Applying A Unified Game-Based Model in A Payment Scheduling Problem and Design of Experiments Sing MOEA Framework
Applying A Unified Game-Based Model in A Payment Scheduling Problem and Design of Experiments Sing MOEA Framework
net/publication/336320218
CITATIONS READS
3 1,491
5 authors, including:
SEE PROFILE
All content following this page was uploaded by Thang Huynh Quyet on 08 October 2019.
Abstract. This paper aims to put a new approach in the picture to the payment
scheduling problem, which looks for a schedule that maximizes the benefit of all
parties in a project. In a project, both sponsor and contractor seek to have a good
payment strategy on their own. The timing of payments and the completion times of
activities in projects are determined simultaneously in order to achieve an equitable
schedule among the sponsor and the development team. In previous research, we
developed a Unified Game-Based Model for conflicts in project management. In
this paper, we applied this model to this problem, implemented in an open sourced
evolutionary computation library named MOEA framework. The use of a Unified
Game-Based Model enables us to figure out a suitable schedule for the problem, and
in the tool, we conducted an experimental test of the model by the used of several
multi-objective optimization algorithms. The experimental results demonstrated that
the presented approach is effective and promising so that both parties could use this
model to choose the proper tactics for each of them in scheduling payment.
1. Introduction
The schedule is a plan for carrying out a process or procedure, giving lists of
intended events and times. Scheduling problems in project management, which is
considered as one of the most important and challenging aspects, have caught much
attention of both theoreticians and practitioners in the research community [1,2,3,4,5,6].
It is a generic name given to a whole class of problems in which the allocation of the
project resources, the optimizing of the project duration and the estimated project costs
are necessary [3]. In [1], the authors introduced the project scheduling problems by
presenting a survey of the general parameters: the resources, and the activities, as well
as the algorithms used to solve the problems and the differences of the variants of the
problems. This matter was also provided an extensive literature review of the models and
solution procedures with non-deterministic activities duration [2]. There is three
1
Quyet-Thang Huynh, Department of Software Engineering, School of Information and
Communication Technology, Hanoi University of Science and Technology, Hanoi, Vietnam; E-mail:
[email protected]
56 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
(i) Based on the Unified Game-Based Model, propose a novel solution including 2
players, their strategies, their payoff function, and Nash Equilibrium
(ii) Demonstrate the feasibility of the model by using the MOEA framework and
some multi-objective optimization algorithms in finding Nash Equilibrium point.
The remainder of this paper is organized as follows. We introduced the basic
knowledge of Game theory, multi-objective algorithms as well as MOEA framework in
Section 2. Section 3 presented the conflicts in project payment scheduling problem. In
this section, we listed all tensions between the sponsor and the development team to have
an objective view on this issue. In Section 4 we applied the Unified Game-Based Model
into the problem mentioned above. We conducted the experimental evaluation in Section
5, and Section 6, concluded with a summary and shows our future work.
2. Background
Game theory, known well as the study of mathematical models of conflict and
cooperation between rational decision-makers, has the ability to address almost the issues
in all fields of social science as well as in computer science. [9]. In project scheduling
problems, game theory plays an essential part in solving its practical applications. Piotr
Skowron and Krzysztof Rzadca [10] modeled the fair scheduling problem as a
cooperative game theory approach to determine the ideal fair schedule. George
Christodoulou et al. [11] studied the mechanism design problem of scheduling tasks at
the intersection of Computer Science and Game Theory. Cyril Briand and J. Billaut [12]
considered a project-scheduling environment assuming that the activities of the project
network are distributed among a set of actors (or agents). Na Fu et al. [13] provided two
new practical mechanisms referred to as Individual Completion based Payments and
Social Completion based Payments no longer only to minimize makespan for the entire
project, but rather, to maximize social welfare. Andres Abeliuk et al. [14] proposed an
independent scheduling game model where each player commands a set of services
scheduled independently.
In the context of the research activities presented in this paper, the focus lays on the
process-related aspects of game theory and players, who are in successive problem-
solving of risk management. Moreover, we take advantage of some extended form of
Genetic algorithm to figure out the solution [8, 15, 21, 24].
There are several approaches in solving the multi-objective problem, such as Integer
linear programming, but in this research, we used evolutionary multi-objective
algorithms because of the following reasons:
First, a free and open-source Java named MOEA framework was widely used to
support in building a fast, reliable implementation of many state-of-the-art multi-
objective evolutionary algorithms [18]. This library supports numerous types of MOEAs
and various metrics to evaluate the results and performances of various MOEAs on a
specific problem [19]. In addition to these pre-defined algorithms, new algorithms can
be easily constructed using existing components. Using algorithms in MOEA framework
can help us to have a faster and better experiment.
58 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
Costs associated with project activities or cash outflows are usually known, but
payments comprising of contractor's cash inflows or sponsor's cash outflows are mostly
made during project execution and their terms can be negotiated within the framework
of the contract agreement. Due to milestones in a contract agreement, there is a pre-
defined payment schedule for both sponsor and contractor, but in fact, a complicated
project involves many changes during their development, planning, and realization. As
a result, the pre-defined schedule has to change regularly too.
There is always a conflict between sponsor and contractor in payment schedule in
terms of time to pay, and the sponsor tends to pay as late as possible. In contrast, the
contractor wants to have a significant amount of money received as soon as possible. Of
course, in terms of financial control, when people have a late payment than usual, they
must pay for an additional interest rate. In the conflict as above, there are some methods
to deal with, but the best way to resolve the conflict is collaborating [26] to find a win-
win solution. In this problem, we utilize the game theory model and Nash equilibrium to
figure out a solution that can satisfy both sides, the sponsor and the contractor to archive
a win-win situation. A directed graph is used for visually presenting the information of
the model, in which: (i) Each node of the graph represents an activity; (ii) Nodes have
no relationship can be performed simultaneously; (iii) Each node contains several
attributes such as: characteristics of resources required to complete the activity, time
needed to finish the activity.
Example:
Assume that in the project is going to be executed there is a set ܣincluding ݊
activities that need to define the order of execution. and ݉ constraints on the order of
distinct activity pairs. There are 4 types of constraints: Finish to Start (FS), Finish to
Finish (FF), Start to Start (SS), Start to Finish (SF). Each activity is characterized by two
attributes:
x Time need to complete ሺݐሻ and
x Number of people or machine needed to calculate the fee of activity ሺݏ݁ݎሻ.
Assume a small project with 5 activities ܣൌ ܽଵ ǡ ܽଶ ǡ ܽଷ ǡ ܽସ ǡ ܽହ in which there is a
constraint between activities as follows: ܽଵ ܽܵܨଶ ǡ ܽଶ ܽܵܨଷ ǡ ܽଵ ܵܵܽହ . As shown in
Table 1, each cell shows attributes of each activity, includes: t in first row, res in second
row.
The project's resource consists of 1 staff and 1 personal computer (PC), and they
cost $10000 per day. The sponsor requests the contractor to follow the initial order of
activities from 1 to 5. Assume that the currency depreciates by 1% each day. When
paying for 5 activities, the money will be $110000. In case, the sponsor chooses the
strategy of arranging activities order such as: ܽଵ ǡ ܽଶ ǡ ܽଷ ǡ ܽସ ǡ ܽହ , only use one person and
one machine for 11 days. Due to the price depreciation, this planned $110000 must be
actuall ̈́ʹͲͲͲͲሺͳ ͳΨሻଶ ̈́ʹͲͲͲͲሺͳ ͳΨሻସ ̈́ʹͲͲͲͲሺͳ ͳΨሻ ̈́ʹͲͲͲͲሺͳ
ͳΨሻ଼ ̈́͵ͲͲͲͲሺͳ ͳΨሻଵଵ ൌ ̈́ͳͳǡͷͳ to achieve the original value for the
contractor. Therefore, the contractor lose ̈́ͳͳǡͷͳwhereas the sponsor actually lose
the less money than expected.
So, the contractor has re-ordered the activity chain in a different order to do activity
5 first to avoid the impact of the devaluation. However, activity 5 still have to stand
behind activity 1 due to the impact of the conflict between activity 1 and 5, these
alternatives would be ܽଵ ̴̴ܽହ Ǣܽଶ Ǣܽଷ . FS is sequence (;) and SS is parallelism (___).
Therefore, the contractor recommends the sequence of activity is ܽଵ ǡ ܽହ ǡ ܽସ ǡ ܽଶ ǡ ܽଷ .
If the project is done as the solution mentioned above, this planned $110000 must be
actuall ̈́ʹͲͲͲͲሺͳ ͳΨሻଶ ̈́ʹͲͲͲͲሺͳ ͳΨሻସ ̈́͵ͲͲͲͲሺͳ ͳΨሻ ̈́ʹͲͲͲͲሺͳ
ͳΨሻ଼ ̈́ʹͲͲͲͲሺͳ ͳΨሻଵଵ ൌ ̈́ͳͳǡͲʹͻ . Therefore, the value of money that the
contractor receives will be more $542 than the plan of the sponsor.
in which,
ܰ: the number of activities in the project;
ݔ: the money needed to complete activities;
ݕ: the time needed to complete activities;
ݎ: banking rate;
ܣǡ ܤ: tuning constants that can be adjust by decision maker of the payment
scheduling problem.
62 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
5. Experimental results
Day 1 2 3 4 5 6
Activity 1 1 2 2 3 3
Activity 4 4 5 5 5
the same, in which implemented the processes of all algorithms. In the output module of
the MOEA framework, we modified some library to show the needed information for
analysis such as: running time, payoff value, best strategy found (Nash Equilibrium).
The MOEA framework provides access to 24 standard multi-objective evolutionary
algorithms, but they have the same execution routine. So that, modification for an
algorithm could apply to others: NSGA-II, PESA-II, ε-MOEA, GDE3, ε-NSGA-II, and
SMPSO. The next step in using the MOEA framework is that customize this framework’s
parameters for the implementation of 6 above algorithms. The process of utilizing
MOEA framework follows the evolutionary algorithms process includes these steps:
initialization, selection, hybridization, mutation. In detail, the modifying in MOEA
framework source code is described as follow:
5.2.1. Initialization
Following the constraints and chromosome principles defined above, it generates
randomly some individuals and arranges the adaptability of the first generation by
evaluating fitness. We modified the MOEA framework library to assign the population
for each generation by 100. This constant can be changed later.
5.2.2. Selection
To improve the fitness of the next generations, the algorithms must elect most
compatible individuals for the hybrid process. The chromosomes, fitness values of which
are minimum in the sorted population, will be the best solutions produced after each
generation.
5.2.3. Hybridization
After population arrangement completes, it takes half of the population which have
a high adaptability value to pair with each other. Then, the new generations, which are
hybridized from grafted pairs, replace ½ populations have less ability to adapt. The
number of chromosomes in the new generation is equal to that in the old population but
only include those with high adaptability and the new generation.
The hybrid process includes: select cutting point randomly, generate 2 children (one
inherits gene fragments from head to cutting point of mother chromosome and from the
cutting point to the end of father chromosome, the other inherits conversely).
5.2.4. Mutation
The mutation will be conducted interchangeably with the hybrid. A mutation is
usually applied with a low probability ௫ to preserve diversity in the genetic population.
There are many kinds of mutation but all of them need to split chromosomes into
segments to carry out mutation on selected ones.
The data used in the experiment are being collected from paper [8] and two
outsourcing software projects which need a fair payment schedule arrangement. It is
being filtered and finally tabulated in SQL format then. All 3 projects are developed in
64 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
different areas, either in different environments. Research has shown that the Unified
Game-Based Model and its application can work with all data sets pretty well.
The computer configuration for algorithms comparison is Windows 10 Pro 64-bit in
CPU 2.80 GHz Intel(R) Core i5. For each particular algorithm, we run continuously 10
times, each time we record the Nash equilibrium results found. Results of each run for
each algorithm contains the following 3 primary information: (i) Payoff function of the
solution; (ii) The running time; (iii) Nash Equilibrium that includes a percentage of
budget spend for each activity.
In which the shorter the running time, the better the algorithm's relevance and the
higher the model. The payoff function value indicates the quality of the answer found.
The smaller this value is, the better the results are found. In Table 6, Table 7, due to the
limit of paper, we only show the first 2 information: payoff value, running time.
27 2 1 12 64 9 10 1
28 10 5 21 65 6 1 1
29 5 10 3 66 3 3 1
30 4 6 3 67 8 4 3
31 3 7 10 68 8 10 1
32 3 9 10 69 8 7 1
33 6 5 14 70 5 4 27
34 9 4 28 71 10 9 15
35 9 6 35 72 5 2 15
36 6 7 6 73 6 2 17
37 1 8 34 74 4 4 10
Table 4. The parameters of the activity network from Project 1
No Activity 1 Activity 2 Relation
1 16 27 FS
2 56 58 FS
3 13 62 FS
4 45 61 FS
5 37 43 FF
6 50 34 FF
7 65 41 FF
8 26 29 FF
9 8 20 FF
10 34 15 FF
11 32 33 SF
12 63 74 SF
13 19 7 SS
14 27 45 SS
15 4 56 SS
16 12 70 SS
17 1 2 FS
18 1 3 FS
19 1 4 FS
20 1 5 FS
Table 5. The parameters of the relationship between activities in Project 1’s data
In Table 6, each row contains the payoff value of the Nash equilibrium point found, and
the running time in second of each algorithm. Eg: in the first run with NSGA-II, the value
in Table 6 is “301343880/7s” means the payoff value of Nash equilibrium point is
301343880, the running time of NSGA-II is 7 seconds.
No NSGA-II ε-NSGA-II ε-MOEA GDE3 PESA2 SMPSO
1 301343880/7s 301343880/9s 301343880/9s 301343880/10s 301343880/9s 301343880/13s
2 301343880/8s 301343880/8s 301343880/10s 301343880/10s 301343880/9s 301343880/13s
3 301343880/9s 301343880/9s 301343880/10s 301343880/10s 301343880/9s 301343880/11s
4 301343880/8s 301343880/8s 301343880/9s 301343880/10s 301343880/9s 301343880/13s
5 301343880/11s 301343880/9s 301343880/10s 301343880/10s 301343880/9s 301343880/11s
6 301343880/7s 301343880/10s 301343880/9s 301343880/10s 301343880/9s 301343880/13s
7 301343880/8s 301343880/10s 301343880/10s 301343880/10s 301343880/9s 301343880/13s
8 301343880/10s 301343880/9s 301343880/9s 301343880/10s 301343880/9s 301343880/11s
9 301343880/9s 301343880/9s 301343880/9s 301343880/10s 301343880/9s 301343880/11s
10 301343880/11s 301343880/8s 301343880/10s 301343880/10s 301343880/9s 301343880/11s
Table 6. Testing results from Project 1’s data bucket after 10 times running
Also using the above model and putting it into the input of MOEAs in this problem,
Table 6 shows that the algorithms as mentioned above are compared in order to check
which of them is most efficient in solving the matter and finding the Nash Equilibrium
point for the Unified Game-Based Model. As demonstrated in Table 6, the payoff values
66 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
are equal, but the running time is dissimilar among these algorithms. NSGA-II is
considered as the fastest MOEA with 8.8s, which is faster than 12.27% than the average
running time of five remain algorithms, following is PESA2 with 9s.
No
NSGA-II ε-NSGA-II ε-MOEA GDE3 PESA2 SMPSO
1 280675880/11s 280675880/9s 259276750/10s 280675880/9s 280675880/8s 280675880/12s
2 280675880/11s 280675880/10s 280675880/9s 280675880/9s 280675880/8s 280675880/13s
3 259276750/10s 280675880/8s 280675880/9s 280675880/9s 280675880/8s 280675880/11s
4 280675880/7s 280675880/10s 280675880/8s 280675880/9s 280675880/8s 280675880/13s
5 280675880/7s 280675880/9s 259276750/10s 280675880/9s 280675880/8s 280675880/12s
6 280675880/10s 280675880/9s 219878880/9s 280675880/9s 280675880/8s 280675880/11s
7 280675880/11s 219878880/9s 280675880/8s 280675880/9s 280675880/8s 280675880/12s
8 280675880/7s 280675880/8s 280675880/8s 280675880/9s 280675880/8s 280675880/13s
9 259276750/10s 280675880/6s 219878880/9s 280675880/9s 280675880/8s 280675880/12s
10 280675880/9s 280675880/8s 219878880/9s 280675880/9s 280675880/8s 280675880/11s
Table 7. Testing results of Project 2’s data bucket in 10 times running
In Table 7, each cell of table contains the payoff value of the Nash equilibrium point
found, and the running time in second of each algorithm. Eg: in the first run with NSGA-
II, the value in Table 7 is “280675880/11s” means the payoff value of Nash equilibrium
point is 280675880, the running time of NSGA-II is 11 seconds.
Different from the two above data bucket, there is a slight change in the payoff
values. The payoff values fluctuate when using NSGA-II, ε-NSGA-II, and ε-MOEA. In
this test, PESA2 is the fastest MOEA in the experiment to find the Nash Equilibrium
point after 10 runs in project 2’s data. In detail, the running time of it is 8.75% higher
than ε-NSGA-II, 7.5% higher than ε-MOEA, 12.5% higher than GDE3, 15% higher than
NSGA-II and 50% than SMPSO. The difference between the effectiveness of MOEAs
B. Ngoc Trinh et al. / Applying a Unified Game-Based Model 67
in 2 tests may come from the small size data. Some randomly step in algorithms such as:
initialize, the mutation has a significant impact on the convergence of the algorithm.
6. Conclusion
Acknowledgments
This research was supported by Vietnam’s National Foundation for Science and
Technology Development (NAFOSTED), funded by the Ministry of Science and
Technology (Project Code 102.03-2019.10).
68 B. Ngoc Trinh et al. / Applying a Unified Game-Based Model
References
[1] Jorge A. Ruiz-Vanoye, et al, Editorial for Volume 9 Number 3: Models, Resources and Activities of
Project Scheduling Problems, International Journal of Combinatorial Optimization Problems and
Informatics, Vol. 9, No. 3, Sep-Dec 2018, pp. 1-17. ISSN: 2007-1558.
[2] Nestor Raul Ortiz-Pimiento, Francisco Javier Diaz-Serna, The Project Scheduling Problem with Non-
Deterministic Activities Duration: A Literature Review, JIEM, 2018 - 11(1): 116-134 - Online ISSN:
2013-0953 - Print ISSN: 2013-8423 https://doi.org/10.3926/jiem.2492
[3] Icmeli, Oya., Scheduling problems in project management, University of Florida, George A. Smathers
Libraries, 1992, pp.7-22.
[4] Dayanand, N. and Padman, R. Payments in projects: a contractor’s model, Technical Report, The Heinz
School, Carnegie Mellon University, Pittsburgh 1993
[5] Richard C. Grinold, The payment scheduling problem, Volume 19, Issue 1,
https://doi.org/10.1002/nav.3800190110
[6] Kazaz, B. and Sepil, C.B. Project scheduling with discounted cash flows and progress the payments,
Journal of Operational Research Society, 47 (1996), 1262-1272
[7] Bao Ngoc Trinh, Quyet Thang Huynh, Xuan Thang Nguyen, Nash Equilibrium model for conflicts in
project management, Journal of Computer Science and Cybernetics, Journal of Computer Science and
Cybernetics, V.35, N.2 (2019), 167–184; DOI: DOI 10.15625/1813-9663/35/2/13095
[8] Deng Ze-min, Gao Chun-ping, Li Zhong-xue, Optimization of project payment schedules with Nash
equilibrium model and genetic algorithm. J Chongqing Univ, ISSN 1671-8224, 2007, 6 (2):107-112.
[9] José Ramón San Cristóbal, The use of Game Theory to solve conflicts in the project management and
construction industry, International Journal of Information Systems and Project Management
[10] Piotr Skowron, Krzysztof Rzadca, Non-monetary fair scheduling — cooperative game theory approach,
in Annual ACM symposium on parallelism in algorithms and architectures, pp 288–297, 2013
[11] George Christodoulou et al., A lower bound for scheduling mechanisms, Article in Algorithmica 55,
December 2009
[12] Cyril Briand, Jean-Charles Billaut, Cooperative project scheduling with controllable processing times:
a game theory framework, IEEE 5-9 Sept. 2011
[13] Pradeep Varakantham, Na Fu, Mechanism Design for Strategic Project Scheduling, in Twenty-Sixth
International Joint Conference on Artificial Intelligence, August 2017
[14] Andres Abeliuk et al., Interdependent Scheduling Games, Proceedings of the Twenty-Fifth International
Joint Conference on Artificial Intelligence
[15] K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, A fast and elitist multiobjective genetic algorithm:
NSGA-II, IEEE Transactions on Evolutionary Computation, vol. 6, no. 2, pp. 182–197, 2002.
[16] Miqing Li, Shengxiang Yang, Xiaohui Liu, Kang Wang, IPESA-II: Improved Pareto Envelope-Based
Selection Algorithm II, EMO 2013: Evolutionary Multi-Criterion Optimization, pp143-155
[17] L.C. Jiao, Handing Wang, R.H. Shang, F. Liu, A Co-evolutionary Multi-Objective Optimization
Algorithm Based on Direction Vectors, Information Sciences 228:90 – 112, April 2013, DOI:
10.1016/j.ins.2012.12.013
[18] MOEA framework, A Java library for multi-objective evolutionary algorithms,
http://www.moeaframework.org/, accessed: March 10, 2-15.
[19] DTIC ADA615240: Tactical AI in Real Time Strategy Games, Defense Technical Information Center,
2015-03-26.
[20] Deb, K. et al., A Fast Multi-Objective Evolutionary Algorithm for Finding Well-Spread Pareto-Optimal
Solutions, KanGAL Report No 2003002, Feb 2003.
[21] Chih-Hao Lin and Pei-Ling Lin (2010), A New Non-dominated Sorting Genetic algorithm for Multi-
Objective Optimization, Modeling Simulation and Optimization Focus on Applications, Shkelzen Cakaj
(Ed.), ISBN: 978- 953-307-055-1.
[22] Lin, C.H. & Chuang, C.C. (2007), A rough set penalty function for marriage selection in multiple-
evaluation genetic algorithms, Lecture Notes in Computer Science, Vol. 4481, pp. 500-507.
[23] Kukkonen and Lampinen (2005), GDE3: The Third Evolution Step of Generalized Differential Evolution,
KanGAL Report Number 2005013.
[24] David W. Corne, Nick R. Jerram, Joshua D. Knowles, Martin J. Oates (2001), PESA-II: Region-based
Selection in Evolutionary Multi-objective Optimization, In Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO’2001), Pages 283-290, ISBN:1-55860-774-9
[25] A.J. Nebro, J.J. Durillo, J. Garcıa-Nieto, C.A. Coello Coello, F. Luna, E. Alba (2009), SMPSO: A New
PSO Metaheuristic for Multi-objective Optimization. Conference: Computational intelligence in multi-
criteria decision-making, DOI: 10.1109/MCDM.2009.4938830
[26] A Guide to the Project Management Body of Knowledge (PMBOK® Guide)—Sixth Edition, Project
Management Institute, 2017.