SYSTEM DYNAMICS MODELING USING MULTIMEDIA
Antdnio S. Camara, Francisco Ferreira, Paulo Diogo, Pedro Gongalves and
Jodo P. Silva
Environmental Systems Analysis Group, New University of Lisbon,
2825 Monte deCaparica, Portugal
ABSTRACT
System dynamics models are tools that allow one to explore the quantitative behavior of
systems through time. However, real systems are usually multidimensional with both
quantitative and qualitative variables. Recent developments in digital video and sound
processing suggest the enhancement of system dynamic models with streams of images
and sounds related to the systems those models try to represent. A framework to
integrate systems dynamics modelling and multimedia technologies is proposed herein:
A multimedia systems dynamics water pollution model. is included for illustrative
purposes.
INTRODUCTION
The main goal of a mathematical modeller is to support decision making with his/her
mathematical models. Typically, these models attempt to describe a natural or artificial
system and relate decision variables with impact variables. A set of values for decision
variables defines a strategy; the set of impact variables. composes the objective function.
Applying a mathematical model with such a structure, any decision maker can test
strategies and, eventually, select the best strategy to be implemented. Pnfortnnapelys
they rarely do so because they mistrust models.
Models tend to not include the abstract and subjective system components, are often,
based on mathematical formulations that are not understood by decision makers, and
their user interfaces are usually poor. In short, models tend to be removed from the
systems they try to describe.
Multimedia technologies relying upon the digital processing of images, video and
sound provide platforms for the development of models more realistic and easier to use.
The interface of these technologies and system dynamic models is addresses herein
with the goal of achieving that tight linkage between visual, auditory, textual and
numerical representations that creates the feeling of immersion common to many
computer games and virtual reality systems as discussed in Laurel (1992).
MULTIMEDIA TECHNOLOGIES
Multimedia technologies may be analyzed in two dimensions: multimodal interaction,
consisting of computer interaction modes using the mouse, natural language, voice or
gestures; and multimedia information, including several data types, presented
concurrently and in a correlated manner Guimardes (1993).
The second dimension is the most relevant for the purpose of this work. The goal is to
develop system dynamics models presented and used as multimedia documents. These
documents have two major features: the integration of video and sound in addition to a
system dynamics model; and a non-linear structure.
SYSTEM DYNAMICS '93, 41
Video constitutes the most faithful means of recording the world of change: Morrison
and Morrison (1982). Video may be used to show backgrounds, point scenes or
transitions. It may be natural or synthetically generated. Also, synthetic video may be.
superimposed on video images of natural sights. Video manipulation includes zooming,
transitions (wipes, dissolves, fades) and highlighting Fox (1991).
The role of sound includes the reinforcement or. replacement of visuals, conveying
patterns and signaling exceptional conditions as discussed in Brown and Hershberger
(1992). Stereo sound may also be used to provide the notion of space and sound icons
to create movement or illustrate point scenes as shown in Gaver and Smith (1990).
The large streams of data resulting from adding video and audio can only be handled
through compression and decompression mechanisms. Standards for data compression
and hardware processing such as JPEG and MPEG are discussed in Fox (1991).
The non-linear structure comes from the typical representation of a multimedia
document as a graph where each node contains information (text, images, video or
sound). The nodes are connected by links, which help the user to move from one node
to another. This is the classical structure associated to hypertext presented in Nielsen
(1990) among others. Standards addressing the structuring of hypertext documents
such as HyTime are also presented in Fox (1991).
SYSTEM DYNAMICS AND MULTIMEDIA
To achieve meaningful and enriching multimedia dynamic models one has to establish a
relationship between the information limited numerical representation and the
information rich multimedia model. While the numerical representation deals with level,
rate and auxiliary variables, the multimedia model deals with image and sound objects.
To establish the relationship between variables and objects one has to analyze the
system at the verbal description and causal diagram levels. It is readily shown that it is
impossible to establish a direct correspondence between- some variables and
relationships between variables, and real objects. Mathematical modelling tradition is to
penetrate behind the external appearances of phenomena and identify their underlying
mechanisms. Instead of representing objects and their interactions, models tend to
depict reality as a set of processes. The use of analogical representations, suchas
visual images of objects, is rarely adopted in mathematical modelling.
Even if one would adopt such analogical view, it is obviously difficult to obtain real
images and sounds on the different system states through time. Finally, the
coordination between thé numerical and multimedia representations (windows) is a
significant implementation problem as system dynamics models may produce
substantially diverse results from slightly different initial conditions.
These problems are addressed herein from both methodological and implementation
standpoints. To overcome the first problem, it is proposed to follow some of the
lessons of movie making. Verbal descriptions of problems are movie scripts with actors
(objects) that have roles. These roles include a definition of their behaviors and
interactions (relationships). The more abstract relationships may be only expressed by
voice or natural language. The parallel system dynamics model stems from the script
skeleton: the causal diagram. At this stage, one has to establish the mapping between
the image objects and the correspondent level, rate and auxiliary variables. The
definition of sound objects should be directed towards the signaling of patterns and
exceptional conditions that may arise in the system dynamic model run.
42 SYSTEM DYNAMICS '93
The multimedia model objects change through time. In principle, only the more
significant changes should be displayed in the multimedia window using images or
video, and sound. If for some of the meaningful events, real footage can not be
obtained one has to resort to animation. Images of transitions may be achieved with
popular morphing programs.
‘The coordination between the system dynamics and multimedia models is only
possible, at this stage, by pre-defining the scenarios to be explored with the system
dynamics model. For each scenario, every significant system change activates the pre-
recorded video and/or audio clips.
Implementation of a multimedia system dynamics model was done in this case in a
Apple Macintosh Quadra workstation using Aldus Supercard®. Video processing was
achieved with Adobe Premiere® and sound editing with SoundEdit®. Similar tools are
also available for DOS and Unix environments.
Future developments include extensive user testing to determine appropriate mixes of
numerical, visual and sound information ‘and the creation of an open multimedia
platform. To accomplish this goal, one has to be able to run both the system dynamics
and the multimedia model without a priori coordination. A possible-method to solve this
problem, involves extended cellular automata formulations. and a sound generating
scheme based on images.
APPLICATION
A simplified model was created to simulate the evolution of the Sado's Estuary area.
Four major variables were considered in the model: population, industry, agriculture,
and pollution (figure 1). Relations between these variables were established in order to
translate the development of this particular area since 1960.
population a
migration
bir death =
death rate
= \ polution
birth rate ie) DE
pol ggneration treatment
industry equivalents
= agriculture equivalents
fert application
Figure I. Model diagram
Each step of the model represents one year in real time. The number of steps to include
in one run of the model are entered through the menu "Simulation". In this menu the
simulation is set On or Off.
SYSTEM DYNAMICS '93 4
Each step of the model represents one year in real time. The number of steps to
include in one run of the model are entered through the menu "Simulation". In this
menu the simulation is set On or Off.
w File Edit View Data Statistics Simulation ( fo]
Multimedia Window - movies and sounds iustrate the
most significant changes in the model
Figure 2. Screen example (model stopped)
Figure 2. represents one possible output while the model is running (in this case the
windows shown are those which were visible after the last run of the model.
The Tool Palette, on the left side of the screen includes icons representing a zoom
option (to be used in the map window) and an icon to stop or continue the running of
the model (in this case the stop icon is visible). Other icons represent the output
information to be presented which can also be selected through the menu "View".
After each run of the model the severa] optional outputs, if selected, are updated and
shown on screen: Video clips, charts, tables and maps.
44 SYSTEM DYNAMICS '93,
w File Edit View Data Statistics Simulation @ Bl
Graphs *1
m2: polution 3: agriculture equiv...
1970.00
ie
ws
©
Figure 3. Screen example (model running)
In Figure 3 an example with two output windows is presented. One for multimedia
information (like video clips) and one for the charts of the several variables modeled.
Notice the tool palette on the left side of the screen. Both the map and charts icon are
hilited, showing that these two windows are selected and presented on screen.
While the model is running it is possible to change the level of the variables modeled.
For this it is necessary to stop the model run, through the Simulation menu or by
using the tool palette. Whenever the model is stopped , another palette appears on
screen. This palette.contains four icons representing the input variables of the model:
population, industry, agriculture and pollution. These input variables can also be
selected through the menu “Data”. Clicking each of these icons or menu items allows
the user to input new values for the variables, its characteristics ( e.g., several types of
industries and water treatments) and location. Selecting the proper icon the user can
insert water treatment facilities, and verify the corresponding multimedia effect.
After all the changes are finished the model can run again, using all the updated data.
SYSTEM DYNAMICS '93 45
SUMMARY AND CONCLUSIONS
Simulation models should be easier to use and closer to reality than they usually are.
Multimedia computing allowing the digital processing of video and sound enables the
development of models with better user interfaces and including images and sounds of
the real systems.
The integration of multimedia and system dynamics modelling was discussed herein.
Three major difficulties were identified: the mapping between system dynamics
variables and relationships and image, video and sound objects; the inexistence of such
objects covering the whole system evolution; and, finally, the coordination between
system dynamics and multimedia windows in a multimedia system dynamics modelling
environment.
A proposal to solve these difficulties based on the lessons from movie making and
research in multimedia was presented. To illustrate this proposal a water pollution
model from the Sado estuary was used.
ACKNOWLEDGMENTS
The support of Direcgao Geral da Qualidade do Ambiente through research grant
number 86/91 is gratefully acknowledged.
REFERENCES
Brown, M.H. and J. Hershberger. 1992. Color and Sound in Algorithm Animation.
IEEE Computer, 25, 12:52-63.
Fox, E. 1991. Advances in Interactive Digital Multimedia Systems. IEEE Computer,
24, 10:9-21.
Gaver, W. and R. Smith. 1990. Auditory Icons in Large-Scale Collaborative
Environments. in D, Diaper, ed. Human-Computer Interaction-Interact 90.
Amsterdam: North Holland.
Guimaraes, N. 1993. Ferramentas para a Produgdo de Material Multimedia. Multimediz
93. Porto.
Laurel, B. 1991. Computers as Theatre. Reading: Addison Wesley.
Morrison, P. amd P. Morrison. 1982. Powers of 10. New York: Scientific American
Library.
Nielsen, J. 1990. The Art of Navigating Through Hypertext. Communications of the
ACM, 33, 3:297-310.
46 SYSTEM DYNAMICS '93,
RESUME
ANTONIO da Nobrega de Sousa da CAMARA
Bom May 13, 1954 in Lisbon, PORTUGAL
EDUCATION
Licenciado in Civil Engineering (IST), 1977
Master of Urban and Regional Planning (Virginia Tech), 1979
PhD in Civil Engineering (Virginia Tech), 1982
Agregado (New University of Lisbon), 1992
ACADEMIC CAREER
Assistant Professor, (New University of Lisbon), 1982-87
Post-Doctoral Associate (MIT), 1983
Nato Fellow and Visiting Fulbright Scholar (Comell University), 1988-89
Visiting Associate Professor (Virginia Tech), 1989
Associate Professor (New University of Lisbon), 1987-1992
MAIN RESEARCH PROJECTS
Principal investigator and co-principal investigator of seven research projects
sponsored by JNICT and one project sponsored by the U.S. NSF in the areas of
simulation, decision support systems and geographic information systems,
Principal investigator in the project “Water Quality Management of the Tagus
Estuary", sponsored by the Ministry of the Environment.
Author of the general methodology for the "Environmental Impact Assessment
of the Alqueva Dam",
Member of the Working Group for the National System of Geographic Information.
PUBLICATIONS
30 refereed papers published in journals such as Water Resources Research,
Journal of Environmental Engineering and Journal of Water Resources
Planning and Management, ASCE, Journal Water Pollution Control
Federation, Water Research, Ecological Modelling, Annals of Regional Science,
Journal of Environmental Management, Journal of Forecasting and Simulation.
SOFTWARE
Main author of the program “The Picture Simulator", used in universities and
tesearch institutes in the U.S., U.K., Germany, Holland, Belgium, Italy, France,
India and Mexico.
OTHER
Invited lecturer at MIT, Cornell, Johns Hopkins, Rensselaer, Georgia Tech,
Virginia Tech, Imperial College, Birbeck College, Free University of Amsterdam
and at several workshops and conferences organized by the European Community,
European Science Foundation and NATO.
SYSTEM DYNAMICS '93. a7
