Group Title: Department of Computer and Information Science and Engineering Technical Reports
Title: The design of a human computer interface for a multimodeling object oriented simulation environment
Full Citation
Permanent Link:
 Material Information
Title: The design of a human computer interface for a multimodeling object oriented simulation environment
Series Title: Department of Computer and Information Science and Engineering Technical Reports
Physical Description: Book
Language: English
Creator: Kim, Youngsup
Fishwick, Paul A.
Affiliation: University of Florida
University of Florida
Publisher: Department of Computer and Information Science and Engineering, University of Florida
Place of Publication: Gainesville, Fla.
Copyright Date: 1998
 Record Information
Bibliographic ID: UF00095417
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.


This item has the following downloads:

1998270 ( PDF )

Full Text

The Design of a Human Computer Interface for a
Multimodeling Object Oriented Simulation Environment

Youngsup Kim and Paul A. Fi-li,' 1:

Computer and Information Science and Engineering Department
University of Florida

MOOSE (., ll11i, I !,. Object Oriented Simulation Environment) is an application framework under the develop-
ment at U! i -if- of Florida, which is used for modeling and simulation. MOOSE is based on Object Oriented
PL- -i. 1 Modeling (OOPM), and consists of a Human Computer Interface (HCI), Translator, and Engine. A human
model author builds the model of a 11 -1. i1 -- -1. i with the help of Graphical User Interface (GUI) and represents
his/her model with a picture. The MOOSE GUI contains two I- i" of modelers: conceptual and dynamic. The
conceptual modeler supports a model author to define classes and relations among classes in a form of class hierarchy
that represents the conceptual model. The dynamic modeler assists a model author to build dynamic models for
each of the classes defined in the conceptual model. The dynamic model I I." supported are Functional Block
Model, I iiI.- I ,I.. Model, Equation Model System Dynamics Model, and Rule Based Model. We are currently
performing research to enlarge the HCI' ,1 il 'l by adopting 3D graphics to provide a more immersive and natural
environment for better interfacing geometry and dynamic models. We suggest 3D GUI with MOOSE Plug-ins and
APIs, a static modeler and 3D scenario. When a user selects a ,'1! -1. ,1 object on the 3D graphic window, they use
this "!i i I"!.-" to get the conceptual model that relates to that object.
Keywords: Simulation, HCI, GUI, Multimodeling, MOOSE

HCI involves designing computer -1. in- that support people so that they can carry out their activities productively
and safely, and its goal is to produce an usable and safe -l. i, as well as a functional -l. i, Visibility and
Functionality are in i. r roles of HCI.1
MOOSE is an ,, I !i ii for Multimodeling Object Oriented Simulation Environment, a modeling and simulation
enabling tool under the development at Uii- i -if- of Florida. The MOOSE is an implementation of OOPM for
multimodeling 1i1- -'., 1 -- -1. i,- into digital models. The MOOSE is composed of HCI, Translator and Engine.
The purpose of the MOOSE HCI is to give model authors a l. ,1 ll- of interacting with the MOOSE -l. i to
build and execute models with the help of a 2D GUI. Model authors can

think clearly about, to better understand, or to elucidate a model ,

participate in a collaborative modeling effort,

repeatedly and painlessly refine a model as required, in order to achieve adequate 6il. lif at minimal develop-
ment cost,

painlessly build large models out of existing working smaller models,

start from a conceptual model which is intuitively to domain experts, and to unambiguously and automatically
convert this to a simulation program,

create or change a simulation program without being a programmer, and
Other author information: (Send correspondence to Y. Kim)
Y. K.: Email:; !. 1 ...... 352-392-1435; Fax: 352-392-1414.
P.A.F: Email:; 1. i....... 352-392-1414; Fax: 352-392-1414


Modeler System Fie> Translator

UsC++ Code

I Scenario
Scenario Executable Engine

Figure 1. The MOOSE components.

perform a simulation model execution and to present simulation results in a meaningful way so as to facilitate
the order objectives above.2

The MOOSE HCI consists of Scenario and two I- i" of Modelers: Conceptual and D,o ...... A model author
can define a model on the modeler, and a model information from the modeler is transferred to the translator.
The translator internally converts the information into l. ....,,I..-.' Target Language (TTL) which was written C++.
Errors made during a conceptual modeling phase on the Modeler can be detected by the translator. Engine source
codes emitted from the translator are compiled and linked directly to create an engine, an executable, by a compiler
such as g++ or Visual C++ depending on a platform environment. The scenario 1i1 -i.1 the results of the model in
graphical forms. See Refs. 2,3 for details (Fig. 1). Since a 2D GUI has restrictions for representing and visualizing
I'11 -.i ,1 -1 -1. in- we are evolving the MOOSE HCI .' 1, 'l.l1i- to 3D graphics to provide a more immersive and
natural environment. We will suggest some design issues for integrating 3D geometry and 2D dynamics as a first
step toward a 3D MOOSE HCI.
Section 2 describes what kinds of features we are using to support OOPM on MOOSi. Section 3 gives brief
introductions about each MOOSE GUI component from the Conceptual modeler to the Scenario, and Section 4
suggests how MOOSE can use 3D graphics for the future.

2.1. Object Oriented Physical Modeling (OOPM)
The OOPM is a design approach for structuring multimodels in an object-oriented framework for 11!. -i1 ,1 modeling.
To build an object oriented digital model for a ,1!i -i. ,1 -l. i, we must create a class graph which consists of classes
and relations among the classes. Furthermore, each class has attributes, methods, and objects.4 ('I .-- and objects
in a digital model being built correspond to those in the i,1! -i, ,1 model being modeled. Objects are instances of a
class so that all the objects of one class must keep the same attribute and method structure of the class but can have
different instance values according to a purpose of each object.
The relations among the classes are generalization, I',,!., ..;!''.,'. and dual.4 In the generalization relation, a child
class is a !i,, of a parent class, and has Inheritance property which allows data passing from the parent to the child.
In the :; i ri ,. i relation, a child class is a part of a parent class and has Composition property which makes data
passing from the child to the parent. C..I' I ,l '.,, of a child class in the :;, --i il,! should be specified. In general,
without such a specification, it is assumed that a class can be composed of ii,, number of objects of the child class.
The dual relation is a combination of the generalization and the :;,. i,., in some specific case. All relations are
repeatedly defined in a class hierarchy.

Once a class hierarchy is constructed, we identify attributes and methods for each class. An attribute is either
a variable or a static model. The variable is one of native data I- i such as integer, real or string, and the static
model is a group of objects to represent a geometry of the model. A method can be a code method or a dynamic
model. The code method is written in a specific object oriented programming language such as C++ or Java. The
dynamic method reflects behaviors of the ,'1! -1. 1 model, operates on the static models and variable attributes to
effect changes, and tiil.. the multimodeling 1, l,11li- to capture various levels of model abstraction.

2.2. Multimodeling
Models that are composed of other models, in a network or a graph, are called a multimodel, and a modeling process
in which we model a -1, i at multiple levels of abstraction is multimodeling.5 It provides a way of structuring
' til, i i I model I- i together under one framework so that each I- I" performs its part, and the behavior is preserved
as levels are mapped.6-9 \\ ii! two .,li , i levels in a multimodel hierarchy, components of the higher level can be
,, 1..1. / by several components of the lower level, and the lower one can be abstracted into the higher level. So the
higher level takes a role of a black box in which the lower level represents a functional information inside. Of course,
the inputs and outputs for both levels should be identical.

3.1. Modeler
The main purpose of the modeler is for a user to create a digital model of a i'1! -i. ,1 -- in by defining all the
features described in OOPM such as classes and objects. The modeler not only provides users with a set of windows
to create and modify models but also produces various I-l.I..1._- Id. for transferring a model information to the
translator. The set of windows are divided into two groups: Modeler window and Dialogue window. A user can
draw his/her models on proper modeler windows, and set specific model information on the dialogue windows. The
MOOSE windows are written in Tcl/Tk. Once the model construction is completed, the model information is saved
into either a class hierarchy definition ti .hdf, or a dynamic model -l.I,,1._- I !. .,1/f depending on the modeler
I- ." The .hdf ti-. keeps information from the conceptual modeler and the .i,.f til! keeps those from the dynamic

3.1.1. Conceptual Modeler
The in iir roles of the Conceptual Modeler are to create and manage a conceptual model of a i1!i -i. ,1 -, Ii by
defining classes, relations among classes, objects, attributes and methods, and to support a control of the whole
modeling process from constructing of a model to executing the model.

1. Windows
a user can draw a class hierarchy on a conceptual modeler window. Symbols defined in the conceptual modeler
are classes, generalizations, :i, -J.- i .- and the connection links. In Fig. 2, a rectangle represents the class,
a small circle represents the generalization, a small square represents the :i -1I11 a white small rectangle
represents the 1 ,1li,1 ili- and an orthogonal line represents connection links among classes in an :;__-i ,lr, ,
or a generalization relationship. The dialogue windows are Class window, Object window and Method window.
A user can define attributes, methods and objects of a class on the class window of the class. A- _i. 1, ,1
classes are automatically set as attributes of the Abstract Data Type on the ; 1 i 1, class. The object
window consists of an attribute section, a method section, an ;, _ .I i. .11 section and a generalization section.
Except for the method section, user can specify specific values on each section. The method window shows
input and output parameters of the method and a t!.- name where the method information is located. All the
symbols in the modeler window are movable to form a well structured class hierarchy.

2. Hierarchy definition File (.hdf)
All the OOPM information and a model execution information should be saved into a fixed .!,/1/ t!. I'* ? If
Whenever symbols are created, idnumbers for each ii !11 i I" are increased and associated with correspond-
ing -- !iiil.1- and these idnumbers are used in the t'....'; !1,f as an identifier. The t'.'.. ." ,f includes (1)
A model execution information, (2) a class information including attributes (1- i- and name), methods (I I,"
name and .!t.f !I.- name) and object list, (3) an :,_ _. ir., /generalization relation information, (4) a toolkit
library information (described in the 1 i;. I section) and (5) initializations of each object.


"t. VIAn U 0

(a) Conceptual modeler

HI % I MsM Va

IIE1 1M I- t

Hydrology [hydrology
int |month
real Itemperature

SCODE IIChangeMonth j
U CODE IChangeTemp
SFBM II xeculass

marsh __ 1i

(b) Class window

Figure 2. The Conceptual Modeler in the Snail model.

3. Control
The directory MOOSE includes sub directories: modeler, trans, engine and prj.

Because putting all the

models built in MOOSE under the prj directory gives more convenient to users as well as MOOSE developers,
when the conceptual modeler is opened, a current directory will be automatically changed to a directory of
0OOSE/prj". Of course, user can select ,I!i directories on a directory window during model loading and
In the menu of project/option, a user can specify a project name, a I!'' 1i directory, a start object name and
a start method name, a simulation time interval and a maximum time limit to control a model execution.
Execution control window can be popped up by clicking an image button of a negative film. If a user clicks
delete all button, the conceptual modeler window is cleared, and an internal data structure is removed so that
the conceptual modeler becomes iild

3.1.2. Dynamic Modeler
Methods are divided into dynamic methods and the code methods. The code method is a C++ code segment for
the method which is not categorized as ,.1- 1 iiii. model. It can be written at the conceptual Modeling phase, and
saved as a .cod ti!.-
The dynamic modelers have been developed for each dynamic model I I" Functional Block Model (FBM), I i i i.
i ,i.- Model (1 S. I, ), E(ul i. Model (i.'M), System Dynamics Model (SDM) and Rule Based Model (l1l;.i By
the definition of the multimodel, components of a dynamic model can be refined into other dynamic models or code
methods. Therefore, all the methods must have multiple input and output parameters so that each outputs of one
method are connected with inputs of other methods in the dynamic modeler. Also, one of outputs can be specified as
a return value of a dynamic method which cannot have a return statement on it while the code method can explicitly
have a return statement. Dynamic modeler windows that are similar to the corresponding windows of the Conceptual
Modeler, and produces ti,-. keeping rI .1- ._- -information of the dynamic method. A name of dynamic method is
specified as Classname::methodname (Example: car::Drive means Dynamic method "Di of a class ", ').

1. FBM Modeler


int month
e temperature F

] CODE ChangeMonth
Q CODE ChangeTemp I
S FBM Execute

"Oc I


(c) Object window


Figure 3. The I I;. l Modeler in the Monorail model.

(a) Windows
Symbols of the 1 1;., I Modeler are Function Block and Data Flow Link among function blocks. In the 1;., I
Modeler window in Fig. 3, a square represents a function block, and a black line segment between two
function blocks denotes the data flow link. When a function block is created, a block number is increased
by one and associated with the function block, and a user can specify a method name for the function
block. The user can explicitly specify the direction of an input and an output between the blocks, and
images can be used for representing complex behaviors of some blocks. The 1 I;., I Modeler also supports
Toolkit library window to select functions in the toolkit library. The toolkit is categorized into groups
depending on their properties. Control toolkit supports Add, Subtract, Multiply, Divide and Integration,
and .)' ., '.', toolkit helps users build queuing models.
(b) Event scheduling
Since an FBM should be executed in a sequential ascending block number order, all the block numbers
should be sequential order of execution. Whenever the FBM is saved, an Automatic renumbering is
performed to renumber block numbers if those are not properly ordered, due to deletions of ;oi- blocks
or creations of blocks in complicated order. In order to draw functional block models, the 1 I;. I Modeler
window supports Grid facility to place blocks at proper locations and automatic line adjustment to redraw
all the line segments for a link to be orthogonal line segments by ;~lil-l[ a start and an end positions.
(c) Topology File
The ,,-I !i' of the 1 I;. I includes function block information, link connections of outputs and inputs among
the blocks and external inputs and outputs of the a functional block model. If the toolkit library was
used, the toolkit names are detected and saved in a ; .; ../ i..!l

2. FSM Modeler

(a) Windows
The main -- i,1 .!- of the 1 S. I are S I .- and Transition indicating a movement from one state to another
states according to a transition condition. In Fig. 4, a state is represented by a circle, the transition
is an arrowed arc, and the condition on the transition arc is a text. When a state is created, a state
number is increased by one and associated with the state, and a method name is specified by a user. All
the attributes and inputs of the 1 S. I can be used in the condition statement. When saving the 1 I a
dialogue window asks for the start state and 1 S. I outputs.
(b) Event scheduling
The 1 S. I execution begins at the start state, proceeding to other I S. I states according to the result of


li-k aLad

I/"" / "' 0,

Figure 4. The 1 S. I Modeler in the Boiling water model.

transition conditions, and ends when it arrives one of the final states or when the simulation time limit is
(c) Topology File
The .:j ti!.- of I S. I includes a state information, transitions and associated conditions among states, out-
puts of I S. 1, and the start state of the model. Whenever saving the 1 S. I state numbers are renumbered
to make all the state numbers be in a consecutive order.

3. EQM Modeler

(a) Windows
The ICQM enables the user to create a constraint model using differential and algebraic equations. Any
number of nth order differential equations 11i be entered by using an intuitive syntax. Differential
equations are represented using symbols such as x, x' as the first derivative of x, x" as the second derivative
of x, and x followed by n single quotes as the nth derivative of x in general. Several variables can be used,
and each variable 11i be one of attributes in a class to which the equation model belongs and input
parameters of the QIM. To give a multimodeling .' 1 -1if.ll- to the MOOSk., code methods and dynamic
methods are used as a variable. The code methods should explicitly include a "i' I [I 1 -I I. i1. iii and
dynamic methods must have an output parameter specified as a return value. Then those return values
will substitute the method variable in the Q IM. If a method has output parameters, equations has to use
attributes as actual output parameters of the method. When saving QIM, a dialogue window asks initial
values of variables used in the differential equations, and the outputs of FQM. The outputs of the model
i i be attributes or iii order derivative of ;,ii variables (Fig. 5).
(b) Event scheduling
The IiIQM executes equations in a top down order. Because every new equation is put at the last line
on the QIM window, an order of creating equations is very important. Algebraic equations are executed
earlier than differential equations by an action of the translator.


iL '=L11 (aux1)- k2(L1 +k3)
IL2'=K2"(L1 +k3)-k3"(L2+k3)
lauxl =Rtn(outl ,out2)

L1(0) 3.7
L2(0) 1.3
11 IL
12 L2

Figure 5. The EIQM Modeler in the Applesnail model. The dialogue window inside is the initialization window.

(c) Topology File
The .i/f ti!-. of the ijQM consists of initial values of derivations of -til, i, ii- i! equations, Programmable
Form of equations converted from the equations on the EQIM window by Automatic Equation Conversion
program, and outputs of the EIQM.

4. SDM Modeler

(a) Windows

The SDM is similar to the 1 I;., I in the view of a modeler window and an execution behavior. The -- 1 .1-
are Source, Rate, Level, Auxiliary, Constant and Sink. Control flow and Data flow ((' i-, .1-1.11. i ) are
links connecting symbols. In Fig. 6, the control flow is a black arc heading to a center of -- ,1,,.1- and
the data flow is a blue arc going to a bottom of symbols. The SDM hass COMA ( ('I --Object-Methods-
Attribute). It displays a list of possible classes from the current class to which the SDM belongs, and
shows the corresponding objects, methods and attributes when a user selects one of the available classes.
Symbols represent attributes as well as methods. All the inputs are gathered into the source, and the
sinks announce the output !1,, .1 of the SDM. The SDM supports a dialogue window for -i f- ;i-i a
connection of inputs of SDM.
(b) Topology File
Our premise is that the semantics of an SDM is equivalent to a set of differential/algebraic equations.
I i-l ,, I: defines the rules that translate an SDM with a set of equations.5'10 We convert an SDM into
an equivalent E ('. I' by following the rules.

5. RBM Modeler

(a) Windows

The RBM windows are composed of three sub windows. A main window on top shows "IF -I I~. I I-
created and activates other two sub windows to edit the if statements. On the predicate window, a user
specifies the predicates by selecting variables listed in the Accessible element list. A consequence window

JE ini i [Ii i M iEEm ii,

b...BI .. B. .......

Figure 6. The SDM Modeler in the Applsnail model.

I RIrMll MImIlilElR 0

If month >. 3 ANW

IF mK.lth >= 10

IFP r'teies1

0,,it it iout ,ve

month < 10O

SnilP'op Nonbreeding-



Save Quit

Accessible eleririts I.ptrailor

.TnrU _l _year .:

RuIle ErIllor


OK ICancel


Figure 7. The RBM Modeler in the Snail model.


00 60 0
.-31 i

normal abnormal stimulafng
.:, ) s.:

(a) xy plot (b) FSM dynamic window

MOIII ..II i.. ..

(c) Animation (d) Histogram in a multiple cells

Figure 8. The Scenario visualization windows.

lists available methods from the class to which the RBM belongs. The else part of the IF statement
should be converted into a next IF statement. The consequence must be a method and it will support the
multimodeling (Fig. 7). RBM evaluates the predicates in a top-down order, and executes a consequence
method when a predicate becomes true.
(b) Topology File
The .i/ -f I !.. of RBM consists of each predicate and consequence in order of creation.

3.2. Scenario
While the Modeler is an interface from the users to the MOOSE -.i the Scenario is the one from the MOOSE
-I. i!, to the users. It consists of an Execution control window and Output visualization window. By the Execution
control window, a user can translate a model to engine source !!!. -, build an engine which is executable code of the
model, modify simulation parameters, change a rate of simulation time, and run the engine.
When running the engine, the Scenario establishes a bidirectional pipe connecting it to the engine, and i! 1!, -
nizes an execution of the engine with the scenario. The scenario writes to the pipe as a standard input of the engine,
and reads a standard output of the engine from the pipe.2
The Output visualization window dynamically shows outputs of the engine in visualization forms: a xy_plot, a
Histogram of multiple cells, a 2D Animation on which objects can move around, a I S I dynamics window showing a
current states of each I S. Method (Fig. 8). Some simulation outputs are not necessarily amenable to the realtime
graphical treatment, and there is a necessary role for traditional methods of .1 -; 2,11 13


We are enhancing the MOOSE HCI to a more immersive and natural environment by i I!!1 i~-J a 3D visualization
technique. Blending the 3D Geometry model representation with the 2D dynamic model representation requires (1)
MOOSE Plug-ins and APIs, (2) 1 tIi, modeler and (3) 3D scenario.

Figure 9. The MOOSE environment.

4.1. MOOSE Environment
We chose Vi :.11, (Virtual ,,1 lI Modeling Language, presently Vi:. ll.. )14 as our 3D graphic programming
language. VI:., 11 allows to create 3D objects in a 3D space and to add behaviors to the objects. For a communication
between the VI:. 11 world t!i.- (.wrl) and its external environment, an interface between the two is needed. The
External Authoring Interface (EAI) is the interface defining a set of functions on the VI:., 11, brower that the external
environment can perform to affect the VI:., 11, world. Currently, a Java applet can dynamically communicate with a
VI:., 11, world embedded in the same HTML page through the EAI. We put the MOOSE on the WEB in a form of a
Plug-in with proper APIs to interact with a Java Applets to communicate with VI:. 11i world (Fig. 9). In the Web
brower, a user can run the MOOSE Plug-in by selecting a MOOSE model embedded in a HTML page, and experience
Vi:., 11, worlds which are dynamically changing according to the results of a model on the MOOSE Plug-in.

4.2. 3D Scenario Window
A communication scheme of the 3D scenario and the MOOSE engine is very similar to that of the 2D scenario. As
the 2D scenario writes to and reads from the engine through "pip."' built when the 2D scenario activates the engine,
the 3D scenario can communicate with the engine via a 3D interface between the two. As discussed in the previous
section, Vi:. 11i worlds can represent 3D scenarios the EAI serves as the 3D interface between the Vi:. 11i world
and Java, and the Java can communicate with MOOSE by using networking protocols such as TCP/IP. Therefore,
the Java is a bridge connecting the 3D scenario and MOOSi.
When a user clicks an object on the 3D scenario window, the EAI passes the object selection information to a
Java code, and the Java code sends it to the MOOSE conceptual modeler. A corresponding 2D conceptual model
including a class to which the 3D object belongs will be opened so that the user can see a geometry of the object or
modify ,ii, attributes or methods of the object.
In the other direction, MOOSE can affect the 3D scenario window to reflect outputs of MOOSE execution over
time. During a model execution on MOOSi. outputs can be sent to the Java code. The Java code can command
the 3D scenario to change its status according to the outputs via the EAI.

4.3. 2D Static modeler
The word "-I i in the static model refers to the ii ,1 ;l; of the model to cause changes of values of attributes; it
does not mean that the model doesn't change. Our primary I- I- of static model is one that specifies the l 1I,. .1 .-_*
or geometry of a 11!- -i A1 object, and the dynamic methods can change the structure of static models over time.15'16
There are a number of representational techniques for modeling geometry and space, ii >ii of which are discussed

Web Browser


Plugins) Audio

( Plugins)

by Samet.17"'8'15'16 Our goal is implementing the 2D static modeler to give a 3D static model 'I' Il.ili- to the
MOOSE instead of extending .oI of static modeling method.
VI:. 11, 3D graphics are the results of executions of .wrl tl. Therefore, the .wrl I. of the 3D static model can
be edited and kept in one of three 2D static modeler l- I" VRML code corresponding to the C++ code method,
,,i: modeler corresponding to the dynamic modeler, and URL of a .wrl ti!. In :,, cases, .:,~/ I! of static models
must keep real VI:;. 11 codes so that they can be converted into a .wrl ti!. and .1l -1 ,- 1 as 3D static models.
For the VI:. 11 code, a user can write VI:. 11. programs as the user can write C++ code method in the concep-
tual modeling. The written VI:. II codes can be copied directly into .!/'f t! On the static modeler, to support
multimodeling ', 1 ,1 i. lI of the MOOSi. static models should be .,_i _i. Il 1 and generalized in a similar way to
dynamic models. Therefore, a static model can be refined into another static models. As we map a method name
into a function block in a FBM Modeler, each square on a static modeler represents a corresponding static model
which can be composed of another static models. The .:./ tI!b.- from the static modeler is a set of VI;:. 11 codes of all
the static models defined in the static modeler. When the static modeler I"- is an URL, the .!i/f t!l. can just copy
a body of a .wrl tib- pointed to by the URL.
To l i-1 .1 3D static models by using VI:. 11. codes, we need enhance the current translator. The translator creates
a VI:. 11, source ti! .wrl, by combining each ../-f !t!.- of static models. Since 3D static models have to communicate
with Java code, each 2D static model should be associated with corresponding Java codes including EAI codes. The
Java code can be managed in the same ways of the static model: Java code, Java modeler and Java URL. After
translating it, an engine builder must compile and link it to generate Java classes.

The MOOSE HCI supports human model authors in easily building Object Oriented Multimodels of I'1 -i; 1i -', i-
without being a simulation specialist. Human model authors creates conceptual models and dynamic behaviors of
'!,! -1. ,1 -, ,in,- based on OOPM concepts, create and modify multimodels with '!tilI i. I model I" execute
them, and finally show the results of model executions. All these activities are assisted by the MOOSE HCI in
order for human model authors to represent their thoughts in a form that matched their mental models of their
11! -;i 1 domains. When we add 3D graphical' '>' i.il1l- to the current 2D MOOSE by building the static modeler,
3D Scenario windows and the MOOSE Plug-ins and appropriate APIs on the \\ 1.1; the MOOSE will provide more
useful functionalities to build a realistic model of complex -1. 11, -

We would like to thank the following funding sources that have contributed towards our study of modeling and
implementation of the MOOSE multimodeling simulation environment: GRCI Incorporated (Gregg Liming) and
Rome Laboratory (Si. - Farr) for web-based simulation and modeling, as well as Rome Laboratory (Al Sisti) for
multimodeling and model abstraction. We also thank the Department of the Interior under a contract under the
ATLSS Project (Don DeAngelis, University of Miami). \\ i 1. LI their help and encouragement, our research would
not be possible.

1. J. Preece, Y. Rogers, H. !I, 1p, D. B. ii .,i S. Holland and T. Carey, HUMAN-COII'I 1 II./ \ 1/I I1:. i.TION,
Addison-WA -1 1994.
2. R. M. Cubert and P. A. I i- 1I !: i I..... An object-oriented multimodeling and simulation application frame-
work," Submitted to Simulation 1997.
3. R. M. Cubert, T. Goktekin and P. A. I i-1!- : "...... Architecture of an object-oriented multimodeling
simulation application framework Submitted to Simulation 1997.
4. P. A. 1 i-.!- i!: "hi! Ii ii!i, continuous and discrete models with object oriented 1,1- -;1. modeling," 1997
Western Simulation Multi ....-f,.... January 1997.
5. P. A. I i-! I: Simulation Model Design and Execution : Building Digital Worlds, prentice Hall, 1997.
6. K. Lee and P. A. i i-l I:, "A semi-automated method for dynamic model abstraction," in Processing of
Enabling T, ..... I,,, for Simulation Science, Pare of SPIE AeroSpace '97 C,-f' . ... 1997.

7. B. P. Zeigler, "Toward a formal theory of modeling and simulation: Si ,.I IiL. preserving morphism," Journal
of the Association for Computing Machinery 19(4), pp. 742-764, 1972.
8. P. A. I i-l!- i !:. "The role of process abstraction in simulation," IEEE I, ...-.; .... on t.(. ..... man and Cy-
bernetics 18, pp. 18-39, January/F. i.,L. i- 'I,
9. B. P. Zeigler, Object Oriented Simulation with Hierarchical, Modular Models: Intelligent Agents and Endomor-
phic .(!. !... Academic Press, 1990.
10. N ,!! Roberts, Davis Andersen, Ralph Deal, Michael Garet, and \\ ili i !!! I! II Introduction to Computer
Simulation: A ', .I... D '....... Approach, Addison-AW. -1. Reading, 1983.
11. J. A. P, !. Introduction to Simulation: programming techniques and methods of .....,/... McGraw-Hill, 1982.
12. A. M. Law and W. D. Kelton, Simulation Modeling and .1...I..'. McGraw-Hill, 1991.
13. G. S. I i-I!!! ii, Concepts and Methods in Discrete Event Digital Simulation, John \\ & Sons, 1973.
14. A. L. Ames, D. R. Nadeau and J. L. Moreland, VRML 2.0 source book, John \\ I!. & Sons, New York, N.Y.,
15. P. A. 1 i-l!;- !I: "A visual object-oriented multimodeling design approach for 1i1i -1i ,1 modeling," Submitted to
I; .1i I,...-.... ... on Modeling and Computer Simulation 1997.
16. P. A. 1 i-lI !: 1 "-1..1 !,I!io object oriented design for ,1. -i. i1 modeling," AC.( .I .... ; ... on Modeling and
Computer Simulation 1996.
17. H. Samet, Applications of Spatial Data ,i ., *., *Computer Graphic, Image I."p ... "..'. and GIS, Addison-
W -1. 1990.
18. H. Samet, Ib!, Design and .1.o... ,' of Spatial Data S' .,. ..:,.., Addison-AW, -1. 1990.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs