Automation in Construction 59 (2015) 24–37

Contents lists available at ScienceDirect

Automation in Construction

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

A BIM-based automated site layout planning framework for congested

construction sites
Srinath S. Kumar, Jack C.P. Cheng ⁎
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: Site layout planning is often performed on construction sites to find the best arrangement of temporary facilities
Received 24 January 2015 so that transportation distances of on-site personnel and equipment are minimized. It could be achieved by cre-
Received in revised form 12 May 2015 ating dynamic layout models, which capture the changing requirements of construction sites. However, formu-
Accepted 13 July 2015
lating such models is extremely tedious because it requires much manual data input and changes to design and
Available online 27 August 2015
construction plans are manually updated by layout planners. This study presents an automated framework of cre-
Keywords:
ating dynamic site layout models by utilizing information from BIM. The A* algorithm is used in conjunction with
Actual travel paths genetic algorithms to develop an optimization framework that considers the actual travel paths of on-site person-
Building information modelling (BIM) nel and equipment. To address the space limitation on site, our model optimizes the dimensions of facilities and
Construction site layout planning (CSLP) also considers interior storage within buildings under construction. A case example is demonstrated to validate
Congested sites this framework and shows a 13.5% reduction in total travel distance compared with conventional methods.
Facility dimensions optimization © 2015 Elsevier B.V. All rights reserved.
Interior storage

1. Introduction models in generating optimum layout plans because they allow layout
planners to cater to the changing site requirements and facilitate site
Construction site layout planning (CSLP) is a crucial step in construc- space to be reused. Currently dynamic models are created specific to a
tion planning that has been proven to reduce material handling costs project, based on the following information – (1) the number and
while improving safety and productivity of a project [1–3]. Construction types of facilities required, (2) the dimensions of each facility, and
projects require a large number of temporary facilities such as material (3) the specific time interval for which each facility would be required
storage areas, fabrication shops, etc. in order to support various con- on the construction site [19]. In most CSLP tools, such information has
struction activities. Traditionally these facilities are set up on unoccu- to be determined by the layout planner and manually entered into the
pied areas, within the boundaries of the construction site. In such software program. However manually determining this information
situations the goal of CSLP is to determine the best arrangement of tem- could be quite laborious, especially for projects with complex schedules
porary facilities such that the travel distances of construction personnel spanning several days. Changes to the design or construction plans
is minimized [4,5]. An obvious solution could be to set up temporary fa- would have to be continuously updated into the site layout models,
cilities on the free areas surrounding the building under construction. resulting in an inefficient workflow that is very time consuming. This
However, this is possible only on construction sites which have ade- severely limits the practicality of current CSLP tools and is one of the
quate amounts of free area to facilitate such an arrangement. In most reasons for their failure to achieve widespread adoption by the con-
urban construction projects, site space is limited and must be used judi- struction industry. There is a need for a practical and generic tool,
ciously in order to avoid problems with accessibility, safety and conges- which not only reduces unnecessary work by the layout planner but
tion. Comprehensive site layout planning can ensure a smooth flow of can also be easily adapted for use on different projects. Several research
materials, equipment, and labour, thereby improving the safety and ef- studies have attempted to improve the ease of use of dynamic CSLP
ficiency of on-site operations. tools. Tommelein et al. [2] developed a dynamic layout tool called
Site layout models fall into two categories – (1) static layout models, MovePlan with a graphical user interface, which took activity relation-
which assume that all of the facilities are assembled at the start and ships as input and generated optimized site layouts. Xing et al. [20] de-
exist for the entire duration of construction [1,4–10], and (2) dynamic veloped a GIS- based construction site material layout evaluation tool
layout models, which consider the actual duration for which facilities which took the resource loaded schedule as input to calculate the mate-
are required [11–18]. Dynamic layout models are far superior to static rial accessibility grade on a construction site. Said and El-Rayes [21] de-
veloped a construction logistics optimization system, which automated
⁎ Corresponding author. Tel.: +852 2358 8186; fax: +852 2358 1534. the retrieval of spatial and temporal data from BIM models and con-
E-mail addresses: [email protected] (S.S. Kumar), [email protected] (J.C.P. Cheng). struction schedules. In this study, we further improve on the practicality

http://dx.doi.org/10.1016/j.autcon.2015.07.008
0926-5805/© 2015 Elsevier B.V. All rights reserved.
 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 25

of current tools, by presenting a BIM based framework that automates [15] proposed a congested construction logistics planning (C2LP)
the creation of mathematical models for dynamic CSLP. BIM models model that generates optimal material logistics and site layout plans.
are rich sources of information and have been used to facilitate site lay- The C2LP model requires input parameters such as the site exterior
out planning [21–26]. The focus of this paper is to leverage information and interior spatial data, dimensions of temporary facilities, their rela-
from BIM models and construction schedules, to estimate the size, tionship with activities on the construction schedule and material
dimensions and number of temporary facilities required during differ- assignment to activities, based on which it optimizes the storage loca-
ent stages of construction. Since this methodology is pivoted on BIM, de- tions in exterior and interior building spaces. In a following study, Said
sign and construction changes can be automatically integrated into the and El-Rayes [21] developed an automated multi-objective construction
mathematical models, significantly reducing redundant work by layout logistics optimization system (AMCLOS), which uses information in BIM
planners. models and schedules to optimize the utilization of interior storage
In almost all of the studies on site layout planning, the optimization spaces in a building. The AMCLOS system uses IFC (Industry Foundation
goal is to determine temporary facility layouts that would minimize on- Classes) files to extract the geometry of interior and exterior site re-
site transportation costs, without compromising the safety or accessibil- gions, thereby automating the computation of available storage space.
ity of the site. [1–9,11–20,24,27–34]. A common formula used to However, the permissible periods of interior storage areas of the mate-
achieve this is rials and dimensions of each temporary facility have to be manually
specified in the AMCLOS system. In our study, we leverage information
Xn−1 Xn from BIM models to develop an automated method for interior and
Min i¼1 d f
j¼iþ1 i j i j
ð1Þ
exterior storage optimization. At any particular instant of time, the
amount of interior storage space is dependent upon the number of com-
where dij and fij represent the distance and frequency of trips between pleted floors, the geometry of the building and the presence of floor-
two facilities i and j, respectively, while n represents the total number level construction activities. By linking material and spatial data from
of facilities. For the sake of simplicity, most early studies on CSLP ap- BIM models to activity data from the schedule, we are able to automate
proximated dij by using linear distances such as the Euclidean (straight the computation of available interior storage space during different
line) or Manhattan (rectangular) distance. However, due to the pres- stages of the project. Our framework also automates the computation
ence of obstacles it is nearly impossible to always follow straight line of required storage amounts for each material, and optimally assigns
paths, due to which the Euclidean and Manhattan distances would be them to different storage locations. Therefore, this study presents an au-
significantly different from the actual travel distances of site personnel. tomated framework for CSLP, which addresses the requirements of
Yahya and Saka [18] introduced the concept of obstruction distance, congested construction sites by utilizing interior building spaces to
which was added to the computed Euclidean path in order to approxi- store materials. Our framework, which relies on BIM, enables us to de-
mate the actual travel distances. Park et al. [30] demonstrated the ben- velop a CSLP model that is generic enough to be useful in a variety of
efits of using actual travel distances instead of linear distances in solving cases. The number of inputs from layout planners is minimized since
the floor-level material layout problem for an indoor environment. In most of the computations are performed on information available
our study, we use the A* algorithm to accurately compute the actual from the BIM model and construction schedule.
travel distances between facilities on a construction site, and use them Our framework for automated CSLP using BIM consists of three mod-
as a basis for site layout optimization. Our method also considers varia- ules (see Fig. 1). In the first module, BIM based facility size estimation is
tions in path widths between construction personnel and machines, used to accurately compute the required size and dimensions for each
thereby resulting in a more accurate representation of on-site transpor- facility. In the second module, we present a methodology framework
tation activities. Another drawback among all of the previous studies on to automate the creation of dynamic layout models. The third module
CSLP is that the dimensions of facilities are taken as input parameters, deals with formulating the objective function and using an actual travel
prior to performing the optimization. As a result, only the position and path driven optimization to generate facility layouts. A demonstrative
orientation of each facility are considered as the decision variables for example is considered to highlight the benefits of this new approach.
optimization. As will be demonstrated in this paper, the previous ap-
proach severely limits the range of possible solutions. In this study, we 2. BIM-based facility size estimation
consider the position, orientation and dimensions of each facility as de-
cision variables, which are then optimized using Genetic Algorithms The CSLP problem consists of optimizing the locations for temporary
(GA). As a result, our CSLP tool determines the optimal dimensions of facilities, which may be approximated as one or more rectangles, on the
each facility, significantly improving the efficiency of generated layouts. unoccupied areas of a construction site. Most CSLP tools require the user
To facilitate the use of GA, a modified crossover and mutation operator to specify what the dimensions of each facility are, prior to performing
has been developed in this study. the layout optimization. However, this approach has a significant draw-
A characteristic of urban construction projects is the lack of on-site back, which leads to an under-utilization of site space. Facilities are used
storage space. To make up for this shortage of space, layout planners for storing materials and equipment, or to provide a working area for
may assign storage facilities to be set up inside the buildings under con- humans. It is thus essential that the facilities be large enough to satisfy
struction [31]. Such an approach increases the total area for setting up of these requirements. For example, it could be specified by the layout
facilities and consequently reduces on-site congestion. However, the in- planner, that a site office with 10 engineers should provide a floor
terior regions of a building are active workspaces for a number of floor- area of 100 m2. However, its exact dimensions can only be decided
level construction activities [35]. This imposes a limitation on the after taking into consideration the dimensions of the available space
amount of space that can be used for interior storage. As a result, interior which it is to be set up on. The facility in this case can take the form of
storage plans must be planned and coordinated carefully, to ensure a variety of rectangles with different values of length and width. The
maximum utilization of the available space. Park et al. [30] developed only constraints imposed are that the facility should provide the neces-
a system framework to optimize the interior storage locations of con- sary floor area while being of reasonable dimensions. Hence, specifying
struction materials on every floor of the building under construction. a fixed dimension for a facility prior to assigning its location on the site
However, their study was limited to optimizing storage locations in in- severely restricts the range of possible solutions. On construction sites
terior spaces only and did not address the storage needs in exterior re- with limited spaces, it is thus essential to optimize the dimensions of
gions of the construction site. Elbeltagi et al. [13] developed a dynamic each facility according to the dimensions of the available site locations.
CSLP tool, which used the constructed space of a building to store tem- In this study, we consider their dimensions as variables, which are opti-
porary facilities, with a view to reduce congestion. Said and El-Rayes mized using Genetic Algorithms.
26 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

Fig. 1. The proposed BIM-based construction site layout planning framework.

Another obvious drawback of having user defined facility sizes is provide adequate storage area for materials while protecting them
that it makes the process extremely tedious, since the user has to calcu- from damage due to sun, wind or rain. Processing facilities are used to
late the required size for several facilities. Changes to the design or con- process the raw materials on a construction site. Examples of these are
struction schedules may require a majority of these calculations to be fabrication shops, rebar cutting shops, steel bending yards, etc. Since
revised, resulting in increased work for the layout planner. To simplify processing facilities usually house humans and machines, it is essential
and improve the practicality of this process, we propose a methodology that they provide enough space to perform the required activities in a
which uses information from BIM to determine facility sizes, as shown safe and efficient manner. Residence facilities provide shelter and
in Fig. 2. In the first step, we use the BIM based facility size estimation cater to the requirements of personnel involved in the construction
module to determine the area required by each facility. The areas are project. Common examples of residence facilities include site offices,
then used as a basis to decide their optimized dimensions. pantries, toilets, etc. Residence facilities should provide adequate
space for humans to work efficiently. Thus, the size of a storage facility
2.1. Facility area estimation is governed by the quantity of material to be stored in it, whereas the
size of processing and residence facilities depends on the number of ma-
The primary role of temporary facilities is to support construction chines and humans, respectively.
operations. These facilities can be classified into three categories, Assuming a fixed order quantity inventory cycle, the maximum in-
(1) storage facilities, (2) processing facilities, and (3) residence facilities. ventory level is related to the rate of consumption of that material and
Storage facilities are used to temporarily store materials that will be the time between successive deliveries as given by
used for construction. Examples of these are steel storage yards, precast
panel storage yards, cement warehouses, etc. Storage facilities should Q ¼R  T ð2Þ
 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 27

Fig. 2. The proposed facility size estimation module.

where Q is the maximum inventory level, R refers to the peak material would provide us the approximate time needed to complete that activ-
demand or rate of consumption of that resource, and T is the time be- ity. To facilitate this, the schedule was exported as a csv file, which could
tween successive deliveries. Thus, we can calculate the required size be read by our program.
of a storage facility if we know the rate of consumption and time be- Step 3: BIM-Schedule linkage. In this step, the activities on the
tween successive deliveries, using the equation: schedule are linked with their corresponding components in the BIM
model. For example the activity “Level 1: Columns 001: Pour Concrete”
As ¼ C  R  T ð3Þ is linked to the quantity of concrete used in “column 001”, resulting in a
4D BIM model. In our study, we use a program to parse each activity
where, As refers to the area of a storage facility, C is a constant which is name on the schedule and link it with the corresponding element in
an indicator of the amount of space needed to store a unit quantity of the BIM model. Each activity in the schedule should therefore follow a
that material, R refers to the peak rate of consumption of the material naming convention, so that the activity can be correctly identified by
stored in that facility, and T refers to the time between deliveries. the parser program. It is also crucial that the granularity of the schedule
In the case of processing facilities, the number of machines used is matches with the level of development (LOD) in the BIM model. For our
dependent on the quantity of material that needs to be processed each study, the architectural, structural and MEP elements are modelled at
day. Considering the case of a rebar bending shop, the number of LOD 300.
machines housed in it will be determined by the quantity of rebar that Step 4: Rate of consumption calculation. By dividing the quantity of
needs to be processed each day and the capacity of each machine. materials consumed by the duration of consumption, we thus obtain the
Hence the size of a processing facility can be determined from the rate rate of consumption of different materials during the different stages of
of consumption of the processed material, given by: construction. A limitation of this approach is that it assumes a linear rate
of consumption in order to simplify the calculations.
Ap ¼ K  R ð4Þ Step 5: Facility size estimation. Using the relationships defined pre-
viously, the size of each facility is thus determined. This information is
where Ap refers to the size of a processing facility, K is a constant which then parsed to the facility dimensions computation module. The only
depends on the material being processed, and R refers to the rate of con- input from the user is the time interval between successive deliveries.
sumption of the dependent resource.
Since BIM models do not usually contain information about the 2.2. Permissible facility dimensions
number of site personnel, the size of residence facilities has to be
manually inputted by the user. Based on this logic, the required size of Once we have the floor area requirement of a facility, we must deter-
a facility could be computed using the following five steps, as shown mine its dimensions. The dimensions should be selected such that it can
in Fig. 2. A drawback of the proposed facility size estimation framework accommodate even the largest object to be stored in it. For example, if
is the assumption that materials will be delivered to the construction the standard length of rebar is 8m, the rebar storage yard must have ei-
site at fixed intervals. ther a length or width greater than 8m. Thus, the dimensions of a facility
Step 1: Material quantities extraction module. The quantities of var- should be selected keeping in mind the nature of objects to be stored in
ious materials used in different components of the building are calculat- it as well as the floor area requirement. Based on this logic, we devel-
ed, from the BIM model. For example, in the component “Column 001” oped a set of mathematical relationships which represent the possible
the quantities to be determined would be (1) volume of concrete and dimensions of a facility.
(2) amount of rebar. This was performed by extracting the material Consider a rectangular facility with an area A, length L and width W.
quantities of individual components from the BIM model. Autodesk Let Lmin and Wmin represent its minimum allowable length and width,
Revit, which provides an Application Programming Interface (API), respectively. Lmin and Wmin are determined from the size of the objects
was used as the modelling software. Using the Revit API, we developed that will be contained in the facility.
a program in Microsoft C# to extract and tabulate the necessary quanti-
ties. This framework also allows us to compute derived quantities such L ¼ α þ Lmin ; where α ≥0; ð5Þ
as the amount of formwork, scaffolding, etc. which may be missing from
in BIM model.
Step 2: Activity duration computation module. In this module, the W ¼ β þ W min ; where β ≥0; ð6Þ
durations of different activities are determined from the planned con-
struction schedule. For example, the scheduled duration of an activity
“Level 1: Columns 001 – 005” is read from the planned schedule and it ðLmin þ α Þ  ðW min þ βÞ ¼ A; ð7Þ
28 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

Rearrange the terms to get, In this study, we developed an automated method of creating a dy-
namic site layout model from the construction schedule of the project.
A−Lmin W min −αW min Activities from the construction schedule were mapped to the corre-
β¼ ð8Þ sponding facilities that they would require. For example, the activity
Lmin þ α
“Floor Slab level 1” is linked with the rebar storage and processing
yards, which are facilities required to store materials used in the slabs.
For each value of α we obtain a corresponding value of β. This gives First, a set of relations is created to determine the facilities that would
us a range of permissible values for length and width of each facility. be required to complete each activity. Next, a corresponding facility re-
Unlike previous studies which specify fixed dimensions for facilities, quirement schedule is generated from the construction schedule by
we represent them as a range of permissible values, and based on the using the previously defined relationships. The underlying logic of this
characteristics of the available site space, determine what their opti- mapping is that facilities will be required as long as activities requiring
mized dimensions should be. As will be demonstrated in Section 5, the them are underway. Generally construction materials are delivered to
permissible range of values are used as the basis for determining the op- the site a couple of days prior to their requirement, to account for deliv-
timized dimensions of each facility ery delays. This information must be accounted for while generating the
facility requirement schedule. Thus, each bar on the facility requirement
3. Dynamic layout planning schedule indicates the total duration for which a facility will occupy
space on the site. Since most commercial scheduling software allow
A construction project commences in various construction stages or the schedule to be exported as a csv file, we were able to write a pro-
phases. Foundation, framing, MEP and finishing works are common ex- gram to determine the facility requirement schedule from the activity
amples of different phases. As a result, the requirement of facilities will schedule (see Fig. 3).
also differ between phases. Facilities required in one phase may not be The next step of dynamic layout planning is to identify the space
required in the next phase. In such cases, unnecessary facilities may be availability on the site. The available spaces on a site can be separated
dismantled after use and the space previously occupied by them could into – (1) interior and (2) exterior storage areas. In most construction
be used to set up other facilities. In certain cases the use of a facility projects, the exterior storage area remains fixed throughout the con-
may be altered to avoid de-mobilization and mobilization costs. On con- struction project and can hence be treated as a constant. Interior spaces,
struction sites with limited space availability, multiple facilities may oc- however, are dynamic in nature and vary as the construction progresses.
cupy the same location on the site at different stages of construction. To In the following section, we describe in detail our methodology for auto-
model these changing facility requirements, dynamic layout models matically computing the available interior storage area in a building.
should be created. Dynamic layouts refer to a sequence of site layouts
that are spread over distinct time intervals and when taken together 3.1. Interior storage processing module
represent the entire construction project [11]. In most approaches, the
construction project is split into discrete phases and partial layouts for The interior storage processing module is designed to compute the
each phase are determined [12]. Since some facilities may be required interior space availability for material storage. This is performed in
in multiple phases, the partial layouts will be interdependent. three main steps.

Fig. 3. Dynamic CSLP model based on the construction schedule.

 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 29

Step 1: Calculation of feasible dates for interior storage. In order material should be stored on the same floor as its activity workspace.
to utilize the interior building spaces for material storage, it is This is mathematically represented as:
important to identify the beginning date when any given floor
becomes suitable for storing materials. In most building construction XN X
T

projects, this is given by the date of slab completion. Thus, we assign Maximize FUEi;t ð11Þ
i¼1 t¼0
the start date of interior storage on a given floor as the date of slab
completion.
where N is the number of floors and T is the number of time steps. Due
Step 2: Computation of available storage space. In the United States
to the presence of floor-level activities and safety regulations, only a
of America, Occupational Safety and Health Administration (OSHA) [36]
limited amount of interior floor space is available for storage and should
rules specify that materials stored inside buildings under construction
hence be used judiciously. The optimization of interior storage matches
shall not be placed within 6 feet of any hoist-way or inside floor open-
the availability of storage space with the storage requirements of mate-
ings, nor within 10 feet of an exterior wall which does not extend
rials. Hence, the optimization has a single objective function subject to
above the top of the material stored. The gross permissible storage
multiple constraints. It is important to note that because of the similar-
area for every floor can thus be determined from the floor plan.
ities between each floor, most building schedules are planned in floor
However, owing to the presence of floor level activities, only a frac-
construction cycles. That is to say, the scheduling of labour, resources
tion of this space can be used for material storage. It is also necessary to
and time is very similar between different storeys and hence the con-
provide sufficient amount of free space on every floor to facilitate the
struction of a multi-storey building can be regarded as the repetitive
unrestricted movement of materials and personnel. Finally, a reduction
construction of a single storey. As a consequence, the interior storage
in the total available space due to floor partitioning must also be con-
schedules will follow the same repetitive process. The complexity of
sidered. The net available area for material storage on a floor at any
the optimization problem therefore becomes 2N where N is the number
given point in time is thus computed by deducting the clear space and
of storage facilities.
workspaces of indoor activities from the gross storage area. A further
reduction in available space is made after the completion of floor
3.2. Exterior storage processing module
partitioning. Mathematically, this is represented as:

X The remaining facilities are assigned to the exterior regions of the

Ai;t ¼ Gi;t − C i;t − FOi;t − FP i;t ð9Þ site. In order to facilitate the creation of a dynamic layout model, the re-
maining activities are grouped into discrete phases. Andayesh and
where, Ai,t is the net space available, Gi,t is the gross floor area, Cit is the Sadeghpour [11] argued that dynamic layout models, which split the
clear area required, FOi,t is the floor space occupied by an activity and FPit project into fewer phases, suffer from under-utilization of available
is the reduction in floor space after floor partitioning for the floor i, at space. However, splitting the construction operation into too many dif-
time t. Said and El. Rayes [21] proposed a method of extracting the ferent phases would result in the generation of several partial layout
gross floor area from BIM models represented in the IFC format. In our plans, which could be difficult to implement in practice. Keeping both
study, we use classes provided by the Revit API to calculate the gross these conflicting factors in mind, we chose a system that splits the pro-
floor area, whereas the clear area and the floor space occupied by activ- ject into three discrete phases. However, the number of phases can be
ities are determined based on prior experience and interviews with changed according to the project complexity.
practitioners. There is scope for further improvement in the calculation This splitting of the project into phases was done to promote re-
of these values. utilization of space. Facilities which are no longer needed should give
The construction schedule which lists all the activities, their start way to new facilities. Hence, we use the logic that a significant change
dates and duration is exported as a csv file in order to be parsed by in the facility space demand should indicate the beginning of a new
phase. Fig. 3 shows the relationship between the construction schedule
our program. The start and end dates of each activity are identified
and using Eq. (9), we calculate the net available storage area for each and the start and end date of each phase.
floor.
Step 3: Interior storage assignment. Since each activity occurring on 4. Formulation of the optimization problem
a floor will require a corresponding material storage area, it is more ef-
ficient to store the required materials on the same floor as opposed to on 4.1. Objective function formulation
another floor. For example, it is advisable to store wallboards and metal
studs, both of which are needed for interior partitioning, on a floor for After determining the facility requirements and space availability,
we formulate the CSLP problem into an optimization problem. On a con-
the duration of the interior partitioning activity on that same floor.
The facility requirement schedule defined in Section 3 allows us to com- struction site, personnel travel from one facility to another either to per-
form certain activities or to transfer materials. Each such trip is assumed
pute the amount of space required for facilities during different con-
struction stages. Based on this, we seek to optimally allocate storage to incur a cost, which is directly proportional to the distance travelled
and also depends on the mode of transportation. The frequency of
facilities to interior regions of the building. To optimize the usage of in-
terior storage, we define a term called floor space utilization efficiency these movements is stored in a trip frequency matrix, whereas the
cost of transportation per unit length is stored in the cost matrix. The
(FUE) as follows:
cost per unit length, which represents the ease of transportation, de-
Space assigned to facilities pends on the mode of transport and the nature of the materials being
FUEi;t ¼ ð10Þ transported. The total transportation cost (TTC) is mathematically rep-
Ai;t
resented as:

where Ai,t is the net space available on the floor i at time t. The floor XN Xn−1 Xn
space utilization efficiency is an indicator of how efficiently the net TTC ¼ p¼1 i¼1
d E f
j¼iþ1 i jp i jp i jp
ð12Þ
available interior space is used for storage. A value of 1 indicates that
the space is utilized to its maximum efficiency, whereas a lower value where N refers to the number of construction phases, n refers to the
signifies under-utilization. The problem now is that of selecting mate- total number of facilities, and dijp, Eijp and fijp refer to the actual travel
rials to be stored inside, such that the FUE is maximized. Material distance, cost per unit length and frequency of transportation between
assignment should, however, comply with the constraints that each facilities i and j during phase p, respectively. The value of dijp is
30 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

dependent upon the position of each facility and the presence of

obstacles on the construction site (See Section 4.2). Eijp and fijp depend
on the mode of transportation and number of trips that have to make,
respectively.
Facilities that are not required on the construction site anymore are
either dismantled, relocated or re-used for storing other materials. As a
result, each facility has an associated assembling, dismantling and
relocation cost, which would depend on the labour and equipment
required.

XN Xn  
FC ¼ p¼1 i¼1
Acip þ Dcip þ Rcip ð13Þ

where FC refers to the total cost incurred in assembling, dismantling or

moving facilities. Acip, Dcip and Rcip are the costs associated with assem-
bling, dismantling and relocating facility i in phase p, respectively.
The total cost incurred due to on-site material storage and transpor-
tation (TC) can thus be represented as

Fig. 4. Actual travel path using the A* algorithm.

TC ¼ TTC þ FC ð14Þ

4.3. Constraints of the facility layout optimization problem

In another words, the optimization objective is to minimize TC, sub-
ject to various constraints (see Section 4.3). Layouts corresponding to 4.3.1. Available site space
the lowest value of the objective function represent optimal solutions Osman et al. [10] proposed a method of extracting the spatial infor-
to the site layout problem. mation of a construction site from CAD files. BIM encompasses all of the
functionalities of CAD, hence by extracting the site layout plans of the
4.2. Computation of actual travel paths construction project, we were able to automatically generate the site
spatial data. This was done by extracting data contained in the layout
A drawback of most studies is that they used linear distances to plans into an excel database and formulating them into a set of mathe-
model the travel paths of site personnel. In reality however, due to the matical inequalities. The closed area bounded by these inequalities rep-
presence of obstacles, it is nearly impossible to always follow linear resented the total site area available for construction. The fixed facilities
paths. In this study, we use the actual travel paths that would be follow- (buildings) were then identified in a similar manner and subtracted
ed instead, as a basis for site layout optimization. It is human nature to from the total area encompassed by the site. The remaining area repre-
take the path of least effort while travelling from one place to another. sents the total space allocated for setting up of temporary facilities.
On a construction site, this is manifested in site personnel following
the shortest possible path between two facilities. The actual travel
path is therefore assumed to be the shortest path from one facility to an-
other, considering the presence of site obstacles [37–39].
To calculate the actual travel paths, the site is converted into a grid
with a spacing of one metre in the X and Y directions. Obstacles such
as trees, boundaries, temporary and permanent facilities are marked
on this grid. A number of heuristics could be used to calculate the
shortest paths between two points on such a grid. Soltani et al. [39]
compared the accuracy and time complexity of Dijkstra’s algorithm, A*
algorithm and Genetic Algorithms in finding the shortest route on con-
struction sites. Andayesh and Sadeghpour [37] conducted a study to
compare direct, grid based and visibility graph approaches in finding
the actual distance between objects on a construction site. Lin et al.
[40] proposed an IFC-based method of finding the shortest path be-
tween two points in a 3D indoor space. In our study, we use the A* algo-
rithm to determine the shortest possible path between two locations,
and use it as a basis for site layout optimization (see Fig. 4). Although
the accuracy of computed paths can be improved by selecting a smaller
grid spacing, the time taken to calculate the path would significantly
increase [30,38,39]. To ensure that accessibility has not been compro-
mised, it is important to consider the mode of transportation used on
site. Light materials may be carried by hand or wheelbarrow whereas
heavier materials may need a forklift. The clear width of the path need-
ed by a forklift will differ from that required by a wheelbarrow or a
human being. As a result, the shortest path available on site might differ
for each mode of transportation. Accordingly we specify a clear width of
1m for labourers, 3m for small hauling machines, and 5m for larger
hauling machines. This approach ensures that the generated site layouts
cater to the use machines in on-site transportation and do not restrict
their accessibility. Fig. 5. Encoding the position of a facility as well as its dimension into a gene.
 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 31

Fig. 6. The set of permissible dimensions for each facility (set D).

4.3.2. Safety constraint tower crane and TR represents the tower crane’s pickup radius. The
In our study, we represent the safety constraint with the following tower crane’s pickup radius is inversely related to the weight of object
inequality: being lifted. This entails that heavier objects must be located within
closer proximity to the tower crane than lighter materials. In our sys-
ei j ≥si j ð15Þ tem, the radius of tower crane pertaining to different lifted materials
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi can be specified by the user. In the demonstrative example, we ensured
 2  2 that structural steel storage yards and metal panel storage yards were
ei j ¼ xi −x j þ yi −y j ð16Þ
within the reachable radius of the tower crane. Currently our system as-
sumes a predefined location for setting up a tower crane. The system
where, eij refers to the Euclidean distance between two facilities and sij is will be extended later to consider different crane types, locations and
the minimum safe distance between facilities i and j, with centroids at operations so as to address this limitation.
coordinates (xi, yi) and (xj, yj) respectively. A large value of sij implies a
safety concern in positioning facilities i and j within close proximity of
one another. 4.3.5. Site accessibility constraints
In this study, we defined a fixed path for material delivery trucks to
4.3.3. Overlapping constraint ensure easy access to the construction site. In order to facilitate smooth
We use a mathematical inequality to ensure that facilities do not site operations, it is important that facilities should not obstruct these
overlap with one another or with the building to be constructed. A facil- paths. For the safety of the site personnel, it is also necessary to ensure
ity is represented by two parameters, (1) the coordinates of one corner that work areas are not too close to these paths. In our study, we en-
point, and (2) its length and width. The constraint to avoid overlapping sured that all facilities had a minimum clearance of 1m from this path.
between facilities thus becomes: This ensured that the paths were never obstructed by facilities while
h minimizing the safety hazard.
    i
max x j −xi −li x j −xi þ l j ; y j −yi −wi y j −yi þ wi ≥0 ð17Þ

4.3.6. Miscellaneous constraints

where (x, y) represent the coordinates of the top left corner, l represents
Based on the layout engineer’s discretion or environmental con-
the length and w represents the width of the facility.
cerns, a range of project specific constraints could be added to the
paper. For example, in some cases the construction site may share one
4.3.4. Tower crane reachability constraint
or more boundaries with schools, hospitals or other environmentally
Certain material storage yards need to be located within the reach-
sensitive zones. In such situations, it must be ensured that the sound
able radius of the tower crane. This is mathematically represented as:
generated due to on site activities are within acceptable limits. Thus, fa-
 2  2 cilities that create a lot of noise would have to be placed at a sufficient
xt −x f þ yt −y f ≤T R ð18Þ distance from these zones. In certain situations from prior experience,
the layout engineer might require a certain relationship in the locations
where (xt, yt) represents the axis of rotation of the tower crane, (xf, yf) of two facilities. Such conditions can be formulated into mathematical
represents the corner of the facility which is farthest away from the relationships and incorporated into the optimization model.

Fig. 7. Crossover and mutation operators.

32 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

Fig. 8. Revit model of the 12-storey office building in the demonstrative example and the developed Revit Plugin.

5. Optimization using genetic algorithms getting stuck on a local minima, crossover and mutation operations
are developed.
The CSLP problem is considered to be ‘NP-hard’ and several research
studies have attempted to arrive at solutions using heuristics [13,14,17, 5.1. Gene encoding
32,34] or mathematical optimization techniques [28]. Genetic
algorithms (GA), due to its ease of implementation, is one of the most In GA, the variables to be optimized are represented as genes. Tradi-
common methods of solving the site layout problem. The essence of tional CSLP approaches optimize only the locations of facilities on the
GA lies in combining elements from two solutions of the same genera- construction site. Hence, in traditional practice only the positional infor-
tion (parents) or mutating individual solutions to produce a third mation of each facility is encoded into a gene. In our study, we seek to
solution which is then evaluated based on the objective function [8]. If optimize the dimensions of each facility in addition to its positions. As
the third solution is better than its parents, it is kept; otherwise, it is a result, the dimensions of each facility are also encoded into a gene
discarded. This process is repeated until there is no significant dif- along with their positions (see Fig. 5). Subsequently modified crossover
ference in the fitness of resulting chromosomes, which represents and mutation operators need to be developed in order to generate var-
convergence on an optimal solution. To prevent the algorithm from iations in facility dimensions as well.

Fig. 9. Construction schedule and corresponding phases in the demonstrative example.

 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 33

Fig. 10. Utilization of interior storage areas in the demonstrative example.

5.2. Modified crossover and mutation operators parent genes, Gene A and Gene B which represent two possible facility
layouts for this problem. The positional information of each gene, repre-
The crossover operator combines the genes of two parents to create sented by the corner points of the facility are (xA, yA) and (xB, yB), respec-
an offspring. In all of the previous studies which used GA to optimize tively. In addition to positional information, Gene A and Gene B also
site layouts, the crossover operator combined only the positional infor- contain information about the facility’s length and width (lA, wA) and
mation of facilities. In our study, we develop a modified crossover oper- (lB, wB), respectively. By definition, (lA, wA) and (lB, wB) belong to set
ator which combines facility dimension information as well. The D, which is the range of permissible facility dimensions. The two parent
crossover operation for a facility’s positional information has been ex- genes are to be crossed and would result in an offspring, represented by
plained numerous times in previous studies and hence we skip the ex- Gene C. We generate Gene C in such a way that it contains information
planation. However, a crossover for facility dimensions has not been from both Gene A and Gene B. As a result, the dimensions of the facility
attempted before. This is explained with the help of the following stored in Gene C lie somewhere in between those of Gene A and Gene
example. B and by definition also belong to set D. Fig. 7 shows the modified
Consider the trivial case in which we just have one temporary facility crossover and mutation operators for a construction site comprising of
of area Ai on a construction site. The range of permissible values for n facilities. This ensures that genes in the offspring Gene C, are a combi-
length and width for this facility are calculated using Eq. (8), and repre- nation of its parents. A random function is used to decide how much ge-
sented by set D = {(l1, w1), (l2, w2) … (ln, wn)}, (see Fig. 6). Consider two netic material is inherited from each parent. The positional information

Table 1
Facilities assigned to the exterior site regions.

Facility number Name Type Area (m2) Phase required Size calculation

1 Tool shed Storage 55 1,2,3 User defined

2 Site office Residence 70 1,2,3 User defined
3 Restrooms Residence 35 1,2,3 User defined
4 Structural steel primary beams Storage 40 1 BIM and Schedule
5 Structural steel secondary beams Storage 65 1 BIM and Schedule
6 Structural steel columns Storage 80 1 BIM and Schedule
7 Structural steel assembly yard Processing 75 1,2 BIM and Schedule
8 Scaffolding pile Storage 40 1 BIM and Schedule
9 Fire proofing materials Storage 50 1 BIM and Schedule
10 Floor metal panels Storage 85 1,2 BIM and Schedule
11 Rebar storage Storage 70 2 BIM and Schedule
12 Formwork storage Processing 50 2 BIM and Schedule
13 Aggregate warehouse Storage 60 2 BIM and Schedule
14 Cement warehouse Storage 35 2 BIM and Schedule
15 Concrete mixer Processing 65 2 BIM and Schedule
16 Plumbing storage Storage 50 3 BIM and Schedule
17 Curtain wall glass Storage 90 3 BIM and Schedule
18 Curtain wall framing storage Storage 70 3 BIM and Schedule
19 Curtain wall assembly yard Processing 75 3 BIM and Schedule
20 Floor tiles storage Storage 65 3 BIM and Schedule
21 Electrical systems storage Storage 75 3 BIM and Schedule
34 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

Table 2
Facilities stored inside the building.

Facility number Name Type Area (m2) Size calculation

22 Mechanical storage Storage 80 BIM and Schedule

23 Interior wallboard and metal studs Storage 100 BIM and Schedule
24 Fixtures storage Storage 90 BIM and Schedule
25 False ceiling panels storage Storage 80 BIM and Schedule
26 Fire safety systems storage Storage 65 BIM and Schedule

in Gene C, is obtained using a simple crossover operator, which com- The modified mutation operator works similar to the crossover op-
bines the positional information of the two parent genes, with a random erator. In this function, genes are randomly picked and their facility di-
function determining the amount of genetic information inherited from mensions are given a random change such that the new dimensions
each parent. would also belong to set D. The crossover and mutation operations try

Fig. 11. 4D visualization of the interior storage space.

Fig. 12. Comparison of layouts obtained from the two cases.

 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 35

Table 3
Comparison of the dimensions of facilities.

Facility number Name Area (m2) Facility dimensions (L,W)(rounded to the nearest metre)

Conventional method Proposed method

(user defined dimensions) (optimized dimensions)

1 Tool shed 55 7,8 5,11

2 Site office 70 10,7 9,8
3 Restrooms 35 9,4 4,9
4 Structural steel primary beams 40 5,8 10,4
5 Structural steel secondary beams 55 8,7 6,9
6 Structural steel columns 80 8,10 7,11
7 Structural steel assembly yard 75 8,9 7,11
8 Scaffolding pile 40 5,8 6,7
9 Fire proofing materials 50 7,7 10,5
10 Floor metal panels 85 8,11 7,12
11 Rebar storage 70 9,8 9,8
12 Formwork storage 50 6,8 10,5
13 Aggregate warehouse 60 7,8 6,10
14 Cement warehouse 35 7,5 9,4
15 Concrete mixer 65 7,9 6,11
16 Plumbing storage 50 6,9 9,6
17 Curtain wall glass 90 7,13 7,13
18 Curtain wall framing storage 70 9,8 9,8
19 Curtain wall assembly yard 80 10,8 7,11
20 Floor tiles storage 75 7,11 10,8
21 Electrical systems storage 75 8,10 7,11

to account for the variations in genetic information between successive would be assigned to exterior storage locations (Table 1) and interior
generations of offspring. storage locations (Table 2) were determined. Since the number of facil-
ities to be considered for interior storage was 13, an optimized solution
6. Demonstrative example could be found using an exhaustive search. Setting up the MEP, tiles,
ceiling and wall board storage inside the built environment was found
We tested the BIM based CSLP framework on an illustrative example to represent the most optimal allocation of interior space.
of a building construction project. The project involves construction of a Once the dynamic model of the site was created, we move over to
12-story steel building with concrete floor slabs on a site area with some optimizing the layouts of facilities on the exterior regions of the site.
existing trees and an access road. Autodesk Revit was used to create the The optimization was performed using genetic algorithms with a popu-
BIM model, which contained the material information of the building lation size of 1500.
(see Fig. 8). A schedule for the construction activities was created
using Microsoft Project (see Fig. 9). 6.1. Results and discussions
The construction site was converted into a square grid of side 80
units, where 1 unit represented 1 m on the site. The building to be In order to demonstrate the benefit of our framework, two cases
constructed, existing trees, existing access road and the site boundary were considered: (1) Euclidean distance based optimization with
were marked on this grid. To facilitate the export of material quantities, fixed facility dimensions (conventional method) and (2) actual travel
we created a plugin named “BIM based CSLP”, using the Revit API and distance driven optimization with variable facility dimensions (the
Microsoft C# (Fig. 8). The schedule was exported from Microsoft Project proposed method) (see Fig. 12). Table 3 compares the dimensions of
as a csv file, which could be taken as input by the Revit plugin. Using the facilities in the two cases. Tables 4 and 5 compare the Euclidean and
previously defined methodology, we were then able to determine the actual travel distances corresponding to the two cases. In order to
area requirement of each temporary facility. The size of the site office compare the two layouts, we calculated what the total distance trav-
was decided to be 100 m2 based on the assumption that 13 engineers elled in each case would be. The conventional method, which assumes
would be needed for the project. The size of the tool shed and toilet, Euclidean or centre-to-centre distances, yielded a total Euclidean dis-
based on similar assumptions, were defined to be 55 m2 and 35 m2, re- tance of 2390 km for the entire project. However, the actual travel dis-
spectively. The frequency of trips made by workers between facilities tance for this case was computed as 3520 km. This suggests that the
was defined with logical assumptions, based on the quantity of work presence of site obstacles lead to a significant under-estimation of travel
each day. For the sake of simplicity we assumed that each path started distances while using a linear distance driven optimization. Our pro-
and ended on the centroid of a rectangular facility. posed method of using actual travel distances and variable facility
Using the schedule file and the information about floor plans, we de- dimensions yielded a total travel distance of 3050 km, a reduction of ap-
veloped an indoor storage utilization schedule (Fig. 10). The schedule proximately 13.5 % in the actual travel distance. This large difference can
tells us how much indoor storage space is available during different be attributed to two causes. Firstly, the conventional algorithm of using
stages of construction. Based on this information the facilities that centre-to-centre distances in the optimization is not intelligent enough

Table 4 Table 5
Comparison of the Euclidean distance in layouts obtained using Euclidean and actual trav- Comparison of the actual travel distance in layouts obtained using Euclidean and actual
el path driven optimization. travel path driven optimization.

Euclidean distance (km) Actual distance travelled (km)

Layout obtained taking dij as Phase 1 Phase 2 Phase 3 Total Layout obtained taking dij as Phase 1 Phase 2 Phase 3 Total

Euclidean 900 670 820 2390 Euclidean 1330 960 1230 3520
Actual path 990 740 980 2710 Actual path 1100 810 1140 3050
36 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37

Table 6 computing the actual travel paths followed on a construction site. Fur-
Comparison of the total transportation cost in the conventional and proposed methods. thermore, this paper presents a method to determine optimized dimen-
Total Transportation Cost ($) sions for each facility, thus allowing for an increase in the efficiency of
Phase 1 Phase 2 Phase 3 Total
layouts. By giving consideration to interior storage within the building
under construction, our model becomes suitable to deal with CSLP on
Conventional method 15,960 11,520 14,740 42,240
congested urban building projects as well. In the demonstrative exam-
Proposed method 13,200 9720 13,680 36,600
ple, our optimization method lead to layouts which reduced the overall
travel distance by 13.5 percent over conventional methods. Our CSLP
to identify the presence of obstructions to the travel paths. As a result, engine can be integrated with graphical visualization tools to facilitate
they fail to accurately represent the on-site transportation activities. 3D interactive site layout planning. This model has the potential to be
The A* algorithm on the other hand identifies obstacles on the site and extended to address supply chain planning as well. Future efforts will
is better equipped to model the actual travel patterns. Modelling the focus on developing a system that incorporates real time construction
path widths of personnel and machines further improves the accuracy schedules in order to track inventories and manage material logistics.
of the computed travel paths. The second reason attributable to the sig-
nificant difference in travel distances is the optimization of facility
dimensions. The conventional method specified fixed dimensions for References
each facility prior to the optimization and as a result, the range of solu- [1] A. Hamiani, C. Popescu, CONSITE: a knowledge-based expert system for site layout, J.
tions was severely restricted. The proposed approach optimizes facility Comput. Civil Eng. Microcomput. Supercomput. ASCE (1988) 248–256.
dimensions based on the dimensions of the available site spaces. By [2] I. Tommelein, P. Zouein, Interactive dynamic layout planning, J. Constr. Eng. Manag.
119 (2) (1993) 266–287.
varying the length and breadth of a facility, they could even be allocated [3] M. Cheng, J. O'Connor, ArcSite: Enhanced GIS for construction site layout, J. Constr.
to constricted spaces. These findings strongly suggest using actual travel Eng. Manag. 122 (4) (1996) 329–336.
distance driven optimization in favour of linear distances. They also sug- [4] H. Li, P.E. Love, Site-level facilities layout using genetic algorithms, J. Comput. Civ.
Eng. 12 (4) (1998) 227–231.
gest that a method that optimizes facility dimensions should be used in [5] I.-C. Yeh, Construction-site layout using annealed neural network, J. Comput. Civ.
favour of the conventional method. Tables 6 and 7 compare the total Eng. 9 (3) (1995) 201–208.
transportation cost and total cost (including facility setup, dismantling [6] E. Elbeltagi, T. Hegazy, A.H. Hosny, A. Eldosouky, Schedule-dependent evolution of
site layout planning, Construct. Manag. Econ. 19 (7) (2001) 689–697.
and relocation cost).
[7] A. Khalafallah, K. El-Rayes, Automated multi-objective optimization system for air-
As can be seen in Fig. 10, the net available storage area undergoes a port site layouts, Autom. Constr. 20 (4) (2011) 313–320.
significant variation with time. The interior storage optimization mod- [8] M.J. Mawdesley, S.H. Al-Jibouri, H. Yang, Genetic algorithms for construction site lay-
out in project planning, J. Constr. Eng. Manag. 128 (5) (2002) 418–426.
ule recommended using interior storage for MEP, tiles, ceiling panels
[9] I. Tommelein, R. Levitt, B. Hayes-Roth, SightPlan model for site layout, J. Constr. Eng.
and wallboards and to store the pipes required for the safety sprinklers. Manag. 118 (4) (1992) 749–766.
The average of the floor space utilization efficiency (FUE) achieved by [10] H.M. Osman, M.E. Georgy, M.E. Ibrahim, A hybrid CAD-based construction site layout
this combination was equal to 0.65, indicating that on average, 65% of planning system using genetic algorithms, Autom. Constr. 12 (6) (2003) 749–764.
[11] M. Andayesh, F. Sadeghpour, Dynamic site layout planning through minimization of
the net available interior space was utilized for storage. Such a low total potential energy, Autom. Constr. 31 (2013) 92–102.
value is attributable to the lack of coordination between material pro- [12] M. Andayesh, F. Sadeghpour, The time dimension in site layout planning, Autom.
curement and interior storage feasibility. Planning material logistics Constr. 44 (2014) 129–139.
[13] E. Elbeltagi, T. Hegazy, A. Eldosouky, Dynamic layout of construction temporary fa-
based on the feasibility of interior storage could improve the FUE. The cilities considering safety, J. Constr. Eng. Manag. 130 (4) (2004) 534–541.
optimized site layout and interior storage assignments can be used as [14] X. Ning, K.-C. Lam, M.C.-K. Lam, Dynamic construction site layout planning using
a basis to generate 4D BIM models of the project (see Fig. 11). Such max-min ant system, Autom. Constr. 19 (1) (2010) 55–65.
[15] H. Said, K. El-Rayes, Optimal utilization of interior building spaces for material pro-
models can then further be used for a variety of analyses, such as clash curement and storage in congested construction sites, Autom. Constr. 31 (2013)
detection and construction sequence coordination. 292–306.
[16] H.M. Sanad, M.A. Ammar, M.E. Ibrahim, Optimal construction site layout considering
safety and environmental aspects, J. Constr. Eng. Manag. 134 (7) (2008) 536–544.
7. Conclusions and future work [17] J. Xu, Z. Li, Multi-objective dynamic construction site layout planning in fuzzy ran-
dom environment, Autom. Constr. 27 (2012) 155–169.
This paper presents a BIM based framework that enables automating [18] M. Yahya, M. Saka, Construction site layout planning using multi-objective artificial
bee colony algorithm with Levy flights, Autom. Constr. 38 (2014) 14–29.
the construction site layout planning (CSLP) process. Based on this, we
[19] P. Zouein, I. Tommelein, Dynamic layout planning using a hybrid incremental solu-
developed a tool to optimize the dynamic layouts of temporary facilities tion method, J. Constr. Eng. Manag. 125 (6) (1999) 400–408.
on a construction site. Our framework allows quick and easy facility [20] X. Su, H. Cai, J. Pan, A. Kandil, H.M. Said, GIS-based dynamic construction site material
sizing and does not require users to manually input project specific in- layout evaluation for building renovation projects, Autom. Constr. 27 (2012) 40–49.
[21] H. Said, K. El-Rayes, Automated multi-objective construction logistics optimization
formation. Since all the calculations are based on information from system, Autom. Constr. 43 (2014) 110–122.
BIM, changes to the design and schedule could be updated at the click [22] B. Choi, H.-S. Lee, M. Park, Y.K. Cho, H. Kim, Framework for work-space planning
of a button, enabling them to be reflected in the final calculations. This using four-dimensional BIM in construction projects, J. Constr. Eng. Manag. 140
(9) (2014).
achieves a significant reduction in effort when coping with frequent [23] E.P. Karan, J. Irizarry, Extending BIM interoperability to preconstruction operations
changes. Our model could be easily incorporated into existing CSLP using geospatial analyses and semantic web services, Autom. Constr. 53 (2015) 1–12.
tools to improve their practicality. The effort spent by layout planners [24] L.-C. Lien, M.-Y. Cheng, Particle bee algorithm for tower crane layout with material
quantity supply and demand optimization, Autom. Constr. 45 (2014) 25–32.
in performing unnecessary calculations would thus be significantly re- [25] S. Zhang, K. Sulankivi, M. Kiviniemi, I. Romo, C.M. Eastman, J. Teizer, BIM-based fall
duced, allowing them to focus on decision making instead. This paper hazard identification and prevention in construction safety planning, Saf. Sci. 72
also develops a more realistic representation of site-level activities, by (2015) 31–45.
[26] Y. Zhou, L. Ding, X. Wang, M. Truijens, H. Luo, Applicability of 4D modeling for re-
source allocation in mega liquefied natural gas plant construction, Autom. Constr.
Table 7 50 (2015) 50–63.
Comparison of the total cost (including facility setup, dismantling and relocation) in the [27] K. Chau, M. Anson, J. Zhang, Four-dimensional visualization of construction schedul-
conventional and proposed methods. ing and site utilization, J. Constr. Eng. Manag. 130 (4) (2004) 598–606.
[28] S.M. Easa, K.M.A. Hossain, New mathematical optimization model for construction
Total Cost ($) site layout, J. Constr. Eng. Manag. 134 (8) (2008) 653–662.
[29] Z. Ma, Q. Shen, J. Zhang, Application of 4D for dynamic site layout and management
Phase 1 Phase 2 Phase 3 Total of construction projects, Autom. Constr. 14 (3) (2005) 369–381.
Conventional method 25,960 22.520 27,740 76,240 [30] M. Park, Y. Yang, H.-S. Lee, S. Han, S.-h. Ji, Floor-level construction material layout
planning model considering actual travel path, J. Constr. Eng. Manag. 138 (7)
Proposed method 23,200 20,720 26,680 70,600
(2011) 905–915.
 S.S. Kumar, J.C.P. Cheng / Automation in Construction 59 (2015) 24–37 37

[31] D.R. Riley, V.E. Sanvido, Patterns of construction-space use in multistory buildings, J. from http://www.osha.gov/pls/oshaweb/owadisp.show_document?p_table=
Constr. Eng. Manag. 121 (4) (1995) 464–473. STANDARDS&p_id=10685.
[32] C. Tam, T.K. Tong, A.W. Leung, G.W. Chiu, Site layout planning using nonstructural [37] M. Andayesh, F. Sadeghpour, A comparative study of different approaches for find-
fuzzy decision support system, J. Constr. Eng. Manag. 128 (3) (2002) 220–231. ing the shortest path on construction sites, Procedia Eng. 85 (2014) 33–41.
[33] H. Wang, J. Zhang, K. Chau, M. Anson, 4D dynamic management for construction [38] A. Soltani, T. Fernando, A fuzzy based multi-objective path planning of construction
planning and resource utilization, Autom. Constr. 13 (5) (2004) 575–589. sites, Autom. Constr. 13 (6) (2004) 717–734.
[34] J. Zhang, L. Liu, R. Coble, Hybrid intelligence utilization for construction site layout, [39] A.R. Soltani, H. Tawfik, J.Y. Goulermas, T. Fernando, Path planning in construction
Autom. Constr. 11 (5) (2002) 511–519. sites: performance evaluation of the Dijkstra, A∗, and GA search algorithms, Adv.
[35] B. Akinci, M. Fischer, J. Kunz, Automated generation of work spaces required by con- Eng. Inform. 16 (4) (2002) 291–303.
struction activities, J. Constr. Eng. Manag. 128 (4) (2002) 306–315. [40] Y.-H. Lin, Y.-S. Liu, G. Gao, X.-G. Han, C.-Y. Lai, M. Gu, The IFC-based path planning for
[36] Occupational Health and Safety Administration (OSHA), OSHA Guidelines for Storage 3D indoor spaces, Adv. Eng. Inform. 27 (2) (2013) 189–205.
on Construction Sites - General Requirements for Storage (No. 1926.250). Retrieved