THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 51
BEYOND DYNAMO : FUTURE SYSTEM DYNAMICS SIMULATION PROGRAMMING
Raj Kumar Bapna
International Software India Limited
SDF Block 1, Units 5 & 6, M.E.P.Z.
Madras-600 045 India
Sujoy Ghose
Department of Computer Science and Engineering
Indian Institute of Technology
Kharagpur-721 302 India
Sushil K. Sharma
Department of Mechanical Engineering
institute of Technology, B.H.U.
Varanasi-221 005 India
ABSTRACT
Advances in computer technology, since DYNAMO first appeared,
promise significant developments for System Dynamics modeling and
simulation. This paper discusses the future System Dynamics
simulation languages as a part of human-engineered, integrated
simulation programming environment. Many features of such
languages and environments, that may become available in the next
five years, are identified and discussed. The relevant advances
in Simulation and Computer Science are presented with appropriate
modifications in the context of System Dynamics methodology. Some
aspects of implementation of such languages and environments are
also discussed.
1 INTRODUCTLON
DYNAMO was designed over 25 years ago for simulation of the
System Dynamics (Sl: models and it continues to be the most
widely used simulation language for SD studies (Forrester 1961,
Richardson and Fugh 1981). It is a compliment to the creators of
DYNAMO that it, has remained largely unchanged over the years, and
that other languages like, DYSMAP (Coyle 1977), DYMOSIM
(Mohapatra and Bora 1983), NDTRAN (Uhran and Davisson 1984) are
similar to DYNAMO.
DYNAMU was initially developed for batch processing and program
entry was done with computer cards. It was later adapted. for
interactive use with computer terminals. However, it did not
expleit the advantages of the interactive mode (and
graphical/visual communication). Computer technology has greatly
developed since DYNAMO first appeared. Hardware has become
cheaper and more powerful and software has become more versatile
and canvenient to use. These developments promise new
possibilities of rapid developments for SD modeling and
52THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
simulation. The future SD simulation will be fundamentally
similar to what it is with DYNAMO. However, significant
developments in SD modeling and simulation in the near future
will emerge primarily in the following two ways:
(1) Application of currently well established computing
techniagues to provide a number of tools in an integrated SD
simulation programming environment. Typical in this category are
dimensional analysis facility and spreadsheet-like program
execution monitoring facility
(2) Development of fundamental concepts and techniques in the
fields of SD modeling and simulation. Typical in this category is
a facility to qualitatively/quantitatively analyze the SD model
performance.
Such developments are likely to happen ata faster pace for
SDSLs(System Dynamics Simulation Languages) than for other
simulation languages (because implementing SDSLs is easier due to
their simple syntax), and make SD and SDSLs more popular and wide
spread in use. Other developments will come from applications of
other fields, e.g., Artificial Intelligence.
Some work has been done on advanced and futuristic concepts in
SD, simulation and software (Aus and Korn 1971, Baltzer 1983,
Coyle and Sharp 1976, Ghose, Chakrabarti and Bapna 1987a, Hooper
1984, Oren and Ziegler 1979, Randers 1980). This paper primarily,
addresses those methods, techniques, features, or facilities
which are’ “likely to become available during the next five years
in the SDSLs.
1.1 Brief History of System Dynamics and DYNAMO
Prof. Forrester pioneered the philosophy, principles, techniques
and applications of System Dynamics (Forrester 196171969, 1973).
DYNAMO was developed at MIT over 25 years ago as the first
simulation language for SD studies. Though DYNAMO is still the
standard language for SD studies, there are other languages like
DYSMAP (Coyle _1977) and DYMOSIM (Mohapatra and Bora 1983). which
also were specifically developed for SD studies. In addition,
general purpose continuous system simulation languages are also
being extensively used for the same purpose (Gordon 1982).
Various types of studies on SD methodology have been conducted.
Such studies include eigenvalue analysis (Forrester 1983),
evitical parameter identification (Starr 1981), sensitivity
analysis (Vermeulen and De Jongh 1977), parameter estimation
(Thillainathan and Price 1981), objective function specification
(Rohrbaugh and Andersen 1983)° and parameter optimization (Birta
1977). There have been a large number of successful applications
of the SD approach. SD is also considered as a useful philosophy
for modeling and simulation of any system in general (Roberts
1983). Use of both SD methodology and DYNAMO has spread to many
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 53
areas over the years - SD and DYNAMO have even been found useful
as tools for teaching simulation to children (Roberts 1983).
1.2 Relevant Trends in Simulation and Computer Science
Computer Science community seems to be directing its efforts for
designing computer systems keeping the user in mind (Raeder
1985). The main reason for this is the wide use of powerful yet
inexpensive micro computers. Better user interfaces are being
designed by using graphics and using results of Artificial
Intelligence research. Typical examples of the current trend are
Macintosh (Benzon 1985) computer, UNIX (Kochan and Wood 1984)
operating system, Ada (Buxton 1980) and Small Talk (Goldberg and
Robson 1983) programming languages. Also there is a trend to
provide a set of tools in an integrated environment, e.g.,. UNIX.
Unlike the traditional simulation languages, like, GPSS, SIMULA,
SIMSCRIPT etc, the current trend in simulation languages attempts
to graphically draw a model on the screen, and then study its
behavior through animation and dynamically evolving statistics.
IDSS (IDSS 1983) and SIMAN (Pegden 1982) allow a user to store a
simulation trace and use it for post simulation run animation.
CINEMA (an extension of SIMAN) allows the user to draw the
simulation scene in color and icons. There has been a trend to
include popular features from other simulation languages, e.g.,
both discrete and continuous » system simulation features are
provided in modern simulation languages (Cellier 1979, Hooper
1984). Another area of current interest in simulation is the
application of model bases and knowledge bases.
1.3 Overview of Future System Dynamics Simulation
Many of the fundamental ideas and techniques for the future SDSLs
will come from DYNAMO itself. Such features include the DYNAMO
style of writing equations, automatic sorting of equations, pre
defined functions, reporting of graphical and tabular outputs
etc. However, there will be development of integrated SD
simulation programming environments. In addition to providing
SDSLs, such environments will also provide supporting software
tools. The primary design consideration will be ease of use as
well as power and flexibility.
Many features can be provided by such SD’ environments in future.
User-friendly interfaces can ease the process of programming.
Syntax-directed editors can help enter programs correctly.
Graphical languages can allow writing programs as diagrams. Run-
time editing, debugging can be important features. Assertion-
directed break points and spreadsheet-like facility can provide
easy debugging and monitoring of program execution. Modular
programming can facilitate modeling and simulation of large and
complex systems by teams. Dimensional analysis, unit conversion
and model validation can enforce reliable modeling and
programming. Program generators can permit flexibility and
54 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
[A] sreded area shows those parts which are currently provided by DYNAIO
oD Ellipses show features / facilities
=> Dark ellipses show those features which are unique to SUSLs, i.e., not
available in general purpose simulation / programming languages
— > Arrows show transfer of contro! during program execution
NY
Fig. 1 SCHEMATIC OF FUTURE SYSTEM DYNAMICS
SIMULATION ENVIRONMENT
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 55.
portability. The user can analyze performance of a model
qualitatively as well as quantitatively.
Figure 1 shows the schematic of such an environment. The features
and facilities provided by DYNAMO are shown shaded in the figure.
Eliipses represent features or facilities. Dark ellipsés
represent those additional features which are unique to the SDSLs
and not available in general purpose simulation or programming =:
languages. The user initiates interaction with the environment *
via editor or via program execution. And, as the program
executes, the control passes to various parts through the paths
of the environment indicated by the arrows.
2 HUMAN-ENGINEERED INTERACTIVE SIMULATION ENVIRONMENT
Human~engineering attempts to develop effective user-interfaces
(Be 1982). The design of such interfaces considers various
factors such as psychology of programming (Weinberg 1971),
operator skill, user reinforcement, feedback, consistency,
demands on human memory etc. Errors are dealt with in a
comprehensive and user-friendly way. Human-engineered designs
make extensive use of interactive style of user-computer
communication, computer graphics, visual programming.
Future SDSLs will be human-engineered throughout the entire life
cycle of the SD simulation project/programming - model design, —
program development, program entry, program execution, debugging;
testing and report generation. Ease of. use is likely to be the
primary design criteria for the future SD simulation systems. The
language can be so designed that: a first time user can learn and
use it easily, but at the same time, an experienced user can have.
flexibility and power of programming.
A future SDSL will not be just a language, but it will rather be
an environment. Because simulation (including SD simulation) is
an experimental technique, a human-engineered interactive
simulation environment is of great importance in making
simulation easier, more powerful and in increasing: user
productivity.
2.1 Interactive User-friendly Interfaces
Interactive user-friendly interfaces for SD can use several
techniques which are already popular. in the computer industry. -
Some of them are discussed below, and others are discussed
elsewhere in this paper. vRB,
Syntax: The syntax of commands as well as the SDSLs can be
uniform and concise. Most commands can also, have abbreviated
forms.
Variety and choice: A rich set of features may be provided so
that most of the operations can be done in more than one way.
56 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
This extra choice enables a user. to work in a way which is
natural to him.
Invocation command with options: .An invocation command, that
accepts options, can allow auser to. select options without
making changes to the program.
On-line help and tutorial: On-line help allows a: user to get
help-information on the screen while he is using the computer
system interactively. On-line tutorial with practice sessions
makes {t easy to learn and use the system.
Report generation: This allows the user to print the SDSL source
program as well as the tabular and graphical outputs in a style
he like. j
Defaults:. To provide power and flexibility of use, a rich set of
features are needed. However, most of the time, the user needs to
know little about language constructs; commands and arguments.
When unspecified by the user, a suitable option is assumed as
default. The defaults are set to normally used values.
Mouse and menus: A menu is a list of various options displayed on
the screen from which a choice can be made. The selection is made
by typing a code or using a special selection device called
mouse. A primary benefit of menus'is that they do not require the
user:to memorize the commands and their syntax. Further, menu
based systems encourage system exploration 4 reminding the user
of various options available: °
Windows: Graphics-based displays allow ‘the screen'to be divided
inte.many. rectangular areas’ called windows. Each window is
independent of other windows. All windows on the screen can
contain different information simultaneously. Windows can be
created, deleted, overlapped, moved and changed in sizes.
Integrated:environment: Most computing environments (including
the future SD simulation environment) consist of several software
tools; each designed for a'sub task within the overall system. An
integrated environment presents a uniform and systematic view of
all the tools to the user. This makes rapid changes of modes easy
and encourages use of parts of the system that would otherwise
remain unused.
2.2 Run-Time Editing, Debugging and Monitoring
The future SDSLs will execute. programs under the user’s control
and provide facilities like run time editing, debugging and
monitoring. Tracing and assertion-directed break points are
important for debugging. Spreadsheet-like monitoring facility
would be very useful in the context of SD
Run-time error handling/and ‘correction: As soon as an error is
detected during program execution, the program will be suspended.
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY, CHINA 57
Then the user is ghown-appropriate statements with suitable error
messages and suggestions for error correction. The user: may
correct the errors, and then may continue the program execution,
stop the program execution, make more changes, or restart the
program. .
Tracing: As the program executes, the output, as soon-as it is
calculated, may be produced in a tabular or graphical form. This
way the user can trace the execution.
Assertion-directed break points: This is a facility to halt a
program when some assertion or condition becomes true during
program execution. The control is transferred to the user
terminal. The user can set new break points, change values of
variables, edit the program, resume execution, or view values of
various variables using a spread-sheet like facility.
Spreadsheet-like monitoring: The concept of spreadsheet was
popularized by a package called Lotus 1-2-3 (Kelley 1983). A
spreadsheet allows viewing two dimensional arrays on the screen
as cells arranged in rows and columns. The spreadsheet can
contain a large number of rows and columns and a portion of that
is shown on the screen at a time. In SD context, each cell can
keep a numeric value. Each column of the spreadsheet can keep
values of one particular variable corresponding to different
times. And each row can keep values of different variables
corresponding to some particular time. Spreadsheet can offer
another important facility, i.e., the values of one or more
variables may be changed and simulation may be automatically
repeated for the necessary iterations to update all the cell
values. This facilitates answering “what-if” questions. It is
possible to insert or remove rows or columns; to instantly create
a graph for certain variables.and view the graph; and to perform
statistical calculations such as max, min, mean etc.
2.3 Syntax-Directed Editing
Text-editors are used for entering and modifying data and
program. Syntax-directed editors are text-editors but they also
have knowledge of the particular language syntax, i.e., the rules
defining the correct sequences of language elements. Thus they
allow only syntactically correct statements, equations or
language elements to be entered.
Syntax-directed editing has several advantages:
(1) It reduces typing effort and minimizes typing errors. This
reduces program development time.
(2) It ensures that a program is syntactically correct.
(3) It provides an interface for debugging tools.
58 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
Syntax-directed editors for S5DaLs can be designed to be
particularly more effective (than those for general simulation or
programming languages) because of the simple syntax of SDSLs.
2.4 Graphical Programming and Visual Programming
All the currently used SDSLs require. the user to convert the
model specifications from some kind of a causal loop diagram to a
set of equations. The future SDSLs may allow programs to be
entered graphically. Suitably modified causal loop diagrams may
be used for this purpose. Such a graphical programming language
offers a more natural way of _SD modeling and simulation. This
should result in reduced program development time, increased
reliability and increased understanding of the program by the
user.
Visual programming (Melamed and Morris 1985) attempts to use
visual (and graphical) means to show on the screen a program and
its execution. This results in increased understanding of the
program and its behavior.
3 RELIABLE MODELING AND PROGRAMMING
The designs of programming languages have evolved over the past
three decades. The design changes reflect our changing
understanding of good methods of writing large and complex
programs that are reliable and maintainable (Pratt 1984). And it
also reflects the changing programming and execution environment.
3.1 Modular Programming
Features like subroutines, procedures and macros in programming
languages facilitate modular programming. Modular programming
(Pressman 1982) means that a large software problem is divided
into a set of smaller software units or elements called modules.
Modularization utilizes the principle of divide and conquer.
Modularization also simplifies program debugging and testing.
In SD context, modular programming offers several benefits.
Modules developed in current SDSLs require all variable names to
be different and unique. Modular programming will require that
variable names within each module need be unique, but the same
variable name may appear as local-variable in more than one
module (Some global .or inter-module variables need to be known
to more than one module). This simplifies model development and
makes it possible to use some previously developed modules
without major alterations. Modular programming will facilitate
development of “algorithmic control modules" (Wolstenholme 1985a)
and make it possible to easily incorporate them in the SD
programs. The algorithmic control modules are standard
submodules or submodels which fit into many SD models.
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 59
Modular programming facilitates top-down design of software
(Mills 1971), i.e., the software system is initially designed
from a global or a total view, and then details for each module
are worked out at a later time during implementation.
Modularization also facilitates development of large programs by
a team (Pressman 1982).
Further, with the availability of modular programming features,
execution-time profiler. may also be provided. An execution-time
profiler is used to determine how much time is taken for
execution by each module. If some module is found to take too
much time, the model of that module may be simplified. It is
obvious that such changes will make the program run faster, but
may also reduce the accuracy of the model.
3.2 Dimensional Analysis and Unit Conversion
Dimensional analysis, a facility currently provided by DYSMAP’
(Coyle 1977), is used to find out if all the equations in an SD
model are dimensionally correct. For this purpose, the user has
to specify the dimensions of each variable. A natural extension
of dimensional analysis concept is to allow the users to specify
different units for the same type of quantity and also specify
the conversion formulae. This facility called unit conversion
was implemented by us in a DYNAMO-like language (Bapna 1985,
Bapna et.al 1987a).
3.3 Model Validation and Error Diagnosis
A syntax-directed editor (See Section. 2.3) will identify and
help correct most syntax errors. The future SDSLs will perform
extensive error diagnosis and generate user-friendly error
messages. Run-time error correction (See Section 2.2) would also
form an important part of error diagnosis and correction.
Model-validation will be performed to a limited extent since
comprehensive model validation is not yet possible in SD or any
other method _ of simulation. Expert systems are expected to solve
some of the problems in this area.
3-4 Software Maintenance Tools
Computer programs are always changing. There are bugs to fix,
enhancements and optimizations to make, versions to change etc.
Certain tools help maintenance of software.
Cross-reference generator: In the context of SD, cross-reference
generator can provide cross-reference listings showing the line
numbers in which a variable gets a value assigned and all the
line numbers where it appears in an equation on the right hand
side. Such lists can be separately generated for each type of
equation, e.g., A-type, L-type, R-type etc. In case of modular
programming, cross-reference listing for inter module variables
may also be generated.
60 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
Source code complexity measure: Software matrices (Gibb 1977) can
be used to find the complexity of a program in terms of. its
modularity, readability and maintainability. A suitable source
code complexity measure may be provided by the future SDSLs.
Source code version control systems: Source code version control
systems can be used to keep track of history of each module and
this facilitates development and maintenance of large programs by
teams. This. facility is available in gooa operating systems like
UNIX.
4 ANALYTICAL TECHNIQUES AND TOOLS
Availability of analytical techniques and tools for studying SD
models will be another important aspect of SDSLs in future. Such
techniques and tools can help the user to -understand the SD
modeling and simulation better, guide the user to design better
models, and give the user more flexibility and control.
4.1 Program Generators
Some languages are implemented as program generators. For
example, a CSMP program is first translated into FORTRAN program
which is then compiled.
A program generator for SDSL will accept as input a program in
SDSL and produce as output an equivalent program in Fortran,
Pascal, Basic, C or some other high level programming language.
Availability of a program generator will offer many advantages:
(i) Rather than writing a complete program ina high level
language for a special application, the user can write the
initial program in SDSL. Then the user can get an equivalent
program generated ina high level language and make necessary
changes to allow flexibility which SDSL may not allow.
(2) The generated program in high level language is correct
because many logical checks are made by the SDSL processor.
(3) The generated program in high level language can be compiled
with the best available optimizing compiler for high run time
efficiency.
(4) Using the generated program, the SD model can be simulated on
a computer even if SDSL is not available on that computer,
4.2 Qualitative/Quantitative Performance Analysis
At present, using DYNAMO or other SDSL, the graphical outputs (or
tabular outputs) have to be studied to know how good the system
performance of an SD model is. The future SDSLs will allow a
user to determine the performance of an SD model qualitatively
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 61
(Ghose et al 1987b, Wolstenholme 1985b) as well as quantitatively
(Bapna and Sharma'1987b). We strongly feel that availability of \
such general purpose QPM (Qualitative/Quantitative Performance
Measure) will be a major development. in the SD methodology.
4.3 Sensitivity Analysis
Availability of QPM will make it possible to provide practical
sensitivity analysis tools. Sensitivity analysis can be used to
study how the system:-pérformance of an SD model is affected by
variations in certain variables or policies in the model.
4,4 Optimization
With the availability of QPM, it will be possible to ‘provide
features to perform optimization. Three types of optimization
will be particularly feasible for SD (Bapna and Sharma 1987b).
(1) The user may specify the range of variables. Then non-linear
programming (Operations Research) can be used by the package to
find ‘a sub-optimal solution with associated values of the
variables.
(2) The user may specify various alternate policies for an SD
model. The SDSL can find QPM for each of the options and choose
the best solution.
(3) A model base may contain several models for the same system.
These models may differ by one or more modules. The SDSL can
find QPM for each model and choose the best solution.
Qualitative optimization (Ghose et al. 1987b, Wolstenholme 1985b)
is another possibility. In all optimization, some of the best
solutions can be given to the user, so the an expert user can
then make the.final choice.
4.5 Applications from Other Fields
Some new ideas and techniques in SD methodology. may come from
other fields, like Simulation, Control Systems, Optimization,
Decision Support Systems, Model-base technologies, Artificial
Intelligence, Expert Systems etc. (Ghose, Chakrabarti and Bapna
1987a, Mohapatra and Sharma 1985, Wilsson- 1972, Weiss and
Kulikowski 1984). The availability of a suitable
qualitative/quantitative performance analysis methods will help
develop applications from these fields. Such applications will
lead to development of new SD techniques and tools.
4.6 System Dynamics Modeling Kits
The future SDSLs may also provide SD modeling kits. These kits
may contain several well written programs for common use for SD
applications. Such kits may serve two purposes. First, it helps a
student or new user to quickly acquaint himself with SD modeling
62 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY, CHINA
and simulation programming. Second, it serves as a reference for
the experienced user when he needs to build SD models. A drawback
of SD modeling kits is that the users may tend to use a model or
program from the kit even without ensuring its suitability for a
specific application.
4.7 Statistical Summary Facility
Summary statistics, such as extreme values, mean, standard
deviation, min, max etc. may also be computed; these may be
provided in a easy to use fashion with a spreadsheet type of
facility (See Section 2.2).
5 IMPLEMENTATION OF FUTURE SDSLs
Programming language design and implementation methods have
evolved rapidly since the earliest languages (including DYNAMO)
appeared in the late 1950s (Aho and Uliman 1977). The
implementation of SDSLs is particularly easy because of its
inherent simplicity. Absence of control structures is the primary
cause of simplicity.
The implementations of SDSLs in future will have to consider
various factors like implementation téchniques, portability
considerations, efficiency considerations and the available
hardware systems. An attempt to provide a user-friendly
interactive environment will also be an important design
consideration.
5.1 Portability Considerations
Portable implementation means that the implementation can be
moved to different machine or computing environment with much
less effort than that required to rewrite it (Wallis 1982). The
main advantage of portability is economy ~ if a software product
is portable, it is cheaper and simpler to implement on other
computers. Portability may have the effect of making the software
better documented, better designed and more thoroughly tested
than it might otherwise have been - these aspects make it more
reliable and easily maintainable. MS-DOS ‘and UNIX operating
systems as well as the C programming language are likely ta he a
.popular choice for achieving portability.
5.2 Efficiency Considerations and Implementation Methods
There are two techniques for translation of programming languages
- compilation and interpretation. Compilation is usually used for
achieving fast program execution and interpretation is used for
providing good interactive program development and debugging
facilities. Many of the current languages (including DYNAMO) have
been designed to maximize the execution-time efficiency. However,
execution-time efficiency is only one side of the proplem. The
other side is that a program also has to be created, modified,
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 63
debugged, tested and maintained by the users The preoious user
time and effort can be saved by providing good interactive
facilities (See Section 2).
The future SDSLs will be implemented to contain both compiled and
interpreted modules to take advantages of both the approaches.
Some implementations may use microprogramming (Hayes 1978) for
achieving higher run-time efficiency.
5.3 Computer Hardware and Technology
The computer hardware has become relatively cheap and powerful.
Inexpensive personal computers may be used in various simple SD
applications. Bigger computers are of course needed for complex
applications. SDSLs will need to be implemented on various types
of computers.
Future SDSLs may also exploit the advanced technology like
parallel processing, distributed processing and computer
networking. The very nature of SD simulation methodology permits
exploitation of parallel computers and array processors. Such
hardware can result in significantly increased computing power.
5.4 Estimate of Time Needed for Implementation
A rough estimate of the time needed for experimental
implementation of each feature described in this paper is 1 to 10
man months. The effort. required may be less if many features are
to be implemented by the same team.
6 AN EXPERIMENTAL IMPLEMENTATION
We implemented a DYNAMO-like language (Bapna 1985, Bapna et.al
1987a). The implementation included usual DYNAMO-like features
like sorting of equations, pre defined functions, detecting
level-less loops and generating associated cross-reference
listing etc and some additional features like dimensional
analysis. unit conversion, local sensitivity analysis (over DT
period).
7 CONCLUSIONS
This paper argued that the advances in the fields of Simulation
and Computer Science offer opportunity for improvements in the
way SD is created, the way user interacts with the program and in
general the way SD studies are carried out. These advances were
Presented in a suitably modified way in the context of System
Dynamics. Various features discussed include human-engineered
interactive simulation environment, reliable modeling and
programming and analytical techniques and tools. Many of the
features discussed are likely to become available in next five
years. Some aspects of implementation were also discussed.
64 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
REFERENCES
Aho, A. V., and Ullman, J. D. (1977) Principles..of Compiler
Design, Addison-Wesley.
Aus, H. M., and Korn, G. A. (1971) "The Future of. On-Line
Continuous System Simulation", Proc. AFIPS/JSCC.
Balzer, R. R. et al.(1983) “Software Technology in 1990s : Using
a New Paradigm", Computer November.
Bapna, R. K. (1985) Design and Implementation of a Simulation
Language for.System. Dynamics, Master’s Thesis, Indian
Institute of Technology, Kharagpur, India.
Bapna, R. K., Ghose, S., Sharma, S. K. (1987a) "Development of
a System Dynamics Simulation Language", Proc. of the Second
National Conference on System Dynamics, India, Jan 1987.
Bapna, R. K., and Sharma, 8S. K. (1987b) “A Quantitative
Performance Measure and its Applications to System
Dynamics", Proc, of the 1987 International Conference of the
Svstem Dynamics Society, China, June 1987.
Be, K. (1982) “Human-Computer Interaction", Computer, November.
Benzon, B. (1985) “The Visual Mind and the Macintosh",. Byte,
January.
Birta, L. G. (1977) “A Parameter Optimization Module for CSSL-
Based Simulation Software", Simulation, Vol 28 No 4.
Brooks, F.. P. (1975) The Mythical Man-Month: Essays’ in Software
Engineering, Addison-Wesley.
Buxton, J. (1980) Stoneman: Reguirement for Ada Programming
Environments, US Dept of Defense.
Cellier, F. E. (1978) Combined Continuous/Discrete System
Simulation by Use of Digital Computers: Techniques and
fools, Ph.D. Dissertation, ETH, Zurich.
Coyle, R. G., and Sharp, J. A. (1976) System Dynamics - Problems.
Cases and Research, John Wiley and Sons.
Coyle, R. G. (1977) Management System Dynamics, John Wiley and
Sons.
Davis, R., and Lenat, D. B. (1982) Knowledge-Based Systems in
Artificial Intelligence, McGraw-Hill.
Forrester, J. W. (1961) Industrial Dynamics, MIT Press.
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 65
Forrester, J. W. (1969) Urban Dynamics, MIT Press.
Forrester, J. W. (1973) World Dynamics, Wright-Allen.
Forrester, N. B. (1983) "Eigenvalue Analysis of Dominant
Feedback Loops", The 1983 International System Dynamics
Gonference Plenary Papers.
Ghose, S., Chakrabarti P. P., Bapna, R..K. (1987a) “Intelligent
Decision Support Systems Using System Dynamics Models",
Proc...of the Second National Conference on System Dynamics,
India, Jan 1987.
Ghose, S. et. al (1987b) “Qualitative Optimization-An AI Approach
to Policy Design in System Dynamics", Proc. of the Second
National Conference on System Dynamics, India, Jan 1987.
Gibb, T. (1977) Software Matrices, Cambridge, Mass.
Goldberg, A and Robson, D. (1983) Smalltalk 80, The Language and
Its Implementation, Addison-Wesley.
Gordon, G. (1982) System Simulation, Prentice-Hall.
Hayes, J. P. (1978) Computer Architecture and Organization,
McGraw-Hill. :
Hooper, J. W. (1984) “Simulation Technology : Advances, Trends,
Challenges”. Simulation, January.
{DSS (1983) Prototype(2.0), IDSS_User’s Reference._Manual,
Pritsker and Associates, Inc.
Kelley, J. E. (1983) The IBM PC and 1-
Keloharju, R. (1987) Use of Optimization Techniques in SD",
Orientation Course _ ure Notes.of the Secon: nal
Conterence on System Dynamics, India, Jan 1987.
-3, Banbury Books.
Kochan, 8. G., and Wood, P. H. (1984) , Exploring. the UNIX
System, Hayden Book Company.
Lucas, J. J., and Wait, oF Vv. (1975) “DAREP - A Portable CSSL-
type Simulation Language”, Simulation, Jan 1975.
Melamed. B., and Morris, R. J.T. (1985) “Visual Simulation: The
Performance Analysis Workstation", Computer, August.
Mills, H. D. (1971). “Top_Down Programming in Large Systems", In
R. Rustin(ed.), Debugging Techniques in Large Systems,
Prentice-Hall, Englewoods Cliffs.
Mohapatra, P. K. J., and Bora, M. C. (1983) DYMOSIM User’s
Manual, Indian Institute of Technology, Kharagpur, India.
66 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
Mohapatra, P. K. J., and Sharma, S. K. (1985) “Synthetic Design
of Policy Decisions in System Dynamics Models - A Modal
Control Theoretic Approach", System Dynamics Review, Vol 1
No 1.
Newman, W. M., and Sproull, R. F. (1979) Principles of Computer
Graphics, McGraw-Hill.
Nilsson N. J. (1980) Principles. of Artificial Intelligence,
ioga.
Oren,.T. I., and Zeigler, B. P. (1979) “Concepts for Advanced
Simulation Methodologies", Simulation, Vol 32 No 3.
Pegden, C. D. (1982) Introduction to SIMAN, System Modeling
Corporation.
Plattner, B., and Nievergelt, J. (1981) “Monitoring Program
Execution : A Survey", Computer, Vol 14 No 11.
Pratt, T. W. (1984) Programming Languages: Design and
Implementation, Prentice-Hall.
Pressman, R. (1982) Software Engineering: A Practitioner's
Approach, McGraw-Hill.
Pugh, A. L. (1976) DYNAMO ITI User’s Manual, MIT Press.
Raeder, G. (1985) “A Survey of Current Graphical Programming
Techniques", Computer, August.
Randers, J., ed. (1980) Elements of System Dynamics Method, MIT
Press.
Roberts, N. et al. (1983) Introduction to Computer Simulation:
The System Dynamics Approach, Addison-Wesley.
Richardson G. P. and Pugh, A. L. (1981) Introduction to System
Dynamics with DYNAMO, MIT Press.
Rohrbaugh, J., and Andersen, D. F. (1983) “Specifying Dynamic
Objective Functions: Problenis and Possibilities”, Dynamica 9
Starr, P. J. (1981) "Identifying Critical Parameters in System
“Dynamics Models" in Ed. Paulre, E. System. Dynamics and
Analysis of Change, North Holland.
Thillainathan, V. and Price, D. H. R. (1981) "Estimating
Parameters Values in Continuous Simulation Model of an
Industry”, European Journal of Operational Research 8.
Uhran, J. ‘J., and Davisson, W. I. (1984) "The Structure of
“CNDTRAN: A System Simulation Language”, IEEE Trans on
Systems, Man and Cybernatics, Vol 14.
THE. 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 67
Uyttenhove, Hq, ad. J. (1978) omputer-Aided . Syatema
Methodologies: An. emblage of _ Methonolozieal Toole for
Systems Problem. Solving, Ph.D. Dissertation, State Univ. of
New York, USA.
Vermeulen, P. J., and De Jongh, C. L. (1977)- “Dynamics of Growth
in Limited World : A Comprehensive Sensitivity Analysis",
Automatica 13.
Wallis, P. S. L. (1982) Portable Programming, Macmillan.
Weinberg, G. (1971) The. Psychology of Computer Programming, Van
Nostrand Reinhold. Z
Weiss, S. M., and Kulikowski, C. A. (1984) A Practical Guide to
Designing Expert Systems, Chapman and Hall.
Wolstenholme, E. F. (1985a) “Algorithmic Control Modules for
System Dynamics Models", Proc...of the 1985 International,
Conference of the System Dynamics Society.
Wolstenholme, E. F. (1985b) "A Methodology for Qualitative
System Dynamics", Proc. of the 1985 International Conference
of the System Dynamics Society.
Zeigler et al. (1979) Methodology in__Systems Modeling and.
Simulation, North-Holland.
