Accuracy Improvement of Real-Time Location Tracking for Construction Workers

sustainability

Article Accuracy Improvement of Real-Time Location Tracking for Construction Workers

Hyunsoo Kim 1 and Sangwon Han 2,*

1 Department of Architectural Engineering, Gyeongnam National University of Science and Technology, 33, Dongjin-ro, Jinju-si 52725, Korea; [email protected] 2 Department of Architectural Engineering, University of Seoul, Seoul 130-743, Korea * Correspondence: [email protected]; Tel.: +82-751-3677

Received: 4 April 2018; Accepted: 7 May 2018; Published: 9 May 2018

Abstract: Extensive research has been conducted on the real-time locating system (RTLS) for tracking construction components, including workers, equipment, and materials, in order to improve construction performance (e.g., productivity improvement or accident prevention). In order to prevent safety accidents and make more sustainable construction job sites, the higher accuracy of RTLS is required. To improve the accuracy of RTLS in construction projects, this paper presents a RTLS using radio frequency identiﬁcation (RFID). For this goal, this paper develops a location tracking error mitigation algorithm and presents the concept of using assistant tags. The applicability and effectiveness of the developed RTLS are tested under eight different construction environments and the test results conﬁrm the system’s strong potential for improving the accuracy of real-time location tracking in construction projects, thus enhancing construction performance.

Keywords: real time locating system; radio frequency identiﬁcation; safety management; assistant tag; tracking error mitigation

1. Introduction Accident prevention is directly linked not only to safety, but also sustainability. From the perspective of sustainability, occupational health and safety (OHS) is critical for sustainable development [1]. Based on the three-pronged model of sustainability suggested by McKeown [2], the role of OHS in promoting sustainability can be discussed in three dimensions: people, planet, and proﬁt. Among these dimensions, people should be directed towards promoting respect for diversity and human rights; health and safety protection; and empowerment and caring [3]. With regard to safety improvement and sustainability, diverse technologies have been applied in the construction industry for reducing accident occurrence. Among various technologies, real-time locating systems (RTLSs) have been widely used in the construction industry. A RTLS for construction components such as labor, equipment, and materials offers numerous opportunities to improve construction performance. These include accident prevention [4–8], productivity measurement and improvement [9,10], progress monitoring [11,12], and material management [13–15]. Among these approaches, accident prevention approaches using RTLS have tried to reduce human error (e.g., near-miss accident, lack of hazard recognition, inexperienced labor) [16,17]. One representative approach is that a safety management system tracks laborers’ location and alerts them when they move near to hazardous areas. Therefore, accident prevention approaches based on RTLS require high accuracy of location tracking, because low accuracy of location tracking can result in system error and fail to prevent an accident. When a RTLS is used in order to prevent an accident, an accurate RTLS for laborers’ locations is one of the most important components. However, several factors (e.g., multi-path, shadow area, signal

Sustainability 2018, 10, 1488; doi:10.3390/su10051488 www.mdpi.com/journal/sustainability Sustainability 2018, 10, 1488 2 of 16 loss) can cause location estimation errors and decrease the accuracy of location tracking. Apart from the general problems of location tracking, the characteristics of construction sites make location tracking difficult. As mentioned above, a construction site usually consists of indoor (building construction work) and outdoor (earthwork phase and outdoor civil construction) activity areas. Therefore, the location tracking technology selected should be able to be used in both areas. Although the accuracy of localization is considerably influenced by signal availability, maintaining enough signal availability is very difficult at construction sites [8]. The accuracy of a location-sensing system depends highly on the environment, which can be classified as a line of sight (LOS) or non-line of sight (NLOS) environment. LOS environments have few obstacles; thus, communications between tags and readers are usually undisturbed. However, LOS environments in a construction site can cause localization errors, because there are diverse materials. Moreover, the dynamic changes in a construction site can cause a shift from LOS environments to non-line of sight NLOS environments. NLOS environments can decrease the accuracy of a location-sensing system. This change, which is one of the most remarkable characteristics of location-sensing on construction sites, should be considered in order to improve the accuracy of a location-sensing system. Therefore, localization errors should be solved under both LOS and NLOS conditions. To address these difficulties in location tracking, this paper develops an accurate and effective real-time locating system (RTLS) for construction sites. In particular, the workers’ location on construction sites is focused on, because construction activities are highly labor intensive. This paper is organized as follows. We first conduct a literature review on location-tracking technologies in construction and discuss the main reasons for the occurrence of localization error. Next, we introduce location tracking error mitigation algorithms and suggest the use of assistant tags in order to overcome NLOS conditions to improve tracking accuracy in construction projects. Based on these ideas, we develop an RFID-based RTLS and test its applicability and effectiveness. Finally, we draw some conclusions based on the test results.

2. Literature Review This chapter introduces a location-sensing selection procedure based on comparative overview. Literature and current approaches to ways in which the accuracy of a given location-sensing technology can be improved are reviewed.

2.1. Location-Sensing Technologies One of the most widely used location-sensing technologies is the global positioning system (GPS), which locates a target object by using a network of 24 satellites [18]. GPS is quite effective in locating and tracking objects in outdoor environments [19], but has an inherent weakness in locating indoor objects. This has limited the effectiveness of GPS in building construction projects, where many activities are conducted indoors. Therefore, there is a need for a method that can be applied to indoor and outdoor conditions. There are diverse locating methods that can determine the location of objects in both indoor and outdoor conditions. These methods consist of location measurement methods (e.g., triangulation, scene analysis, and proximity), distance measurement methods (e.g., angle of arrival (AOA), received signal strength indication (RSSI), time of arrival (TOA), and time difference of arrival (TDOA)), and technologies (e.g., Infrared, IEEE 802.11, Ultrasonic, ultra wide band (UWB), and RFID). These methods have problems, such as multi-path, obstacle (shadow area), and signal loss [20,21]. IEEE 802.11b products are easy to set up for general indoor conditions, as well as construction sites. Nevertheless, they do not guarantee high-accuracy location sensing. Ultrasonic has a relatively higher accuracy than other technologies. Despite its higher accuracy, an ultrasonic-based system requires a great deal of infrastructure and cost in order to be highly accurate [22]. Another problem is Sustainability 2018, 10, 1488 3 of 16 that it is difficult to install this system on construction sites. It cannot penetrate obstacles such as walls or installed forms, which are usually the main components of ongoing construction projects [23]. Although UWB presents the highest accuracy, it is difficult to apply to outdoor construction activities because the signal transmission distance is approximately 20 m [24]. This short signal transmission distance means that a number of readers are required in order to cover a whole construction site, and it is not cost-effective. RFID has been widely used for checking and estimating accessibility in urban spaces [25,26]. These methods are operated by capturing the signal from a user. The captured signal is considered as a user’s access to a specific reader (which is a reference location already known). This approach is operated based on the one-on-one matching between a tag and a reader [27]. Thus, it can estimate a tag’s location by means of range from a reader. This method can be expanded from location identification to positioning by using multiple readers, like GPS. Theoretically, at least three readers are required to identify a target’s location (positioning). In this approach (using more than three readers), RFID is still relatively cost-effective and has high capability for information transfer. Despite its strengths, it is heavily influenced by environments, which cause multi-paths, obstacles (i.e., shadow areas), or signal loss [20,21]. This is because the accuracy of location sensing largely relies on the strength of the tracking signal [8], which can vary depending on the tracking environment. RFID is a small electronic device that consists of a small chip and an antenna. An RFID-based RTLS can store and retrieve location data by utilizing a number of tags (which are attached to target objects) and readers (i.e., tracking equipment). To locate a target object, a reader receives data emitted from a tag attached to the target object. After transmitting and receiving the location data, an RFID-based RTLS can calculate the coordinates of a target tag. Many researchers have proposed the use of RFID for location tracking because of its good tracking performance and cost-effectiveness [13,19,28]. For example, Ni et al. [19] developed an RFID-based indoor location-sensing system that can improve the overall accuracy of location sensing by applying reference tags. Song et al. [13] developed a mathematical model that estimates the signal strength within a construction site. Similarly, Skibniewski and Jang [28] suggested a framework for automated real-time location tracking of construction materials based on RFID signals combined with ultrasonic waves. By providing supplementary tracking tools (e.g., a signal strength fingerprint map), these approaches are effective in tracking indoor objects when site conditions are well pre-determined. When installing the location-sensing systems (i.e., determining the suitable location of the readers), a signal strength fingerprint map has strong potential to improve the overall accuracy of RTLS. However, the tracking environments in a construction site can vary significantly as the project progresses. Therefore, this paper suggests an enhanced RTLS that can improve the accuracy of location tracking in construction projects.

2.2. Movement Chacrateristics Tracked objects have their own characteristics (e.g., direction, range, and velocity). Thus, if we understand the characteristics of the target, then it can be involved in an RTLS [29]. There are a few researchers dealing with movement characteristics of a target for location-sensing [29,30]. One main example is the research of Son et al. [29]. They suggested an enhanced asset-locating system for vehicle pooling in a port terminal. This system focused on the range and speed of a transfer crane. A transfer crane has three parameters: path, range and velocity (x-, y-, and z-axes). Moreover, each axis has a range that can be movable. Unlike vehicles in a port terminal, construction workers do not have a specific direction. A worker can move in all directions. One similar characteristic between a construction worker and a vehicle in a port terminal is having a velocity range. If information about the velocity range of a worker is incorporated into the location-sensing engine, a log—which is defined as data that records a target’s location by specific time interval—located on outside of a target’s range means that a measured log can be considered an error. Based on the consideration of this characteristic of tracking a worker’s movement characteristics, this paper selects the velocity of workers as a primary factor to mitigate estimation errors. SustainabilitySustainability2018 2018, ,10 10,, 1488 x FOR PEER REVIEW 44 ofof 1616

2.3. Location-Tracking Error Caused by Multi-Path and NLOS 2.3. Location-Tracking Error Caused by Multi-Path and NLOS There are two main reasons for location-tracking error. One is multi-path of signals and the other is NLOS.There First, are two in an main RFID-based reasons for RTLS, location-tracking a tag repeatedly error. emits One radio is multi-path frequency ofsignals signals and and the a reader other isreceives NLOS. these First, signals. in an RFID-based The location RTLS, of the a tag is repeatedly estimated emits based radio on the frequency signal flight signals time and from a readerthe tag receivesto the reader. these signals.A tag transmits The location omnidirectional of the tag is signal estimateds, and based can make on the both signal direct flight and time reflected from the paths. tag toAccordingly, the reader. Athe tag flight transmits times omnidirectionalof signals emitted signals, from a and tag canat a make certain both moment direct can and vary reflected depending paths. Accordingly,on the type of the signal flight path. times For of example, signals emitted as shown from in aFigure tag at 1, a certainthe distance moment of the can direct vary path depending (D1) is onshorter the type than of that signal of the path. reflected For example, path (D2). as shown In this in case, Figure a reader1, the distancereceives two of the different direct path signals, (D1) and is shorterthe difference than that can of cause the reflected an error. path If the (D2). reader In this receives case, a a reader signal receivesconveyed two through different the signals, reflected and path, the differencethe reader can estimates cause anthe error. location If the of readerthe tag receives based on a signalthe distance conveyed of the through reflected the path reflected (D2), path, which the is readerlonger estimates than the actual the location distance of the (D1) tag. In based detail, on the the velocity distance of of a the sign reflectedal emitted path from (D2), a whichtag is constant. is longer thanThus, the if actualthere distanceare two (D1).different In detail, paths the(see velocity direct ofpa ath signal and reflected emitted from path a in tag Figure is constant. 1), the Thus, reader if thereregards are them two different as having paths been (see emitted direct pathfrom anda diffe reflectedrent location, path in Figurebecause1), the readerreader regardscalculates them the asdistance having between been emitted a tag from and aa differentreader using location, only because signal travel the reader time. calculates Thus, when the distancea reflected between signal aarrives tag and at athe reader reader, using the onlyreader signal estimates travel the time. distance Thus, as when D2 without a reflected any signal consideration arrives at of the reflection reader, the(straight reader line). estimates Thus, the if distancethe tracking as D2 signal without throug anyh consideration the direct path of reflection is not successfully (straight line). delivered, Thus, ifa thereflected tracking path signal can cause through an theestimation direct path error is as not grea successfullyt as the distance delivered, difference a reflected between path the can direct cause and an estimationreflected paths error (D2-D1). as great as the distance difference between the direct and reflected paths (D2-D1).

FigureFigure 1.1. ExampleExample ofof location-trackinglocation-tracking errorerror causedcaused byby multi-path.multi-path.

NLOSNLOS environmentsenvironments cancan heavilyheavily decreasedecrease thethe accuracyaccuracyof oflocation locationsensing. sensing. UsesUses ofof thethe referencereference tagstags andand thethe assistantassistant tagstags areare twotwo primaryprimary examplesexamples thatthat cancan increaseincrease thethe accuracyaccuracy ofof locationlocation sensingsensing withoutwithout thethe additionaladditional installationinstallation ofof readers.readers. TheThe conceptconcept ofof referencereference tagstags isis introducedintroduced inin LANDMARCLANDMARC [[19].19]. It It serves serves as as a reference a reference point point in the in system. the system. Its use Its does use not does require not require a large number a large numberof expensive of expensive RFID readers; RFID rather, readers; its rather, use requir its usees extra requires RFID extra tags, RFID which tags, are whichcheaper are than cheaper RFID thanreaders. RFID It can readers. also make It can location-sensing also make location-sensing results more accurate results more and reliable. accurate In and spite reliable. of reference In spite tags ofhaving reference advantages, tags having the installation advantages, of theadditional installation tags ofcan additional interrupt tagsconstruction can interrupt activities. construction In terms activities.of constructability, In terms of an constructability, assistant tag (active an assistant tag attached tag (active to other tag attached workers) to othercan be workers) an alternative. can be anAlthough alternative. the concept Although of thethe conceptassistant of tag the is assistant very similar tag is to very that similar of the toreference that of the tag, reference the primary tag, thedifference primary is differencethe installation is the place. installation In this paper, place. Inassistant this paper, tags are assistant attached tags to areworkers’ attached safety to workers’ helmets, safetyand tags helmets, follow and workers’ tags follow movements. workers’ When movements. a tag is When in an aNLOS tag is environment, in an NLOS environment, another tag anotherthat can tagbe detected that can bein an detected LOS environment in an LOS environment can communicate can communicate with the first withtag (in the the ﬁrst NLOS tag (inenvironment) the NLOS environment)and readers. andThis readers. approach This has approach been attempt has beened attemptedto convert to an convert NLOS an environment NLOS environment into a intoLOS aenvironment LOS environment [8,31]. [By8,31 utilizing]. By utilizing the concept the concept of assistant of assistant tags, interrupt tags, interruptionsions for construction for construction activities can be minimized and additional effort for installing reference tags or readers is not required. However, there are two remaining problem that can be described as determination of environment Sustainability 2018, 10, 1488 5 of 16 Sustainability 2018, 10, x FOR PEER REVIEW 5 of 16

condition and selection of proper RFID readers. If two problems remain, there may be less benefit to activities can be minimized and additional effort for installing reference tags or readers is not required. applying assistant tags. Thus, current problems for application of assistant tag concept will be However, there are two remaining problem that can be described as determination of environment addressed in this paper. condition and selection of proper RFID readers. If two problems remain, there may be less beneﬁt to applying assistant tags. Thus, current problems for application of assistant tag concept will be 3. Methodology addressed in this paper.

3.3.1. Methodology Algorithm Development to Overcome Multi-Path Generally, moving objects have their own movement characteristics in terms of direction, range 3.1.and Algorithm velocity. DevelopmentEffective use to of Overcome these characteristics Multi-Path of a target object can afford valuable information for Generally,filtering out moving location-estimation objects have theirerrors, own an movementd thereby characteristicsimprove tracking in terms accuracy of direction, [27,30]. rangeConsequently, and velocity. some notable Effective research use of has these applied characteristics the movement of a characteristics target object can of a affordtarget valuableobject for informationlocation sensing for ﬁltering [29–32]. out For location-estimation example, Son et al. errors, [29] introduced and thereby an improve enhanced tracking asset accuracy locating [system27,30]. Consequently,for vehicle pooling some in notable a port researchterminal has by appliedusing th thee information movement on characteristics the range and of speed a target of objecta transfer for locationcrane. sensing [29–32]. For example, Son et al. [29] introduced an enhanced asset locating system for vehicleSince pooling construction in a port equipment terminal by and using workers the information also have ontheir the own range movement and speed characteristics of a transfer crane. (e.g., operatingSince range, construction velocity, equipment and path), and the workersmovement also characteristics have their ownof a target movement object in characteristics construction (e.g.,projects operating are expected range, velocity, to improve and path), the theaccuracy movement of real-time characteristics location of a targettracking. object Use in construction of possible projectsmovement are expectedrange information to improve is the the accuracy simplest of way real-time to detect location location tracking. estimation Use of possibleerrors. If movement the arrow rangeline is information a worker’s isactual the simplest path and way the to area detect between location the estimation two dotted errors. lines Ifis thea possible arrow line range is a where worker’s the actualworker path could and move the area (see between Figure the2), twoapplication dotted lines of the is ainformation possible range of possible where the movement worker could range move can (seeeasily Figure detect2), applicationfour points of(A, the B, informationC, and D) as of locati possibleon estimation movement errors, range since can easily they are detect located four pointsoutside (A,the B, possible C, and range. D) as location estimation errors, since they are located outside the possible range.

FigureFigure 2.2.Elimination Elimination ofof trackingtrackingerrors errorsusing usingpossible possiblerange range information. information.

AsAs such,such, thethe possiblepossible movementmovement rangerange informationinformation isis helpfulhelpful toto improveimprove thethe accuracyaccuracy ofof aa location-sensinglocation-sensing system.system. However, However, this this kind of informationinformation isis notnot veryvery effectiveeffective forfor detectingdetecting localizationlocalization errorserrors withinwithin thethe range.range. Furthermore,Furthermore, itit isis tedioustedious andand time-consumingtime-consuming toto deﬁnedefine andand updateupdate thethe informationinformation ofof allall possiblepossible movementmovement ranges,ranges, particularlyparticularly inin aa constructionconstruction projectproject wherewhere possiblepossible movement movement range range can can vary vary signiﬁcantly significantly as as the the project project progresses. progresses. AcknowledgingAcknowledgingthese these challenges, challenges, this this paper paper proposes proposes the concept the concept of reducing of locationreducing estimation location errorsestimation by calculating errors by calculating distances between distances two between consecutive two consecutive logs (i.e., signals). logs (i.e., As signals). mentioned As mentioned above, a targetabove, object a target moving object in moving a construction in a construction site has its own site movementhas its own speed movement range. speed Several range. studies Several have suggestedstudies have [33, 34suggested] that many [33,34] people that tend many to walk people at about tend 1.4to walk m/s whenat about they 1.4 walk m/s on when general they pedestrian walk on walkways.general pedestrian Considering walkways. construction Considering sites usually constructi provideon harsh sites conditions, usually provide such as harsh poorly conditions, paved surfaces, such wetas poorly and oily paved surfaces, surfaces, and slope wet ways,and oily construction surfaces, workersand slope may ways, walk construction slower than generalworkers pedestrians. may walk Withslower this than in mind, general an experimentpedestrians. was With conducted this in mi onnd, a an construction experiment site. was The conducted participants on werea construction asked to allowsite. The recording participants (using videowere asked record) to of allow their recording walking during (using their video tasks. record) In this of experiment,their walking a total during number their oftasks. 1124 In steps this (strides) experiment, were analyzed.a total number Among of these 1124 1124 steps steps, (strides) only 27were steps analyzed. were over Among 70 cm/s. these Thus, 1124 in thissteps, study, only 70 27 cm/s steps is usedwere as over the threshold70 cm/s. ofThus, a possible in this movement study, 70 range. cm/s Figureis used3 illustratesas the threshold the concept of a ofpossible how to movement determine an range. error. Figure A worker 3 illustrates cannot move the moreconcept than of the how maximum to determine distance an calculatederror. A worker based cannot move more than the maximum distance calculated based on the worker’s maximum velocity and time (time difference between two consecutive logs). In Figure 3, n + 1th location log is within the Sustainability 2018, 10, 1488 6 of 16 on theSustainability worker’s 2018 maximum, 10, x FOR PEER velocity REVIEW and time (time difference between two consecutive logs). In Figure6 of 163 , n + 1th location log is within the range and can be determined as a proper log. However, n + 2th location range and can be determined as a proper log. However, n + 2th location log is located on outside of log isSustainability located on 2018 outside, 10, x FOR of thePEER range. REVIEW The n + 2th location log means that a worker moves beyond6 of his/her16 the range. The n + 2th location log means that a worker moves beyond his/her maximum speed, so maximum speed, so that this log can be determined as an error. that this log can be determined as an error. range and can be determined as a proper log. However, n + 2th location log is located on outside of the range. The n + 2th location log means that a worker moves beyond his/her maximum speed, so that this log can be determined as an error.

Figure 3. Concept of error log determination method. Figure 3. Concept of error log determination method. Figure 3. Concept of error log determination method. Figure 4 presents the procedure for how the suggested RTLS calculates every log. At first, the Figure4 presents the procedure for how the suggested RTLS calculates every log. At first, the distanceFigure and 4time presents difference the procedure between for two how consecutive the sugge logssted RTLS(nth and calculates n + 1th log) every are log. calculated. At first, the Based distanceon this and information, time difference the moving between velocity two consecutive between log logsn (defined (n and by n x +n, 1yn, log)zn, tn are) and calculated. logn+1 (defined Based by on distance and time difference between two consecutive logs (nthth and n + 1thth log) are calculated. Based n+1 n+1 n+1 n+1 thisx information,on, ythis, information,z , t the) is moving obtained the moving velocity using velocity Equation between between (1). log logn (definedn (defined by by x xnn,, yn,, z zn,n t,n t)n and) and log logn+1 (definedn+1 (defined by by xn+1,x yn+1n+1, y,n+1 zn+1, zn+1, t, n+1tn+1) is obtained obtained using using Equation Equation (1). (1).

FigureFigure 4. 4.4.Error Error Error determinationdetermination determination method. method.method. Sustainability 2018, 10, 1488 7 of 16

q 2 2 2 Sustainability 2018, 10, x FOR PEER REVIEW (xn+1 − xn) + (yn+1 − yn) + (zn+1 − zn) 7 of 16 Moving Velocity = (1) tn+1 − tn Moving Velocity (1) Then, if the moving velocity between a log (logn) and the next log (logn+1) exceeds the maximum velocity of a target object, the next log is regarded as a localization error and is eliminated in the Then, if the moving velocity between a log (logn) and the next log (logn+1) exceeds the maximum locatingvelocity process. of a target object, the next log is regarded as a localization error and is eliminated in the locatingAlthough process. an error can be identified by the aforementioned method, there is still a problem that is caused byAlthough serial errors. an error More can thanbe identified two consecutive by the afor errorementioned logs cannot method, be there dealt is with still usinga problem the suggestedthat method,is caused since by the serial above errors. method More isthan to determinetwo consecutive that logerrorn+1 logsexceeds cannot the be maximumdealt with using velocity the when lognsuggestedis within method, the maximum since the above velocity. method Therefore, is to determine error logsthat log aren+1 estimated exceeds the sequentially, maximum velocity and serial when logn is within the maximum velocity. Therefore, error logs are estimated sequentially, and serial errors should be eliminated. As shown in Figure5, log n+1 is eliminated, because this log is a certain errors should be eliminated. As shown in Figure 5, logn+1 is eliminated, because this log is a certain distance away from logn (exceeding the maximum velocity). Then, the locating system calculates distance away from logn (exceeding the maximum velocity). Then, the locating system calculates the the moving velocity between logn and logn+2. By repeating this filtering out process, logs that are moving velocity between logn and logn+2. By repeating this filtering out process, logs that are measured as moving faster than the maximum velocity are eliminated. The suggested process shown measured as moving faster than the maximum velocity are eliminated. The suggested process shown in Figurein Figure5 may 5 may provide provide too too few few valid valid logs logs ifif manymany logs logs are are filtered filtered out out as aserrors. errors. However, However, more more thanthan ten thousandten thousand signals signals are are transmitted transmitted per second second,, which which provides provides enough enough logs logs to locate to locate a moving a moving object,object, even even when when many many logs logs are are identified identified as errors.

FigureFigure 5. Elimination 5. Eliminationof of tracking tracking errors errorsusing using maximummaximum distance distance information: information: (a) ( aStep) Step 1; and 1; and (b) Step (b) Step 2. 2.

TimeTime synchronization synchronization between between tags tags and and readers readers is important is important when whenapplying applying this approach this approach for for filteringfiltering out location estimation estimation errors. errors. Since Since the the suggested suggested approach approach determines determines the location the location estimationestimation errors errors based based on on the the moving moving velocity of of a atarget target object, object, a slight a slight time time difference difference between between a a tag andtag and a reader a reader may may result result in in significant significant location-estimationlocation-estimation errors. errors. In fact, In fact, an RFID-based an RFID-based wireless wireless networknetwork is sensitive is sensitive to to this this type type of of errorerror since th thee network network transmits transmits a number a number of signals of signals for location for location tracking. To address this issue, several time synchronization protocols that have been suggested for tracking. To address this issue, several time synchronization protocols that have been suggested for wireless sensor network include reference broadcast synchronization (RBS) [35], timing-sync protocol wireless sensor network include reference broadcast synchronization (RBS) [35], timing-sync protocol for sensor network (TPSN) [36], and flooding time synchronization protocol (FTSP) [37]. TPSN was for sensorselected network as the time (TPSN) synchronization [36], and floodingprotocol for time our synchronization study because of its protocol effectiveness (FTSP) in addressing [37]. TPSN was selectedthe propagation as the time delay synchronization time that can protocoloften occur for in our a construction study because site. of its effectiveness in addressing the propagation delay time that can often occur in a construction site. 3.2. Application of Assistant Tags 3.2. Application of Assistant Tags In order to locate a target object using an RFID tag, at least three readers should receive signals fromIn order the tag. to locateSince an a RFID target tag object sends using signals an omnidirectionally, RFID tag, at least it is three difficult readers to accurately should receivelocate the signals fromtag the with tag. fewer Since than an RFIDthree readers. tag sends Thus, signals successful omnidirectionally, tracking of a moving it is difficult object requires to accurately careful locate Sustainability 2018, 10, 1488 8 of 16

Sustainability 2018, 10, x FOR PEER REVIEW 8 of 16 the tag with fewer than three readers. Thus, successful tracking of a moving object requires careful determination of the readers’ installationinstallation location so that at least three readers can receive the signals from aa tagtag whereverwherever itit isis inin thethe trackingtracking environment.environment. Obstacles betweenbetween a taga tag and and readers readers may causemay signalcause attenuationsignal attenuation and multi-path, and multi-path, thus preventing thus successfulpreventing communication successful communication between the between tag and thethe readers. tag and Multi-paththe readers. is Multi-path a propagation is a phenomenonpropagation inphenomenon which two orin which more paths two or of more radio paths signals of reachradio asi readergnals reach due toa reader reflection due from to reflection walls or from floors walls [36]. Signalor floors attenuation [36]. Signal is attenuation a gradual loss is a in gradual radio signalloss in intensity radio signal as it intensity propagates as it through propagates space through due to manyspace due effects, to many including effects, free-space including loss, free-space diffraction, loss, and diffraction, absorption and [38 absorption]. [38]. Suppose thatthat fourfour readersreaders areare installed installed in in each each corner corner of of the the room room but but only only two two readers readers (readers (readers C andC and D) D) receive receive signals signals from from the the tag, tag, while while the the other other two two (readers (readers A A and and B) B)do do notnot duedue toto anan obstacle (Figure(Figure6 6).). Thus, Thus, while while more more than than three three readers readers are are installed installed in in this this case, case, itit isis stillstill difficultdifficult toto accuratelyaccurately locate thethe tag,tag, sincesince onlyonly twotwo readersreaders receivereceive thethe signalssignals fromfrom thethe tag.tag.

Figure 6. Assistant tag for tracking a targettarget object in shadow area.

To address this problem, Lee et al. [8] suggest the idea of assistant tags, which are fundamentally To address this problem, Lee et al. [8] suggest the idea of assistant tags, which are fundamentally similar to reference tags, but differ in terms of their installation location. Lee et al. [8] suggest similar to reference tags, but differ in terms of their installation location. Lee et al. [8] suggest attaching attaching assistant tags to workers’ safety helmets, instead of on floors. Hence, tags are not installed assistant tags to workers’ safety helmets, instead of on floors. Hence, tags are not installed on some on some pre-fixed places on floors but attached to moving workers, which significantly increases the pre-fixed places on floors but attached to moving workers, which significantly increases the accuracy accuracy and applicability of RTLS. In the example shown in Figure 6, an assistant tag works as a and applicability of RTLS. In the example shown in Figure6, an assistant tag works as a virtual reader. virtual reader. That is, three readers (readers A, C and D) receive the signals from the assistant tag That is, three readers (readers A, C and D) receive the signals from the assistant tag and the moving and the moving location of the assistant tag is identified. Then, the assistant tag and the two readers location of the assistant tag is identified. Then, the assistant tag and the two readers (readers C and D) (readers C and D) can communicate with the tag in the shadow area. The location of the tag in the can communicate with the tag in the shadow area. The location of the tag in the shadow area can thus shadow area can thus be identified by using an assistant tag. be identified by using an assistant tag. However, their method has two main limitations. First, they did not suggest method to However, their method has two main limitations. First, they did not suggest method to determine determine whether the environment is LOS or NLOS. Thus, a tag that is not in a shadow area whether the environment is LOS or NLOS. Thus, a tag that is not in a shadow area continuously continuously communicates with an assistant tag. If there is no environment determination method, communicates with an assistant tag. If there is no environment determination method, the RTLS the RTLS engine can be influenced by multi-paths from assistant tags and other readers. As a result, engine can be influenced by multi-paths from assistant tags and other readers. As a result, the accuracy the accuracy of the location sensing decreases. Moreover, a reader tries to search for signals from a of the location sensing decreases. Moreover, a reader tries to search for signals from a target tag, which target tag, which can burden the system. This searching can also excessively consume the batteries of can burden the system. This searching can also excessively consume the batteries of readers and tags. readers and tags. To address this issue, this study suggests a system that can search blink signals from a tag and To address this issue, this study suggests a system that can search blink signals from a tag and counts the number of signals. More than three signals mean that the tag can be located without the counts the number of signals. More than three signals mean that the tag can be located without the help of an assistant tag, and the suggested system determines the location of the target tag based on help of an assistant tag, and the suggested system determines the location of the target tag based on the three readers. On the other hand, when fewer than three signals are detected, the suggested system the three readers. On the other hand, when fewer than three signals are detected, the suggested searches the nearest assistant tag to the target tag and estimates its location based on the two readers system searches the nearest assistant tag to the target tag and estimates its location based on the two and the nearest assistant tag. readers and the nearest assistant tag. Second, although Lee et al. [8] suggested the assistant tag concept for converting NLOS to LOS environment, they did not how to select a proper assistant tag from among various neighboring tags. Sustainability 2018, 10, 1488 9 of 16

SustainabilitySecond, 2018 although, 10, x FOR Lee PEER et REVIEW al. [8] suggested the assistant tag concept for converting NLOS to9 LOS of 16 environment, they did not how to select a proper assistant tag from among various neighboring tags. InIn this paper, paper, to to select select a aspecific speciﬁc assistant assistant tag, tag, geometric geometric dilution dilution of precision of precision (GDOP) (GDOP) is used. is used.GDOP GDOP has been has widely been widely used as used an accuracy as an accuracy metric metricfor navigation for navigation and tracking and tracking systems systems [39]. GDOP [39]. GDOPdescribes describes error caused error caused by the by relative the relative position position of GPS of GPS satellites. satellites. In Inthis this study, study, GPS GPS satellites satellites are convertedconverted toto thethe readers. This method is firstly ﬁrstly used to select proper readers for accurate positioning. TheThe arrangementarrangement of of multiple multiple readers readers according according to to the th relativee relative position position of theof the readers readers can can determine determine the levelthe level of precision of precision in estimating in estimating a target a target location. locati Figureon. Figure7 presents 7 presents examples examples of good of (low)good and(low) poor and (high)poor (high) GDOP. GDOP. The measurement The measurement has error has ranges, error andranges the, trueand locationthe true oflocation the target of the will target lie anywhere will lie inanywhere the red area.in the In red Figure area.7b, In the Figure error 7b, range the iserror the samerange as is in the Figure same7 a,as but in Figure the possible 7a, but position the possible area (redposition area) area has considerably(red area) has grown considerably due to the grown arrangement due to the of thearrangement two readers. of Thesethe two results readers. mean These that theresults close mean arrangement that the close of readers arrangement can increase of readers the uncertainty can increase of locationthe uncertainty sensing. of location sensing.

Figure 7. Examples of GDOP: (a) good (low) GDOP; ((b)) poorpoor (high)(high) GDOP.GDOP.

At first, the most recently updated tags among tags with connectivity are extracted. Each At ﬁrst, the most recently updated tags among tags with connectivity are extracted. Each selected selected tag and reader receives a blink message from a target tag, which is used to calculate GDOP tag and reader receives a blink message from a target tag, which is used to calculate GDOP rate. rate. After calculating the GDOP rate of each connected assistant tag, an assistant tag with the best After calculating the GDOP rate of each connected assistant tag, an assistant tag with the best GDOP GDOP rate (lowest value) is selected and used to estimate a target location. By selecting an rate (lowest value) is selected and used to estimate a target location. By selecting an appropriate appropriate assistant tag, NLOS environments can be converted to LOS environments and the assistant tag, NLOS environments can be converted to LOS environments and the accuracy of location accuracy of location sensing can be also improved. sensing can be also improved.

3.3.3.3. Experimental Settings TagsTags and readers are are two two main main devices devices required required for for locating locating a target a target object. object. A tag A tagattached attached to a totarget a target object object sends sends signals signals at a at regular a regular interval. interval. A Atag tag consists consists of of a a microcontroller, microcontroller, a wirelesswireless communicationcommunication module, and a sensorsensor module.module. In In this this paper, the MSP430F5xxx series by Texas Instruments,Instruments, whichwhich givesgives satisfactorysatisfactory performanceperformance and low powerpower consumption,consumption, was selectedselected as aa microcontroller.microcontroller. The The radio radio transceiver transceiver is based is based on IEEE on 802.15.4a,IEEE 802.15.4a, which useswhich the uses chirp the spread chirp spectrum spread (CSS)spectrum on the (CSS) 2.4 GHzon the industrial 2.4 GHz science industrial medical scienc (ISM)e medical band. (ISM) In addition, band. In the addition, tag uses the acceleration tag uses sensorsacceleration to calculate sensors the to velocitycalculate of the a target velocity object, of a sotarget that aobject, candidate so that location a candidate range canlocation be determined. range can be determined.A reader consists of a core process module and a wireless communication module. It receives signalsA transmittedreader consists from of aa target core process tag and module sends them and backa wireless to the targetcommunication tag to estimate module. the tag-readerIt receives distance.signals transmitted In this paper, from the a readerstarget tag were and selected sends them as the back Cortex-M3 to the target produced tag to by estimate the ARM the corporation, tag-reader whichdistance. can In receive this paper, location the readers logs in were 72 MHz. selected This as reader the Cortex-M3 also includes produced 128 KB by ﬂash the ARM and 20corporation, KB RAM. Therewhich arecan two receive types location of readers: logs stationaryin 72 MHz. and This portable. reader also In general, includes the 128 stationary KB flash typeand 20 is preferredKB RAM. whenThere thereare two is a types power of supply. readers: Otherwise, stationary the and portable portable. type In isgeneral, preferred thesince stationary it works type with is preferred a battery. Thewhen battery there lifeis a ofpower the reader supply. (portable Otherwise, type) the selected portable in type this paperis preferred was more since than it works three with years. a battery. The batteryIn order life to of locate the reader a target (portable object using type) an selected RFID tagin this and paper readers, was a more target than tag ﬁrstthree sends years. a signal in orderIn order to search to locate for adjacenta target object readers. using A tag an sendsRFID tag a signal and readers, omnidirectionally, a target tag becausefirst sends it discerns a signal in order to search for adjacent readers. A tag sends a signal omnidirectionally, because it discerns in which direction adjacent readers are placed. Once readers successfully receive the signal, they transmit an acknowledgement (ACK) signal to the tags (or assistant tags) in order to confirm that they successfully received the signal for the tags. Then, the readers calculate the distance to the tags Sustainability 2018, 10, 1488 10 of 16

inSustainability which direction 2018, 10, x adjacent FOR PEERreaders REVIEW are placed. Once readers successfully receive the signal, 10 they of 16 transmit an acknowledgement (ACK) signal to the tags (or assistant tags) in order to confirm that theyand send successfully the distance received list to the the signalRTLS engine for the that tags. locates Then, a target the readers object based calculate on distance the distance information to the tagsfrom and at least send three the distance readers listadjacent to the to RTLS a target engine tag. thatFor locatesthe distance a target measurement, object based the on symmetric distance informationdouble-sided from two-way at least ranging three readers(SDS-TWR) adjacent method—a to a target ranging tag. method For the distancethat utilizes measurement, two delays thethat symmetricnaturally occur double-sided in the signal two-way transmission ranging (SDS-TWR)process for method—adetermining ranging the range method between that two utilizes wireless two delaysdevices—was that naturally selected occur in inthe the RTLS signal engine transmission because process it can formeasure determining the round-trip the range betweentime without two wirelesssynchronizing devices—was the measurement selected in theclocks RTLS [40], engine thereby because significantly it can measure reducing the the round-trip possibility time of withouttracking synchronizingerrors due to clock the measurement synchronization. clocks [40], thereby significantly reducing the possibility of tracking errorsUnder due to these clock synchronization.experimental settings, the suggested system was tested in terms of its location trackingUnder accuracy these experimental and location estimation settings, the time, suggested since these system are two was primary tested in factors terms in of assessing its location the trackingeffectiveness accuracy of an and RTLS location [41]. estimation time, since these are two primary factors in assessing the effectiveness of an RTLS [41]. 4. Results 4. Results 4.1. Location Tracking Accuracy 4.1. Location Tracking Accuracy Tests for the accuracy of the suggested system were conducted in two different tracking environments.Tests for the The accuracy first test of was the performed suggested in system a basement were conductedparking lot inof twoan ongoing different apartment tracking environments.project. The test The area first had test no wasobstacles performed except in so ame basement structural parking columns. lot The of an test ongoing analyzed apartment tracking project.logs of Thea worker test area moving had noaround obstacles a 7.5 except m × 5.5 some m rectangle. structural The columns. moving The velocity test analyzed of the worker tracking was logs set ofin a the worker range moving of 0 cm/s around (stop a 7.5state) m × to5.5 70 m cm/s. rectangle. Figure The 8a movingshows the velocity tracking of the logs worker when was the set location in the rangetracking of 0 error cm/s mitigation (stop state) algorithms to 70 cm/s. suggested Figure8a in shows this paper the tracking were not logs applied. when theIn this location case, tracking2648 logs errorwere mitigationtracked, and algorithms the average suggested error was in this around paper 38 were cm. notOn applied.the other In hand, this case,Figure 2648 8b logsshows were the tracked,tracking and logs the after average eliminating error wasthe tracking around 38errors cm. identified On the other by the hand, location Figure tracking8b shows error the mitigation tracking logsalgorithms after eliminating suggested the in trackingthis paper. errors In comparison identified by with the location the results tracking shown error in Figure mitigation 8a, the algorithms average suggestedtracking error in this was paper. reduced In comparison to around with21 cm; the a results44% improvement shown in Figure in location8a, the averagetracking trackingaccuracy. error This wasimprovement reduced to was around obtained 21 cm; by a eliminating 44% improvement the logs inthat location exceeded tracking the maximum accuracy. distance This improvement that a target wasobject obtained can move by eliminating in a specific the time logs interval. that exceeded In this test, the maximum708 out of 2648 distance logs that were a targetfiltered object out as can tracking move inerrors a specific by the time suggested interval. algorithms. In this test, These 708 out results of 2648 demonstrate logs were filteredthe strong out potential as tracking of the errors suggested by the suggestedlocation tracking algorithms. error These mitigati resultson demonstrate algorithms the to strongimprove potential the accuracy of the suggested of location-tracking location tracking in errorconstruction mitigation projects. algorithms to improve the accuracy of location-tracking in construction projects.

FigureFigure 8. 8.Elimination Elimination of of tracking tracking errors errors using using possible possible range range information: information: ( a(a)) Before Before eliminating eliminating the the trackingtracking errors; errors; and and ( b(b)) After After eliminating eliminating the the tracking tracking errors. errors.

The suggested system was further applied in eight different construction environments in order to test its applicability and effectiveness under various situations, as shown in Figure9. In each testing environment, four different experiments (A to D) were conducted in order to measure the effect of the location tracking error mitigation algorithms and the application of the assistant tag, as shown in Table1. The four different experiments are as follows. Sustainability 2018, 10, x FOR PEER REVIEW 11 of 16 • Accuracy experiment A: location tracking applying neither the error mitigation algorithms nor  theAccuracy assistant experiment tags; A: location tracking applying neither the error mitigation algorithms nor • Accuracythe assistant experiment tags; B: location tracking applying the error mitigation algorithms only • Accuracy experiment C:B: location tracking applyingapplying the assistanterror mitigation tags only algorithms only • Accuracy experiment D:C: locationlocation tracking applyingapplying boththe assistant the error tags mitigation only algorithms and the  assistantAccuracy tags. experiment D: location tracking applying both the error mitigation algorithms and the assistant tags.

Figure 9. Location tracking accuracy tests.

Table1 shows the trackingTable results 1. Location of the tracking four accuracy different test experiments results. under eight different construction environments. As expected, the average trackingLocation Tracking error (256 Accuracy cm) was greatest in experimentTests A. InTracking contrast, Environments experiment D showedExperiment the lowest Experiment tracking errorsExperiment (58 cm), Experiment resulting in a 77.3% improvement in location accuracy in comparisonA with experimentB A. C D Earth Work Area Tracking Error 138 cm 61 cm 47 cm 38 cm TestIn 1 particular, experiment A showed a signiﬁcant tracking error (791 cm) in stock yard environments(Outdoor) where H-beams Improvement were piled (i.e., experiment- A 55.8% under test 4).65.9% This is because72.5% the Steel Work Area Tracking Error 158 cm 120 cm 96 cm 70 cm steelTest of 2 the H-beams easily causes heavy multi-path and signal attenuation. In this case, applying both (Outdoor) Improvement - 24.1% 39.2% 55.7% the algorithms and the assistant tags (experiment D) can reduce the location tracking error to 96.5 cm, Underground Tracking Error 116 cm 73 cm 76 cm 51 cm 87.9%Test 3 of the improvement in tracking accuracy from experiment A. The test results showed that the Parking Lot Improvement - 37.1% 34.5% 56.0% Stock Yard Tracking Error 791 cm 321 cm 148 cm 96 cm Test 4 (H-Beam piled) Improvement - 59.4% 81.3% 87.9% Stock Yard Tracking Error 549 cm 243 cm 152 cm 88 cm Test 5 (Brick piled) Improvement - 55.7% 72.3% 84.0% Frame Structure Tracking Error 127 cm 80 cm 78 cm 53 cm Test 6 (Indoor) Improvement - 37.0% 38.6% 58.3% Wall Structure Tracking Error 78 cm 71 cm 57 cm 35 cm Test 7 (Indoor) Improvement - 9.0% 26.9% 55.1% Test 8 Tracking Error 86 cm 46 cm 61 cm 33 cm Sustainability 2018, 10, 1488 12 of 16

65% average improvement in experiment C was slightly higher than that in experiment B (50.3%), which suggests that assistant tags are more effective than the algorithms in tracking environments where there are many obstacles between readers and tags.

Table 1. Location tracking accuracy test results.

Location Tracking Accuracy Tests Tracking Environments Experiment Experiment Experiment Experiment A B C D Earth Work Area Tracking Error 138 cm 61 cm 47 cm 38 cm Test 1 (Outdoor) Improvement - 55.8% 65.9% 72.5% Steel Work Area Tracking Error 158 cm 120 cm 96 cm 70 cm Test 2 (Outdoor) Improvement - 24.1% 39.2% 55.7% Underground Tracking Error 116 cm 73 cm 76 cm 51 cm Test 3 Parking Lot Improvement - 37.1% 34.5% 56.0% Stock Yard Tracking Error 791 cm 321 cm 148 cm 96 cm Test 4 (H-Beam piled) Improvement - 59.4% 81.3% 87.9% Stock Yard Tracking Error 549 cm 243 cm 152 cm 88 cm Test 5 (Brick piled) Improvement - 55.7% 72.3% 84.0% Frame Structure Tracking Error 127 cm 80 cm 78 cm 53 cm Test 6 (Indoor) Improvement - 37.0% 38.6% 58.3% Wall Structure Tracking Error 78 cm 71 cm 57 cm 35 cm Test 7 (Indoor) Improvement - 9.0% 26.9% 55.1% Finishing Work Tracking Error 86 cm 46 cm 61 cm 33 cm Test 8 Area Improvement - 46.5% 29.1% 61.6% Tracking Error 255 cm 127 cm 89 cm 58 cm Average Improvement - 50.3% 65.0% 77.3%

4.2. Location Tracking Estimation Time Location tracking estimation time is an important factor in evaluating performance of an RTLS. Even though RTLS is very accurate, if locating a moving object is too slow, the RTLS is meaningless. The two methods suggested herein for improving the accuracy of RTLS (i.e., location tracking error mitigation algorithms and assistant tags) can significantly increase the location tracking estimation time. To address this issue, this paper suggests a floor classification method that does not measure the z-coordinate but uses the information of floor level in order to reduce the location tracking estimation time while maintaining the accuracy of location tracking. That is, two-dimensional coordinates (x- and y-), rather than three-dimensional coordinates (x-, y- and z-), are used for tracking the location of the target object by arbitrarily assigning the information of the floor number to the z-coordinate. In order to measure the impact of the floor classification method in reducing location tracking estimation time, the test was conducted under the following three different settings.

• Time experiment A: measuring the estimation time when applying neither location tracking error mitigation algorithms nor assistant tags (i.e., the same to the accuracy experiment A in Table1). • Time experiment B: measuring the estimation time when applying both location tracking error mitigation algorithms and assistant tags (i.e., the same to the accuracy experiment D in Table1). • Time experiment C: measuring the estimation time when applying the ﬂoor classiﬁcation method in addition to the error mitigation algorithms and assistant.

These three experiments were conducted in three different settings: one floor with four readers, three floors with twelve readers (four readers on each floor), and five floors with twenty readers (four readers on each floor). Furthermore, each experiment was conducted to measure the estimation time in tracking a different number of target tags (i.e., 10 to 50 tags), as shown in Table2. Sustainability 2018, 10, 1488 13 of 16

Table 2. Location tracking estimation time test results.

Experimental Settings Location Tracking Estimation Time (ms) Time Reduction Number of Number of Number of Experiment A Experiment B Experiment C ((B–C)/B) Floors Readers Tags 10 157 227 173 23.8% 1 4 20 282 348 309 11.2% 30 418 525 445 15.2% 10 284 377 307 18.6% 3 12 30 376 497 424 14.7% 50 483 687 519 24.5% 10 254 409 298 27.1% 5 20 30 448 564 420 25.5% 50 756 863 628 27.2% Average 384.2 499.7 391.4 21.7%

As expected, the time experiment A required the shortest time for location estimation and the time experiment B required the longest time. This was mainly attributed to the increased level of computational time and effort to apply the location tracking error mitigation algorithms and assistant tags. This result reveals a trade-off between location tracking accuracy and estimation time. That is, application of the error mitigation algorithms and assistant tags can increase the tracking accuracy by 77.3% (see Table1) at the cost of 115.5 ms extra computational time. Although seemingly negligible, 115.5 ms may be significant, particularly in RTLS. There is one thing to be noticed. According to the results, there is a clear trade-off between accuracy and computational time by comparing experiment A and B. Greater accuracy required more computational time. This study used up to 50 tags. In these cases (three floors and five floors case), the time differences are approximately 0.2 s (204 ms) and 0.1 s (107 ms) respectively. These differences seem to suggest that the additional computational time may not be crucial. However, as the number of tags increases, the computational time also increases. If the required time increases excessively when the number of tags is more than that of this experiment (i.e., more than 50), the increase of required time cancels the benefit of increased accuracy. For example, an increased time of more than 3 s—in which workers can move up to approximately 2 m—leads to a safety management system potentially being late to respond to a hazardous situation. In comparison with time experiment B, time experiment C required a similar time (391.4 ms) to that of time experiment A, while maintaining the same level of accuracy as experiment B. This result confirmed the effectiveness of the suggested floor classification algorithms in reducing location tracking estimation time, thus expanding the applicability of the suggested system for real-time location tracking purposes.

5. Conclusions RTLS technology has been widely used for various purposes, including measuring productivity, monitoring equipment and workers, and preventing safety accidents. Among various objectives for using RTLS technologies, preventing safety accidents requires much higher accuracy because the accuracy is directly related to a human’s life. This paper suggested an RFID-based RTLS capable of enhancing the accuracy of real-time location tracking in construction projects. Specifically, location tracking error mitigation algorithms were introduced in order to filter out erroneous logs due to multi-path, and assistant tags were used to increase the accuracy in tracking an object in a shadow area. In particular, to address the multi-path problem, the construction workers’ maximum moving velocity is applied to develop the suggested algorithm. Additionally, in the application of assistant tags, the suggested method of how to select proper readers is incorporated in the current assistant method. The developed system was applied to a construction project and tested in terms of its location tracking accuracy and location estimation time. The test results showed that the application of error Sustainability 2018, 10, 1488 14 of 16 mitigation algorithms and assistant tags could increase location tracking accuracy by 77.3%, thereby demonstrating the strong potential of the algorithms and assistant tags to enhance location tracking accuracy in construction projects. However, this increased accuracy was gained at the expense of extended location estimation time, which is another important aspect in assessing the performance of an RTLS. In order to minimize the extra computational time for RTLS while maintaining its accuracy, a floor classification method that does not calculate the z-axis coordinate was proposed, thereby decreasing the computational time and effort. Although the suggested system proved its applicability and effectiveness for real-time location tracking in construction sites, the following limitations need to be addressed in further research. First, the suggested method could be carried out in combination with Kalman filtering considering geometric constraints. Second, the performance of the system needs to be further validated in many different construction environments. In particular, the application of assistant tags may not be operated when a construction worker does his task solely without other workers who can take a role as an assistant tag. However, considering the fact that construction workers usually perform tasks as a group effort, this problem may not be critical. Lastly, the maximum speed of a target object should be tested in various situations in order to improve the effectiveness of the location tracking error mitigation algorithms.

Author Contributions: H.K. and S.H. conceived and designed the experiments; H.K. performed the experiments; H.K. developed the error mitigation algorithm; S.H. developed the assistant tag selection method; H.K. and S.H. wrote the paper. Acknowledgments: This work was supported by Gyeongnam National University of Science and Technology Grant in 2017. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Amponsah-Tawiah, K. Occupational health and safety and sustainable development in Ghana. Int. J. Bus. Adm. 2013, 4, 74. [CrossRef] 2. McKeown, R. Education for Sustainable Development Toolkit; Portland State University: Portland, OR, USA, 2002; Available online: http://www.esdtoolkit.org/ (accessed on 27 April 2018). 3. Willard, B. The New Sustainability Advantage: Seven Business Case Beneﬁts of a Triple Bottom Line; New Society Publishers: Gabriola, BC, Canada, 2012. 4. Navon, R.; Kolton, O. Model for Automated Monitoring of Fall Hazards in Building Construction. J. Constr. Eng. Manag. 2006, 132, 733–740. [CrossRef] 5. Fullerton, C.; Allread, B.; Teizer, J. Pro-Active-Real-Time Personnel Warning System. Available online: https://ascelibrary.org/doi/abs/10.1061/41020(339)4 (accessed on 8 April 2018). 6. Wu, W.; Yang, H.; Chew, D.A.S.; Yang, S.; Gibb, A.G.F.; Li, Q. Towards an autonomous real-time tracking system of near-miss accidents on construction sites. Autom. Constr. 2010, 19, 134–141. [CrossRef] 7. Chae, S.; Yoshida, T. Application of RFID technology to prevention of collision accident with heavy equipment. Autom. Constr. 2010, 19, 368–374. [CrossRef] 8. Lee, H.; Lee, K.; Park, M.; Baek, Y.; Lee, S. RFID-Based Real-Time Locating System for Construction Safety Management. J. Comput. Civ. Eng. 2012, 26, 366–377. [CrossRef] 9. Teizer, J.; Castro-Lacouture, D. Combined Ultra-Wideband Positioning and Range Imaging Sensing for Productivity and Safety Monitoring in Building Construction. In Proceedings of the 2007 ASCE International Workshop on Computing in Engineering, Reston, VA, USA, 24–27 July 2007; pp. 681–688. 10. Teizer, J.; Caldas, C.; Haas, C. Real-Time Three-Dimensional Occupancy Grid Modeling for the Detection and Tracking of Construction Resources. J. Constr. Eng. Manag. 2007, 133, 880–888. [CrossRef] 11. Chin, S.; Yoon, S.; Choi, C.; Cho, C. RFID+4D CAD for Progress Management of Structural Steel Works in High-Rise Buildings. J. Comput. Civ. Eng. 2008, 22, 74–89. [CrossRef] 12. Cheng, T.; Teizer, J. Real-Time Data Collection and Visualization Technology in Construction. In Proceedings of the Construction Research Congress, ASCE, Reston, VA, USA, 8–10 May 2010; pp. 339–348. 13. Song, J.; Haas, C.; Caldas, C. Tracking the Location of Materials on Construction Job Sites. J. Constr. Eng. Manag. 2006, 132, 911–918. [CrossRef] Sustainability 2018, 10, 1488 15 of 16

14. Torrent, D.; Caldas, C. Methodology for Automating the Identification and Localization of Construction Components on Industrial Projects. J. Comput. Civ. Eng. 2009, 23, 3–13. [CrossRef] 15. Saidi, K.S.; Teizer, J.; Franaszek, M.; Lytle, A.M. Static and dynamic performance evaluation of a commercially-available ultra wideband tracking system. Autom. Constr. 2011, 20, 519–530. [CrossRef] 16. Kim, H.; Lee, H.; Park, M.; Chung, B.; Hwang, S. Automated hazardous area identification using laborers’ actual and optimal routes. Autom. Constr. 2016, 65, 21–32. [CrossRef] 17. Kim, H.; Ahn, C.R.; Yang, K. Identifying Safety Hazards Using Collective Bodily Responses of Workers. J. Constr. Eng. Manag. 2017, 143, 04016090–1-12. [CrossRef] 18. Garmin Corporation, About GPS. Available online: http://www8.garmin.com/aboutGPS/ (accessed on 4 April 2018). 19. Ni, L.M.; Liu, Y.; Lau, Y.C.; Patil, A.P. LANDMARC: Indoor location sensing using active RFID. Wirel. Netw. 2004, 10, 701–710. [CrossRef] 20. Jin, G.Y.; Lu, X.Y.; Park, M.S. An indoor localization mechanism using active RFID tag. In Proceedings of the IEEE International Confernational on Sensor Networks, Ubiquitous, and Trustworthy Computing, Taichung, Taiwan, 5–7 June 2006; pp. 1–4. 21. Han, J.; Zhao, Y.; Cheng, Y.S.; Wong, T.L.; Wong, C.H. Improving Accuracy for 3D RFID Localization. Int. J. Distrib. Sens. Netw. 2012, 2012, 865184. [CrossRef] 22. Ijaz, F.; Yang, H.K.; Ahmad, A.W.; Lee, C. Indoor positioning: A review of indoor ultrasonic positioning systems. In Proceedings of the IEEE 15th International Conference on Advanced Communication Technology (ICACT), PyeongChang, Korea, 27–30 January 2013; pp. 1146–1150. 23. Zhang, D.; Xia, F.; Yang, Z.; Yao, L.; Zhao, W. Localization technologies for indoor human tracking. In Proceedings of the IEEE 5th International Conference on Future Information Technology (FutureTech), Busan, Korea, 21–23 May 2010; pp. 1–6. 24. Yarovoy, A.G.; Zhuge, X.; Savelyev, T.G.; Ligthart, L.P. Comparison of UWB technologies for human being detection with radar. In Proceedings of the IEEE European Microwave Conference, Munich, Germany, 9–12 October 2007; pp. 1574–1577. 25. Mora, H.; Gilart-Iglesias, V.; Pérez-del Hoyo, R.; Andújar-Montoya, M.D. A Comprehensive System for Monitoring Urban Accessibility in Smart Cities. Sensors 2017, 17, 1834. [CrossRef][PubMed] 26. Mora-Mora, H.; Gilart-Iglesias, V.; Gil, D.; Sirvent-Llamas, A. A computational architecture based on RFID sensors for traceability in smart cities. Sensors 2015, 15, 13591–13626. [CrossRef][PubMed] 27. Gilart-Iglesias, V.; Mora, H.; Pérez-delHoyo, R.; García-Mayor, C. A computational method based on radio frequency technologies for the analysis of accessibility of disabled people in sustainable cities. Sustainability 2015, 7, 14935–14963. [CrossRef] 28. Skibniewski, M.J.; Jang, W.S. Localization Technique for Automated Tracking of Construction Materials Utilizing Combined RF and Ultrasound Sensor Interfaces. In Proceedings of the ASCE Computing in Civil Engineering, Reston, VA, USA, 24–27 July 2007; pp. 657–664. 29. Son, S.; Park, Y.; Baek, Y. Design and Implementation of Enhanced Asset Locating System for Vehicle Pooling in a Port Terminal. In Proceedings of the 2013 IEEE 27th International Conference on Advanced Information Networking and Applications (AINA), Barcelona, Spain, 25–28 March 2013; pp. 710–715. 30. Lei, Z.; Taghaddos, H.; Olearczyk, J.; Al-Hussein, M.; Hermann, U. Automated Method for Checking Crane Paths for Heavy Lifts in Industrial Projects. J. Constr. Eng. Manag. 2013, 139, 04013011. [CrossRef] 31. Cho, H.; Jung, Y.; Choi, H.; Jang, H.; Son, S.; Baek, Y. Precise location tracking system based on time difference of arrival over LR-WPAN. In Proceedings of the First ACM International Workshop on Mobile Entity Localization and Tracking in GPS-Less Environments, San Francisco, CA, USA, 19 September 2008; pp. 67–72. 32. Mullen, D. The application of RFID technology in a Port. Port Technol. Int. 2005, 22, 181–182. 33. Browning, R.C.; Baker, E.A.; Herron, J.A.; Kram, R. Effects of obesity and sex on the energetic cost and preferred speed of walking. J. Appl. Physiol. 2006, 100, 390–398. [CrossRef][PubMed] 34. Mohler, B.J.; Thompson, W.B.; Creem-Regehr, S.H.; Pick, H.L., Jr.; Warren, W.H., Jr. Visual flow influences gait transition speed and preferred walking speed. Exp. Brain Res. 2007, 181, 221–228. [CrossRef][PubMed] 35. Elson, J.; Girod, L.; Estrin, D. Fine-grained network time synchronization using reference broadcasts. ACM SIGOPS Oper. Syst. Rev. 2002, 36, 147–163. [CrossRef] Sustainability 2018, 10, 1488 16 of 16

36. Ganeriwal, S.; Kumar, R.; Srivastava, M.B. Timing-sync protocol for sensor networks. In Proceedings of the ACM 1st International Conference on Embedded Networked Sensor Systems, Los Angeles, CA, USA, 5–7 November 2003; pp. 138–149. 37. Maróti, M.; Kusy, B.; Simon, G.; Lédeczi, Á. The ﬂooding time synchronization protocol. In Proceedings of the ACM 2nd International Conference on Embedded Networked Sensor Systems, Baltimore, MD, USA, 3–5 November 2004; pp. 39–49. 38. Rappaport, T.S. Wireless Communications: Principles and Practice; Prentice Hall PTR: Upper Saddle River, NJ, USA, 1996; Volume 2. 39. Yu, K.; Sharp, I.; Guo, Y.J. Ground-Based Wireless Positioning; John Wiley & Sons: Hoboken, NJ, USA, 2009; Volume 5. 40. Rohrig, C.; Muller, M. Indoor location tracking in non-line-of-sight environments using a IEEE 802.15. 4a wireless network. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems IROS, St. Louis, MO, USA, 10–15 October 2009; pp. 552–557. 41. Park, S.H.; Kim, Y.J.; Lee, H.C.; Lim, M.T. Improved adaptive particle ﬁlter using adjusted variance and gradient data. In Proceedings of the IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, Seoul, Korea, 20–22 August 2008; pp. 650–655.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).