global GDP, it continues to underperform, 23 although the industry has not been in recession [2]. Construc-24 tion efficiency gains of 50%-60% or more are expected to 25 add $1.6 trillion to the industry's value and increase global 26 GDP [3]. Construction equipment is a company's greatest 27 asset during a time crunch because it streamlines and provides 28 the most assistance. As the global construction equipment 29 industry is expected to develop at a compound annual growth 30 The associate editor coordinating the review of this manuscript and approving it for publication was Gongbo Zhou. rate (CAGR) of 3.9 percent from 2022 to 2030, the sector 31 has enormous potential to add value to construction [4]. This 32 is because the global increase in construction activity is pro-33 jected to stimulate the demand for machines. Mega projects, 34 such as roads, mines, dams, and open-pit mining, mostly rely 35 on earth moving operations [2]. The performance of heavy 36 machinery, such as excavators, loaders, and dump trucks, 37 has a significant impact on the overall project efficiency. 38 The growing popularity of electric construction equipment 39 is projected to provide new income streams for original 40 equipment manufacturers (OEMs) in the coming years [4]. 41 Increasing heavy machinery utilization is crucial not only 42 for productivity but also for cost management. Therefore, 43 it is important to evaluate and improve the performance of 44 construction machines to increase the productivity of the 45 construction operations.
• RT1: To identify the areas of machine operation in 114 which emerging technologies can potentially contribute 115 to adding value? 116 • RT2: Which technology can be used for which specific 117 area of machine operation? 118 • RT3: What are the existing challenges, and how can they 119 be addressed using new technologies? 120 The remainder of this paper is structured as follows: The 121 review technique and process are presented in Section 2 122 ''Review Method.'' In Section 3, ''Emerging technologies'' 123 for machine operation optimization (RT 1 , RT 2 ) are discussed.

130
Firstly, a systematic search of the literature was conducted 131 using PRISMA protocol guidelines and bibliometric analy-132 sis. This was done by arranging the collected papers based 133 on their keywords, document types, language, and so on. 134 Second, bibliometric analysis was performed to improve the 135 quality of the research and find research trends via quantita-136 tive analysis. Fig 2 illustrates the major phases of the review 137 process. to measure the productivity of the construction equipment. 143 The scope was set to articles in English published between 144 2002 and 2020 that dealt with heavy machinery productivity 145 and technology in various construction projects. 146 2) DATABASE AND KEYWORDS 147 Five databases, Web of Science (WoS), Scopus, Science 148 Direct, ACM, and IEEE Explorer, were chosen to compile the 149 literature on innovative solutions for construction machinery. 150 Google Scholar was used to obtain the citation counts of the 151 collected articles. 152 The keywords were split into two categories: those con-153 nected to technology and those related to construction. 154 Construction-related words are autonomous, construction, 155 equipment, operation, excavator, monitoring, management, 156 building, project, efficiency, productivity, earthmoving, track-157 ing, and optimization. Data visualization, information, 158 modeling, mobile, machinery, digital, models, sensing, and 159 analytics are words used for technologies.  Table 2 lists the keywords used to perform these queries. Each 162 keyword within a group was paired using the OR operator, 163 whereas the groups were paired using the AND operator 164 ( Table 2). The last row of Table 2  Using the established technique, a total of 1,980 papers 171 published between 2002 and 2022 were collected. The max-172 imum number of articles was published in the last five years, 173 indicating that the subject is active in academia, and the 174 publishing trend is growing (Fig 3). Maximum data collection 175 was obtained from the Web of Science (WoS) and Scopus 176 databases. This stage was divided into four steps, as shown in Fig 4: 179 identification, screening, article eligibility, and inclusion. The 180 first step was to determine the number of publications in 181 each database. The second step was screening, where the 182 publications were screened for duplication, title, and abstract. 183 Duplication was based on articles with the same titles that 184 were re-selected by alternative keyword combinations from 185 were eliminated. Subsequently, 440 irrelevant articles were 191 eliminated by reviewing the abstract of each paper.

192
In the third phase, the full text eligibility for each article 193 was carefully examined considering the tools and methodol-194 ogy used, resulting in a total of 88 articles being included and 195 87 publications being removed. Following the filtering pro-196 cedures stated above, the number of publications included in 197 each database yielded the following synopsis: Fig 4 displays  The articles from the selected publications were analyzed 202 based on the keywords, sources of articles, and technological 203 tools used for experimentation. The map of the selected pub-204 lications was then evaluated based on the title and abstract to 205 establish the correlation between words and the most frequent 206 terms.  Collected publications were also analyzed for their sources to 218 determine the distribution by journal/conference, as shown 219 in Fig 6 and 7, respectively. Only 15 of the 88 papers 220 were chosen from conferences, and the remaining 73 pub-221 lications were found in journals. In addition, it was found 222 that ''Automation in Construction'' had the highest num-223 ber of articles among the selected journals and conferences 224 with 18 records, followed by ''Information Technology in 225 Construction'', ''Journal of Computing in Civil Engineer-226 ing'', and ''Advanced Engineering Informatics'', with seven, 227 five, and five records, respectively. In the conference cat-228 egory Construction Research Congress, IEEE, and ASCE 229 execution of operation, and 37 papers related to technologies 239 and innovation. Fig 8 shows

245
The use of information and communication technology (ICT) 246 in construction projects has the potential to improve project 247 efficiency [3]. Automation and robotics have recently been 248 introduced in the construction industry. These approaches 249 use a combination of computers, machine components, and 250 software to systematically run the equipment [7]. These tech-251 nologies are used to improve the work environment, health 252 and safety, scheduling, product quality, and reduce on-site 253 However, it is important to determine which technologies 258 are used, how often they are used, and the environment in 259 which they are used because on-site job circumstances vary 260 in construction projects (RT 2 ). We have studied several tech-261 nologies used in construction machinery to improve produc-262 tivity. Fig 9 depicts the classification of technologies based on 263 their potential for improving machine performance, which is 264 covered in detail in the following subsections (RT 1 and RT 2 ). 265

266
In comparison to construction, the manufacturing sec-267 tor is more advanced in utilizing digital solutions, and 268 research shows that most digitization processes in con-269 struction share challenges like those of manufacturing. 270 The construction equipment industry has a well-established 271 distribution network that employs comparable production 272 methods [7]. Although researchers and practitioners are 273 working on launching digitally equipped machinery, they 274 have not succeeded in developing fully automated machines. 275 attractive owing to its distinct features, such as bidirectional   dimensions, shapes, and other features that correspond to the 331 real objects. DT has the potential to map the logic and rules 332 used by physical entities and determine the past, present, 333 and future of a physical entity (assets or processes) [27]. 334 Moreover, in extreme circumstances during project execu-335 tion or remote operations, where manual monitoring is chal-336 lenging, DT can be anticipated to gain self-awareness and 337 self-optimization as it offers two-way data exchange and 338 control [18]. Thus, contractors can meet their daily objectives 339 on dangerous or crowded building sites with the support of 340 digital versions of machinery [27]. In addition to improving 341 robustness, DT can assist practitioners and managers to save 342 energy and other resources.

343
Although DT is a new technology, it is continually evolving 344 and has not yet been developed in the construction indus-345 try. However, the literature reveals that researchers have 346 suggested several DT designs to utilize this technology in 347 automating machine operations in the manufacturing indus-348 try. Table 4 provides a summary of the scholarly research 349 on DT in manufacturing management systems. Research on 350 DT is still in its early stages, and numerous challenges and 351 shortcomings must be resolved before DT can be broadly 352 used in the construction sector. Modeling every unit of 353 a real-world system requires significant computing power, 354 data storage, and continuous data transmission and process-355 ing [28]. The infrastructure required to achieve a high degree 356 of performance has not yet been achieved. Precision, accu-357 racy, data collection, and synchronization are important for 358 reliable simulations [29]. Therefore, real-time communica-359 tion, sophistication, precision, connectivity, and architectural 360 foundations are challenges that must be overcome before DTs 361 can be used in practice [28]. The use of a secure, reliable, 362 and rapid connection to transmit data in real-time is a well-363 known requirement. In addition, the technical components, 364 protocols, and tools needed to build a DT or describe it as an 365 all-encompassing technology are yet to be agreed upon [30]. 366 Sophisticated technologies, such as wireless sensor networks, 367 industrial artificial intelligence, blockchains, and transfer 368 learning algorithms, can also be integrated to increase the 369 functioning and capabilities of DT in different areas. 370 2) CYBER PHYSICAL SYSTEMs (CPS) 371 CPS combines informatics, real-time control subsystems, 372 components, and human operators to affect physical pro-373 cesses through collaboration and a partially automated con-374 trol mechanism [5]. One of the primary distinctions between 375 DT and CPS is that CPS embeds a system-level thinking 376 approach based on networked products and functions (inte-377 grating the connectivity principle), whereas DT is an engi-378 neering system that drives new abilities to design, execute, 379 maintain, and develop new services to optimize its worth [30]. 380 CPS includes the coupling of physical systems with their dig-381 ital replicas via sensors and actuators, whereas DT provides a 382 digital copy of the facility as designed and implemented [34]. 383 Table 5 presents the typical CPS architecture. As auto-384 mated mobile machinery is more appealing for operation in 385 VOLUME 10, 2022 upgradation in construction machinery can potentially lower 420 overall construction expenses by 20-30% [34], [35]. The 421 machine moving times by 40% and 50%, respectively [35]. 423 Compared to other sectors that have already embraced Indus-424 try 4.0, CPS can increase construction productivity by a factor 425 of three [35]. This technology can also promote small farms      The construction industry is one of the wealthiest industries 529 that produces massive amounts of data [3]. Therefore, there 530 is great potential for using such data to enhance production.  . These sensors are used to identify mechanical issues 561 with tracked equipment and measure specific features, such 562 as temperature, pressure, and control lever position, which 563 affect the equipment cycle time and productivity [49]. For 564 example, the OBI of a crane has an encoder sensor and load 565 moment indicator. Encoder sensors, which are used to mea-566 sure the rotation angle, are either installed between the body 567 of the crane and truck base or between the boom and crane 568 boom. To collect obstacle data, [73] used a mobile crane's 569 boom head and accessible sensors (boom length, rotation 570 angle, and elevation angle). In addition, the crane overturning 571 motion (load multiplied by radius) can be calculated and dis-572 played using the load moment indicator (LMI) [21]. OBI and 573 LMI are well-established technologies. It is currently used 574 as an overload mechanism for large cranes. This technology 575 displays the crane's rated capacity and the percentage of the 576 lifted object's moment to the operator, as well as sounding an 577 alert if the moment exceeds the limit [15].

578
The OBI offers efficient cost savings and is used for various 579 tasks in a number of earthmoving processes, such as avoiding 580 loader rollovers, regulating scraper wheel slippage and gear-581 box shifting, and increasing dozer productivity [49]. Caterpil-582 lar, a pioneer in the field of operational business intelligence, 583 developed a vital information management system (VIMS) 584 to monitor real productivity and machine conditions [49]. 585 Because OBI is already used as a standard tool for most heavy 586 equipment, no additional installation costs are required. How-587 ever, systems based on OBI are expensive and cannot antici-588 pate fleet performance, project costs, or completion times in 589 a deterministic or stochastic manner [49]. The use of OBI 590 to track progress has several drawbacks. The information 591 quality is proven to be poor in encoder-based cable length 592 measurements, which are approximated by cable stretch and 593 winch spooling [74]. In addition, data collected through OBI 594 is not useful for tracking productivity [1]. Therefore, to make 595 it more effective, it must be combined with information 596 regarding material quality and site environment, as presented 597 in [48]. It is difficult to avoid accidents involving other cranes 598 and building structures, as well as swaying loads, in terms 599 of workplace safety [21]. In addition to OBI, no additional 600 sensors are available, which is a limitation [73]. RFID is a wireless-based technology that uses radio fre-603 quency (RF) to interact with distinctively traceable tags. 604 RFID systems comprise tags and readers. Tags connected 605 to the equipment were used to collect and transmit digital 606 data to readers via radio waves. These tags are used to track 607 and identify objects [75]. RFID can be used to measure dis-608 tance, identify targets, and determine proximity [14]. Using 609 triangulation and signal propagation time, the location of tag 610 can be calculated. Event detection begins with a fingerprint 611 pattern [14]. To estimate the tag placement, a biometric is 612 matched to a digital scan. The transmitter density is used to 613 VOLUME 10, 2022 determine the distance. Readers can set a specific voltage technology that offers high-speed communication over small 669 distances. It can be used to track the locations of multiple 670 pieces of equipment and to obtain information about them. 671 It can also locate and identify multiple dynamic pieces of 672 equipment on a job site. Like, [62] used a UWB positioning 673 system to collect job data from a worksite. By periodically 674 attaching a tag to the deployed component, workers and 675 supervisors can trace operations. Similarly, [78] developed 676 a system that uses UWB to help operators find equipment 677 and assess hazards. Wearable gadgets and connected IoT 678 devices were developed in [79], which can alert a worker 679 to potentially harmful areas.
[80] developed a method for 680 compiling site equipment and creating management analyt-681 ics. To do this, they proposed ''Smart Connected Objects,'' 682 which include machinery, tools, materials, and even buildings 683 with sensing, processing, and communication capabilities. 684 This edge computing method provides machines with auton-685 omy and awareness to make better decisions. A closed-loop 686 lifecycle management system framework based on the IoT 687 was developed by [81]. To improve the efficiency and safety 688 of mining equipment, [82] used coal mine safety monitor-689 ing and maintenance to construct an IoT-based predictive 690 maintenance system. Using IoT and RFID, [83] developed a 691 warning system to inform employees of the possible dangers. 692 An IoT-based system was also employed on-site in [84] to 693 enable data collection, supervision, and analytics. To ensure 694 the safety of each tower crane during operation, [85] uti-695 lized the IoT to capture the status information on the crane 696 arms and designed an anti-collision algorithm. The issue and 697 intricacy of tracking job status and productivity assessment 698 can also be resolved by integrating multiple wireless tech-699 nologies [42].  Telematics refers to the use of wireless technology to connect 708 equipment-monitoring systems. Telematics includes wireless 709 communications, vehicle monitoring systems, and position-710 ing sensors that provide real-time location and operation 711 data [79]. The sensors used in these devices capture and 712 transmit data through cellular and GPS networks [42]. This 713 specific data depends on the machine type and telematics unit. 714 Today's equipment rental companies depend on telematics 715 to collect real-time machine data [1]. Telematics can be used 716 to improve the efficiency of the work locations. Implementing 717 telematics in an equipment footprint can enhance productiv-718 ity, reduce expenses, and provide data-driven patterns [86]. 719 Working hours, location, fuel consumption, and productivity 720 are significant features of heavy-equipment efficiency [2]. 721 Compared to RFID, telematics has a reduced transmission 722 frequency by default, which restricts its use for precise 723 operation tracking. However, it is possible to increase this 724 frequency, but doing so would require a robust data storage 725 system to support a high volume of information [1], [79]. 726 A detailed description of the RFID, IoT, and telematics is 727 provided in Table 8.  . However, each method has its 756 own set of advantages and disadvantages, making it diffi-757 cult to determine which one is the most appropriate and 758 effective, but it mostly depends on the quality and integrity 759 of the data collected on the field side [3]. Fig 11 shows 760 how machine learning algorithms have been used recently to 761 address issues regarding equipment management and earth-762 moving operations. CV-based methods have the potential to become new data col-785 lection and analytics technologies for earthmoving machine 786 optimization; however, there is still a long way to go before 787 a reliable and automated system can be employed in con-788 struction projects. Although some researchers have suggested 789 methods for extracting cycle time from video, which is used 790 as an input for process simulation tools, the number of articles 791 that combined construction-process simulation and vision-792 based monitoring to analyze the existing earthwork produc-793 tivity and suggest an optimal resource allocation plan were 794 found minimal [95]. Table 9 summarizes the deep learning 795 and computer vision algorithms proposed by various authors 796 for construction machines.

798
The integration of information and communication tech-799 nology (ICT) has a positive influence on the progression 800 VOLUME 10, 2022 acceleration data collected from excavators. [112] created a 827 system that uses accelerometers to monitor equipment effi-828 ciency and environmental performance. They used vibration 829 signal analysis as their primary technique to identify and 830 track equipment operations. Supervised learning algorithms 831 were used to classify the accelerometer data into equipment 832 activities (working, idling, and engine-off). Table 10 provides 833 a detailed breakdown of these studies. Although researchers 834 have integrated multiple machine learning techniques, such 835 as ANN and support vector machine (SVM), with BIM to 836 automate construction productivity tracking, these existing 837 techniques are applied on a very limited scale to different con-838 struction areas, such as workers, processes, and machinery, 839 and do not cover large-scale job sites [103]. Furthermore, the 840 number of surveillance tools (such as tags and cameras) are 841 limited [48]. Although many AI-related construction activi-842 ties are still in the development stage, they are anticipated to 843 allow the seamless integration of automated equipment in a 844 5D BIM planning environment [113].

895
As construction operations grow on a large scale, the com-896 plexity also increases. To achieve the accuracy requirements, 897 the techniques and technologies must be employed to exe-898 cute them also be upgraded. Because too many machines 899 are simultaneously used on the worksite to complete the 900 job, the connectivity of single value-adding processes is 901 important [10]. In terms of architectural needs, timing, and 902 commitment to project completion, each stage must be care-903 fully planned and scheduled in the initial phases to avoid 904 divergence from the expected outcomes that can affect all 905 subsequent processes. Ideally, the goal is to optimize the 906 workflow rather than the individual phases. Because work 907 is being done simultaneously, better synchronizations are 908 required to avoid serious deadlocks if the activity outputs 909 diverge. There must be a constant flow of information to the 910 equipment that is performing other jobs and to the supervi-911 sory system, preferably in parallel. Parameters such as time, 912 fuel usage, and the quantity of equipment and personnel are 913 also involved in determining the optimum performance [53]. 914 Hence, a multicriteria performance assessment tailored to 915 the specific conditions of the entire construction process 916 is required to achieve the desired results. In other words, 917 smarter machines require better methods to measure their 918 performance. Digital transformation is the first step toward 919 implementing a production model, and digitalization of the 920 operation is the most important step [64]. Once the process 921 is digitized, all essential data can be accessed and collected 922 for the application of sophisticated technologies. One of the 923 initial phases of digitization in certain industrial upgrades is 924 data harvesting and transfer [74]. Subsequently, the acquired 925 data can be used to optimize machine performance, stream-926 line operations, control defects, and make smart decisions to 927 build a digital infrastructure. Hence, a solid platform for data 928 collection and exchange is required [3]. 929 Power, transmission, and control systems, as well as pri-930 mary support technologies for construction machines, have 931 all seen significant technical advancements [17]. However, 932 complex structures, difficult construction site environments, 933 and substantial load fluctuations for construction machines 934 cause technical problems. Therefore, detailed construction 935 site data collection, more automated solutions, and effective 936 condition-monitoring systems are essential. One of the most 937 desirable pieces of information from heavy equipment is 938 the downtime. Telematics can be used to record the engine 939 status during idle time, depending on the configuration set-940 ting of the machine [86]. Minimizing the idle time of a 941 machine can reduce the emissions and fuel usage. Two major 942 challenges with automated productivity data collection are 943 the potential overflow of low-level data or information that 944 might burden project managers [79], and the lack of direc-945 tion in evaluating workflow in construction operations [48]. 946 Because increased data access and communication between 947 sensors, devices, machinery, monitoring, and control sys-948 tems are required to optimize the entire operation, a digital 949 twin and a cyber-physical system can be useful solutions, 950 as physical objects (equipment, actuators, and tools) can link 951 with cyber entities that provide data storage, processing, and 952 analysis. Moreover, the availability of wireless technologies 953 that allow the Internet of Things (IoT) has increased dramatically, and this is expected to be a major advancement in 955 the future integration of online services [72]. Images, videos, 956 and audio books were among the collected data, leading to 957 the creation of large metropolitan datasets. Documentation 958 from machines, maintenance, and subscriptions has previ-959 ously been overlooked, but it has now been brought to light, 960 like a gold mine, providing for the extraction of previously 961 unknown data, invisibility, and other important details to 962 improve prediction and productivity [84]. Rather than spend-963 ing time on data to be produced, continuous data techniques 964 must be devised to collect, retain, filter, and analyze large 965 amounts of data to maximize their utility [42]. This is espe- as the cost of time and potential losses [29]. Real-time sensor 995 data is therefore highly beneficial for virtual solutions [118]. 996 With the use of a digital twin, enhanced simulation results are 997 required for better decision-making and monitoring of data 998 coming from sensors [28]. It is also feasible to support the 999 lifecycle of a machine process such as material recovery by 1000 establishing an IBMS [43]. Likewise, advances in construc- where the machine is placed next to its working area and 1007 digging can be performed via sensors and control, have been 1008 growing [113]. In 2015, Komatsu, a well-known producer 1009 of equipment that specializes in data-driven and machine-1010 learning-enhanced analysis, introduced its smart construction 1011 system. Automated systems enhance the analytical efficiency, 1012 accuracy, and quality [13]. INSITE is currently working on 1013 a system that combines computer vision, deep learning, and 1014 aerospace algorithms to make a machine smarter by predict-1015 ing its location and visual perceptions [113]. Such solutions 1016 will make machine operation more reliable, productive, and 1017 efficient. Further digitization will improve the equipment 1018 performance and provide more opportunities to achieve the 1019 overall aim of process optimization.

1021
In this section, we discuss major gaps in the applications 1022 of new technologies employed in construction projects to 1023 evaluate the performance of the equipment. [85], [42]. Adopting these techniques with machine 1038 learning algorithms has improved their ability to mon-1039 itor construction productivity and performance; how-1040 ever, anomalies, monitoring range, and tagged device 1041 privacy problems still need to be addressed.

1042
3) Recently, sensor technology has been used in con-1043 struction machines. However, owing to the complex 1044 mechanical structure, power and transmission systems, 1045 and operational environment of construction machines, 1046 sensor-based applications experience tougher require-1047 ments [17]. Therefore, sensor technologies cannot be 1048 instantly applied to construction machines, and exten-1049 sive testing and optimization are required to ensure 1050 their suitability for construction machinery.