Вы находитесь на странице: 1из 12

Advanced Engineering Informatics 25 (2011) 535546

Contents lists available at ScienceDirect

Advanced Engineering Informatics

journal homepage: www.elsevier.com/locate/aei

Performance-based evaluation of RFID-based indoor location sensing solutions for the built environment
Nan Li, Burcin Becerik-Gerber
Department of Civil and Environmental Engineering, University of Southern California, Los Angeles, CA 90089, United States

a r t i c l e

i n f o

a b s t r a c t
Indoor location information is of great value to the building industry in improving the utilization and maintenance of facilities. The paper identies previous academic accomplishments of radio frequency identication (RFID)-based indoor location sensing (ILS) solutions. The paper summarizes the major location sensing methods used in previous RFID-based solutions, and provides a review of 21 research projects, with their algorithm design, devices, test setup, and performance evaluation presented in detail. Based on this review, the paper summarizes the intensive use of the proximity method in RFID-based ILS, and analyzes the underlying rationale. The ndings point out that no single solution satises all criteria for widespread implementations, and that the adaptability of these solutions to built environments need to be further justied. Finally, the paper outlines the gaps for future research, including modifying ILS solution design, developing a seamless outdoor/indoor location sensing solution, and building a context-aware information delivery mechanism for the building industry. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 19 July 2010 Received in revised form 15 February 2011 Accepted 18 February 2011 Available online 23 March 2011 Keywords: Indoor Location sensing RFID Built environment Accuracy Affordability

1. Introduction Location information is crucial to a variety of standard and personalized applications in a wide range of industries such as transportation, manufacturing, logistics, and healthcare, and it is the basis for delivery of personalized and location-based services [1]. Examples include applications that show drivers their vicinity and guide them to their destinations [2], applications that enable the user to search for published services within his/her immediate vicinity [3], and applications that monitor the users indoor location, and adjust the music played in rooms based on user preferences [4]. While global positioning system (GPS) technology has well met the need for outdoor location sensing (OLS), so far no single solution to the indoor location sensing (ILS) problem has been universally adopted. Various technologies have been proposed and tested for ILS, among which are indoor GPS, motion and rotation sensors, infrared, ultrasound, ultra wide band (UWB), wireless local area network (WLAN), and radio frequency identication (RFID). Indoor location information is also of great value to the building industry, and has potential to improve the utilization and maintenance of facilities. For example, occupants unfamiliar with a built environment, ranging in scale from a building to a neighborhood to large-scale civic surroundings, could be provided with location

Corresponding author.
E-mail address: becerik@usc.edu (B. Becerik-Gerber). 1474-0346/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.aei.2011.02.004

information to navigate around and nd their destinations; facility management (FM) personnel could be provided with locations of building components or equipment they need to maintain or repair; locations of tools and on-site FM personnel and the length of time they spend at each location could be analyzed to monitor the work procedures and improve productivity; changes in building occupancy could be detected in real time through location sensing, and energy conservation measures, such as adjustment of lighting and air conditioning, could be automated. Moreover, ILS also lays the basis for context awareness within the built environment, which relies on automatic recognition of both the users location and activity [5]. Context-aware information can automate the delivery of spatial information to on-site mobile personnel, with which targets, including building components, equipment, tools and people, can be easily located and target-specic information can be accessed on-site. This study focuses on ILS solutions that are built on RFID technology. Prior research has proven RFID technologys capability of providing accurate and cost efcient indoor location information. Moreover, RFID technology is applicable to the built environment because of its non-line-of-sight characteristic, wireless communication and on-board data storage capacities, and its wide use in the building industry [68]. However, the focus building industry researchers currently place on the development of an applicable RFID-based ILS solution could be further strengthened. The reasons are the research efforts in this area have been mostly concentrated in electrical engineering (EE) and computer science (CS) elds, with a focus on solutions technical improvement. However, the


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546

collaboration of multiple disciplines including civil engineering (CE) is required to identify RFID-based ILS solutions value in practice, increase their adaptability to the built environment, and develop extended applications built on the ILS solutions. Besides, future research directions are needed and they may drive technical and methodological development to support a wider and more sophisticated adoption by the building industry. In order to address these challenges and provide a comprehensive review of research in this area, 21 RFID-based ILS solutions are included in this paper. Algorithm design, devices, test setup, and performance evaluation of these solutions are described in detail. In addition, the potential value of these solutions and of services built using them are analyzed, and future research directions are presented. 2. Overview of RFID technology A typical RFID system consists of two components, a reader and a tag, and operates at a certain frequency, as shown in Fig. 1. A tag contains a microchip and an internal antenna. Attached to an object, a tag stores a specic ID and other object-related data, and sends the data to a reader upon its request. Tags can be distinguished as passive or active, according to their power source. Passive tags need to be activated by the electromagnetic energy the reader emits and depend on that for power to operate. Therefore, they have shorter read ranges and smaller data storage capacities. Active tags rely on internal batteries for power supply, which en-

Fig. 1. RFID system components.

hances the read ranges signicantly and enables additional onboard memory and local sensing and processing capacities. However, using a local power source also limits active tags lifetime to 510 years [9] and increases the cost. To bridge the gap between passive and active tags, a third type of tag, battery-assisted passive (BAP) tags or semi-passive tags, has been introduced; these tags use internal batteries to power the chips, but they are only activated when in the readers read range. An RFID reader, composed of a transceiver and an antenna, reads data from and writes data to tags, and transfers the collected data to a host computer for future retrieval and analysis. The antenna establishes the communication between the transponder and the transceiver, and its shape and dimensions determine the performance characteristics such as the frequency range [10]. Larger antenna loops tend to yield wider coverage areas, but the signal-to-noise ratio decreases at the same time; therefore a careful balance in reader design must be attained between the coverage area and reception reliability [11]. The frequency on which the RFID system operates is another important element, which determines the characteristics of the signals traveling between reader and tags. Available frequencies include low frequency (LF), high frequency (HF), and ultra-high frequency (UHF) [12]. Super-high frequency (SHF) or microwave is also used. Presently UHF is the most widely used, because UHF passive tags offer simple and inexpensive solutions, and most active tags operate on UHF. Characteristics of each frequency are summarized in Table 1. Choices of tags, readers, and frequencies and their combinations offer users exibility in building customized RFID systems to meet their requirements. Jaselskis and El-Misalami [14] provided detailed guidelines with 37 steps to help contractors and owners in the building industry determine the conguration of the RFID systems that best t their applications. As the RFID technology gains wider adoption and is used throughout the supply chain or across industries, the level of compatibility becomes an important issue, which has given birth to various worldwide technical standards. The International Organization for Standardization (ISO) has created standards for data structure (ISO 11784) and air protocol interface (ISO 11785) of tags used in tracking animals. It also has standardized the air interface protocol of tags used in proximity cards for access control and payment systems (ISO 14443) and in vicinity cards that can be read from a greater distance (ISO 15693). Also have been established are the standards for testing the conformance and performance of tags and readers (ISO 18047, ISO 18046). As the use of RFID technology for goods tracking becomes prevalent, ISO has also developed relevant standards to cover related air interface protocol (ISO 18000 series) [15].

Table 1 Characteristics of radio frequencies. LF Frequency range Read range Standards Metal/uid impact Data transfer rate Power and data transmission for passive tags Typical industry 125135 kHz <0.5 m (passive) ISO 11784/5, 14223, 180002 Very low Low Inductive couplinga Farming, security, brewery HF 13.56 MHz <1.0 m (passive) ISO 14443, 15693, 180003 Low Medium Inductive couplingb Pharmaceutical, health care UHF 400960 MHz <10 m (passive), >10 m (semi-passive and active) ISO 180006/7, EPCGen1 and 2 High High Propagation coupling Manufacturing, logistics, construction Microwave 2.455.8 GHz >100 m (active) ISO 180004/5 High High N/A Army, shipping, airlines

a Tag draws power from the electromagnetic eld that reader creates, and changes the electric load on the antenna, which inuences the electromagnetic eld and is sensed and analyzed by the reader. b Tag uses electromagnetic power that reader emits, changes the load on the antenna, and reects back an altered signal. Also called backscatter [13].

N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546


Auto-ID Center, an non-prot organization that focused on technologies for tracking goods globally, developed the Electronic Product Code (EPC) standard, which identied an object with a unique number consisting of a head, manage number, object class, and serial number. Auto-ID Center then applied EPC to RFID tags, and developed its own RFID air interface protocol standard, which included six classes of tags (class 05) covering passive, semi-passive, and active tags. The EPC technology was licensed to EPCglobal and made royalty-free to manufacturers and end users. Because of the incompatibility between class 0 and class 1 tags, which were both read-only passive tags but used different protocols, EPCglobal developed the second release in 2004, known as Gen 2, in a collaboration of leading RFID technology users and vendors. It has been agreed upon and widely conformed in a myriad of industries [15]. While Gen 1 standard was not aligned with ISO standards, Gen 2 was created to be compatible, and has been incorporated into ISO 180006 as an international standard.

3.2. Selection of ILS technology The authors have completed a thorough investigation of ILS technologies researched to date. The following paragraph provides a brief overview of the ILS technologies considered for the built environment. Various GPS-based solutions have been introduced to extend the use of GPS to indoor environments, such as high-sensitivity GPS that uses highly signal-sensitive receivers developed for weak-signal conditions [16], and Assisted GPS (A-GPS) that sends assisting information, such as satellite orbit information, to the receiver to speed its satellite acquisition time and improve its performance [17]. Inertial navigation systems (INS) enable a user carrying a package of motion and rotation sensors and a microprocessor to sense his/her own location independently of external infrastructure. It has two ways to reduce its large and time-proportional error. These are calibrating the system periodically with location-known devices such as RFID tags [18], and calculating a route by counting the users steps and measuring the step length, whose error is proportional to the distance of travel, and is thus smaller in indoor scenarios. Infrared-based solutions, such as Active Badge [19], use portable infrared beacons and xed infrared sensors to provide zone-level ILS service. Badges attached to targets transit signals through a wired network of sensors, providing information about targets locations to a central server. Ultrasound is used to facilitate some radio frequency (RF)-based solutions, such as Cricket Location Support System [20] and Active Bat location system [21]. They compare the time of arrival of ultrasonic signals with that of RF signals to determine the distance between the signal transmitter and the receiver. Such solutions can locate static or mobile targets accurately in a zone level or within centimeters. UWB-based solutions use very short pulses for communication between tags and receivers, and ensure a high accuracy by requiring deployment of large amount of infrastructure [22,23]. WLAN-based solutions, such as the RADAR system [24], can be easily set up and require few base stations. They use the RF signal strength, the magnitude of the electric eld, as an indicator of the distance between a transmitter and a receiver. This distance information is then used to obtain the users location by triangulation. RFID-based solutions require installation of RFID readers and tags around the building, and rely on the signals traveling between them to estimate the distance or nearness of the tags. A comparison of all ILS technologies based on the above criteria is presented in Table 2. Ultrasound is excluded from the table because it is always used as complement to RF-based solutions in the literature. Accuracy and affordability are associated with the specic test bed and devices used and therefore, vary signicantly among different solutions. Since the focus of this paper is on the built environment, accuracy, and affordability information is based on a representative solution from a recent publication by building industry researchers. If no solution has been published in the building industry, then a representative solution from the EE/CS eld is referenced. Accuracy is measured in terms of error distance, and affordability is measured in terms of cost per square meter. The cost information was obtained by either reviewing the publications, contacting the authors, or getting quotes from the vendors of the equipment referenced in the publications. Based on the analysis, RFID technology is chosen as the most suitable ILS technology for the built environment due to its proven capability of providing accurate and affordable ILS services. The achievable accuracy could meet most location sensing needs that may arise in the building industry. Typical cost of RFID technology is higher than that of WLAN, partly because the scale of deployment of the chosen solution (44.87 m2) is smaller than that of WLANs (157.3 m2), but it is still competitive against other technologies. In terms of accuracy and affordability, RFID is comparable

3. Research methodology The authors have adopted a three-stepped methodology in this study. First, the authors proposed a set of criteria for selecting the most suitable ILS technology for the built environment. Second, based on the criteria, the most suitable ILS technology is selected. Third, the selection of studies is carried on and two criteria are employed to evaluate them.

3.1. Technology evaluation criteria While various technologies have been tested to date for ILS, there is still no agreement with regard to which technology is the most suitable for the built environment. The authors proposed a set of criteria that a successful ILS technology should meet. These criteria are selected in such a way that (1) users need for indoor location information can be satised, (2) solutions built on the technology are implementable in complex built environment rather than merely ideal scenarios, and (3) the solutions are adaptive to various and changing environments and can provide continuous ILS services. The authors suggested that a successful ILS technology should be able to support ILS solutions (1) with adequate accuracy (within meters) as determined by particular use scenarios. This criterion has been extensively examined in previous research [4756] and is closely related to the needs of the building industry [59,60,62]; (2) that are affordable to the majority in the industry, so that the solution is economically competitive and can be widely implemented in practice; (3) that requires no need for line of sight, which accommodates the complexity of the built environment where various obstructions exist; (4) that has wireless communication capabilities, which allows the identication and data transfer to be done in a non-contact fashion so that the location of a target could be obtained by an observer without additional communication modules; (5) that is context-independent, which ensures that the accuracy is not sensitive to changes in the work environment and the solution can provide consistent services; (6) that has on-board memory capability, so that location and other context-specic information could be stored and accessed on site; (7) that has independent power supplies, so that the solution could provide continuous services in case of lost of external power supplies due to situations such as emergencies; (8) that builds on a technology which is widely used in the building industry. This would justify the applicability of the technology for complex and dynamic work environments, and provide the possibility of sharing equipment for multipurposes.


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546

Table 2 Comparison of ILS technologies based on the technology evaluation criteria. Technology Accuracy Affordability ($/m2) 380 20 17 140 3 25 Reference for accuracy and affordability a [25] [26] [27] [25] [28] [29] No line of sight required X p X p p p Wireless communication X X p p p p Contextindependence X p X X X X On-board data storage p p X X X p Built-in power supplies p p X X X X Wide application in the building industry p X X X p p

GPS variance INS Infrared UWB WLAN RFID


12 cm 1.104.15 m 3050 cm 650 cm 4.536.89 m 1.553.11 m

Refs. [25,28,29] are from the CE eld, [26,27] are from the EE/CS eld.

with INS, however, one major limitation with INS is the need for calibrating the system periodically and also the dependence of accuracy on the distance of travel as explained above. An important strength of RFID technology over WLAN, UWB and infrared is the on-board data storage capacity of RFID tags, which enables the tags to act as distributed databases and decentralizes the tagged objects information, making it available wherever the objects exist. The RFID technology can work in harsh environments such as construction sites with noise, contaminants, glare, and dirt [30]. Therefore, it is being widely used and has the potential to become pervasive [7] in the building industry for various purposes such as prefabrication control and management [9,31], supply chain management [12,30,32,33], locating components in lay down yards [3436] or underground [11], enhancing jobsite tool availability [37], preventing accidents [38], supporting assembly processes [39], improving quality inspection work [40], monitoring project progress [41], providing on-site FM personnel with access to real-time and up-to-date equipment information and optimized maintenance schedules [42], and ensuring complete, accurate, and available component information during the life cycle of a building [43]. In addition, unlike GPS and infrared, RFID technology does not require line of sight, which is essential for ILS as obstructions are present in built environments. It also supports wireless communication, which allows the identication and data transfer to be done in a non-contact fashion. On the other hand, RFID technology has some limitations, including its sensitivity to changes in the environment that may result in inconsistent performance. The proximity of metals and liquids can signicantly reduce the read range and lower the data transfer rate of the RFID system, due to their interference with the radio waves traveling between readers and tags. Encapsulation of tags helps ameliorate the performance reduction [43]. Another solution, which aims to align the tags antenna and a non-insulated metal material in such a way that the metal actually amplies the signal of the tag, is under investigation [8]. In fact, only INS technology is independent of environmental factors in the built environment, however, its error in accuracy is proportional to the distance of travel, which relies on external infrastructure for periodic calibration. In addition, the dependence of RFID readers on external power supplies may prevent RFID-based solutions from providing continuous services during emergencies. However, this limitation is also shared among the majority of ILS solutions included in the analysis. 3.3. Selection and evaluation of studies The authors searched electronic databases of scientic publication from mostly the past 5 years, including the Engineering Village, ScienceDirect, and ASCE Library databases. Using a variety of search algorithms and combinations of keywords, the authors sought publications related to indoor location sensing and either

performance criteria used to evaluate RFID-based ILS solutions: accuracy and affordability. In addition, the authors manually searched the following publications from the past 5 years for relevant articles: the Journal of Advanced Engineering Informatics, Journal of Computing in Civil Engineering, Automation in Construction, Journal of Construction Engineering and Management, Journal of Information Technology in Construction, and proceedings of the past 2 years of the International Conference on Computing in Civil and Building Engineering (ICCCBE) and the International Symposium on Automation and Robotics in Construction (ISARC). Finally, other articles as appropriate to provide background information on ILS solutions were reviewed. The authors identied and reviewed articles considered of primary relevance those focusing on RFID-based ILS solutions. Two performance criteria are used to synthesize and evaluate the reviewed solutions, namely accuracy and affordability. The former determines how well the location information addresses users needs. A satisfactory accuracy reects the technical viability of a solution and lays the basis of its overall effectiveness; the later has a crucial impact on the practical implementation of a solution. Prohibitively high cost prevents a solution from being practically used. Both criteria are frequently used in prior research, and can be quantied to support a justiable comparison. The rest of the criteria used to choose the most suitable ILS technology are not included in the analysis because they apply in the same way to all RFID-based solutions. Only peer-reviewed journal articles and articles published in proceedings of scientic conferences are included in the study. Twenty-one RFID based studies were reviewed and included to the analysis. 4. Review of RFID-based ILS solutions This section provides a comprehensive review of the work selected for the analysis, which ends with Table 3 that summarizes all selected solutions. RFID-based ILS solutions have been categorized into the following three areas based on their ILS methods and scale of deployment of their prototypes: triangulation-based ILS solutions, LANDMARC-based solutions, and zone or building level ILS solutions. 4.1. ILS methods Three major methods have been developed to locate a target in an indoor environment utilizing RFID technology: (1) Triangulation. The triangulation method is performed by lateration using distance measurements between reference positions. Theoretical or empirical models are used to map the collected signal strength or time of arrival to the distance between the reader and tag, and the distances are used for triangulation to obtain the 2D or 3D position of the tag.

N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546 Table 3 Summary of ILS solutions.a. Area covered Accuracy Cost (RFID equipment)b Reader/antenna N/A 4m5m Targets were located N/A within 3 m 3 m zones 2m 2 Tag


Other ndings/notes

Field of study EE/CS


N/A 16


[44] [47]

9/49 m2

0.41.0 m


8m6m 9.6 m 4.8 m (1) Semi-closed area; (2) Spacious closed area; (3) Typical university ofce with furniture 4 m 4 md

0.83 m 2.5 m (90% probability) 0.29 m (non-boundary), 1.5 m (boundary)

4 8 4

20 30 16

(1) The use of reference tags could offset dynamic EE/CS environmental inuence (2) More readers or reference tags increased accuracy (3) Problems included lack of direct signal strength report, long latency, and variation of tags (1) Deploying readers on each side of and with dif- EE/CS ferent distances to the sensing area led to higher accuracy (2) More readers led to higher accuracy (3) Lower computation load was achieved N/A EE/CS (1) RSI was more sensitive to reference tag density EE/CS than RSSI (1) Solution worked best in test bed (1) and worst in EE/CS test bed (3)


[55] [48] [1]

1.5 m


Test bed built in a university

No results were reported from the test 93% (area level localization) N/A



384 m2 including 4 rooms

15 m 10 m

10 m 10 m

4.8 m 4.8 m 5 m 5 md 5.63 m 11.7 m 4.58 m 9.6 m 10 10 10 md 5 m 6.5 md

0.1 m (noise-free), 0.2 m (with noise) 0.72 m 0.15 m 1.402.69 m 1.382.76 m 0.5 m 52.5% (area level localization) 1.2 m N/A 1.4 m (90% probability) 10.7 m (93% probability)

4 4 26 4 4 3 4

No reference tags used (1) Time to locate a lost item reduced from 11 to EE/CS No 1 min reference tags used 49 N/A EE/CS 9 36/20 20/48 12/10 11 16

(1) High accuracy and noise resistance were EE/CS achieved (2) Removal of abnormal data increased accuracy CE (1) UWB, WLAN and RFID were integrated for ILS (2) BIM provided information of particular locations (3) An integration platform managed all information (1) Accuracy for xed targets was lower than mov- EE/CS ing targets






5 m 10 m No tests 3 m 4md 115 m 75 m

A public hospital building

No quantitative results were reported 0.0524 m (error distance standard deviation) 0.53 m (error distance standard deviation)

4 N/A 4 Reader(s) was used in creating a signal strength ngerprint map N/A

16 N/A 16 8

(1) RPL2 and RPL3 had better performance in terms EE/CS of average and worst location errors (2) Signal strength preprocessing decreased the worst location error by 0.1 m N/A EE/CS (1) Efciency in area level localization had up to EE/CS 96.66% improvement over LANDMARC when the RSSI variance increased N/A EE/CS (1) The paper only discussed the methodology EE/CS N/A EE/CE (1) The solution was not meant for orientation rec- CE ognition and time invariance


[54] [53]

[56] [57] [58] [61]


1 m2d

4, 6, 8, 10

100 m2d

4, 6, 8, 10

No reference tags used No reference tags used

7 m 6.4 m (in building) 6.3 m 5.1 m (on construction site)

1.553.11 m 1.223.79 m

1 1

4 3/6

(1) Solution could identify and monitor a particular EE/CS target, detect violation and communication failures (1) Best practice: circular or regular polygonal dis- CE tribution of reference tags and even distribution of readers, or rectangular distribution of reference tags and symmetrical distribution of readers (2) Centralized target placement increased accuracy (3) A larger number or a more centralized placement of readers increased accuracy (1) Accuracy was higher in operating buildings than CE construction sites (2) MinMax algorithm had the best performance (3) ROCRSSI reduced implementation difculties (4) Scene analysis was sensitive to beacon positioning




Solutions are listed based on the year of publication. Typical cost of an RFID reader ranges between $600 and $1400. Typical cost of an RFID active tag ranges between $15 and $50. All eld tests used active tags. c Only the number of reference tags is included in the table. Number of tracking tags is determined by the number of targets to locate rather than the solution design and test bed. d Tests were done under simulation. Computer programs are used to simulate the propagation of signals, estimate the locations of targets, and calculate the error distances.


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546

(2) Proximity. The proximity method requires the measurement of the nearness of a set of neighboring points, which are close to the target and whose locations are xed and known. When an RFID tag or a reader is attached to a target, it continuously communicates with the readers or tags deployed in the environment through radio waves, the strength or time of arrival of which is observed and used to measure the nearness of these reference positions. The measured nearness, along with the corresponding known locations, is used to estimate the location of a target. (3) Scene analysis. The scene analysis method captures the signal strength in the sensing area and compares it with a preexisting signal strength database to map the target to its location. Pre-application mapping is required, during which the location sensing infrastructure, such as RFID readers or tags, is deployed in the sensing area, and the signal strengths of the reference positions are measured and recorded to establish a signal strength ngerprint map. During deployment, the signal strength of the radio waves that travel between the target and the infrastructure is measured and compared with the ngerprint map. The known locations of reference positions that best match the database are then used to estimate the location of a target. 4.2. Triangulation-based ILS solutions The rst reported prototype of RFID-based ILS solution is SpotOn [44]. SpotOn relies on the triangulation method for calculating locations of targets. This solution consists of RFID equipment, an aggregation algorithm, a server, and a client for visualization. Tags are attached to mobile targets; as a target moves, three or more readers deployed in the environment detect the target. Based on an empirical function, signal strength data is used to estimate the distance of the target to each reader. The location of the target is then triangulated. The visualization client then displays the results on screen with multiple functions such as pan, tilt, zoom, and 3-axis motion. In the test, targets were located within 3 m 3 m grids. The major limitation of the SpotOn solution is its low accuracy, mainly due to its reliance on the empirical function, whose accuracy is subject to the impact of multi-path effect, a phenomenon where signals reach the receivers by more than one path. This limitation has signicantly limited the application of triangulationbased solutions. However, a recent research project applied the triangulation method to a clean room in semiconductor production, where there were few partition walls and sponges were laid out on the walls to absorb magnetic wave reection [45]. This simple and clean environment maximized the effectiveness of the triangulation method and improved the accuracy. In addition, researchers established the signal strength-distance relationship through offline experiments instead of using an empirical function, which contributed to the solutions success. The solution was used to help locate misplaced cassettes in a clean room of 15 m 10 m, and was reported to have saved about 90% of reallocation time and machine idle time due to missing cassettes. Zhou and Shi [46] proposed another solution, whose algorithm is similar to triangulation in that both are based on a signal propagation model to get the distance between readers and tags using signal strength data. After obtaining the estimated distances between the tag attached to the target and a number of readers deployed in the sensing area, the solution estimates the targets location by minimizing the sum of error distances between the tag and all readers using a multilateration method. Based on this algorithm, the authors completed simulations to test the impact of various factors on the accuracy, including geometry of the sensing area, target distribution, error distribution of the signal propagation model, and reader placement. The results showed that: for

circular or regular polygonal geometries, error is minimized with readers evenly distributed along a concentric circle or polygon of the region boundary; for rectangular geometry, error is minimized with readers symmetrically placed with regard to the centerlines. The authors also reported that accuracy improves when placement of targets is more centralized, and that a larger number or a more centralized placement of readers increases accuracy. Luo et al. [29] added to the literature by testing the effectiveness of four different algorithms that are based on the received signal strength indication (RSSI), a measurement of signal strength at the receivers location. These are MinMax, Maximum Likelihood, ROCRSSI, and scene analysis. The rst three algorithms are variances of triangulation. In MinMax, the boundaries around beacons for triangulating a target are squares instead of circles, whose side lengths are twice of the estimated distances between the beacons and a target using a path-loss model. The Maximum Likelihood algorithm maximizes the probability of a targets location by minimizing the variance of estimated error, given locations of the beacons and their estimated distances to the target. ROCRSSI draws two circles around each beacon, whose radiuses are determined by RSSI values, and assumes a target is located between the circles. This lays the basis of triangulation. Field tests were done in a 7 m 6.4 m test bed in a building, and a 6.3 m 5.1 m test bed on a construction site. Findings include: solutions work better in operating buildings than on construction sites; MinMax has better performance than other algorithms with an average error of 1.2 m; ROCRSSI can avoid implementation difculties of other algorithms; scene analysis may be particularly sensitive to beacon positioning. 4.3. LANDMARC-based solutions LANDMARC [47], a proximity-based solution as shown in Fig. 2, has been a foundation for many RFID-based solutions due to the advantages of the proximity method over triangulation in adapting to complex and dynamic built environments. Unlike SpotOn, LANDMARC uses active tags in two different ways: tracking tags, which are attached to the targets so that locating the targets equals locating the tracking tags; and reference tags, which are xed in the sensing area in known locations, so that they can provide reference locations. As the target moves within an area, the tracking tag communicates with readers/antennae, while the solution measures and records RSSI values. All the reference tags do likewise. Readers then transfer the collected data to a computer, which identies the nearest reference tags to the target, and measures their nearness. In addition, with tag IDs, the known locations of the identied nearest neighboring reference tags are extracted from the database. The neighbors locations and nearness to the target are then used in computing the targets location. This algorithm is called the k nearest neighbor (KNN) algorithm. Field tests done within an area of 4 m 10 m reported an accuracy of within 1 m with 50% probability, and within 2 m with 90% probability. According to the researchers, the optimal number of nearest neighbors used in computation was four, and that the solution could offset the inuence of dynamic environmental factors. They also found that using more readers or reference tags would yield better results. They also noticed several limitations, including the low accuracy in the signal strength report, high latency or time delay, and variation in tags behavior. The satisfactory performance and potential for improvement of LANDMARC have drawn considerable attention in the research community. A number of research projects followed, which are grouped below according to the aspects they aimed to improve upon. 4.3.1. Improvements to accuracy To improve accuracy, Polito et al. [48] used an algorithm called RSSI spatial interpolation (RSI), which interpolated the distance

N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546


vector obtained in the RSSI algorithm to build a denser two-dimensional map to estimate the nearness of reference tags. In indoor tests done in an area of 9.6 m 4.8 m, an accuracy of 2.5 m with 90% probability was reported for both RSSI and RSI algorithms. When the spacing between reference tags increased from 60 to 180 cm, the accuracy of RSSI was not affected much, while that of RSI increased signicantly, with maximum error decreasing from 6 to 2.5 m. The authors listed the following as limitations: the need for a large number of readers and tags, problems of signal propagation in the sensing area, and the difculty in predicting best operational parameters. In their VIRE solution, Zhao et al. [1] used imaginary reference tags, referred to as virtual tags, to achieve a higher accuracy. These tags are virtually distributed linearly between real reference tags, increasing the density of the reference tag grids. Locations of the virtual reference tags are known and recorded, and their signal strength is estimated by linear interpolation of that of the real tags next to them. The sensing area is divided into small regions and each reader maintains its own proximity map, so that the most probable location can be estimated. Field tests were done with four readers and 16 reference tags in three distinct test beds: (1) a semiclosed area that is not surrounded by concrete walls and furniture, (2) a spacious closed area that does not have many metallic objects, and (3) a typical university ofce with many desks and chairs. The overall results showed an average estimation error of less than 1.5 m for locations at the boundary of the sensing area and 0.29 m for other locations. The difference was due to the fact that targets in the boundary region were not completely covered by neighboring reference tags. Also reported was a reduction of error ranging from 17% to 73% over LANDMARC. The solution worked best in test bed (1) and worst in test bed (3). Future research plans included building a large-scale deployment to study the effects of different grid spacing distances, the size of boundary regions that should be avoided, and the number and placement of readers. Zhang et al. [49] attempted to improve the accuracy by eliminating the impact of the diversity of reference tags that might result from different tag types or used-time of built-in batteries, which would result in dissimilarity of tag performances and therefore lower accuracy. Zhang et al. used the RFIDiffFreeLoc algorithm to offset this impact, applying diversity elimination to all reference tags RSSI using the tracking tag as a criterion. The proposed algorithm was tested by simulating four readers in corners of a noisefree area of 10 m 10 m and 49 reference tags in grids, which revealed noticeable improvement of accuracy over LANDMARC, with error distance reduced from 0.45 to 0.1 m with 50% probability.

However, the improvement was less signicant in the simulated environment with noise, or in practical experiments that were done with four readers and nine reference tags in a room of 4.9 m 4.9 m. Hsu et al. [50] tried to improve the accuracy by introducing a preprocessing phase to the location calculation. The preprocessing phase uses a moving average lter to smooth the RSSI values for each tag; the latest RSSI values are stored in a buffer for dynamic average. These values are updated every time a new reading is added to the buffer replacing a previous one. The underlying rationale is that the environmental changes in the sensing area may cause a single RSSI value to vary signicantly, and using a dynamic average could minimize this impact. The largest error of estimated locations of all targets in the eld test was reduced by 0.1 m over LANDMARC. The authors also cross-tested two reference tag layouts (RTLs) and four reader position layouts (RPLs), and reported that RPL2 (four readers in corners of the sensing area) and RPL3 (two of the four readers moved outside the boundary of the sensing area) had better performance in terms of average and worst location errors for both RTLs. For RTL1 (36 reference tags placed in grids), the best average location result was obtained using RPL3, while RPL4 (all four readers in corners outside the sensing area) had better performance in terms of the worst location error. For RTL2 (20 reference tags placed along the perimeter of the sensing area), RPL4 performed best in terms of both location errors. 4.3.2. Extension of the RSSI algorithm capabilities Sue et al. [51] introduced improvements in two ways: by reducing the computation load, and by offering users the exibility of determining the level of detail of the location information. To develop their solution, named FLEXOR, a special reference tag layout was used with boundary tags deployed in hexagons and one cell tag in the center of each hexagon. Based on this layout, a region localization mode was built, which locates a target within a hexagon area whose cell tag is closest to the target according to the RSSI values. If the targets coordinates are further requested, FLEXOR can turn to a coordinate mode, nding the two boundary tags in that hexagon which are closest to the target, and estimate the targets coordinates using their known coordinates as well as those of the cell tag. Simulations included 16 reference tags in 9 m2 regions and 64 reference tags in 49 m2 regions. Results showed that this two-mode design added to the exibility in location sensing, and reduced the computation load and time. Best accuracy was achieved with shorter reader power level intervals and enough read range, more readers, and each of them deployed on each side of and with different distances to the sensing area. Huang et al. [52] contributed to this discussion by proposing an improved Bayesian-based algorithm. The general Bayesian-based algorithm uses a certain probability model to estimate the most likely position of the target. Its accuracy could be very high but decreases fast when the measurement error is increased. On the other hand, the error of various LANDMARC-based solutions is generally limited to an acceptable range. Huang et al. made their contribution by combining these two algorithms. The algorithm has an additional function to remove abnormal RSSI readings that might be caused by environmental changes. Simulation results showed that the combined algorithm yielded better performances than LANDMARC or the general Bayesian-based algorithm, with the 60th-percentile error distance decreased from around 1 to 0.5 m. Silva and Goncalves [53] proposed another solution, called LANDMARC+, to solve the following problem: if the LANDMARC solution reports an inaccurate location and the user cannot nd the target in the suggested room, then the user has no information on which room to search next and thus faces a blind search situation. LANDMARC+ addresses this problem by adding a secondary-possible room to the original LANDMARC results. Assuming

Fig. 2. Schematic components of LANDMARC.


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546

that each room has the same number of readers, it calculates the mean Euclidean distance of all tags in each room, and reports the room with the smallest distance to the user as the secondary-possible room. Validation was presented by a simulation that was based on a signal propagation model with the inuence of walls and oors. Two test scenarios were used, an open-area suburban ofce and a suburban ofce with soft partitions. Results showed an improvement of up to 96.66% over original LANDMARC in terms of reporting the correct room location. Although previous solutions locate targets in a 2D format, it can be easily extended to estimate 3D locations by incorporating a z coordinate into the RSSI algorithm. Khan and Antiwal [54] addressed this issue in their research, and validated the modied algorithm through a simulation in which three readers, 11 reference tags, and two tracking tags were deployed in a 10 m 10 m 10 m space. The reported average error distance was 0.5 m. 4.3.3. Calibrating the error value Two research projects focused on calibrating the reported locations. Jin et al. [55] proposed a more efcient, effective and accurate solution that supplements the proximity method with triangulation method. After identifying the k nearest neighbors based on signal strength, this solution estimates their locations with the triangulation method. Since the actual locations of these neighbors are known, the estimated errors can be calculated. The mean error is then added to the estimated location of the target, which is calculated using the proximity method. A test was done in a room of 6 m 8 m, with four readers and 20 active reference tags. Ten tracking tags were located 100 times each. The reported maximum error in the tests was about 0.7 m, which was a remarkable improvement. Wang et al.s approach [56] is slightly different. After the location of a tracking tag is estimated with the KNN algorithm, the tracking tag and its k nearest neighbors are put in a set. Then, for every reference tag in the set, its location is estimated via the known locations of other reference tags and the estimated location of the tracking tag. The result is compared with its actual location, and the error is recorded. Then the error for all reference tags in the set is averaged and added to the estimated location of the tracking tag for a nal estimation. Tests were done in a room of 5 m 10 m with four readers and 16 active tags, which reported an accuracy within 1 m over 90% of the time, an improvement by around 50% over LANDMARC. 4.3.4. Improvement to the reference tags The use of reference tags distinguishes the proximity method from triangulation and scene analysis methods, and is a key factor in the design of LANDMARC. Two issues with reference tags have aroused the interest of researchers. First, the success of LANDMARC-based solutions relies on using large quantities of reference tags, which increases the computation load and latency. A large density of reference tags also requires more investment in the hardware and causes difculties for real-world deployments. Therefore, reducing the density of reference tags is of signicant value. The second issue has to do with the placement of reference tags in deployment. In most of the previous work discussed above, reference tags are deployed in rectangular mesh by default, without any justication. Li et al. [57] addressed the rst issue by introducing a key reference tag concept. To use fewer reference tags while retaining accuracy, their solution works as follows: In an ofine training phase, the same quantity of reference tags as in LANDMARC is deployed, and their RSSI values are all recorded in the form of tuples. This sampling process can be done under various environments to gain more tuples. Based on a divide-and-conquer technique, a sub-

set of tags can be found whose RSSI values are sufcient to identify the RSSI values of all other reference tags using the recorded tuples. Tags in this subset are called key reference tags. Once this process is completed and key reference tags are found, all other tags are no longer needed, while the location sensing accuracy is retained. Huang et al. [58] challenged the default rectangular deployment by arguing that a different deployment formation may achieve better performance. Through geometrical deduction, they demonstrated that a fourth nearest neighbor was the main source of error; therefore only three nearest neighbors should be used for location computation. On this basis, they proposed to deploy the reference tags in triangular mesh, and justied this idea through simulation, with the 90th-percentile error distance decreased from about 1.6 to 1.4 m.

4.4. Zone or building level ILS solutions This section summarizes further recent efforts that have experimented with other location sensing methods, used RFID technology together with other technologies for more integrated ILS solutions, and tested in zone or building-scale deployments. Zhen et al. [59] proposed a solution to estimate the region where the target was located. They tried to mitigate the problems of using radio frequency in location sensing, including multiplepath effect, dynamic environment, and unreliable communication, by developing a supporting vector machine (SVM)-aided algorithm, which follows similar rules to the round-robin competition used in sports games: each region is compared with each of the others, using the RSSI data to determine in which region the target is located. Historical data and geometric relationships among rooms are used to further improve the robustness and accuracy of the solution. To locate a tagged user with this solution, all rooms are numbered in advance. Then as a target moves around, multiple readers pick up the tag, and the location is estimated with the round-robin comparison algorithm. In the eld tests, which were done in four rooms with a total area of 382 m2, 240 samples were collected, and the accuracy was reported to be 93% with a standard deviation of 1.5%. The researchers also reported that the delay due to hardware and software was 56 and 23 s, respectively. According to the researchers, deployment at large scale could be achieved by reducing the number of readers required and grouping the readers and running the solution in parallel for different subregions. Rueppel and Stuebbe [60] published an indoor navigation solution to support rescuers in emergency or daily maintenance work. For ILS purposes, the solution utilizes three types of infrastructure depending on the built environment. UWB is used in halls, as it is less inuenced by metals and high humidity and requires only a few sensors; WLAN is used in an ofce area to make use of the existing WLAN network. Finally, RFID technology is used in rooms that have little technical building infrastructure, such as cellar rooms or underground garages. Active tags are preferred. The solution also integrates building information modeling (BIM) technology, which models the geometry of the building and stores related data, to provide rescuers with the building information of particular locations. A key component of the solution is the navigation integration platform (NIP), which administrates all the information including the actual positions of the rescuers, the organizational structure, the building information, generated routing networks for navigation, and information sent to the rescuers. When personnel are out of ofce to perform daily maintenance work or deal with emergencies such as res, their positions are monitored by the NIP in real time, from which they can also get support, such as suggested routes, or information on the building components within their spatial context. When their results were

N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546


published, the solution was being evaluated in a university test bed, and was planned to be implemented at the Frankfurt airport. Pradhan et al. [61] published the rst RFID-based ILS solution to be implemented in the CE eld. Instead of using the proximity method, this solution uses the scene analysis method for location sensing, which requires that a signal strength ngerprint map be created before the solutions application. As a user carrying an RFID reader walks in the sensing area, the reader collects the signal strength of the tags that are deployed in the environment and that are within the readers read range, which is then compared with the reference points in the ngerprint map to determine the users current location. A set of ten experiments were done on a whole oor of a typical campus building (115 m 76 m), with the reported accuracy ranging from 10.7 m with 93% probability and to 1.5 m with 40% probability to 10.7 m with 80% probability and 1.5 m with 20% probability. Taneja et al. [62] repeated the same ILS tests at the same space 4 years later and reported a reduced accuracy. The authors concluded that reduced accuracy may be due to the tags age, which resulted in a shorter read range. The authors also examined whether the solution could provide the orientation of the targets and monitored the time invariance performance in the tests, but reported negative results. Low accuracy and high sensitivity to environmental changes are found to be the two major limitations of this solution. Silva et al. [63] proposed a comprehensive ILS solution by adding middleware, database, and web-based applications to the ILS solution, and by conducting a large-scale deployment. Their solution, named TraceMe, consists of RFID tags and readers, a server, middleware, a database, a website, and a host computer. The ILS is accomplished by detecting tracking tags within xed readers read ranges, and triangulating the tags locations based on RSSI values. Any detection of a tag is reported by the reader to the server, which searches the database for information about the reader, tag, and room, and makes it accessible to users through the website. Tests inside a hospital building showed the solutions ability to detect tag violation, identify, locate and monitor a particular target, and detect communication failures.

to tag positions when compared with triangulation-based algorithms [29], which reduces its reliability. On the other hand, the proximity method can accommodate the impacts of the complexity and dynamic nature of built environments. It does not use the RSSI value to estimate the travel distance of the signal. Instead, it uses RSSI value to compare the nearness of reference tags to the target, which is qualitative and less sensitive to the obstructions that impact the propagation of the signal. Moreover, tracking tags and reference tags are exposed to the same environments, so the impacts of environmental changes exerted on them are the same and can be offset. Though the proximity method also has its limitations. It requires a large number of reference tags, which inevitably adds to the cost and lowers the ease of use. The need to deploy these reference tags in the environment ahead of time is another downside. In this, it is similar to the scene analysis method, requiring a time- and labor-consuming pre-application phase.

5.2. The criteria, used by prior research in solution design and evaluation, are not persuasive Accuracy and affordability are the two criteria used most frequently by prior research in designing and evaluating the solutions. However, there are other criteria that need to be considered in order for the solutions to address various user demands, and meet with real-world challenges. For example, robustness measures the capability of the solution to survive unexpected malfunctions and damages. Given the complexity and dynamic nature of built environments that may signicantly impact the effectiveness of the location algorithm and performance of the sensing infrastructure, it is necessary to evaluate the solutions ability to provide a consistent and steady service. Besides, during emergencies, location information may be needed for evacuation and rescuing; therefore a successful solution is expected to work in harsh environments even with a loss of part of the sensing infrastructure. The RFID technology may not be the best choice in such situations where the tags themselves may be dislocated or destroyed, and a solution should be designed to minimize the chances of occurrence of such dislocation or destruction. Another important criterion is the ease of use, which measures the effort required for users to deploy and maintain the solution. Installation and maintenance of the solution not only accounts for part of the investment but also requires certain expertise, which affects how users can efciently employ the solution. Besides the ease of deployment, it is also necessary to improve the way the solution is designed to operate, such as having users carry a tag instead of a reader, incorporating a graphical user interface, or extending wireless communication to enable delivery of location sensing results to both on-site and online users. Likewise, other criteria, such as scalability, reliability, interoperability, latency, and generality also should be dened and considered in designing and evaluating various ILS solutions. Nevertheless, the unbalanced weight of each criterion and the difculty of dealing with the tradeoffs between them have resulted in no single solution satisfying all criteria mentioned above. Solutions such as the one proposed by Li et al. [57] require less infrastructure and are thus cost efcient, but they are sensitive to the intactness of the infrastructure; solutions such as LANDMARC [47] are robust, and would continue to work even with some damage to the infrastructure, but they require deployment of a large number of readers or tags, which renders them inconvenient to practical applications in built environments; solutions such as the one proposed by Pradhan et al. [61] require less effort to deploy the infrastructure, but the accuracy might change with changes in the environment.

5. Discussion 5.1. The proximity method is the most frequently used and suitable method for complex and dynamic built environments A noticeable phenomenon in the previous RFID-based ILS research is the concentration on the proximity method, which was used by 13 solutions out of the total 21 solutions reviewed. The triangulation method was the rst method tested but was rarely used afterwards. The scene analysis method was not used until recently, although it was used in combination with other technologies such as WLAN for ILS before [24]. While researchers did not provide any explicit explanation for their chosen method, the dominance of the proximity method has most likely resulted from the complexity and dynamic nature of the built environment with obstructions, multiple materials, ambient changes, occupancy, and so on. The effectiveness of the triangulation method inherently decreases because it depends largely on an accurate signal propagation model. In complex and dynamic environments, the signal is inuenced by the multi-path effect and the existence of liquids and metals, so it is highly unlikely to accurately estimate the travel distance of the signal using its RSSI value. The scene analysis method requires a time-consuming mapping of a ngerprint, which is used repeatedly afterwards. However, due to the continuous environmental changes and also the tag battery life span, the ngerprint has to be updated periodically, which requires considerable time and labor. Recent research also reported the sensitivity of scene analysis


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546

5.3. Adaptability of these solutions within the built environment is uncertain Some of the research projects veried the proposed solutions with simulations, which allow easy manipulation of parameters such as dimensions of sensing area, positions of readers and tags, and noise level. However, simulation relies heavily on a signal propagation model, whose accuracy is unknown. Accordingly, the simulation test results are not a reliable justication of the solutions applicability for practical use. Moreover, although a large portion of the reviewed solutions were validated with eld tests, in which RFID readers and tags were deployed and eld data were collected, a close examination of the test setup of these solutions reveals that the majority of the test scenarios were as simple as a single empty room. Such simplication makes it questionable that the solutions would yield the same results in real-world applications. Because most of the previous research was done in the EE and CS elds, efforts understandably focused more on technical improvement than on adaptability and applications. However, from the building industrys point of view, tests in more complex and dynamic environments are crucial to proving the value of these solutions and transforming them into day-to-day applications. In addition, the return of investment of these solutions must be assessed. In the long run, this can be done through case studies of actual implementations; in the short run, a Markov chain model similar to the one proposed in [64] could be helpful.

6. Recommendations for future research 6.1. ILS solution design should be further modied to achieve better performance Previous work has focused on testing different location sensing methods and improving the algorithms. More work could be done in the following ways to improve current solutions: First, different algorithms can be combined to utilize each algorithms advantages. Several research projects have demonstrated the value of integrated algorithms. For example, triangulation is not reliable due to its dependence on a signal propagation model, but it could be used to calibrate the results of proximity-based methods [55]; the general Bayesian algorithm is sensitive to the accuracy of signal measurement, but it can be incorporated into proximity-based methods to achieve high and stable accuracy [52]. Other potential algorithm integrations might be the following: scene analysis could be used to establish an empirical signal propagation model, which could be used for simulation, or get updated periodically by a solution that uses the proximity method with reference tags; triangular tag deployment could be combined with the cell reference tag concept to see if the computation load could be further reduced. Some other algorithms, such as the one that uses read time of RFID system to obtain orientation and location information [65], could also be investigated and integrated. Second, previous work mainly aimed to achieve high accuracy, but did not pay detailed attention to the other criteria. More work is needed to develop a solution that satises the other criteria outlined in the discussion section. For example, the dominant LANDMARC solutions require a large number of reference tags, which impedes their practical application. A tradeoff between accuracy and ease of use should also be leveraged. Tag-based solutions that attach tags instead of readers to the targets are generally easier to use; proximity-based solutions are more accurate and robust; scene analysis-based solutions do not require much infrastructure and are more affordable and scalable. How to leverage the advantages of different types of ILS solutions requires extensive research.

Third, the ILS solutions should be further developed to better adapt to built environments. For this, the complexity and dynamic nature of the built environments should be taken into consideration when the solution is designed, which might lead to valuable features such as rugged equipment, less infrastructure, and an easy-to-maintain deployment plan. In addition, the prototypes need to be tested in environments closer to the real world rather than in a single unoccupied space, so that impacts of the obstructions and environmental changes could be studied and mitigated. Contrast tests could also be done in various environments to understand the impacts of environmental factors, such as building types, temperature and humidity, and level of activity, on the performance of the ILS solutions. These test results would effectively guide users to apply the solutions in various environments to meet their practical needs. It is important to note that the proposed improvement of solution design will benet from closer collaboration between researchers in the CE and EE/CS elds. On one hand, CE researchers have better insights into the potential applications of indoor location information in the building industry, including HVAC equipment localization, onsite FM personnel monitoring, tool localization and optimal allocation, asset management, building security control, occupancy-driven building energy conservation, emergency response, and so on. Moreover, CE researchers have a deeper understanding of challenges in implementing ILS solutions in built environments such as dynamic occupancy ows and prevalent metal obstructions, therefore, they can revise the solutions accordingly, making them more adaptable to the built environment. On the other hand, EE/CS researchers have the expertise to contribute to improvement of location sensing algorithms, assembly and calibration of hardware, development of supporting software, and processing and distribution of data. In short, effective collaboration between researchers in the CE and EE/CS elds will further advance the research in ILS area. 6.2. ILS could be integrated with OLS to achieve seamless location sensing solutions Technologies used for OLS and ILS are usually different; so are the associated algorithms and infrastructure. This may cause a gap in location sensing services when users move from indoors to outdoors or vice versa. Therefore, there is a need for integrated OLS and ILS solutions to ensure continuity of location sensing and provision of location-based services, such as information-rich surrounding sensing, driver and passenger navigation, location-based gaming, and assistance to the disabled people [66]. Attempts were made to integrate location-sensing technologies such as GPS and UWB [67], GPS and indoor GPS [68], or GPS, sensors, and WLAN [69]. Prior research has also discussed the integration of the RFID technology and GPS for applications such as navigation for the impaired [70] and human resources tracking [71]. However, to the authors best knowledge, no prototype has been built and tested that can integrate the RFID technology with GPS in such a way that accuracy of both indoor and outdoor locations is as high as that achieved in individual RFID and GPS-based solutions, that location information is exchangeable between technologies, and that verication of the solution is done under real-world built environments. To fulll these goals, the following research questions need to be answered: (1) Whats the technology to be used for OLS; (2) What OLS and ILS infrastructure is needed, and how should it be deployed; (3) What the data format is used by ILS and OLS infrastructure, and how the data is fused; (4) What the communication protocol is if wireless data transfer is required; (5) What the location sensing algorithm should be for both OLS and ILS; (6) How to dene the interface between indoor and outdoor environments; (7) How to present the sensed locations, whether by a host computer

N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546


or mobile devices; (8) What the tradeoffs are between the criteria of accuracy, affordability, and other potential criteria listed in the discussion section in the design and deployment of a solution; (9) If the solution is interoperable with technologies that can accommodate semantic data in addition to the location information, such as CAD and GIS [72], and if the solution is extendable to incorporate additional location-based services, such as navigation? 6.3. ILS solutions could be extended to develop a context-aware information delivery mechanism to be used in the building industry Navigation information, as demonstrated in Rueppel and Stuebbes research [60], is an example of important information that can be delivered. Also to be delivered is a wide range of information that forms the basis of context awareness in construction and FM, leading to improvement in project safety, schedule, cost [73], and decision-making [25]. Furthermore, integrated with OLS solutions, seamless information delivery could be achieved. Such an information delivery mechanism should include four basic components, each of which requires future research efforts: an ILS component, which identies the users context; an information management component, which stores and updates the information to be delivered; an information lter component, which selects the information based on user preference and context; and an information transfer component, which communicates with the user to send the information. Future research should focus on applying this information delivery mechanism to multiple areas including (1) execution and management of construction activities, e.g. assembly instructions are delivered to the onsite crew; (2) safety and security, e.g. locations of onsite workers are monitored in real time to avoid invasions or collisions, or alerts are sent when assets are removed without authorization; (3) FM, e.g. an onsite worker is located and his/her context analyzed, so that information such as maintenance history, work orders, or inspection records is delivered to him/her to facilitate the maintenance work; and (4) emergency response, e.g. rescuers are guided to the shortest route inside a building. 7. Conclusion ILS is important to the building industry because it improves the utilization and maintenance of facilities and lays the basis of context awareness within built environments. The objective of this paper was to review the academic accomplishments of RFID-based ILS solutions, and provide directions for future research. Twentyone research projects were presented, with the algorithm design, devices, test setup, and performance evaluation of each solution described in detail. It was concluded that the proximity method is the most frequently used and suitable ILS method for the built environments, that no single solution meets all the criteria for successful ILS solutions, and that the adaptability of these ILS solutions within the built environments is uncertain. Modifying and combining different algorithms, studying the tradeoffs between criteria for success, and improving the adaptability to complex and dynamic built environments, can further improve ILS solution design. In addition, a context-aware information delivery mechanism should be built to deliver value to the building industry through improvements to execution and management of construction-related activities, safety and security, FM, and emergency response. References
[1] Y. Zhao, Y. Liu, L.M. Ni, VIRE: active RFID-based localization using virtual reference elimination, in: Proceedings of the 2007 International Conference on Parallel Processing, IEEE, Piscataway, NJ, USA, 2007.

[2] S. Chumkamon, P. Tuvaphanthaphiphat, P. Keeratiwintakorn, A blind navigation system using RFID for indoor environments, in: Proceedings of ECTI-CON 2008, Inst. of Elec. and Elec. Eng. Computer Society, Krabi, Thailand, 2008, pp. 765768. [3] J. Nord, K. Synnes, P. Parnes, An architecture for location aware applications, in: Proceedings of the 35th Annual Hawaii International Conference on System Sciences, IEEE Comput. Soc, Los Alamitos, CA, USA, 2002, pp. 38053810. [4] K. Eustice, V. Ramakrishna, N. Nguyen, P. Reiher, The smart party: a personalized location-aware multimedia experience, in: Proceedings of the 5th IEEE Consumer Communications and Networking Conference, Institute of Electrical and Electronics Engineers Computer Society, Las Vegas, NV, United states, 2008, pp. 873877. [5] Z. Aziz, C.J. Anumba, D. Ruikar, P.M. Carrillo, N.M. Bouchlaghem, Context aware information delivery for on-site construction operations, in: Proceedings of the 22nd CIB-W78 Conference on Information Technology in Construction, 2005, pp. 321327. [6] E. Ergen, B. Akinci, An overview of approaches for utilizing RFID in construction industry, RFID Eurasia, 2007 1st Annual, 2007, pp. 15. [7] K. Domdouzis, B. Kumar, C. Anumba, Radio-frequency identication (RFID) applications: a brief introduction, Adv. Eng. Inform. 21 (2007) 350355. [8] R. Wing, RFID applications in construction and Facilities Management, ITcon. 11, Special Issue IT in Facility Management, 2006, pp. 711721. [9] E. Ergen, B. Akinci, R. Sacks, Tracking and locating components in a precast storage yard utilizing radio frequency identication technology and GPS, Autom. Constr. 16 (2007) 354367. [10] T. Hassan, S. Chatterjee, A taxonomy for RFID, in: Proceedings of the 39th Annual Hawaii International Conference on System Sciences (HICSS06), Institute of Electrical and Electronics Engineers Computer Society, Kauai, HI, United states, 2006, p. 184b. [11] K. Dziadak, B. Kumar, J. Sommerville, Model for the 3D location of buried assets based on RFID technology, J. Comp. Civil Eng. 23 (2009) 148159. [12] L. Wang, Y. Lin, P.H. Lin, Dynamic mobile RFID-based supply chain control and management system in construction, Adv. Eng. Inform. 21 (2007) 377390. [13] ERA BUILD, Review of the current state of Radio Frequency Identication (RFID) technology, its use and potential future use in construction, 2006, pp. 1100. [14] E.J. Jaselskis, T. El-Misalami, Implementing radio frequency identication in the construction process, J. Constr. Eng. Manag. 129 (2003) 680688. [15] RFID Journal, A summary of RFID standards. http://www.rdjournal.com/ article/view/1335/1, 2005 (accessed May 20, 2010). [16] S. Schon, O. Bielenberg, On the capability of high sensitivity GPS for precise indoor positioning, in: Proceedings of the 5th Workshop on Positioning, Navigation and Communication 2008 (WPNC08), Inst. of Elec. and Elec. Eng. Computer Society, Hannover, Germany, 2008, pp. 121127. [17] F.V. Diggelen, Indoor GPS theory and implementation, in: Proceedings of the IEEE Position Location and Navigation Symposium, IEEE, Piscataway, NJ, USA, 2002, pp. 240247. [18] J.R. Guerrieri, M.H. Francis, P.F. Wilson, T. Kos, L.E. Miller, N.P. Bryner, et al. RFID-assisted indoor localization and communication for rst responders, in: Proceedings of the 2006 First European Conference on Antennas and Propagation Conference, IEEE, Piscataway, NJ, USA, 2008. [19] R. Want, A. Hopper, V. Falcao, J. Gibbons, The active badge location system, ACM Trans. Info. Syst. 10 (1992) 91102. [20] N.B. Priyantha, A. Chakraborty, H. Balakrishnan, The Cricket location-support system, in: Proceedings of the 6th Annual International Conference on Mobile Computing and Networking, ACM, New York, NY, USA, 2000, pp. 3243. [21] A. Harter, A. Hopper, P. Steggles, A. Ward, P. Webster, The anatomy of a context-aware application, in: Proceedings of the 5th Annual Joint ACM/IEEE International Conference on Mobile Computing and Networking (MOBICOM99), ACM, New York, NY, USA, 1999, pp. 5968. [22] B. Becker, Application range of UWB-technologies, in: Proceedings of the SMART Solutions 2008, Isernhagen, Germany, 2008. [23] Y.K. Cho, H.Y. Jong, D. Martinez, Error modeling for an untethered ultrawideband system for construction indoor asset tracking, Autom. Constr. 19 (2010) 4354. [24] P. Bahl, V.N. Padmanabhan, RADAR: an in-building RF-based user location and tracking system, in: Proceedings of the 9th Annual Joint Conference of the IEEE Computer and Communications Societies, IEEE, Piscataway, NJ, USA, 2000, pp. 775784. [25] H.M. Khoury, V.R. Kamat, Evaluation of position tracking technologies for user localization in indoor construction environments, Autom. Constr. 18 (2009) 444457. [26] A.R. Jimenez, F. Seco, C. Prieto, J. Guevara, A comparison of pedestrian deadreckoning algorithms using a low-cost MEMS IMU, in: Proceedings of the 6th IEEE International Symposium on Intelligent Signal Processing (WISP 2009), IEEE, Piscataway, NJ, USA, 2009, pp. 3742. [27] S. Lee, K.N. Ha, K.C. Lee, A pyroelectric infrared sensor-based indoor locationaware system for the smart home, IEEE Trans. Consum. Electron. 52 (2006) 13111317. [28] S. Woo, S. Jeong, E. Mok, L. Xia, C. Choi, M. Pyeon, et al., Application of WiFibased indoor positioning system for labor tracking at construction sites: a case study in Guangzhou MTR, Autom. Constr. 20 (2011) 313. [29] X. Luo, W.J. OBrien, C.L. Julien, Comparative evaluation of Received SignalStrength Index (RSSI) based indoor localization techniques for construction jobsites, Adv. Eng. Inform., in press. [30] J. Song, C.T. Haas, C. Caldas, E. Ergen, B. Akinci, Automating the task of tracking the delivery and receipt of fabricated pipe spools in industrial projects, Autom. Constr. 15 (2006) 166177.


N. Li, B. Becerik-Gerber / Advanced Engineering Informatics 25 (2011) 535546 [54] M.A. Khan, V.K. Antiwal, Location estimation technique using extended 3-D LANDMARC algorithm for passive RFID tag, in: Proceedings of the 2009 IEEE International Advance Computing Conference (IACC 2009), IEEE, Piscataway, NJ, USA, 2009, pp. 249253. [55] G. Jin, X. Lu, M. Park, An indoor localization mechanism using active RFID tag, in: Proceedings of the IEEE International Conference on Sensor Networks, Ubiquitous, and Trustworthy Computing, Institute of Electrical and Electronics Engineers Computer Society, 2006 II Taichung, Taiwan, 2006, pp. 4043. [56] X. Wang, X. Jiang, Y. Liu, An enhanced approach of indoor location sensing using active RFID, in: Proceedings of the 2009 WASE International Conference on Information Engineering (ICIE), IEEE, Piscataway, NJ, USA, 2009, pp. 169172. [57] W. Li, J. Wu, D. Wang, A novel indoor positioning method based on key reference RFID tags, in: Proceedings of the 2009 IEEE Youth Conference on Information, Computing and Telecommunication (YC-ICT 2009), IEEE, Piscataway, NJ, USA, 2009, pp. 4245. [58] Y. Huang, S. Lv, Z. Liu, W. Jun, S. Jun, The topology analysis of reference tags of RFID indoor location system, in: Proceedings of the 2009 3rd IEEE International Conference on Digital Ecosystems and Technologies (DEST), IEEE, Piscataway, NJ, USA, 2009, pp. 313317. [59] Z. Zhen, Q. Jia, C. Song, X. Guan, An indoor localization algorithm for lighting control using RFID, in: Proceedings of the 2008 IEEE Energy 2030 Conference (ENERGY 2008), Inst. of Elec. and Elec. Eng. Computer Society, Atlanta, GA, United states, 2008. [60] U. Rueppel, K.M. Stuebbe, BIM-based indoor-emergency-navigation-system for complex buildings, Tsinghua Sci. Technol. 13 (2008) 362367. [61] A. Pradhan, E. Ergen, B. Akinci, Technological assessment of radio frequency identication technology for indoor localization, J. Comp. Civil Eng. 23 (2009) 230238. [62] S. Taneja, A. Akcamete, B. Akinci, J. Garrett, L. Soibelman, E.W. East, Analysis of three indoor localization technologies to support facility management eld activities, in: Proceedings of ICCCBE10. Nottingham, UK, June 30July 2, 2010. [63] P.M.M.A. Silva, M. Paralta, R. Caldeirinha, J. Rodrigues, C.M.J.A. Serodio, Traceme: indoor real-time location system, in: Proceedings of the 35th Annual Conference of IEEE Industrial Electronics (IECON 2009), IEEE, Piscataway, NJ, USA, 2009, pp. 27212725. [64] X. Qu, L.T. Simpson, P. Staneld, A model for quantifying the value of RFIDenabled equipment tracking in hospitals, Adv. Eng. Inform. 25 (2011) 2331. [65] S. Park, S. Hashimoto, An intelligent localization algorithm using read time of RFID system, Adv. Eng. Inform. 24 (2010) 490497. [66] J. Raper, G. Gartner, H. Karimi, C. Rizos, Applications of location-based services: a selected review, J Location Based Serv. 1 (2007) 89111. [67] D.S. Chiu, K.P. OKeefe, Seamless outdoor-to-indoor pedestrian navigation using GPS and UWB, in: Proceedings of the 21st International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS 2008), The Institute of Navigation, 1 Savannah, GA, United states, 2008, pp. 322333. [68] H.M. Khoury, V.R. Kamat, High-precision identication of contextual information in location-aware engineering applications, Adv. Eng. Inform. 23 (2009) 483496. [69] L. Pei, R. Chen, Y. Chen, H. Leppakoski, A Perttula, Indoor/outdoor seamless positioning technologies integrated on smart phone, in: Proceedings of the 2009 First International Conference on Advances in Satellite and Space Communications (SPACOMM), IEEE, Piscataway, NJ, USA, 2009, pp. 141145. [70] K. Yelamarthi, D. Haas, D. Nielsen, S. Mothersell, RFID and GPS integrated navigation system for the visually impaired, in: Proceedings of the 53rd IEEE International Midwest Symposium on Circuits and Systems (MWSCAS 2010) Institute of Electrical and Electronics Engineers Inc., Seattle, WA, United states, 2010, pp. 11491152. [71] M.G. Guillemette, I. Fontaine, C. Caron, Hybrid RFID-GPS real-time location system for human resources: Development, impacts and perspectives, in: Proceedings of the 41st Annual Hawaii International Conference on System Sciences 2008, Inst. of Elec. and Elec. Eng. Computer Society, Big Island, HI, United states, 2008. [72] B. Akinci, H. Karimi, A. Pradhan, C. Wu, G. Fichtl, CAD and GIS interoperability through semantic web services, Electron. J. Inf. Technol. Constr. 13 (2008) 39 55. [73] C.H. Caldas, D.G. Torrent, C.T. Haas, Using global positioning system to improve materials-locating processes on industrial projects, J. Constr. Eng. Manag. 132 (2006) 741749.

[31] S.Y.L. Yin, H.P. Tserng, J.C. Wang, S.C. Tsai, Developing a precast production management system using RFID technology, Autom. Constr. 18 (2009) 677 691. [32] T. Kaneko, K. Hamada, T. Kondo, Development of construction logistics system using radio frequency identication, in: Proceedings of ISARC 2007, pp. 5562. [33] Z. Ren, C.J. Anumba, J. Tah, RFID-facilitated construction materials management (RFID-CMM) a case study of water-supply project, Adv. Eng. Inform., in press. [34] J. Song, C.T. Haas, C.H. Caldas, Tracking the location of materials on construction job sites, J. Constr. Eng. Manag. 132 (2006) 911918. [35] D.G. Torrent, C.H. Caldas, Methodology for automating the identication and localization of construction components on industrial projects, J. Comp. Civil Eng. 23 (2009) 313. [36] J. Song, C.T. Haas, C.H. Caldas, A proximity-based method for locating RFID tagged objects, Adv. Eng. Inform. 21 (2007) 367376. [37] P.M. Goodrum, M.A. McLaren, A. Durfee, The application of active radio frequency identication technology for tool tracking on construction job sites, Autom. Constr. 15 (2006) 292302. [38] S. Chae, T. Yoshida, Application of RFID technology to prevention of collision accident with heavy equipment, Autom. Constr. 19 (2010) 368374. [39] T. Umetani, T. Arai, Y. Mae, K. Inoue, J. Maeda, Construction automation based on parts and packets unication, Autom. Constr. 15 (2006) 777784. [40] L. Wang, Enhancing construction quality inspection and management using RFID technology, Autom. Constr. 17 (2008) 467479. [41] S. Chin, S. Yoon, C. Choi, C. Cho, RFID + 4D CAD for progress management of structural steel works in high-rise buildings, J. Comp. Civil Eng. 22 (2008) 7489. [42] C. Ko, RFID-based building maintenance system, Autom. Constr. 18 (2009) 275284. [43] E. Ergen, B. Akinci, R. Sacks, Life-cycle data management of engineered-toorder components using radio frequency identication, Adv. Eng. Inform. 21 (2007) 356366. [44] J. Hightower, G. Borriello, R. Want, SpotON: an indoor 3D location sensing technology based on RF signal strength, Department of Computer Science and Engineering, University of Washington, 2000-02-02 Seattle, WA, 2000, pp. 1 16. [45] K. Yu, C. Liao, M. Lee, H. Lin, Design and implementation of a RFID based realtime location-aware system in clean room, in: Proceedings of the 2009 IEEE International Symposium on Parallel and Distributed Processing with Applications (ISPA 2009), IEEE Computer Society, Chengdu, Sichuan, China, 2009, pp. 382388. [46] J. Zhou, J. Shi, A comprehensive multi-factor analysis on RFID localization capability, Adv. Eng. Inform. 25 (2011) 3240. [47] L.M. Ni, Y. Liu, C.L. Yiu, A.P. Patil, LANDMARC: indoor location sensing using active RFID, Wireless Netw. 10 (2004) 701710. [48] S. Polito, D. Biondo, A. Iera, M. Mattei, A. Molinaro, Performance evaluation of active RFID location systems based on RF power measures, in: Proceedings of the 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC07), Institute of Electrical and Electronics Engineers Inc., Athens, Greece, 2007. [49] T. Zhang, Z. Chen, Y. Ouyang, J. Hao, Z. Xiong, An improved RFID-based locating algorithm by eliminating diversity of active tags for indoor environment, Comput. J. 52 (2009) 902909. [50] P.W. Hsu, T.H. Lin, H.H. Chang, Y.T. Chen, C.Y. Yen, Y.J. Tseng, et al., Practicability study on the improvement of the indoor location tracking accuracy with active RFID, in: Proceedings of the 2009 WRI International Conference on Communications and Mobile Computing, IEEE, Piscataway, NJ, USA, 2009, pp. 165169. [51] K. Sue, C. Tsai, M. Lin, FLEXOR: a exible localization scheme based on RFID, in: Proceedings of the International Conference on Information Networking (ICOIN 2006), Springer Verlag, 3961 LNCS Sendai, Japan, 2006, pp. 306316. [52] Y. Huang, Z. Lui, G. Ling, An improved Bayesian-based RFID indoor location algorithm, in: Proceedings of the 2008 International Conference on Computer Science and Software Engineering (CSSE 2008), IEEE, Piscataway, NJ, USA, 2008, pp. 511514. [53] R. Silva, P.A. Goncalves, Enhancing the efciency of active RFID-based indoor location systems, in: Proceedings of the 2009 IEEE Wireless Communications and Networking Conference (WCNC 2009), Institute of Electrical and Electronics Engineers Inc., Budapest, Hungary, 2009.