Sei sulla pagina 1di 14

Automation in Construction 16 (2007) 354 – 367

www.elsevier.com/locate/autcon

Tracking and locating components in a precast storage yard utilizing


radio frequency identification technology and GPS
Esin Ergen a , Burcu Akinci b,⁎, Rafael Sacks c
a
Department of Civil Engineering, Istanbul Technical University, Istanbul, 34469, Turkey
b
Department of Civil and Environmental Engineering, Carnegie Mellon University, Pittsburgh, PA, 15213, USA
c
Faculty of Civil and Environmental Engineering, Technion - Israel Institute of Technology, Haifa, Israel
Accepted 25 July 2006

Abstract

Problems in existing manual methods of identifying, tracking and locating highly customized prefabricated components result in late deliveries,
double-handling and misplacement of components, and incorrect installations that lead to schedule delays and increased labor costs. To eliminate these
deficiencies, an automated system using radio frequency identification technology combined with GPS technology, requiring minimal worker input, is
proposed. The requirements and approaches needed to utilize the system for locating precast concrete components with minimal worker input in the
storage yard of a manufacturing plant were developed. Based on the requirements identified and approaches formalized, a prototype system was
developed, assembled and tested in the field at a precast storage yard. The prototype system succeeded in automatically identifying pieces that were
relocated, demonstrating feasibility of the approach.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Automated tracking; Locating; Radio frequency identification; Precast; Component

1. Introduction sources to track and locate components using manual approaches


[1,2].
In construction supply chains, problems in poorly identifying, The objective of the study described in this research was to
tracking and locating highly customized prefabricated compo- identify the requirements for and develop an appropriate auto-
nents result in late deliveries, double-handling and misplacement mated approach to tracking and locating precast concrete (hence-
of components, and incorrect installations that lead to schedule forth termed simply ‘precast’) components in a manufacturer's
delays and increased labor costs [1]. Tracking and locating prefa- storage yard using advanced tracking technologies. Since just-in-
bricated components individually at any point across a construc- time delivery is required for precast pieces at construction sites, it
tion supply chain is a challenging task. Manufacturing plants of is critical for a precast manufacturer to know exactly where any
prefabricated components have a dynamic nature where many individual piece is located in a storage yard, so that any com-
unique components are frequently relocated and shipped daily. ponent that is requested from a construction site can be located
Similarly, prefabricated components are frequently received, and supplied quickly [2–4]. Precast storage yards typically cover
stored, relocated and installed at construction sites. The cus- several thousands of square yards; in the case of large-scale
tomized nature of prefabricated components amplifies the chal- precast companies, yards commonly contain up to 4500 pieces at
lenge of identifying and locating them along their supply-chain any given point in time. Accurately and reliably tracking such a
due to the fact that most of them are unique and therefore need to large number of components in a wide area is a challenging task.
be tracked individually. Under these challenging conditions, both Existing approaches for tracking components utilize manual
prefabrication companies and contractors invest significant re- tracking by paper-based documents or barcodes. Information
collected using such labor-intensive methods is not reliable or
⁎ Corresponding author. Tel.: +1 412 268 2959; fax: +1 412 268 7813. complete since these data collection methods rely on workers'
E-mail address: bakinci@cmu.edu (B. Akinci). motivation. Unreliability in material tracking might result in
0926-5805/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.autcon.2006.07.004
E. Ergen et al. / Automation in Construction 16 (2007) 354–367 355

delays in shipping due to pieces that cannot be located on time or information, and thus have limited identification range requi-
misplaced pieces that have to be manufactured again [2]. Current ring that a labor-intensive scanning activity be performed on
practice highlights the need for a tracking system that provides each object. In addition, some of these technologies cannot
up-to-date location information for precast pieces in the storage survive in harsh environmental conditions (e.g., rain, dust,
yard with minimum human input. impact), typically encountered in precast storage yards. RFID
Radio frequency identification (RFID) integrated with the technology is durable, does not theoretically require line-of-
global positioning system (GPS) provides an opportunity to sight and some types of RFID technology have relatively long
uniquely identify precast components and to track and locate reading/detection ranges up to 30–90 m.
them using minimal or no worker input. Although RFID tech- RFID has two main components: a reader and a tag. The tag,
nology is being used to track assets in other industries, very little which consists of an electronic chip coupled with an antenna, is
research has been conducted in the architecture, engineering and attached to an object and stores data about the object. The
construction (AEC) industry or in facility management (FM) on reader, combined with an external antenna, reads/writes data
the utilization of RFID technology for tracking components from/to a tag via radio frequency and transfers data to a host
[1,3,5–8]. Among these studies, only Song et al. [8] provided a computer. Some tags have LED notifying the user with a blin-
reasoning mechanism based on a proximity method for locating king light during communication. Reading and writing ranges
materials in a construction site using RFID and GPS. In this depend on the operation frequency (low, high, ultra high and
approach, construction site is scanned daily in detail to identify microwave), whether the tag needs a battery to operate (active)
the location of materials on a given site. Whereas, in the or not (passive) and whether some materials, which can interfere
research described in this paper, the goal has been to track with radio signals, exist in the environment (e.g., concrete,
components as they are moved to different locations within a steel). Finally, unlike barcodes, RFID tags can withstand harsh
storage yard. conditions [1,5].
To identify the requirements and to develop reasoning me- Tags operating at ultra high frequency (UHF) typically have
chanisms, an initial field test was conducted to test the per- longer reading ranges than tags operating at other frequencies.
formance of RFID on precast components and GPS in a precast Similarly, active tags have typically longer reading ranges than
yard. Based on the results of such an initial test, a detailed list of passive tags. However, a limitation of active RFID technology
requirements and corresponding approaches was formalized. A is that it requires battery management since the lifetime of an
prototype system, which utilizes the approaches, was then internal battery is approximately 5–10 years. In addition, the
developed, assembled and tested in the same precast storage cost of active tags and readers is relatively high compared to
yard. barcode technology. However, since it is an evolving technol-
ogy, new RFID systems have consistently decreasing costs and
2. Background increasing data storage capacities.
In recent years, different industries started to use RFID
Baldwin et al. [2] identified the need for improvement in technology to track a variety of discrete components, such as
locating precast components in a storage yard and conducted a vehicles, containers, assets and people. Tracking containers in a
field test at a precast plant to determine the feasibility of using port is similar to tracking precast components in a storage yard:
barcodes at the storage yard and during shipping. In spite of the Both containers and precast components are large in size, are
labor-intensive data collection activities associated with the stacked on top of each other in large storage areas and similar
utilization of barcodes, during which pieces are manually scan- cranes are used to relocate the pieces. The current applications
ned at each and every relocation in the storage yard, typical for tracking containers at a port via RFID do not utilize RFID
savings were estimated to be very high: 70% time saving during for automated identification of containers, but only for
checking out precast beams while loading to a trailer, 30% time automated tracking of containers. In those applications, con-
saving in locating a beam in a storage yard, and 85% time tainer identification processes rely on optical character recog-
saving in clerical time for entering beam data to the company's nition (OCR) [13] or manual identification [14]. RFID is only
computer system. However, our observations at a precast used for automatically tracking the cranes that carry the iden-
manufacturing plant that utilizes barcodes for tracking compo- tified containers' in the storage yard via real-time locating
nents during storage showed that workers try to avoid labor- systems. However, in this research explained in this paper, the
intensive barcode scanning activities and as a result, the data is goal is to use RFID during the entire tracking and locating
not complete. If the data is not complete, then it is not con- processes including identification of components.
sidered reliable and therefore not considered worth maintaining. Previous research on RFID in the construction industry
This creates a detrimental spiral of decreasing trust on the part mostly focused on three areas: (1) identification of conceptual
of the workers to the data provided in such databases. In prac- applications of RFID technology in construction [14,15], (2)
tice, it has been observed that the expected savings of using performance of RFID for automated identification during deli-
barcodes have not been fully realized. very and receipt of components [1,5,16], and (3) for tool
Besides barcodes, various other automated identification tracking [7] at construction sites. In all these studies, com-
technologies, such as two dimensional barcodes, RFID, optical ponents were either manually scanned with handheld readers
character recognition and touch probes exist [5,9–12]. Most of [5,7,16] or scanned at a fixed portal as a trailer with components
these technologies require line-of-sight to capture identification passed through a gate [1]. Only one research effort focused on
356 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

locating the materials that are scattered on a construction site, large, they are almost indistinguishable externally (despite
and combined active UHF RFID and GPS technologies using being unique in internal design), and generally represent the
proximity techniques [8]. This approach provides approximate large majority of the pieces in buildings of which they are a part.
locations of materials on construction site, and it can be used as A double-tee is a type of precast concrete beam that is used
a front-end solution for the research described in this paper to as a slab in precast structures. Typical dimensions of a double-
identify the components' initial locations in the storage yard. tee produced by this manufacturer are 0.6 m in height, 4 to 5 m
GPS, a relatively mature technology, is also used for tracking in width and 10 to 15 m in length. They are transported within a
objects in outdoor environments. GPS was used to track the storage yard by mobile gantry cranes. Double-tees are placed on
real-time location of equipment on construction sites in several top of each other when stored; each vertical group of compo-
research studies [17–19]. Although Navon and Shpatnitsky [20] nents is called a “stack”. A stack can have up to five double-
identified GPS technology as an accurate and robust technology tees. A storage yard is typically divided into zones, commonly
for automated data collection for road construction control, they called as “aisles”. Each aisle is composed of rows and each row
acknowledged the inaccuracy of GPS data observed which was has two stacks (Fig. 1).
caused by objects that shielded the communication of the GPS
receiver with satellites. 3.1. Opportunities for improvement in the current storage
In combination, RFID and GPS technologies present an processes
opportunity to track large precast components with minimum
labor input in a large storage yard. However, prior to the current Effective tracking and locating of pieces in a storage yard is
work, no research has been pursued to identify the requirements primarily needed to provide just-in-time (JIT) deliveries to
and reasoning mechanisms needed to exploit these advanced construction sites. Precast components are large in size and
tracking technologies for this particular application. delivered incrementally, almost everyday, to job sites; thus,
everyday many pieces need to be located at a precast storage
3. Existing conditions and high-level requirements yard. Currently, anecdotal evidence shows that locating a piece
takes from 5 min to 5 h depending on the accuracy and com-
The research described in this paper was conducted in pleteness of the location information. Delays in locating pieces
collaboration with a large-scale precast manufacturer. The com- require utilization of additional resources to find pieces, while a
pany operates a state-of-the-art information system with bar- failure to locate a piece has a more significant impact in terms of
code labeling. The storage yard at the facility where the field time and cost since that piece has to be manufactured again.
tests were carried out has a total area of 190,000 m2, where Moreover, delays in shipping of pieces result in wasted re-
pieces are typically stored from 3 weeks to 6 months before sources (cranes and crews) at the site and consequent delays in
shipping. The tests were performed in the double-tee compo- construction schedules.
nent storage yard, where the majority of problems in existing In addition to locating pieces for shipping, precast manu-
practice are observed: double tee pieces are comparatively very facturers also need to locate pieces periodically prior to shipping,

Fig. 1. A portion of the double-tee storage yard.


E. Ergen et al. / Automation in Construction 16 (2007) 354–367 357

Fig. 2. Schematic layout used to record locations of double-tees in a storage yard.

for inspection, surface treatment, patching, etc. For example, Finally, when an erector confirms the delivery date for this type
owners' representatives visit manufacturing plants to check of piece in the daily load list, material handling personnel
pieces in terms of quality and quantity. During these visits, all retrieve the location records for the storage yard and locate the
pieces in a building (usually hundreds and often thousands) are piece to load it onto a trailer for shipping.
inspected; delays or failures to locate pieces not only result in To locate pieces effectively, the location data must be collected
waste of time for both manufacturers' and owners' staff, but also and recorded accurately, and they should be easily retrieved by the
foster mistrust between the two parties. material handling personnel. In the cases observed at two plants,
The current tracking process includes a data collection both paper-based methods and barcodes were used. In the paper-
activity that has to be repeated each time a piece is relocated in a based approach, the ID of the piece is written on the piece and the
storage yard. The storage process for a double-tee starts with a aisle and row IDs are written at each row. Once a piece is placed in
piece being taken to the storage yard once it is cast and cured. a stack, a worker identifies in which aisle and row the piece was
The piece is placed on top of one of the stacks in the yard and its placed using the storage yard's schematic layout plan, and
ID and its location are recorded. During storage, a double-tee recording the ID of the piece for that location on the layout plan
might be moved several times if a piece underneath that double- (Fig. 2). This process is repeated for each piece that is relocated in
tee in the stack needs to be treated or shipped. Each time the the storage yard, and the storage layout plan is manually updated
piece is moved, its new location must be collected and recorded. accordingly. When a piece needs to be located, workers scan the

Fig. 3. Proposed approach for automating tracking and locating of precast pieces.
358 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

layout and manually search for the ID of the required piece. This GPS receiver, which is also mounted on the crane. The ID and
method is error-prone and time-consuming since data collection, location information of the piece is then sent to a database
entry and retrieval are done manually. for retrieval for shipping or for inspection by the owner or architect
In the barcode approach, a barcode label is attached to each later.
piece after it is cast. In addition, each stack location has a It is expected that automatic collection of location informa-
barcode. As a piece is placed in a stack, a worker locates the tion of the pieces can minimize inaccurate or incomplete data
barcode on the piece and climbs up the stack to scan its barcode. due to minimum human involvement as long as the accuracy of
Then, the same worker walks up to one end of the stack, which the automated approach exceeds the accuracy and completeness
could be 12–15 m away, to scan the barcode of the stack. This of the manual and barcode-based approach. Accurate location
data is then sent to a database using wireless communication. In information will be used to locate pieces in a timely manner.
this approach, data entry is performed automatically and data is This will reduce the resources used for locating misplaced
retrieved more effectively since IDs are stored in a database in a pieces, delays in shipping, eliminate the need for replacing
searchable format. However, data collection is still manual. missing pieces, and thus improve the reliability and stability of
Both paper-based and barcode methods are not very effective just-in-time deliveries to construction sites.
because of their dependence on manual data collection. Manual Exploration of the feasibility of this conceptual approach to
data collection is labor-intensive and thus expensive and automation of tracking of precast pieces using existing techno-
inefficient. Since workers have to spend extra time to collect logies was pursued in the following steps:
information and this is considered as a secondary task, data
collection is sometimes neglected or not fully performed. This (a) selection of appropriate technologies, statement of de-
results in incomplete or inaccurate data that leads to misplaced tailed requirements and identification of appropriate hard-
pieces in the storage yard. ware, including preliminary field testing,
(b) development of a prototype and of the necessary data
3.2. High-level requirements for tracking precast pieces processing algorithms,
(c) design and assembly of a physical prototype for proof of
Based on the description of the current practice described in concept and
the previous section, data collection was identified as the (d) field tests.
primary focus for improvement by utilizing advanced tracking
technologies. In light of the failures of the existing system, the The first three steps are described in this section; the results
following high-level requirements were defined: of the field tests are reported in the following section.

(1) The new location of a piece delivered to the storage yard 4.1. Functional requirements and technology selection
shall be identified with minimal or no worker input.
(2) Any piece that is relocated in the storage yard shall be The identified requirements emphasize the need for mini-
identified with minimal or no worker input. mizing or eliminating worker input when tracking and locating
(3) The accuracy with which a piece is identified and located components. Among available identification technologies,
shall be higher than with the current system; ideally, it barcode (one-dimensional and two-dimensional), optical char-
should get close to zero error. acter recognition and touch probes do not meet this requirement
(4) Performance reduction of any selected technology under because they require human action to effect a reading (they all
harsh construction conditions (e.g., noise, dust, harsh need line-of-sight or direct contact at short range). They also
light) and under the presence of metal and concrete shall require clean environments. RFID technologies that have longer
be minimal: Pieces are stored in an open air environment, reading ranges (i.e., UHF and microwave frequencies) meet the
therefore visibility may be obscured due to sunlight. minimum worker input requirement since they enable identi-
Metal cranes are used, and the pieces themselves have fication of objects from relatively long distances without line of
large volumes of steel and concrete; metal and concrete sight. Thus, RFID technology operating at UHF or microwave
are commonly observed in the environment. frequencies was selected for identification of pieces.
In addition to RFID, GPS technology was selected since it is
4. Automated tracking a mature technology used for tracking objects in open air. What
makes GPS technology applicable and feasible in this case is
To meet the requirements for minimum worker input and that precast pieces can only be moved by cranes. It is therefore
minimum change to the current storage process, an automated data ensured that the location of the crane can be used to establish the
collection approach was developed. In this approach, RFID tags location of the pieces, using only a single GPS receiver mounted
that contain unique ID numbers are placed on precast pieces, and an on the crane instead of a receiver placed on each piece.
RFID reader is mounted on the mobile gantry crane, as indicated in
Fig. 3. Each time a piece is picked up and moved, the ID informa- 4.1.1. RFID technology
tion of the piece is captured by the RFID reader. At the times of pick The technology-specific requirements for the RFID system
up and release of the load–which are identified by a load cell on the in the proposed approach, based on the high-level requirements,
crane–the location of the crane (and thus the piece) is read from a were the following:
E. Ergen et al. / Automation in Construction 16 (2007) 354–367 359

antenna would be placed at the center of gravity of the picking


bar (a steel girder used to distribute the load of the piece).
• To prevent “multiple simultaneous tag readings”, the maximum
reading range should be limited to 5 m. This requirement
derives from the geometry of the pieces and the layout of the
storage yard (see Fig. 4). Given the typical width of pieces, the
shortest distance between the centers of gravity of adjacent
pieces is at least 5 m. When a piece is placed on a stack, it is
desirable that the tags on neighboring pieces should not be
readable, as this would create uncertainty as to which ID is the
correct one. This requirement can be relaxed by using an algo-
rithm to distinguish between multiple readings (as described
below) or if the signal strength can be measured, in which case
the strongest signal read would be assumed to be the correct ID.
• To minimize the performance reduction of selected technol-
ogy under harsh conditions (e.g., rain or possible impacts
from different pieces of equipment) and while in contact with
Fig. 4. Distances between aisles and rows in a storage yard. metal and concrete, RFID tags will be encapsulated or
insulated. At the chosen frequencies (i.e., UHF and micro-
wave), the reading range of the tags decreases if the tags are
• To automatically identify each piece that is picked up by the attached to metal or concrete objects. The degree of degra-
crane, the reading range must be at least 3 m. This requirement dation depends on how much of the area of a tag is in contact
derives from the geometry of the crane and the pieces (Fig. 4). with the object. Thus, encapsulated or insulated tags are
It is the distance between the optimal location of the reader needed to prevent direct contact with precast pieces. In
with respect to the location of the tag on a typical piece that is addition, RFID tags contain a chip and an antenna; if the tag
being carried. The RFID tag on the piece should always be in is not encapsulated with a durable material, both could be
the range of the RFID reader during identification to enable damaged by moisture or impact (workers continuously climb
automatic identification on demand. In the case of the mobile on double-tees to attach them to the crane).
gantry cranes, the optimal location for the tag is the center of • The RFID antenna should have wireless communication with
gravity point on each piece, while the corresponding reader the host computer. The reason for this requirement is that the

Fig. 5. (a) Passive UHF RFID system, (b) active UHF RFID system.
360 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

picking bar is suspended by cables from the frame of the the prototype application. Since no RFID equipment with
gantry crane and moves in relation to the frame, making a wireless communication with the antenna is available, the RFID
cable connection highly undesirable.Various products were tags were placed at the center points of the long sides of the
evaluated in terms of meeting these functionalities and two pieces and the antenna was placed on the operator's cabin. To
specific candidate RFID systems were selected: a passive test the RFID systems, a piece was picked up by the crane and
UHF RFID system (Fig. 5a) and an active UHF RFID system carried to another stack; tags were read continuously during the
(Fig. 5b). Both systems had the longest reading ranges in process to determine if the piece that was picked up could be
their respective categories. The active system was claimed to detected all the time. In addition, the reading success for tags
have up to 30 m reading range in open air, and the passive that were placed at different locations on a component was
system was claimed to have a corresponding 4.5–7.5 m investigated.
reading range. Both active and passive tags were encapsu- The results of the preliminary test demonstrated that the
lated or insulated to perform effectively around metal and in active RFID system performed well in-real-life conditions;
harsh environments. However, both technology suppliers however, the passive RFID system did not meet the require-
stated that the RFID systems need to be tested in real-life ments. For the passive system, the reading range was measured
conditions to determine the actual reading range when the to be between 1 and 5.5 m, i.e., it was not consistently and
tags are in contact with concrete and metallic objects that reliably greater than the minimum requirement (3 m). In addi-
exist in the environment. tion, detection of the tags was not reliable because tags were
observed to fall out of reading range if they were not in line with
A preliminary field test was conducted to evaluate the per- the antenna both horizontally and in elevation. This demon-
formance of the two selected RFID systems in real conditions at strated that passive RFID systems are not reliable in a dynamic
a precast storage yard, to refine the technology-specific functio- environment, where the antenna and the tags are not closely in
nalities and to identify the reasoning mechanisms needed for line with each other all the time.

Fig. 6. Hardware setup in the prototype system.


E. Ergen et al. / Automation in Construction 16 (2007) 354–367 361

The observed reading range for the active UHF RFID system of pieces, the shortest distance between the centers of gravity
was approximately 6–7.5 m, which is 20%–25% of the nominal of adjacent pieces is at least 5 m.
reading range. This result meets the minimum requirement. The • The GPS receiver shall have a wireless communication with
reading range of the active RFID technology was sufficient to the host computer. As explained above, the approach for
read the piece ID regardless of the location of the picking bar. identifying the location requires that location data be collec-
However, it was not possible to read through a component. ted at the center points of pieces. Thus, the GPS receiver
Another interesting result of the field test of active RFID tags must be placed at the center point of the picking bar of the
was that, during relocation, the reader not only detected the tag crane. As in the case of the RFID antenna, since the picking
that was attached to the piece being relocated, but also other tags bar is a mobile component of the crane, wires connecting the
that were close by. The inability to restrict the reading range, GPS to the host computer would limit the movement of the
violating the maximum range requirement, demonstrated the picking bar.
need for a reasoning mechanism that can filter all the tag IDs • The GPS unit will be durable to function in open air conditions.
that are collected during relocation in order to uniquely identify Since the GPS unit will be placed on the picking bar of the
the piece that is actually picked up. crane, it should be durable to open air conditions such as rain.

4.1.2. GPS Budget constraints dictated that a GPS receiver with less than
The technology-specific requirements for the GPS were the required accuracy was used for the field tests and the
identified as follows, based on the high-level requirements and prototype. The GPS receiver used had a nominal accuracy of
the proposed approach given in the previous section: 5 m with WAAS (Wide Area Augmentation System) data cor-
rection and 15 m without the WAAS capability. WAAS is based
• To uniquely identify the locations of pieces that are placed in on a series of satellites and ground antennas that correct data
different rows or aisles, the GPS unit will have an accuracy transmitted between satellite and GPS devices. The GPS re-
of at least 2.5 m. The distance between the aisles is appro- ceiver used in the test used wireless communication with the
ximately 6 m and the distance between two adjacent rows is host computer using Bluetooth technology. It was also wea-
approximately half a meter (Fig. 4). Given the typical width therproof. In the preliminary field test, the GPS unit was

Fig. 7. (a) Schematic top view and (b) schematic side view of a gantry crane carrying a double-tee at two extreme locations (positions 1 and 2).
362 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

4.2. Prototype system design

The prototype system was developed by integrating the


active RFID system with a GPS unit using a laptop as a host
computer. The components of the system and their installation
locations are given in Fig. 6.
The RFID system was composed of a reader, an antenna and
active tags. The reader is in the form of a PCMCIA card plugged
into a laptop, which was placed in the operator's cabin. Because
no RFID equipment with wireless communication with the
antenna is available, the antenna was placed on the windshield
of the operator's cabin and attached to the reader using an RF
cable, and the tags were mounted at the center of one of the long
sides of the piece. The operator's cabin and the center of the
long sides of a piece were observed to be approximately in line
with each other both horizontally (Fig. 7a) and in elevation
(Fig. 7b), most of the time during which a piece is being carried
by the crane. This alignment would ensure that the tag attached
to the piece is in the range of the reader.
Based on the relative locations of the tag and reader in the
prototype setup, and the topography of the storage yard, the
practical required reading range was identified as approximately
Fig. 8. An example explaining the method for unique identification. The figure between 0.3 m and 6 m–8 m to be able to read the piece that is
shows a plan view of a storage yard and movement of the crane along with the
piece IDs detected.
on the far side of the crane (Fig. 7a). The tags were UHF active
tags with a battery of 5 years of lifetime and a LED that can be
flashed as the corresponding ID is read. The tags were hung
mounted on the picking bar. No problems observed in com- loosely at the edge of the flange of each piece opposite its center
municating with the GPS receiver placed on a metallic picking using plastic wire and tape. The GPS receiver was placed on the
bar or receiving GPS signals, and the accuracy proved to be picking bar using its magnetic mount and it communicated with
close to the nominal accuracy. the laptop using a Bluetooth connection.
The inaccuracy of the GPS unit was overcome by using an
approximation approach to identify the location and by assu- 4.3. Prototype application
ming that row-level accuracy for location is acceptable (since
currently the company tracks pieces by row numbers instead of A prototype application was developed to demonstrate how
stack numbers). The approach, which is explained in the follo- the system proposed in this research will: (1) communicate with
wing sections, acquires location data at center points of pieces the GPS receiver and RFID reader to acquire coordinates and
and compares them with the closest stack location. The closest tag ID data, (2) identify the tag ID that belongs to the piece that
distance between center points of two stacks is approximately is picked, (3) determine the location of the pieces that are
11 m for the pieces that are placed at two stacks in adjacent moved, and (4) store the new location with the ID of the piece
aisles. Thus, the accuracy required was half of this minimum that is moved in a database. During prototype implementation,
distance, which is 5.5 m. two reasoning methods were developed. The first was for

Fig. 9. Identification of the location of a piece at point A using a geo-referenced site plan.
E. Ergen et al. / Automation in Construction 16 (2007) 354–367 363

Fig. 10. The user interface of the prototype application for creating a geo-referenced map.

identifying the piece that is picked up by the crane, and the The method requires that reading the tags within the proximity
second was for identifying which stack and row the piece was of the reader must start as a piece is picked up by the crane and
placed in. The application that integrates GPS and RFID continue frequently (every 2 s) as the piece is carried to another
information was developed in Visual Basic .NET. stack. Reading stops when the piece is lowered onto its new
location. Once the piece is released, the proportional occurrence
4.3.1. Identifying a piece that is relocated of each ID in the total number of readings is calculated, and the ID
This method was developed to identify a piece that is re- with the highest occurrence rate is identified as belonging to the
located in the storage yard. It filters all of the IDs that are piece being picked up. Possible limitations of this approach and
detected in the environment as they come within the range of the how it is overcome are explained in the following paragraphs.
reader to identify which among them belongs to the piece that is An example can be seen in Fig. 8, where piece 421 is carried
moved. The method is based on the observation that only the ID by a mobile crane from stack C to stack A. As piece 421 is picked
of the piece carried appears in all readings performed as it is up, the reader detects it and also piece 103 (which is in the same
being transported by the crane, while readings of all the other stack) and pieces 422 and 424 (which are behind it). As the crane
piece IDs are transient. carries piece 421 to stack A, the reader on the crane detects

Fig. 11. The user interface of the prototype application for collecting ID and location information for a relocated piece.
364 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

Fig. 12. Picture and plan view of the test area.

pieces 101 and 102 as well as piece 421. Similar behavior is reading) to visually identify the tag and confirm that it belongs to
observed along the way as seen in Fig. 8. Finally, when piece 421 the piece that was carried.
is lowered onto stack A, the reader detects piece 100, which is in
the same stack. At this point, the method calculates that ID 421 4.3.2. Determining the location of a piece
had a 100% occurrence rate, while all others were less than 100% This approach is developed to determine the location of a piece
and thus identifies 421 as the piece carried. in the storage yard in terms of aisle and row numbers. In the field
The method overcomes two possible exceptions that could test case, the storage yard is divided into aisles and aisles are
occur due to possible obstructions or interferences in the envi- divided into rows. In each row, there are two stacks, where pieces
ronment. The first occurs when no IDs are detected during a are stacked on top of each other. In the current practice, the piece
reading. In this case, the reading is ignored. The second ex- locations are tracked by aisle and row only; stack numbers are not
ception occurs when the piece that is picked up is not detected in recorded. Nevertheless, in order to explore the boundaries of the
isolated readings due to random obstructions or interferences. In prototype system, the tests were setup to attempt identification of
this case, the rate of occurrence of the piece carried is less than location at the level of the stack position.
100%, but still significantly higher than that for other pieces. In this method, a geo-referenced map of the storage yard is
Conceivably, in rare cases where the occurrence rate of some created once, and then used to identify locations of the piece by
other piece or pieces is less than a user-calibrated threshold rate, comparing the coordinates received from the GPS with those in
the crane operator can be asked to flash the tag (perform a single the geo-referenced map. To generate the geo-referenced map,
E. Ergen et al. / Automation in Construction 16 (2007) 354–367 365

the user collects data at two opposite corners of each stack. The tags was attached to the piece that was to be picked up, and the
method then calculates and stores the middle point and corner last one was attached to the piece at the second highest level in
points of each stack. Once a geo-referenced map is created for a the stack immediately behind the piece that was to be picked up.
storage yard, it is used to identify the stack in which a piece is Eighteen relocations of components were performed and they
placed. This is done by comparing the GPS data collected to the involved three types of moves (Fig. 12a): (1) three same-row
geo-referenced map each time a piece is relocated. Since some moves as a piece is moved from one stack to another one in the
inaccuracy exists in the GPS device, this comparison is per- same row, (2) six adjacent-row moves as a piece is moved from
formed by finding the distance between the current location of one stack to another in an adjacent row, (3) nine adjacent-aisle
the piece to each stack, and identifying the stack that is the moves as a piece is moved from one aisle to an adjacent aisle. As
closest to the location of the piece, as illustrated in Fig. 9. explained earlier, same-row relocations are not tracked in current
practice. The three type 1 moves (as defined above) were
4.3.3. User interface included to explore the boundaries of the prototype system.
The user interfaces implemented for the prototype are shown The results of the tests are given in Table 1. The prototype was
in Figs. 10 and 11. The first interface enables calibration of the 100% successful when identifying the piece during all 15
geo-referenced map of the storage yard (Fig. 10). To collect adjacent-row or adjacent-aisle moves, but of the three same-row
stack coordinates, the user stands at the first corner, enters the moves, only in one instance could the piece ID be identified. In the
stack name and clicks on the “refresh” button. Once the inco- other two instances, there were no false-negatives in the list, and
ming GPS coordinate readings converge to stable numbers (i.e., the user was able to correctly select the piece ID from the short-
when the difference in the fourth digit after the decimal point list. This result was expected for the same row move, because,
varies less than 3), the user confirms capture of the first corner unlike the other moves, in the same-row move, the crane is
and progresses to the opposite corner. The procedure is repeated stationary; only the picking bar that holds the piece moves from
for all the stacks in the yard. one end of the crane to the other. Since the lack of wireless
The second interface screen (Fig. 11) was developed to enable communication between the RFID antenna and reader in the
the crane operator to trigger identification of a piece that is prototype prevented mounting the antenna on the picking bar,
relocated and to collect its new location data. This interface is reader was attached to the crane itself. Therefore, the tags in the
only applicable in the prototype: in a commercial application, the range of the reader do not change during relocation of the piece
triggering action would be automated. For example, a load sensor unless other tags on other pieces are blocked by the carried piece.
on the crane can be used to issue “start and stop reading” signals During the evaluation of the test results related to the location
to the GPS and RFID sub-systems when the load is lifted or when of the components, if the identified location of a component was
it is released, thus obviating the need for any user action at all. in the same aisle and row as the actual location of the component,
this result was considered acceptable. Of the 18 relocations, 61%
5. Field tests and results were identified within the acceptable limits. This was due to the
inaccuracies associated with the inexpensive GPS used in the test
The prototype system was tested in a real-life scenario, in and cloudy and stormy weather, which might block the GPS
which a piece was picked up by a mobile gantry crane and carried signals, during the test. Utilizing a GPS with sub-meter accuracy,
to another location. A person in the operator's cabin collected and installing a reference base station for correcting the GPS readings,
recorded the ID and location information of the piece using the are expected to solve this problem.
prototype application. The gantry crane is a critical resource for
the precast company since precast components can only be 6. Evaluation of the prototype system
relocated by gantry cranes. Also, operating cost of the crane is
high. Thus, the crane was available for a day for the test. The first and second high-level requirements for an auto-
The worker, who operated the crane, hooked and unhooked mated precast tracking system stated that data collection should
the precast components to/from the crane. In the process of
moving one component from one location to another, the
worker climbed down the crane and up to the stack to hook the Table 1
Field test results
component to the crane and repeated the same steps to unhook
the component. Including the set-up and initial testing time, the Move type Number RFID tag readings GPS location results
of tests
test could be repeated 18 times in one day. Correct List of Exact Same Next
A densely populated area of the double-tee storage yard was ID possible row aisle
IDs (∼ 5 m) (∼ 15 m)
chosen as the test location to test the effect of tags attached to
other pieces in close proximity with the precast component that 1. Same row 3 1 2 – 1 2
2. Adjacent row 6 6 – 1 3 2
will be moved (Fig. 12). In preparation for the tests, the GPS
3. Adjacent aisle 9 9 – 5 1 3
coordinates for the stacks in the test area (shown shaded in Fig. Acceptable results 18 16 (89%) 11 (61%)
12b) were retrieved and stored in a database. The hardware was (all moves)
installed on the crane as shown in Fig. 6, and eight tags were Acceptable results 15 15 (100%) 10 (66%)
attached to eight pieces in the test area. Six of those eight tags (move types
2 and 3)
were attached to the pieces at the top of each stack. One of the
366 E. Ergen et al. / Automation in Construction 16 (2007) 354–367

require minimal or no user action. In the prototype system, the or concrete in the environment, even while using Bluetooth
crane operator is required to press several buttons in the user communication. Utilizing a GPS with sub-meter accuracy or
interface in the crane cabin to collect ID and location data. No installing a base station for correcting the GPS readings (using
other data collection activities, such as climbing on top of a differential GPS technology) are expected to increase the rate of
stack or walking along rows to scan barcodes, or manual data successful location identification.
entry, are needed. In a production system, the RFID and GPS In the prototype system, data collection activities did not
subsystems could also be integrated with a load sensor placed require any change to the current storage process (except requi-
on the crane to eliminate the activities of the crane operator for ring the operator to press buttons to activate data the readers).
data collection, thus achieving a fully automated system. Moreover, all data collection activities previously required, such
The third requirement stated that the accuracy with which a as scanning a barcode or manually entering data to the database
piece is identified and located shall be higher than possible with system, were eliminated.
the current system. The piece was identified semi-automatically In the short term, similar tests should be conducted to collect
in all cases and fully automatically in cases where the crane was more data points and further research should focus on refining
not stationary. If the reader was placed closer to the tags (e.g., on the prototype system as suggested, to further validate the hypo-
the picking bar), and the reading range was limited accordingly thesis that a fully automated and reliable system is feasible. In
to read only the tag on the piece, fully automated reliable the long term, application of the same concept to monitoring of
identification could be achieved also for the cases where the precast pieces at the construction site should also be explored.
crane does not move. The location was identified with 61%
success for relocations that are currently tracked in the system. Acknowledgments
The failure of the location identification would probably be
corrected using a GPS with better accuracy, and by combining This work was partially funded by High Concrete and the
the approach with tag ID and signal strength data in the envi- Precast/Prestressed Concrete Institute (PCI). The authors grate-
ronment to increase the confidence. fully acknowledge this sponsorship. The authors also great-
The fourth requirement related to the need for consistent lyappreciate the assistance and support received from High
performance under harsh construction conditions and in the Concrete.
presence of metal and concrete. Both the RFID tags and GPS
were designed for harsh environments, and no problems were References
observed in terms of receiving the signals from RFID tags and
[1] J. Song, C.T. Haas, C. Caldas, E. Ergen, B. Akinci, Automating the task of
GPS unit in the field conditions. Apart from a small number of
tracking the delivery and receipt of fabricated pipe spools in industrial
individual readings, the RFID tag of the piece that was carried projects, Automation in Construction 15 (2) (2006) 166–177.
could be consistently detected in the required range. [2] A.N. Baldwin, A. Thorpe, J.A. Alkaabi, Improved materials management
through bar-coding: results and implications from a feasibility study, Pro-
7. Conclusions ceedings of the Institution of Civil Engineers, Civil Engineering 102 (4)
(November 1994) 156–162.
[3] E. Ergen, B. Akinci, R. Sacks, Formalization and automation of effective
The field tests demonstrated the basic feasibility of an tracking and locating of precast components in a storage yard, Proceedings
automated system that integrates RFID and GPS technologies of EIA-9, E-Activities and Intelligent Support in Design and the Built
for tracking precast pieces in a storage yard. The prototype Environment, 9th EuropIA International Conference, Istanbul, Turkey
system used for the tests had all the features of the conceptual (October 2003) 31–37.
[4] L.S. Pheng, C.J. Chuan, Just-in-time management of precast concrete
system, with three exceptions: the RFID antenna was mounted
components, Journal of Construction Engineering and Management 127
on the crane cabin rather than on the picking bar in close (6) (2001) 494–501.
proximity to the piece centers, the GPS used did not meet the [5] E.J. Jaselskis, T. El-Misalami, Implementing radio frequency identification
minimal accuracy requirement, and user intervention was re- in the construction process, Journal of Construction Engineering and
quired to activate the readers of both subsystems. Management 129 (6) (2003) 680–688.
[6] R. Sacks, B. Akinci, E. Ergen, 3D modeling and real-time monitoring in
Despite these restrictions, the prototype system was suc-
support of lean production of engineered-to-order precast concrete
cessful in semi-automatically identifying all of the pieces that building, in: J. Martinez, C. Formoso (Eds.), Proceed. of the 11th Annual
were relocated. The reading range of the RFID tags was Conference on Lean Construction, July 2003, pp. 398–410, Blacksburg,
reduced to 1/4th–1/5th of the nominal reading range for open VA USA.
air environments; however, this reading range was still [7] P.M. Goodrum, M.A. McLaren, A. Durfee, The application of radio
frequency identification technology for tool tracking on construction job
sufficient for the precast tracking case. Mounting a wireless
sites, Automation in Construction 15 (3) (2006) 292–302.
antenna on the picking bar, and using a load sensor to activate [8] J. Song, C.T. Haas, C. Caldas, K. Liapi, Locating materials on construction
reading, would enable the system to operate fully automatically site using proximity techniques, Proceedings of the Construction Research
in all cases. Congress, ASCE, April 5–7, San Diego, US, 2005, pp. 552–557.
Analysis of the location information shows that approxi- [9] W.J. Rasdorf, M.J. Herbert, Automated identification systems—focus on
bar coding, Journal of Computing in Civil Engineering 4 (3) (1990)
mately 60% of the location information was within the accept-
279–296.
able limits. The low success rate was attributed to inaccuracies [10] B. McCullouch, G.K. Lueprasert, 2D bar-code applications in construc-
of the inexpensive GPS that was used in this test. No per- tion, Journal of Construction Engineering and Management 120 (4) (1994)
formance reduction was observed due to the presence of metal 739–752.
E. Ergen et al. / Automation in Construction 16 (2007) 354–367 367

[11] E. Finch, R. Flanagan, L. Marsh, Auto-ID application in construction, and Robotics in Construction, 19th (ISARC), NIST, Gaithersburg,
Construction Management & Economics 14 (1996) 121–129. Maryland, September 23–25 2002, pp. 155–160.
[12] L. Marsh, E. Finch, Attitudes towards auto-ID technologies within the UK [20] R. Navon, Y. Shpatnisky, Field experiments in automated monitoring of
construction industry, Construction Management & Economics 16 (1998) road construction, Journal of Construction Engineering and Management
383–388. 131 (4) (2005) 487–493.
[13] RFID Journal, Dockside Cranes Get Brains, http://www.rfidjournal.com,
last accessed June 2006, October 29, 2003. Esin Ergen is an Assistant Professor at the Civil Engineering Department at
[14] Sensor Magazine, Active RFID solution manages imports at the port of Istanbul Technical University. Her research interests are information flow in
Los Angeles, http://www.sensorsmag.com, last accessed June 2006, April construction supply chains and application of advanced tracking and data storage
18, 2006. technologies to improve the information flow process.
[15] E.J. Jaselskis, M.R. Anderson, C.T. Jahren, Y. Rodrigues, S. Njos, Radio-
frequency identification applications in construction industry, Journal of Burcu Akinci is Associate Professor at the Department of Civil and Environmental
Construction Engineering and Management 121 (2) (1995) 189–196. Engineering at Carnegie Mellon University. Her research area is information-rich
[16] J. Yagi, E. Arai, T. Arai, Parts and packets unification radio frequency history of a project and a facility supported by emerging sensing and data capture
identification application for construction, Automation in Construction 14 technologies. Her web address is bhttp://www.ce.cmu.edu/∼bakinci/N.
(2005) 477–490.
[17] Y. Aksoy, I. Chan, K. Guidry, J. Jones, C. Wood, Materials and Asset Rafael Sacks is a Senior Lecturer at Faculty of Civil and Environmental Engi-
Tracking Using RFID: a Preparatory Field Pilot Study, FIATECH Smart neering at Technion-Israel Institute of Technology. His research interests are
Chips Project Report, September 2004 http://www.fiatech.org/, last information technology enabled process change in civil engineering design and
accessed September 2005. construction, building information modeling and automated building systems. His
[18] A.A. Oloufa, I. Masaaki, O. Hiroshi, Situational awareness of construction web address is bhttp://www.technion.ac.il/∼cvsacks/N.
equipment using GPS, wireless and web technologies, Automation in
Construction 12 (6) (2003) 737–748.
[19] F. Peyret, R. Tasky, Asphalt quality parameters tracability using electronic
tags and GPS, Proceedings of International Symposium on Automation

Potrebbero piacerti anche