Proceedings of the 2000 Winter Simulation Conference

J. A. Joines, R. R. Barton, K . Kang, and P. A. Fishwick, eds.

ON SIMULATION MODEL COMPLEXITY

Leonard0 Chwif Ray J. Paul

Marcos Ribeiro Pereira Barretto

Mechatronics Lab Centre for Applied Simulation Modelling

University of S5o Paulo Department of Information Systems and Computing
AV.Prof. Mello Moraes 223 1 Brunel University
Sao Paulo, 05508-900, Brazil Uxbridge, Middlesex, UB8 3PH, U.K.

ABSTRACT research into simplification in simulation is surprising,

despite the importance of the subject. This still is a “green
Nowadays the size and complexity of models is growing field” of simulation research and much has to be done.
more and more, forcing modelers to face some problems This concern could be seen by some people as a
that they were not accustomed to. Before trying to study historical worry since “nowadays we have computational
ways to deal with complex models, a more important and power to deal with huge and complex simulation models”.
primary question to explore is, is there any means to avoid This may be true, but complex models not only have an
the generation of complex models? The primary purpose of impact on computer performance, but also on all aspects of
this paper is to discuss several issues regarding the simulation modeling, such as managing a simulation
complexity of simulation models, summarizing the project, communication time and resource constraints, etc.
findings in this area so far, and calling attention to this area The discussion in this paper is organized as follows. In
that, despite its importance, appears to remain at the section 2 we examine definitions of a complex model and
bottom of simulation research agendas. measures of complexity. Section 3 tries to pinpoint some
possible reasons for the increasing number of complex
1 INTRODUCTION simulation models. In Section 4, we show several
relationships between some properties of the model,
There is a consensus amongst the simulation community simulation software and modelers (such as validity,
that a simple model is mostly preferable to a complex one. computational performance, animation capabilities,
In fact, “ Model Simple - Think Complicated”, is one of modeler expertise, etc) to model complexity. Based on this,
Pidd’s (Pidd 1996) “Five Principles of Simulation in section 5 we present the advantages and disadvantages
Modelling”. Several other authors have reinforced this of a simple model. In section 6 we show how we can try to
point of view through time (Ward (1989); Yin and Zhou better deal with complexity. Finally, section 7 makes a
(1989), Innis and Rexstad (1983); Law et. a1 (1993); brief summary and draws conclusions.
Musselman (1 993); Robinson (1994); Pedgren, Shannon
and Sadowski (1995)). Salt (1993) asserts that 2 COMPLEXITY OF SIMULATION MODELS
“simplification is the essence of simulation” and Pidd
(1996) is conclusive in his declaration: “complicated Although complexity is in some sense an intuitive concept,
models have no divine right of acceptance”. there is no general definition or single accepted definition
Despite this, for several reasons that we are going to of complexity when applied to a model (Brook and Tobias
discuss further, a proliferation of complex and large 1996). Some definitions (Golay, Seong and Manno 1989)
models has taken place, forcing modelers to face problems relate model complexity with the cognitive aspect, i.e. the
that they were not accustomed to (e.g. how to express, difficulty of understanding the system being modeled.
validate, solve and understand the results of complex Others (Simon 1964) associates the complexity of a system
models). These problems were called by Nicol (Page et al. with the number of parts and elements that it contains.
1999) “Problems of Scale”, and he indicated that they Using the same line of thinking Ward (1989) defines model
should be part of the roadmap of simulation research. simplicity with regard to the concepts of “transparency”
This paper’s primary intention is to contribute to a (related to understanding) and “constructive simplicity”
necessary discussion about complex simulation models. (related to the model itself). Here we focus on the
According to Brooks and Tobias (1999) the scarcity of

449
 Chwij: Bawetto, and Paul

constructive complexity aspect of the model, although incorporation of more objects and elements (we
some aspects of “transparency” will also be discussed. will discuss this matter in the next section).
Very often the complexity of a simulation model is 2. “Include all” Syndrome: it attacks more - ,
confounded with “level of detail” (Webster et al. 1984). In inexperienced modelers. Since they feel
fact “level of detail” is one of complexity’s components “insecure” about what to include in the simulation
whereas “scope” is another. As an example, consider when model, they follow the maxim “It is better the
simulating manufacturing systems, then either the entire excess than the lack” having the tendency to
facility or just one work-center can be modeled. “Scope” is include everything that is possible in the model.
reduced in the latter case when compared to the former. On 3. “Possibility” factor: Because of increasing
the other hand, the work-center could be modeled with an computation power (approximately the speed of
associated processing time or processing times, CPU and memory size doubles its capacity every
breakdowns, shift patterns, material handling equipment, 18 months, according to the old but still in effect
etc. In this case the “level of detail” is much higher Moore’s law), complexity and size is not a
although the scope remains the same (the same work- constraint on building a simulation model
center). The reader should refer to Robinson (1994) for a - anymore. So some people, as amazing as it may

detailed discussion on “Scope” and “Level”. seem, can create complex models, just because
Complexity measures were defined in order to computer power allows it. The increase of
objectively quantify the complexity of a simulation model. computational power allows better model
Wallace (1987) defined a metric called CAT, which performances during the runs. It also gives birth to
measures the complexity of a simulation model represented better software tools. Paradoxically it is also one
using Condition Specification [see Overstreet and Nance factor that is affecting the growing of complex
(1985) for details]. Yucesan and Schruben (1998) proposed models.
metrics based on Event Graphs representations; one of
them following McCabe’s (1 976) work on the complexity Among technical factors we outline:
within software engineering.
All these measures share common issues. First, they Lack of understanding of the real system: The
are associated with a specific model representation system being modeled must be very well
technique i.e. the simulation model should be described in understood by the modeller in order for him to
the given technique in order to be able to measure the formulate correctly the hypothesis and to consider
complexity. Second, none of them covers all aspects of proper levels of detail. If the system is poorly
complexity in a model. Finally, there are no standard understood, the model will be a result of
measures widely accepted. modeler’s misunderstandings, possibly adding
This clearly leaves room for discussion and research to complications that could be unnecessary.
better define model complexity and the components of Inability to model correctly the problem
complexity. (conceptual model): The lack of ability to model
(or abstract) correctly the problem is common.
3 REASONS FOR INCREASING Modelers who tend to build the model “as close to
COMPLEXITY reality as possible”, including practically
everything seen in real world. Since modeling is
Complex and large models are growing at a significant rate an abstraction of reality, model results (not the
(Arthur et al. 1999). This goes against what the general model itself) should be close to reality, not exactly
simulation community preaches. In the following, we the same. Who needs the complexity of reality in
enumerate some reasons for this fact, dividing them into the model when that complexity was what started
technical and non-technical. Non-technical reasons are the project?
related to human nature and we can cite: Inability to translate or code correctly the
conceptual model into a computerized model
“Show off’ factor: A complex model when or lack of the simulation software knowledge:
shown to managers has more impact than a If the modeler is not totally acquainted with the
simpler one, even if they both could perform the functionality of the simulation software or he has
same job. Furthermore since it is more complex it a lack of good programming skills, he can
has a connotation that it was more difficult to generate the code in a very complex way. This
build, valorizing in some sense the modeler’s job. can .lead to a more complex computerized
Besides this, the great focus on animation features simulation model.
could lead to a complex model due to the Unclear simulation objectives: There is
unanimity among several authors [Innis and

450
 Chwif; Barratto, and Paul

Rexstad (1983); Yin and Zhou (1989); Salt plausible reason that their models are better “polished” and
(1993)l that this is one factor that contributes the thus simpler.
most to the growth of complex models. Unclear or
poorly defined objectives directly affects 4.2 Complex Models and Model Validity
complexity “scope” component of a given model.
As Salt (1993) points out: “Where the overall aim It seems there is no conclusive answer about this
(of the simulation model) is poorly defined, the relationship reported in literature. According to Zeigler
anxious simulationist may draw the bounds of the (1984), a more complex model can apparently represent
model too wide, in the hope of including whatever reality better, although we have to agree with Salt’s (1993)
it is that the user is really interested in”. point of view that it is possible to create a complex and
fully detailed model that is completely imprecise.
Another point related to the last item is the fact that we According to Lobiio and Porto (1997) the relationship
are usually tempted to focus on the system itself instead of between “level of detail” (which is a component of
having always in mind the objectives of the system. In complexity) and model confidence (which is directly
order to try to reinforce this fact, we conducted an related to validity) follows a tendency shown in figure 1.
experiment involving 17 people with various knowledge of Nevertheless, all of these statements are hypothesis based
simulation modeling, from very little knowledge (less than on “experience” and “good sense” on how validity is
1 year) to experienced (more than 5 years). We distributed related to complexity, because no extensive experiments
questionnaires containing a written description of a have been reported. There is a possibility to explore this
problem. Then the respondent was asked to draw a relationship further, since it receives so little attention in
communicative model in an ACD (Activity Cycle the literature.
Diagram) from the description. The problem consists
basically of three machines in serial with infinite buffers 4.3 Complex Models and Computer Performance
between them. The objective of this model is to determine
the utilization of machine two (the machine in the middle). Even though there are no conclusive studies, this
All the people involved in this test (including beginners, relationship is quite obvious: computer performance
intermediate and expert modelers) included all the three decreases as model complexity increases. However, the
machines. Since this model has no constraint on buffer shape of this relationship (linear, polynomial, exponential,
size, it could be modeled with only the first two machines, etc) is unknown, as are those components of complexity
because the behavior of machine three does not affect that mostly affect computer performance (Brooks and
machine two. Although this example is very simple and the Tobias 1996). We did some experiments (reported in
number of persons involved did not allow us to reach a Chwif, Barretto and Santoro 1998) in which a simplified
conclusive response in statistical terms, it reinforces our version of a given model (yielding the statistically
hypothesis that modelers have a tendency to focus on the equivalent results) ran 8 times faster.
“system structure” instead of on the “simulation

T
Model ConJiience
objectives”.

4 RELATIONSHIPS OF COMPLEXITY

In this section we discuss briefly some of the relationships

Confidence Confidence
of complex models with regard to several aspects within increases Decrcses
the M&S area (Modeling and Simulation). These will help
Level of Derail
us to draw conclusions about the advantages and
disadvantages of a simple model (section 5) and raise some Figure 1 : Relationship Between
discussions about model complexity. “Level of Details” and “Model
Confidence” (Lobiio and Porto 1997).
4.1 Complex Models and Expert Modelers
4.4 Complex Models and Simulation Study Total Time
As we saw in section 3, inexperienced modelers have a
tendency to develop more complex models than expert It is intuitive that a simple model is easier to code, validate
modelers mostly due to the “Include all Syndrome”. It has and analyze. Thus in general complexity will increase the
been shown (Willemain 1994, Willemain 1995) that expert time to perform a simulation study (including
modelers tend to concentrate their .modeling efforts mostly conceptualization, implementation and analysis). In
at the conceptual phase of modeling, and this should be a another experience reported in Chwif, Barretto and Santoro
(1998), we created two models: a complex and a simple
model of the same system. A reduction of more than 50%

451
 Chwif; Barretto. and Paul

of the total simulation study time was found (39 hours to method for determining the best complexity level
1’7 hours). This kind of reduction could be achieved when of a given model that still maintains its validity.
the model is “born” simple, but if we think about a model 2. Problems of Scope Reduction: In some cases a
that we attempt to simplify later, since a great effort could simpler model could be achieved by reducing its
be spent on this process (Rextad and Innis 1985), this result scope. For example in an AGV system, a model
would not hold. that can deal with up to five AGVs could be
simplified to deal with only two. In Zeigler’s
4.5 Complex Models and Graphical Animation Terminology this means to reduce the
“Experimental Frame” of a model (Zeigler 1976).
Despite the importance of animation and visualization on This could be translated also into less flexibility
model understanding, a model could have its complexity of the simpler model.
increased if it is “driven” for animation. Hence a series of 3. Difficulties in understanding: In some case a
elements and building blocks have to be put in the model very simple model could be achieved by applying
for “easing understanding” and “closing the model to one great human ability: abstraction. This is great
reality”. For instance, one transport process could be for the modeler because he understands his model
modeled simply by assigning a timed distribution to it with by knowing exactly what kind of abstractions he
practically no animation, or a complete path of a forklift made. Then when he shows it to another person,
truck has to be built in order to show the forklift truck the model could be considered totally
movement. . “unintelligible” and it then becomes necessary to
decrease the level of abstraction of the model (to
5 ADVANTAGES/DISADVANTAGES ease the peoples’ part understanding). So a
OF SIMPLER MODELS simpler model is not always easy to understand.

Robinson (1994) reported two real cases of simulation The complexity of a simulation model has two sides.
modeling which made him conclude that “small is In the next section we present some “rules” of thumb to try
beautiful”. In fact, based on our previous discussions, we to cope with complexity.
can enumerate some of the advantages of a simpler model:
6 HOW TO TACKLE COMPLEXITY
1. It is easier to implement, validate and analyze.
Salt (1 993) also pointed out that it is much easier This section provides some rules of thumb based on the
to “throw away” a simpler model if it is wrong or literature and our own experience.
not reliable, because “it is much harder to admit
the failure of a million dollar systems than a 1. Keep it Simple: Always try to pursue the simpler
thousand dollar one”. model for your purposes (having also in mind the
2. It is easier to “change” a simpler model than a problems described in the previous section).
complex one if the conditions and hypothesis of Remember: if the model is born simple, there will
the systems change (as they usually do!). be gains throughout all the simulation study. In
3. The time to complete a simulation study could be the literature we found some possible guidelines
reduced with a simpler model. In fact according to to make a model simpler (Yi and Zhou 1989,
Pedgren, Shannon and Sadowski (1995), it is Innis and Rextad 1983, Robinson 1994, Pedgren,
infinitely better to have results (even Shannon e Sadowski 1995, Frantz 1995).
approximate) of a simpler model before the 2. Add complexity later: If you are in doubt of
deadline of the simulation study than to have the including some element or factor in a model, do
results of a highly complex model after the the following: do not include it; assume the
deadline. hypothesis it will not affect model results. Then
only after you validate, analyze and have the
Simple models, however, are not always the best results, include this if you feel it is really
choice. Below we summarize a list of problems of simple necessary. Always ask the question: Is this strictly
models: necessary? Instead of using “It is better the excess
than the lack” try this “It is worthier the necessary
1. Problems of Validity: Paraphrasing Einstein: A minimum than the possible maximum”. This
model must be as simple as possible, but not complies with Pidd’s (Pidd 1996) second
simpler. It must be complicated if necessary, but principle of modeling which is “be parsimonious,
not so much. So an over simplified model could start small and add”.
lead to a validity loss. Unfortunately there is no

452
 Chwg Barretto, and Paul

3. The complex model was already’built. Is there 3. There is a substantial increase in the complexity
any means to simplify it? Unfortunately, once a in our current simulation models. One factor that
complex model is created, the efforts to simplify it surely contributes to this is the development of
could be unworthy. Moreover there is a lack of powerful computer hardware.
simplification procedures related to simulation 4. Several relationships that involve complexity are
models reported in the literature. Chwif, 1999 not explored in detail in the literature.
developed an automated method that tries to 5 . Despite some very good “guidelines” there is a
simplify an existing model based on its objectives. lack of methodologies to lead a modeler to obtain
Although his algorithm reached in some case a simpler model.
reductions in complexity of almost SO%, the
simulation model has to be described using a Since complexity will be a constant in our simulation
specific simulation model representation models, Nicol (Page et al. 1999) mentions one possible
technique. future line of research which would try to answer the
4. “Reduce” the Level of Detail using Hierarchy: question “HOW can we deal with complex and huge
Hierarchy is considered one method of model simulation models?” On the other hand we have to keep in
abstraction (Luna 1993) and thus can “simplify” mind that there are a number of advantages if we were able
the simulation model. The use of hierarchical to obtain a simpler model. Therefore the real question
modeling can be crucial for the manageability of might be “How can we avoid the generation of complex
complex models (Daum and Sargent 1999). Since models?”
hierarchy sometimes could not simplify a model
in terms of “constructive simplicity”, we use the ACKNOWLEDGMENTS
words “Reduce” and “Simplify” in quotation
marks. That is because the number of elements in The authors would like to thank the Department of
a model that is hierarchically constructed could be Information Systems and Computing, Brunel University,
the same as in a “flat” model, with the difference who provided the resources necessary for part of this
that some of them are “hidden” by the hierarchy. research, whilst Leonard0 Chwif was a Research Visitor.
A simplification could be achieved if it is possible Thanks also to the Brazilian’s Research Funding Agencies
to aggregate some portions of the model (e.g. a set CAPES who supported partially this research (under
of machines becomes one “big” machine). reference no. 0439198-8) and FAPESP.
5. Reduce the Scope of the Model: Another
possibility to try to obtain a simpler model (or REFERENCES
many, various simpler models) is to attack the
scope component of complexity. In this case, Arthur J. D. et al. 1999. Verification and validation: what
divide your system into parts and model each part impact should project size and complexity have on
separately creating a series of simpler models attendant V&V activities and supporting
instead of one “huge” model. Once these parts infrastructure? In Proceedings of the 1999 Winter
pass through all phases of the simulation study Simulation Conference, ed., P. A. Farrington, H. B.
and if and only if there is a need, integrate these Nembhard, D. T. Sturrock, and G. W. Evans, 148 -
models into a bigger one. Another possibility is to 155. Institute of Electrical and Electronics Engineers,
reduce the scope of a model for analyzing a more Piscataway, New Jersey.
specific and urgent question. But remember this Brooks, R. J. and Tobias, A. M. 1996. Choosing the best
coarsening in scope can lead to less flexibility of model: level of detail, complexity and model
the model. performance, Mathl. Comput. Modelling, 24(4): 1-14.
Brooks, R. J. and Tobias, A. M. 1999. Methods and
7 SUMMARY AND CONCLUSIONS benefits of simplification in simulation, In:
Proceedings of UKSIM, United Kingdom Simulation
Based on our discussions the most important issues on Society, 88-92.
simulation model complexity are the following: Chwif, L. 1999. Discrete event simulation model reduction
in its conceptualization: a causal approach. Ph.D.
1. The simulation community is pro to the simplicity Thesis (Portuguese), Department of Mechatronics
of simulation models. Engineering, University of Siio Paulo.
2. There is no widely accepted definition of what a Chwif, L., Barretto M.R. and Santoro, M. C. 1998. Model
complex model is. There is also no general Reduction Some Results, In: Proceedings of the 31’‘
complexity measure of a given simulation model. Annual Simulation Symposium, 120-125. IEEE
Comput. Society,. Los Alamitos, CA, USA.

453
 Chwg Barretto, and Paul

Daum, T and Sargent, R.G. 1999. Scaling, hierarchical Pegden, C.D., R.E. Shannon and R.P. Sadowski, 1995.
modeling, and reuse in an object-oriented modeling Introduction to simulation using SIMAN, 2nd
and simulation system, In Proceedings of the 1999 Ed.,New York: McGraw-Hill.
Winter Simulation Conference, ed., P. A. Farrington, Pidd, M. 1996. Five simple principles of modelling, In
H. B. Nembhard, D. T. Sturrock, and G. W. Evans, Proceedings of the 1996, Winter Simulation
1509-1520. Institute of Electrical and Electronics Conference, ed. J. M. Charnes, D.M. Morrice, D. T.
Engineers, Piscataway, New Jersey. Brunner, and J. J. Swain, p. 721-728. Institute of
Frantz, F. K. 1995. A taxonomy of model abstraction Electrical and Electronics Engineers, Piscataway, N.J.
techniques, In: Proceedings of the 1995 Winter Rextad, E. and Innis, G.S. 1985. Model simplification -
Simulation Conference , ed. C.Alexopoulos, K. Kang, three applications”, Ecological Modelling, 27(%): 1-13.
W. R. Lilegdon, and D. Goldsman,. 1413-1420. Robinson, S. 1994. Successful simulation - a practical
Institute of Electrical and Electronics Engineers, approach to simulation projects, Maidenhead, UK.
Piscataway, New Jersey. McGraw-Hill.
Golay, M. P., Seong, P.H. and Manno, V.P. 1989. A Salt, J. D. 1993. Keynote address: simulation should be
measure of the difficulty of systems diagnosis and its . easy and fun!, In Proceedings of the 1993 Winter
relationship to complexity, International Journal of Simulation Conference,. ed. G. W. Evans, M.
General Systems, 16(1): 1-23. Mollaghasemi, E. C. Russell, and W. E. Biles, 1-5.
Innis, G . and Rexstad, E. 1983. Simulation model Institute of Electrical and Electronics Engineers,
simplification, Simulation,41(1): 7-15, July. Piscataway, New Jersey.
Law, A.M., et al. 1993 A forum on crucial issues in the Simon, H.A. 1964. The architecture of complexity”,
simulation of manufacturing systems, In Proceedings General Systems Yearbook, 10: 63-76.
of the 1993 Winter Simulation Conference ed. G. W. Wallace, J.C. 1987. The control and transformation metric:
Evans, M. Mollaghasemi, E. C. Russell, and W. E. toward the measurement of simulation model
Biles, 91 6-922. Institute of Electrical and Electronics complexity, In Proceedings of the 1987 Winter
.Engineers, Piscataway, New Jersey. Simulation Conference, ed. A. Thesen, H.Grant, and
Loblo, E.C. and Porto, A.J.V. 1997. A simulation study W. D. Kelton, 597-603.Institute of Electrical and
systematization. (Portuguese) In: Proceedings of the Electronics Engineers, Piscataway, New Jersey.
XVII ENEGEP - NATIONAL CONGRESS OF Ward, S.C. 1989. Arguments for constructively simple
INDUSTRIAL ENGINEERING, Gramado, Rio Grande models. J. Opl. Res. Soc. 40(2): 141-153.
do Sul, Brazil. Webster et al., 1984. Determining the level of detail in a
Luna, J. J. 1993. Hierarchical Relations in simulation simulation model - A case study. Computers and
models. In: Proceedings of the 1993 Winter Simulation Industrial Engineering, 8(3/4): 2 15-225.
Conference, ed. G. W. Evans, M. Mollaghasemi, E. C. Willemain, T.R. 1994. Insights on modeling from a dozen
Russell, and W. E. Biles, 132-137. Institute of experts. Operations Research, 42 (2): 213-222.
Electrical and Electronics Engineers, Piscataway, New Willemain, T.R. 1995. Model formulation: what experts think
Jersey. about and when. Operations Research, 43 (6): 916-932.
Mc Cabe, T.J. 1976. A Complexity Measure, IEEE Yin H.Y. and Zhou Z.N. 1989. Simplification techniques
Transactionson Software Engineering, 2(4): 308-320. of simulation models, In Proceedings of Beijing
Musselman, K.J. 1993. Guidelines for simulation project International Conference on System Simulation and
success. In Proceedings of 1993 Winter Simulation Scientific Computing, 782-783.
Conference, ed. G. W. Evans, M. Mollaghasemi, E. C. Yucesan, E. and Schruben, L. 1998. Complexity of
Russell, and W. E. Biles, p.58 - 64. Institute of simulation models: A graph theoretic approach,
Electrical and Electronics Engineers, Piscataway, New INFORMS Journal on Computing, lO(2): 94-106.
Jersey. Zeigler, B.P. 1976. Theory of Modelling and Simulation,
Overstreet, C.M. and Nance, R.E. 1985. A specification John Wiley and Sons, Inc., New York.
language to assist in analysis of discrete event simu- Zeigler, B.P. 1984 Multifacetted Modelling and Discrete
lation models, Communications of the ACM, Event Simulation. London: Academic Press.
28(2):190-201, February.
Page, E. H. et al. 1999. Panel: strategic directions in AUTHOR BIOGRAPHIES
simulation research, In Proceedings of the 1999
Winter Simulation Conference, ed., P. A. Farrington, LEONARD0 CHWIF graduated in Mechanical
H. B. Nembhard, D. T. Sturrock, and G. W. Evans, Engineering (Mechatronic Specialization) in 1992 at the
1509-1520. Institute of Electrical and Electronics University of Sao Paulo and got his MSc degree in 1994
Engineers, Piscataway, New Jersey. and his Ph.D. in Simulation in 1999 from the same
University. .He spent a session at Brunel University as a

454
 Chwif: Barretto, and Paul

research visitor at the Centre for Applied Simulation

Modelling. Upon graduation Dr. Chwif joined the Brazilian
branch of Mercedes-Benz truck manufacturer and then he
joined the Brazilian branch of Whirlpool Corporation. His
research interests includes simulation, optimization and
simulation model representation techniques. His email is
<[email protected]>.

MARCOS RIBEIRO PEREIRA BARRETTO graduated

in Electronics Engineering in 1983. He got his MSc degree
in 1988 and his Ph.D. in 1993, all from the University of
Sao Paulo. He is currently Assistant Professor at the
Mechanical Engineering Department of the Polytechnic
School of the University of Sao Paulo. His research
interests include manufacturing integration and CIM
technology. His email is <mrpbarre@usp. br>.

RAY J. PAUL has been Professor of Simulation

Modelling at Brunel University since 1992, Head of the
Information Systems and Computing Department 1993-
1998, and Dean of the Faculty of Science since 1999. He
previously taught information systems and operational
research for 21 years at the London School of Economics.
He has published widely in book and paper form (three
books, over 200 papers in journals, books and conference
proceedings), mainly in the areas of the simulation
modeling process and in software environments for
simulation modeling. He has acted as a consultant for a
variety of U.K. government departments, software
companies, and commercial companies in the tobacco and
oil industries. He is currently running a simulation research
group, the Centre for Applied Simulation Modelling, made
up of five faculty members and eight research students. His
email and web addresses are R a y . Paul@
brunel.ac.uk and <www.brunel.ac.uk:
/-csstrjp>.

455