A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS
IN VIRTUAL 3D ENVIRONMENTS
D. Biella and W. Luther
Institute of Informatics and Cognitive Sciences,University of Duisburg-Essen, Lotharstr. 65, Duisburg, Germany
Keywords: Modelling of virtual 3D environments, reuse of parameterizable framework, automated code generation.
Abstract: This paper reports on a parameterizable 3D framework that provides 3D content developers with an initial
spatial starting configuration, metaphorical connectors for accessing exhibits or interactive 3D learning ob-
jects or experiments, and other optional 3D extensions, such as a multimedia room, a gallery, username
identification tools and an avatar selection room. The framework is implemented in X3D and uses a Web-
based content management system. It has been successfully used for an interactive virtual museum for key
historical experiments and in two additional interactive e-learning implementations: an African arts museum
and a virtual science centre. It can be shown that, by reusing the framework, the production costs for the lat-
ter two implementations can be significantly reduced and content designers can focus on developing educa-
tional content instead of producing cost-intensive out-of-focus 3D objects.
1 INTRODUCTION
Web-based 3D applications are very popular in the
learning, commercial and entertainment sector. De-
spite the success of these products, the cost of de-
signing and producing content-rich 3D learning en-
vironments is still a major issue, especially if realis-
tic models, non-deterministic simulations and spe-
cial user interaction are required.
Our work focuses on the production of learning
content with a high degree of interactivity, photo-
realism, reversibility and non-deterministic simula-
tion models, which allows learners to understand the
design and methodology of an experiment through
an individualized, interactive, Web-based applica-
tion (Biella and Luther, 2007).
In a recent interdisciplinary project concerned
with the historical replication of interactive key
experiments in psychology and education, 2D and
3D implementations of B.F. Skinner’s historical
experiment on operant conditioning were imple-
mented, tested and evaluated by students (Biella,
2006). Results showed that
o the 3D version effected a higher degree of
spatial and cognitive immersion,
o the 3D visualization of the experimental
setup was more realistic in terms of graphi-
cal quality,
o users tend to accept mixed (2D and 3D) con-
tent despite media discontinuities in favor of
knowledge creation,
o the experiment was easier to comprehend in
3D.
The question of whether to use 3D or 2D depends on
the design requirements, the target user group and
the model data to be presented. Design considera-
tions are listed in the section “Modelling pipeline”
and discussed in the section “Discussion”.
This paper reports on a parameterizable 3D
framework for Web-based experiments and identi-
fies reusability options.
It highlights three case studies that prove the fea-
sibility of the concept and its cost efficiency, and
gives an overview of recent work to enhance the
framework.
2 MODELLING PIPELINE
The modelling pipeline for replicated experiments in
virtual 3D environments presented by Biella (Biella,
2006) includes a general framework and several
workflow components that are suited for reusability
(Figure 1).
316
Biella D. and Luther W. (2008).
A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS IN VIRTUAL 3D ENVIRONMENTS.
In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 316-323
DOI: 10.5220/0001519503160323
Copyright
c
SciTePress
Figure 1: Reusability options of the framework within the
modelling pipeline.
Initially, a real-world experiment is analyzed from
an authentic setting or from primary and/or secon-
dary sources with a focus on the 3D geometrical
model of both the experimental assets and the his-
torical surrounding, an animation model, an interac-
tion model, and an simulation model.
The input/output hardware interfaces and driv-
ers, the rendering software, the modelling languages
and tools, and, optionally, a framework development
for the integration of multiple experiments are de-
termined by specifying conceptional design re-
quirements that depend on the target user group and
the complexity of the geometrical, interaction and
simulation models.
The formal model description describes the
process of defining the animation, interaction and
simulation models in abstract notions, such as
mathematical functions, statistical models or state
machines. These models are digitized by using suit-
able model description languages, such as unified
modeling language (UML). The implementation of
these models (Interaction/Simulation logic) depends
on the programming and scripting languages deter-
mined as a result of the conceptional design re-
quirements.
The following design considerations have been for-
mulated in (Biella, 2006) with regard to Web-based
museums for replicated experiments.
Historical context representation: Although
2D visualizations may suffice for the pres-
entation of an abstract theoretical model, the
visualization of a historical laboratory envi-
ronment is challenging. In a 2D visualiza-
tion, such information has to be provided
separately through additional sources, such
as a sequence of photographs or text de-
scriptions. A 3D model allows the designer
to integrate contextual information in the
space surrounding the experimental setup.
Impact of occluded surfaces: Surface occlu-
sion of objects with crucial functionalities
within an experimental setup can disturb the
knowledge transfer. In this case, 3D imple-
mentation must offer sufficient viewpoints
or other techniques (for example, surface
transparency).
Implicit 3D experiments: Experiments that
involve implicit 3D setups are best suited for
implementation based on geometrical mod-
elling and free real-time user navigation.
Model complexity: Animation, interaction
and simulation models vary in their degree
of complexity and may require high-level
programming languages instead of common
Web-based scripting languages.
I/O Interfaces: Input/output interface re-
quirements must be defined according to
user’s perception channels.
Data format consistency: A set of multiple
3D worlds should be implemented in a con-
sistent data format.
3D model asset acquisition describes the process of
gathering 3D raw data usually by 3D scanning or 3D
reconstruction. The post-production of 3D data im-
plies storage of high-definition 3D data in a flexible
file format (e.g.: X3D, XML), polygon and texture
size reduction and export in a Web-based data for-
mat, as defined in the conceptional design require-
ments. Both Web-compatible logic and the 3D
model represent a single replicated experiment (3D
content).
The presentation of multiple experiments re-
quires the development of a metaphorical model,
which defines a navigation scheme, spatial meta-
phors, and an exhibition layout. With regard to a
(real world) building metaphor, experiments are
connected through vertical and horizontal metaphors
that represent thematic and temporal selections of
experiments and allow the user to navigate between
them.
Cost-efficiency aspects with regard to the creation of
particular 3D exhibits suited for learning have al-
ready been presented (Biella and Luther, 2007).
A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS IN VIRTUAL 3D ENVIRONMENTS
317
The virtual laboratory realization presented here
underlines the feasibility of the theoretical approach
and stipulates that the production cost of 3D content
could essentally be reduced by code reusability.
In this paper, we want to focus on reusability as-
pects of the surrounding museum framework, which
is intended to allow 3D content designers to effi-
ciently create several standard 3D museum assets, so
they can focus on actual content development, espe-
cially for Web-based learning museums and virtual
science centres. The modelling pipeline, the frame-
work and reusability options are depicted in Fig-
ure 1. They concern the spatial design, including
lighting and texturing, interaction logic, navigational
schemes and spatial or temporal metaphors.
3 THE FRAMEWORK
The framework for the replication of experiments in
virtual environments (referred to below as the Repli-
cave framework) was first developed for an interac-
tive virtual museum for key historical experiments
and is implemented in X3D and PHP (Biella, 2006).
It features both pre-defined and partly parameteriz-
able routines for the automated generation of an en-
trance hall, a gallery, a multimedia room, graphical
user interface (GUI) components and metaphor-
based connectors leading toward the laboratory
rooms in which the interactive experiments are lo-
cated.
Figure 2: Conceptional metaphor with maximal number of
laboratories (wireframe view including position descrip-
tors).
The framework references a real-world building
metaphor. Basic framework assets are visualized as
building parts at the ground-floor level. Content-
related locations can be accessed through meta-
phorical connectors (Figure 2). A connector’s X3D
code consists of static and dynamic 3D objects that
are combined into a single X3D scene graph. For the
dynamic placement of thresholds, seven positions
(N1-N3, E1, E2, W1, W2) are reserved to which
laboratory doors, a small media room door or walls
are allocated according to a certain layout matrix.
The dynamic generation of the wings is achieved by
including static X3D code fragments that define
either a wall, which renders the wing’s entrance
closed, or the full wing layout.
In analogy with the building metaphor, the tem-
poral metaphor is the vertical connector. On each
level, the doors of this allocation place form a meta-
threshold and give access to meta-paths. The desired
path and destination threshold is defined by the
user’s selection of a different content or experiment
on an appropriate navigation panel. The panel is
rendered dynamically at runtime and lists all avail-
able presents.
3.1 Entrance Hall
As the user’s initial starting point in the 3D world,
the entrance hall is an obligatory asset. By default,
the main hall is cylindrical with a transparent domed
ceiling. The design is intended to reproduce the at-
mosphere of a modern museum building located in
an urban environment, which is visualized by a blue
sky texture and externally placed building fragments
that suggest the urban context (Figure 3).
Figure 3: Entrance hall (with 3D data progress indicator).
The exit door and the thresholds leading toward the
main media room, the gallery, and the vertical con-
nector are exclusively and equidistantly connected to
the entrance hall, virtually dividing it into four quar-
ter pie sections. The reception desk, with a recep-
tionist’s avatar and a login terminal, is located in the
section between the gallery and the metaphorical
connector. Dynamically generated information pan-
els listing the available experiments are located at
the reception desk and in the elevator (Section 2.5).
WEBIST 2008 - International Conference on Web Information Systems and Technologies
318
The three other sections are decorated with items
that underline the context (research-related images)
and 3D furniture objects.
The exit door is an interactive plane that allows
the user to leave the system and that can be config-
ured to initialize post-visit routines, such as opening
a new browser window for an evaluation.
3.2 Gallery
The mainstream of psychology is presented in the
gallery, which is octagonal in shape. This design
allows the installation of seven information-related
walls and an eighth wall that provides space for a
connector to the main hall. The gallery’s static 3D
model defines outer walls and a transparent roof in
accordance with the design of the main hall.
Large coloured information boards display the
various scientific research fields following the con-
tent and colour design of an optional 2D-based time-
line. 3D information plates are created by dynamic
scripts and display short informational texts (for
example, a curriculum vitae) and a photograph. Each
plate has an interactive event sensor and an individ-
ual viewpoint configuration that allow the user to
navigate there by simply clicking on them (jump
navigation). The text-based content and photographs
displayed in the gallery are stored separately from
the multimedia content of the laboratories and the
media rooms.
3.3 Multimedia Rooms
Replicave uses a two-tier approach: General multi-
media content about a given topic or person (image,
video, text) can be managed separately from the
multimedia data that describes the virtual experi-
ments. Hence, the framework contains pre-defined
designs for two multimedia room layouts: a large
central multimedia room located on the ground floor
and a small layout for topic-related content.
As a result of this approach, learners can access
all available 2D multimedia documents in the central
multimedia room, which can serve as both a library
and an appetizer for a potential 3D implementation.
Furthermore, it allows content creators to integrate
existing 2D content without the requirement of pre-
senting a 3D experiment. If at least one interactive
experiment is present, a small version of the multi-
media room is automatically generated in the 3D
content section that contains only topic-related
documents.
Both room layouts share the same functional
elements and visualization metaphors: For each
topic or person, there is a book shelf that contains
the corresponding multimedia documents repre-
sented by interactive 3D icons distinguished by their
multimedia document types—book (text based
document), film roll (video document) or picture
frame with preview (image file).
Figure 4: Central media room with dynamic 3D architec-
tural design for multimedia content visualization.
All multimedia assets can be directly administered
via a file-based document management system
(DMS). The 3D icons representing the assets are
automatically generated at run-time.
The parameterizable layout in the central multi-
media room is achieved through an automated shelf
creation, which originates in a dedicated corner of
the room and is designed as a dynamically sized
bulge. For large numbers of shelves (that is, topics
presented in the multimedia database), it extends as
a corridor of fixed width in a 45-degree angle from
the adjacent static walls. This architectural design
pattern leads to a library area with convex outer
walls, so that the shelf objects are occlusion-free for
any user located in front of them (Figure 4).
3.4 GUI Components
Replicave features the following GUI components:
Five different 3D interactive buttons (exit, con-
text-sensitive help, information on 3D naviga-
tion, up, down);
Single HTML capable overlaid 2D frame.
The GUI components can be reused and individually
extended in experiment-related 3D environments.
A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS IN VIRTUAL 3D ENVIRONMENTS
319
3.5 Metaphorical Connectors
Connectors allow the user to navigate between the
entrance hall and content-related spaces as well as
between different experiments. Two connectors are
used: The vertical connector (here, an elevator) is
used to navigate between the different topics, while
the horizontal metaphor (here, a floor) is used to
navigate between different 3D experiments on the
same topic.
Dynamic content is generated in both connec-
tors. An interactive topic selection panel in the ele-
vator lets the user chose the desired content. The
horizontal connector is based on parameterizable 3D
model templates. Two templates are available: the
building floor (default) and a small-scaled simplified
model of the building floor that serves as an interac-
tive 3D map.
The default template’s shape depends on the to-
tal number of doors that are to be visualized. For a
minimal visualization, the centre of the floor, the
door leading to the small media room and one labo-
ratory door are displayed. For a complete (maximal)
layout, two side wings are added, each of which
provides access to two laboratory doors, while three
doors (a single media room door and two laboratory
doors) are visualized opposite the elevator.
3.6 Other Assets
The framework has recently been extended by an
avatar selection room metaphorically designed as a
dressing room and a parameterizable template for
experimental rooms (Hiller, 2007). Furthermore,
new X3D nodes were created to partly replace ex-
ternal scripting.
3.7 The Document Management
System
Replicave uses the open-source software Philex
(http://sourceforge.net/projects/philex/) for the file
and folder administration. It features user admini-
stration, Web-based file and folder management
with basic operations, editing of text-based files,
configurable access restriction regardless of file con-
tent and a user interface with tree-based folder visu-
alization. Due to the hierarchical folder structure and
multi-user support, content is grouped in dedicated
folders and managed via the DMS. At least five ad-
ministrator or curator roles can be deduced by set-
ting user-specific root folders. Each laboratory or
exhibition room is initially loaded by opening a
script file. Specific initialization and content files
offer a high degree of flexibility
4 IMPLEMENTATION
The Replicave framework is implemented in Exten-
sible 3D (X3D) and the PHP scripting language.
Interaction is implemented in ECMAScript. The
system requires a PHP5-capable Web server. On the
client PC, an X3D browser plug-in is required. The
framework has been successfully tested with the BS
Contact VRML/X3D plug-in by Bitmanagement
Software GmbH and Microsoft Internet Explorer 5.
The following case studies work with the BS Con-
tact VRML/X3D plugin, version 6.2 or higher, and
Microsoft Internet Explorer 7. Together with
DirectX 9.0c, the high-level shading language HLSL
is supported and was used for a soap bubble experi-
ment in a virtual science centre (Hiller, 2007).
5 CASE STUDIES
The feasibility of the concept presented here has
been tested in three case studies. While the Repli-
cave framework has been reused in two implementa-
tions, the author of the third implementation decided
to create an entirely new framework.
5.1 Virtual Science Center
The virtual science center contains several interac-
tive scientific 3D experiments that refer to various
mathematical theories. Users are expected to learn
through interaction with virtual installations.
Hiller (2007) used the Replicave framework and
showed that he could significantly reduce the pro-
duction time required for framework design and
visualization. Decorative 3D-objects, furniture and
plants were used or slightly modified for the desk,
the wardrobe and the exhibition rooms. With this
savings, resources could be spent primarily on the
production of learning content, and a total of five
new interactive 3D experiments with simulations
were developed and implemented.
First, the experiments were classified with regard
to their spatial appearance, simulation model type,
interaction logic and manipulation features. This will
enable parts of the code to be reused to implement
extensions of experiments or similar experiments,
such as the brachistochrone and tautochrone prob-
lem (Figures 5 and 6).
WEBIST 2008 - International Conference on Web Information Systems and Technologies
320
Figures 5 and 6: Brachistrochrone—inverted cycloid ver-
sus a straight line with the same endpoints; tautochrone—
the ball will take a constant time to roll to the end point,
regardless of its starting position.
At first glance, visual representation of the experi-
mental environment enhanced by avatars (Figures 5
and 6) do not deliver the outcome in the form of
textual results or functional dependencies among
variables written in 2D interface elements. We invite
the user to find physical laws behind the experiment
by watching and trying out the simulation. However,
it may be worthwhile to include additional kinds of
text displays or help desks in a future version.
Figure 7: “Lights on” experiment with initial 2D instruc-
tion screen.
Figure 7 depicts the “Lights on” experiment, which
uses complex action logic. The figure shows a table
with seven small lights and individual commutators
for switching the lights on and off. However, each
switch influences not only its own lamp but also
both adjacent lights. The goal of the experiment is to
achieve a situation in which all lamps are lit with a
minimum of switching operations. This kind of ex-
periment requires a mathematical model or an im-
plemented action or interaction logic to enable the
computer to simulate a player and to find and dis-
play the correct solution. In a recent project dealing
with interactive cryptographic protocols, we devel-
oped a methodology to create a PetriStateMachine
that executes the action logic of the players involved
(Baloian et al, 2007). It was shown that Petri net
editors and simulation engines such as Renew
(available at http://www.renew.de/), support time-
efficient modeling of action and interaction logics.
5.2 African Art Museum
The Replicave framework was also successfully
used in an implementation of an African arts
(“grassland”) museum (Mafo, 2007), in which the
framework’s gallery assets were used as a primary
exhibition space (Figure 8). Changes could be lim-
ited to adapting texture image files in order to com-
ply with an appropriate design for the exhibition
context.
Existing interaction facilities allow for manipu-
lating 2D image-based content as well as 3D sculp-
tures. Information panels were updated, and simple
new signalling elements were introduced to help
users navigate within the adapted room installation.
This work could be done by any nonspecialist with-
out deep knowledge in 3D modelling languages.
Again, it has been shown that resources could be
spent on the exhibition content rather than on the
framework design surrounding it.
A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS IN VIRTUAL 3D ENVIRONMENTS
321
Figure 8: Gallery room in the grassland museum.
5.3 Virtual Art Museum
Unlike the two aforementioned implementations, the
virtual art museum created by W. Liu (Liu, 2007)
required a specific visual and geometrical frame-
work design (Figure 9). Instead of using the Repli-
cave framework, an individual framework was de-
signed and implemented.
It could be shown that the framework-related
work took 50% of the entire content production
time, underlining the advantage of reusable frame-
work designs.
Figure 9: Sculpture room in the virtual art museum.
At the same time, the implementation suggested
possible extensions of the framework, including the
implementation of a graphical editor using drag-and-
drop mouse actions to modify the X3D content
graph on the fly, encompassing
spatial design,
lighting and texturing of walls and furniture,
modification of assets.
Such a new framework can be used to create virtual
versions of existing museum installations.
6 DISCUSSION
Today, the design of virtual 3D applications is very
popular, and there is abundant literature on this
topic. However, the realization of virtual laborato-
ries and museums is long and expensive. There are
several papers (Kleinermann, 2006) that outline the
procedure to follow in the design of a VR applica-
tion and number important tools for modelling the
scene, defining the objects and means of interacting
with them. These papers provide a first impression,
whereas the inherent complexity of the creational
process makes it necessary to consult special pur-
pose literature.
In a recent paper, Hendricks et al. (2003) report
on virtual African art galleries. In a comparative
study, 2D and 3D environments were evaluated. The
results of the user study showed that users have a
clear preference for 3D environments only if they
are not too complex and provide the users with a
high level of navigational support, whereas 2D set-
tings are better suited to convey a large amount of
information that exists in sequential form.
These results encouraged us to use the Replicave
framework in the realization of the grassland mu-
seum. An initial evaluation with a small number of
participants partly confirmed observations made by
Hendricks et al. concerning interaction and 3D navi-
gation.
3D modelling languages, like X3D, use scene
graphs to build a 3D spatial design. A notion that is
key to reusing code consists in rewriting the scene
graph.
Unfortunately, changes in the scene graph in-
tended to bring about a local modification of an ob-
ject with respect to its geometric shape, texture or
position should take into account the context of the
relevant node. In Reitmayr and Schmalstieg (2005),
the authors present the idea of adding a context ele-
ment to the traversal states of a scene graph that al-
lows the scene graph to be parameterized and reused
for different purposes. An annotated context-sensi-
tive scene graph improves its own inherent flexibil-
ity when acting as a template with parameters set
during traversal of the graph. Using the new con-
cepts, a general framework called “Studierstube” is
presented, together with a dedicated model server
component containing the scene graph of the build-
ing model and an interface that allows users to inte-
grate already designed components into their own
scene graph realizations.
Automatic generation of user-specific content is
addressed in a paper by Chittaro et al. (2004) and in
further papers cited therein. The authors propose a
novel tool that provides automatic code generation
WEBIST 2008 - International Conference on Web Information Systems and Technologies
322
for adding personalized guided tours to 3D virtual
environments that were developed using frameworks
able to dynamically generate X3D content.
7 CONCLUSIONS
We have presented a general framework for repli-
cated experiments in virtual 3D environments that is
parameterizable and reusable. The framework is
embedded in a new generalized modelling pipeline
that promotes cost-efficiency by reusing existing
design patterns and 3D assets.
The approach was successfully applied in two
implementations of interactive Web-based virtual
3D exhibitions: a grassland museum for African art
and a virtual science centre. In both cases, designers
were freed to focus on content development. A third
implementation illustrated the significant amount of
additional resources required for individual frame-
work design without using an existing framework
library.
8 OUTLOOK
A further extension of the number of parameteriz-
able variables and a complete conversation to XML-
based model and content description languages are
part of ongoing research. Another focus of our future
work concerns the inclusion of conversational agents
in analogy to existing solutions for real museums
(Kopp, 2007).
ACKNOWLEDGEMENTS
We thank our anonymous reviewers, whose detailed
and helpful reports have helped us to improve this
paper.
REFERENCES
Baloian, N., Breuer, H., Luther, W,. 2007. Concept key-
boards in the animation of standard algorithms, Jour-
nal of Visual Language and Computing, 2007; doi:
10.1016/j.jvlc.2007.07.005
Biella, D., 2006. Replication of Classical Psychological
Experiments in Virtual Environments, Logos. Berlin.
Biella, D., Luther, W., 2007. Reusability of 2D and 3D
Course Content for Replicated Experiments in Virtual
3D Environments. In Proceedings of the 6
th
European
Conference on e-Learning – 2007. ACL Press.
Chittaro, L., Ieronutti, L.,Ranon, R., 2004. Navigating 3D
Virtual Environments by Following Embodied Agents:
a Proposal and Its Informal Evaluation on a Virtual
Museum Application. PsychNology Journal, Vol. 2,
(1), pp. 24–42.
Hendricks, Z., Tangkuampien, J., and Malan, K., 2003.
Virtual galleries: Is 3D better? In Proceedings of the
2
nd
International Conference on Computer Graphics,
Virtual Reality, Visualization and Interaction in Af-
rica. ACM Press, pp. 17–24.
Hiller, A., 2007. Interaktive naturwissenschaftliche
Experimente in virtuellen Laboratorien. Entwurf und
Implementation in einem Rahmensystem, Thesis (MA).
University of Duisburg-Essen.
Kleinermann, F., 2006. Methods and Tools for Designing
VR Applications for The Internet. The European
Journal for the Informatics Professional, Vol. VII, No.
2, pp. 12–16.
Kopp, S., Gesellensetter, L., Krämer, N., Wachsmuth, I.,
2005. A conversational agent as museum guide—
design and evaluation of a real-world application. In
Panayiotopoulos et al. (eds.): Intelligent Virtual
Agents, LNAI 3661, pp. 329-343, Berlin, Springer.
Liu, W., 2007. 3D-Modellierung virtueller Museen –
Interaktion mit geometrischen Objekten, Thesis (MA).
University of Duisburg-Essen.
Mafo, L., 2007. Interaktion mit Exponaten in virtuellen
Museen anhand von ausgewählten Beispielen. Thesis
(MA), University of Duisburg-Essen.
Reitmayr, G. and Schmalstieg, D., 2005. Flexible Parame-
terization of Scene Graphs. In Proceedings of IEEE
Virtual Reality 2005, Bonn, Germany, pp. 51–58.
A GENERAL FRAMEWORK FOR REPLICATED EXPERIMENTS IN VIRTUAL 3D ENVIRONMENTS
323