ASM Sc. J.

, 14, Special Issue 1, 2021 for ICSTSS2018, 19-24

A Mathematical Programming Model for Exam-

Invigilator Assignment Problem
Mazura Mokhtar1*, Saharani Abdul Rashid1, Nur Fatihah Haron1 and Mohamad Affendi Abdul Malek1

1
Faculty of Computer and Mathematical Sciences,Universiti Teknologi MARA Cawangan Pahang, Raub Campus, 27600
Raub, Pahang, Malaysia

At the end of each term or semester academic institutions must assign invigilators to exams as it is
an important administrative activity that must be performed. To create a good exam invigilator
schedule manually is a complex and time-consuming process as it must satisfy various requirements
and constraints. Therefore, the aim of this study is to propose a mathematical programming model
to solve the exam invigilator assignment problem at Universiti Teknologi Mara (UiTM) Pahang,
Raub Campus. The model, which was formulated by using integer programming, assigns lecturers
to time slots and rooms. The objectives of the proposed model are to fairly assign duties to the
invigilators, chief invigilator and standby invigilators among the academic staff. This study also
proposes new constraints, which are the chief invigilator in a large room should be a senior lecturer
and chief invigilator’s position can only be assigned to a lecturer once. These two constraints have
never been considered in any studies. The model is sufficiently flexible to be used with various
operational requirements in most academic institutions. Computational experiment was conducted
by using real data from UiTM Pahang Raub Campus. Results from the experiment demonstrated that
the proposed model can produce a feasible and optimal timetable that satisfies all th e constraints
within a reasonably short time as compared to the manual assignment procedure.
Keywords: examination timetabling; integer programming; invigilator assignment

I. INTRODUCTION constraints that are used in British universities can be

found in Burke et al. (1996).
Examination timetabling problems are among the most
Invigilator assignment is one aspect of the examination
important and difficult tasks faced by many academic
timetabling problems. The problem deals with scheduling
institutions worldwide. The problems are combinatorial
of invigilators to a given number of examinations and
optimisation problems in which a set of examinations are
rooms so that there are no conflicts or clashes (Ozturk et
required to be scheduled within a fixed number of time slots or
al., 2010). Invigilator assignment is often done separately
periods and rooms so that no student has to take multiple
prior to or after the examination timetabling phase.
examinations on the same time slot (Carter, 1986). The
Constructing an invigilator’s schedule manually is a
examination timetabling problems are subjected to various
difficult task and it cannot be done quickly. In addition, the
types of hard constraints that must be met at all costs and soft
manual process is vulnerable to errors and may require
constraints that may be violated but should be satisfied as many
several corrections and amendments before a satisfactory
as possible. Both hard and soft constraints vary among
solution is obtained. The complexity of this problem
institutions, depending on their particular needs and limited
depends mainly on the number of available invigilators,
resources. Examples and detailed explanations of hard and soft
number of examinations and number of examination
rooms. Although several models have been proposed in

*Corresponding author’s e-mail: [email protected]

 ASM Science Journal, Volume 14, Special Issue 1, 2021 for ICSTSS2018

literature to solve the invigilator assignment problems, it is complete description of the hard and soft constraints
difficult to find a model or solution that can solve this problem imposed on the model. The objective function of the model
generally because the needs and requirements vary is also presented in Section 4. Meanwhile, the
significantly across different institutions. implementation of the proposed model by using real data
In practice, there are various constraints that should be is presented in Section 5. Finally, conclusion and possible
overcome to solve the invigilator assignment problem. future research studies are highlighted in Section 6.
Although the constraints differ among institutions, there are
some common requirements that serve as basis for the general II. LITERATURE REVIEW
model. Among the constraints are invigilators must not be
Over the past several decades, a number of approaches were
scheduled for more than one room in a time slot, invigilators
proposed to solve a variety of invigilator timetabling
cannot invigilate their own exam papers, and invigilation
problems. A common approach to solve the problem is to
duties must be assigned fairly among invigilators. Unfair
formulate the problem by using mathematical
invigilation duties may generate conflicts between invigilators
programming. Among the most interesting studies which
and the administration. On top of that, a lecturer preference
had used this approach were those by Kahar and Kendall
survey conducted by Cowling et al. (2002) suggested that
(2014) and Marti et al. (2000). Kahar and Kendall (2014)
invigilators preferred to have two to three invigilation duties
formulated a mathematical programming model based on
with one or two days gap, and lecturers with other
integer programming for Universiti Malaysia Pahang
responsibilities, such as administrative or research work,
(UMP). The model considers three other hard constraints
should be given less invigilation duties.
in addition to the ones presented in Cowling et al. (2002).
This study is concerned with the invigilator assignment
The constraints are the chief invigilator must be a lecturer,
problem at Universiti Teknologi MARA (UiTM) Pahang, Raub
all staff must invigilate not more than three examinations
Campus. Currently, the problem is manually solved by an
within the exam period, and the total number of invigilators
examination timetable committee. The process requires two to
who are assigned to each room has to equal the number of
three days of work and sometimes the obtained solution failed
invigilators required for each room. On top of that, a
to adhere to some requirements imposed by the scheduler. To
constructive algorithm that can produce good quality
improve the assignment process, a mathematical
solutions was also proposed as compared to the software
programming model based on integer linear programming
used by UMP.
approach was developed. The model aims to satisfy a set of
Marti et al. (2000) formulated an exam invigilator
constraints, which are lecturers cannot invigilate their own
assignment problem as a multi-objective integer
subjects, no lecturer is scheduled to invigilate two or more
programming model with a weighted objective function
examinations on the same time slot, only one chief invigilator
that integrated a preference function with a workload-
is required in a room, chief invigilator in a large room should
fairness function. The model used the concept of combining
be a senior lecturer and a lecturer can only be assigned as the
good solutions to obtain a better solution. To solve the
chief invigilator once. the proposed approach can be applied
formulated model, a solution technique was used based on
to invigilator scheduling issues at any other educational
a scatter search. Koide and Iwata (2014) formulated a
institutions which encounter the same type of problems.
mathematical programming model for the invigilator
The remaining part of this paper proceeds as follows: Section
assignment problem at Konan University by employing a
2 reviews previous related works on the exam invigilator
mixed integer programming approach. The authors also
assignment problem. Section 3 presents a description of the
built a prototype system by using a spreadsheet tool to find
invigilator scheduling problem at UiTM Pahang, Raub
the solution. Koide (2015) later extended and revised the
Campus, including the assumptions of the adopted model. The
work by Koide and Iwata (2014) to deal with several new
formulation of the integer linear programming model for the
practical conditions for invigilator assignment at Konan
problem is discussed in Section 4, which also comprises a
University. However, the proposed model was not able to

20
 ASM Science Journal, Volume 14, Special Issue 1, 2021 for ICSTSS2018

yield a feasible solution in an acceptable time for the system III. INVIGILATOR ASSIGNMENT
users. Recently, Hanum et al. (2015) formulated an exam AT THE UiTM PAHANG RAUB
CAMPUS
invigilator timetabling problem by using the non-pre-emptive
goal programming approach. The model offers more fairness UiTM Pahang Raub Campus is one of the UiTM campuses

by incorporating several preferences related to the equity of with almost 3,000 students. The university has three

invigilating task number. The proposed model was successfully faculties which offer five diploma programmes in Business,

applied to a simple case of exam invigilator assignment at the Banking, Public Administration, Computer Science and

Department of Mathematics in Bogor Agricultural University. Statistics. At the end of each semester, students must attend

Several researchers have employed metaheuristic algorithms examinations for a couple of weeks. The examination

to solve the exam invigilator assignment problem. Erden et al. timetable is prepared by the Examination Unit of UiTM

(2016), for instance, used a genetic algorithm to find a solution main campus. Once the examination timetable is ready, all

that did not have overlaps in exams or invigilation duties, while faculties and branch campuses need to assign examinations

the invigilator preferences were satisfied as much as possible. and invigilators to rooms and time slots. At UiTM Pahang

In Awad and Chinneck (1998), a basic genetic algorithm Raub Campus this task is done manually by an examination

framework was combined with a simple user interface based on timetable committee which consists of eight to nine

readily available software tools to develop a computer-based members and normally it takes three to four days to

system for assigning invigilators. Pokudom et al. (2010) used complete.

the ant colony system to generate exam invigilator schedules To formulate a mathematical programming model for the

for educational institutes. The aim of the study was to reduce exam invigilator assignment problem, the following

the time for organising each staff’s invigilating schedule. The assumptions were considered in this study.

proposed method was able to produce equal invigilation duties

among staff during normal and extra workload times and avoid Assumptions:

exam proctoring on weekends and on any staff’s engaged hours. a) The examination timetable is already available.

There are also studies which focused on developing a decision b) There are two time slots per day, morning and

support system for solving the invigilator assignment problem. afternoon.

Ozturk et al. (2010), for example, developed a user-friendly c) Examination rooms with sufficient capacity are

web-based automated system based on a multi-objective mixed available.

integer programming model for exam invigilator assignment The model is focused on scheduling academic staff for the

problem. The system optimised objectives related to exams.

assignment cost, total assignment on individual loads and total

IV. MODEL FORMULATION
assignment on undesired timeslots. The system was tested with
real data provided by the Industrial Engineering Department of In this section, the formulation of the proposed integer
Eskisehir Osamngazi University. Another computer-based programming model for invigilator-exam assignment
system for the exam invigilator assignment was developed by problem is described in detail. Model by Kahar and Kendall
Ong et al. (2009). The system which optimised lecturers’ (2014) constitutes the core of the study’s model, although
preferences allowed lecturers to view the examination there are a few differences stemming from specific
timetable, choose their preferred invigilation timeslots, specify requirements in this study. To provide a better overview of
the examination date and time of their own subjects and view the notations used, the following list contains all the sets,
their individual schedules. In addition, the system enabled indices, parameters and decision variables.
lecturers to give their feedback and any other relevant
information to the invigilation scheduling committee. Sets:
𝑁 Set of all examinations
𝑆 Set of academic staff

21
 ASM Science Journal, Volume 14, Special Issue 1, 2021 for ICSTSS2018

𝑆𝑝 Set of academic staff with administrative post (𝑆𝑝 ⊆ 𝑆) where

𝑆𝑏 Set of academic staff without administrative post

∑𝑇 𝑅
𝑡=1 ∑𝑟=1 𝑞𝑟𝑡 𝑙𝑟
∑𝑇 ∑𝑅 (𝑥 )
(𝑆𝑏 ⊆ 𝑆) 𝑓(𝑥𝑠𝑡𝑟 , 𝑦𝑠𝑡𝑟 ) = {0, 𝑖𝑓 𝑡=1 𝑟=1 𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ≤ 𝑆
1, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑅 Set of rooms
(2)
𝑇 Set of time slots and

Indices: 0, 𝑖𝑓 ∑𝑇𝑡=1 𝑧𝑠𝑡 ≤

∑𝑇
𝑡=1 𝑛𝑡
<0
𝑔(𝑧𝑠𝑡 ) = { 𝑆𝑏 (3)
𝑖 Index for exams, 𝑖 ∈ {1, … , 𝑁} 1, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑠 Index for staff, 𝑠 ∈ {1, … , 𝑆}
The objective function (1) can be rewritten as
𝑟 Index for rooms, 𝑟 ∈ {1, … , 𝑅}
𝑡 Index for time slots, 𝑡 ∈ {1, … , 𝑇}
min ∑𝑆𝑠=1(𝑢𝑠 + 𝑣𝑠 ) (4)

Parameters: subject to
𝑙𝑟 The number of invigilators required in each room 𝑟.
∑𝑇 𝑅
𝑡=1 ∑𝑟=1 𝑞𝑟𝑡 𝑙𝑟
𝑛𝑡 The number of standby invigilators required in each ∑𝑇𝑡=1 ∑𝑅𝑟=1(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) − 𝑀𝑢𝑠 ≤ , ∀𝑠 ∈ 𝑆 (5)
𝑆

time slot 𝑡.
∑𝑇
𝑡=1 𝑛𝑡
ℎ𝑠 1 denotes a senior academic staff and 0 otherwise. ∑𝑇𝑡=1 𝑧𝑠𝑡 − 𝑀𝑣𝑠 ≤ , ∀𝑠 ∈ 𝑆𝑏 (6)
𝑆𝑏

𝑎𝑖𝑠 The exam-staff matrix. 1 denotes that the lecturer

teaches the course in that semester, 0 otherwise. where 𝑀 is a large positive number, 𝑢𝑠 (𝑠 = 1,2, … , 𝑆) and

𝑐𝑖𝑡 1 if examination 𝑖 is scheduled on time slot 𝑡 , 0 𝑣𝑠 (𝑠 = 1,2, … , 𝑆) are indicator variables restricted to be

otherwise. either zero or one.

𝑤𝑖𝑟 1 if examination 𝑖 is assigned to room 𝑟, 0 otherwise.

Constraints:
𝑞𝑟𝑡 1 if room 𝑟 is assigned to time slot 𝑡, 0 otherwise.
The constraints for the examination invigilator assignment
model are briefly listed as follows.
Decision Variables:
𝑥𝑠𝑡𝑟 1 if staff 𝑠 is assigned to invigilate time slot 𝑡 in room 𝑟
a) Invigilators or chief invigilators are not allowed to
as an invigilator, and 0 otherwise.
invigilate examinations of the courses they had taught.
𝑦𝑠𝑡𝑟 1 if staff 𝑠 is assigned to invigilate timeslot 𝑡 in room 𝑟
as a chief invigilator, and 0 otherwise. ∑𝑁 𝑇 𝑅
𝑖=1 ∑𝑡=1 ∑𝑟=1(𝑎𝑖𝑠 𝑐𝑖𝑡 𝑤𝑖𝑟 )(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) = 0 , ∀𝑠 ∈ 𝑆 (7)
𝑧𝑠𝑡 1 if staff 𝑠 is assigned as a standby invigilator for
timeslot 𝑡, and 0 otherwise. b) The total number of staffs to invigilate room 𝑟 in time
slot 𝑡 must be equal to the number of invigilators

Objective function: required for each room.

The objective function of the model is constructed with the aim
∑𝑆𝑠=1(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) = 𝑞𝑟𝑡 𝑙𝑟 , ∀𝑟 ∈ 𝑅, ∀𝑡 ∈ 𝑇 (8)
of finding a schedule that balances the number of invigilation
duties among the staff. It can be formulated as in Kahar and
c) There is only one chief invigilator for room 𝑟 in time slot
Kendal (2014):
𝑡.

min ∑𝑆𝑠=1 𝑓(𝑥𝑠𝑡𝑟 , 𝑦𝑠𝑡𝑟 ) + 𝑔(𝑧𝑠𝑡 ) (1) ∑𝑆𝑠=1 𝑦𝑠𝑡𝑟 = 1 , ∀𝑟 ∈ 𝑅 , ∀𝑡 ∈ 𝑇 (9)

d) Staff without any administrative post are required to be

a standby invigilator not more than once.

∑𝑇𝑡=1 𝑧𝑠𝑡 ≤ 1 , ∀𝑠 ∈ 𝑆𝑏 (10)

22
 ASM Science Journal, Volume 14, Special Issue 1, 2021 for ICSTSS2018

e) No staff is assigned to multiple rooms at the same time. the solution produced by the proposed model while the
third column shows the solution obtained by solving the
∑𝑅𝑟=1(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) + 𝑧𝑠𝑡 ≤ 1 , ∀𝑠 ∈ 𝑆𝑏 , ∀𝑡 ∈ 𝑇 (11) problem manually. It can be clearly seen from the table that
the proposed exam invigilator assignment model could
f) Staff without any administrative post are required to
yield a solution that satisfies all constraints without any cost
invigilate exams not more than 𝑘𝑏 times within the exam
to the objective function. When the solution was compared
period.
to the manually prepared schedule, it was observed that the

∑𝑇𝑡=1 ∑𝑅𝑟=1(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) ≤ 𝑘𝑏 , ∀𝑠 ∈ 𝑆𝑏 (12) schedule from the proposed model was much better in
satisfying all constraints. Note that in the manually
g) Staff with administrative post can only be assigned to one
prepared schedule, several important constraints were
invigilation duty within the exam period.
violated. For example, there were four cases, whereby staff
without administrative post had to be on standby as
∑𝑇𝑡=1 ∑𝑅𝑟=1(𝑥𝑠𝑡𝑟 + 𝑦𝑠𝑡𝑟 ) = 1 , ∀𝑠 ∈ 𝑆𝑝 (13)
invigilators and one case in which a staff had to invigilate

h) Staff can only be assigned as a chief invigilator not more more than two examinations. There were also three staff

than once. with administrative post who had to invigilate more than
once during the exam period. In addition, the requirement
∑𝑇𝑡=1 ∑𝑅𝑟=1 𝑦𝑠𝑡𝑟 ≤ 1 , ∀𝑠 ∈ 𝑆 (14) of avoiding the chief invigilators or invigilators to invigilate
the courses they taught was violated 6 times and the
i) The chief invigilator in a large room must be a senior
constraint of each staff can be assigned as an invigilator
lecturer.
once was violated 7 times. However, by using the proposed
model all these constraints were fully satisfied. Therefore,
∑𝑇𝑡=1 𝑞1𝑡 𝑦𝑠𝑡𝑟 ≤ ℎ𝑠 , ∀𝑠 ∈ 𝑆 (15)
it can be concluded that the proposed model is able to
j) The total number of standby invigilators in time slot 𝑡 must produce a superior solution as compared to the manually

be equal to 𝑛𝑡 . prepared solution.

∑𝑆𝑠=1 𝑧𝑠𝑡 = 𝑛𝑡 , ∀𝑡 ∈ 𝑇 (16)

Table 1. Invigilator timetabling result

V. MODEL IMPLEMENTATION

In this section, the researchers present the implementation of

the model into a case of exam invigilator scheduling problem at
Universiti Teknologi MARA (UiTM) Raub Campus. The
problem (for March 2016) consisted of planning 87 different
examinations within 19 days. Two time slots were reserved for
examinations per day. The university has 105 academic staff
who could serve as invigilators during the exams. Eight rooms
of different capacities were available for examinations, and the
total number of invigilators required in each room varied from
a minimum of two to a maximum of five. In the experiment, the
researchers set both 𝑘𝑏 in Equation (12) and 𝑛𝑡 in Equation (16)
as 2.
The optimal solution was obtained in a few seconds by using
the MATLAB R2017b software programme. The results are
summarised in Table 1. The second column of this table shows

23
 ASM Science Journal, Volume 14, Special Issue 1, 2021 for ICSTSS2018

VI. CONCLUSION students and invigilation duties on weekends should be

fairly assigned among invigilators. However, the addition of
The process of assigning invigilators to examinations is a very
these constraints may create certain complexities in terms
challenging task. It is difficult to produce the optimum exam
of achieving an optimal solution. Therefore, further
invigilator assignment schedule by solving the problem
research is required into the use of heuristic or
manually. In this paper, a mathematical programming model
metaheuristic techniques to solve this problem.
for the exam invigilator assignment problem was presented.
The model was formulated by using integer linear
VII. ACKNOWLEDGEMENT
programming approach and consisted of several common hard
constraints that must be satisfied. This model was successfully The authors would like to express their sincere thanks to the

tested by using real data from UiTM Pahang Raub Campus. The Universiti Teknologi MARA (UiTM) Pahang for the

results demonstrated that the model could find an optimal financial support. We are also grateful to the reviewers for

solution that adhered to all requirements. their useful comments and suggestions, which have greatly

Concerning the model constraints, there are several practical helped in improving the paper.

requirements which can be included in the future such as the

number of invigilators must be proportional to the number of

VIII. REFERENCES

Awad, RM & Chinneck, JW 1998, ‘Proctor assignment at examination proctor assignment problem using mixed
Carleton University’, Interfaces, vol. 28, no. 2, pp. 58-71. integer programming’, in Proceedings of the World
Burke, EK, Elliman, DG, Ford, PH & Weare, RF 1996, Congress on Engineering and Computer Science, vol. 2,
‘Examination timetabling in British universities: a survey’, in pp. 961-964.
International Conference on the Practice and Theory of Koide, T 2015, ‘Mixed integer programming approach on
Automated Timetabling, pp. 76-90, Springer, Berlin, examination proctor assignment problem’, Procedia
Heidelberg. Computer Science, vol. 60, pp. 818-823.
Carter, MW 1986, ‘OR practice—a survey of practical Marti, R, Lourenco, H & Laguna, M 2000, ‘Assigning
applications of examination timetabling algorithms’, proctors to exams with scatter search’, in Computing
Operations Research, vol. 34, no. 2, pp. 193-202. Tools for Modelling, Optimisation and Simulation, pp.
Cowling, P, Kendall, G & Hussin, NM 2002, ‘A survey and case 215-227, US:Springer.
study of practical examination timetabling problem’, in Ozturk, ZK, Ozturk, G & Sagir, M 2010, ‘An automated
Proceedings of the 4th International Conference on the multi-objective invigilator-exam assignment system’,
Practice and Theory of Automated Timetabling, pp. 258-261. International Journal of Information Technology &
Erden, C, Kokcam, AH & Demir, HI 2016, ‘A two-phase Decision Making, vol. 9, no. 02, pp. 223-238.
modelling with genetic algorithm for solving exam-invigilator Ong, ML, Liew, LH & Sim, J 2009, ‘Examination
timetabling problem’, International Journal of Engineering invigilation scheduling system in optimising lecturers’
and Technology Research, vol. 1, no.1, pp. 13-22. preference’, in Conference on Scientific & Social Research.
Hanum, F, Romliyah, MA & Bakhtiar, T 2015, ‘Exam Pokudom, N, Boonplob, N & Hond, P 2010, ‘Ant colony
invigilators assignment problem: a goal programming system for exam proctor schedule’, in SCIS & ISIS SCIS &
approach’, Applied Mathematical Sciences, vol. 9, no.58, pp. ISIS 2010, pp. 758-761, Japan Society for Fuzzy Theory
2871-2880. and Intelligent Informatics.
Kahar, MM & Kendall, G 2014, ‘Universiti Malaysia Pahang
examination timetabling problem: scheduling invigilators’,
Journal of the Operational Research Society, vo. 65, no. 2, pp.
214-226.
Koide, T & Iwata, K 2014, ‘Prototype system development for

24