Methods of Modelling People using Discrete-event Simulation
Andrew Greasley
Operations and Information Management Group, Aston University, Birmingham, U.K.
Keywords: Discrete-event, Simulation, People.
Abstract: Discrete-event simulation (DES) is a developed technology used to model manufacturing and service systems.
However, although the importance of modelling people in a DES has been recognised, there is little guidance
on how this can be achieved in practice. The results from a literature review were used in order to identify
examples of the use of DES to model people. Each article was examined in order to determine the method
used to model people within the simulation study. It was found that there are no common methods but a
diverse range of approaches used to model human behaviour in DES. This paper provides an outline of the
approaches used to model people in terms of their decision making, availability for work, task performance
and arrival rate. The outcome brings together the current knowledge in this area and will be of interest to
researchers considering developing a methodology for modelling people in DES and to practitioners engaged
with a simulation project involving the modelling of people’s behaviour.
1 INTRODUCTION
The need to incorporate people, when modelling
systems, is demonstrated by Baines et al. (2004), who
found that the results from a simulation study, when
incorporating human factors, could vary by 35%
compared to a traditional study, when no human
factors were considered. Juran and Schruben (2004)
found that individual difference variables explain as
much as 80% of the variability in the productivity of
serial work-sharing teams. These differences can
either be across individuals or be changes in the
performance of a person over time. An example of the
latter is a change in performance of an individual in
response to direct changes such as training or
environmental changes such as working conditions.
The ability to model people provides the ability to
avoid abstracting away individual differences and
thus achieve a more accurate model for prediction.
The article encompasses the application of
simulation to modelling people using the widely used
techniques of discrete-event simulation (DES).
Although other simulation methods such as Agent
based simulation (ABS) are used to model human
behavior, and indeed are considered by some authors
as more suited to this task (Elkosantini, 2015; Siebers
et al., 2010), the scope of this study is restricted to
DES.
Papers taken from a structured literature review
which reports on academic publications regarding
discrete-event simulation applications that model
people over the 10 years from 2005 to 2014 forms the
basis of this review. The review followed the steps of
a search of the Scopus citation database and filtering
of papers for relevancy using the CiteSpace
visualisation tool, abstract reviewing and full-text
reviewing. The final sample was supplemented by
reference chasing to identify additional papers of
relevance, some of which fall outside of the original
search period of 2005 to 2014.
The data requirements for modelling people in
DES are now defined and used to categorise the
methods employed to model people in the papers
identified in the literature review. The methods are
then assessed in terms of the approaches of human
performance modelling and human behaviour
modelling.
2 METHODS OF MODELLING
PEOPLE IN DES
In order to consider the different aspects of people’s
behaviour we wish to model we define the data
requirements to model a person in a DES model.
These can be categorised as for the data requirements
312
Greasley, A.
Methods of Modelling People using Discrete-event Simulation.
DOI: 10.5220/0006005803120317
In Proceedings of the 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2016), pages 312-317
ISBN: 978-989-758-199-1
Copyright
c
 2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

for any DES model as outlined in Greasley (2004).
Data requirements can be classified as:
• Logic data defining the process flow undertaken
by people in the model including decision points.
Decision points may be modelled using
conditional (if.. then, else) rules or probability
distributions defining the probability of
following alternative process flow routes.
• Process or task durations which define the time
taken by a person to undertake a task.
• Resource Availability – this defines the
availability of a person over time, such as a work
schedule.
• Demand Pattern – the arrival of people into the
simulation, such as customer arrivals.
• Process Layout – a diagram/schematic of the
process which can be used to develop the
simulation animation display.
The methods of modelling people will be
categorised under these data requirements and
considered under four main headings of ‘Modelling
People’s Decisions’ which relates to the logic data
that controls the flow of people through the model,
‘Modelling People’s Availability’ which relates to
resource availability, in this case people’s availability
to do work, ‘Modelling People’s Task Performance’
which relates to the process/task durations which are
defined in the model and ‘Modelling People’s
Arrivals’ which relates to the demand pattern of
people entering the simulation model. The final
category of process layout relates to the requirements
of the simulation animation facilities to display
people which is not considered here.
2.1 Modelling People’s Decisions
The traditional method of modelling people’s
decisions in a DES model is to either implement a
conditional if..then..else rule or to assign a probability
to the decision outcomes. Both of these methods are
normally implemented as generalised to an average
person and do not take into account changes in
behaviour over time. The following articles provide
examples of methods that attempt to model how
people make decisions.
Kokkinou and Cranage (2011) use an online
scenario-based survey to identify relevant variables in
the decision of customers when choosing between
self-service and manual facilities in a hotel check-in
process. A regression equation is derived that
describes an individual’s decision to select self-
service or manual service. Hannah and Neal (2014)
investigate the decision making process for air traffic
controllers faced with multiple tasks referred to as
“on-the-fly” scheduling. The decision process is
treated as a 2-stage process. Tasks are selected for
execution but then considered for either immediate
execution or deferral until a later time. The initial
selection rule used was “first come-first served” but
the model uses an equation incorporating variables
for airspace complexity, conflict duration, workload
and time to deadline for the deferral decision. In
Brailsford at al. (2006) the Theory of Planned
Behaviour (TPB) cognitive model (Ajzen, 1991) was
used for breast cancer screening policies. The TPB
takes empirical data on demographic variables and
personality traits and transforms these to attitudes,
subject norms and perceived behavioural control.
These 3 aspects then lead to behaviour. Majid and
Herawan (2013) investigate the behaviour of staff and
customers in a customer processing system. Proactive
behaviour in response to a busy operation is
considered in terms of staff refusing customer entry,
staff speeding up service and customers skipping the
queue. Elliman et al (2005) looks at the nature of how
people schedule their tasks. In particular people often
have “at-will” tasks which they can decide
themselves when to execute. In this study 4 factors of
task deadline, length of task, customer importance of
task and importance to the business of the task are
identified in the task scheduling process. In summary
the methods identified are as follows:
• Fit empirical data to a probability distribution as
in Majid and Herawan (2013).
• Use empirical survey data to undertake a
regression analysis to form an equation which
can be used to formulate the decision point
(Kokkinou and Cranage, 2011)
• Develop an equation from theory and test by
comparing with real system data (Hannah and
Neal, 2014)
• Use cognitive models and other data to derive a
decision probability distribution (Brailsford et
al., 2006)
• Use empirical data gathered on task and
organisational factors to derive a work schedule
(Elliman et al, 2005)
2.2 Modelling People’s Availability
The traditional method of modelling people’s
availability in a DES model is to treat them as a
resource and assign them as available or unavailable
for time periods during the simulation run. The
following articles provide examples of methods that
attempt to model the factors that affect worker
Methods of Modelling People using Discrete-event Simulation
313

availability.
Lassila et al (2005) investigate the operation of
assembly lines in an automotive plant. Operators
were modelled using a triangular distribution to
represent task durations and a further random
distribution was used to represent the unavailability
of workers due to off-station work tasks or not work
related activities. Neumann and Medbo (2009)
investigate human factors (HF) also called
ergonomics using DES. 2 kinds of HF are considered;
operators work autonomy and operators work
capacity (ability to work at a standard pace).
Autonomy is modelled as the ability to take breaks
(which occur randomly) and operator capacity is
modelled at 50% pace to represent a new employee,
an older employee or an employee returning to work
from injury. Silva et al (2014) models a mixed
automatic/manual assembly line. The variation in
performance of assembly line operators uses
empirical data regarding each operator. Each
operator’s task mean and standard deviation are used
as parameters in a log-normal distribution. Operator
non-availability for shifts plans and lunch breaks are
also modelled. Freudenberg and Herper (1998) use a
central worker disposition mechanism to assign
workers to a task dependent on their availability and
qualification for the task. Freudenberg and Herper
(1998) states one of the main elements of modelling
human behaviour is to explicitly model machine
availability and worker availability separately.
Resources such as equipment and machinery is
normally available permanently (assuming
maintenance and breakdowns are not being
modelled), but a distinction is to model worker
unavailability due to factors such as other tasks or
lunch breaks is incorporated into the model. In
summary the methods identified are as follows:
• Worker availability is modelled as a schedule
and workers are allocated work when available
(Freudenberg and Herper, 1998).
• Workers unavailability due to shift plans and
lunch breaks are modelled as a schedule of
availability within the model (Silva et al, 2014).
• Model worker unavailability by the use of a
probability distribution derived from data on
worker behaviour when undertaking tasks not
related to the scope of the model (Lassila et al.,
2005).
• Factors such as worker autonomy can be
operationalised as having control over the timing
of rest breaks. These breaks are then modelled as
worker unavailability (Neumann and Medbo,
2009).
2.3 Modelling People’s Task
Performance
The traditional method of modelling people’s task
performance in a DES model is to model task duration
as a probability distribution derived from a sample of
process times. The following articles provide
examples of methods that attempt to model the factors
that affect people’s task performance and thus the
task duration.
Mason et al (2005) investigates the operation of
assembly lines in a factory. Empirical data was
collected on operator performance on 10 operations
within the factory. Curve fitting software was used to
fit a distribution to the activity data and the curve that
gave the most reliable fit was the Pearson Type IV.
Baines et al. (2004) investigates the effect of age and
circadian rhythm on worker performance in a
production system. Equations are used to quantify the
decrease in performance due to age and work time in
terms of task durations. The impact on throughput
performance is measured. Colombi and Ward (2010)
assess the task load on people when controlling
unmanned aircraft from a computer terminal.
Operator tasks are decomposed into a series of mouse
and keyboard inputs. The task time for these
keystroke-level inputs are estimated using the
Keystroke-level model (Card et al. 1983). For the
timing of transition movements between tasks around
the computer screen Fitt’s Law was employed (Keele,
1986). For the time taken to choose options on the
computer screen the Hick-Hyman Law (Wickens and
Hollands, 2000) is employed which takes into the
consideration the number of options (pages, menus,
links) available on the screen at any one time. Ilar
(2008) studies the impact of worker competence on
productivity in a highly automated press line. The
model covers both the main processes but also
support processes such as tool preparation, setup and
maintenance processes. Each operator has an
assigned competence level at a particular task based
on empirical data such as interviews with personnel.
Each competence level is adjusted when the operator
performs the task using a learning curve equation.
Wang et al (2013) investigates the potential loss of
output due to training when attempting to increase the
flexibility of workers. The output of workers during
training has been modelled using a learning curve
equation. The performance of a worker is initially set
to a value measuring assembly time per unit. This
value falls, as a task is repeated, until a steady-state
working speed is reached. Empirical data related to
variables such as experience, age and dexterity are
used as parameters for a worker’s learning curve.
SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications
314

Wang et al (2007) models the variation of
performance due to fatigue and skill level of assembly
line operators. A normal distribution is used to model
worker assembly times. The time to walk between
assembly areas is also modelled. Juran and Schruben
(2004) model individual differences by assigning a
probability distribution to the parameters of the
probability distribution for the task. This can be done
indirectly from a sample of workers or directly from
empirical data on factors such as personality and age
for an individual. In summary the methods identified
are as follows:
• Fit human task performance to a generalised
distribution, in this case Pearson Type IV (Mason
et al., 2005).
• Factor in a decrease in performance using an
equation expressing age and circadian rhythm
parameters (Baines et al, 2004).
• Decompose tasks and estimate task time using a
theoretical model (Colombi and Ward, 2010)
• Use empirical data to assign work competence
level to a task (Ilar, 2008)
• Use a learning curve equation to adjust task
performance over time (Ilar, 2008; Wang et al.,
2013).
• Model variation of individual differences as a
probability distribution of the parameter of the
task duration probability distribution (Juran and
Schruben, 2004).
2.4 Modelling People’s Arrivals
The traditional method of modelling people’s arrivals
is as a probability distribution derived from a sample
of interarrival times. The following articles provide
example of methods that attempt to model the factors
that affect people’s arrival behaviour. Brailsford and
Schmidt (2003) investigate the probability of
attendance at a clinic for diabetes from individual
factors such as stage of process, anxiety, knowledge
of disease and educational level. These where
classified into the components of the PECS cognitive
model (Schmidt, 2000) and given a score. A
compliance factor for attendance was then calculated
from the PECS score in combination with the number
of previous visits and a motivation score. Knight et al
(2012) covers the decision making of individual
patients when choosing a hospital to attend and
deciding whether to actually attend that hospital. A
cost function is assigned to each facility-patient pair
dependent on the hospital reputation, waiting list at
the hospital and travel distance to that particular
hospital from a demand node. The human behaviour
element of the model is the assignment of a level of
irrational attraction or repulsion of a patient for a
particular unit. This is derived from a normal
distribution. In summary the methods identified are
as follows:
• Use cognitive models and other data to derive a
probability of arrival (Brailsford and Schmidt,
2003)
• Use a normal distribution to model the
attraction/repulsion to a hospital unit and thus
probability of arrival (Knight et al., 2012)
3 DISCUSSION
As can be seen a variety of methods have been utilised
to model people using DES with no one method being
favoured. The methods have been implemented in a
range of manufacturing and service applications and
attempt to model differences across individuals and
differences in behaviour within an individual over
time. A variety of individual, task and organisational
variables are used to model people’s behaviour. The
methods identified cover the use of empirical data,
mathematical equations, theoretical distributions and
cognitive models. In order to assess the challenge of
implementing these methods their implementation is
considered within two modelling approaches.
The first approach involves modelling the action
of humans in response to a pre-defined sequence of
tasks and is often associated with the term human
performance modelling. Human performance
modelling relates to the simulation of purposeful
actions of a human as generated by well-understood
psychological phenomenon, rather than modelling in
detail all aspects of human behaviour not driven by
purpose (Shaw and Prichett, 2005). In order to
undertake this we will need to identify either the
characteristics of individuals that are affecting the
performance of the organization (e.g. age) or those
characteristics of the task such as workload or those
characteristics of the organisation or environment
such as ambient temperature. A combination of
individual, task and organizational characteristics
may be incorporated in the model. The key challenge
of the human performance modelling method is the
collection of the empirical data required to model the
actions of humans. The difficulty in practice of
collecting this data is reported in Benedettini et al.
(2006). Another issue is model validation, with
Neumann and Medbo (2009) finding difficulty in
obtaining empirical evidence to validate the
operationalization of human performance in their
Methods of Modelling People using Discrete-event Simulation
315

model. This approach covers the use of the methods
of empirical data, mathematical equations and
theoretical distributions.
The second approach to modelling people makes
use of cognitive architectures to represent the
cognitive process of human beings. This involves
modelling how humans actually behave based on
their individual attributes such as perception and
attention and attempts to model the internal cognitive
processes that lead to human behavior. This approach
can be termed human behavior modelling. A number
of architectures that model human cognition, such as
PECS (Schmidt, 2000) and TPB (Ajzen, 1991) have
been developed. The challenge of modelling people’s
human behavior by modelling their internal cognitive
processes is even greater than that of modelling
human performance. Silverman (2004) states ‘there
are well over one million pages of peer-reviewed,
published studies on human behavior and
performance as a function of demographics,
personality differences, cognitive style, situational
and emotive variables, task elements, group and
organizational dynamics and culture’ but goes on to
state ‘unfortunately, almost none of the existing
literature addresses how to interpret and translate
reported findings as principles and methods suitable
for implementation or synthetic agent development’.
Another barrier is the issue of the context of the
behavior represented in the simulation. Silverman
(1991) states ‘many first principle models from the
behavioral science literature have been derived within
a particular setting, whereas simulation developers
may wish to deploy these models in different
contexts’. Further issues are the difficulty of use of
these architectures (Pew, 2008) and the difficulty of
validation of multiple factors of human behavior
when the research literature is largely limited to the
study of the independent rather than the interactive
effects of these factors.
It is clear that modelling people using either
approach presents challenges in terms of gathering
the empirical data necessary in order to drive and
validate these models. Furthermore the need to be
aware of what human performance and human
behavior methods are appropriate and to understand
how they can be deployed for a particular simulation
application adds another skill to the already wide
skillset of the simulation practitioner. These
challenges may imply a team approach to simulation
development when modelling people. In respect to
the challenge of modelling human behavior,
Bruzzone et al. (2007) discuss the need to evaluate the
modelling impact in terms of the cost and workload
required to introduce these aspects.
4 CONCLUSION
This article presents a summary of published work in
the area of modelling people in a DES, categorised
into the main data requirements for this task. The
methods employed are then discussed in terms of the
approaches of human performance modelling and
human behaviour modelling. Methods identified that
implement a human performance modelling approach
include the use of empirical data directly, derived
mathematical equations and derived theoretical
distributions. A method identified that implements a
human behaviour modelling approach is the use of a
cognitive model.
Further work is needed to provide a critical
assessment of the appropriateness and validity of
these methods and to derive a methodology for their
use in a DES study.
REFERENCES
Ajzen, A., 1991. The theory of planned behaviour,
Organizational Behaviour and Decision Processes, 50:
179-211.
Baines, T., Mason, S., Siebers, P. and Ladbrook, J., 2004.
“Humans: the missing link in manufacturing
simulation?” Simulation Modelling Practice and
Theory, Vol. 12, pp. 515-526.
Benedettini O, Baines T and Ladbrook J., 2006. Human
Performance Modelling within Manufacturing Systems
Design: From Theory to Practice. Proceedings of the
2006 OR Society Simulation Workshop.
Brailsford, S., Schmidt B., 2003. Towards Incorporating
Human Behavior in Models of Health Care Systems: An
Approach using Discrete-Event Simulation. European
Journal of Operational Research 150: 19-31.
Brailsford, S. C., Sykes, J., Harper, P. R., 2006.
Incorporating Human Behavior in Healthcare
Simulation Models, Proceedings of the 2006 Winter
Simulation Conference, 466-472.
Bruzzone, A. G., Briano, E., Bocca, E., Massei, M., 2007.
Evaluation of the impact of different human factor
models on industrial and business processes, Simulation
Modelling Theory and Practice, 15, 199-218.
Card, S. K.; Moran, T. P.; Newell, A., 1983. The
Psychology of Human-Computer Interaction, Hillsdale,
NJ:Lawrence Erlbaum Associates.
Colombi, J. M., Ward, J. R., 2010. Human systems
integration: A discrete event simulation of operational
utility for control of multiple small unmanned aerial
systems, Proceedings of the Human Factors and
Ergonomics Society, 3, pp. 2145-2149.
Elkosantini, S., 2015. Toward a new generic behavior model
for human centred system simulation, Simulation
Modelling Practice and Theory, 52, 108-122.
SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications
316

Elliman, T., Eatock, J., Spencer, N., 2005. Modelling
knowledge worker behaviour in business process
studies, The Journal of Enterprise Information
Management, Vol. 18 No. 1, pp. 79-94.
Freudenberg, R., Herper, H., 1998. Simulation of Workers
in Manufacturing Systems, Proceedings of the 1998
Winter Simulation Conference, 951-956.
Greasley, A., 2004. Simulation Modelling for Business,
Ashgate Limited
Hannah, S. D., Neal, A., 2014. On-The-Fly Scheduling as a
Manifestation of Partial-Order Planning and Dynamic
Task Values, Human Factors, 56 (6), pp. 1093-1112.
Ilar, T. B. E., 2008., A simplified modelling approach for
human system interaction, Proceedings of the 2008
Winter Simulation Conference, SCS, pp. 903-908.
Juran, D. C., Schruben, L. W., 2004. Using worker
personality and demographic information to improve
system performance prediction, Journal of Operations
Management, 22, 355-367.
Keele, S. W., 1986. Motor Control, Handbook of
Perception and Human Performance, vol. 2, John
Wiley and Sons, Inc. New York, NY.
Knight, V. A., Williams, J. E., Reynolds, I., 2012.
Modelling patient choice in healthcare systems:
development and application of a discrete event
simulation with agent-based decision making, Journal
of Simulation, 6, 92-102.
Kokkinou, A., Cranage, D. A., 2011. Modeling Human
Behavior in Customer-Based processes: The Use of
Scenario-Based Surveys, Proceedings of the 2011
Winter Simulation Conference, SCS, 683-689.
Lassila, A. M., Saad, S. M., Perera, T., Koch, T., Chrobot,
J., 2005. Modelling and Simulation of Human-Centred
Assembly Systems – A Real Case Study, IFIP
Advances in Information and Communication
Technology, 159, pp. 405-412.
Majid, M. A., Herawan, T., 2013. Modelling Reactive and
Proactive Behaviour in Simulation: A Case Study in a
University Organisation, International Journal of
Multimedia and Ubiquitous Engineering, Vol. 8, No. 6,
pp. 329-338.
Mason, S., Baines, T., Kay, J. M., Ladbrook, J., 2005.
Improving the Design Process for Factories: Modeling
Human Performance Variation, Journal of
Manufacturing Systems, Vol. 24 No. 1, pp. 47-54.
Neumann, W. P., Medbo, P., 2009. Integrating human
factors into discrete event simulations of parallel flow
strategies, Production Planning and Control, Vol. 20
No. 1, pp 3-16.
Pew R. W., 2008. More than 50 Years of History and
Accomplishments in Human Performance Model
Development. Human Factors 50: 489-496.
Schmidt, B., 2000. The Modelling of Human Behaviour,
SCS-Europe.
Siebers, P. O., Macal, C. M., Garnett, J., Buxton, D., Pidd,
M., 2010. Discrete-event simulation is dead, long live
agent-based simulation!, Journal of Simulation Vol. 4
No. 3, pp. 204–210.
Shaw A. P., Pritchett, R., 2005. Agent-Based Modeling and
Simulation of Socio-Technical Systems. In: Rouse WB
and Boff K R (eds). Organizational Simulation, John
Wiley & Sons, Inc: New Jersey. 333.
Silva, E., Donauer, M., Azevedo, A., Pecas, P., Henriques, E.
2014. A case study evaluating the impact of human
behavior on a manufacturing process in-line with
automatic processes by means of a simulation
model, IEEE International Conference on Industrial
Engineering and Engineering Management, pp. 145-149
Silverman, B. G., 2004. Toward Realism in Human
Performance Simulation. In: Ness J W, Tepe V and
Rizer D R (eds). The Science and Simulation of Human
Performance, Elsevier Ltd: Oxford. 472.
Silverman B. G., 1991. Expert critics: Operationalising the
Judgement/Decision Making Literature as a Theory of
“Bugs” and Repair Strategies. Knowledge Acquisition
3: 175-214.
Wang, Q., Sowden, M., Mileham, A. R.
2013. Modelling
human performance within an automotive engine
assembly line, International Journal of Advanced
Manufacturing Technology, 68 (1-4), pp. 141-148
Wang, Q., Owen, G. W., Mileham, A. R. 2007.
Determining numbers of workstations and operators for
a linear walking-worker assembly line, International
Journal of Computer Integrated Manufacturing, 20(1),
1-10.
Wickens, C. D., Hollands, J. G. 2000. Engineering
Psychology and Human Performance (3
rd
Ed.) Upper
Saddle River, NJ: Prentice Hall.
Methods of Modelling People using Discrete-event Simulation
317