Critical Sequence Crashing Heuristic for

Resource-Constrained Discrete Time–Cost

Trade-Off Problem
Rifat Sonmez 1; Mahdi Abbasi Iranagh 2; and Furkan Uysal 3
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Despite the importance of project deadlines and resource constraints in construction scheduling, very little success has been
achieved in solving the resource-constrained discrete time–cost trade-off problem (RCDTCTP), especially for large-scale projects. In this
paper a new heuristic method is designed and developed to achieve fast and high-quality solutions for the large-scale RCDTCTP. The pro-
posed method is based on the novel principles to enable effective exploration of the search space through adequate selection of the activities to
be crashed for a resource constrained schedule, by only crashing the activities with zero float in a resource constrained-schedule, which form
the critical sequence. The computational experiment results reveal that the new critical sequence crashing heuristic outperforms the state-
of-the-art methods, both in terms of the solution quality concerning project cost and computation time. Solutions with a deviation of 0.25%
from the best known solutions are achieved within seconds for the first time, for a large-scale project including up to 2,000 activities. The
main contribution of the new heuristic to practitioners and researchers is that it provides a fast and effective method for optimal scheduling of
real-life-size construction projects with project deadlines and resource constraints. DOI: 10.1061/(ASCE)CO.1943-7862.0001077. © 2015
American Society of Civil Engineers.
Author keywords: Project management; Resource management; Construction planning; Scheduling; Costs; Optimization; Cost and
schedule.

Introduction
The critical path method (CPM) is commonly used for scheduling
of construction projects. However, the critical path method is not
capable of optimal scheduling of projects when there are resource
constraints or project deadlines. Hence, extensive research efforts
have focused on the resource-constrained project scheduling prob-
lem (RCPSP) and the time-cost trade-off problem. The general
RCPSP aims to achieve the minimum project duration that satisfies
both the precedence and resource constraints, whereas the time-cost
trade-off problem involves minimizing the total direct and indirect
costs without exceeding the project deadline. Because in practice
many resources (e.g., crews, equipment) are available in discrete
units, much research has focused on the discrete version of the
time-cost trade-off problem called the discrete time–cost trade-
off problem (DTCTP).
RCPSP and DTCTP are both crucial for planning and manage-
ment of construction projects because there are resource constraints
and project deadlines in the majority of projects. However, popular
commercial project management software has very limited capabil-
ities for solving the RCPSP (Mellentien and Trautmann 2001;
Hekimoglu 2007; Lu et al. 2008; Bettemir and Sonmez 2014)
1
Associate Professor, Dept. of Civil Engineering, Middle East Technical
Univ., Ankara 06531, Turkey (corresponding author). E-mail:
rsonmez@metu.edu.tr
2
Ph.D. Candidate, Dept. of Civil Engineering, Middle East Technical
Univ., Ankara 06531, Turkey.
3
Expert, Ministry of Development, Necatibey Caddesi No: 110-A,
Kızılay, Ankara 06570, Turkey.
Note. This manuscript was submitted on March 11, 2015; approved on
August 26, 2015; published online on October 23, 2015. Discussion period
open until March 23, 2016; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Construction En-
gineering and Management, © ASCE, ISSN 0733-9364.
and does not provide any options for the time-cost trade-off
problem (Menesi et al. 2013).
Few studies have focused on the resource-constrained time-cost
trade-off problem, which combines the time-cost trade-off problem
with the RCPSP. The majority of the resource-constrained discrete
time–cost trade-off problem (RCDTCTP) studies used problem
instances, including up to 50 activities in computational experi-
ments. Hegazy and Menesi (2012) reported the performance of
a heuristic method for 360 activities. Menesi et al. (2013) used
large-size instances, including up to 2,000 activities in computa-
tional experiments.
Despite a large amount of the research being concentrated on
designing heuristics and metaheuristics for the RCPSP and DTCTP,
very few of the proposed methods can be applied to real-life con-
struction projects, which typically include more than 300 activities
(Liberatore et al. 2001). Besides, a few methods that are capable of
solving large-scale problems usually require a significant amount
of computation time to achieve high-quality solutions. The parallel
genetic algorithm (GA) of Kandil and El-Rayes (2006) required
136.5 h on a single processor, and 19.7 h over a cluster of 20 pro-
cessors to obtain the Pareto front for a DTCTP including 720
activities. The metaheuristics of Bettemir (2009) were able to
achieve a 2% deviation from the optimal in 73 min for DTCTP
instances including 630 activities. The heuristic of Hegazy and
Menesi (2012) required 32 min for a RCDTCTP including 360 ac-
tivities (Menesi et al. 2013). The constraint programming model
of Menesi et al. (2013) achieved a solution with 6.39% deviation
from the upper bound (best known solution) in 120 min. Hence,
for the time-cost trade-off problem there is a significant gap be-
tween the literature and the needs of real-life construction project
management.
The main objective of this research is to design and develop a
heuristic that can achieve high-quality solutions in a short amount
of computation time for the large-scale RCDTCTP. The research

© ASCE 04015090-1 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 attempts to provide a fast method for optimal scheduling of real-
life-size projects with project deadlines and resource constraints.
Hence, the paper aims to contribute to both researchers and profes-
sionals by narrowing the gap between the literature and the needs of
the practice.
Literature Review
RCPSP and DTCTP both belong to the nondeterministic
polynomial–time hard (NP-hard) (Blazewicz et al. 1983; De et al.
1997) computational complexity class, and exact methods can
solve these problems only for small- to medium-size networks.
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
Hence numerous heuristic and metaheuristic methods were pro-
posed for optimal scheduling of projects under resource constraints
or project deadlines. Priority rule-based scheduling heuristics
(Özdamar and Ulusoy 1994; Hegazy et al. 2000; Tormos and Lova
2001) and metaheuristics, including genetic algorithms (Chan et al.
1996; Hartmann 1998; Hegazy 1999; Kim and Ellis 2008; Chen
and Shahandashti 2009; Kim and Ellis 2010; Sonmez and Uysal
2014), simulated annealing (Lee and Kim 1996; Bouleimen and
Lecocq 2003; Valls et al. 2005), tabu search (Deblaere et al.
2011), and particle swarm optimization (Lu et al. 2008; Wang
and Qi 2009; Chen 2011), are among the methods proposed for
the RCPSP.
The methods proposed for the DTCTP include Siemens
approximation method (Siemens 1971), genetic algorithms (Feng
et al. 1997; Hegazy 1999; Zheng et al. 2005; Kandil and El-Rayes
2006; Fallah-Mehdipour et al. 2012; Sonmez and Bettemir 2012),
ant colony optimization (Ng and Zhang 2008; Xiong and Kuang
2008; Afshar et al. 2009), particle swarm optimization (Yang
2007; Bettemir 2009; Fallah-Mehdipour et al. 2012), shuffled frog
leaping (Elbeltagi et al. 2007), and tabu search (Vanhoucke and
Debels 2007).
In an early attempt to integrate resource-constrained scheduling
with the time-cost trade-off problem, Chua et al. (1997) proposed a
GA-based model. Leu and Yang (1999) presented a multicriteria
genetic algorithm for the RCDTCTP. Ahn and Erenguc (1998)
developed a multipass heuristic procedure for the resource-
constrained time-cost trade-off problem. Chen and Weng (2009)
adopted a GA-based time-cost trade-off analysis for considering
resource-constrained scheduling along with time-cost trade-off.
Wuliang and Chengen (2009) developed a GA for the RCDTCTP.
Hegazy and Menesi (2012) presented a heuristic method that
crashes the lowest-cost critical activities that are determined by
the critical path method and resolves any resource overallocation
by imposing start-delay values to the activities to meet both project
deadlines and resource limits. In a recent study, Menesi et al. (2013)
proposed a constraint programming model for the RCDTCTP and
implemented the model for large-scale projects including up to
2,000 activities. The constraint programming approach provides
a flexible method for the RCDTCTP and allows various practical
constraints such as different types of precedence relations.
Resource-Constrained Discrete Time–Cost Trade-Off
Problem
The objective of the resource-constrained time-cost trade-off prob-
lem is to determine a time, cost, or resource mode (option) and a
start date for each activity in such a way that the precedence and
resource constraints are satisfied, and the total direct costs, indirect
costs, and delay penalties (liquidated damages) are minimized.
In the discrete version of this problem the relation between the
duration of activities and the resources committed is discrete.
© ASCE 04015090-2
J. Constr. Eng. Manage., 04015090
Different versions of the resource-constrained time-cost trade-
off problem have been studied in the literature. Chua et al.
(1997) considered exceeding the resource constraints at an addi-
tional cost for optimizing the resource-constrained time-cost
trade-off problem. Ahn and Erenguc (1998) minimized the sum
of direct costs and the penalty costs for the resource-constrained
project-scheduling problem in which the duration reduction (crash-
ing) can be performed. Hegazy and Menesi (2012) and Menesi et al.
(2013) focused on minimizing the sum of direct and indirect costs
and the penalties and incentives.
Few studies aimed to achieve the complete nondominated set of
the time, cost, or resource modes and the start dates over the set of
feasible project durations called the Pareto front, while considering
the resource constraints. Leu and Yang (1999) obtained the nondo-
minated solutions that minimized the sum of direct and indirect
costs for the RCDTCTP. Chen and Weng (2009) also focused
on Pareto front optimization for the RCDTCTP and considered ac-
tivity interruption. Wuliang and Chengen (2009) presented a multi-
mode resource-constrained discrete time–cost trade-off model to
achieve the Pareto front for the RCDTCTP.
Critical Sequence
In the critical path method, the project duration is calculated by
adding the durations of the activities on the longest path in the
project network called the critical path, which is determined by
the precedence relations. When there are resource constraints, the
critical path method is not sufficient to identify the sequence of
activities with zero total float so-called critical sequence (Wiest
1964) or critical chain (Goldratt 1997). Wiest (1964) presented
a procedure for calculation of floats in an early study to define
and identify the critical sequence. Lu and Li (2003) proposed
the resource-activity critical-path method to calculate the floats
and to determine the sequence of critical activities for resource-
constrained scheduling. Lu et al. (2008) developed a simplified
simulation-based scheduling system to provide valid floats and
optimum schedules for the RCPSP.
The significance of critical sequence in resource-constrained
scheduling (RCS) is similar to the importance of critical path in
the critical path method. The precedence and resource feasible
project duration can be shortened by crashing the activities that
are on the critical sequence(s). Unlike the critical path method,
in resource-constrained project scheduling it is sometimes possible
to shorten the project duration by crashing the activities that are not
on the critical sequence (Wiest 1964). However, an efficient heu-
ristic method can be designed for the RCDTCTP by only consid-
ering crashing of the activities that are on the critical sequence(s),
which is the main focus of this research.
Critical Sequence Crashing Heuristic
A novel heuristic method that is based on crashing of the critical
sequence is designed and developed especially for the large-scale
RCDTCTP. The heuristic method consists of two parts: backward-
forward resource-constrained scheduling, and critical sequence
crashing.
Backward-Forward Resource-Constrained Scheduling
The critical sequence crashing heuristic (CSCH) starts the search
by using the normal (uncrashed) modes for the activities. Once
the modes are selected, the start dates of the activities and project
completion can be determined by using a resource-constrained
J. Constr. Eng. Manage.
 project scheduling method. The resource-constrained project
scheduling method used to determine the start dates of activities
has a very significant impact on the project duration and the critical
sequence(s) (Herroelen and Leus 2005). The backward-forward
resource-constrained scheduling method is integrated into the pro-
posed critical sequence heuristic to achieve an adequate and fast
solution for the RCPSP.
The backward-forward scheduling method was proposed by Li
and Willis (1992) to improve a feasible resource-constrained sched-
ule by increasing the resource utilization. Lova and Tormos (2002)
developed a heuristic using the backward-forward scheduling
method for a resource-constrained multiproject scheduling prob-
lem, and showed that backward-forward scheduling improved
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
the multiproject duration. In a recent study, the backward-forward
scheduling method integrated hybrid genetic algorithm has outper-
formed the state-of-the-art methods for a resource-constrained
multiproject scheduling problem (Sonmez and Uysal 2014).
The backward-forward scheduling method performs resource-
constrained scheduling twice by using the serial scheduling scheme
(Kelley 1963). The serial scheduling scheme sequentially schedules
the activities (one by one) at their earliest precedence and resource
feasible start time, according to a priority list. In backward-forward
scheduling, first backward scheduling is executed in the reverse
time direction, then forward scheduling is performed. An arbitrary
project completion time is selected to start backward scheduling
because the exact duration of the feasible schedule is not known
at the beginning. The resulting backward schedule is adjusted such
that the project completion start is equal to time instant zero.
In the backward scheduling phase of the proposed CSCH, total
floats of activities that are calculated by the critical path method are
used to determine the priority list. The activity with the smallest
CPM float is backward scheduled first, and in case of a tie the ac-
tivity with the larger activity number is selected. The CPM total
float priority enables the activities on the critical path to be re-
source-constrained scheduled first, and usually works well when
the resource constraints are not tight. The forward scheduling is
performed in the order of start times that are obtained in backward
scheduling, and in case of a tie the activity with the larger activity
number is selected.
Critical Sequence Crashing
The critical sequence(s) is identified for the schedule determined in
backward-forward scheduling to start crashing. In the proposed
heuristic method, the critical sequence is defined as the sequence
of activities that has zero total float for a precedence and resource
feasible schedule. The critical sequence(s) is identified through the
following steps (Wiest 1964):
1. The activities of the schedule obtained by forward improvement
(forward schedule) are started as late as possible (right-shifted)
one by one, in descending order of their finish times in the for-
ward schedule (early finish times), without changing the project
completion time to calculate the late start and late finish times.
2. If several activities have the same early finish time, the initial
late start times for these set of activities are calculated by right-
shifting the activities so that they start as late as possible
(assuming no other activity in the set has been shifted); then
the activities are right shifted in descending order of their initial
late start times, and their late start and late finish times are
calculated.
3. In case of a tie in the initial late start times, the activities are
right-shifted in ascending order of their resource requirements.
4. If the tie is not broken, the tied activities are right-shifted in as-
cending order of their activity numbers.
© ASCE 04015090-3
J. Constr. Eng. Manage., 04015090
After the late start times of each activity are determined by right-
shifting, RCS total floats are calculated by subtracting the late start
times of the activities from their early start times (their start times in
the forward schedule). The critical sequence(s) is identified by
tracking the sequence(s) of the activities with zero RCS total float.
Removal of local suboptimalities is not performed to identify the
critical sequence in a short amount of computation time.
The crashing is performed only to the activities that have zero
RCS total float. Among the activities that are on the critical seqeu-
ence(s), the activity with the least daily crashing cost is crashed
first, considering one activity crashing option at a time. In case
of a tie, the activity with the least resource impact (least crashing
resource difference) is selected. If the tie is not broken, the activity
with the larger activity number is selected as the third criterion.
Backward-forward resource-constrained scheduling is per-
formed to determine the project duration once the activity to be
crashed is determined. The project duration obtained by the latest
mode selections is compared with the project duration obtained by
the previous mode selections (in the first cycle previous mode se-
lections includes the normal modes). Crashing is not executed and
the selected activity is not crashed further if the project duration of
the latest mode selections is larger than the project duration of the
previous mode selections. Finally, the crashing and backward-
forward scheduling stages are executed until all of the activities
in the latest critical sequence(s) are considered for crashing, and
the solution with the minimum cost is reported. The flow chart
of the critical sequence crashing heuristic method is illustrated
in Fig. 1.
Case Example
A case example is presented in Fig. 2 and Table 1 to illustrate the
proposed CSCH. The modes in Table 1 are the time, cost, and re-
source options of the activities for different execution alternatives.
For example, for Activity 1 the first mode (M-1) has a duration
of 6 days, a resource requirement of 12 men, and a direct cost
of $17,000. The deadline for the case example is 36 days, and
the indirect costs and the liquidated damages are $2,500=day
and $5,000=day, respectively. The backward-forward resource-
constrained scheduling is initiated by selecting the normal modes
for all of the activities. The critical path method is performed to
determine the CPM floats of the activities as shown in Fig. 3.
The backward scheduling priority list is determined as <9, 8, 7,
3, 5, 2, 4, 1, 6 > based on the CPM total floats that are calculated
according to the critical path method and by selecting the activity
with the larger activity number in case of a tie. The backward
scheduling is performed according to the priority list by scheduling
the activities in the reverse time direction (one by one) at their latest
precedence and resource feasible finish time, using an arbitrary
project completion time of 50 days. The resulting backward sched-
ule is adjusted such that the project start time is equal to Day 0, as
shown in Fig. 4, and the project duration is obtained as 47 days.
A priority list of <2, 6, 1, 3, 4, 7, 5, 8, 9 > is obtained for the
forward scheduling phase using the start times of the activities in
the backward schedule of Fig. 4. The forward scheduling is also
performed by using the serial scheduling scheme, according to
the priority list obtained from the backward scheduling phase.
The objective of the forward scheduling phase is to improve the
schedule obtained in the backward scheduling phase. The project
duration of the resulting schedule (Schedule 1), has decreased to
40 days, at the end of forward scheduling improvement as shown
in Fig. 5.
In Schedule 1, none of the activities can be right-shifted (started
at a later start time without changing the project completion time),
J. Constr. Eng. Manage.
 Select the normal modes for
the activities

Perform critical path method

scheduling and determine
CPM total floats
Crash the critical sequence
activity with the least daily
crashing cost Implement backward
resource-constrained
scheduling based on CPM
total float priority
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

Identify the critical sequence

Perform forward
improvement based on start
dates obtained in backward
scheduling
Calculate RCS total floats

Right shift the forward Is the schedule the first

schedule to determine RCS Yes schedule with normal
late start times modes?

No

Does the crashing Do not execute the latest

increase the project Yes crashing and do not crash
duration? the activity further

No
No

Are all lof the latest

critical sequence activities
considered for crashing?

Yes

Report the best solution

Fig. 1. Flowchart of critical sequence crashing heuristic

so all of the activities are identified to be on the critical sequence.
The crashing options for the activities that are on the critical se-
quence in Schedule 1 are summarized in Table 2. Activity 5 was
crashed first by changing the mode of this activity to Mode 1 (M-1)
because this activity had the least daily crashing cost. The critical
path method is performed to determine the CPM floats for the new
activity durations in which the mode of Activity 5 is changed to
M-1. The next backward scheduling priority list is determined
based on the revised CPM floats. Backward scheduling and forward
scheduling improvement are performed to obtain the next schedule
(Schedule 2) as shown in Fig. 6.
In Schedule 2 (Fig. 6), the activities are right-shifted one by one,
in descending order of their early finish times to identify the critical
sequence. Activity 9 has an early finish time of 40 days and is right-
shifted first, followed by Activity 8, which has an early finish time
of 36 days. Activity 9 cannot finish later than Day 40 and Activity 8
cannot finish later than Day 36 without changing the project com-
pletion time. Hence, for Activity 9 and Activity 8 the late finish

© ASCE 04015090-4 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 1
4
1. (6,12,$17000) 1. (4,6,$5200)
2. (8,10,$12500)

6 9
1. (5,13,$19600) 1. (2,6,$6900)
2 2. (8,10,$14500) 2. (4,2,$5800)

1. (7,5,$16500)

7
8
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

3 1. (13,12,$22500)
1. (3,4,$11500)
2. (5,2,$9000)
1. (6,13, $25500)
2. (8,9,$22500) NO
3. (10,6,$21000)
5
1. (Dur., Res., Cost)
1. (9,7,$19600) n. (Dur., Res., Cost)
2. (12,4,$18400)

Indirect costs=$2,500/Day
Liquidated damages=$5,000/Day
Deadline = 36 Days
Resource limit=20/Day

Fig. 2. Case example

times are equal to the early finish times. Activity 5 and Activity 7 RCS total floats of the activities are calculated by subtracting
have the same early finish time of 31 days. Activity 5 and Activity 7 their late finish times from their early finish times. In Schedule
are right-shifted to calculate the initial late start times for these ac- 2, Activity 2 and Activity 6 are not on the critical sequence.
tivities, assuming that no other activity in the set of activities with The crashing options for the activities that are on the critical
an early finish time of 31 days has been right-shifted. The initial sequence in Schedule 2 are given in Table 3. Hence, the activity
late start time of Activity 5 is calculated as 22, and the initial late selected for next crashing is Activity 9, and Activity 9 was crashed
start time of Activity 7 is calculated as 18. Hence, Activity 5 is by changing the mode of this activity to M-1. Backward scheduling
right-shifted first because it has a later initial late start time, and and forward scheduling improvement are performed to obtain the
the late finish time of Activity 5 is calculated as 31. Activity 7 duration of the next schedule (Schedule 3) as 38 days. In Schedule
is right-shifted next, and the late finish of Activity 7 is also calcu- 3, Activity 3 has the least daily crashing cost, and the mode of this
lated as 31. Activity 4, Activity 1, and Activity 6 are respectively activity is changed to M-2 to obtain Schedule 4, which has a du-
right-shifted next. The late finish times for Activity 4 and Activity 1 ration of 36 days. Similarly, in Schedule 4, the mode of Activity 8 is
are also equal to their early finish times. However, Activity 6 can be changed to M-1 to obtain Schedule 5 with a duration of 34 days.
right-shifted 3 days and can finish at Day 18 without changing the Next, Activity 3 is considered for crashing, however, crashing
project completion time. Finally Activity 3 and Activity 2 are right- Activity 3 to its first mode increased the duration to 39 days in
shifted 0 and 3 days, respectively. Schedule 6, hence this crashing is not executed and Activity 3
is not crashed further. Next, the mode of Activity 1 is changed to
M-1 to obtain Schedule 7 with a duration of 33 days. Finally, the
mode of Activity 6 is changed to M-1; however, crashing Activity 6
Table 1. Activity Data of the Case Example to its first mode in Schedule 8 increased the duration to 38 days,
Duration Resource Direct hence this crashing is not executed. Because at this point all of the
Activity Mode (days) requirement (people) cost ($) activities in the latest critical sequence are considered for crashing,
1 1 6 12 17,000 the heuristic is terminated and the minimum cost solution is re-
2 8 10 12,500 ported. The proposed critical sequence crashing heuristic was able
2 1 7 5 16,500 to achieve the minimum cost of $216,700 in Schedule 5 for the case
3 1 6 13 25,500 example. The time, cost, and resource modes and the start dates of
2 8 9 22,500 the minimum cost solution are shown in Fig. 7.
3 10 6 21,000
4 1 4 6 5,200
5 1 9 7 19,600 Input-Output Interface
2 12 4 18,400
6 1 5 13 19,600 An input-output interface was developed in Microsoft Excel 2013
2 8 10 14,500 to enable simplified data input and output and to facilitate data
7 1 13 12 22,500 exchange with the commercial project management software to en-
8 1 3 4 11,500 hance the use of the proposed critical sequence crashing heuristic in
2 5 2 9,000 practice. The input screen of the interface for the case example is
9 1 2 6 6,900 illustrated in Fig. 8. The heuristic requires a dummy start and a
2 4 2 5,800
dummy finish activity. The successor information and time, cost,

© ASCE 04015090-5 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 0 8 8 8 4 12 ES DUR EF
1 4 Act. No
11 11 19 19 11 23 LS TF LF

Critical Path

0 7 7 7 8 15
2 6
3 3 10 20 13 28
28 4 32
9
28 0 32
10 13 23
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

7
0 10 10 10 0 23 23 5 28
3 8
0 0 10 23 0 28

10 12 22
5
11 1 23

Fig. 3. Critical path method schedule

20
19
18
17
16
1-M2 4-M1
15
5-M2
14
Resource Usage

13
12
11
10
9
8
7
7-M1
6
6-M2
5
4
3-M3
3 2-M1
2
8-M2 9-M2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
Time (Days)

Fig. 4. Resource-constrained backward schedule

and resource modes, the project deadline, the daily indirect cost,
incentives, and the liquidated damages are entered in the input sheet
of the interface. Once the heuristic is executed, the time, cost, and
resource modes and the start dates of activities for the minimum
cost solution that satisfies the resource constraints can be obtained
in the output sheet as shown in Fig. 9. An executable version of
the critical sequence crashing heuristic and the input/output inter-
face can be obtained from http://www2.ce.metu.edu.tr/∼sonmez/
research/csch.
Computational Experiments
Computational experiments are conducted to evaluate the perfor-
mance of the proposed critical sequence crashing heuristic for
the RCDTCTP using benchmark instances. The proposed algo-
rithm is coded in C# and is built within Visual Studio 2013
for both 32- and 64-bit platforms. All of the tests were carried
out on two different configurations. The first configuration (C1)
consisted of a desktop computer with a P9X79 Chipset (ASUSTeK
Computer, Taiwan) motherboard, 16 GB 667 MHz DDR3 RAM,
Intel Core i7-3.40 GHz CPU (Intel Corporation, Santa Clara,
California), and 64-bit Windows 8.1 operating system. The second
configuration (C2) included a laptop computer with a GM45 Chip-
set (Intel Corporation, Santa Clara, California) motherboard, 4 GB
400 MHz DDR2 RAM, Intel Core 2 Duo P8600 2.40-GHz CPU
(Intel Corporation, Santa Clara, California), and 32-bit Windows
8.1 operating system. In the tests CSCH is executed solely (no other
application is ran simultaneously), and overclocking was not
performed.

© ASCE 04015090-6 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 20
19 Critical
18 sequence
17 activity
16 1-M2
15
14
3-M3
Resource Usage

13
7-M1
12
11
10
9
8
7
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

6
6-M2
5
4
4-M1
3 2-M1
5-M2
2
8-M2 9-M2
1

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Time (Days)

Fig. 5. Critical sequence for Schedule 1

Deviation from the upper bound (best known solution) is used to

evaluate the performance of the different methods along with the
Table 2. Crashing Options for Activities on Critical Sequence in
CPU time. The best known lowest-cost solutions for the small-test
Schedule 1
instances were obtained from the literature (Leu and Yang 1999;
Daily crashing Crashing resource Chen and Weng 2009; Hegazy and Menesi 2012; Menesi et al.
Activity Crashing option cost ($=day) difference 2013). Because the medium- and large-size benchmark RCDTCTP
1 M-2 to M-1 2,250 2 instances were created by copying the test instance of Chen and
3 M-3 to M-2 750 3 Weng (2009) in serial several times, the upper bounds for medium-
5 M-2 to M-1 400 3 and large-scale instances were calculated by multiplying the best
6 M-2 to M-1 1,700 3
known solution of the test instance in Chen and Weng (2009) with
8 M-2 to M-1 1,250 2
9 M-2 to M-1 550 4
the number of times that it was copied in serial. Deviation from the
upper bound (DUP) is calculated as follows:

20
19 Critical
18 sequence
17 activity
16 5-M1
6-M2 4-M1
15
14
Resource Usage

13
12
11
10
9
2-M1
8
7
7-M1
6
1-M2
5
4
3-M3
3
2
8-M2 9-M2
1

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Time (Days)

Fig. 6. Critical sequence for Schedule 2

© ASCE 04015090-7 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 Solution − Upper Bound Small-Scale Test Instances
DUP ¼ × 100 ð1Þ
Upper Bound The proposed heuristic is initially tested with the small-scale
RCDTCTP test instances. The first test instance included a project,
in which Solution = minimum cost solution obtained that satisfies
including nine activities with up to four modes and three resources
the resource constraints; and Upper Bound = best known solution
(Leu and Yang 1999). The problem is solved for the deadline of
for the problem.
64 days. CSCH obtained the best known solution of $7,400 in
0 s for both C1 and C2 as shown in Table 4.
Table 3. Crashing Options for Activities on Critical Sequence in The second small-scale RCDTCTP test instance, which is
Schedule 2 shown in Fig. 10, consisted of a project, including 10 activities
up to four modes and a single resource (Chen and Weng 2009).
Daily crashing Crashing resource
Activity Crashing option cost ($=day) difference The objective of the second problem was to determine the mini-
mum total cost solution for an indirect expense of $2,200=day,
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

1 M-2 to M-1 2,250 2 while considering a daily resource constraint of 30. The proposed
3 M-3 to M-2 750 3
heuristic achieved the best known solution of $244,000 in 0 s for
8 M-2 to M-1 1,250 2
9 M-2 to M-1 550 4
both C1 and C2. The best solution was also obtained by the genetic
algorithm of Chen and Weng (2009), which identified a Pareto front

20
19 Critical
18 sequence
17 activity
16 5-M1
6-M2 4-M1
15
14
13
Resource Usage

2-M1
12
11
10
9
8
7
7-M1
6
1-M2
5 3-M2
4
9-M1
3
8-M1
2
1

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
Time (Days)

Fig. 7. Minimum cost solution for the case example

Fig. 8. Input screen of the input-output interface

© ASCE 04015090-8 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 solution of $245,900 in 2 s, which had a 0.78% deviation from
the upper bound. The performances of the four methods for the
second problem are summarized in Table 5.

Medium- and Large-Scale Test Instances

Hegazy and Menesi (2012) and Menesi et al. (2013) created
medium- and large-scale test instances for the RCDTCTP by copy-
ing the test instance of Chen and Weng (2009) in serial several
times. The test instances included 100, 300, 1,000, and 2,000 ac-
tivities, and reflected the size of real-life construction projects.
Table 6 compares the performance of the proposed CSCH with
the performance of the heuristic developed by Hegazy and Menesi
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

(2012) and constraint programming model presented by Menesi

et al. (2013). CSCH achieved a DUP value of 0.22 and 0.24%
for 100 and 300 activity problems in 0 s for C1, and in 0 and
Fig. 9. Output screen of the input-output interface 1 s for C2. The heuristic of Hegazy and Menesi (2012) obtained
solutions with DUP values of 0.78% for both of the problems in 1
and 21 min. The constraint programming model (Menesi et al.
2013) achieved a DUP value of 2.34% in 15 s, and a DUP value
Table 4. Comparison of Results for Small-Size Project 1 of 0.32% in 10 min, for the problem including 100 activities. For
Source Method Solutiona DUP (%) CPU time the problem, including 300 activities, the model was able to obtain
a solution with a DUP value of 5.90% in 15 s, and a solution with a
Leu and Genetic algorithm $7,400 0.00 NA
Yang (1999) DUP value of 0.88% in 20 min. All of the experiments in Hegazy
Hegazy and Heuristic $7,400 0.00 2s and Menesi (2012) and Menesi et al. (2013) were performed on a
Menesi (2012) laptop with a 2.4-GHz CPU and 3 GB of RAM. The proposed heu-
Menesi Constraint $7,400 0.00 1s ristic method achieved better solutions than the heuristic of Hegazy
et al. (2013) programming (CP) and Menesi (2012) and the constraint programming model (Menesi
This study CSCH $7,400 0.00 0s et al. 2013) at a significantly lower computation time for medium-
Note: NA = not available. scale test instance, including 100 and 300 activities.
a
Solutions are for project deadline of 64 days. The performance of the CSCH was consistent for the large-scale
test instances as shown in Table 7. The proposed heuristic achieved
minimal deviations from the best known solutions with DUP values
solution for the problem with an average processing time of 8 min. of 0.24 and 0.25% for 1,000 and 2,000 activity problems in 5 and
The constraint programming model presented by Menesi et al. 33 s for C1, and in 12 and 83 s for C2. For the problem, including
(2013) also achieved the best known solution in 1 s. However, 1,000 activities, the constraint programming model (Menesi et al.
the heuristic of Hegazy and Menesi (2012) was able to obtain a 2013) was able to obtain a solution with a DUP value of 6.24% in

1 4
1. (12,16,$192)
1. (5,15,$75) 2. (15,10,$150)
3. (18,8,$144)

2 5 8
10
1. (4,16,$64) 1. (22,18,$396)
1. (14,7,$98)
2. (6,10,$60) 2. (24,16,$384) 1. (3,4,$12)
2. (15,6,$90)
3. (8,7,$56) 3. (26,14, $364) 2. (5,2,$10)
3. (16,4, $64)
4. (9,6,$54) 4. (28,12, $336)

3 6
1. (6,13, $78) 1. (14,20,$280)
2. (8,9,$72) 2. (18,15,$270)
3. (10,7,$70) 3. (24,8,$192)

7 9 NO
1. (15,5,$75)
1. (9,17,$153)
2. (18,4,$72) 1. (Dur., Res., Cost)
2. (10,14,$140)
3. (20,3, $60) n. (Dur., Res., Cost)

Costs are in hundreds

Indirect costs=$2,200/Day
Resource limit=30/Day

Fig. 10. Small-scale Test Instance 2

© ASCE 04015090-9 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 Table 5. Comparison of Results for Small-Size Project 2 methods could obtain a solution with a total cost of $51,916,400
Source Method Solution DUP (%) CPU time for the same project. The proposed new heuristic enabled a potential
cost saving in the amount of $2.997 million by providing high-
Chen and Genetic algorithm $244,000 0.00 8 mina
quality solutions for the large-size project. The proposed critical
Weng (2009)
Hegazy and Heuristic $245,900 0.78 2s sequence crashing heuristic not only outperformed state-of-the-art
Menesi (2012) methods, but was also able to achieve high-quality solutions for
Menesi Constraint $244,000 0.00 1s the large-scale RCDTCTP within seconds for the first time.
et al. (2013) programming (CP)
This study CSCH $244,000 0.00 0s
a
Average CPU time for pareto front optimization. Conclusions

A new heuristic was designed and developed for optimal planning

Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.

15 s, and a solution with a DUP value of 4.18% in 120 min. The
performance of the model for the problem including 2,000 activities
worsened and had the model obtained a solution with a DUP value
of 6.67% in 40 s, and a solution with a DUP value of 6.39% in
120 min.
The computational results indicate that CSCH can achieve
high-quality solutions for the large-scale RCDTCTP in a short
computation time by searching a very small portion of the solution
space. CSCH’s effective exploration of the search space can be
mainly attributed to its adequate selection of the activities to be
crashed for a resource-constrained schedule by only considering
crashing of the activities that are on the critical sequence(s). The
backward-forward resource-constrained scheduling method also
enables achieving an adequate project duration after the mode se-
lection for the activities are made. However, CSCH is developed for
the general RCDTCTP, integrating additional assumptions, such as
allowing splitting of resources, which may increase the processing
time of the proposed method. In order to decrease the computation
time, only finish to start precedence relations are allowed in CSCH.
Hence, CSCH requires transformation of the networks, including
start to start, start to finish, and finish to finish relations, to finish
to start networks (Lu and Lam 2009).
CSCH was able to determine a solution with a total cost of
$48,919,400 for the project with 2,000 activities. The state-of-the-art
of large-scale construction projects with project deadlines and re-
source constraints. In the proposed heuristic, the backward-forward
scheduling technique is adopted and crashing of the activities on
the critical sequence are considered to present an effective method
for the resource-constrained discrete time–cost trade-off problem.
The computational experiments confirmed the effectiveness of the
proposed critical sequence crashing heuristic.
The computational tests reveal that the new heuristic achieved
successful results for small-, medium-, and large-scale projects
with project deadlines and resource constraints, and outperformed
the state-of-the-art methods with respect to both solution quality
and computation time requirement. High-quality solutions with mi-
nor deviations from the best known solutions are obtained within
seconds for the large-scale resource-constrained discrete time–cost
trade-off problem for the first time. The main contribution of the
new heuristic is that it provides adequate solutions for the real-
life-size projects within seconds, and enables significant savings
during planning of construction projects with project deadlines
and resource constraints.
The test instances that are used to evaluate the performance
of the proposed critical sequence crashing heuristic included up
to 2,000 activities and reflect the size of the real-life construction
projects. However, the instances do not reflect the complexity of
the real-life construction projects because they are produced by

Table 6. Comparison of Results for Medium-Size Projects

Constraint programming Heuristic
(Menesi et al. 2013) (Hegazy and Menesi 2012) CSCH (this study)
Project size Solution DUP (%) CPU time Solution DUP (%) CPU time Solution DUP (%) CPU time
100 activities $2,497,000 2.34 15 s $2,459,000 0.78 1 min $2,445,400 0.22 0 s (C1)
$2,452,900 0.53 5 min — — — — — 0 s (C2)
$2,447,900 0.32 10 min — — — — — —
300 activities $7,751,700 5.90 15 s $7,377,000 0.78 21 min $7,337,400 0.24 0 s (C1)
$7,479,400 2.18 5 min — — — — — 1 s (C2)
$7,429,600 1.46 10 min — — — — — —
$7,348,900 0.88 20 min — — — — — —

Table 7. Comparison of Results for Large-Size Projects

Constraint programming (Menesi et al. 2013) CSCH (this sudy)
Project size Solution DUP (%) CPU time Solution DUP (%) CPU time
1,000 activities 25,923,800 6.24 15 s $24,459,400 0.24 5 s (C1)
25,882,700 6.07 5 min — — 12 s (C2)
25,571,700 4.80 20 min — — —
25,419,700 4.18 120 min — — —
2,000 activities 52,053,100 6.67 40 s $48,919,400 0.25 33 s (C1)
52,002,200 6.56 10 min — — 83 s (C2)
51,969,200 6.50 30 min — — —
51,916,400 6.39 120 min — — —

© ASCE 04015090-10 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 04015090

 copying small networks in serial. Generation of large-scale com-
plex problem instances and development of methods that can pro-
vide high-quality solutions for the large-scale complex projects
with resource constraints appears to be promising areas for future
research. The quality of the solutions of the proposed heuristic can
be improved by removal of local suboptimalities or by considera-
tion of multipass methods during resource-constrained scheduling
and by inclusion of activities that are not on the critical sequence in
crashing, but these improvements will come at the expense of
increased computational time.
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
Acknowledgments
The authors gratefully acknowledge the financial support provided
by the Scientific and Technological Research Council of Turkey
(TÜBİTAK) under Grant #213M253.
References
Afshar, A., Ziaraty, A., Kaveh, A., and Sharifi, F. (2009). “Nondominated
archiving multicolony ant algorithm in time-cost trade-off optimiza-
tion.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2009)
135:7(668), 668–674.
Ahn, T., and Erenguc, S. S. (1998). “The resource constrained project
scheduling problem with multiple crashable modes: A heuristic pro-
cedure.” Eur. J. Oper. Res., 107(2), 250–259.
Bettemir, O., and Sonmez, R. (2014). “Hybrid genetic algorithm with
simulated annealing for resource-constrained project scheduling.” J.
Manage. Eng., 10.1061/(ASCE)ME.1943-5479.0000323, 04014082.
Bettemir, Ö. H. (2009). “Optimization of time–cost–resource trade-off
problems in project scheduling using meta-heuristic algorithms.”
Ph.D. thesis, Middle East Technical Univ., Ankara, Turkey.
Blazewicz, J., Lenstra, J., and Rinnooy Kan, A. H. G. (1983). “Scheduling
subject to resource constraints: classification and complexity.” Discrete
Appl. Math., 5(1), 11–24.
Bouleimen, K., and Lecocq, H. (2003). “A new efficient simulated
annealing algorithm for the resource-constrained project scheduling
problem and its multiple modes version.” Eur. J. Oper. Res., 149(2),
268–281.
Chan, W. T., Chua, D. K. H., and Kannan, G. (1996). “Construction re-
source scheduling with genetic algorithms.” J. Constr. Eng. Manage.,
10.1061/(ASCE)0733-9364(1996)122:2(125), 125–132.
Chen, P. H., and Shahandashti, S. M. (2009). “Hybrid of genetic algorithm
and simulated annealing for multiple project scheduling with multiple
resource constraints.” Autom. Constr., 18(4), 434–443.
Chen, P.-H., and Weng, H. (2009). “A two-phase GA model for resource-
constrained project scheduling.” Autom. Constr., 18(4), 485–498.
Chen, R. M. (2011). “Particle swarm optimization with justification and
designed mechanisms for resource-constrained project scheduling
problem.” Exp. Syst. Appl., 38(6), 7102–7111.
Chua, D. K. H., Chan, W. T., and Govtndan, K. (1997). “A time-cost trade-
off model with resource consideration using genetic algorithm.” Civ.
Eng. Syst., 14(4), 291–311.
De, P., Dunne, E. J., Ghosh, J. B., and Wells, C. E. (1997). “Complexity of
the discrete time-cost tradeoff problem for project networks.” Oper.
Res., 45(2), 302–306.
Deblaere, F., Demeulemeester, E., and Herroelen, W. (2011). “RESCON:
Educational project scheduling software.” Comput. Appl. Eng. Educ.,
19(2), 327–336.
Elbeltagi, E., Hegazy, T., and Grierson, D. (2007). “A modified shuffled
frog-leaping optimization algorithm: applications to project manage-
ment.” Struct. Infrastruct. Eng., 3(1), 53–60.
Fallah-Mehdipour, E., Bozorg Haddad, O., Rezapour Tabari, M. M., and
Mariño, M. A. (2012). “Extraction of decision alternatives in construc-
tion management projects: Application and adaptation of NSGA-II and
MOPSO.” Exp. Syst. Appl., 39(3), 2794–2803.
© ASCE 04015090-11
J. Constr. Eng. Manage., 04015090
Feng, C. W., Liu, L., and Burns, S. A. (1997). “Using genetic algorithms to
solve construction time-cost trade-off problems.” J. Comput. Civ. Eng.,
10.1061/(ASCE)0887-3801(1997)11:3(184), 184–189.
Goldratt, E. (1997). Critical chain, North River Press, Great Barrington,
MA.
Hartmann, S. (1998). “A competitive genetic algorithm for resource con-
strained project scheduling.” Naval Res. Logist., 45(7), 733–750.
Hegazy, T. (1999). “Optimization of resource allocation and leveling using
genetic algorithms.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-
9364(1999)125:3(167), 167–175.
Hegazy, T., and Menesi, W. (2012). “Heuristic method for satisfying both
deadlines and resource constraints.” J. Constr. Eng. Manage., 10.1061/
(ASCE)CO.1943-7862.0000483, 688–696.
Hegazy, T., Shabeeb, A., El-Beltagi, E., and Cheema, T. (2000). “Algorithm
for scheduling with multi-skilled constrained resources.” J. Constr. Eng.
Manage., 10.1061/(ASCE)0733-9364(2000)126:6(414), 414–421.
Hekimoglu, O. (2007). “Comparison of the resource allocation capabilities
of project management software packages in resource constrained
project scheduling problems.” M.S. thesis, Middle East Technical
Univ., Ankara, Turkey.
Herroelen, W., and Leus, R. (2005). “Identification and illumination of
popular misconception about project scheduling and time buffering
in a resource-constrained environment.” J. Oper. Res. Soc., 56(1),
102–109.
Kandil, A., and El-Rayes, K. (2006). “Parallel genetic algorithms for opti-
mizing resource utilization in large-scale construction projects.” J.
Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2006)132:5(491),
491–498.
Kelley, J. E., Jr. (1963). “The critical-path method: Resources planning and
scheduling.” Industrial scheduling, J. F. Muth and G. L. Thompson,
eds., Prentice-Hall, Englewood Cliffs, NJ, 347–365.
Kim, J. L., and Ellis, R. D. (2008). “Permutation-based elitist genetic algo-
rithm for optimization of large-sized resource-constrained project
scheduling.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364
(2008)134:11(904), 904–913.
Kim, J. L., and Ellis, R. D. (2010). “Comparing schedule generation
schemes in resource-constrained project scheduling using elitist genetic
algorithm.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364
(2010)136:2(160), 160–169.
Lee, J. K., and Kim, Y. D. (1996). “Search heuristics for resource con-
strained project scheduling.” J. Oper. Res. Soc., 47(5), 678–689.
Leu, S. S., and Yang, C. H. (1999). “A genetic-algorithm-based resource-
constrained construction scheduling system.” Constr. Manage. Econ.,
17(6), 767–776.
Li, K., and Willis, R. (1992). “An iterative scheduling technique for re-
source-constrained project Scheduling.” Eur. J. Oper. Res., 56(3),
370–379.
Liberatore, M., Pollack-Johnson, B., and Smith, C. (2001). “Project man-
agement in construction: Software use and research directions.” J.
Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2001)127:2(101),
101–107.
Lova, A., and Tormos, P. (2002). “Combining random sampling and
backward–forward heuristics for resource-constrained multi-project
scheduling.” Proc., 8th Int. Workshop on Project Management and
Scheduling, EURO Working Group, 244–248.
Lu, M., and Lam, H-C. (2009). “Transform schemes applied on nonfinish-
to-start logical relationships in project network diagrams.” J. Constr.
Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000062, 863–873.
Lu, M., Lam, H-C., and Dai, F. (2008). “Resource-constrained critical path
analysis based on discrete event simulation and particle swarm optimi-
zation.” Autom. Constr., 17(6), 670–681.
Lu, M., and Li, H. (2003). “Resource-activity critical-path method for con-
struction planning.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-
9364(2003)129:4(412), 412–420.
Mellentien, C., and Trautmann, N. (2001). “Resource allocation with
project management software.” OR Spektrum, 23(3), 383–394.
Menesi, W., Golzarpoor, B., and Hegazy, T. (2013). “Fast and near-
optimum schedule optimization for large-scale projects.” J. Constr.
Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000722, 1117–1124.
J. Constr. Eng. Manage.
 Ng, S., and Zhang, Y. (2008). “Optimizing construction time and cost using
ant colony optimization approach.” J. Constr. Eng. Manage., 10.1061/
(ASCE)0733-9364(2008)134:9(721), 721–728.
Özdamar, L., and Ulusoy, G. (1994). “A local constraint based analysis
approach to project scheduling under general resource constraints.”
Eur. J. Oper. Res., 79(2), 287–298.
Siemens, N. (1971). “A simple CPM time-cost tradeoff algorithm.”
Manage. Sci., 17(6), B–354–B–363.
Sonmez, R., and Bettemir, Ö. H. (2012). “A hybrid genetic algorithm for
the discrete time-cost trade-off problem.” Exp. Syst. Appl., 39(13),
11428–11434.
Sonmez, R., and Uysal, F. (2014). “Backward-forward hybrid genetic
algorithm for resource-constrained multiproject scheduling problem.”
J. Comput. Civ. Eng., 10.1061/(ASCE)CP.1943-5487.0000382,
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
04014072.
Tormos, P., and Lova, A. (2001). “A competitive heuristic solution tech-
nique for resource-constrained project scheduling.” Ann. Oper. Res.,
102(1/4), 65–81.
Valls, V., Ballestin, F., and Quintanilla, M. S. (2005). “Justification
and RCPSP: A technique that pays.” Eur. J. Oper. Res., 165(2),
375–386.
© ASCE 04015090-12
J. Constr. Eng. Manage., 04015090
Vanhoucke, M., and Debels, D. (2007). “The discrete time/cost trade-off
problem: Extensions and heuristic procedures.” J. Scheduling, 10(4–5),
311–326.
Wang, Q., and Qi, J. (2009). “Improved particle swarm optimization for
RCP scheduling problem.” 6th Int. Symp. on Neural Networks,
Springer, Berlin, 49–57.
Wiest, J. D. (1964). “Some properties of schedules for large projects with
limited resources.” Oper. Res., 12(3), 395–418.
Wuliang, P., and Chengen, W. (2009). “A multi-mode resource-constrained
discrete time-cost tradeoff problem and its genetic algorithm based
solution” Int. J. Project Manage., 27(6), 600–609.
Xiong, Y., and Kuang, Y. (2008). “Applying an ant colony optimization algo-
rithm-based multiobjective approach for time-cost trade-off.” J. Constr.
Eng. Manage., 10.1061/(ASCE)0733-9364(2008)134:2(153), 153–156.
Yang, I. (2007). “Using elitist particle swarm optimization to facilitate
bicriterion time-cost trade-off analysis.” J. Constr. Eng. Manage.,
10.1061/(ASCE)0733-9364(2007)133:7(498), 498–505.
Zheng, D., Ng, S., and Kumaraswamy, M. (2005). “Applying Pareto
ranking and niche formation to genetic algorithm-based multiobjective
time-cost optimization.” J. Constr. Eng. Manage., 10.1061/(ASCE)
0733-9364(2005)131:1(81), 81–91.
J. Constr. Eng. Manage.